Optical coherence tomography and multiphoton microscopy offer new options for the quantification of fibrotic aortic valve disease in ApoE−/− mice

Aortic valve sclerosis is characterized as the thickening of the aortic valve without obstruction of the left ventricular outflow. It has a prevalence of 30% in people over 65 years old. Aortic valve sclerosis represents a cardiovascular risk marker because it may progress to moderate or severe aortic valve stenosis. Thus, the early recognition and management of aortic valve sclerosis are of cardinal importance. We examined the aortic valve geometry and structure from healthy C57Bl6 wild type and age-matched hyperlipidemic ApoE−/− mice with aortic valve sclerosis using optical coherence tomography (OCT) and multiphoton microscopy (MPM) and compared results with histological analyses. Early fibrotic thickening, especially in the tip region of the native aortic valve leaflets from the ApoE−/− mice, was detectable in a precise spatial resolution using OCT. Evaluation of the second harmonic generation signal using MPM demonstrated that collagen content decreased in all aortic valve leaflet regions in the ApoE−/− mice. Lipid droplets and cholesterol crystals were detected using coherent anti-Stokes Raman scattering in the tissue from the ApoE−/− mice. Here, we demonstrated that OCT and MPM, which are fast and precise contactless imaging approaches, are suitable for defining early morphological and structural alterations of sclerotic murine aortic valves.


Results
The aortic valve leaflets were separated into three previously described regions for the characterization 26 . Region 1 encompassed the free edge of the tip of the leaflet (nodulus valvulae semilunaris). Region 2 covered the middle of the leaflet, while region 3 covered the base of the leaflet, which included the aortic root, but excluded the aortic wall (Fig. 1a,b).

Determination of leaflet dimensions: OCT versus histological analysis.
The OCT cross-sections and the histological sections of the valvular leaflets matched very well (Fig. 1a,b). Depending on the tissue processing and slicing procedures, histological sections can be hampered by structural changes. The analysis of the OCT scans revealed that the ApoE −/− mice had a significantly increased mean leaflet area (69,003 µm 2 ± 7,138 µm 2 ) compared with the WT mice (53,386 µm 2 ± 4,546 µm 2 , P = 0.047, Fig. 1c). In a subsequent histological analysis, the leaflet area of the ApoE −/− mice compared with the WT mice was 41,264 µm 2 ± 8,421 µm 2 and 19,321 µm 2 ± 4831 µm 2 , respectively (P = 0.0001, Fig. 1d). The OCT scans showed that the leaflet thickness in region 1 increased to 137 µm ± 31 µm in the ApoE −/− mice from 105 µm ± 18 µm in the WT mice (P = 0.031, Fig. 1e). There were no differences in the average thickness in region 2 or region 3 between the ApoE −/− mice and the WT mice ( Fig. 1f/g, i/j). These findings correlated with the descriptions in the histological sections: the thickness of region 1 in the WT mice was wild type 58 µm ± 21 µm, while the thickness in the ApoE −/− mice was 112 µm ± 42 µm (P = 0.013, Fig. 1h). The individual absolute area values of the histologically processed leaflets were found to be 2.4-fold smaller than the area values of the OCT processed native leaflets (Supplementary Table S1).
Bleaching of the melanin. Murine heart valves contain melanocytes; the pigmentation level of murine heart valves may correlate with coat color 54 . Because of the genetic background, the aortic valve leaflets of C57BL/6 J WT mice and ApoE −/− mice are pigmented. The melanin is highly efficient at absorbing light, including UV and near infrared light and, thus, hinders MPM by absorbing both the generated signals and the excitation laser light. The latter phenomenon often leads to strong photodamage, which can include burning of the pigmented tissue. For this reason, it was necessary to eliminate the melanin before imaging by bleaching the isolated and formalin-fixed valves in a hydrogen peroxide (H 2 O 2 ) solution in water. This approach has been used for years for the histopathology of melanoma, although different protocols have been employed, which range from incubation in 10% H 2 O 2 for 24 h at room temperature 55 to incubation in 3% H 2 O 2 for 2 h at 55°C 56 . As this method had not been tested for MPM of murine valves, we investigated incubation in different concentrations (1.25% to 15% H 2 O 2 ) for 24 h at room temperature. Incubation in 2.5% H 2 O 2 was found to remove all melanin without producing significant alterations in the SHG signal of the collagen fibers or the CARS signal of the lipids (Fig. 2). Thus, this protocol was used in all of the following experiments. We found that higher concentrations of H 2 O 2 altered the tissue morphology, which was visible both in the SHG and the CARS images, while lower concentrations were not able to bleach the melanin completely.
