A Narrowband 635 nm Autofluorescence Peak in Albino Mouse Eyes Found With Multi-Modal Imaging Reveals the Presence of Protoporphyrin IX in the Choroid

Abstract

Purpose: To investigate differences in fundus autofluorescence (AF) spectra of pigmented (C57Bl/6) and albino (Balb/c) mouse retinas.

Methods: AF spectra were measured with a scanning laser ophthalmoscope (SLO) with a high-resolution spectrometer. The action spectrum of a 635 nm AF “spike” in albino mice was measured to estimate the underlying absorption spectrum. Optical coherence tomography (OCT) and OCT angiography were used to determine the most likely depth location of the source of the AF feature.

Results: Two narrowband emission peaks centered at 635 nm and ∼705 nm were observed in the AF spectra of albino (but not pigmented) mouse eyes. The dual-peak emission spectrum of the albino eye was extracted by subtracting a broadband emission that was similar in pigmented and albino mice: the two peaks correspond to emission spectra peaks of protoporphyrin IX, an obligate precursor in the biosynthesis of heme and cytochrome c. The action spectrum of the prominent 635 nm emission component corresponded with the PPIX absorption spectrum. The spatial distribution of the 635 nm emission did not correspond to that of the retinal vasculature but had a pattern more consistent with a choroidal origin.

Conclusions: Our results reveal that substantial PPIX is present in the posterior tissues of albino (Balb/c) mouse eyes, and measurable with high-resolution spectral measurements, and suggest that the absence of the narrow band PPIX emission peaks in pigmented eyes arises at least in part from screening by RPE and choroidal melanin.

Mice play a pivotal role in biomedical research, offering valuable insight into various aspects of biology, disease mechanisms, and therapeutic interventions.¹ The eye is a natural window through which neurons, blood vessels, flow, and other retinal structures and activities can be visualized.² Given the substantial number of mice employed in fundamental research, noninvasive in vivo imaging and measurement of the mouse eye have the potential to yield substantial resource savings in terms of both the number of mice utilized and researchers' time.³ 

Autofluorescence (AF) measurement in the mouse retina is an important tool in vision research due to its ability to provide insights into retinal health and disease processes.^4,5 This noninvasive technique allows researchers to assess the fluorescence emitted by various naturally occurring retinal molecules, such as lipofuscin, which accumulates with age and in certain retinal diseases.⁵ By analyzing autofluorescence patterns, it has been possible to detect early signs of retinal degeneration, monitor disease progression, and evaluate the efficacy of therapeutic interventions.^6,7 In mouse models, which are widely used to study human retinal diseases, AF measurement serves as a bridge between basic research and clinical applications.^8–10 It offers a valuable means to understand the pathophysiology of conditions like age-related macular degeneration and retinitis pigmentosa, thereby advancing the development of new treatments and preventive strategies.^10,11 There are material differences between mice and human fundus AF, particularly regarding the source(s) of AF. Unlike humans, mice lack a fovea, which significantly impacts the topography and constituents of the degradative load on the RPE.^12,13 Nevertheless, insights garnered from mouse retinal imaging studies have potential utility for human research. For example, discoveries made in mouse animal models can inform the development of innovative diagnostic and therapeutic strategies for retinal diseases.¹⁴ 

Optical coherence tomography (OCT)¹⁵ and scanning laser ophthalmoscopy (SLO)¹⁶ have transformed vision science. SLO, unlike OCT, affords the distinct capability of measuring fundus AF,¹⁷ an invaluable tool in ophthalmology and vision research.⁶ This noninvasive methodology measures fluorescence emitted by endogenous fluorophores within the retina, furnishing insights into retinal health and a myriad of ocular conditions.¹⁸ This technique serves as a cornerstone in preclinical research, facilitating the evaluation of experimental treatments and interventions in animal models.¹⁹ Recognizing that OCT and SLO represent complementary imaging modalities, it has become increasingly common to integrate them into multi-modal imaging systems.^20–24 Such integrated systems offer enhanced capabilities for investigating ocular biology, disease mechanisms, and therapeutic interventions,²⁵ facilitating the discovery of novel biomarkers. 

Traditional SLO fluorescence measurements usually focus on imaging with high spatial resolution, but often lack the high spectral resolution optimum for the identification of fluorochromes. To address this latter limitation, we developed an SLO/OCT multimodal imaging platform equipped with a high spectral resolution spectrometer capable of capturing reflectance and fluorescence spectra during SLO excitation.²⁶ Here we present an investigation into disparities in AF between pigmented and albino mice using this system. Albino mice lack melanin, which can influence light absorption and scattering within the eye, potentially affecting retinal responses and susceptibility to light-induced damage.^27,28 

In this investigation, we found previously unreported narrowband spectral peaks at 635 nm and ∼705 nm in fundus AF spectra of albino mice. Based on an analysis of these spectra that removes a broadband emission component, and on action spectrum measurements, we attribute these narrowband emissions to protoporphyrin IX (PPIX).²⁹ The broadband AF was found to be largely attributable to A2E,³⁰ the best-characterized fluorochrome belonging to the bisretinoid chemical class among several reported to be present in the RPE. 

