From convolution to clarity: effect of different point spread functions for deconvolution in CLSM and STED microscopy images of the nuclear lamina

2.1 Cell culture

Normal human dermal fibroblasts (NHDF) were obtained from PromoCell (Heidelberg, Germany). NHDF were cultured in Dulbecco’s Modified Eagle Medium (DMEM; Biowest, Nuaillé, France) containing 10 % fetal calf serum (FCS) (Gibco, Waltham, MA, USA) and 50 μg/ml Gentamycin (Dechra, Northwich, UK) at 37 °C and 5 % CO₂ in a humified incubator. At near-confluency, the cells were trypsinized using 0.125 % Trypsin/0.02 % EDTA/0.02 % glucose (Gibco, Waltham, MA, USA) solution in Phosphate Buffered Saline (PBS) and passaged at a 1:2 or 1:3 ratio.

2.2 Cell fixation and immunofluorescence staining

Cells were seeded onto 18 mm round glass coverslips (#1.5; VWR, Radnor, USA) at a 1:2 or 1:3 ratio, grown for at least 48 h, and fixed with 4 % formaldehyde (Merck, Darmstadt, Germany) in PBS at room temperature (RT) for 15 min. The cells were stored at 4 °C in PBS containing 0.01 % sodium azide for a maximum of 7 days.

Prior to single immunofluorescence (IF) staining, formaldehyde fixed cells were permeabilized in 0.1 % Triton X-100 (AppliChem, Darmstadt, Germany) in PBS for 15 min at RT, followed by washing with PBS (3 × 3 min). Primary antibodies, diluted in PBS containing 3 % Bovine Serum Albumin (BSA; Roche Diagnostics, Basel, Switzerland), were applied to the coverslips and incubated for 1 h at RT. The following primary lamin antibodies were used: 1) Mouse monoclonal IgG1 anti-lamins A/C culture supernatant (Jol2; dilution 1:50; provided by Prof. dr. C. Hutchison, Durham University, UK) for immunostaining of lamins A and C; 2) Rabbit polyclonal IgG anti-lamin B1 (ab16048; 1 mg/mL; dilution 1:800; Abcam, Cambridge, UK) for immunostaining of lamin B1. After incubation with primary antibodies, the coverslips were washed again with PBS (3 × 3 min) and the secondary antibodies, diluted in PBS containing 3 % BSA, were applied to the coverslips and incubated for 1 h at RT. The following secondary antibodies were used: 1) Goat anti-mouse IgG Abberior Star Green (1 mg/mL; dilution 1:500; Abberior Instruments, Göttingen, Germany) as secondary antibody for detection of lamins A and C; 2) Goat anti-rabbit IgG Abberior Star 512 (1 mg/mL; dilution 1:250; Abberior Instruments, Göttingen, Germany) for detection of lamin B1. A final washing step with PBS (3 × 3 min) was performed before the coverslips were mounted on a microscopy glass slide with Tris-Glycerol DABCO mounting medium (90 % glycerol, 20 mM Tris-HCl pH 8.0, 2 % 1,4-di-azo-bicyclo-2(2,2,2)-octane (Merck, Darmstadt, Germany)), and sealed with nail polish. The slides were stored at 4 °C until imaging within 7 days after immunofluorescence staining.

2.3 Confocal laser scanning microscopy (CLSM) and stimulated emission depletion (STED) microscopy

The immunostained cell samples were imaged using CLSM and 2D-STED microscopy (Leica TSC SP8 STED microscope with LAS X software (Leica, Wetzlar, Germany)). 2D-STED microscopy applies emission depletion only in the xy-plane and not in the z-plane. Throughout this study, when mentioning STED microscopy, we refer to 2D-STED microscopy. Imaging was performed with an HC PL APO CS2 100x/1.40 oil lens. Image acquisition was performed in xyz mode, with gating of 0.2–7.0 ns, format of 1,024 × 1,024 pixels, speed of 400 Hz, and pixel size of 30 × 30 nm. The cells were imaged in photon counting mode, with a gain of 10 %, and line accumulation of 6 for CLSM images and 16 for STED microscopy images. Abberior Star Green was excited with a white light laser (WLL) at 488 nm and its emission detected at 493–545 nm. Abberior Star 512 was excited at 521 nm and detected at 526–565 nm, respectively. Depletion in the xy-plane was done with a 592 nm STED laser, using 50 % laser power. For all images, a small z-stack of three sections (step size 0.10 µm) was generated to fulfil the Nyquist criterion (i.e. the minimal sampling density needed to capture all information from the microscope into the image [18]), which is necessary for optimal image deconvolution.

