1.

Introduction

Corneal dysfunctions are the second major cause of blindness worldwide, surpassed only by cataracts.¹ Precise and early diagnoses to enable adequate treatments of corneal pathologies are essential to reduce visual impairment. Still widely used are imaging techniques, such as the slit lamp microscope, developed over a century ago, to evaluate the anterior segment of the eye.² The rather limited resolution with no cellular information has been surpassed by specular³ and confocal microscopy.⁴ Specular microscopy visualizes the morphologies of healthy and disease-affected endothelial cells,⁵ whereas confocal microscopy opens the access to the morphology of cells from all cornea layers and nerve densities. It is widely used in clinical practice to diagnose cornea diseases and to monitor tissue recovery following surgery.⁶ In combination with further corneal diagnostic methods like anterior segment optical coherence tomography and corneal topography, these techniques have greatly improved corneal diagnosis. However, they fail to provide information on the cell metabolism and disregard the morphological organization of the major layer of the cornea—the stroma.

A corneal clinical imaging modality based on two-photon imaging (TPI) could overcome these limitations and provide information based on earlier detectable signs of pathology, thereby further improving the diagnostic of corneal pathologies. TPI can be used to evaluate the metabolic state of cells by fluorescence lifetime imaging microscopy (FLIM).⁷ With FLIM, the free and protein-bound components of the metabolic cofactors can be quantified, and thereby the oxidation–reduction ratio and consequently the metabolic activity of cells can be indirectly determined.⁸ Furthermore, based on second-harmonic generation (SHG) generated by structures with noncentrosymmetric molecular organization,⁹ the corneal stroma collagen fibers can be visualized, and alterations induced by pathologies to the stroma microstructure can be assessed. The feasibility of TPI to image and characterize the cornea has been previously demonstrated ex vivo^10–15 and in the cornea of rabbits in vivo.¹⁶

In this proof of principle study, we demonstrate the feasibility of using TPI to distinguish between nonpathological (NP) and diseased corneas, based on cell morphology, cell metabolism, and the stromal autofluorescence (AF) lifetime and fibers organization. For that purpose, TPI was used to evaluate differences between NP corneas and samples affected by keratoconus (KC), a disorder that affects the corneal shape, characterized by a progressive corneal thinning, stepping, and formation of a cone.^5,17 Individual samples affected by Acanthamoeba keratitis (AK), a corneal pathology caused by an amoeba that causes ulceration and necrosis of the epithelium and the stroma,^5,18,19 and stromal corneal scars (SCS), which can be formed during atypical wound healing following cornea injury, infection, or refractive surgery,²⁰ were also evaluated.

The cells’ AF intensity images revealed alterations of the corneal epithelial cell morphology, whereas the AF lifetime imaging of the metabolic cofactor nicotinamide adenine dinucleotide (NADH)/nicotinamide adenine dinucleotide phosphate (NADPH)—NAD(P)H was used to assess changes to the cells’ metabolism. Alterations of the corneal stroma structural organization as well as to the AF lifetime were also investigated. To the best of our knowledge, this is the first time that TPI of the corneal epithelium and stroma is jointly used to characterize these pathologies. We demonstrate that TPI can provide information on the cornea that is inaccessible by current clinical devices. This information could be used for the diagnostic of corneal diseases, as well as for a better understanding of their mechanisms of action.

2.

Methods

2.2.

Two-Photon Microscope

A near-infrared (NIR) femtosecond (fs) five-dimensional (5-D) multiphoton laser-scanning microscope with axial, lateral, temporal, and spectral resolutions of (1 to 2) $\mu \mathrm{m}$, 300 nm, 100 ps, and 12.5 nm, respectively, was used for image acquisition. The system (Fig. 1) consisted of an inverted microscope with a galvanometric scanner unit, a beam expander, a dichroic mirror, and a $40\times$ NA 1.3 oil immersion objective for laser light focusing and signal collection in reflection geometry. Sample excitation was accomplished using a mode-locked titanium-sapphire laser (Integral Pro 400; FemtoLasers Produktions GmbH, Vienna, Austria) generating $\sim 10\text{-}\mathrm{fs}$ pulses at a frequency of 85 MHz. The laser emission with an M-shaped spectral profile was centered at $\sim 800\mathrm{nm}$, i.e., it contained two main peaks at 775 and 830 nm and had a full width at half maximum (FWHM) of about 95 nm (Fig. 1), allowing the simultaneous excitation of multiple endogenous fluorophores. The group velocity dispersion of the optics was precompensated using chirped mirrors so that laser pulse durations were below 15 fs behind the microscope objective. Up to 13 images with $256\times 256\text{pixels}$ with a scanning speed of 7.4 s per frame were summed to collect enough photons for data analysis. The corneas were imaged using mean laser powers between 10 and 15 mW.

