The MIET studies presented in the previous sections were concerned with structures situated in increasingly more complex environments: From microtubules on a coated glass cover slip surrounded by a buffer, via the outer membrane of a cell, to focal adhesions (which are close to the outer membrane of the cell) and finally the challenging regime of the nuclear envelope. As we stated in the introduction of chapter two, it is not feasible to model every detail of a real system. The main simplification which is used in all MIET calculations is the assumption of a stratified system. This means that the sample (excluding the dipole emitter itself) is modeled as a system of parallel planar layers with different thicknesses and refractive indices. As the studied structures get more complex, they deviate more from this model. Therefore, an important question when using MIET to study biological samples is which influence a deviation of the real structure from this ideal model has on the calculated height values. The second important parameter set that affects the MIET calibration curve consists of the optical properties of the fluorophore, namely the free space lifetime τ0 and the quantum yield Φ. The former can be readily obtained by measuring the same sample on an untreated cover slip12, therefore we assume that it is always correct. The latter is often only known in a different environment and with some measurement uncertainty, and has to be converted using the measured value of τ0 and the assumed value of the refractive index. Thus, it may also suffer from inaccuracies and introduce a bias in the calculated height values.
In this section, we study the effect of both the refractive index (RI) structure and the quantum yield values by calculating MIET curves for different scenarios and comparing the thus obtained height values zcalc with the true heights ztrue. Depending on the experiment, we are sometimes interested in the absolute height of a fluorophore above the substrate, and sometimes only in thez-position of two different types of fluorophores relative to each other. Therefore, both absolute height values z and height differences
∆z =zupper−zlower will be examined.
The top row of figure 4.26 illustrates the situation studied in the first part of this section. A MIET substrate consisting of a glass cover slip (nglass = 1.52) coated with thin metallic films – namely 2 nm titanium, 10 nm gold and 2 nm titanium – and a 10 nm thick silica spacer (nsilica = 1.52) is covered completely by a homogeneous halfspace with refractive index ncell = 1.37. The composition of the MIET substrate is the same as that used for the examination of the nuclear envelope in the previous section, while 1.37 is the average RI of a HeLa cell. We now simulate a MIET curve for a hypothetical
12If the fluorophores are very close to the glass cover slip, e.g. when working with cells whose plasma membrane has been labeled, one has to take into account that a water/glass interface also induces a lifetime change, albeit a much smaller one of typically less than 10 %. See figure 2.18a for a plot ofStot versuszat a water/glass interface. In this case, one has to roughly estimate the height of the fluorophores, and use a curve such as that in figure 2.18a to determine a guess of the free space lifetime. This value may then be used to calculate the actual MIET calibration curve, and the refined height estimate can then be used to iterate the procedure. Due to the much weakerz-dependence of the lifetime at a water/glass rather than a water/metal interface, this calculation usually converges quickly.
-15 -10 -5 0 5
a) b) c)
d) e) f)
0.8 0.9 1.0 1.1 1.2 1.3 1.4
0 50 100 150 200
a) b) c)
0 50 100 150 200
d) e) f)
ztrue [nm]
ztrue [nm]
(∆zcalc−∆ztrue)/∆ztrue[-]zcalc−ztrue[nm]
1.37
a) b) c) d) e) f)
correct
1.33 1.40 1.46
1.52 1.52
Au Ti
Figure 4.26: Impact of incorrectly modeled sample parameters onz-localization accuracy.
Top row: Schematic of the true sample (left) and six “wrong” models (a-f) used to calculate MIET curves. Models (a-c) correspond to incorrect assumptions about the sample itself, while (d-f) concern wrong modeling of the MIET substrate, see text for more details. Central row: Deviation between the height values zcalc calculated using the MIET curves (a-f) and the true height valuesztruefor various emitter positions. Bottom row: Relative error of height differences ∆z between an upper and a lower fluorophore, plotted against thez-coordinate of the lower fluorophore. Calculations were done for true distances ∆ztrue= 5 nm (solid lines), 15 nm (dashed lines) and 25 nm (dotted lines). In all four panels, the emission wavelength is λ= 600 nm and the quantum yield Φ = 1.
fluorophore emitting at λ = 600 nm and having a quantum yield of Φ = 1 that is situated at many different heights ztrue above the silica layer13. The thus obtained lifetimes are then transformed back into height valueszcalcusing MIET curves calculated for the following “wrong” sample models (see top row of figure 4.26):
a) Homogeneous halfspace filled with water (nwater= 1.33).
b) Homogeneous halfspace filled with mounting medium Fluoroshield (nFS = 1.40).
c) Homogeneous halfspace filled with mounting medium Mowiol, using the average refractive index found in the literature (nMow = 1.46).
d) Replacing the three-layered metal system Ti-Au-Ti by a single gold layer whose thickness (14 nm) equals the sum of the three metal layers (2 nm-10 nm-2 nm).
e) Assuming the gold layer has a thickness of 12 nm instead of 10 nm.
f) Assuming the silica layer is only 5 nm thick.
Thus, a)-c) model the situation that the refractive index of the sample itself was guessed incorrectly, while d)-f) treat errors in the composition of the MIET substrate, namely an oversimplification of the sample by ignoring the presence of the titanium, or wrong thickness assumptions as they might occur due to errors during the sample preparation.
The central row of figure 4.26 depicts the error zcalc−ztrue for all six situations. An incorrect refractive index in the sample halfspace leads to a bias of the height, where a too small RI leads to larger zcalc (a) and a too large RI results in smaller zcalc (b, c).