2.4 Preparation of bead samples for determination of the experimental and experimentally supported theoretical (EST) point spread function (PSF)

Coverslips (18 mm round, #1.5; VWR, Radnor, USA) were coated with 0.01 % Poly-l-Lysine (Sigma Diagnostics, St. Louis, USA) by a 5-min incubation at RT, followed by 1 h of drying at RT. For CLSM PSF samples, 175 ± 5 nm green-fluorescent microspheres (PS-Speck Microscope Point Source Kit; Molecular Probes, Eugene, USA) were diluted 1:1,000 (dilution determined after initial serial dilution tests) in MilliQ and sonicated for 2 min. For STED microscopy PSF samples, 50 nm green-fluorescent nanoparticles (polydispersity index <0.2) (DiagPolyTM Fluorescent Polystyrene Nanoparticles Green 50 nm; CD Bioparticles, Shirley, New York, USA) were diluted 1:10⁵ (dilution determined after initial serial dilution tests) in PBS and sonicated for 5 min. The diluted PSF beads were added to the coated coverslip and dried for 4 h up to overnight at RT, followed by mounting onto a microscopy glass slide as described for the NHDF cell line.

2.5 Determination and characterization of the theoretical, experimental, and EST PSF

The theoretical PSF was calculated based on the sample and microscope imaging parameters using Huygens Professional (Scientific Volume Imaging, The Netherlands, http://svi.nl) [19].

Z-stacks of the beads were made by applying the imaging settings used for the secondary antibody Abberior Star Green (Section 2.3), for both the CLSM and STED microscopy PSF measurements. The z-stack (step size 0.1 µm) recording was started at the focal plane below the beads and ended above it, to ensure that the beads were completely recorded. Then, the experimental PSF was distilled from the bead images with the Huygens PSF Distiller Wizard in Huygens Professional version 19.10 (Scientific Volume Imaging, The Netherlands, http://svi.nl). Appropriate beads, exlcuding doublets, were automatically selected by the software.

For STED, next to the experimental PSF, also important microscopic parameters, specifically the NA, excitation beam fill factor, STED saturation, and STED immunity fraction, were distilled based on the bead recordings. Based on these extracted microscopic parameters, a new theoretical PSF was generated that we refer to as the EST PSF. Thus, this PSF was based on microscopic parameters of experimental bead measurements and not on regular images, as is done during the determination of the regular theoretical PSF.

For all PSF’s, the FWHM of the PSF (i.e., the width of the peak at the half of its maximum value) was determined with the PSF FWHM Estimator in the Huygens software, using a Lorentzian fit (Scientific Volume Imaging, The Netherlands, http://svi.nl).

2.6 Deconvolution using different PSFs

All cellular images were deconvolved with Huygens Professional, using the Classic Maximum Likelihood Estimation (CMLE) algorithm, a quality threshold of 0.05, and maximum iterations of 100. All deconvolved images reached the quality threshold within the 100 iterations. The used PSF for each specific application is indicated in the Results section. The microscopic parameters were extracted from the metadata of the images, except for some parameters for STED microscopy that were estimated on basis of the sub-resolution bead measurements (i.e. NA: 1.26, excitation beam fill factor: 0.60, STED saturation factor 9.85, STED immunity fraction: 14 %). For CLSM images, a signal-to-noise ratio (SNR) of 9.3 was used, while for STED we used a SNR of 10, which was estimated using Huygens Professional with representative images.

