Quantitative and Rapid In Vivo Imaging of Human Lenticular Fluorescence

Abstract

Purpose: To quantitatively investigate the chemical origins of near-UV excited fluorescence in the crystalline lens, and demonstrate the potential usefulness of a rapid and noninvasive diagnostic approach for screening and monitoring of lens damage.

Methods: Anterior segment UV fluorescence imaging was applied to a population of 30 healthy adults, ages 18 to 64 years. Absolute fluorescence intensities and intensity ratios were compared across the population as a function of age. Fluorescence quantum yield (FQY) was calculated from imaging results based on a previous radiometric characterization.

Results: Typical FQYs at 365 nm excitation are approximately 0.2% for healthy adults. Intensity and FQY were observed to increase significantly with age, consistent with ex vivo and confocal microscopy studies. The ratio of blue to green fluorescence is strongly correlated with FQY and age, suggesting that both increases in fluorophore concentration and changes in composition occur with age. Fluorescence data is quantitatively and qualitatively consistent with pyridine nucleotides in young adults, and changes with age are consistent with formation of β-carbolines or advanced glycation end products. Intralens variation is consistent with increased oxidation or glycation in the lens nucleus relative to the cortex.

Conclusions: Lenticular fluorescence can be measured rapidly, accurately, and quantitatively in vivo which provides a spatially resolved, quantitative measure of lens chemistry, including damage from oxidation and glycation. Careful interpretation of fluorescence intensities and intensity ratios can provide chemical insight into lens health. Anterior segment UV fluorescence imaging can thus serve as a useful tool for screening, monitoring, and research of lens damage and cataract formation.

Despite cataracts being responsible for more than 90 million cases of moderate to severe visual impairment or blindness worldwide in 2020,¹ there remain significant gaps in our understanding of the chemical pathways of cataract formation. Although it is generally believed that oxidation plays an important role in cataract formation,² a variety of other processes can cause damage in the lens proteins that accumulate over a lifetime.³ Lens aging, damage, and cataract formation are multifactorial cumulative processes. Some risk factors, such as exposure to UV light, are well-known and associated with the formation of radical oxygen species. A variety of other risk factors have been identified as well, including smoking and high cholesterol,⁴ however, given the multifactorial nature of cataract it is difficult to interpret these data. 

Although there has long been a desire to develop pharmaceutical treatments for cataracts, efforts have not yet been successful.⁵ The development of practices to reduce cataract risk on an individual level could also assist in reducing the disease burden.⁶ A significant limitation of both approaches is the lack of a well-understood chemical mechanism of lens damage and cataract formation. Without knowing the damage mechanism, it is not clear how to prevent or repair damage. This limitation is further exacerbated by limitations in diagnostic tools. There is currently no accepted method to observe the “invisible” damage to the lens in vivo as it occurs over a lifetime. 

Here, we use anterior segment UV fluorescence imaging to investigate the chemical origins of visible fluorescence in human ocular lenses, which is believed to be related to oxidation and other damage mechanisms,⁷ in a population of healthy human volunteers. The method has been applied to model eyes recently,⁸ including a detailed safety analysis. The technique is rapid, low cost, and amenable to automation, making it suitable for routine ophthalmic imaging without requiring additional training or analysis. This method also provides a quantitative measurement of lens chemistry that in principle can be used to detect lens oxidation and damage over an individual lifetime. We propose the method to be suitable to track oxidative damage accumulation in the lens, and thus further probe the chemical pathways of cataract formation. This manuscript reports on the initial application of the technique to a set of 30 healthy human adults, and provides a detailed discussion of the measurement principle in relation to lens chemistry. 