Collagen fiber orientation. The orientation of collagen in the WT mice and the ApoE −/− mice is displayed in Fig. 5. The collagen structure appeared disordered in the ApoE −/− mice (Fig. 5b) compared with WT mice (Fig. 5a). Because the orientation direction of collagen fibers is arranged circumferential to the annulus, images of aortic valve leaflets were divided in the middle resulting in the left and right halves (Fig. 5a subregions 1 and 2). The analysis revealed that there was a significantly lower coherency in base regions 3.1 and 3.2 in the ApoE −/− mice (n = 6) compared with the WT mice (n = 6): 0.23 ± 0.03 versus 0.39 ± 0.03 (P = 0.0018, Fig. 5f) and 0.22 ± 0.03 versus 0.35 ± 0.04 (P = 0.0095, Fig. 5i), respectively. There was no difference in coherency between the WT mice and the ApoE −/− mice within the region 1 and the region 2 of the leaflet structure (Fig. 5d, e and g, h).

Lipid content.
A qualitative evaluation of lipid accumulations in the form of cholesterol crystals and lipid droplets was performed using image stacks acquired using CARS. A maximum intensity projection that displays the total amount of lipids in the leaflet regions is shown in Fig. 6.
No lipids were found in the region 1 and 2 of all analyzed WT mice. Only sporadic and very small lipid droplets were observed in four of the six WT mice in the region 3 near the annulus. In contrast, the leaflets of the ApoE −/− mice showed a massive accumulation of lipids in the form of cholesterol crystals and lipid droplets. All ApoE −/− mice displayed lipid accumulation in the region 2 and region 3. Three of the six ApoE −/− mice also www.nature.com/scientificreports/ showed lipid droplets in the region 1. From the morphology, lipid droplets often seemed associated with cells, which were identified as foam cells (Fig. 7a,b). Cholesterol crystals, identified by the crystal structures, were observed inside the cells (Fig. 7b), and accumulated in plaques within the tissue (Fig. 7c). In the latter case, a cytoplasmic or extracellular localization could not be determined from the multiphoton images.

Discussion
Currently, the only treatment for CAVD is interventional or surgical valve replacement. Potential pharmacological interventions should target early alterations, such as AVS, although this would require the precise identification of disease onset and progression. Small animal models have been used in the development of experimental therapy approaches 8 . The current methods used for the visualization of aortic valve morphology and function in animal models are limited because of the inadequate resolution, the need for markers, and the excessive histological processing 24,57 . Although some molecular imaging 58 and micro-computed tomography 59 approaches offer increased resolution and the opportunity to study pathophysiological processes in vivo, these techniques are not broadly available and are associated with high costs. OCT and MPM exhibit resolutions that are in the micrometer range; both techniques are contactless and do not use markers. OCT has been used in a study on the quantification of murine ventricular volume and mass 20 . We previously showed that 4D OCT high-speed video microscopy is suitable for measuring the valve opening area in an artificially stimulated heart model 60 . MPM has been used to characterize human explanted aortic valves 43,61 , cardiac valve allografts 62 , and tissue-engineered heart valves 63 in which the intensity of the SHG signal was used to quantify the collagen fibers 42,61 . This is the first study that examined the aortic valve geometry and structure of C57Bl6 WT mice and age-matched ApoE −/− mice using OCT and MPM. ApoE −/− mice have been used in animal models to study AVS 18,64 . We found that fibrotic thickening, collagen reduction, and lipid accumulation were significantly different in the ApoE −/− mice compared to WT mice using OCT and MPM.
The flexibility, shape, and surface structure of aortic valve leaflets are essential for the proper function and sufficient closure of the valve 65 . The base and tip regions of the leaflet are particularly prone to an instantaneous increase in tensile strength during the cardiac cycle 66 . Accordingly, we found an increase in the leaflet area and thickness of the leaflet tip in ApoE −/− mice compared with WT mice using OCT; these results were confirmed www.nature.com/scientificreports/ by histology. These findings agreed with the histological studies of Aikawa et al. 64 and Hinton et al. 26 in which the transversal and sagittal directions, respectively, were chosen for sectioning. Comparing these studies with our observations showed that different directions of sections and tissue shrinkage from essential pretreatments present major limitations to histological examinations; these limitations make it difficult to rank results from different studies. Furthermore, many studies use transversal probes 27,67 , which are associated with the risk of overestimating the extent of the leaflet structures. The degree of tissue shrinkage in our study was 40-64% based on leaflet area, which is in accordance with the previously measured values of data on murine aortic valve tissue 26 .