2.7 Full width at half maximum (FWHM) determination of the lamin layer

To estimate the lamin layer thickness in deconvolved images, the average FWHM of the lamina was determined at the mid-level of the nucleus, using Fiji [20]. For the quantitative analysis, the layer thickness was determined at five positions in each cell. Intensity curves were plotted through lines drawn perpendicular to the lamina and fitting this with a Gaussian distribution. The FWHM of the lamin layer could not be correctly assessed in images that were not deconvolved, as their Gaussian fitting is systematically affected by high background levels, particularly in the STED images (Figures A1 E-H). Therefore, the thickness was determined at the same position for the CLSM and STED images only after deconvolution with different PSFs. Formula 2√2(ln2 σ), with σ as Gaussian width parameter, was used to calculate the FWHM [21]. The interobserver variation was assessed in our earlier study [17].

3.1 Determination of the different PSFs

A theoretical PSF was derived from the image parameters of both images immunostained with the secondary antibody Abberior Star Green and Abberior Star 512 and these PSFs are visualized in Figure 1A–F for CLSM and in Figure 2A–F for STED. The experimental PSF was distilled after averaging two separate z-stacks of multiple sub-resolution fluorescent beads (exact numbers in Table 1). The experimental PSF of CLSM did not reveal significant aberrations (e.g. spherical aberration) in the x-, y-, and z-plane (Figure 1G–I). The experimental STED microscopy PSF (Figure 2G–I) has a similar shape as compared to the CLSM PSF but is obviously much narrower in the x- and y-direction. However, around this oval shaped PSF, a low intensity blur can be seen, which is best visible in the x, y-image plane (Figure 2G), and is most probably caused by background noise in the bead recording. Based on the microscopic parameters extracted from the experimental PSF of STED (as described in Section 2.5), a new theoretical PSF was generated (Figure 2J–L), which we call the experimentally supported theoretical (EST) STED PSF. Visually comparing the EST STED PSF to the experimental STED PSF, shows an oval shape that is further elongated in the z-direction of the EST PSF, similar as seen for the theoretical PSFs (Figure 2A–F). Importantly, there is no blur surrounding this EST PSF and it contains no noise because it is calculated and not measured. Note again that the EST PSF was only used specifically for STED microscopy.

Figure 1:

PSFs of CLSM. A–C) theoretical PSF based on images with as secondary antibody abberior star green in the x, y- (A), x, z- (B), and y, z- (C) plane. D–F) theoretical PSF based on images with as secondary antibody abberior star 512 in the x, y- (D), x, z- (E), and y, z- (F) plane. G–I) experimental PSF, distilled form z-stacks of green-fluorescent microspheres (175 ± 5 nm), in the x, y- (G), x, z- (H), and y, z- (I) plane. Scale bars: 0.5 µm. Calibration bar displays fluorescence intensities (scaled to maximum intensity value of 65,534).

Figure 2:

PSFs of STED. A–C) theoretical PSF based on images with as secondary antibody abberior star green in the x, y- (A), x, z- (B), and y, z- (C) plane. D–F) theoretical PSF based on images with as secondary antibody abberior star 512 in the x, y- (D), x, z- (E), and y, z- (F) plane. G–I) experimental PSF of STED microscopy, distilled from z-stacks of green-fluorescent polystyrene nanoparticles (50 nm, polydispersity index <0.2), in the x, y- (G), x, z- (H), and y, z- (I) plane. J–L) EST PSF of STED microscopy (i.e. theoretical PSF based on experimental microscopic parameters extracted from bead measurements) in the x, y- (J), x, z- (K), and y, z- (L) plane. Scale bars: 0.5 µm. Calibration bar displays fluorescence intensities (scaled to maximum intensity value of 65,534).

