Second harmonic generation imaging reveals asymmetry in the rotational helicity of collagen lamellae in chicken corneas

: High tensile strength and optical clarity are unique properties of the cornea. These features are dictated by the three-dimensional architecture of corneal lamellae. Therefore, understanding the microscopic details of the cornea’s structural organization may contribute to the development of artiﬁcial cornea for the treatment of corneal diseases. In this study, the combination of forward second harmonic generation (SHG) microcopy and fast Fourier-transform based image analysis was used to characterize the depth-dependent superstructure of chicken corneal stroma. Our results show that from the surface, adjacent lamellae of anterior chicken cornea lamella rotate in a counterclockwise direction, and the same rotational helicity is observed in left and right corneas. Furthermore, the overall average rotational pitch of lamellae is 0.92 ± 0.11 degree/µm which persists for 176 ± 14 µm in the anterior stroma. As depth further increased, the rate of lamellar rotation decreases. Upon reaching posterior stroma, lamellar orientation remains constant. Throughout the stroma, collagen lamellae in chicken rotate a total of 169 ± 21 degrees. The lack of lamellar rotation in posterior stroma suggests that packing eﬃciency cannot be used to explain the helicity of depth-dependent rotation of anterior stroma. In addition, although the right cornea has a higher rotational pitch (0.95 ± 11 vs 0.90 ± 10 degrees/µm) and thinner anterior stroma (173 ± 13 vs 179 ± 14 µm) than the left cornea, the two eﬀects cancel each other out and result in similar total angular rotation of anterior stroma (161 ± 23 and 165 degrees ± 21). Finally, our observation of a total angular rotation of 169 ± 21 degrees shows that within experimental error, chicken cornea lamellae rotate around 180 degrees or half of a complete turn. Additional studies are needed to arrive at an explanation of chicken superstructure in three dimensions. is 169 ± 21 degrees. The ﬁndings in this study may provide useful bioengineering clues in the further design of artiﬁcial cornea.


Introduction
Spatial arrangement of collagen and its interaction with cells in tissue is important in wound healing, tissue engineering and embryonic development [1][2][3][4]. In the case of cornea, stromal collagen is organized into lamellae that act as a highly transparent, protective layer for the eye. However, how the organization of stromal collagen contribute to the cornea's biomechanical and optical properties is not completely understood. Structurally, the cornea is organized into five layers: epithelium, Bowman's layer, stroma, Descemet's membrane and endothelium. Stroma, the collagen-rich matrix derived from the periocular sulcus, is the thickest layer. Keratocytes located between lamellar collagen fibers secrete collagen and proteoglycans to produce a crystalline architecture which maintains corneal transparency [5]. During wound healing, keratocytes are transformed into fibroblasts and myofibroblasts [6]. Injuries caused by conditions such as infection and trauma are usually accompanied by corneal scarring and can affect corneal clarity. Corneal swelling in conditions such as Fuchs malnutrition contribute to increased scattering and can also lead to vision degradation. Finally, corneal thickness is an important index in the preservation of structure integrity and clinical evaluation of diseases [7][8][9][10]. Therefore, understanding the structural features of cornea in three dimensions is invaluable for basic research and clinical evaluation of corneal diseases.
In order to study corneal structure, a variety of imaging techniques such as optical microscopy, second harmonic generation (SHG) imaging, electron microscopy and X-ray scattering have been used to study normal and abnormal samples in humans and other species [11][12][13][14][15][16]. For example, SHG imaging showed that cross-linking did not significantly affect structures of corneas with well-organized lamellae [17]. Moreover, optical and electron microscopies were used to study corneal development in the chick model and it was found that rotation of collagen lamellae evolves with maturation. [16]. Electron microscopy was also used to measure lamellar number in human corneal specimens [8]. In the optical domain, confocal imaging was used to measure corneal thickness and keratocyte density in the rabbit model [7]. Among these techniques, electron microscopy is laborious in that sample preparation is time-consuming and the information available from the tissue slice is limited to a small volume. X-ray scattering has its limitations in that the structural information obtained represents an integrated average over the entire stromal thickness.
In recent years, SHG microscopy has emerged as an effective technique for studying collagencontaining connective tissues [18][19][20][21]. Second order non-linear polarization properties in rat tissues have been studied with polarization SHG microscopy [22]. In the case of cornea, SHG imaging was applied to studying structural features such as the density and width of collagen lamellae in human [10]. The effects of keratoconus, mechanical pressure, and photothermal modification on corneal lamellar structure have also been investigated [23][24][25][26][27][28][29][30][31]. Finally, corneal lamellar structure in different species including chicken, fish, bullfrog and others were imaged and compared with SHG imaging [32,33]. Although earlier studies demonstrated a persistent rotational pattern of corneal lamellae in some species, detailed measurement of the lamellar helicity between left and right corneas were not compared. Clearly, a detailed characterization of cornea's structure in three dimensions is important for understanding self-assembly of extracellular matrix and contribute to improved methodology in corneal tissue engineering.
In this study, we used a combination of a second harmonic generation (SHG) microscopy and fast Fourier transform-based image analysis to image and quantify the three-dimensional structure of cornea in the chicken model.

