Multimodal photoacoustic microscopy and optical coherence tomography ocular biomarker imaging in Alzheimer's disease in mice

Abstract

Alzheimer's disease (AD) is a neurodegenerative disease characterized by amyloid beta (Aβ)-containing extracellular plaques and tau-containing intracellular neurofibrillary tangles. Reliable and more accessible biomarkers along with associated imaging methods are essential for early diagnosis and to develop effective therapeutic interventions. Described here is an integrated photoacoustic microscopy (PAM) and optical coherence tomography (OCT) dual-modality imaging system for multiple ocular biomarker imaging in an AD mouse model. Anti-Aβ-conjugated Au nanochains (AuNCs) were engineered and administered to the mice to provide molecular contrast of Aβ. The retinal vasculature structure and Aβ deposition in AD mice and wild-type (WT) mice were imaged simultaneously by dual-wavelength PAM. OCT distinguished significant differences in retinal layer thickness between AD and WT animals. With the unique ability of imaging the multiple ocular biomarkers via a coaxial multimodality imaging system, the proposed system provides a new tool for investigating the progression of AD in animal models, which could contribute to preclinical studies of AD.

1. Introduction

As the most common form of dementia worldwide, Alzheimer's disease (AD) is a progressive neurodegenerative disease characterized by amyloid beta (Aβ)-containing extracellular plaques and tau-containing intracellular neurofibrillary tangles [1–3]. Although no effective treatment for AD currently exists, interventions in the prodromal stages could significantly benefit disease management. Thus, biomarkers that enable the identification and monitoring of AD progression during its prodromal period are crucial [4,5]. Studies have demonstrated that the accumulation onset of Aβ and tau occurs decades before the structural abnormalities and the cognitive decline [6]. Recent studies on FDA-approved drug targeting on clearing brain amyloid plaque suggested that early intervention on AD could slow pathological progress and cognition decline [7,8]. The existing gold standard imaging modality, positron emission tomography (PET), uses Aβ-specific radiotracers to detect the biomarker in the brain. Additionally, cerebrospinal fluid (CSF) test is also used clinically to substantiate the diagnosis of AD [9,10].However, the invasiveness of CSF test and the high cost and limited access associated with PET pose considerable challenges for screening larger populations and predicting the progression of AD [11]. Thus, more accessible and non-invasive imaging technologies together with reliable imaging biomarkers are essential for early diagnosis and effective therapeutic interventions.

As an anatomical and developmental extension of the brain, the retina shares many parallels with it, including vasculature, neurons, and immune responses [12]. For many non-invasive and high-resolution optical imaging methods, retina offers a unique area of optical accessibility for the central neural system [13]. Both human post-mortem histopathological studies and studies using transgenic mouse model, such as APPswe/PS1dE9 [14], have demonstrated the quantitative and temporal correlations between brain and retina in AD. Capillary deposition of Aβ has been demonstrated in both retinal vasculature of AD mouse models such as Tg2567 and postmortem retinas of AD patients [15,16]. In addition, Aβ deposits were also found in multiple retinal layers, including ganglion cell layer (GCL), inner nuclear layer (INL), and nerve fiber layer (NFL). The correlation between Aβ deposits in the retina and AD progression has been demonstrated in transgenic AD mouse models [17,18]. However, it is important to note that different AD mouse models may exhibit varying degrees of pathology, the complexity of AD pathology across different models suggests that findings in one model may not be fully generalizable to others.

Evidence from histology and optical coherence tomography (OCT) has disclosed the abnormal retinal structure in AD disease, including ganglion cell losses (the output neurons of the eye), retinal NFL thinning (ganglion cell axons), and optic nerve degeneration [19]. A decrease of total retinal thickness is noted with AD [20–22]. It should be noted, however, that the structural thickness changes in AD patients are inconclusive and dependent on the cohorts, measurement region [23]. Some studies have shown non-significant changes in AD patients regarding the retinal NFL thickness, while some studies showed the association of structural changes between AD and syndromes including global and posterior cortical atrophy [23–26]. In addition, the structure and functional changes of retinal microvasculature were also found to have correlations with the progression of AD [27]. The potential for using vascular biomarkers in early diagnosis of AD has been demonstrated with reduced retinal vascular density and vascular perfusion detected in both AD patients and mouse models having early cognitive impairment associated with AD [28]. Although several ocular biomarkers have been demonstrated to have correlation with AD, some of them are also related to eye diseases, such as Aβ deposition in macular degeneration [29]. Thus, a multimodality imaging system that can image multiple ocular biomarkers simultaneously or sequentially will be highly beneficial to the diagnosis, progression, and treatment monitoring of AD.