To determine the exact size of the various PSFs, the FWHM of intensity plots of different planes (x, y, and z) of the PSF was determined (Table 1). The FWHM of the theoretical PSFs of CLSM are smaller as compared to the experimental PSF in all planes, with the FWHM of the theoretical PSF of Star 512 being bigger compared to that of Star Green. As expected, the FWHM of the STED microscopy PSFs is much smaller in the x- and y-plane compared to the FWHM of CLSM PSFs. This is in line with the results seen in Figures 1 and 2. Since no 3D emission depletion was applied in the z-planes during STED image acquisition, these z-plane FWHM values of the CLSM and STED microscopy PSF are similar. The FHWM of the STED theoretical PSFs is bigger compared to the experimental PSF, but similar to that of the EST PSF. Also, for STED, the FWHM of the Star 512 theoretical PSF is bigger compared to the Star Green PSF. The FWHM of the experimental PSF of STED microscopy is slightly smaller in the x- (16 %) and y-plane (15 %) compared to that of the EST PSF, which is also visible as the smaller center of the PSF shapes in Figure 2G–I as compared to Figure 2J–L. As mentioned above, the EST and theoretical PSFs are elongated in the z-direction compared to the experimental PSF, which is visible as a 45 % larger FWHM in the z-plane when the EST PSF is compared to the experimental PSF.

3.2 Effect of different PSFs on deconvolution of CLSM images

The difference between performing no deconvolution, performing deconvolution with a theoretical PSF, or with an experimental PSF for CLSM images is displayed in Figures 3 and 4. These images show NHDF nuclei stained with antibodies for lamins A/C or lamin B1, imaged at the mid-level (Figure 3) or top (Figure 4) of the nucleus. The used theoretical PSF is computed solely based on microscopic parameters extracted from the image and is thus not based on bead measurements, while the experimental PSF is based on the bead measurements as described above. Comparing non-deconvolved images to deconvolved images of the mid-level of the nucleus by visual inspection (Figure 3), either deconvolved with a theoretical or an experimental PSF, reveals that the lamin layer is less blurry and contains less noise after performing these deconvolution steps. This is also clearly visible from the lamina intensity profile (Figures A1 A-D). The images of the top of the nucleus (Figure 4) show a lamina network that can be more readily distinguished after performing deconvolution. In addition, the tubular invaginations of the nucleoplasmic reticulum (NR) visible in the dense lamin network are better delineated after deconvolution and as expected, the interior space of the NR becomes devoid of fluorescence (arrows in Figure 4A’–C’ and D’–F’).

Figure 3:

Effect of deconvolution using a theoretical or experimental PSF on CLSM images at the mid-level of the nucleus of normal human dermal fibroblasts (NHDF) stained with antibodies against lamins A/C or lamin B1. A–F) CLSM image of lamins A/C (A–C) or lamin B1 (D–F), either non-deconvolved (A, D), deconvolved with a theoretical PSF (B, E), or with an experimental PSF (C, F). A’-F’) higher magnification of the ROIs (white rectangles) in (A–F). Scale bars a–F: 5 µm, A’-F’: 1 µm.

Figure 4:

Effect of deconvolution using a theoretical or experimental PSF on CLSM images at the top of the nucleus of normal human dermal fibroblasts (NHDF) stained with antibodies against lamins A/C or lamin B1. A–F) CLSM image of lamins A/C (A–C) or lamin B1 (D–F), either non-deconvolved (A, D), deconvolved with a theoretical PSF (B, E), or with an experimental PSF (C, F). A’–F’) higher magnification of the ROIs (white rectangles) in (A–F). Arrows point towards tubules in the lamin network that arise due to nucleoplasmic reticulum invaginations. Scale bars A–F: 5 µm, A’-F’: 1 µm.

Measuring the lamina thickness of the deconvolved images at the mid-level of the nuclei reveals a 4 % thinner lamin layer for lamins A/C in the images deconvolved with the experimental PSF compared to those deconvolved with the theoretical PSF (Supplementary Table A1). For lamin B1, the lamin layer is found to be 8 % thinner in the images deconvolved with the experimental PSF, which is a statistically significant (p = 1*10⁻⁴) difference as compared to the images deconvolved with the theoretical PSF. More importantly, the experimental PSF corrects for (potential) setup-specific physical deviations, while the theoretical PSF does not.