Methods

Human lens fluorescence was investigated in vivo in 30 healthy adults (22 female and 8 male). The study was approved (approval #HUM00210338) and overseen by the Institutional Review Boards of the University of Michigan Medical School (IRBMED). All activities were conducted exclusively on the University of Michigan’s College of Engineering Campus in Ann Arbor, Michigan, and conformed to the ethical standards of the IRBMED. All participants provided written informed consent before participation. Participants were ages 18 to 65 years with no acute eye disease and were recruited using the University of Michigan’s UMHealthResearch.org webportal, which is an IRBMED-approved electronic recruiting platform. Additional summary statistics and raw data are available in the published dataset.⁹ 

Eyes were imaged using a custom imaging device that was described in detail in Section 3.2 of Herzog et al.¹⁰ The system consists of two cameras coupled to a near-UV LED, separated by dichroic beamsplitters. The cameras and LED share a common optical axis, making the measurement a line-of-sight technique. One fluorescence image pair is taken for each eye with a 50-ms duration, along with a series of 10 background and 10 reference image pairs. The corrected image intensity for subject i is calculated as¹¹ 

\begin{eqnarray} I_{c,i} = \frac{I_i - \bar{I}_{b,i}}{\bar{I}_{r,i} - \bar{I}_{rb,i}} \bar{I}_s \end{eqnarray}

(1)

where I is intensity, the overbar (\(\bar{\cdot }\)) denotes an average, and the subscripts c, b, r, and s represent the corrected, background, reference, and standard values, respectively. A fluorescent glass window (Edmund Optics, #21-171) was imaged as the reference. The standard value was taken to be the average intensity in the center 50-pixel square of the reference image from participant #1. Each intensity measurement is thus corrected for background and variation in LED irradiance and sensor response. The units of the corrected intensity are counts or analog-to-digital units, and are used in further radiometric analyses. 

The corrected images were registered using a custom gradient phase-correlation approach based on,¹² and then software-binned 2 × 2 to a pixel size of approximately 30 µm to reduce computation time and the influence of subpixel registration artifacts. The pupil boundary was identified in the green band images based on statistical thresholding and smoothing via morphological operations, followed by elliptical fitting as described by Herzog et al.⁸ The fluorescence intensity ratio R is then calculated as the ratio of the green- to blue-band fluorescence intensity. The fluorescence intensity is calculated as the sum of the green- and blue-band intensities and is the total visible intensity neglecting any ultraviolet emission. The effective fluorescence quantum yield (FQY) is calculated as the ratio of the total intensity to the incident UV flux per pixel. All image processing and analysis were conducted in Matlab r2020b. In some of the subsequent analyses, it is more convenient to use the multiplicative inverse of the ratio R⁻¹, which gives the ratio of blue-to-green fluorescence. However, both quantities intrinsically contain the same chemical information. 

Radiometric analyses are conducted on the thresholded pupil region to avoid regions obstructed by the eyelids or eyelashes, and reduce bias stemming from the interpretation of inhomogenity. Additional analyses were performed that considered the total pupil area and the center 1-mm diameter disk, but were not significantly different from the results presented here except for small but systematic shifts caused by inhomogeneity in the lenses. Inhomogeneity in the lenses will also be discussed in detail in Results. Pupils were not dilated before imaging, and mean pupil radii varied from approximately 2.2 to approximately 3.7 mm. 

Results

Fluorescence intensities, FQY, and ratios are plotted as a function of age in Figure 1. The intensity and ratio data both show an initial linear growth from 18 to approximately 40 years of age, and then tapers off to a constant value. For both measurements, the left and right eyes of each individual are strongly correlated (ρ > 0.95), and the vast majority of the observed variation occurs between individuals. There is an approximately 3% (ratio) and an approximately 8% (intensity) variation between left and right eyes, compared with approximately 100% across the entire dataset. 

Figure 1.

 

Fluorescence intensity and ratio increase with age in healthy eyes. Median corrected intensity and fluorescence intensity ratios are plotted for each eye (left and middle, respectively) as a function of age, with best fit curves superimposed. Fluorescence intensity and ratio (right) have a nearly linear relationship. Errorbars represent five standard-deviations of the mean. Circled points indicate patients with diabetes or prediabetes.