Preparation of chicken cornea specimens
Corneas from adult chicken of the Arbor Acres breed were acquired from the animal farm of National Taiwan University. Four pairs of left and right corneas from male chicken were used in this study. The diameter of each specimen was close to 8 mm. Prior to the imaging experiment, corneas were cut into strips along the temporal-nasal direction and marked with scissors on the temporal side to identify the specimen orientation. Each chicken cornea strip was stored in a home-made, specimen-containing chamber with 10% formalin. During imaging, the samples were enclosed in a confinement chamber attached to a concave cover slide and sealed with a cover glass and high-vacuum grease. While performing the imaging experiments, the formalin solution was replaced with PBS.

SHG image acquisition
Cornea specimens each approximately 10 × 10 × 0.7 mm 3 in dimension were imaged using a homemade, multiphoton imaging system based on a commercial inverted microscope (TE2000U, Nikon, Japan). The excitation source was a titanium-sapphire laser (Tsunami, Spectra Physics, Mountain View, CA) pumped by a diode-pumped, solid-state (DPSS) laser system (Millennia Pro, Spectra Physics).
The excitation wavelength used was 780 nm. Focusing of the excitation source was achieved through an objective (S Flour, 20x, NA 0.75, WD 1.0, Nikon). The corresponding axial resolution was approximately 5.7 µm. Since the lamellae thickness is around 2-3 µm, we found that two adjacent lamellae can be visualized. The focusing objective is used to collect epi-illuminated signals which were filtered by appropriate band-pass filters. Specifically, the collected signals were spectrally resolved by a combination of dichroic mirrors (405dcxr, 530dcxr Chroma Technology) and filters (HQ390/10, HQ540/70, HQ630/70). In this manner, spectral signals centered at 390, 540, and 630 nm can be respectively collected. In the transmission geometry, forward second harmonic generation (FSHG) was collected by a second objective (S Flour, 20x, NA 0.75, WD 1.0, Nikon) and spectrally resolved by a dichroic mirror (535dcxr) and filter (HQ390/10). Each image 227 × 227 µm 2 in area was scanned at 512 × 512 pixels resolution. All signals were detected by single-photon-counting photomultiplier tubes (R7400P, Hamamatsu, Hamamatsu City, Japan).
Image stacks were collected at five different positions, each along the temporal-nasal axis and at a distance 2 mm apart ( Fig. 1A, B). At each position, scanning was performed in the anterior-posterior direction. To locate the boundary of the corneal stroma, signals from collagen and corneal cells were simultaneously collected. In this manner, forward SHG signal was used to visualize collagen fibrils and auto-fluorescence indicates the presence of corneal cells. A representative scanned volume is illustrated as an image stack (Fig. 1C). Variation of corneal stroma structure with depth was illustrated as an image series. As can be seen, the angle between two lamellae is close to being orthogonal and remains constant throughout stroma (Fig. 1D).

Determination of imaging positions
For the five imaged positions, Positions 1 and 5 were chosen to be close to the interface between cornea and limbus, where Position 1 was closest to the temporal side and Position 5 was nearest to the nasal side. Once the Positions 1 and 5 were determined, Position 3 (between Positions 1 and 5) was marked as the central position of the cornea. Positions 2 and 4 were then determined, as the middle points between 1 and 3, and between 3 and 5, respectively. The distance of the adjacent layer in the image stack was 5 µm.

