1. Introduction

Pseudoxanthoma elasticum (PXE, OMIM#264800) is an autosomal recessive metabolic disorder characterized by progressive ectopic mineralization of soft connective tissue [1]. To date, diagnosis of PXE is confirmed by skin biopsy and/or genotyping. Histopathology findings include fragmented, clumped and mineralized elastic fibers and calcium deposits in the mid-dermis, mainly consisting of calcium hydroxyapatite and calcium hydrogen phosphate [2]. According to ultrastructural studies, fine precipitates inside the elastic fibers appear which then build up to form bulky deposits; this process eventually leads to deformation and breakage of the elastic fibers [3]. Coiled, split collagen fibers with irregular diameter and flower-like collagen fibrils are also observed sometimes [4].

Nonlinear microscopy (NLM) is a high-spatial resolution imaging technique that has been increasingly used in dermatological research including non-invasive detection of skin cancer [5] and stain-free examination of dermal connective tissue alterations [6–8]. Among the different NLM modalities, two-photon excitation fluorescence (TPEF) is suitable for the visualization of endogenous fluorophores, such as elastin and keratin, whereas collagen with its chiral structure generates second-harmonic generation (SHG) signal [9].

Recently, we have shown that NLM is able to visualize the histopathological alterations of the mid-dermis in PXE-affected skin using formalin fixed, paraffin embedded, deparaffinized PXE sections. Besides SHG imaging of collagen, we successfully visualized calcification, elastic fiber fragmentation and calcified elastic fibers with TPEF, that are histopathological hallmarks of PXE. Based on our results, NLM visualized the similar features of PXE, compared to conventional histopathological alterations, and proved to provide a highly detailed imaging of specific alterations of PXE-affected skin [10]. For the diagnosis of PXE, the distinct assessment of elastic and collagen fibers and the presence of calcification is necessary. However, in this preliminary study, we captured two-channel NLM images (SHG and TPEF), where TPEF signals originating from elastin and calcification were detected by a single, green (525/50) NDD channel. Thus, their optical differentiation was not feasible and solely was based on their morphology.

Here, we aimed to increase our imaging contrast in NLM by developing a mathematical apparatus (spectral decomposition) to create high spatial resolution, high contrast mosaic images for low concentration Phloxine B stained, deparaffinized PXE sections, where elastin, calcification and collagen is visualized with distinct colors reflecting their actual chemical concentration.

2. Methods

2.1 Sample preparation

Skin biopsy samples from PXE-affected areas of five patients diagnosed with PXE were involved in this study. The sample from patient 1 was prepared from fresh-frozen PXE sample stored in CryoMatrix at -80°C. 50 μm thick unstained cryosections were prepared for NLM imaging (PXE cryosection). The other four samples from patients 2-5 were prepared from formalin-fixed, and paraffin-embedded tissue blocks. For the NLM imaging, 20 μm thick, deparaffinized sections were used. After deparaffinization, sections were soaked in xylene for 10 minutes, then in distilled water for 3 × 3 minutes, then stained with low concentration Phloxine B (standard 1g Phloxine B / 100 ml alcohol staining solution was 15.000 times deluded in 70% (V/V) alcohol) for 2 × 10 minutes and then cleaned again with xylene for another 10 minutes. We refer to these samples as Phloxine B stained PXE sections. For conventional histopathology, sections from the same skin biopsies were stained with Weigert’s elastic (WE) and von Kossa (VK) stains. Histopathologic evaluation was performed by an expert dermatopathologist. PXE patients were diagnosed and managed at Semmelweis University, Budapest, Hungary and at the PXE National Reference Center (MAGEC Nord), Angers University Hospital, Angers, France, as a part of the protocol for phenotyping the French PXE cohort (ClinicalTrials.gov Identifier: NCT01446380). Our present study was approved by the local Ethics Committee in Budapest, Hungary (SE TUKEB no. 193/2017).

