1. Introduction

Optical microscopy techniques based on the control of the polarization of the light have demonstrated their capability to image samples in a non-invasive way without using any fluorescent dye [1–7]. More specifically, it has been consolidated that, by analyzing the interaction between the left and right circular polarization states within an optically active sample, the emerging Circular Dichroism (CD) signal carries structural information at the single molecule level [8]. Numerous early works referred that outside the absorption bands of the sample a weak differential scattering signal can be measured. This is called Circular Intensity Differential Scattering (CIDS) and is originated from long-range chiral structures at a scale of 1/20th the wavelength of the incident light [9–12]. CIDS emission is angularly dependent [9] and is induced by (1) the characteristics of chiral structures, such as their radius and pitch and (2) the compaction of chiral groups. The sharp control of polarization states at the excitation stage can be implemented by the use of photo-elastic modulators [13] combined with a lock-in detection. This allows to extract the weak scattering signal from the interaction with the sample at a high rate [13,14]. This was demonstrated for a dedicated application by improving the imaging contrast of high-order biomolecular organization compared to most conventional microscopy techniques [9,15–22]. Recently, our group has proposed to implement the CIDS technique into a commercial confocal laser scanning microscope, showing the possibility to image the cellular organization and isolated nuclei at the interphase in a label-free way [5]. We have shown that coupling CIDS with a fluorescence modality gives us the ability to decipher the polarimetric fingerprint of the chromatin-DNA in cell nuclei. However, this study was performed on isolated nuclei extracted from the cell in a chemical way, introducing a possible denaturation of the chromatin-DNA organization. Furthermore, we have the drawback of losing the benefit of enhanced spatial resolution at the nanoscale level, achieved by the recent development of super resolution microscopy, limiting the use of such a technique.

Expansion Microscopy (ExM) is a method providing super-resolved images with conventional fluorescence microscopy techniques. To achieve this goal, the expansion process is based on the use of a dense polyelectrolyte hydrogel characterized by high swelling force [23,24]. These features allow an isotropic expansion of the biological sample of about 4.5-fold and, hence, resolve molecules before spaced under the diffraction limit of a conventional microscope [25]. Recently, many variants have developed of such technique, able to obtain a linear expansion in the range of 3-20-fold [26,27]. In addition to the expansion factor (EF), these versions can be classified in two different groups, according to the homogenization process and the introduction of the fluorescent probes before or after expansion [28–30]. The combination of such expansion techniques with other super resolution methods, including STED [25,31,32], STORM [33] and SIM [34,35], have allowed deciphering finer details of several biological structures. The main challenge of investigating the chromatin organization is to preserve, as much as possible, its native organization without isolating the nuclei from live specimens using a chemical method or modifying the molecular density by adding multiple different fluorescent proteins. As introduced by Boyden et al. in the first protocol, ExM allows easily labeling and preserving the DNA organization after the digestion and the expansion process [28]. Using fixative and crosslinking molecules, we are able to maintain and decipher finer details inside the nuclear compartment using a diffracted optical microscopy [20,24,36]. However, it is extremely important to functionalize the biomolecules with specific anchors and homogenize the specimens. Indeed, if these steps are not correctly achieved, the sample can resist the expansion causing the formation of artefacts in the final imaging step [23,28]. In addition, by increasing the relative distance between the labelled features in the biological sample, the fluorescent signal decreases, requesting high-light dose illumination.

Here, we propose to improve the resolution and imaging contrast of our CIDS scanning microscope by introducing expanded samples presenting chromatin-DNA in a combined approach, which we have named ExCIDS. The idea is comparatively simple: when the sample buffer is expanded, the density of the chiral structures is reduced. This leads to a reduction of overall scattered intensity and a lower intensity averaging of different chiral structures per voxel. This results in a better localization of the chiral groups bringing a better accuracy to determine the chromatin-DNA fiber compaction. An improvement of the ExCIDS contrast is expected since it depends on the spatial increasing of radius and pitch of DNA helices [37]; however, the Signal-to-Noise ratio (SNR) decreases dramatically. For this reason, we proposed to understand how the scattering effect could influence the resolution and the imaging contrast mechanisms of the sample embedded in a swellable hydrogel. To do this, we monitored two parameters of the sample preparation influencing the polarization-based imaging contrast: the expansion factor, EF (Pre-Ex, Post-Dig and Post-Ex) and the digestion time, demonstrating the capacity of this novel approach to image the DNA organization in an expanded state.