Material and Methods

Experimental Setup

As described previously,²⁶ our custom SLO/OCT multimodal system facilitates a wide range of investigations of the murine retina. For the scope of the present investigation, the SLO part was the principal portion of the apparatus employed, as shown in Figure 1. A super-continuum laser (“SC laser”; NKT Photonics, Birkerød, Denmark) served as the light source. To provide tunable, narrowband excitation from the SC laser, a tunable filter (F1; the SuperK VARIA; NKT Photonics) was incorporated, enabling the precise selection of different 4 nm wave bands from 400 nm to 800 nm for excitation of fundus autofluorescence. 

Figure 1.

 

Schematic of the SLO/OCT system equipped with a spectrometer (see text for details).

The light emitted by the SC laser passed through two beam splitters (denoted as BS1 and BS2, with ratios between reflectance and transmission of 70:30 and 30:70, respectively), configured to partition the incident light, and directed toward an XY galvo-scanner to generate the scanning beam. A cascaded optical configuration, comprising a scan lens and a tube lens, relays the scanning pattern onto the murine retina. The interaction between incident light and the retinal tissue results in both back-reflectance and fluorescence emissions. These optical signals are recorded using three Photomultiplier Tubes, PMT1 (H7422-20), PMT2 (H7422-40) and PMT3 (H7422-50 (Hamamatsu Photonics K.K, Hamamatsu, Japan). PMT1 was used to capture the 488 nm reflective component, while PMT2 with filter F2 (FF01-525/45; Semrock, West Henrietta, NY, USA) and PMT3 with filter F3 (FF01-635/18) recorded fluorescence emission in two non-overlapping spectral bands. A flip mirror (FM) was used to direct back-reflected and fluorescent light toward the spectrometer (QE65000; Ocean Optics, Orlando, FL, USA), with a long-pass filter (F4, BLP01-514R or BLP02-561R; Semrock) ensuring that only autofluorescence (AF) was recorded by the spectrometer from the selected region of interest (ROI). 

The integration of the OCT into the multi-modal system was achieved by means of a dichroic mirror (DM1), as illustrated in Figure 1. The OCT system used an 860 nm superluminescent diode light source (T-860-HP; Superlum, Cork, Ireland) with an effective full width at half maximum bandwidth of approximately 82 nm and an empirically determined axial resolution of 3.6 µm within biological tissue.³¹ 

SLO Reflectance and Fluorescence Imaging and Autofluorescence Emission Spectrum Measurements of Pigmented and Albino Mice

The SLO system captured both reflectance and AF images concurrently. Following a preliminary alignment, wherein the optic nerve is approximately centered in the image field, a dataset comprising 100 pairs of unprocessed reflectance and autofluorescence images is gathered. After this acquisition, alignment, averaging, and contrast enhancement protocols are executed to yield composite averaged images. Illustrative examples for pigmented and albino mice are presented in Figures 2a and 2b and 2d and 2e, respectively. In the pursuit of acquiring AF emission spectra, a designated ROI, such as that marked by the green rectangle in Figures 2a and 2d, is singled out, and the scanning restricted to the ROI, increasing the linear spatial sampling density 20-fold. By toggling the flip mirror, light emerging from the eye is deflected into the spectrometer's path. To ensure sufficient signal strength of the fluorescence spectrum, an integration time of one second was used, accompanied by the concurrent storage of 50 spectra. These spectra were subsequently averaged, substantially reducing noise. Averaged spectra for pigmented and albino mice are depicted in Figure 2: these reveal a distinctive “spike” centered at 635 nm in AF of the albino eye (Fig. 2f) that is not seen in the spectrum of the pigmented mouse (Fig. 2c). 

Figure 2.

 

Example SLO reflectance images (a, d) and autofluorescence images (b, e) from pigmented (C57Bl/6, top row) and albino (Balb/c, bottom row) mice, and the corresponding autofluorescence emission spectrum measurements (c, f). (The spectra are spatiotemporal averages from the ROIs indicated by the green rectangles in a and d). Scale bar: 300 µm. Average spectral resolution: 6.52 nm.

Action Spectrum Measurement and Data Analysis Protocol

To conduct action spectrum measurements, precise control of wavelength and power is imperative. As detailed in the Table, a tunable filter was employed to select a sequence of excitation wavelengths, each with a narrow bandwidth (4 nm). For each of these selected wavelengths, the power was precisely set at three distinct levels, enabling measurement of the dependence of the emission amplitude on excitation power for each wavelength. We found that monitoring the laser input power was important, as it could vary by ±20% within half an hour. To address this, a power meter was placed after BS2 (Fig. 1) to measure the power in real time. To ensure consistency and account for potential fluctuations, we measured the ratio between the power measured after BS2 and the power at the pupil prior to the experiment. This ratio served as a reference to correct for any laser output fluctuations. The final power at the mouse pupil was calculated by dividing the power after BS2 by this predetermined ratio. 

Table.

