Cornea is a transparent tissue covering the front surface of the eye. It protects the eye from foreign matter and delivers two thirds of the eye’s refractive power. To remain clear, cornea is devoid of any blood vessels, and is nourished by a tear film and an aqueous humor filling the chamber behind it. The arrangement, spacing, and size of collagen fibers in corneal stroma are critical in maintaining corneal transparency.1 Microbial keratitis, ocular herpes, a range of corneal dystrophies, and complications due to refractive surgeries are among many causes of corneal scars leading to permanent vision loss. Therefore, being able to visualize individual cells and collagen matrix at submicron resolution noninvasively in vivo is important in understanding the biology of normal corneas and evaluating their transformation during disease and wound healing.
In the last few years, a combination of nonlinear signals such as intrinsic two-photon fluorescence (TPF) and second harmonic generation (SHG) has been employed to form high contrast images of cells and extracellular matrix. 2, 3, 4, 5, 6, 7, 8, 9, 10 Previous nonlinear optical microscopy (NLOM) studies hinted at the possibility of employing endogenous fluorescence11 and second harmonic generation (SHG) signals to analyze corneal health,12 particularly during/after intrastromal corneal surgery.13, 14 Yet currently, no clear relationship between the observed intrinsic NLOM signals and a histologically documented corneal irregularity has been established. The sensitivity and value of these signals in elucidating cellular/extracellular matrix changes in cornea and other thick tissues resulting from aberrant gene expression are also largely unknown.
We explored the potential of NLOM to detect genetically induced phenotypic alterations in the transgenic mouse model of corneal epithelial dysfunction. The modulation of Clim [co-factor of LIM (a combination of first letters of Lin-11 (C. elegans), ISL1 (rat), and Mec-3 (C. elegans) gene names) domain proteins, a diverse group of transcription factors] activity in epithelial tissues of mice using a keratin 14 (K14) promoter resulted in widespread epithelial abnormalities throughout the animal. The generation and characterization of these mice is discussed in a separate paper.15 The K14 dominant negative (DN)-Clim mice develop corneal abnormalities such that the corneas undergo an ultimate transformation into epidermis-like structures. These corneas become highly neovascularized, inflamed, and exhibit down-regulation of the corneal-specific protein keratin 12, concomitant with up-regulation of epidermal markers keratin 10, filaggrin, and loricrin in the epithelium.15 Eighty percent of mice are blind due to corneal opacities. In severe cases, the affected mice exhibit abnormal overgrowths of the stratified squamous epithelium, which are visually detectable and confirmed by histology.
In this work, endogenous cellular fluorescence and second harmonic signals from collagen showed dysplasia in the squamous epithelium, irregular collagen arrays in the stroma, and a compromised posterior endothelium in K14-DN-Clim mice corneas. A two-photon ratiometric redox fluorometry method based on cellular fluorescence from reduced nicotinamide adenine dinucleotide (NADH) and oxidized flavin adenine dinucleotide (FAD) indicated about 30% higher metabolic activity in the endothelial layer compared to an epithelial layer in both normal and aberrant corneas. Overall, the K14-DN-Clim mouse is an excellent model for NLOM to explore genetically induced corneal abnormalities ranging from cellular dysplasia to stromal matrix structural irregularities and neovascularization—processes commonly observed in cancer and fibrosis.
Materials and Methods
All animal procedures were performed in accordance with an animal protocol approved by the University of California at Irvine. Mice strain CB6F1 was used in all experiments. Transgenic mice were generated by microinjection of the suitable plasmid into fertilized eggs implanted into healthy CB6F1 animals. These mice expressed the transgene in the basal cell layer of the epidermis, the outer root sheath of hair follicles, the basal cell layer of neonatal corneal epithelium, and limbal cells of adult corneal epithelium.16 RNA isolation for microarray experiments were performed as described in Ref. 17. The microarrays data were analyzed using the software Cyber-T accessible through a Web interface18 (http://www.genomics.uci.edu/software.html) using a stringent statistical cut off value of 0.01.
All animals studied were euthanized by asphyxiation with . The ages of the mice were from two months up to one year. Mice eyes were removed, placed into 0.9% saline solution, and imaged immediately following euthanasia. The corneas were not dissected out of the eyes and were amenable to imaging for one or two hours. About 20 normal and 15 K14-DN-Clim mice eyes of matching ages were imaged.