Some previous studies^13,14 have observed a linear trend in fluorescence intensity with age, although these studies are not directly comparable and use different methodologies. We instead hypothesize that the tapering observed here is a characteristic of healthy lenses: we do not expect to see large increases in older patients, because these patients have healthy lenses. Several additional factors could also influence this relationship. We consider only visible fluorescence in this measurement. As will be discussed in Spectral Analysis, increased intensity is associated with a blue shift in the fluorescence emission. At the highest intensities, it is likely that a portion of the emission is ultraviolet and thus not detected here. Similarly, cornea transmission, particularly at UV wavelengths, tends to decrease with age.¹⁵ Neither effect should influence the fluorescence intensity ratio as the cornea transmission is typically flat at visible wavelengths.¹⁶ In support of these arguments, previous studies have also observed a similar trend in the ratio of fluorescence from different fluorophores with age.¹³ 

The intensity data were fit to the function  

\begin{equation} y = y_0 \frac{x^c}{\sqrt{x^{2c} + x_0^{2c}}}, \end{equation}

(2)

which was selected to describe an initial power law growth with an exponent of c, which eventually asymptotes to a limiting intensity y₀. The parameters x₀ and y₀ can be thought of as a characteristic age and intensity, respectively, which separate the power law growth and constant intensity regimes. The parameter c is a characteristic exponent for the growth. The ratio inverse was fit to a linear function, \(R^{-1} = \beta S + R_0^{-1}\). The fit results are summarized in Table 1. 

Table 1.

 

Least-Squares Optimized Fit Coefficients for Fluorescence Intensity and Ratio

 

Detailed subgroup analyses are not possible owing to the limited sample size. However, three patients were previously diagnosed with diabetes or prediabetes (labeled in Fig. 1), a disease that is expected to accelerate lens damage through glycation. It is interesting to note that the two largest fluorescence intensities measured here are from patients with diabetes or prediabetes. The impact of diabetes and other measures of blood glucose concentration will be investigated in detail in future work. We also note that for young eyes (age < 34 years), both intensity and ratio are strongly correlated with age (ρ > 0.8), whereas for older eyes they are not correlated with age (ρ ≈ 0.2 and ρ < 0.1, respectively). 

Spectral Analysis

The ratio measurement is spectrally-resolved by design. The green band collects fluorescence emission at wavelengths larger than ∼460 nm, while the blue band collects wavelengths from ∼400–460 nm. The fluorescence intensity ratio can provide a measurement of spectral shift, which is estimated as follows. An illustration of the calculation is provided in Figure 2. 

Figure 2.

 

Illustration of fluorescence spectrum, simplified collection bands, and parameters for spectral shift calculation. The parameters d, w, and f represent the location of the fluorescence median relative to the cutoff wavelength, a width parameter of the spectrum, and the shaded area fraction of the spectrum, respectively.

A symmetric fluorescence spectrum with a mean wavelength equal to the cutoff wavelength (λ_c ≈ 460 nm) would be split evenly between the image bands and have a ratio of unity. As the spectrum shifts toward shorter wavelengths by a wavelength d, a fraction f of the emission spectrum is removed from the green band and placed into the blue band. Thus, the ratio becomes  

\begin{equation} R = \frac{1-2f}{1+2f}. \end{equation}

(3)

If the spectrum is uniform or broad with characteristic width w and small shifts in mean wavelength d, then f ≈ d/w. The spectral shift can be estimated from the ratio as  

\begin{equation} d = f w = \frac{w}{2}\frac{1-R}{1+R}. \end{equation}

(4)

 

Strictly speaking, f = d/w and the remaining analysis is valid when the cumulative distribution function of the fluorescence emission has a slope of w⁻¹. Based on a previous study of model eyes,⁸ this condition is satisfied to within approximately 5% from 410 to 490 nm with w = 100 nm. The calculated values |f| ≲ 0.4 (and |d| ≲ 40 nm) are believed to be accurate and well-represented by this analysis. 