Our previous publications have described a high-resolution multimodality ocular imaging system, which integrated optical coherence tomography (OCT), photoacoustic microscopy (PAM), and fluorescence microscopy (FM). Its performance in imaging retinal structure changes, vasculature changes, and molecular biomarkers have been demonstrated in rabbit eyes with various disease models [30–32]. In this study, an upgraded system with both mouse eye imaging capability and Aβ deposition molecular imaging capability using anti-Aβ antibody-conjugated Au nanochains (AuNCs) as contrast agents was developed to provide multiple ocular biomarkers imaging for AD disease. We conducted the assessments of both retinal structural change and Aβ deposition using TG2576 mice, a well-established AD mouse model.

2. Method

2.1. Antibody-conjugated AuNCs

2.1.1. Chemicals

All chemicals were used as received without further purification and were at a minimum of ACS grade. The bulk gold target (16 mm long, 8 mm wide, 0.5 mm thick, and 99.99% purity) used in the fabrication of 20 nm diameter Au nanoparticle (NP) monomers by pulsed laser ablation method [33] was a product of Alfa Aesar (Ward Hill, MA). Pentapeptide with an amino acid sequence CALNN having purity higher than 95% were custom synthesized by RS synthesis LLC (Louisville, KY). The thiol-terminated methoxy-poly (ethylene glycol) with a molar mass of $2000\text{g}\cdot \text{mo}{\text{l}}^{-1}$ (mPEG 2k-SH) was purchased from Creative PEGWorks (Chapel Hill, NC, catalog number: PLS-605). Cysteamine (CAS number: 60-23-1, catalog number: 30070), ethylenediaminetetraacetic acid disodium salt dihydrate (EDTA-Na, CAS number: 6381-92-6, catalog number: E4884), Traut’s Reagent (2-Iminothiolane·HCl, CAS number: 4781-83-3, catalog number: I6256), bovine serum albumin (BSA, CAS number: 9048-46-8, catalog number: A7906), and monoclonal anti-Aβ antibody clone BAM-10 (catalog number: A3981) came from Sigma-Aldrich (St. Louis, MO). The BAM-10 antibody recognizes the epitope within amino acids 1-12 of Aβ, and has been shown to bind with Aβ in living Tg2576 mice brain [34,35]. Phosphate buffered saline (PBS, pH 7.4, catalog number: 10010023), sodium phosphate monobasic monohydrate (CAS number: 10049-21-5, catalog number: 424395000), and Zeba spin desalting columns (40 kDa molecular weight cut off and 0.5 mL, catalog number: 87766) were purchased from ThermoFisher Scientific (Waltham, MA). Peptide, PEG, cysteamine, EDTA-Na, Traut’s Reagent, and BSA were in powder form. Prior to the use of these chemicals, they would be dissolved in deionized (DI) water having an electric conductivity less than 0.7 $\mu \mathrm{S}\cdot \text{c}{\text{m}}^{-1}$ . All solutions were freshly made and used within twelve hours.

2.1.2. Instrumentations

Images of the colloidal gold nanoparticles (AuNPs) were recorded using a TEM (JEOL 2010F, Japan) at an accelerating voltage of 100 kV. Hydrodynamic diameter size and zeta potential of AuNPs were characterized via DLS analyses using a Nano-ZS90 Zetasizer (Malvern Instrument, Westborough, MA). UV–VIS absorption spectra were recorded by a spectrophotometer (UV-3600, Shimadzu Corp., Japan).

2.1.3. Fabrication of AuNP monomers

We first produced raw capping agent-free spherical colloidal AuNP monomers used for the fabrication of AuNCs in the present study via femtosecond pulsed laser ablation (PLA) of a bulk Au target immersed in DI water as previously described [33]. This method uses tightly focused micro-joule (µJ) femtosecond laser pulses to produce colloidal solution of NPs. The size distribution of generated NPs can be precisely controlled by optimizing laser parameters, such as wavelength, pulse energy, duration, and repetition rate.

Briefly, a laser beam emitted from the ytterbium-doped femtosecond fiber laser (FCPA µJewel D-1000, IMRA America, Ann Arbor, MI) operating at 1045 nm wavelength with a repetition rate of 100 kHz, 10 µJ pulse energy, and 700 fs pulse duration was first focused by an objective lens and then reflected by a scanning mirror to the surface of the bulk Au target immersed in flowing DI water (18 MΩcm). The size of the laser spot on the Au target was estimated to be 50 µm and its position was precisely controlled by the scanning mirror. AuNPs produced by PLA were partially oxidized by oxygen present in solution. These Au-O compounds were hydroxylated, followed by a proton transfer to give a surface of Au-O^- as described by Sylvestre, J.-P. et al. [36]. Therefore, the generated AuNPs were naturally negatively charged, and no capping agents and stabilizing ligands were required for maintaining their colloidal stability. This unique feature of having capping-agent free surface for the AuNPs produced by the PLA allows versatile surface modification to obtain controllable surface chemistry [33], which is crucial for self-assembling them into one dimensional (1D) AuNCs (to be explained later on in this paper). Colloidal solutions of AuNP monomers with an average diameter of 20 nm were produced and used in our experiments. The generated AuNPs have a narrow size distribution and an absorption peak at 520 nm due to the localized surface plasmon resonance (LSPR).