2.2 Nonlinear microscope imaging

NLM imaging was carried out with a ∼ 20 MHz repetition rate, sub-ps Ti:sapphire laser (FemtoRose TUN LC GTI, R&D Ultrafast Lasers Ltd., Budapest, Hungary, (R&D Ltd.)) operating at around 805 nm [11]. To capture images, an Axio Examiner LSM 7 MP laser scanning two photon microscope system (Carl Zeiss, Germany) with modified detection optics was used with a 20x water immersion objective (W-Plan – APOCHROMAT 20x/1.0 DIC (UV) VIS-IR, Carl Zeiss, Germany). Individual image size was 420 × 420 μm². Further details about the imaging setup can be found in Ref. [6]. TPEF and SHG signals were collected by two NDD detectors and visualized by the ZEN 3.0 SR software (Carl Zeiss AG, Germany). 460/50 nm (cyan) and 590/45 nm (orange) bandpass filters were used to capture TPEF signals of elastin and calcification, respectively, whereas for SHG, a 405/20 nm (violet, displayed as magenta) bandpass filter was used. Mosaic images were captured with the use of a custom-made stepping motor system (Mosaic, R&D Ltd.). The acquired TPEF and SHG images were processed and assembled into three-color mosaic images with ImageJ v1.46 software (NIH, Bethesda, MD, USA).

2.3 Nonlinear optical signal normalization

Our Zeiss LSM 7MP microscope was equipped with a set of emission filters, which were placed in front of the two NDD detectors in pairs. For selective SHG imaging of collagen, we must use a relatively narrowband bandpass filter in order to minimize disturbing TPEF signals that arise from other tissue components. In practice, we use a 405/20 filter for SHG detection, which limits the two-photon excitation wavelength in the 790-830 nm range. During our measurements, laser central wavelength was set in the 800-810 nm range, depending on the actual settings of the birefringent tuning element and a Gires–Tournois interferometer type dispersion compensation element, which slightly shifts the operation wavelength depending on its actual work-point setting [11]. For TPEF detection, we can use 460/50 nm (cyan), 525/50 nm (green) and 590/45 nm (orange) bandpass filters. In order to find the highest signal level for each tissue components and reach highest selectivity of imaging, first we measured the (TPEF or SHG) optical signal level for collagen, elastin and calcification in each spectral detection bands (violet (displayed as magenta), cyan, green and orange).

For high quality imaging, one has to adjust the gain (e.g. PMT voltage) of the NDD detectors in order to avoid saturation of the detector or too low signal levels. To be able to compare the optical signal levels at different gain settings, we have made a calibration for the gain: by using a fixed value of excitation power and pulse duration, we have measured the average optical signal level (of a fluorescent microscope slide) as the function of gain described by the I_average(gain) function. We have found that it can be approximated by a

2.4 Determination of transfer matrix and three-channel decomposition

Overlapping emission spectra result in correlated detected signals, where signal intensities appearing in different detection channels are proportional to concentration of a given tissue component. We have developed an algorithm for decomposition of mixed TPEF and SHG signals and deduced contribution of each substance.

Briefly, first we acquired TPEF signals with cyan and orange emission filters and SHG signal with a violet filter from 132 × 132 μm² regions with the highest SHG (collagen), TPEF_cyan (elastin) and TPEF_orange (calcium deposits) signal intensities, respectively (I₄₀₅, I₄₆₀, I₅₉₀). Selected regions correspond to areas where concentrations of collagen, elastin and calcium deposits are the highest, respectively. I₄₀₅, I₄₆₀, I₅₉₀ are linear combinations of light emission from tissue components described by their concentrations (collagen (C_Co), elastin (C_E) and calcium deposits (C_Ca)). Contribution of each tissue component to measured SHG and TPEF intensities is described by matrix coefficients a_i,j. In each region, a 3 × 3 pixel area with the highest intensity value pixel was identified. Average I₄₀₅, I₄₆₀, I₅₉₀ signal intensities were measured within these three areas, respectively. Measured average intensities were then normalized to their maximum. These computed intensity ratios were used as matrix coefficients in transfer matrix [A]:

The measured optical signal vector $\underline I$ (at each pixel) can be calculated as a corresponding concentration vector $\underline C$ multiplied by transfer matrix [A], i.e. $\underline I = \underline{\underline A} \underline C$. From this, an inverse matrix of [A] can be defined ([A]⁻¹). After calculation of matrix elements of [A]⁻¹, relative concentrations of tissue components (at each pixel) can be deduced such as $\underline C = {\underline{\underline A} ^{ - 1}}\underline I$. Image post-processing was applied to each pixel (x,y) of our three-channel images $\underline I (x,y)$ in order to get concentration maps $\underline C (x,y)$, using the ImageJ v1.46 software (NIH, Bethesda, MD, USA). Decomposed single channel concentration maps have been merged and assembled into three-color mosaic images.