Fiber orientation analysis
After acquisition of depth-resolved SHG images, fast Fourier transform (FFT) analysis was performed on individual images and used to determine and track the rotational pattern of corneal lamellae in three dimensions. This process is illustrated in Fig. 2(A). The SHG image of corneal lamellae (Fig. 2(A), 1) was fast Fourier transformed into the corresponding frequency domain image (Fig. 2(A), 2). Next, thresholding techniques were applied to reveal the dominant collagen lamellar orientation (Fig. 2(A), 3). The value of thresholding used in the entire stack was between 140∼160 in the 8-bit intensity image. The FFT filtering method filtered out large structures (shading correction) and small structures (smoothing) of the specified size by Gaussian filtering. The optimized FFT images were then processed with custom-programs based on MATLAB R2010a (Mathworks, Cambridge, UK). Principal directions of lamellar orientations can be determined through an angular meter which is the frequency magnitudes plotted as a function of orientation (Fig. 2(A), 4). The distribution of lamellar orientation is shown in the histogram (Fig. 2(A), 5). ImageJ (National Institutes of Health, Bethesda, MD, USA) was used for reading raw data and reconstructing 3D image stacks. Following FFT analysis, the two intensity peaks (marked by circles) in the first two quadrants of the histograms (0-180°) were used to determine the principal directions of collagen lamellae. At succeeding depths, angular orientations of two major directions were determined relative to those at the previous depth. In this manner, depth-dependent variation of dominant lamellar orientations can be determined. In the case that total angular rotation is greater than 180°, our approach would limit a total angular change of the lamellae to be between 0-360°( Fig. 2(B)).

Statistical analysis
From the measured angular orientations of collagen lamellae, we determined corneal thickness and rotational pitches for left and right corneas. Furthermore, Student's t-test was used to determine the correlation of the data between left and right corneas. The resultant p values are shown in Tables 1-4.  Fig. 3 is a representative example of depth-dependent SHG images, corresponding FFT results, and lamellar orientations of a chicken cornea specimen. Depth-dependent variations in principal axes of collagen lamellae obtained from SHG images and FFT analysis are shown in Fig. 3(A) and 3(B). Our results show that starting from the corneal surface, angular orientations of the two primary axes vary monotonically in the anterior region ( Fig. 3(C)). To analyze our results further, we obtained the depth-dependent profiles of lamellar orientations at Positions 1-5 for both left and right corneas (Fig. 4). With the results normalized to stroma thickness, we found that at each position, collagen lamellae rotate in a counterclockwise direction in the anterior stroma to approximately half of the cornea thickness. As depth further increases, the rotational pitch of lamellae gradually decreases until the lamella orientation remains constant. Therefore, changes in the rotational pitch of corneal lamellae can be categorized into three zones (Fig. 5(A)). Zone 1 (monotonic change in rotational pitch) is anterior stroma, Zone 2 is the transition region (slowing down of lamella rotation), and Zone 3 (constant lamella orientations) is posterior stroma.