Our data demonstrated the superiority of OCT and an important refinement in determining leaflet thickening during the development of AVS in small animal models because it was possible to observe the native tissue in a fast, precise, and tridimensional fashion. AVS is an actively regulated cellular process 68 in which resident valvular interstitial cells differentiate into an activated myofibroblast state in response to an altered mechanical stress 69 and pro-fibrotic cytokines. Myofibroblasts produce extracellular matrix components and remodeling enzymes 70 that cause structural changes. In ApoE −/− mice on a hypercholesterolemic diet, the thickening of valve structures is associated with the presence of macrophage-rich lesions 64 . It is well known that macrophages and activated valvular interstitial cells acquire excessive levels of proteolytic enzymes, such as collagenase-1 (MMP-1), collagenase-3 (MMP-13), gelatinase A (MMP-2), and gelatinase B(MMP-9), that contribute to collagen and elastin degradation and lead to valvular remodeling [71][72][73] . www.nature.com/scientificreports/ In this study, we determined the collagen content by analyzing the acquired SHG signal in MPM and compared the results with the collagen content in equivalent histological sections. We found that the collagen content decreased in the aortic valves in ApoE −/− mice compared with the WT mice in all three of the analyzed regions. This is in accordance with the work of Trapeaux et al. who demonstrated a decreased collagen content in the cross-sections of the aortic valves from 32-week-old ApoE −/− mice that had received a high fat diet 74 . The study found that the reduction in collagen content was most pronounced in the middle region of the leaflet; this distribution pattern was confirmed in the current study. It should be noted that a lower amount of collagen was detected by MPM compared with the histological results. This may be explained by the fact that the amount of collagen was based on the fraction of sample area displaying the SHG signal or picrosirius red staining and by the inherent confocality of MPM. While the SHG signal was collected from a laser focus depth of about 2 µm, the staining signal was collected from the entire thickness of the slide by light transmission microscopy. We also studied the orientation of the collagen meshwork in the three analyzed regions and found less orientation in the base region of the ApoE −/− mice. This agreed with a study by Chu et al. 23 who found an altered orientation of collagen fibers in a hypercholesterolemic/hypertensive mice model at the base of the cusps. They assumed that, although the total collagen content in the valve was not increased or further reduced, the remodeling of the collagen contributed to the restricted valve opening. It needs to be mentioned that Chu and coworkers could only describe the orientation of the base on transversal histological sections. Using the SHG imaging, we demonstrated an analysis of collagen fibers over the entire valve dimension. This allows a more detailed analysis of the collagen meshwork.
Lipid accumulation represents another hallmark of early aortic valve lesions 75 . It is desirable to use pharmacological interventions that target lipid accumulation before there is evidence that fibrocalcific aortic valve disease has developed 57 . One study showed that the accumulation of oxidized phospholipids to large cholesterol crystal structures in mice fed a high cholesterol diet was involved in the inflammatory processes within aortic valve disease 76 . We denoted an increase in both the lipid content and the cholesterol crystal content in the base  23 . Although we did no lipid staining, we detected more pronounced holes, which resulted from cut adipocytes and lipid particles, in the middle region and especially in the annulus region of the leaflet from ApoE −/− mice. This observation is in accordance with the respective results of CARS signal analysis. Our data demonstrated that OCT and MPM are suitable for quantifying early pathomechanisms in aortic valve disease in a murine model, such as fibrotic thickening and fatty degradation. OCT and MPM showed collagen reduction, lipid accumulation, and thickening of the aortic valve structure, which correlated with the results from conventional histological examinations. We concluded that the implementation of OCT and MPM has great potential in the early diagnosis of CAVD in murine models.
Limitations. The bleaching process, necessary before MPM, made the valve tissue porous and hindered cryosectioning. Therefore the murine aortic valves were embedded in paraffin for the histological analysis and lipid staining was not possible.