 

Laser Power Used for the Action Spectrum Measurements

 

The maximum permissible light exposure for the mouse eye was calculated assuming that the mouse retina possessed the same susceptibility to light damage as the human retina.³² Taking the numerical aperture (NA) of the mouse in this imaging system to be roughly the same as that of the human eye with a fully dilated pupil, there are no differences in the spot size and retinal illuminance for the purpose of the simplified calculation. During the spectrum measurements, the smallest field of view used was 2.5° × 2.5°. These measurements were conducted during SLO at a resolution of 256 × 256 pixels and a frame rate of approximately 2 Hz. The spectrum measurements were accumulated over 20 SLO scans, totaling 10 seconds. Including precheck time, the total duration for each measurement is less than 20 seconds. The laser light is turned off for an average of 20 seconds between different measurements. Therefore the commonly used "continuous wave limits" for the SLO are applied, as they offer the greatest degree of protection for small fields at short wavelengths.³³ In our case, the wavelengths range from 450 nm to 560 nm. The maximum permissible beam power (in watts) for the thermal limit is calculated by:³³ 

\begin{eqnarray}M{{P}_{th}} = 6.93{\rm{\ }} \times {{10}^{ - 4}}{{C}_T}\left( {\frac{{4{{\alpha }_F}}}{{\pi {{\alpha }_{min}}}}P_{20}^{ - 1}{{t}^{ - 0.25}}} \right)\end{eqnarray}

(1)

where C_T = 1 for wavelength within 400–700 nm, α_F = 43.6 rad, α_F = 1.5 rad, the pupil factor for t > 0.7 s and wavelength within 400–600 nm is P₂₀ = 5.44, t = 20s; thus the thermal limit MP beam power is 2.23mW. 

The photochemical limit MP beam power (in watts) is calculated by:³³ 

\begin{eqnarray}M{{P}_{ph}} = 5.56{\rm{\ }} \times {{10}^{ - 6}}{{C}_B}\frac{4}{\pi }\alpha _F^2{{t}^{ - 1}}\end{eqnarray}

(2)

where C_B = 10^{0.02(λ − 450)} for wavelength within 450–600 nm, α_F = 43.6 rad, t = 20s. Thus the photochemical limit is wavelength-dependent; the calculated results are presented in the Table. 

To capture AF emission spectra, the scanner was positioned for an ROI in the retina for which the AF showed a prominent peak centered at 635 nm (e.g., Fig. 2f). Each ROI was approximately 100 µm × 100 µm at the retina. Before beginning data acquisition, an initial “background” spectrum was recorded without the mouse in the apparatus. This spectrum was obtained using identical imaging conditions but in the absence of any retinal tissue, in effect, a readout of dark noise with a DC offset from the line sensor of the spectrometer. This background spectrum subsequently served as a reference. By subtracting this background spectrum from all subsequent measurements, we obviated any fluorescence that could be attributed to instrumental factors. 

In extracting the action spectrum of the 635 nm AF peak, the following approach was taken. The averaged AF spectra were examined within two distinct wavelength bands: one band ranged from 588 nm to 593 nm on the shoulder to the left of the 635 nm peak, while a second band, containing the peak, spanned from 632 nm to 636 nm (blue and red lines in Fig. 2f, respectively). From each excitation wavelength (Table), the averaged emission fluorescence intensities of these two bands were extracted for every measurement. For each wavelength and band, three emission intensities were measured, and the slope between emission strength and excitation power was extracted. In subsequent analysis, the slope of the emission from the 588 to 593 nm band was subtracted from the slope of the 632 to 636 nm band. On the assumption that the broadband emission is present “beneath” the 635 nm spike, this adjusted slope was designated as the “action” for the molecular species underlying the spike. Finally, the action spectrum was compared with predictions based on established reference absorbance spectra, specifically for A2E³⁴ and PPIX.²⁹ Other bisretinoid fluorophores in addition to A2E have been identified, each with distinct excitation spectra.³⁵ These latter spectra vary considerably and include excitation maxima reported around 430 nm, 490 nm, and 510 nm. Although acknowledging that these and other fluorochromes may have affected the measurements, we found that the A2E spectrum³⁶ nonetheless provided a reasonable account of the broadband AF excited with 488 nm light. 

Animal Preparations and Imaging

All mouse husbandry and handling protocols were approved by the University of California's Institutional Animal Care and Use Committee, which strictly adheres to all guidelines of the NIH and satisfies those of the Association for Research in Vision and Ophthalmology for animal use. Adult pigmented (C57BL/6(J), n = 3) and albino (Balb/c, n = 10) mice, aged two to six months, were obtained from the Jackson Laboratory, and maintained on a 12:12, −100-lux light cycle. During measurements, they were anesthetized with inhalational isoflurane (2% to 2.5% in O₂), and their pupils were dilated with medical-grade tropicamide and phenylephrine. A contact lens³⁷ and gel (GelTeal Tears; Alcon, Geneva, Switzerland) were used to maintain good tear film quality and corneal transparency during imaging, with a heating pad used to maintain the mouse's body temperature. For the experiments measuring narrowband autofluorescence (AF) centered at 635 nm (Figs. 2–4), eight albino and three pigmented mice were recorded (Supplementry Appendix 1.1); For the action spectrum measurement of the 635 nm peak, illustrated in Figures 5 and 6, one albino mouse was used. For examining the spatial relation between fluorescence and optical coherence tomography angiography (OCTA) vascular images, one albino mouse was used. As the anesthesia and imaging protocols were non-invasive, some mice were imaged in more than one experiment; in such cases we allowed at least three days between experiments. 

Figure 3.