Fig. 1

Schematic setup of the 5-D laser scanning microscope used for TPI of the cornea. The system has axial, lateral (three-dimensional), temporal (four-dimensional), and spectral (5-D) resolutions of (1 to 2) $\mu \mathrm{m}$, 300 nm, 170 ps, and 12.5 nm, respectively. The laser spectrum is shown. The pulse width at the sample was below 15 fs.

AF and SHG photons were detected between 400 and 600 nm using a 16 channels photomultiplier tube (PMT) detector with 12.5-nm spectral resolution (16PML-PMT, Hamamatsu, Japan) associated with a time-correlated single photon counting SPC-150 module (Becker & Hickl GmbH, Berlin, Germany).

3.

Results

3.1.

Pathology-Induced Changes of the Corneal Epithelium

Pathology-induced morphological alterations to epithelial cells of different types (superficial, wing, and basal cells) were observed in AF images (Fig. 2). In pathological samples, the cells are irregularly shaped within each layer. Furthermore, an enlargement of the basal cells was observed for KC and SCS corneas when compared with NP samples. In KC corneas, an elongation of the cells was also observed (Fig. 2). AF from the cell membranes was detected for the AK cornea. Moreover, regions without corneal basal cells appear (Fig. 2, red arrows). In the SCS cornea, bright autofluorescent spots, possibly related to debris deposition, are visible throughout the epithelium (Fig. 2, white arrows). A decrease in the epithelial thickness was observed for all pathological conditions. The epithelial thickness decreased from approximately $(48\pm 6)\mu \mathrm{m}$ in NP corneas to $(31\pm 9)\mu \mathrm{m}$, $(25\pm 5)\mu \mathrm{m}$, and $(30\pm 4)\mu \mathrm{m}$ in KC, AK, and SCS, respectively.

Fig. 2

AF intensity images of the epithelial layers of NP human corneas and corneas with KC, AK, and SCS. The different cell types of the epithelium (superficial, wing, and basal cells) and their respective depths are shown for each condition. In AK, red arrows point to regions where basal cells are absent. In SCS, white arrows point to autofluorescent deposits. $\text{Scale bar}=30\mu \mathrm{m}$.

Alterations induced by the pathologies can be highlighted in the AF lifetime images and their corresponding lifetime distributions (Fig. 3). The epithelial cells in NP corneas have a homogenous AF lifetime distribution, i.e., all cells in the field-of-view have similar ${\tau }_m$ values, whereas in KC, AK, and SCS corneas heterogenous AF lifetime distributions can be seen (Fig. 3). Cells with altered/abnormal morphology in KC have a longer ${\tau }_m$ than the surrounding cells with normal morphology. In the AK cornea, the membrane AF, which is absent in NP and the other pathological conditions, has a longer mean lifetime than the AF of the cytoplasm of the same cells. In the SCS cornea, the bright autofluorescent spots (debris) have a homogenous and short lifetime (Fig. 3).

Fig. 3

AF lifetime images and corresponding lifetime distributions of nonpathological (NP) corneas, and corneas diagnosed with keratoconus (KC), Acanthamoeba keratitis (AK), and stromal corneal scars (SCS). For all conditions, the lifetime distributions of the entire field-of-view and regions of interest (ROIs) including single cells are shown in dashed line (blue) and in full line (black), respectively. For KC, AK, and SCS corneas, the lifetime distribution of ROIs of cells with altered/abnormal morphology, the cellular membrane, and debris, respectively, is additionally shown in dotted line (red). The ROI boundaries are indicated in the lifetime images with black and yellow lines. $\text{Scale bar}=30\mu \mathrm{m}$.

The NAD(P)H AF lifetime was used to assess alterations induced by the pathologies to the epithelial cell metabolism (Fig. 4). KC corneas have significantly higher ratios between free ($a_1$) and protein-bound ($a_2$) components of NAD(P)H ($a_1/a_2$ ratio) than the NP corneas (Fig. 4). For AK and SCS corneas, a tendency for an increase in the metabolic cofactor $a_1/a_2$ ratio is also observed (Fig. 4).

Fig. 4

Epithelial cell NAD(P)H free to protein-bound ($a_1/a_2$) ratios of nonpathological and diseased corneas. Values represent $\text{mean}\pm \text{standard deviation}$. ****Statistically significant at $p<0.0001$.

3.2.

Pathology-Induced Changes to the Cornea Stroma

SHG images were used to assess morphological alterations induced by pathologies to the stroma, the main layer of the human cornea. Figure 5 shows SHG images of the anterior stroma at multiple depths for NP, AK, KC, and SCS corneas. Qualitative differences between all pathologies can be observed (Fig. 5). The FFT analysis of the SHG images was used to quantitatively assess alterations of organization of collagen fibers [Fig. 6(a)]. For KC, a significant decrease of PP was observed, indicating a decrease in the orientation of the collagen fibers. A tendency for a decrease was also observed for AK and SCS samples [Fig. 6(a)].