Methods
Animals and tissue processing. Five-week-old female ApoE −/− (B6.129P2-Apoe tm1Unc /J) mice were purchased from Charles River Laboratories (Sulzfeld, Germany). Following 1 week of acclimatization, the mice were fed a Western-type diet (TD88137 mod.; Ssniff, Soest, Germany), which contained 21.2% total fat, 2.071 mg/ www.nature.com/scientificreports/ kg cholesterol, and 17.5% protein for 16 weeks. Aged-matched female C57BL/6JRj mice, purchased from Janvier Labs (Le Genest-Saint-Isle, France), served as controls and received a standard diet (V1534 R/M-H; Ssniff, Soest, Germany) for 16 weeks. After the feeding period, the mice were sacrificed by cervical dislocation, and the ascending aorta, including the aortic valves, was excised. The aortic valve structure was opened using a longitudinal incision between the right and left coronary leaflet. This preparation was used to visualize native, non-fixed aortic valve tissue under ex vivo conditions with OCT, as described in a previous study 60 . Prior to the MPM scans, the aortic valve tissue was fixed in 4% formaldehyde and bleached after incubation in 2.5% H 2 O 2 for 24 h at room temperature. Subsequently the aortic valve tissue was embedded in paraffin for histological analysis (see Supplementary Fig. S1). The animal research ethics committee of the Technische Universität Dresden and the Regional Council (Regierungspräsidium) Dresden approved the experiment according to the institutional guidelines and the German animal welfare regulations (AZ: 24-9168.24-1/2013-10).
Optical coherence tomography (OCT). The OCT system used in this study is a custom-made frequencydomain spectrometer-based device with a 12-kHz A-scan rate and a lateral and axial resolution of 11 µm and 6.4 µm, respectively as described in detail in a previous study 60 . and analyses. In OCT, the lateral and axial resolution are not directly linked. While the axial resolution depends on the used wavelength, bandwidth, and the spectrometer design, the lateral resolution depends on beam shaping and focusing. Therefore, the first step in image processing involved scaling the image data stack to achieve uniform voxel sizes in all three dimensions. Furthermore, as an optical imaging modality, OCT measures the optical path length. For the measurement of geometric lengths, such as leaflet thickness, the optical refractive index of the tissue must be taken into account. Although the exact refractive index of aortic valve tissue is not known. Since the heart valves generally correspond to an endothelial duplication, the refractive index could be in the range of 1.33 to 1.44, but nothing is known about the change of refractive index due to the different age, diet and stage of disease in wildtype and ApoE knockout mice. Any assumption about this lead to additional inaccuracies in thickness measurements. Therefore, we assumed a mean refractive index of 1.33 (water) because all thickness measurements were compared relative to each other. After scaling, the voxel size was determined to be 2.255 µm in each direction. The tissue thickness was measured in a cross-sectional view perpendicular from the base to the tip (Fig. 8). To enhance the tissue contrast and the signal-to-noise ratio, the intensity values of ten adjacent cross-sections were summed. The leaflet area was measured by manual segmentation in the same cross-section.
Multiphoton microscopy (MPM). The MPM system has been described in detail elsewhere 80 . In short, the MPM system consists of 2 picosecond lasers emitting at 781 nm and 1,005 nm coupled to a laser scanning microscope. The SHG signal generated by the first laser can be detected in transmission mode with a band-pass filter centered at 390 nm and a bandwidth of 18 nm. The CARS signal of the symmetric stretching vibration of methylene groups, which corresponded to the Raman band at 2,850 cm −1 , was detected in transmission mode using a band-pass filter centered at 647 nm and a bandwidth of 57 nm. Additionally, the two-photon excited fluorescence (TPEF) signal was acquired in the 500-550 nm range in the reflection configuration, which worked as a reference tissue morphology by visualizing the cellular structures. The excitation laser beams were focused with a C-Apochromat 32 × /0.85 objective.
For imaging, the isolated and prepared aortic valve was placed in a drop of phosphate-buffered saline between two coverslips with the leaflet up and the aorta below. Stacks of images were acquired from the tip, the middle region, and the base of the right posterior/acoronary leaflet ( Fig. 1a and b) for the subsequent evaluation of the collagen and lipid content from the SHG and CARS images, respectively. Each stack of images spanned the entire thickness of the leaflet and was composed of 14 images; the z-step was varied depending on the local thickness of each leaflet in order to acquire only the leaflet and not the underlying aorta. The same acquisition parameters, such as laser power and photomultiplier gain, a pixel dimension of 0.28 µm, and 65,536 Gy levels (16-bit image), were used for all images.

MPM image analysis.