Second Harmonic Generation/Two-Photon Fluorescence Imaging and Spectra
The images and spectra were obtained using a combined two-photon fluorescence (TPF)/second harmonic(SHG) setup previously described. 9, 12, 19, 20, 21, 22, 23, 24, 25, 26, 27 The excitation was linearly polarized at . Focusing objective (Zeiss, water immersion, NA 1.2) was used to collect spectra and images. For photomultiplier tube (PMT) ratio imaging, a dichroic filter and bandpass filters [ ; NAD(P)H and ; flavoprotein] were used in front of corresponding PMTs. To minimize background signal from the laser, a broadband filter was placed in front of the spectrograph. The time for a typical scan was one to three seconds. The spectra were collected from the single optical sections with the typical integration time of ; monochromator slit was , and the laser beam was scanned continuously during spectral acquisition. The spectra were not corrected for the instrument response. The thickness of the corneas was measured with a long working distance focusing objective (Zeiss, water immersion, NA 0.8).
After remaining in saline for approximately during the NLOM imaging, specimens were fixed in 10% buffered formalin (Sigma-Aldrich, St. Louis, MO) overnight, dehydrated by an ethanol gradient, and embedded in paraffin wax (Fisher Scientific, Hampton, NH). The six micron sections closest to the center of the eye were stained with hematoxylin and eosin.
Second Harmonic Generation/Two-Photon Fluorescence Signals and Spectra
Figure 1 shows a hematoxylin and eosin (H and E) stained, en face histology. Epithelium (the outer region of corneal tissue, top) is a layer about five cellular layers deep and contains 98% of nerve endings, making cornea very sensitive to pain.28 Stroma (a layer behind epithelium) is the main component of cornea. Its composition is water, collagen fibers, and proteoglycans. The arrangement, spacing, and sizes of collagen fibers are critical in maintaining corneal transparency.29 Endothelium (a single layer of cells located between the stroma and the aqueous humor) mainly functions to regulate water balance. All the topographic features seen in NLOM scans (Fig. 2 ) have excellent coregistration with (H and E) stained sections; however, structural organization of collagen stroma evident in SHG/TPF images is not apparent in H and E processed tissues. Regardless of animal age, the corneal size was estimated to be around from NLOM data, with the epithelial layer making up the first , stroma with parallel arrangement of collagen layers comprising the next (Fig. 2), and endothelial layer at . The endothelial layer was the last layer amenable to imaging in reflectance, and no further SHG or TPF signals were detected with a physiology long-working-distance objective (Neofluar, model 440090, Zeiss, water immersion, NA 0.8). The spectrum obtained for the normal mice eyes is a broad fluorescence band (Fig. 3 ). The two peaks at around 450 and possibly result from NADH and fluorescence, respectively (spectra not shown). The PMT ratio of the peaks is 30% higher for the epithelial versus endothelial layer, which we believe is due to differences in metabolic activity between the two cellular layers. The spectrum of stroma is dominated by a second harmonic signal from collagen and was described earlier.12
Figure 4 shows a hematoxylin and eosin (H and E) stained, en face histology section for the moderately affected K14-DN-Clim mice cornea. The high resolution NLOM images (Fig. 5 ) are highly effective in highlighting the dysplastic areas within the epithelial layer and irregular organization of collagen arrays within the stroma. In several severe cases, partial absence of the endothelial layer (differently sized cells show up in the same optical section) was also noted and confirmed by histology. When present, the endothelial cells were irregularly shaped. The spectrum for the K14-DN-Clim mice cornea is broad, with the PMT ratio of peaks being 30% higher for the epithelial versus endothelial layer, similar to normal corneas.
This work describes normal and pathological corneas in mice using reflectance NLOM (TPF and SHG combined). The epithelial layer dysplasia, irregularities in a collagen structure within stroma not apparent in H and E processed tissues, and compromised endothelial layer are reproducibly detected at high magnification in K14-DN-Clim mice corneas and validated with a conventional histology.