2.1.4. Self-assembly of 20 nm diameter AuNP monomers into near-infrared (NIR)-absorbing AuNCs

The self-assembly was performed by modifying surface of AuNP monomers with two different types of ligands, pentapeptide with an amino acid sequence CALNN and cysteamine, in a sequential manner by first mixing the colloidal solution of AuNPs with CALNN and then cysteamine. The binding of CALNN peptides and cysteamine to the AuNPs relies on strong anchoring of the Au-sulfur bonds, which covalently attach two ligands onto the surface of AuNPs. The function of bound CALNN peptides was to improve AuNP colloidal stability via enhancing interparticle electrostatic repulson, which is crucial to achieve a right balance between the repulsive potential and attractive potential after the addition of cysteamine molecules, a governing factor in the linker-mediated self-assembly of NPs [37,38]. Cysteamine molecules, containing two reactive terminal groups, sulfhydryl (-SH) and amine (-NH2), can link AuNPs via attaching to their surface by either covalent bonds (-SH groups) or electrostatic attraction (-NH2) thereby forming 1D AuNCs. It is worth mentioning that only minimum amount of CALNN peptides and cysteamine molecules required for inducing structurally stable chain formation were used for the surface modification of AuNPs. In this way, enough space will be left on the surface of AuNCs for subsequent anti-Aβ antibody conjugation.

In a typical process, 50 mL colloidal solution of 20 nm diameter AuNPs with OD 1 at 520 nm was mixed with 100 µL CALNN peptide solution with concentration of 1 milliMolar (mM) to achieve a defined molar ratio of 2000:1 between CALNN peptides and AuNPs. Then, the mixture was kept undisturbed for 2 hrs at room temperature (RT) to allow sufficient conjugation of CALNN peptides to the AuNPs via Au-sulfur bonds. Following the conjugation of CALNN peptides, surface of AuNPs was further modified with cysteamine molecules by adding 90 µL cysteamine solution with concentration of 1 mM to achieve a molar ratio of 1800 between cysteamine molecules to AuNPs. The mixture was kept undisturbed until the observation of significant color change from red-pink to blue, typically occurring at 24 h or serval days after addition of cysteamine molecules, which is a clear evidence of a successful self-assembly of AuNP monomers into AuNCs. After the formation of NIR-absorbing AuNCs with an absorption peak at 700 nm, they were spun down to a pellet using a centrifuge and the final mass concentration was adjusted to 0.5 $\text{mg}\cdot \text{m}{\text{L}}^{-1}$ via adding DI water to the pellet after removing the supernatant.

2.1.5. Conjugation of anti-Aβ antibody onto AuNCs

For effective and selective targeting AD biomarkers in mouse’s eyes, anti-Aβ antibodies were covalently conjugated onto the surface of AuNCswith the synthetization shown in Fig. 1(A). Before this covalent conjugation, thiolation of anti-Aβ antibodies was performed by using Traut’s Reagent, which reacted with antibody primary amines to introduce sulfhydryl (-SH) groups. In this process, first, 200 µL of anti-Aβ antibody in 100 mM phosphate buffer (pH 8.0) containing 4 mM EDTA-Na at a concentration of 1 mg/ml was prepared. To initiate the reaction of introducing -SH groups to the antibody, 50-fold molar excess of Traut’s Reagent was added to the antibody by rapidly pipetting into the solution a necessary volume of freshly prepared 14 mM Traut’s Reagent and allowing the mixture to be gently shaken at 100 rpm for one hour at RT. Upon completion of this reaction, excess Traut’s Reagent was removed from the thiolated antibody via running through desalting columns equilibrated with 100 mM phosphate buffer (pH 8.0) containing 4 mM EDTA-Na for twice. Then, 200 µL thiolated antibody at a concentration of 1 $\text{mg}\cdot \text{m}{\text{L}}^{-1}$ was transferred to a fresh microfuge tube that had been rinsed with DI water to remove trace impurities or contaminants.

Fig. 1.

(A) 2 Method of conjugating Anti-Aβ antibodies onto Au nanochains. (B) The optical absorption spectrum of Au nanochain. (C) The sizes distribution of Au nanochains.