Zonal definition and rotational pitch of corneal lamellae
To further analyze our results, we quantified zonal thicknesses of corneal collagen lamellae and rotational pitch of the anterior stroma. In determining Zone 1, we noticed that angular orientation as a function of depth fits well to a line. Therefore, we applied linear fitting to the lamellar angle as a function of depth for 40% of the stroma thickness from the corneal surface. When the difference of measured and fitted angles is more than 10 degrees at a given depth, that position would be defined as the end of Zone 1 and start of Zone 2, the transition region. To determine the other end point of Zone 2 and start of Zone 3, another linear model is used to fit angles measured from other 40% of the stromal thickness near the posterior stroma. When a 10-degrees difference is found between measured data and fitted result, the corresponding position is defined as the ending position of Zone 2. The rotational pitch of the anterior stroma is obtained from the slope of fitted angle vs. depth curve. The average rotational pitch at each position for the two principal axes of collagen lamellae are shown in Table 2. In addition to the thickness measurements, we also determined the ratios of the thickness of each zone to the overall stroma thickness (Table 3). We found the overall thickness for Zones 1, 2, and 3 are 176 ± 14, 31 ± 3, and 170 ± 22 µm, respectively.
By averaging the rotational pitch at each position, we found that the rotational pitches for the left cornea is 0.90 ± 0.10 degree/µm and 0.95 ± 0.11 degree/µm for the right cornea. These results differ from the 0.68 degree/µm result obtained in our earlier study [32]. A number of factors may contribute to such differences. First, the adult corneas in the previous study, were obtained from a local market without full knowledge of the species. In comparison, the corneas in the current study were obtained from a single species (Arbor Acres). Secondly, a more significant factor may be the state of the cornea tissues. For the cornea specimens obtained from the market, we did not know the precise time of chicken slaughter. In addition, since these chickens are for consumption, they are treated, possibly with hot water, for feather removal. In comparison, for the chicken obtain from our university farm, we made efforts to obtain the cornea specimens quickly after slaughter. In addition, no additional treatment was performed on the chicken. Therefore, tissue degradation and feather removal procedure from our previous study may contribute to differences in measured rotational pitches. Furthermore, by multiplying the rotational pitch and the thickness for Zone 1, we found that collagen lamellae in chicken cornea rotate 161 ± 23 degrees for the left cornea and 165 ± 21 degrees for the right cornea. However, if we computed the total angular change of Zones 1 and 2, collagen lamellae in chicken cornea rotate 166 ± 22 degrees for the left cornea and 173 ± 19 degrees for the right cornea. The total rotation including Zone 1 and Zone 2 is 169 ± 21 degrees. The results of left and right corneas were tabulated in the Table 4 and the overall values for all corneas were listed in the Table 5. Previous studies had shown that posterior stroma was more ordered, better hydrated, more easily swollen, and has a lower refractive index than the anterior stroma [34]. However, the boundary between the anterior and posterior stroma was not clearly defined. In this study, we quantified the orientation of collagen lamellae as a function of depth. Our findings show that the counterclockwise rotation of corneal lamellae exists in both left and right corneas. In addition, the overall average rotation pitch, total Zone 1 rotational angle, and total Zone 1 thickness are 0.92 ± 0.11 degrees/µm, 163 ± 22 degrees, and 176 ± 14 µm, respectively. The total rotation on average between anterior and posterior stroma is 169 ± 21 degrees. The findings in this study may provide useful bioengineering clues in the further design of artificial cornea.

Conclusion
Previously, SHG imaging and Fourier analysis have been used to study the structures of various connective tissues such as trachea, ear and cornea [35,36]. Backward SHG combined with Fourier and aspect ratio analysis was used to investigate corneal collagen orientation [37]. Since forward SHG provides orientation information of corneal collagen fiber, the use of aspect ratio is not needed to define the two principal axes of cornea collagen lamella [28]. Furthermore, earlier electron and optical microscopy studies demonstrated changes in chick lamellar rotation during development [16]. Unlike previous studies, our work on chicken cornea superstructure found the same counterclockwise rotational helicity with depth persisted in both left and right corneas. The reflection symmetry of physical features typically observed along an organism's sagittal plane is not conserved in the case of depth-dependent helicity of chicken corneal lamellae. While it was suggested to us that geometric packing of collagen molecules may favor such orientation, packing efficiency cannot be used to explain the lack of lamellar rotation in posterior stroma. In addition, we defined and quantitatively determined physical parameters that characterized lamellar geometry. Specifically, we found that the right cornea has a higher rotational pitch (0.95 ± 11 vs 0.90 ± 10 degrees/µm) and thinner Zone 1 (173 ± 13 vs 179 ± 14 µm). Interestingly, the two effects cancel each other out and result in similar angular rotation of Zone 1 (161 ± 23 and 165 degrees ± 21). Furthermore, the p values in Tables 1-4 show, except for Zone 1, point 2 where the p value is 0.047 (Table 1), all other p values are larger than 0.05, suggesting that differences of the measured thicknesses of different zones and rotational pitches between left and right corneas are insignificant. Finally, our observation of a total angular rotation of 169 ± 21 degrees (Table 5) shows that within experimental error, chicken cornea lamellae rotates around 180 degrees or half of a complete turn. As the current study focused on lamellar orientation and thickness of chicken cornea, additional studies are needed to arrive at an explanation of chicken superstructure in three dimensions.

Disclosures
The authors declare that there are no conflicts of interest related to the article