 

AF spectra taken at multiple locations from a pigmented mouse (C57Bl/6, top row) and from an albino mouse (Balb/c, bottom row) revealing notably a pronounced AF “spike” at 635 nm in albino mice not detected in the pigmented mouse. (a, d) SLO 488 nm reflectance image with ROIs marked by green, numbered boxes; (b, c, e, f) spectra obtained from each of the locations, as identified by the numbers marked in a and b, respectively. A 514 nm long-pass filter blocked reflected light. (AF spectra of additional mice are presented in Supplementry Appendix 1.1 “AF Emission Spectra of Additional C57Bl/6 (pigmented) and Balb/c (albino) mice.”) Scale bar: 300 µm.

Figure 4.

 

Protoporphyrin IX (PPIX) emission signature in the AF spectra of albino mice. (a) AF spectra, derived from the measurements of Figs. 3(e) and (f) and normalized by the average emission in the 590 to 610 nm range. (b) The normalized difference spectrum between the spectrum with the most pronounced 635 nm spike and the spectrum with the least pronounced 635 nm spike. (c) Protoporphyrin IX (PPIX) chemical structure (Jmol/JSmol Interactive Script Documentation, St. Olaf College). (d) Published PPIX absorbance and (e) PPIX emission spectra (SearchLight Spectra Viewer, Semrock). (f) Comparison of the albino difference spectrum (panel b; blue) with the published PPIX emission spectrum (red), both normalized at 635 nm.

Figure 5.

 

Action spectrum measurements. (a) Plot of the unusual “undulating” absorption spectrum of PPIX in the spectral range 440 nm to 560 nm, plotted together with the 11 selected excitation wavelengths (red circles, Table 1). (b) Raw emission spectrum of the measurements, with two bands ([582, 586] nm (blue thick line) & [632,636] nm (red thick line) indicating spectral regions that were further analyzed. (c) For the [582, 586] nm band, the measurements with three different excitation powers for each wavelength were plotted to extract the slopes. (d) For the [632, 636] nm band, the measurements with three different excitation powers for each wavelength were plotted to extract the slopes. (e, f) Slopes of the lines from the [582, 586] nm and [632,636] nm emission bands respectively plotted as a function of excitation wavelength. The black stars plotted in erepresent slopes extracted from each of the three pairs of the measurements for each condition, whereas the symbols plot the mean value. The goodness of fit of the slopes, indicated by R² values, was very close to unity (Supplementry Appendix 1.3), underscoring reliable linear dependence on excitation power at all wavelengths.

Figure 6.

 

Comparison of the action spectra of the [582, 586] nm band (b) and the [632, 636] nm band with the media-corrected absorption spectrum of A2E (a) and PPIX (b).

Results

AF Spectra of Albino Mouse Eyes Have a Retinal Location-Dependent, Narrowband Emission Peak at −635 nm

We compared the AF spectra of a pigmented mouse and an albino mouse excited with 488 nm light, with reflectance images serving as a fundus reference map (Figs. 3a, 3d). The AF spectra were taken from randomly selected fundus locations, and were separated into two groups for clarity: one from the left portion of the retina (Figs. 3b, 3e) and the other from the right portion (Figs. 3c, 3f). All AF spectra from this and other pigmented mice (Supplementry Appendix 1.1) exhibited a smooth profile with no discernible narrowband peaks. In contrast, the spectra taken from most locations in this eye and those of other albino mice (Supplementry Appendix 1.1) exhibited a distinct narrowband AF peak (“spike”) at 635 nm whose amplitude in this instance was greatest at locations 4 and 10. The facts that the spikes were not present at all locations under otherwise identical stimulation and measurement conditions and, when present, varied in amplitude relative to the broadband AF emission, indicate a nonuniform fundus distribution of the underlying fluorochrome. 

The ROIs in these experiments were randomly selected, in part because it became clear early in the investigation that the 635 nm AF spike varied in amplitude across the albino fundus and indeed in some locations was undetectable (e.g., trace 2 in Fig. 3e). We found that the 635 nm spikes were only present in AF spectra of albino mice, and never in the spectra of pigmented mice, in this experiment and numerous additional experiments (Supplementry Appendix 1.1). 

The Presence of a 705 nm Emission Peak Together With the 635 nm AF Spike Suggests That the Narrowband Emission Arises From PPIX

To examine the AF spectra of the albino eye in more detail, the measurements presented in Figures 3d through 3f were first normalized to their average intensity within the 590 to 610 nm range (Fig. 4a). Next, the spectrum with the least pronounced 635 nm spike (no. 2) was subtracted from the spectrum with the most pronounced spike (no. 10), on the hypothesis that the spectrum lacking the spike represents a ubiquitous broadband emission (Fig. 4a), whereas that with the highest amplitude spike represents the purest spectrum of the narrowband emission. This difference spectrum revealed a second narrowband peak at ∼705 nm (Fig. 4b). A search of the fluorophore database in the SearchLight Spectra Viewer provided by Semrock yielded a matching candidate: PPIX (Supplementry Appendix 1.2), whose chemical structure, absorbance spectrum and emission spectrum are presented in Figures 4c through 4e, respectively, with the latter exhibiting narrowband 635 nm and 705 nm emission peaks. A comparison of the measured albino AF difference spectrum with the published PPIX emission spectrum, revealed them to be virtually identical (Fig. 4f). The subtraction of the broadband spectra from the composite spectra was rationalized by the facts the broadband spectrum was ubiquitous (if not uniform in amplitude) in the AF of albino mice and resembled that seen in pigmented mice, whereas the narrowband peak varied in its presence, location and amplitude, consistent with the broadband and narrowband components arising from distinct underlying fluorochromes. 