Once the thiolated anti-Aβ antibodies were obtained, they would be covalently conjugated onto AuNCs right away. Such conjugation was facile due to the high affinity of gold-thiol bonds. Briefly, after rinsing a 15 mL centrifugal tube with DI water, it was filled with 0.2 mL of 100 mM phosphate buffer (pH 7.4) and 5 mL of AuNCs with a mass concentration of 0.5 $\text{mg}\cdot \text{m}{\text{L}}^{-1}$ . Then, freshly prepared 0.2 mL thiolated anti-Aβ antibody with mass concentration of 1 $\text{mg}\cdot \text{m}{\text{L}}^{-1}$ was rapidly added to the tube and immediately mixed by pipetting up and down or vortex. After allowing the mixture to shake at 100 rpm for one hour at RT, 100 µL of mPEG 2k-SH with concentration of 1 mM was added to passivate the surface area of the AuNCs which was not covered by antibodies. PEG molecules were used because they can improve stability, biocompatibility, and simultaneously minimize nonspecific interactions with biological tissues under physiological conditions by providing a hydrophilic steric barrier. The mixture of antibody-conjugated AuNCs and mPEG 2k-SH was shaken at 100 rpm for two hours at RT to enable sufficient binding of mPEG 2k-SH molecules to the AuNCs. After this reaction, antibody-conjugated AuNCs were further blocked by adding to them 5 mL of 4 mM phosphate buffer (pH 8.0) containing 10 $\text{mg}\cdot \text{m}{\text{L}}^{-1}$ BSA. The resultant solution was allowed to shake at 100 rpm for an additional one hour before being centrifuged at 2000g for 0.5 hour until a pellet was formed. After removing the supernatant, the mass concentration of the antibody-conjugated AuNCs was adjusted to 2.5 $\text{mg}\cdot \text{m}{\text{L}}^{-1}$ by resuspending the pellet with 1.0 mL of 4 mM phosphate buffer (pH 8.0) containing 5 $\text{mg}\cdot \text{m}{\text{L}}^{-1}$ BSA. Finally, the antibody-conjugated AuNCs were filtrated through a 0.22 µm filter into a sterile 2 mL tube for future use. The obtained anti-Aβ antibody-conjugated AuNCs can be stored at 4 ^oC for the life of the antibody. A 10 nm red shift in the peak absorption wavelength of the AuNCs from 700 nm to 710 nm was observed after the covalent conjugation of anti-Aβ antibodies.

2.1.6. Characterization of AuNCs

The obtained AuNC and anti-Aβ antibody-conjugated AuNCs were characterized by an array of analytic instruments and techniques, including TEM, UV-VIS absorption spectroscopy, and dynamic light scattering (DLS) measurements. TEM images were recorded at an accelerating voltage of 100 kV. UV-VIS absorption spectra of AuNCs recorded from 350 nm to 800 nm were shown in Fig. 1(B). DLS mea­surements were employed in Fig. 1(C) to determine hydrodynamic diameters of AuNCs before and after the conjugation of anti-Aβ antibodies. All measurements and processes were carried out at RT, approximately 20°C.

2.2. Imaging system

The experimental setup for dual-wavelength PAM and OCT dual-modality imaging system are shown in Fig. 2 [30,31]. Two pulsed nanosecond fiber lasers (GLPM-16-1-10-M, IPG Photonics) working at a wavelength of 532 nm with pulse repetition rate up to 600 kHz were employed as the illumination sources for PAM. The output from one of the lasers was coupled into a 40-meter polarization maintaining single model fiber (PM-SMF) through an achromatic doublet lens (AC254-040-A, Thorlabs). The incident 532 nm laser pulses went through Raman scattering in the PM-SMF, where discreate longer wavelength peaks, i.e., Stokes wave with around 440 cm^-1 frequency separation were generated [39]. After passing through a fiber collimator, the PM-SMF output beam was filtered by a bandpass filter (FB710-10, Thorlabs) to extract the 11^th Stokes peak with 710 nm wavelength and a broadened 20 nm bandwidth, which matched with the optical absorption peak of antibody-conjugated AuNCs. Then, the PAM-SMF output beam was combined with the 532 nm beam from another laser through a dichroic mirror (DM1 in Fig. 2, FF556-SDi01-25 × 36, Semrock).

Fig. 2.

The schematic of the dual-wavelength multimodality imaging system.

After passing through a spatial filter, the two laser beams were coaxially aligned. Part of the mixed beam went to a photodiode (PD in Fig. 2, PDA36A, Thorlabs) to monitor the laser energy fluctuations, which was digitized by the DAQ card (RazorMax PCIe CSE161G4, Dynamic Signal Inc) with a 250-MHz sampling rate. The OCT illumination light (Spectral Domain OCT System, TEL321, Thorlabs) with the central wavelength of 1300 nm was coaxially aligned with the PAM excitation light via a multiphoton single-edge dichroic beam splitter (DM2 in Fig. 2, FF925-Di01-25 × 36, Semrock). Sharing the same galvanometer, the excitation lights of the two different imaging modalities were delivered and focused on the same area of the retina through a telescope configuration, which comprises a scan lens (SL in Fig. 2, LSM03-BB, Thorlabs) and an ophthalmic lens (OL in Fig. 2, AC080-010-A, Thorlabs).