3. Results

3.1 Color vector representation of nonlinear optical signals

In Table 1, normalized optical signal levels of each tissue component in different detection channels are displayed for a PXE cryosection (left) and Phloxine B stained, deparaffinized sections (right). For the latter ones, data is expressed as mean and standard deviation of these values. We used these data in the following for color vector representation of optical emission spectra from different tissue components.

Table 1. Sets of color vector coordinates describing signal intensities of collagen, elastin and calcium deposits in PXE cryosections and Phloxine B stained PXE sections. Detection channels are supplied with bandpass filters 405/20 (magenta), 460/50 (cyan) and 590/45 (orange), respectively. Color highlighting indicates the detection channel in which a given compound emits the highest nonlinear optical (SHG or TPEF) signal. For the Phloxine B stained deparaffinized PXE sections, data is expressed as mean ± standard deviation of measured values.

View Table

We observed that collagen has a strong SHG signal that can be detected in the violet channel, as expected. In other detection channels, collagen emits a low level of TPEF signal both in case of cryosections and Phloxine B stained PXE sections. Similarly, elastin and calcification has a low TPEF signal in the violet channel. In case of elastin and calcification in the PXE cryosections, however, we have found that both of them have a broad, overlapping emission spectrum all over the cyan-green-orange (435-610 nm) spectral range.

When we used color vector representation of nonlinear optical signal emission of the tissue components (Fig. 1.), we found that direction of the color vectors of elastin and calcification nearly matches in cryosections (see Fig. 1, green and red solid vectors), which means that these two components cannot be distinguished based on their TPEF signal. For the green (525/50 nm) detection channel, we experienced the same (data not shown). We also found that using a dark-red filter (650/50 nm) we can still detect elastin, but the signal level is too low for quality imaging. For calcification, we could not detect any signal in this dark-red spectral range (data not shown). In case of the Phloxine B stained, deparaffinized section the color vectors of elastin and calcification are different (see Fig. 1, dashed green and red vectors). The color vector of elastin has a similarly high value in the cyan channel, but considerably decreased in the orange channel. On the other hand, color vector of calcification exhibits a considerable shift towards the orange spectrum and decreased in the cyan channel.

 

Fig. 1. Color vector representation of normalized nonlinear optical signals measured in three detection channels for the three tissue components. Solid vectors refer to the PXE cryosection, dashed vectors refer to the mean of the color vector coordinates of the Phloxine B stained PXE sections. Colors indicate the different compounds: collagen (blue), elastin (green) and calcium deposits (red). Axis labels indicate color encodings and central wavelengths of the different bandpass filters in front of NDD detectors. Vector coordinates correspond to values listed in Table 1.

Download Full Size | PDF

3.2 Choosing emission filters for three-channel PXE measurements

Based on the results in the previous section, we have chosen a cyan (460/50 nm) filter for NLM imaging of elastin and an orange (590/45 nm) filter for the detection of calcium deposits. For collagen detection, an SHG (405/20 nm) filter was applied. Accordingly, two sets of measurements were performed for each sample, one with a SHG filter paired with the orange filter and one with a cyan filter paired with the orange filter. Images with the orange filter were then used to check spatial overlapping of the images and to compare intensity levels during the measurements. For spectral decomposition, pairing the cyan filter with the orange filter assured that corresponding two images were taken exactly from the same physical locations.

3.3 Inverse matrices and three-channel decomposition

In frozen PXE sections, we found that emission spectra of elastin and calcification can hardly be distinguished in a four color (viola/cyan/green/orange) vector space. However, by applying a low concentration Phloxine B solution after deparaffinization, elastic fibers and calcification could be optically distinguished by their fluorescence spectra: elastin exhibits a relatively higher TPEF signal in the cyan, whereas calcium deposits have relatively high TPEF signal in the orange detection channel. Reference images for determination of inverse matrix coefficients are shown in Fig. 2. Linear combination of signal intensities detected in a single channel was described in Section 2.4. To exemplify, by substituting calculated coefficients into Eq. (1), we obtained transfer matrix [A] for our Phloxine B stained PXE sample from patient 2 (Eq. (2)), and then derived the inverse matrix [A]⁻¹ (Eq. (3)).