The open-source software Fiji 79 was used to analyze all images. The collagen content was based on the percentage of tissue area that had a SHG signal. First, a mean filter with a radius of 1 pixel was used to smooth the SHG images; this filter smooths the image by replacing each pixel with the neighborhood mean, where the size of the neighborhood is specified by the radius in pixels. Smoothing was followed by a linear contrast enhancement (min-max with 0.4% saturated pixels). The area in the image with a signal intensity above a threshold of 20,000 was measured and used to determine the fraction of tissue containing collagen; the original images of a representative subset were used to define the threshold value. A visual inspection of the binary images took place after thresholding. The procedure was applied to all the images in each stack and was used to calculate the percentage fraction of the area with a SHG signal for each image. However, this area was not always www.nature.com/scientificreports/ representative of the collagen content because the cholesterol crystals also had a SHG signal. Therefore, a correction was performed to subtract the contribution of the cholesterol crystals, which was based on the fact that the crystals also generated intense CARS signal. For this purpose, all lipid structures were identified on the CARS images. First, a mean filter with a radius of 1 pixel was used to smooth the CARS images, followed by a linear contrast enhancement (min-max with 0.4% saturated pixels). A background subtraction was performed using a radius of 30 pixels. A threshold was then applied to create a binary image using the triangle method for automatic threshold setting. Finally, the binary CARS images were multiplied with the binary SHG images. In this way, only the structures that were both SHG-active and CARS-active (i.e., the cholesterol crystals) were retained. The percentage fraction of the area was then measured and subtracted from the percentage fraction of the area with the SHG signal for each image. A flow-chart of the whole analysis procedure with example images is displayed in Fig. 9. Finally, a mean value for each leaflet region was obtained by averaging the percentage fraction of the area with the corrected SHG signal of all the images in the stack. The degree of organization of collagen fibers was retrieved using the plugin OrientationJ for ImageJ. The coherency parameter was used to quantify the orientation order of the fiber 81 . The coherency in an image has a value of one if the fibers are perfectly oriented in one direction and a value of zero for an isotropic structure without any orientation. Because the orientation direction of collagen fibers was different in the two halves of images, which were acquired from the midline of the leaflet, OrientationJ was applied to the left and right halves of the images (subregions 1 and 2).  OCT scan of the right coronary (triangle), acoronary (star) and left coronary leaflet (pentagon). Further image processing and measurements were performed on acoronary leaflet only (b) and (c)). The base of the acoronary leaflet is marked with the star and the tip of the nodulus is marked with the rhombus. To further support visual image correlation, the asterisk (a) and (e)) marks the pin used to keep tissue in place. www.nature.com/scientificreports/ 12,061; Morphisto GmbH, Frankfurt, Germany). All stained sections were acquired using a Zeiss Axio Scan. Z1 slide scanner (magnification 10 ×) and were manually segmented to determine the leaflet area and thickness using ZEN 3.1 software (Carl Zeiss Microscopy GmbH, Jena, Germany).

Measurement of leaflet area and leaflet thickness and assessment of collagen content in histological samples.
The acoronary aortic leaflets (valvula semilunaris posterior) were used for all histological analyses. The section with the maximum dimension of leaflet structure was determined after careful evaluation of the serial sections for each sample. The leaflet area and thickness were manually segmented using ZEN 3.1. software (Carl Zeiss Microscopy GmbH, Jena, Germany). The leaflet thickness was determined using a grid pattern in the three regions 26 : Region 1 was at the free edge of the tip (nodulus valvulae semilunaris); region 2 was the middle of the leaflet and region 3 was at the hinge or the base of the semilunar valve and included the aortic root, but excluded the aortic wall. A line was scattered perpendicular to the leaflet surface in the above mentioned three regions (Fig. 1a,b). The total collagen content was assessed in the positive picrosirius red-stained sections using the ImageJ plugin Colour Deconvolution and a manually defined threshold 82 . Data represented the percentage of valve area that displayed positive staining.
Statistics. Data were presented as the means ± standard deviation (SD) and the normal distribution was tested with the Kolmogorov-Smirnov test. The Student's unpaired one-tailed t-test was used to compare the two experimental groups when the data had a normal distribution; otherwise, the Mann-Whitney U test was www.nature.com/scientificreports/ employed. P-values < 0.05 were considered to be significant. GraphPad Prism 6.0 software (GraphPad Software, Inc., La Jolla, USA) was used for the statistical analysis and visualization.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request. www.nature.com/scientificreports/