The OCT reflection light from the sample directly went through the second dichroic mirror (i.e. DM2), and combined with the reference light from the reference arm to provide interference, which was detected by the line scan camera with up to 146-kHz repetition rate. The acoustic wave induced by the PAM illumination light was acquired by a custom-built needle-shaped ultrasound transducer with a central frequency of 30 MHz (Optosonic Inc., Arcadia, CA, USA) and amplified by a homemade 70 dB amplification circuit before digitization. A four-channel delay generator (DG535, Stanford Research Systems) triggered by the internal clock with 150 kHz pulse repetition rate was used to precisely trigger the GLPM laser, the galvanometer, and the DAQ card. With a scanning area of 512 × 512 points, it took about 1.75 seconds to obtain one PAM or 3D OCT image.

2.3. Animal preparation

All the experimental procedures were performed in accordance with the ARVO (The Association for Research in Vision and Ophthalmology) Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Institutional Animal Care & Use Committee (IACUC) of the University of Michigan (Protocol PRO00010388, Multimodal Imaging and Treatment of the Eye).

Six Tg2576 male AD mice (APPswe, B6; SJL-Tg(APPSWE)2576Kha, Taconic, Europe) and four B6 albino male wild type (WT) mice (Jackson lab, USA) were involved in this study. All of them are 44 weeks old. The retinal vessels of both groups were imaged in vivo. The animals were anesthetized with a mixture of ketamine (80 mg·kg^-1) and xylazine (5 mg·kg^-1) via intraperitoneal (IP) injection. The pupils of the eyes were dilated with topical application of phenylephrine hydrochloride 2.5% and tropicamide 1% eye drops. Topical tetracaine drops were applied for additional topical anesthesia prior to initiation of the experiments. After the anesthesia, the mouse was placed on the homemade holder with integrated resistance heating system to maintain the body temperature.

The retinal structure images were acquired using OCT, followed by manual image segmentation to separate different retinal layers. To evaluate in vivo structural changes, enface OCT images were taken from both eyes of the six AD mice and the four WT mice. For each eye, the OCT scan was performed around the optic nerve head and with a scanning area of around 1.4 mm by 1.4 mm. Subsequently, measurements of GCL plus the inner plexiform layer (IPL), INL plus outer plexiform layer (OPL), and outer nuclear layer (ONL) were obtained for both TG2576 mice and WT mice.

The PAM images of the retinal vasculature were obtained before the tail vein injection of anti-Aβ antibody-conjugated AuNCs (0.1 ml, 1 mg/ml) as the baseline. Follow-up imaging was conducted at three different time points (24 hours, 48 hours, and 72 hours) after contrast agent administration. A pulse energy of around 40 nJ, which was lower than the ANSI safety limit [40], was applied for PAM imaging. During anesthesia, both left and right eyes of the mice experienced opacification overtime, which could lower the image quality, especially for PAM imaging due to the shorter wavelengths being used. To minimize the difference in eye conditions, the right eye was the first eye imaged for each mouse involved in PAM imaging. Three AD mice were injected with anti-Aβ antibody-conjugated AuNCs (i.e., targeted group) and another three AD mice were injected with non-targeted AuNCs as a negative control group (i.e., control group 1). In addition, three mice in the WT group were injected with anti-Aβ antibody-conjugated AuNCs as another control (i.e., control group 2). One mouse in the WT group was excluded from PAM imaging due to the cloudified eye condition. Buffered saline solution (BSS, Altaire Pharmaceuticals, Inc., Aquebogue, NY) was applied to rinse the cornea to keep it moist while coupling the ultrasound transducer to the sclera.

2.4. Data analysis

The 3D OCT data were acquired through ThorImageOCT software (Thorlabs), and processed in MATLAB (Mathworks Inc., USA) and Image J (National Institutes of Health, USA). Each data contains 512 × 512 depth-resolved data points corresponding to a 1.4 mm by 1.4 mm area. To calculate the layer thickness, five B-scan cross sections were chosen, namely, one that passed through the center of the optic nerve head, two that were 0.2 mm away from the first B-scan on both sides, two that were 0.3 mm away from the first B-scan on both sides. The same procedure was repeated in the perpendicular direction, yielding 25 intersections on the retina. Except for the optic nerve head, the other 24 data points were used for thickness calculation of each layer following manual segmentation of each B-scan. The average of the thickness from the 24 data points was used as the thickness for statistical analysis. Comparison of the group-averaged layer thickness between AD group and WT group was performed by two-sample t-test, p < 0.01 was considered significant. The effect size is defined by Cohen’s d. Results are presented as mean ± SD.