 

Fig. 2. Illustration of three-channel spectral decomposition for the Phloxine B stained PXE sections. (A) First column: representative area after spectral decomposition in patient 2 in the three different emission channels, respectively (405/20 nm, 460/50 nm and 590/45 nm from top to bottom). Color intensities in these channels were modified according to their pertaining coefficients. These images reveal the relative amounts of collagen, elastin and calcification, respectively. Second column: Composite images of the three channels from the first column unprocessed (top), after spectral decomposition (middle) using inverse matrix shown in Eq. (3)., and RGB color coding (bottom) after decomposition. (B) Representative composite images of the three channels from patient 3 (left column), patient 4 (middle column) and patient 5 (right column) unprocessed (top), after spectral decomposition (middle) using their corresponding inverse matrix and RGB color coding (bottom) after decomposition. Red: calcification, green: elastin, blue: collagen. White arrows: collagen fibres, red arrows: elastic fibres, yellow arrows: calcification. Individual image size is 132 × 132 μm².

Download Full Size | PDF

As the SHG signal of collagen is measured with a narrowband bandpass filter, both elastin and calcification TPEF results in a minimal contribution to SHG signal. However, collagen does exhibit low TPEF signal that appears in the cyan and orange channels. Even in case of Phloxine B stained sections, calcification and elastin has wide, overlapping TPEF spectra that manifests in relatively high coefficients in both cyan and orange channels. Executing image post-processing using the retrieved inverse matrix resulted in spectrally decomposed images, where crosstalk arising from the other components is considerably reduced and hence optical signals, which originate from each tissue component, are well separated. Figure 2(A) displays individual channels and composite of reference images before and after decomposition in patient 2. Figure 2(B) shows reference images before and after decomposition in patients 3-5. Furthermore, we have also applied an artificial blue, green, red (RGB) color encoding for collagen, elastin and calcification for better optical contrast of these tissue components, which also enhanced mapping of co-localization of mineralization and elastic fibers.

3.4 NLM imaging and histopathology

Composite (three-color) mosaic image of the fresh-frozen PXE section from patient 1 is shown in Fig. 3(A) without any modification or spectral decomposition. In this sample, elastin and calcification appear in the same greenish color due to the addition of the cyan color of elastin and orange color of calcification. Phloxine B stained PXE section from patient 2 unprocessed can be seen in Fig. 3(B). The effect of Phloxine B staining is manifested in that the color of calcification and elastin is diverged, although spectral overlap is still present. The effect of spectral decomposition of Fig. 3(B) is demonstrated in Fig. 3(C). One can observe that spectral decomposition of this mosaic image enabled us to separate collagen and elastic fibers. Furthermore, the contrast between the TPEF signal of elastin and calcification is considerably improved and areas, which are affected by calcification, are clearly circumscribed. The papillary dermis was not affected. Elastic fibers in the mid-dermis are fragmented and polymorphous, as shown by the TPEF_cyan channel. Calcium deposits revealed by the TPEF_orange channel occupy large areas in the mid-dermis and dislocate clumped elastic fibers. Finally, in order to visualize calcified elastic fibers, we applied artificial red, green and blue colors to display calcification, elastin and collagen, respectively, after decomposition (Fig. 3(D)). With standard histopathology, the epidermis, the papillary dermis and the deep layers of the dermis was unaffected in all PXE patients. With the WE stain, fragmented, clumped and mineralized elastic fibers were displayed (Fig. 3(E)). VK staining revealed mineral deposits and calcified elastic fibers (Fig. 3(F)).

 

Fig. 3. Composite of two-photon excitation fluorescence (TPEF) and second-harmonic generation (SHG) mosaic images of fresh-frozen PXE-affected skin section of patient 1 (A) and Phloxine B stained, deparaffinized PXE-affected skin sections of patient 2 (B-D) and histopathological images of patient 2 (E-F). Excitation wavelength is 805 nm. Cyan (460/50 nm) and orange (590/45 nm) emission filters were chosen to detect high portion of TPEF signals originating from elastin and calcification, respectively, with maximum contrast for spectral decomposition in case of Phloxine B stained, deparaffinized PXE sections. SHG signal of collagen was spectrally selected by a narrowband 405/20 nm filter and displayed by magenta color. Note that keratin has a broad TPEF signal appearing in the orange, and to some extent, in cyan channel. Individual images of 420 × 420 μm² were assembled into mosaic images of PXE-affected skin sections. (A) Three-channel detection, composite mosaic image of TPEF_cyan, TPEF_orange and SHG signals in case of fresh-frozen PXE section from patient 1. (B)-(D): Composite mosaic images of a Phloxine B stained, deparaffinized PXE section from patient 2: (B) Three-channel detection, composite mosaic image of TPEF_cyan, TPEF_orange and SHG signals. (C) Three-channel composite of TPEF_cyan, TPEF_orange and SHG mosaic images after spectral decomposition using the inverse matrix described in Eq. (3). (D) Artificial red, green and blue colored image of mosaic image 3C (red: calcification, green: elastin, blue: collagen). Scale bar displays 400 μm. Representative histology images of the mid-dermis of PXE-affected skin of patient 2, stained with Weigert’s elastic (WE) (E) and von Kossa (VK) (F) staining. Scale bars display 200 μm.