3.3 Effect of different PSFs on deconvolution of STED microscopy images

The effect of applying deconvolution with different PSFs on STED microscopy images is visualized in Figures 5 and 6. Comparing non-deconvolved to deconvolved images, regardless of which PSF is used, demonstrates a less blurry lamina in the deconvolved images at the mid-level of the nucleus (Figure 5) and a more defined network at the top-level of the nucleus (Figure 6), as visible by eye. The different PSFs that are compared include the theoretical PSF (based on microscopic parameters extracted from the image), the experimental PSF (distilled from sub-resolution bead measurements), and the EST PSF (based on microscopic parameters extracted from sub-resolution bead measurements).

Figure 5:

Effect of deconvolution using different PSFs on STED microscopy images at the mid-level of the nucleus of normal human dermal fibroblasts (NHDF) stained with antibodies against lamins A/C or lamin B1. The nuclei are the same as those depicted in Figure 3. A–H) STED microscopy image of lamins A/C (A–D) or lamin B1 (E–H), either non-deconvolved (A, E), deconvolved with a theoretical PSF based (i.e. microscopic parameters extracted from the image) (B, F), deconvolved with an experimental PSF distilled from bead measurements (C, G), or deconvolved with the EST PSF (i.e. based on experimental microscopic parameters extracted from bead measurements) (D, H). A’–H’) higher magnification and enhanced brightness of the ROIs (white rectangles) in (A–H). Arrows in C’ and G’ indicate a ring with lack of signal. Scale bars A–H: 5 µm, A’-H’: 1 µm.

Figure 6:

Effect of deconvolution using different PSFs on STED microscopy images at the top of the nucleus of normal human dermal fibroblasts (NHDF), stained with antibodies against lamins A/C or lamin B1. The nuclei are the same as those depicted in Figure 4. A–H) STED microscopy image of lamins A/C (A–D) or lamin B1 (E–H), either non-deconvolved (A, E), deconvolved with a theoretical PSF based (i.e. microscopic parameters extracted from the image) (B, F), deconvolved with an experimental PSF distilled from bead measurements (C, G), or deconvolved with the EST PSF (i.e. based on experimental microscopic parameters extracted from bead measurements) (D, H). A’–H’) higher magnifications and enhanced brightness of the ROIs (white rectangles) in (A–H). White arrows point towards tubules in the lamin network that arise due to nucleoplasmic reticulum invaginations. Yellow arrow points towards an additional artificial gap devoid of fluorescence, not visible after EST PSF deconvolution. Scale bars A–H: 5 µm, A’–H’: 1 µm.

Deconvolution performed with different PSFs has a different visible outcome. Most strikingly, deconvolution performed with the experimental PSF reveals an artificial intranuclear ring close to the lamina lacking a fluorescent signal (arrows in Figure 5C’ and G’). At the top of the nucleus, deconvolution with the experimental PSF seems to lead to over-emphasis of brighter structures and elimination of finer detailed structures, in contrast to the theoretical PSF or the EST PSF (Figure 6). In addition, for lamin B1, the extranuclear noise becomes more prominently visible using the experimental PSF (Figure 6G). The NR tubules in the lamin network, as seen in the CLSM images (Figure 4), are less obvious in the STED images before and after deconvolution but are still present in the images of the nuclei immunostained for lamins A/C (arrows in Figure 6A’–D’). For images of lamin B1 these tubules can no longer be clearly distinguished, except in the image deconvolved with the experimental PSF (white arrow in Figure 6G’), but additional artificial gaps devoid of fluorescence become apparent by applying this deconvolution method (yellow arrow in Figure 6G’). Comparing deconvolution using the theoretical PSF and the EST PSF for both lamins A/C and B1 reveals only minor differences.

Comparing the lamina thickness in the deconvolved images at the mid-level of the nucleus demonstrates no significant differences between the lamin layer of both lamins A/C and lamin B1 when deconvolved with the theoretical PSF compared to the EST PSF (Table A1). Comparing the thickness of the lamin layer in images deconvolved with the experimental PSF to those deconvolved with the EST PSF reveals a slightly, but not significantly, thinner (6 %) lamins A/C layer in the images deconvolved with the experimental PSF. In contrast, the lamin B1 layer is significantly (p = 7*10⁻¹⁰) thicker in the images deconvolved with the experimental PSF as compared to the EST PSF.