The PA signal is processed by MATLAB (Mathworks Inc.), the intensity of antibody-conjugated AuNC was quantified by averaging the peak-to-peak intensity of the PA signal throughout the entire image. The quantified intensity of the AD group injected with anti-Aβ antibody-conjugated AuNCs was presented as mean ± SD.

2.5. Histological analysis

To evaluate the Aβ deposition in the retina, the mice were euthanized 72 hours after intravenous administration of AuNCs, and the eyes were harvested for further histological analysis. Mice were euthanized with carbon dioxide overdose. The eyes from the targeted group and the control groups were removed aseptically from the euthanized mice. The isolated samples were fixed in Davidson fixation solution (A3200, ITW Reagents) for 24 hours. The fixed tissues were embedded in paraffin. Subsequently, the paraffin-embedded tissues were sliced to a thickness of 4 µm and stained with hematoxylin and eosin (H&E), CD31, and Recombinant Anti-Aβ antibody (mOC64, Abcam). The antibody recognizes Aβ1-42 peptide in a variety morphology, including monomeric, oligomeric and fibrillar forms, and it presented extensive colocalization with conventional Aβ antibody, 6E10, in human brain tissue [41]. The slides were imagined by using a Leica DM600 light microscope (Leica Biosystems, Nussloch, Germany).

3. Result

3.1. Retinal structure imaging

The retinal structure of the AD mice and the WT mice were evaluated with OCT imaging. Five B-scan images near the optical nerve were extracted from 3D OCT followed by segmentation to distinguish retinal layers. The average thickness from five B-scan images were present as the thickness for each animal. As presented by the statics results with 12 eyes in AD group and 8 eyes in WT group, the retina thickness for each retinal layer of AD mice is significantly lower compared with the WT mice. The thickness of GCL + IPL in AD mice was 0.083 mm, decreasing significantly from 0.090 mm in the WT mice evaluated by a 2-sample t-test (p < 0.01). The effect size, defined by Cohen’s d, is 2.28, indicating a large differentiating between the two groups. In the meantime, no significant difference in the thickness of INL + OPL was noticed between the AD mice and the WT mice, with an effect size of 0.65. Conversely, a significant increase in the thickness of ONL was observed in AD mice when comparing with the WT mice evaluated by a 2-sample t-test (p < 0.01). The effect size is calculated to be 2.20. In addition, as shown in Figs. 3(A) and 3(B), a decrease in total retinal thickness from the WT mice to the AD mice was observed.

Fig. 3.

The retinal structure evaluated with OCT. (A) OCT B-scan image from a WT mouse; (B) OCT B-scan image from an AD mouse; Vertical scale bar: 100 µm, horizontal scale bar: 200 µm. (C) The comparison of retinal layer thicknesses between the AD mice and the WT mice (n = 12 eyes in AD, n = 8 eyes in WT). “**” p < 0.01 for a two-sample t-test, error bar represents standard deviation.

3.2. Aβ accumulation imaging

The PAM images of the AD mice were obtained with two laser wavelengths at different time points before and after injecting anti-Aβ antibody-conjugated AuNCs. The retinal vasculature was imaged at 532 nm wavelength at which hemoglobin has strong optical absorption. The spatially distributed AuNCs were imaged at 715 nm, where AuNCs showed an optical absorption peak (Fig. 1(B)). In order to confirm the PAM signal at 715 nm was from AuNCs, the PAM images at both wavelengths before AuNCs injection were also acquired as the baselines. While retinal vasculature can be clearly distinguished in the image at 532 nm, no PAM signals can be seen in the image at 715 nm. As shown in Fig. 4(A), no significant changes occurred in retinal vasculature at 24 hours, 48 hours, or 72 hours after AuNCs injection. For the PAM images at 715 nm, the AuNCs started to show accumulation at 24 hours after injection. As shown in Fig. 4(B), the quantified PA signal intensity from the AuNCs reached the peak at 48 hours and then decreased at 72 hours. At 48 hours, the quantified PA signal intensity is about five times higher than the background, indicating that AuNCs can provide sufficient contrast enhancement for imaging the molecular information of Aβ deposition.

Fig. 4.

(A) The dual-wavelength (532 nm and 715 nm) PAM imaging of AD mice retinal vasculature and antibody-conjugated AuNCs at different time points before and after AuNC injection. The fusion images show both the locations of the AuNCs (in pseudo green color) and the retinal vasculature. Scale bar: 500 µm. (B) The photoacoustic (PA) signal intensity at 715 nm over time indicating the time-dependent accumulation of the anti-Aβ antibody-conjugated AuNCs.