2. Materials and methods

2.1 Multimodal optical scanning microscopy

The imaging setup was a Nikon multimodal scanning microscope described in our previous work [5]. It couples fluorescence and four polarization resolved channels, detected at two different wavelengths (below 680 nm for the fluorescence and 800 nm for the CIDS modality). We used a 100X/1.3NA (oil immersion) Nikon objective (Plan Fluor DIC H, Nikon Instruments, Yokohama, JP) to focus the light on the sample with a diffraction-limited lateral resolution of ≈ λ/2NA ≈ 0.3 µm. To prevent possible polarimetric effects from using high NA objective, and guarantee the circular polarization states preservation until the sample plane, the Field of View (FOV) is centered and reduced down to 25 × 25µm². The light is collected using a 40X/0.6NA Nikon objective (Plan Fluor ELWD, Nikon Instruments, Yokohama, JP) as a condenser. The method, based on the use of a photoelastic modulator (PEM), is able to generate all the polarization states at the pixel-dwell time rate without moving parts [38]. It was shown using Mueller-matrix formalism that by an adequate orientation of the optical axis of each device and by locking-in on the signal at the reference frequency of the PEM (ω = 50 kHz), specific polarimetric physical effects can be recorded. To avoid moving parts, the detection arm is equipped with a Glan-Taylor prism (GT10-A calcite polarizers Thorlabs, Inc., USA), which splits the beam in two orthogonally polarized beams in two different optical directions. Each one is collected by a tunable gain photodiode (PDA36A-EC Thorlabs, Inc., USA). The modulated signal is then acquired by a four-channel Nikon control unit (C2+, Nikon Instruments, Yokohama, JP) synchronously coupled to the two demodulated signals at the first harmonic of the PEM via a lock-in amplifier (HF2LI Zurich Instruments AG, SUI). The two DC signals are also measured by directly connecting the outputs of each detector to the two other channels. The conventional confocal fluorescence modality was collected at a wavelength below 680 nm in the backward direction by switching one channel of the unit and without moving the sample. In this paper, we focus our study on the scattering process into the PSF volume, which means we measured the intensities to recover a CIDS 512 × 512 pixels image. A calibration step is performed to take into account (1) the different of sensitivities and gains between the two photodiodes and LA channels, (2) the misalignment of the optical orientations of the devices of the microscope, (3) the reflection on the glass interface of the sample and (4) the presence of residual biological materials. Additional systematic errors in the scanning configuration come from the loss of the generated circular polarization in all the field of view (FOV), explained by the Fresnel laws. They have been compensated by reducing the centered FOV from 6 to 10 times.

2.2 Expansion microscopy preparation

The biological samples are related for preserving the right localization of the biomolecules. The key labeling strategy in this work is to use Hoechst to label chromatin DNA in Pre-Ex, Post-Dig, and Post-Ex samples. For antibodies and fluorescent proteins labeling, we adopted the ExM variant developed by Chozinsky et al. [24]. Hek cells are cultured in DMEM supplemented with 10% Fetal Bovine Serum and 1% pen/strep and glutamine. For CIDS and fluorescence imaging modality, the cells are fixed with 4% PFA for 10 min at room temperature. After fixation, the sample is incubated with Hoechst (dilution 1:1000) for 10 min. After functionalization with 25 mM Methacrylic acid N-hydroxy succinimidyl ester (MA-NHS) for 60 min at room temperature, the sample is incubated in the gelation solution consisting of 2M NaCl, 2.5% (Wt/Wt) acrylamide, 0.15% (Wt/Wt) N,N′-methylenebisacrylamide, 8.625% (Wt/Wt) sodium acrylate, 0.2% (Wt/Wt) tetramethylene diamine (TEMED) and 0.2% (Wt/Wt) ammonium persulfate (APS) in DI water. After gelation, the sample is incubated in the digestion buffer (1x TAE buffer, 0.5% Triton X-100, 0.8 M guanidine HCl and 8 units.mL⁻¹ Proteinase K) at a different time (1 h, 4 h and overnight). This step “cleaned” the biological sample from lipids and proteins, preserving the DNA organization. For post-expansion imaging, after incubation overnight in the digestion buffer, the sample is incubated with Hoechst for 10 min, before expanding in milliQ water.

2.3 Image processing and data analysis

The CIDS images, presented in Figs. 1 and 4, were formed in accordance to our previous work [5]. This means that we normalized the m₀₃ images by the total transmitted light (the m₀₀ element of the Mueller-matrix), encoded between -1 and 1 according to the Mueller-Stokes formalism [39]. This allows comparable and reproducible data for all the samples. For the 3D pixel histogram distribution presented in Figs. 3 and 5, only the nucleus area is considered and is identified by manually cropping the region of interest in ImageJ, obtained from the fluorescence image. Then, for comparative analysis, the pixel histograms of the two imaging contrasts were normalized between 0 and 1 by removing the background contribution (around a zero signal) for the two modalities.