We only study the height range where the MIET curves are monotonically increasing, i.e. up to ztrue = 200 nm. Within this range, the absolute error rises with z, however, it levels off at approximately z = 100 nm, leading to a decreasing relative error. In contrast, the error due to incorrectly modeled metal layers can lead to both over- and underestimation of the height. Finally, assuming that the silica layer is only 5 nm instead of 10 nm thick results in a MIET curve whose shape is almost unchanged but that is shifted by 5 nm. Thus, when adding the silica spacer thickness andzcalc to obtain the distance of the emitter from the top of the metal layers, the result is almost correct.
This indicates that the refractive indices of dielectric materials which are not in direct contact with the emitter do not have a large impact on the MIET curve. The same calculations were repeated for a quantum yield of 30 % but are not shown here because they only differ in the fourth significant digit from the results for Φ = 1.
In co-localization experiments such as the study of focal adhesions in section 4.1.3 or the distance determination of INM and ONM in the nuclear envelope in section 4.1.4, the main quantity of interest is the height difference ∆z between two fluorophores.
Neglecting that these are usually dual-colour measurements, we present the relative error of ∆zcalc for ∆ztrue= 5 nm, 15 nm and 25 nm in the bottom row of figure 4.26. For all three fluorophore distances, the relative error of a model with incorrect sample RI (a-c) shows a similar behaviour that is consistent with the plots in the central row: At small (ztrue <30 nm) and large (ztrue >120 nm) distances from the substrate, the height
13Since the MIET curve is proportional to the free space lifetime and the latter is assumed to be known, its value is arbitrary for these calculations.
bias is almost constant, leading to a small relative error of ∆z in the order of 5 % or less.
At intermediate distances, where the height biaszcalc−ztruechanges rapidly, the relative error of ∆z has a maximum around 6 % to 20 %, depending on the difference between the correct and the actually used refractive index. If the sample is modeled correctly but the metal layers are represented wrongly (d-e), very large relative deviations up to 40 % can occur. Contrarily, assuming a smaller thickness of the silica spacer (f) results in relative errors that are always smaller than 4 %, consistent with the almost parallel curve in the central row.
These results explain why we only saw a very small impact of the four different sample geometries on the calculated average distance between the two nuclear membranes in section 4.1.4. Including a 10 nm thick layer of Mowiol with a higher RI than the cell in the model is similar to changing the thickness of the silica layer, thus only small deviations of ∆zcalc between models are expected. Similarly, the inclusion of the plasma membrane (i.e. a thin dielectric layer) in the model does not result in large changes. In the study of focal adhesions in section 4.1.3, vinculin was found at heights around 10 nm to 20 nm, while actin was located at approximately 20 nm to 50 nm. At these small distances from the substrate, even when comparing the smallest and largest refractive indices used in the simulation, namely 1.33 and 1.46, the calculated heights differ by less than 6 nm (difference between a) and c) in the top row of figure 4.26), or the relative height difference between actin and vinculin by less than 6 % (bottom row). Since the true refractive index inside the cells is probably in between the RI of cytosol (n = 1.37) and the RI of the mounting medium (n = 1.4), using the latter as sample RI is a valid approximation that should not introduce significant errors.
With that, we turn to the influence of incorrect quantum yield values on the deter-mination of height values. The sample we study is the same as before, but now the fluorophore either has a quantum yield of Φtrue= 1.0 or Φtrue = 0.3. Figure 4.27 shows the same kind of height deviation plots as figure 4.26, now for varying assumed quantum yields. The curves in the top row demonstrate that there is one point at ztrue= 161 nm where all MIET curves predict the same lifetime, independent of the quantum yield. At this height, the total energy Stot(z) emitted by an ideal dipole emitter with λ= 600 nm equals the free-space value S0, thus,τ(z) =τ0/(1−Φ + Φ·Stot(z)/S0) =τ0 irrespective of Φ. For ztrue <161 nm, an underestimation of the quantum yield leads to an underes-timation of the height, and vice versa. For ztrue >161 nm, the opposite is true. The effect is stronger for small quantum yield values than for large Φtrue. For example, the height error is always less than 1 nm if one assumes a quantum yield of 98 % instead of 100 %. The plots in the lower row of figure 4.27 illustrate the relative error of the determined relative distance ∆zcalc for ∆ttrue= 5 nm, 15 nm or 25 nm. Here, a complex behaviour can be observed, but the general trend is the same for all seven cases: For fluorophores that are close to the substrate (ztrue, lower ≤ 30 nm), the relative error is quite small, while it increases drastically at heights above 150 nm. As before, the effect is smaller for fluorophores with a higher quantum yield.
Overall, the height deviation observed when employing a wrong quantum yield is larger than the height bias introduced by using a wrong refractive index. This highlights the importance of knowing the correct quantum yield of a fluorophore. However, this quantity is often not published by the manufacturers of fluorescent substances, because it is harder to determine than, say, a fluorescent lifetime or an emission spectrum.
ztrue [nm] ztrue [nm]
Figure 4.27: Impact of incorrect quantum yield values on z-localization accuracy. In the left column, the “true” quantum yield is Φtrue= 0.3, in the right column Φtrue= 1.0.
Top row: Deviation between the height valueszcalc calculated using MIET curves with
“wrong” quantum yield values Φ as indicated and the true height valuesztrue for various emitter positions. Bottom row: Relative error of height differences ∆z between an upper and a lower fluorophore, plotted against the z-coordinate of the lower fluorophore.
Calculations were done for true distances ∆ztrue= 5 nm (solid lines), 15 nm (dashed lines) and 25 nm (dotted lines). In all four panels, the emission wavelength isλ= 600 nm.
Therefore, the next section introduces a method for determining quantum yield values using a metal nanocavity.