To further validate that the signal enhancement in the 715 nm PAM images observed from the targeted group, as shown in Fig. 4, is indeed due to the targeting of the antibody-conjugated AuNCs toward the Aβ in the retinal microvasculature, PAM imaging experiments were also conducted in the two control groups. Figure 5(A) shows the results from the control group 2, i.e., WT mice injected with antibody-conjugated AuNCs. Although the retinal vasculature was imaged successfully at 532 nm at different time points before and after AuNCs injection, no PA signal enhancement was noticed in the images at 715 nm after injection, indicating that there was no accumulation of AuNCs in the retinal microvasculature. Figure 5(B) shows the results from control group 1, i.e., AD mice injected with non-targeted AuNCs. Again, no PA signal enhancement was noticed in the images at 715 nm after injection, indicating that there was no accumulation of AuNCs in the retinal microvasculature.

Fig. 5.

The dual-wavelength (532 nm and 715 nm) PAM imaging of retinal vasculature and AuNC accumulation at different time points before and after injection. Scale bar: 500 µm. (A) Imaging results from a WT mouse injected with antibody-conjugated AuNCs(control group 2). (B) Imaging results from an AD mouse injected with non-targeted AuNCs (control group 1).

3.3. Histology analysis

The H&E staining, CD31 staining, and mOC64 staining, as shown in Fig. 6, were used to demonstrate the anatomical structure of retinal layer, the retinal vasculature wall, and the Aβ deposition in the retinal microvasculature, respectively. The H&E staining result in Fig. 6(A) clearly exhibited the different layers of the mouse retina. With the CD31 staining, the retinal blood vessel wall, as marked by the blue arrows in Fig. 6(B), could be distinguished with black color. The Aβ staining with Recombinant Anti-Aβ antibody (mOC64) showed the Aβ depositions in different layers with brown color. We can see that the Aβ staining exhibits some overlap with the CD31 staining (marked by the blue arrows), demonstrating the Aβ’s accumulation in the retinal vasculature. In addition, the Aβ accumulation was also found in the OPL.

Fig. 6.

Histological images showing retinal layer structures and Aβ deposition. (A) The H&E staining image of the retinal structure of an AD mouse. (B) The CD31 staining image of an AD mouse retina with the blue arrows indicating the endothelial cells of retinal microvacsculature. (C) The recombinant anti-Aβ antibody staining image of the AD mouse indicating the depostion of Aβ in the retinal microvasculature. (D) The H&E staining image of the retina of an WT mouse. (E) The CD31 staining image of the WT mouse retina, blue arrows indicate the endothelial cells. (F) The recombinant anti-Aβ antibody staining image of the WT mouse retina. Scale bar: 50 µm.

As comparison, the staining images for WT mouse are shown in Fig. 6(D), (E), (F). Due to the limited number of samples we acquired, few samples with CD31 staining showing well shaped blood vessel were obtained. In Fig. 6(E), the blue arrow indicates a distorted vessel in the GCL, the distortion might be due to the slicing process. As opposed to the AD group, at the same position in figure (F), the anti-Aβ staining response was negative.

4. Discussion

This study presents our initial work utilizing a multimodality OCT and PAM system to identify the imaging biomarkers of AD in a mouse model. The relatively small size of the mouse eye requires specific selection of telescope lenses to match the entrance beam with the pupil size. In our system, a collimated Gaussian beam of about 1 mm in diameter was acquired at the entrance of the pupil. Based on PAM, we demonstrated, for the first time, in vivo imaging of both microvasculature and Aβ deposition in mouse retina. Based on OCT, we also imaged the structural information including the thicknesses of multiple retinal layers. From the imaging results, we found that the GCL + IPL layer was thinner in the AD mice than the WT mice, while the ONL layer was thicker in the AD mice than the WT mice. The result here is consistent with the findings in previous publications [20,42,43]. We also successfully demonstrated the labeling and visualization of Aβ deposition in the retina of the AD mice by using the AuNCs as the contrast agent. The images from the AD mice showed enhanced contrast in the PAM images at 715 nm after the injection of the targeted AuNCs, which was not observed in any of the control groups. The PAM imaging findings were confirmed by histology and agreed with previous publications [44].