4.1 PSF as indicator of the resolution of the microscope

The FWHM of the experimental PSF provides information about the resolution of the microscope used [3]. For CLSM, this experimental-based FWHM was found to be 181 nm in the x-direction and 180 nm in the y-direction (Table 1). For STED microscopy the values were 82 and 80 nm, respectively. This decrease is as expected because of the use of the STED depletion laser [22]. The FWHM in the x- and y-direction of the STED experimental PSF is also close to a previously reported STED lateral resolution between 30 and 80 nm for biological samples, the exact value being dependent on the properties of the sample and the applied depletion laser power [23]. It should be noted that 50 % STED depletion laser power was applied in this study and a higher resolution can be achieved with higher depletion laser power, but with the disadvantage of increased photobleaching. The relatively low STED resolution can also be explained by the effective NA of 1.26 that was reported based on the bead measurements, which is lower compared to the nominal NA of 1.4.

Because we used 2D-STED microscopy, where emission depletion is only applied in the x- and y-plane and not in the z-direction, there was no enhanced resolution in the z-plane for STED microscopy as compared to CLSM (405 nm for STED microscopy vs. 397 nm for CLSM) [24]. Although imaging with 3D-STED would enhance the z-resolution, it would at the same time decrease the resolution in the x- and y-direction [25].

The FWHM of the EST and theoretical STED PSF were found to be slightly larger in x- and y-plane compared to the experimental PSF. This difference was more prominent in the FWHM value in the z-axis. This could mean that the theoretical and EST PSF underestimate specifically the z-resolution or that the experimental PSF overestimates it. However, since the experimental PSF is based on the bead measurements it likely represents the actual situation better. The effect of using these PSFs for deconvolution is described below.

4.2 PSF for deconvolution

Deconvolution performed with an experimental PSF leads to a clear improvement of resolution and SNR for CLSM microscopy images (Figures 3 and 4). For STED microscopy images the same holds true, but in addition an artificial, ring shaped area lacking signal appears close to the lamin layer in images at the mid-level of the nucleus (Figure 5C’ and G’). Such a region without lamin signal is not visible before deconvolution and not seen using other microscopy techniques [10], [11], [12], [13], [26]. The EST PSF in STED avoids this artifact. A similar phenomenon can sometimes be seen in MRI images, here often referred to as truncation, Gibbs artifact, or ringing artifact [27]. The latter term is also used in microscopy and has the appearance of dark and light ripples around bright features of an image. Ringing artifacts often occur near sharp signal transitions in an image and are known to occur by either inadequate spatial sampling of the raw image or PSF, or a noisy image or PSF [28].

Indeed, the experimental PSF of STED microscopy contains a low intensity signal around the central maximum (Figure 2G), while this is not seen in the images of the theoretical PSF or the EST PSF (Figure 2A-F, J-L). The STED images acquired at the top of the nucleus show no obvious ringing artifact when deconvolved with the experimental PSF, although for the lamin B1 staining (Figure 6G) there seems to be a narrow ring with a lack of signal between the nucleus and the noise in the surrounding cytoplasm. The difference in occurrence of the ringing artifact at the mid-level as compared to the top of the nucleus can likely be explained by the relatively continuous signal at the top of the nucleus in contrast to the high intensity, linear signal at the mid-level, resulting in a sharp transition of fluorescence intensity. Furthermore, at the top of the nucleus deconvolution with the experimental PSF seems to cluster structures (Figure 6C–G), revealing more open structures in the dense network than are to be expected from the CLSM images and the raw STED images. The larger open structures seen in the images at the top of the nucleus are most likely the start of tubular invaginations of the NR and not nuclear pore complexes, which appear much smaller as seen in the study of Xie et al. [10]. For lamin B1, the NR tubules visible in the CLSM images at the top of the nucleus were less well visible in the STED images, except for the ones deconvolved with the experimental PSF. The lack of the visibility of the tubules in the STED images can most likely be explained by a decreased fluorescence intensity at the borders of the tubules caused by a lower SNR in the STED images. The images deconvolved with the experimental PSF do reveal a tubule at a position where it could be expected, but additional regions with lack of signal and regions with clusters of high intensity signals are visible. Together with the perceived ringing artefact, deconvolution with this experimental PSF is therefore not found to be reliable for STED microscopy.