Fig. 5

SHG images of nonpathological human corneas and corneas with AK, KC, and SCS at depths of 0, 20, 45, and $5\mu \mathrm{m}$. $\text{Scale bar}=30\mu \mathrm{m}$.

Fig. 6

(a) Degree of orientation of the stroma collagen fibers in nonpathological human corneas and corneas with AK, KC, and SCS assessed using the PP. (b) Mean stroma AF lifetimes of nonpathological and disease-affected corneas. Values represent $\text{mean}\pm \text{standard deviation}$. **Statistically significant difference ($p<0.01$); ****Statistically significant at $p<0.0001$.

Along with changes of the collagen stroma organization, significant differences were observed for the AF lifetime ${\tau }_m$. It was significantly decreased in KC corneas when compared with NP corneas [Fig. 6(b)]. SCS corneas also show a tendency for a decrease, while AK shows a tendency for an increase in the stroma ${\tau }_m$ [Fig. 6(b)].

4.

Discussion

The early diagnosis of corneal pathologies is an important precondition for a good prognosis. Until now, advances of cornea imaging enabled earlier and earlier detection. The development of confocal microscopy provided additional information to the clinical features of AK, particularly the in vivo visualization of amoeba cysts, and aided AK diagnosis.¹⁹ However, AK is still often misdiagnosed.⁵ Corneal topography greatly improved the detection of KC; however, the results can occasionally be unclear.⁵ Moreover, the biological mechanisms of this disease are not fully understood.¹⁷ Other tools are required to understand them and provide earlier diagnosis. SCS can be formed following cornea injury, infection, or even after refractive surgery and leads to cornea opacification, which can be observed through slit lamp biomicroscopy.²⁰ Instruments capable of detecting structural modifications prior to cornea opacification are needed.

NP human corneas as well as corneas diagnosed with KC, AK, and SCS were characterized with high-resolution and without external labels using TPI. We could characterize not only the specific morphological alterations induced by these pathologies but also alterations to the epithelial cells’ metabolism as well as to the stroma AF and structural organization.

The effect of different pathologies on the epithelial cells’ morphology was revealed using the cells’ AF intensity (Fig. 2). Spectral analysis of the cornea epithelial cells’ AF revealed NAD(P)H as the major sources of AF. Other sources of AF, including flavins, were negligible possibly due to the low concentration of this fluorophore in the cornea.²⁴ Thus, a spectral detection range suitable for NAD(P)H detection was chosen. This metabolic cofactor, mainly distributed in the mitochondria and the cell cytoplasm,⁸ allows visualization of single cells and their morphological characteristics (Fig. 2). Furthermore, the tissue AF lifetime provides additional insight on the tissue alterations due to pathologies (Fig. 3).

In corneas diagnosed with KC, an increase in the epithelial cell size accompanied by an elongation of the cells was observed (Fig. 2). This is consistent with the reported anatomic modifications induced by KC to the epithelial layer.¹⁷ The AF lifetime images also showed an increase in the cells’ heterogeneity, and cells with different NAD(P)H ${\tau }_m$ were visualized (Fig. 3). As anticipated, a decrease in the corneal epithelial thickness was also observed.^5,17

Interestingly, a unique cellular membrane AF was seen in the case of AK cornea (Fig. 2). The cellular membrane AF exhibited a longer fluorescence lifetime than the remaining cellular structures, which provides an additional means of discrimination (Fig. 3). This alteration observed in the epithelial layer is likely related to the gradual infiltration of pathogenic organisms into the cornea. AK results from an infestation by the free-living amoeba, which starts with the attachment of the amoeba to the corneal epithelial cells and leads to degradation of the epithelial layer and stroma and infiltration of deeper layers of the cornea.^5,18 A decrease in the corneal epithelial thickness along with regions where basal cells were absent was also observed, possibly due to cell degradation promoted by the local triple therapy consisting of propamidine isetionate, polyhexanide, and antibiotic eye drops (Fig. 2, red arrows). However, further studies with more AK samples are needed to understand better the AF modifications.

In the SCS cornea, a slight increase in the epithelial cell sizes was observed, particularly for basal cells. A possible explanation is that the epithelial cell layer can smooth the anterior surface of the eye by augmented proliferation and fill gaps induced by the scarring process.²⁵ Furthermore, autofluorescent spots, possibly due to deposition of debris, were observed in the epithelial cells (Fig. 2). These debris AF had a short mean lifetime and homogenous distribution, which can be used to better discriminate between debris and cellular components (Fig. 3).