Applying this analysis to the range of ratio measurements presented in Figure 1, R⁻¹ = 0.4 to R⁻¹ = 0.6, we observe a shift in the mean fluorescence wavelength of approximately 10 nm across the population, as illustrated in Figure 3. 

Figure 3.

 

Calculated mean fluorescence wavelength shift as a function of age.

Spatial Inhomogeneity

All images exhibit some amount of spatial inhomogeneity in fluorescence intensity and ratio inside the pupil. Some of this variation, particularly isolated features, can be attributed to cornea aberrations and have been analyzed in detail previously.¹⁹ Cornea aberrations are typically localized refractive features and have a negligible effect on the integrated fluorescence intensity. However, some spatial inhomogeneity is expected owing to nonuniformity in the lens structure itself. 

Example fluorescence intensity and ratio images for a 30-year-old participant are shown in Figure 4. From the figure, there is a significant increase in intensity near the center of the lens on the order of 20% of the mean, coupled with reduced fluorescence intensity ratio of approximately 2% for this participant. On average over the study population, the fluorescence intensity ratio within 1 mm of the pupil center is approximately 2% smaller than the average over the entire pupil. Similarly, the fluorescence intensity in the center 1 mm of the pupil is approximately 12% larger than the average over the entire pupil. The ratio minima tend to be located very near the center of the pupil, and the fluorescence intensity peak varies between eyes and is not always located at the center. 

Figure 4.

 

Sample spatial variation in fluorescence intensity and ratio. Maps of relative fluorescence intensity and fluorescence intensity ratio are shown (left and middle, respectively). The average ratio is also plotted as a function of radial coordinate averaged over concentric rings with width of approximately 170 µm (right). Contours on the ratio maps and vertical lines on the plot correspond to radii of 1.25 and 2.65 mm for reference. The shaded region on the radial profile indicates the 95% confidence interval of the mean. The images were first smoothed with a Gaussian filter (σ = 2 pixels, or approximately 60 µm) to make the radial pattern more apparent.

In most cases, the shift of the peak fluorescence intensity away from the center of the eye can be largely attributed to differences in eye orientation and lens tilt, which influence the irradiance of the UV light in the lens plane. However, the ratio measurement is formulated to be independent of the excitation irradiance. As such, the spatial distribution of the ratio measurement is primarily a result of inhomogeneity in the lens. The relationship between chemistry and ratio will be further discussed in the Discussion. 

Discussion

Fluorescence intensity and ratio are determined by multiple factors, including concentrations of fluorescent and nonfluorescent species. Multiple fluorophores are known to exist in the lens including kynurenine and derivatives,^7,20 pyridine nucleotides,²¹ β-carbolines and derivatives,^7,22 advanced glycation end products (AGEs),²³ and flavins.²⁴ The approximate fluorescence characteristics of a selection of these fluorophores are detailed in Table 2. It should be noted that the properties listed in Table 2 are approximate. Multiple measurements of some parameters can be found in the literature that differ by an order of magnitude, and should be examined in more detail in the future. 

Table 2.

 

Typical Near-UV Fluorescence Chemicals Found in Human Lens Tissue

 

Our results suggest that young, healthy participants’ eyes contain fluorophores that emit at relatively long (λ₀ ≳ 450 nm) wavelengths, which we believe to be dominated by pyridine nucleotides. This finding is in contrast with some other sources which suggest 3-hydroxykynurenine glucoside (3OHKG), a primary UV filter component in the lens, is dominant at near-UV excitation. Although 3OHKG is perhaps the most abundant fluorescent chemical, the FQY is more than an order of magnitude lower than that of NADH, a dominant pyridine nucleotide. Further, 3OHKG concentration is known to decrease significantly with age,^37,36 whereas we observe a strong increase in intensity and blue-shift with age. Finally, both the measured FQY (∼0.1%–0.4%) and measurements of fluorophore concentrations are consistent with a ∼10:1 mixture of 3OHKG to NADH in healthy young adults. Because the FQY of 3OHKG is so low, the hypothetical 10:1 mixture of 3OHKG to NADH would be dominated by NADH fluorescence. 