A multimodality ocular imaging system, as employed in this work, is able to combine the advantages offered by each modality. By leveraging the strong optical absorption of AuNCs as a contrast agent for labeling Aβ, PAM is capable of mapping the spatially distributed Aβ as a molecular biomarker of AD. The optical absorption peak of AuNCs at 715 nm avoids the strong optical absorption of hemoglobin, which facilitates highly sensitive imaging of Aβ deposition in retinal blood vessels. OCT is a standard imaging tool in ophthalmology clinic for non-invasive measurement of the changes in cross-sectional retinal structure. In previous work, GCL-IPL thinning was found to have a strong correlation with AD in both human and mouse [20,42]. This structural change, however, was also found to be correlated with some optic nerve diseases [45]. Although less emphasized, ONL thickening was also observed in both human and mice with AD [42,43]. The OCT results reported in this work agreed with those previous findings, and further demonstrated that OCT could provide valuable imaging biomarkers for AD. Most importantly, given that each of the observed biomarkers could be induced by other eye diseases, a multimodality system providing a comprehensive view of different aspects of AD would be highly beneficial to the diagnosis and treatment monitoring of this disease.

There are several limitations of this study. First, because the involved AD mouse model is very costly in both time and money, this study included only a limited number of animals, and control group used mice from a different vendor. Ideally, control group should use mice with the same genetic background as the experiment group. Additionally, in the retinal layer thickness analysis, two eyes of the same animal were included as two independent variables while they were dependent to some extent. As a result, the study is limited to the feasibility of detecting structural and pathological biomarkers with the multimodality system. A more comprehensive study is warranted where a designed number and strain of animals shall be chosen based on a power calculation to represent the studied phenotype. Second, limited investigation into the type, morphology and location in retina of Aβ peptide was covered in this study. The choice of antibodies affects the targeted regions of Aβ peptide, and further affects the various forms of Aβ being detected. Studies have shown that antibodies such as 6E10 (targeting Aβ, amino acid (aa) 1-16), 4G8 (targeting Aβ, aa 17-24), and 12F4 (targeting C-terminus of Aβ, aa 34-40) could stain differently in mouse and human retina [44]. Some antibodies (e.g., 6E10 and 12F4) exhibit overlap with amyloid precursor protein (APP) specific staining in retina, where APP could also be expressed in normal retina cells [46]. In addition, the large size of the antibody conjugated AuNC (∼100 nm) could affect its ability to penetrate blood retina barrier, thus limiting the detection of Aβ deposits out of blood vessels. Immunohistochemical or immunofluorescent studies utilizing antibodies that target specific types of Aβ could help substantiating the identification of Aβ in the retina. These studies, combined with in-vivo PAM imaging, could provide a more comprehensive understanding of how the PAM images correspond with Aβ morphology and its distribution in the retina. Third, the instrumentation could be improved to obtain higher resolution and more consistent results. In this work we were not able to observe any changes in density of retinal micro vasculatures in AD mice as reported by a previous work in humans [47]. These microvasculature changes usually belong to the superficial vessel plexus or deep vessel plexus, with vessel diameters typically smaller than 25 µm. The small pupil size and overall dimensions of the mouse eye pose challenges for alignment and beam focusing on the retina, which contributes to the limited spatial resolution of our current PAM system. To assess morphological changes in retinal micro vasculatures as another potential biomarker of AD, optics that are optimized for ocular imaging could be used to improve the lateral resolution of PAM. In addition, the varying optical power of the mice lens could lead to increased imaging time of focus adjustment and inconsistency in the sharpness of PAM images. The limited field of view of the needle transducer also poses challenges for a more homogeneous imaging of the entire retina. To address these issue, a real-time fundus camera guided PAM could be employed for faster imaging process [48], while a ring shaped transducer designed for mouse eye might provide better image quality. At last, in this work, in addition to Aβ deposition, PAM was only utilized to image the retinal vascular structure without assessing their functional properties such as blood flow and blood oxygen saturation. We expect that these functional measurements of retinal micro vasculatures, which can be realized by advanced PAM technologies [49–52], may also reflect the onset and progress of AD and hence can be additional imaging biomarkers.

Our future plan includes further optimization of our imaging system as well as workflow, including employing automatic volumetric segmentation to provide region specific structural information. In addition, a longitudinal study involving a larger number of animals could substantiate the study with enhanced statistical significance. Our goal is to establish and validate a multimodality imaging platform for non-invasively and sensitively assessing a group of imaging biomarkers which can form a comprehensive matrix for AD diagnoses and monitoring.

5. Summary

In summary, we conducted proof-of-concept multimodality imaging of mouse retina and presented a non-invasive way of this imaging system that may offer a low-cost and practical tool to investigate multiple AD associated-biomarkers, including, but not limited to, retinal thickness and Aβ deposition. This system can potentially contribute to the understanding of the pathophysiology of AD as well as the future clinical management of this disease.

Funding

Fight for Sight10.13039/100002089 (FFSGIA16002); National Eye Institute10.13039/100000053 (1R01EY034325, R01EY033000).

Disclosures

The following authors have previously disclosed a patent application (no. US12029490B2) that is relevant to this manuscript: W.Z., X.W. and Y.M.P. Other authors declare no conflicts of interest.

Data availability

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.