Action Spectrum Measurements Confirm That the Albino Narrowband AF Emission Spikes Originate From PPIX

To further test the hypothesis that PPIX is the chromophore whose light-capture underlies the narrowband AF emission spikes, we conducted action spectrum measurements. On the hypothesis that a single PPIX electronic transition underlies the narrowband 635 nm emission “spike,” the excitation intensity (I_ex(λ)) of wavelength λ and the emission intensity (I_em635(λ), the measured “action”) are expected to be linearly proportional to one another as follows:  

\begin{eqnarray}{{I}_{em635}}\left( \lambda \right) = k\ \tau \left( \lambda \right)\ Ab{{s}_{PPIX}}\left( \lambda \right)\ {{I}_{ex}}\left( \lambda \right)\end{eqnarray}

(3)

where k is a system constant, τ(λ) is the transmission of ocular media between the cornea and the PPIX, and Abs_PPIX(λ) is the absorption spectrum of PPIX (Fig. 4d). Thus I_em635(λ), the AF intensity measured when the excitation has wavelength λ, is predicted to be a linear function of I_ex(λ), with the wavelength-dependent slope k τ(λ) Abs_PPIX(λ), the media-corrected absorption spectrum. An important feature of Equation 3 is that the slope dI_em635(λ)/dI_ex(λ) of the relation between excitation and emission is directly proportional to Abs_PPIX(λ), and thus independent of the range of I_ex(λ). The “emission versus power” relation is expected to follow a linear relationship (Eq. 3), providing a robust measurement independent of the absolute level of the absorbing molecular species present in the ROI from which the measurements are made (Supplementry Appendix 1.3). 

Given the limitation of the apparatus, we used excitation in the spectral range 440 nm ≤ λ ≤ 560 nm, with the excitation wavelengths selected to include the peaks and troughs of the strikingly undulating absorption spectrum of PPIX (Fig. 5a, red circles). For each of the selected wavelengths, we measured AF spectra with three different excitation levels (Table). The emission spectra were limited to the spectral region beyond 561 nm, and obtained from a single retinal locus displaying a robust 635 nm spike. The resulting 33 measurements are displayed in Figure 5b. We examined two specific bands of the emission spectrum: one between 588 nm and 593 nm and the other between 632 nm to 636 nm. For these bands, we extracted the spatially averaged fluorescence intensities for the three power levels across the 11 wavelengths and plotted them in Figures 5c and 5d, respectively. For each wavelength, the slope was extracted from the emission measurements obtained with the three excitation levels. The slopes obtained for the 11 excitation wavelengths in these two bands are presented in Figures 5e and 5f, respectively. The action spectra underlying the broadband AF and the 635 nm AF spike are highly distinct: the former peaks at about 460 nm and falls off at higher wavelengths (Fig. 5e), whereas the latter falls off at wavelengths shorter than 500 nm and has twin peaks at about 500 and 540 nm (Fig. 5f). Notably, this twin-peak pattern matches that of the PPIX absorption spectrum illustrated in Figure 5a, whereas the absorption spectrum underlying the broadband AF (Fig. 5e) resembles that of A2E,³⁶ known to accumulate in the RPE of both pigmented and albino mice³⁸ (Fig. 6a). 

On the assumption that the broadband AF is also present in the spectral region containing the 635 nm spike, to obtain the proper action spectrum for the latter, we subtracted the action spectrum of the broadband AF (Fig. 5e) from the “raw” 635 nm spike's AF spectrum (Fig. 5f). Given that the A2E is present in the RPE layer, and that the 635 nm spike AF likely arises from a deeper, choroidal source, the action spectrum of the narrowband containing 635 nm was corrected both for preretinal transmission and A2E absorption (Fig. 6a): the corrected spectrum closely resembles the absorption spectrum of PPIX, and with both spectra having narrowband peaks at 506 nm and 542 nm (Fig. 6b). In sum, the media transmission-corrected action spectrum of the narrowband 635 nm AF closely aligns with the unusual undulating absorption spectrum of PPIX (Fig. 6b), confirming a prediction of the hypothesis that PPIX is the underlying fluorochrome. 