It has been suggested that the increase in fluorescence intensity with age is caused by formation of β-carbolines and AGEs,^7,38 which result from oxidation and glycation. Our results support this conclusion. Many β-carbolines (e.g., harmine) have large FQYs and fluorescence peaks that are between 400 and 450 nm in both acidic and basic aqueous solutions.^25,39 Thus, even small concentrations on the order of a few parts per million or less may be sufficient to eclipse any NADH and 3OHKG fluorescence, and cause the observed blue shift and increased intensity. Although AGE compositions and fluorescence properties are not well-known, one fluorescent AGE species, argpyrimidine, has been observed to contribute to fluorescence in aging and diabetic eyes. Argpyrimidine has a shorter wavelength peak near 400 nm and can occur in concentrations on the order of parts per million and could have a similar effect, although the FQY of arpyrimidine has not been measured to the authors’ knowledge. 

Cataract formation is associated with a depletion in 3OHKG, 3-hydroxykynurenine (3OHKN; the nonglucoside form of 3OHKG), and NADH.^31,33 There is also much evidence to suggest that oxidation products of 3OHKN are responsible for coloration of the lens and are related to nuclear cataract formation.⁴⁰ As such, we hypothesize lenticular fluorescence to be a quantitative and accurate indicator 3OHKN oxidation, and hence cataract formation, before pigmentation. 

After significant pigmentation occurs, the concentration of fluorescent flavins such as FAD and FMN increase significantly.²⁴ Although even the cataractous flavin concentrations are typically low, flavins are known to absorb strongly at blue wavelengths making reabsorption of blue fluorescence likely. We thus hypothesize that significant pigmentation would be associated with a red shift in fluorescence. Other sources of ocular pigmentation could have a similar effect. However, the current study is restricted to noncataractous and nonpigmented lenses. Future work will include further focus on common disease states, including diabetic, prediabetic, and cataractous lenses. 

Spatial Inhomogeneity

The spatial inhomogeneity in the lens can be summarized as showing a approximately 3% increase in ratio and approximately 10% to 20% decrease in intensity from the pupil center to the periphery. As described in Discussion, a decrease in the ratio is consistent with a blue shift in fluorescence. This finding indicates that the relative concentrations of NADH, 3OHKG, β-carbolines, and AGEs vary spatially within the lens. Since the measurement is path integrated along the optical axis through an unknown depth of the lens (on the order of 100 of µm), the observed changes could result from multiple processes. However, we believe our results, including observed spatial distributions, are largely indicative of increased oxidation and glycation in the lens nuclear region. 

It is believed⁴¹ that long-lived proteins, like those in the lens nucleus, accumulate more AGEs compared with other tissues, suggesting that AGE concentrations are higher in the nucleus compared to the cortex. Similarly, there is clear evidence of localized nuclear oxidation in aging and cataracts,⁴² so we expect a larger concentration of protein oxidation products (OPs) such as β-carbolines in the nucleus relative to the cortex. In contrast, NADH is known to be “nucleus-depleted”, having a low concentration in the nucleus compared to the cortex.⁴³ Finally, we note that 3OHKG is homogeneously distributed,⁴³ and thus the penetration depth δ of a ray into the lens is constant. Figure 5 shows an illustration of these properties. 

Figure 5.

 

Illustration of lens geometry, ray paths through the lens, and variation in fluorophore concentrations along central cross-sections. Arrows indicate typical ray paths, including initial focusing by the cornea and have constant penetration depth. Shading indicates the relative concentration of fluorescent chemicals including AGEs and OPs. Gridlines are parabolic contours, where horizontal curves indicate constant distance from the anterior surface.