 

Fig. 1. Normalized images of different Hek cells labeled with Hoechst for (a, b, c, d) the pre-expansion and (e, f, g, h) at the maximum expansion. More particularly, (a, e) are the CIDS images; (b, f) are the Hoechst images; (c, g) are the superposed CIDS and Hoechst images to better gauge the position of the nucleus; (d, h) are the line profiles corresponding to the yellow lines traced in (c) and (g) respectively.

Download Full Size | PDF

3. Results

3.1 Effect of the expansion process

In our previous work, we demonstrated the capability of CIDS microscopy to image the chromatin-DNA in situ on isolated nuclei by applying a chemical treatment digesting the non-nucleus components. This allows removing all the potential artifacts and reducing edge effects between interfaces, common for polarimetric imaging. Thus, in this work, the combined effects between the expansion and the digestion process participates to limiting such artifacts for each imaging voxels. Furthermore, the following EFs are estimated by correlating the increasing of the hydrogel diameter and the distance between NPC, previously determined in our work [25]. Thereby, only the hydrogel diameter measurement is given in this work for characterizing the expansion process. We investigate the CIDS signal of Hek cells expanded at two different EFs: pre-expansion (Pre-Ex, EF = 1X) and at the maximum expansion (Post-Ex, EF = (4.9 ± 0.1)X ≈ 5X) [28]. In this way, using a 100X/1.3NA objective, it is possible to image the sample with an optical resolution from R_obj ≈ 300 nm (Pre-Ex) and R_obj/5 ≈ 60 nm (Post-Ex).

Figure 1 presents the CIDS and the correspondent fluorescence images of Hek cells labelled with Hoechst, in order to assign properly the chromatin-DNA fingerprint. Figures 1(a)–1(b) and Figs. 1(c)–1(d) present the Pre-Ex and the Post-Ex images, respectively. Figures 1(c)–1(g) correspond to the merged CIDS and Hoechst images to allow better identification of the position of the nucleus. For the same reason, Figs. 1(d)–1(h) show the pixel intensity values of the both modalities from the yellow line profile in the CIDS and fluorescence images Figs. 1(c)–1(g). In this work, the specimens at the two expansion stages are different. The acquisition of a large portion of samples and the difficulty of handling the same Post-Ex specimen are limiting factors in this experiment. For this reason, we acquired different labelled cell nuclei at the different expansion stages.

As explained in our previous work, the firm intensity change showing the chromatin-DNA location comes from the nuclear membrane fingerprint, followed by a higher compacted DNA region. The signal is higher at the periphery of the nucleus, corresponding to the region transcriptionally inactive composed by the heterochromatin structure [40]. In the central part of the nucleus, the less organized and compacted area corresponds to the euchromatin region. In the Mueller-matrix formalism, the amplitude and sign of the CIDS images are related to the compaction, to the radius and pitch of the chromatin helix, and left/right hand chirality, providing a sensitive contrast from the interaction between the circular polarization and the chiral groups containing in the PSF volume [5]. It is worth to note that using high NA objective for polarimetric imaging brings edge effects in the m₀₃ element due to reflection and scattering at the interface such as the membrane of the nucleus. This was observed in our previous work [5] related to the intensity profile, which gives a strong CIDS signal change between the heterochromatin and the membrane. After the functionalization with linkable groups and the gelation process, the sample is homogenized with detergents and proteinase K [23,24], removing the polarization-based signal from lipids and proteins, as can be observed in Fig. 1(e) compared to Fig. 1(a). This process is extremely important in this label-free technique, because it allows to reduce the polarimetric imaging artifacts from the refractive index mismatch between the nuclei, the cellular membrane and the buffer environment. After this, the sample is isotropically expanded, enhancing the final achievable resolution. Comparing the two imaging modalities, the CIDS method offered more sensitivity to the chiral groups. A reduction of the collected CIDS intensity is shown for the image at EF 5X, influencing dramatically the contrast. This can be seen also in Fig. 1(d) and 1(h), which represent the histograms relative to the pixel intensity in the line profiles traced in yellow in Fig. 1(c) and 1(g) respectively. The green line represents the fluorescence signal, while the blue one is relative to CIDS. A double sign peak structure described by a rough negative and positive CIDS sign, can be observed at the edge of the nucleus (localized by the fluorescence signal) in Fig. 1(d). This pattern is consistent with what we have observed in our previous work [5]. We propose to attribute the external peak to edge effects that cause polarization loss at the nucleus membrane, while the internal peak, more closely related to the fluorescence signal, is attributed to the more compacted heterochromatin zone at the nuclear periphery. In Fig. 1(h), the reduced SNR makes it harder to distinguish the nuclear boundaries in the CIDS modality.