Spatial Distribution of PPIX Across the Fundus Obtained by Spectral Unmixing

The largest 635 nm AF recorded originated at a notable distance from the optic nerve head. Accordingly, to investigate the spatial distribution of PPIX within the fundus in more detail experiments were conducted with the optic nerve head positioned in the bottom right corner of the image (Fig. 7). The fluorescence generated by the 488 nm excitation was split into two channels by a dichroic mirror, filtered by bandpass filters centered on 525 nm and 635 nm and recorded simultaneously by two separate photomultiplier tubes (Fig. 1, PMT2 and PMT3). Resultant 50-frame-averaged fluorescence images are shown in Figures 7d and 7e. In the same experiment, high spectral resolution data were collected from three randomly selected ROIs (Fig. 7b); from the latter, AF signals attributable to A2E (broadband AF, Fig. 7b) and PPIX (narrowband 635 nm AF, Fig. 7c) were extracted. Sharing the common 488 nm excitation, the AF of each fluorochrome recovered in the two PMT channels is governed by six factors: the local retinal concentration of each component, their 488 nm absorption coefficients and quantum efficiencies of fluorescence, the overlap of the emission spectra with the bandpass filters and the spectral sensitivities of the two PMTs. Because all factors except the concentrations are known from the literature, a spectral unmixing algorithm enabled the extraction of two signals proportional to the local concentrations of A2E and PPIX (Figs. 7f, 7g; Supplementry Appendix 1.4). For validation of the unmixing algorithm, we made scatterplots of the ROI data: the “unmixed” A2E and PPIX signals are proportional to the high-resolution spectral signals from the corresponding ROIs (Fig. 7i), but the raw 2-color fluorescence data were not (Fig. 7h). The average A2E signal over the entire field of view was 4.5-fold greater than the average PPIX signal in this experiment, and 7.5-fold greater in a second experiment. The spatial distributions of A2E and PPIX AF (Figs. 7f, 7g) are similar, with a pixel-by-pixel correlation coefficient of 0.79. 

Figure 7.

 

Spatial distribution of autofluorescence (AF) in the Balb/c fundus. (a) SLO reflectance image captured with 488 nm light. (b) Spectral distributions of AF from 3 ROIs whose locations are identified in a by squares having the same edge color. Smooth curves (gray) have been fitted to the broadband component of AF. (c) Spectral distributions of the narrowband 635 nm component of AF from the three ROIs obtained by subtraction of the gray curves from the spectra in b. (d) AF image obtained by PMT2 with bandpass filter centered on a 525 nm. (e) AF image obtained by PMT3 with bandpass filter centered on 635 nm, acquired simultaneously with that in (d). (f, g) Data from the paired images were transformed pixel-by-pixel to extract signals predicted to be linearly proportional to the local concentrations of A2E and PPIX, respectively (cf. Supplementry Appendix 1.4). (The fluorescence images were automatically contrast-adjusted by the Matlab “imagesc” command). (h) Scatterplot of the AF from the three ROIs in the fluorescence images (d) and (e) paired with the broadband and narrowband AF signals measured with the high-resolution spectrometer from the same ROIs. (i) Scatterplot of A2E and PPIX signals from the three ROIs with the spectral unmixing algorithm (ordinate) paired with the A2E and PPIX signals measured with the high-resolution spectrometer (abscissa) from the same ROIs. The ordering of the fluorescence intensities attributed to each fluorochrome is preserved by the unmixed signals (i) but not the “raw” two-color fluorescence data (h). (The ordinate values of the unmixed PPIX signals were scaled by 3X for clarity). The diameter of the field of view was 50 deg, about 1.8 mm across.

Depth Distribution of PPIX in the Fundus: Insights From Vascular Imaging

Since PPIX represents the final intermediate in the heme biosynthetic pathway, it is reasonable to examine whether its fluorescence might be associated with fundus vasculature. Because the size and branching of blood vessels vary with depth, we performed OCTA using the phase variance method,³⁹ generating vessel maps for the anterior and posterior retina (Figs. 8a–c), and compared them with the derived A2E and PPIX signals (Figs. 7d–g). The low AF from image areas having large anterior vessels reveals that the sources of both 525 nm and 635 nm AF emission lie in the posterior retina. Diminished AF in image locations having large anterior blood vessels is expected because the concentration of blood hemoglobin is about 15 g L⁻¹, and hemoglobin (HbO₂) absorbance is strong: 24175 L (mol cm)⁻¹ at 488 nm and 30882 at 525 nm. In contrast, the AF distributions of both A2E and PPIX are strongest in the regions of highest choroidal vessels (Figs. 8f, 8g), implying that the choroidal vasculature lies posterior to both the depth location of both fluorochromes. The apparent lack of reduction of PPIX emission by the choroidal vasculature suggests that the depth of the PPIX sources may be near the RPE layer, where A2E is well established to localize.⁵ 

Figure 8.

 

Autofluorescence of A2E and PPIX compared with vasculature imaged with OCT angiography (OCTA). (a) OCT b-scan taken at the location indicated by the green arrow in (b). (b) OCTA enface image of the anterior vasculature. (c) OCTA enface image of the posterior vasculature. (d, e) A2E and PPIX AF signals (repeated from Figs. 7f and 7g). (f, g) Overlay of the posterior vasculature image from c in 50% transparency on the corresponding A2E and PPIX AF distributions in panels d and e. The regions with the strongest A2E and PPIX signals also exhibit the strongest OCTA signals. Scale bar in b: 300 µm.

Discussion and Conclusions

Summary of the Key Findings

In ongoing investigations of the fundus autofluorescence (AF) of mice using 488 nm excitation, we observed a hitherto uncharacterized narrowband emission “spike” with a peak at 635 nm (Figs. 2f, 3f). The 635 nm AF spike was detected in albino (Balb/c) mice but not in pigmented (C57Bl/6) controls. The 635 nm AF was not uniformly distributed across the fundus (Fig. 4a; Figs. 7f, 7f), and spectral decomposition of broadband and narrowband components revealed it to be associated with a second narrowband AF peak at ∼705 nm (Fig. 4b). Searches of AF spectra databases led to the hypothesis that the fluorochrome underlying the dual-spike AF emission spectrum is PPIX (Figs. 4c–f). This hypothesis was further confirmed by measurement of the action spectrum for the 635 nm spike (Figs. 5, 6): corrected for ocular transmission losses, including that from A2E in the RPE, the action spectrum was found to have absorption peaks corresponding to those of PPIX (Fig. 6). The PPIX AF did not co-localize with major anterior blood vessels, but rather has a distribution similar to that of the choroidal vasculature (Figs. 7, 8). Further studies are clearly required to determine whether PPIX AF is associated with choroidal vasculature (e.g., in perivascular macrophages). 