Taken together, the ringing effect is most likely caused by the noise around the experimental PSF (Figure 2G) of STED microscopy. This noise might originate from an incomplete depletion of fluorophores [29] or might be related to the low fluorescent intensity of the beads used. The 50 nm beads necessary for sub-resolution bead measurements in STED microscopy suffer from a low fluorescent signal intensity and/or fast bleaching during acquisition of the z-stacks. This is more problematic for STED bead measurements as compared to CLSM bead measurements, where larger beads are used, as the number of fluorescent molecules scales with the size of the beads [6]. Thus, there is a low SNR in the bead recordings with STED microscopy, which could lead to noise around the distilled experimental PSF. A comparison by Klemm et al. [5] demonstrated that gold beads have a much higher intensity and better SNR compared to fluorescent beads. However, using gold beads for PSF measurements requires imaging in reflection mode, which is not representative of the microscope settings when imaging a biological sample with fluorescent labeling. For optimal deconvolution, it is best to use microscopic imaging parameters that are as similar as possible to those used for the actual imaging of the sample [6]. In case PSF measurements are done solely to acquire information about the resolution of the optical set-up, gold beads may be preferred for a better SNR. However, if the PSF is also used for deconvolution, fluorescent beads are preferred. For fluorescent beads it should be noted that these are commercially available in a small variety of fluorescent probes and that these may not have the same spectral characteristics as the probes used in the imaging study. In this study the fluorophores in the beads are not the same as those used for the lamin labelling, although the spectra are very similar. These spectral differences could also have a small effect on the STED saturation factor determined with the bead measurements and used for the EST PSF. However, the EST PSF takes into account other important microscope-specific parameters (NA, excitation beam fill factor, and STED immunity fraction) that are based on experimental data not represented in the theoretical PSF and therefore increases the reliability of deconvolution.

4.3 Image versus reality

In this study, the effect of using different PSFs for deconvolution has been assessed with fluorescent images of the nuclear lamina. The aim of deconvolution is to reverse convolution and retrieve the original object as much as possible [1]. In an optimal setting this would imply that the actual structures of the lamina network can be retrieved after deconvolution. However, it remains a challenge to determine which structures in an image are real, or artifacts introduced by deconvolution. In the underlying study the lamin layer was assessed visually and by quantifying its thickness. Earlier studies, employing different fluorescence super-resolution imaging techniques or electron microscopy, have also visualized the lamin network [10], [11], [12], [14], but due to variations in resolution, image processing, and analysis methods across these studies, direct comparisons are not feasible, making it difficult to use them to validate our chosen deconvolution approach. The lamina thickness has been reported earlier to be ∼14 ± 2 nm [13], [14], [30], which is below the resolution limit of STED microscopy, also ruling out using lamin thickness itself as an exact indicator for the correct deconvolution approach. However, due to their sub-resolution thickness, measurements of the lamin thickness can be used as a parameter for the resolution when assessing the effect of the different deconvolution approaches.

4.4 Conclusions

The experimental PSF of CLSM can be used to determine the resolution of the microscope and to deconvolve images of this microscope. For STED microscopy, the experimental PSF can be used to determine the microscope’s resolution, but for deconvolution it suffers from a ringing artifact due to intensity issues. Therefore, for deconvolution of STED images the EST PSF is better suitable, both as compared to the experimental one, since it does not suffer from ringing, and as compared to the theoretical one, since it uses STED microscope parameters estimated from the experimental PSF and therefore enables a more robust PSF estimation [31]. These findings can further improve the super-resolution study of the nuclear lamina, but the method can also be extended to other microscopy studies. While suggesting a new approach for deconvolution of STED images by using an EST PSF, its application should be further explored with other samples, imaging settings, and microscopes.