Based on the AF lifetime of the metabolic cofactor NAD(P)H, the metabolic activity of epithelial cells was assessed for NP as well as for KC, AK, and SCS corneas. Significant differences of the NAD(P)H $a_1/a_2$ ratios were observed between NP and KC corneas (Fig. 4). In KC, NAD(P)H $a_1/a_2$ ratios of epithelial cells were significantly higher than in NP. Since oxidation and reduction are catalyzed by enzymes (proteins), the ratio between free ($a_1$) and protein-bound ($a_2$) components of NAD(P)H provides an indirect measure of the oxidation–reduction ratio and, therefore, of the cell metabolism. An increase in the NAD(P)H $a_1/a_2$ ratio indicates an increase of free NAD(P)H and/or a decrease of the protein-bound component. This has been previously associated with a decrease in the metabolic activity of cells.²⁶ Although an increase in the NAD(P)H $a_1/a_2$ ratio was also observed for AK and SCS, statistical significance could not be assessed due to the low number of samples with these diseases. This indicates that pathological conditions may induce alterations to the metabolic activity of epithelial cells.

Changes of the stroma AF and SHG signals resulting from pathological conditions were evaluated. Alterations of the structural organization of the corneal stroma due to AK, KC, and SCS were observed based on SHG imaging [Figs. 5 and 6(a)]. The orientation of the collagen fibers was analyzed by an FFT-based method. In KC corneas, a significant decrease in the degree of the collagen fiber orientation was observed [Fig. 6(a)]. Corneal ectatic disorders such as KC are known to induce alterations to the cornea stroma. For KC corneas, a decrease in the stroma thickness and a significant decrease in the number of collagen lamellae at the cone have been reported.^5,17 Hence, alterations of the collagen fibers organization could be anticipated [Fig. 6(a)]. SHG imaging has already shown alterations of the collagen lamellae distribution patterns induced by KC²⁷ and their levels of interweaving,²⁸ as well as their orientations.^29,30

The SHG images of AK corneas show visible alterations when compared with NP corneas (Fig. 5). Moreover, a decrease of PP was observed for this pathology, indicating a tendency for a decrease in the collagen fibers organization [Fig. 6(a)]. In AK, the contaminating organisms degrade the corneal stroma to facilitate their penetration into deeper layers.^5,17 Therefore, a rearrangement of the collagen fibers with a lower degree of organization could be anticipated. The feasibility of AF and SHG imaging to assess morphological alterations induced by bacterial and fungal infections of the corneas has also been reported.^31,32

SCS are formed during abnormal wound healing. When the cornea is injured, the keratocytes are stimulated and migrate to the affected area. These cells are usually in a quiescent state and maintain the corneal stroma through the production of extracellular matrix components including collagen.⁵ However, during wound healing, keratocytes may differentiate into myofibroblasts, which are themselves more opaque than keratocytes and produce high quantities of disorganized extracellular matrix components finally leading to cornea opacification.²⁰ Therefore, alterations of the organization of collagen fibers were expected [Fig. 6(a)]. The feasibility of TPI to monitor the scar tissue formation during wound healing has been demonstrated.³³

In addition to SHG, the collagen and keratocytes in the corneal stroma generate AF signals. In this study, differences of the stroma mean AF lifetimes were observed between NP and pathological corneas as well as between different pathologies [Fig. 6(b)]. This shows that the stroma AF can provide additional information on this layer and help to better discriminate between NP and pathological as well as between different pathologies.

5.

Conclusion

Using TPI, we characterized ex vivo human corneas based on the morphology and metabolism of the epithelial cells as well as on the stroma AF and the structural arrangement of its collagen fibers. Based on these characteristics, which are, except for the cell morphology, inaccessible with current clinical devices, non-pathological and pathological corneas could be discriminated. For KC corneas, the characteristic morphological alterations were observed. We demonstrated that TPI can provide additional information and may eventually enable a better evaluation of this corneal pathology. Although alterations of the morphology of cells as well as indications of altered cell metabolism, corneal stroma AF lifetime, and structural organization were observed for AK and SCS corneas, the analysis of further samples is necessary to be able to draw definite conclusions.

The results indicate that TPI can provide additional information on the cornea to improve the diagnosis of corneal pathologies as well as to understand the pathophysiology of these conditions. To the best of our knowledge, this was the first time that epithelial cells and stromal AF lifetime, combined with stroma SHG imaging, were used to evaluate alterations induced by these pathologies. In the future, we intent to use mobile TPI systems in a clinical environment to evaluate human cornea samples immediately after keratoplasty and set new diagnostic markers and improve and anticipate the diagnosis of the corneal diseases.