Justification for the Spectral Analysis Without Complementary Biological Assays

We identified the source of the two peaks attributed to PPIX using only spectral analysis, foregoing complementary biological assays like histology, for several reasons. First, much previous research has shown that specific fluorescent compounds can be distinguished through their unique spectral signatures. For example, one study utilized fluorescence spectral analysis to identify and characterize dissolved organic matter leached from various plastics, demonstrating significant correlations between specific fluorescent components and plastic additives.⁴⁰ Second, PPIX is well known for its unique twin peak signature, which is widely used in photodynamic therapy (PDT) to identify and monitor PPIX during treatment.⁴¹ Third, our spectral analysis involved comparing the measured spectrum with the PPIX emission spectrum (Fig. 4f). Additionally, the action spectrum was measured, and its distinct twin-peak excitation further supports the hypothesis that the underlying component is PPIX (Fig. 6b). Finally, the OCTA data provides suggestive evidence that the PPIX emission more likely arises from sources near the posterior vasculature, also consistent with its absence in pigmented mice (next section). Histological data (e.g., PPIX immunohistochemistry) might strengthen these findings, but would be challenging because of the possibility that the redox state or other aspects of PPIX would be affected by fixation artifacts. Future studies will aim to integrate such assays to provide a more comprehensive validation. 

Potential Explanation for the Differences in PPIX AF Between Pigmented and Albino Mice

A plausible explanation for the absence of the PPIX AF in pigmented mice is that melanin and lipofuscin in the RPE and choroid screen PPIX from 488 nm excitation. Our previous investigation of fundus reflection spectra in the context of rhodopsin measurement⁴² concluded that the average double-pass optical density at 488 nm of melanin anterior to the sclera in C57Bl/6 mice is about 1.0 log₁₀ units, and given that most melanin absorption is in the RPE cell bodies and apical processes, excitation of PPIX localized posterior to the RPE would be reduced at least 10-fold. An alternative and not mutually exclusive hypothesis is that PPIX is elevated in Balb/c relative to C57Bl/6 mice. A rough estimate of the concentration of PPIX in albino eyes (Fig. 7, Location #1) can be made by comparing the broadband AF measured in our experiments with similar broadband measurements attributable to A2E.⁴³ A straightforward calculation based on the published extinction spectra of A2E and PPIX (Supplementry Appendix 1.5) yields an estimate of the PPIX concentration at this particular location of 230 µM. Recent studies using quantitative intraoperative probes indicate that over 95% of normal tissue contains PPIX levels below 0.1 mg/mL,^44–46 which corresponds to approximately 0.2 µM. Thus the estimated PPIX concentration of 230 µM at the location of maximal AF is much higher than that normally present in tissue. PPIX in disease or therapeutically altered states can be materially elevated above 0.2 µM. For example, the PPIX concentration in malignant brain tumors was found to be 5.8 µM following orally administered 5-aminolevulinic acid,⁴⁷ and PPIX concentrations of 0.6 mg/mL (1666 µM) have been measured in malignant gliomas.⁴⁴ 

Exploring PPIX Accumulation and Its Implications in Mouse Eye Physiology

PPIX (Fig. 3c) is a precursor in the synthesis of several essential biomolecules, including heme and cytochrome c.^29,48 Reactions involving PPIX are highly conserved and regulated as to their location in tissues and organelles: thus, for example, 85% of heme synthesis occurs in the bone marrow and most of the rest in the liver, with typically negligible PPIX present in the circulating blood.²⁹ Excess PPIX is toxic, as found in several diseases associated with iron deficiency, including X-linked protoporphyria.⁴⁹ Lowered levels of iron reduce the rate of PPIX conversion to heme, leading directly to PPIX accumulation in cells. Moreover, photic excitation of PPIX produces reactive oxygen species, and elevated PPIX can interfere with a number of important reactions—including transport of heme into mitochondria by the ABC transporter ABCG2.²⁹ An intriguing hypothesis about the high PPIX level in the apparently healthy albino Balb/c eye is that the abundant, strongly iron-binding isomerohydrolase RPE65⁵⁰ competes against PPIX for iron: Rpe65 has been measured to be 14-fold more abundant in Balb/c mice (which express Rpe^{450Leu/450Leu}) than in C57Bl/6J mice (which express Rpe^{450Met/450Met}).⁵¹ Although the Rpe65 polymorphism of the two strains offers a potential contributing factor for explaining the overall AF differences between the strains, alternative explanations need to also be considered for the observed local variation in PPIX spectra. These might involve differences in metabolic pathways or the influence of other proteins in the albino eye. Addressing these hypotheses would involve the use of other mouse models, such as the albino C57/BL6-c2J mouse,⁵² which could serve to determine whether the strong PPIX signals are specific to the Balb/c strain or arise mainly from the lack of melanin. 