In many tumors, transformed epithelial cells are found to “prime” or provoke the microenvironment of extracellular matrix, which in turn undergoes dynamic alterations leading to structural changes, severe immune responses, and the formation of new blood vessels.30 During this process, the fibrillar organization of extracellular matrix proteins such as fibronectin, collagen I, III, and V, elastin, and proteoglycans is modified,31 and their overall production is increased.32 In breast, the oncofetal extracellular matrix is usually thick, flaky, and largely disorganized.33, 34 Altering the expression of the Clim gene in the epithelial tissues of mice resulted in corneas undergoing a continuous inflammatory/stromagenic/wound healing process concomitant with changes in the epithelial and endothelial cell layers. In line with up-regulation of many inflammatory markers15 and neovascularization in K14-DN-Clim mice corneas, based on mRNA microarray data (Fig. 6 ), procollagen III production is increased at least 2.5 fold compared to a healthy mouse. We therefore propose that the irregular appearance of the stromal collagen in the NLOM images of K14-DN-Clim mice corneas is due to reorganization of extracellualr matrix and deposition of smaller, disorganized fibrils perhaps associated with collagen III. While collagen I is the major fibrillar component of the cornea, it is not actively synthesized in healthy adult corneas. Collagen III is commonly overproduced in response to inflammation/wound healing in skin35 and in cornea36, 37, 38, 39 by the stromal keratocytes that transdefferentiate into fibroblasts and myofibroblasts.40 It is deposited with disorganized architecture, providing a framework for collagen I fibers41 observed in the later stages of wound repair.
Most K14-DN-Clim mice develop obvious macroscopic evidence of corneal abnormalities around six weeks of age. However, partial loss of Clim function leads to a complex phenotype with a severity of the condition varying greatly among individuals of the same age. To overcome the challenge of detecting early stages in K14-DN-Clim mice corneal pathology, we are currently implementing live animal imaging.
Two-photon ratiometric redox fluorometry based on cellular fluorescence from reduced nicotinamide adenine dinucleotide (NADH) and oxidized flavin adenine dinucleotide (FAD) has been proposed as a tool to study mitocondrial energy metabolism.42 Only reduced NADH and oxidized FAD states of these cofactors are significantly fluorescent and linked through the NADH-Q-reductase and various dehydrogenases43 in a process known as oxidative phosphorylation (OXPHOS). They will respond differently to changes in mitochondrial metabolic states, and a ratio will be a measure of mitochondrial energetics. Several earlier studies used fluorescence of NAD(P)H11, 44 and flavoprotein to address the mitochondrial energy metabolism in the biological systems. 45, 46, 47, 48, 49, 50, 51, 52, 53 Our two-photon ratiometric imaging results indicate a higher metabolic activity in the endothelial layer of cornea compared to an epithelial layer located further away from metabolites. An observed difference of 30% might be even greater in vivo, where the endothelial layer conducts metabolite exchange and actively regulates water balance at physiological temperatures. Earlier observations of flavoprotein/NADH and NAD(P)H signals obtained from single photon excitation of perfused corneas detected higher metabolic rate for endothelium relative to that of epithelium.11, 44, 47 Tissue handling associated with ex-vivo measurements is likely to influence our two-photon ratiometric redox fluorometry measurements by inducing a shut down of endothelial layer function in about an hour at room temperature. Therefore, we are currently investigating the utility of functional corneal imaging on live animals.
High-resolution reflectance-based NLOM methods offer novel approaches to probe the downstream effects of gene expression and their influence on phenotypes in vivo without use of exogenous dyes. In this study, the Clim gene is partially blocked in the epithelial compartment of mice tissues: however, it causes significant structural changes in corneal stroma, indicating that Clims may be involved in epithelial-mesenchymal interactions. We detect these changes at high magnification using nondestructive NLOM imaging, and on comparison to normal stromas shows that the method is effective in detailing structural pathologies in reorganized extracellular matrix. These extracellular matrix structural alterations are likely due to collagen III synthesis and deposition during healing and remodeling of corneal stroma. Because no exogenous dyes are employed, NLOM visualization can be used in selecting tissues for gene expression profiling. In addition, NLOM ratiometric redox fluorometry results provide preliminary confirmation that cellular metabolic function can be characterized and imaged in thick tissues.
Julia G. Lyubovitsky thanks George E. Hewitt and the George E. Hewitt Foundation for Medical Research for the fellowship. This work was made possible, in part, through the NIH Laser Microbeam and Medical Program (LAMMP) at the University of California, Irvine (P41-RR01192) and by the Air Force Office of Scientific Research (AFOSR), under agreement number FA9550-04-1-0101. Bogi Andersen is supported by NIH grant AR44882.