Download Full Size | PDF

4. Discussion

Histological alterations in PXE can be observed not just in affected areas, but also in clinically healthy appearing skin, and also in patients without characteristic skin signs [12]. As subtle skin changes seen in histopathology of PXE often precede organ damage, the assessment of skin status should be of primary importance [3].

In our present work we have investigated fluorescent properties of the main tissue components of the dermis of PXE affected skin samples - collagen, elastin and the disease indicating calcification. We found that in fresh-frozen PXE sections, similarly to unstained, deparaffinized sections, elastin and calcification display a broad, overlapping emission spectrum, which is in agreement with the results described in the literature [13]. When we introduced a normalized 3D color vector representation of their emission spectra, the distinction of elastin and calcification for high contrast imaging was found to be impossible. By applying low concentration, (hence low absorption) Phloxine B staining after the deparaffinization process, however, we observed that their spectral distribution considerably differs from each other. As a consequence, these two tissue components can be optically distinguished by laser scanning NLM, even without damaging the tissue due to laser light absorption in stained sections.

Ectopic calcification in PXE mainly consists of calcium hydroxyapatite [2]. According to previous studies, the pure chemical form of this component does not absorb nor emit fluorescent light, as they typically appear as white material [14]. In PXE, calcification appears mostly within the elastic fibers, and gives fluorescence signal since it contains fragmented elastic fibers. Thus, calcium deposits in PXE exhibit similar fluorescent spectral properties to that of the elastic fibers.

We found, however, that Phloxine B was taken up by calcium deposits in a relatively high concentration. Phloxine B is a derivative of fluorescein, a red acid dye, that binds to acidophilic structures [15]. As calcium is a strong base ion, it will bind Phloxine B. In our study, Phloxine B proved to be a proper agent to increase the contrast between elastin and calcification. As a consequence, it has become possible to optically distinguish elastin and calcification in Phloxine B stained PXE sections by laser scanning NLM. In the literature eosin was already used for fluorescent labelling of elastin fibers and calcification also exhibited brown-colored fluorescence in PXE skin sections [16]. As Phloxine B and eosin have comparable chemical properties [15,17], similar results might be achieved with the application of low concentration eosin solution during the deparaffinization process, although this needs further investigations. In order to increase our imaging contrast in NLM, we applied an inverse matrix based method and carried out spectral decomposition to obtain three-color mosaic images, where the distinct colors represent the actual presence and concentration of collagen, elastin and calcification. Comparing NLM results with standard histopathological stainings, we found that NLM images clearly showed histological hallmarks of PXE. Furthermore, distinguishing between elastic and collagen fibers, that was challenging on the WE stained sections, became simple with NLM due to the created contrast between these tissue components. Moreover, we found the NLM images are richer in detail concerning the fine structure of calcium deposits and fine structural alterations of the fragmented elastic fibers.

In conclusion, we have examined nonlinear optical signals of elastin and calcium deposits by TPEF and of collagen by SHG applying different bandpass filters. By determining the contribution of these different tissue components for SHG and TPEF signals at each detection channels, we spectrally separated their overlapping emission spectra. Our results suggest that this method has important incremental features compared to conventional histopathology since spectrally decomposed NLM images can be directly used for quantitative analysis of ectopic calcification. Furthermore, fine structures and alterations to elastic fibers can be better identified due to higher chemical selectivity and higher spatial resolution. In the future, three-channel imaging and color vector representation of calcification in ex vivo deparaffinized samples might be used for quantitative analysis of PXE-affected skin during the diagnostics and severity assessment. This method could be utilized also in case of other diseases with cutaneous calcification.