PPIX Autofluorescence as a Biomarker for Disease Detection and Monitoring

PPIX AF is a potential biomarker for several serious and widespread disease conditions, including cancer and Covid-19. SARS-CoV-2 is known to damage hemoglobin, “pulling” iron from the hemoglobin-associated protein ORF3a for use in the viral reproduction process, resulting in elevated PPIX. PPIX AF has been proposed for serological and dermatological detection of Covid-19 and in the assessment of therapeutic interventions for these and other conditions, including iron deficiencies of various etiology.⁵³ Although detection of PPIX in fundus AF of humans has not yet been reported, this may, in part, be due to inadequate spectral resolution, as well as to screening by melanin. Future studies could explore hypotheses regarding the differences between pigmented and nonpigmented mice by using paradigms similar to those used in past investigations of outer retinal pathology in experimental protoporphyria.^54,55 For human applications, future research could aim to enhance the technique's sensitivity, assess its applicability across diverse populations, and determine its cost-effectiveness in clinical settings. By addressing these challenges, PPIX fundus AF has the potential to become a valuable tool in personalized medicine and public health (e.g., for detecting and monitoring iron deficiency syndromes and Covid-19). Moreover, fundus AF is known to be extinguished in both humans and mice under certain conditions, such as RPE65 mutations.⁵⁶ This presents an opportunity to assess the role of PPIX in our current findings. By examining its significance in scenarios involving retinal degeneration—particularly distinguishing between degenerating RPE and neural retina—we can better understand its potential impact. If PPIX proves to be significant in these contexts, it could lead to broader clinical applications, enhancing our ability to diagnose and monitor retinal diseases. 

Limitations of the Action Spectrum Experiment

The action spectrum experiment (Figs. 5, 6) was designed to test a prediction of the hypothesis that PPIX is the molecular species whose light absorption underlies the highly replicable 635 nm AF “spike” evoked by 488 nm light (Fig. 2f, Figs. 3d– f; Fig. 5b; Fig. 7b; Supplementary Fig. S1c; total of 10 eyes of nine mice). The absorption spectrum of PPIX (Fig. 4d) is highly distinct from its emission spectrum (Fig. 4e), and unusual because of the series of four “undulations” beginning at about 505 nm, and the experiment was designed to use narrowband excitation at two of the peaks and three of the troughs of these undulations (Fig. 5a). The action spectra data revealed the broadband and 635 nm AF spectra to have qualitatively distinct shapes (Figs. 6a, 6b) and indeed showed the latter to have peaks and troughs corresponding to the undulations in the PPIX absorption spectrum. Although the action spectrum of the 635 nm AF spike was obtained from a single mouse, they involved extensive measurements (693 total spectra) with multiple replications (asterisks plotted in Figs. 5e, 5f; cf., also Supplementary Fig. S3 in Supplementry Appendix 1.3) and confirm the prediction of the highly distinctive undulatory PPIX absorption spectrum (Fig. 6b). 

Potential for the 635 nm Peak to Be an Artifact Resulting From Light Damage, Photoreceptor Bleaching, or the Diurnal Cycle

The MP laser damage power (Table) is ∼11 times higher than the power at the mouse pupil in these experiments. Consequently, the likelihood of light-induced toxicity is very low. Additionally, we examined the retinas of all mice used in the study with OCT and found no evidence of damage in the OCT images. 

Another potential spectral artifact might arise from photopigment bleaching, as the AF excitation levels used in most cases are calculated to cause 100% bleaching of rhodopsin.⁴² Specifically, during the excitation of AF with 488 nm light with the intensities and small ROIs used, we estimate that all the rhodopsin would be bleached in a few scans. Thus bleaching would be expected to initially attenuate the 488 nm excitation of fluorochromes in the RPE and choroid by up to twofold or so. However, because dozens of scans at each ROI were always taken, the overall effect would be negligible. Similarly, the emission spectra of fluorochromes in the RPE and choroid would only be affected by rhodopsin absorption during the return through the rod layer for one or two scans, so distortion would be negligible. Moreover, any potential distortion by bleaching will have a negligible effect on emission spectra in the region above 600 nm, where rhodopsin absorption is always negligible. Our experiments were typically conducted in the afternoon. At present we have no information as to whether there might be diurnal variations in the narrowband (or broadband) AF signals. However, the methods employed could potentially yield information about diurnal variation in the biosynthesis from the PPIX precursor of cytochrome c and essential heme-containing compounds in the posterior eye. 

Overall, our findings provide evidence for a novel fundus AF signal arising from ocular PPIX in vivo. This signal could serve as a valuable tool for investigating reactions in which PPIX participates, including iron metabolism and transport in the murine posterior eye, and disease conditions that affect these molecular mechanisms. 

Acknowledgments

The authors are grateful for the generous support of Marie Burns of UC Davis. We thank Xiaoting Yin at Dalian University of Technology for helping in preparing the first draft of Fig. 7(g). 

Supported by the National Eye Institute grants (P30 EY012576, R01 EY026556, R01 EY02660, R01 EY031098) and by the National Natural Science Foundation of China (62175024); Dalian University of Technology (DUT21RC(3)001). 

Disclosure: P. Zhang, None; S.K. Manna, None; M. Goswami, None; R.J. Zawadzki, None; E.N. Pugh, None