Rays entering at the lens apex are normally incident and point toward the nuclear center, exciting fluorescence that is characteristic of the tissue layer at a depth δ behind the anterior surface. Rays entering near the periphery penetrate with a non-normal incidence angle that is generally not coincident with the chemical concentration gradients. The peripheral rays excite fluorescence that is more characteristic of the cortex and contains a larger share of NADH and smaller share of OPs and AGEs. The net effect is largely what is observed here: the central portion of the lens image has higher fluorescence intensity and quantum yield (neglecting imaging effects owing to lens tilt and eye orientation), and a consistently lower ratio value, compared with the peripheral portion of the lens image. 

Conclusions

Despite cataracts being a leading cause of blindness globally, the chemical processes associated with aging and cataract formation in the lens is poorly understood. New noninvasive in vivo diagnostic methods that can probe chemical concentrations relevant to aging, oxidation, and cataract formation are needed to investigate these processes, and to aid in making clinical decisions. To that end, anterior segment ultraviolet fluorescence imaging was applied to investigate lens fluorescence and chemistry in a set of healthy volunteers. The imaging method is rapid, low-cost, easy to interpret, and can be automated, making it attractive for both research and clinical use. 

In this work, we demonstrated that the technique can be used to discern chemical information about the lens. Our results show a general increase in near UV excited fluorescence intensity with age, consistent with previous reports, and a fluorescence blue shift with increasing age. Typical FQY values for healthy adults, neglecting cornea transmission, are on the order of 0.2%. This is the first report of FQY of whole, healthy lenses measured in vivo. Our results and analyses suggest that the majority of this fluorescence in healthy young adults can be attributed to pyridine nucleotides such as NADH. In older participants, the increased fluorescence intensity and blue shift were attributed to advanced glycation end products (e.g., argpyrimide) and tryptophan oxidation products (e.g., β-carbolines). Systematic spatial variation within most eyes was observed as well, with fluorescence emission from the apex of the lens being blue shifted compared with the periphery. The observed spatial variation was attributed to inhomogeneity in lens fluorophore concentrations, and is consistent with increased damage or oxidation in the lens nucleus relative to the cortex. 

Collectively, we achieved two goals. First, we demonstrated the application of anterior-segment ultraviolet fluorescence imaging in vivo in healthy adults outside of a clinical environment. By design, the method is low-cost, portable, and easy to use. Our results show that we can use this method to provide a quantitative and accurate measurement of visible lenticular fluorescence, which may have both clinical and research utility. Second, we used our results to investigate the age dependence of fluorescence emission in healthy adults with near-UV excitation, in particular focusing on understanding chemical changes in the eye. We used our results to show that the fluorescence emission shifts toward blue wavelengths with increasing age. The blue-shift may be related to formation of oxidation or glycation products, although it was not possible to uniquely identify fluorescent compounds here. As fluorescence is directly related to molecular structure and concentration, the fluorescence intensity ratio described here is a direct measurement of chemical changes in the lens, including oxidation and glycation. Although our results are promising, future work is needed to improve our understanding of the fluorescent species that appear in the lens. We are hopeful that this additional effort will provide the foundation needed to rigorously interpret data and enable researchers and clinicians to better understand and observe oxidation, glycation, and cataract formation in vivo. 

Acknowledgments

Regulatory assistance for human subjects study approval was provided by the Michigan Institute for Clinical & Health Research (MICHR), which is supported by the National Center for Advancing Translational Sciences of the National Institutes of Health under award number UM1TR004404. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. 

Disclosure: J.M. Herzog, patent pending to the Regents of the University of Michigan for “systems and methods for fluorescence-based imaging” (P); A. Verkade, None; V. Sick, patent pending to the Regents of the University of Michigan for “systems and methods for fluorescence-based imaging” (P)