from each image its average intensity, as previously
demonstrated [24].- Calculate the spatiotemporal correlation (Eq. 1, G(ξ,χ,τ)).
Remove G(ξ,χ,0) because the correlation due to the shot
noise in low-light regime dominatesG(0,0,0). The correlation
due to the detector dominates theG(1,0,0), and molecular
movement during the exposure time could deformG(ξ,χ,τ) for
τ¼0 by producing an apparent increase in the measured PSF
waist (this effect disappears forτ>0). InterpolateG(ξ,χ,τ)bya
Gaussian fit (Eq.3) (e.g., by the Matlab “gaussfit” function) to
recover theiMSD. - Plot the obtained waistσ(τ)^2 (iMSD) as a function of time (see
Note 8). The first few points can be fitted to extrapolate the
intercept at zero time delay (σ 02 ) (5 points are usually enough
but more points can be fitted if they show a linear behavior) and
compare the obtained value with the previously calibrated value
of the PSF (seeNote 9). - Transient confinement: the case of TfR-GFP. Many studies
showed that the cytoplasmic tail of this receptor interacts with
the membrane-associated F-actin skeleton, which in turn acts
as a fence for the receptor mobility [15, 25] (Fig.4a). A
representative TIRF image of a CHO-K1 cell expressing
TfR-GFP is presented in Fig.2b (seeNote 10). The temporal
evolution of the correlation function with the corresponding
Gaussian fit and residues shows the expected decrease in height
and increase in width of the correlation, due to molecular
movement (Fig.2c). Also, the autocorrelation plot shows
that the characteristic time of the fluctuations is shorter than
the total length of the measurement (arrow) (Fig.2d). Thus,
immobile fraction removal is a safe operation. As expected, the
measured diffusion law (Fig.2e, red curve) for TfR-GFP shows
a first flat behavior below 100 nm, with an averageDappof
about 0.7μm^2 /s, followed by consequent rapid decrease in
apparent diffusivity down to 0.2μm^2 /s (the value typically
measured by diffraction-limited FCS [15]). This result shows
that our approach can easily measure the average displacement
of GFP labeled proteins with a resolution of few tens of nan-
ometers (seeNote 11and ref.26). Moreover, the spatial scale at
which theiMSD starts to decrease its slope sets the characteris-
tic spatial scale of protein (partial) spatial confinement induced
by the membrane skeleton at around 120 nm, in keeping with
previous estimates [1]. F-actin digestion by Latrunculin B
(Fig.2f) produces the expected change in the TfR-GFP diffu-
sion law (Fig.2e, green curve). - Dynamic partitioning: the case of GFP-GPI. The iMSD
approach has the ability to discriminate between transiently
284 Francesco Cardarelli