This loss of imaging contrast is explained in Fig. 2, where it is assumed that the condenser collected the CIDS signal mainly in the forward direction and, hence, averaged the angular emission of the chiral structures with a high enough NA (estimated in this work at around α ≈ sin−¹(NA/n) ≈ 25° from the 0.4NA air condenser).

 

Fig. 2. Schematic principle of the scattering process into the PSF volume (a) without digestion and (b) after n hours of digestion. (c) Principle of the SNR decreasing and imaging contrast quality improvement after the expansion process.

Download Full Size | PDF

In the case of the undigested sample Fig. 2(a), the concentration of the chiral structures inside the PSF volume allows measuring the CIDS signal with an acceptable SNR. Despite this, the chirality is not well resolved, lacking the structural information in the nuclear region. When the sample under illumination is expanded Fig. 2(b), the concentration of biomolecules in the PSF volume decreases improving the optical resolution. Thanks to the swelling of the hydrogel, the Post-Ex image Fig. 1(f) showed an enhancement of about 5x in the optical resolution, as well as a removal of all the scattering cellular components (membrane and in general proteins). Although the nanoscale information in ExM of the chromatin-DNA organization is still not elucidated, we assume that a partial digestion of the histone protein complexes allows crosslinking some chromatin fiber to the mesh gel, carrying out an isotropic expansion. Indeed, the nucleosomes are complex and dense structures, and the proteinase action might not work efficiently in such condition. In addition, as demonstrated by Boyden at al [23], the gel mesh size of 1-2 nm may efficiently preserve the chromatin structure after expansion in a crowded environment like the cell nucleus. As described in Fig. 2(c), CIDS needs a minimum quantity of chiral groups inside the PSF volume to measure the polarimetric signal. Thus, the expansion process leads to an increase in contrast, but at the same time the SNR decreases. At the maximum expansion, a single chiral group is not able to provide a measureable signal above the noise level to discriminate properly the nucleus location, reducing dramatically the visibility factor. It’s also important to keep in mind that CIDS signal has a strong angular dependence [11]. Considering the limited angular integration of the detectors in our system, a part of the signal is lost (acceptance angle of the condenser: 25°). For this reason, the loss of SNR when the expansion factor becomes too important can also be explained by the peak in CIDS signal moving outside the angular acceptance of our detectors.

To take into account the whole sample inhomogeneity we have performed a new analysis not present in our previous work. Figure 3 presents the histogram pixel distribution of the CIDS in function of the fluorescence signal for both EF, by removing the background contribution following an imaging analysis described in the Materials and Methods section.

 

Fig. 3. Normalized 3D histograms pixels distribution of the fluorescence in function of the CIDS modality from the images Fig. 1 at each EF. The red and blue plots correspond to the Pre-Ex and the Post-Ex stages, respectively.

Download Full Size | PDF

Both histograms of the nucleus area are normalized for comparative analysis, which means that a zero value CIDS corresponds to an absence of chirality or a not well-resolved signal coming from the chiral groups and 1 values correspond to a pure deterministic chirality. From the pixel distribution reported in Fig. 3, clear differences between both the expansion stages are visible. First of all, the fluorescence distribution becomes Gaussian at 5X EF contrary to the first stage. This comes from an improvement of the fluorescence contrast after expansion, which brings a better identification of the chromatin location. Concerning the CIDS intensity distribution, the chirality is weak (around 0.2) at EF = 1X because the signal, coming from a high number of chiral structures having different orientations, is averaged all over the PSF volume. At EF = 5X, these structures are more discriminated, owing to the lower number of chiral groups in the PSF, but for the same reasons, the SNR decreases drastically. These histogram representations could also present the enhancement of correlation between the CIDS and the fluorescence modality when the EF increases. Indeed, the large range of maximum amplitudes comes from the same signal location all over the images for the two modalities, directly linked to a better correlation. For this reason, an increase of the pixel amplitudes dispersion around the mean value is a good marker to highlight this correlation. Comparing both steps of the expansion process, we observe that the correlation increases, thanks to the reduction of the scattering elements composed by the cellular structures after digestion.

3.2 Effect of the digestion time

Figure 4 presents the images obtained for different digestion times of Hek cells labeled with Hoechst to target the chromatin-DNA regions. Considering the weak CIDS signal at the maximum expansion (EF = 5X, post-Ex), we considered here the intermediate EF = 2X, mainly composed by the cellular nucleus and only a few residual not digested biological components.

 

Fig. 4. Fig. 4. Images of Hek cells labelled with Hoechst in CIDS (a, e, i) and fluorescence (b, f, j) modalities. A merged image of both modalities is presented in (c, g, k) and a line profile relative to the yellow line in those pictures is shown in (d, h, l). Here CIDS is represented in blue while the fluorescence signal is represented in green.

Download Full Size | PDF

The Figs. 4(a)–4(e)–4(i) are the CIDS images, while Figs. 4(b)–4(f)–4(j) correspond to the fluorescence images. Figures 4(c)–4(g)–4(k) are the respective merged CIDS and fluorescence images, while Figs. 4(d)–4(h)–4(l) are the line profiles for both modalities corresponding to the yellow lines in the merged images. The images for different expansion processes correspond to no digestion process (Figs. 4(a) to (d)), after 4 h (Figs. 4(e) to (h)) and overnight (Figs. 4(i) to (l)) of digestion.

The biological components from the cellular membrane and the cytoplasm are increasingly removed after the digestion process, as it is shown in Fig. 1(i). Comparing the CIDS images between the two extreme stages of digestion Fig. 4(a) and Fig. 4(i), the intensity dynamics into the nuclear region are stronger in the second case, which indicates a better resolution of the chiral group structures, explained by the schematic in Fig. 2. Without digestion (Fig. 4(a)), all the biological components of the cell, such as the cytoplasm and the membrane, increase the scattering of the excited photons through the PSF volume. The consequence is a loss of intensity collected in the forward direction after interaction with the sample and a decrease of the SNR. After the digestion process (Fig. 4(i)), these biological elements are absent and the number of photons collected in the forward direction after interaction with the sample is increased, which contributes to obtaining a higher contrast. This can be also be observed in the relative line profiles shown in Figs. 4(d) and 4(l). Both figures present the previously mentioned double peak structure at the nuclear periphery, but in the case of the sample digested overnight, the correspondence between the two modalities and the internal dynamics of the CIDS signal result better reproduced.

In order to compare the two imaging modalities, we proposed Fig. 5 to plot a 3D histogram of the pixels distribution of the fluorescence in function of the CIDS signal at the three digestion stages.

 

Fig. 5. Normalized 3D histograms of the pixels distribution of the fluorescence in function of the CIDS modality from the images Fig. 4 for the three digestion times. The red, green and blue plots correspond to 0 h, 4 h and overnight digestion, respectively.

Download Full Size | PDF

Comparing the different stages, we note that the fluorescence values present an evident Gaussian distribution after an overnight digestion, due to a decreasing of scattering components all over the field of view as we show in the previous section. The CIDS intensity follows the same trend, explained by an improvement of the sensitivity to image separately the different ordered chromatin orientations. Furthermore, it can be noted that the correlation after 4 h of digestion is stronger than without digestion. The reason is that the CIDS image contains the information of the whole cell contrary to the fluorescence image, which shows only the expected chromatin-DNA labeling contrast. Indeed, for the label free modality, the presence of the volumetric structures surrounding the nucleus blurred the signal coming from the chiral groups, explaining a lower correlation with the confocal fluorescence, which is well known for its axial optical sectioning capability. After digestion, the two modalities finally present similar contrast, only localized at the chromatin DNA area, improving the optical resolution from the chiral organization and the correlation. As expected, the overnight digestion, corresponding to a total removal of the scattered biological components surrounding the nucleus, presents the best correlation in comparison to the previous stages.

4. Conclusion

We presented a study to improve the optical contrast of chromatin-DNA imaging into entire cells by taking advantage of coupling two microscopy methods. The first (ExM) brings a higher lateral optical resolution and the second (CIDS) a better polarization-based label-free sensitivity to the chiral molecular organization of the sample. By combining these two methods, we proposed a new way (1) to reduce the scattering preserving the structure of interest for the polarization-based microscopy contrast, and (2) to extract additional information by correlating the label-free and the fluorescence approaches. In addition, we investigated the DNA organization in a label-free way without extracting the nuclei from the cellular environment. We believe that the combination of these two techniques can be interesting for the investigation of the chromatin-DNA organization for different cellular phases, such as the chromosomic patterns and some genetic pathology cellular disorders.