Imaging rhodopsin degeneration in vivo in a new model of ocular ischemia in living mice

Abstract

Delivery of antibodies to monitor key biomarkers of retinopathy in vivo represents a significant challenge because living cells do not take up immunoglobulins to cellular antigens. We met this challenge by developing novel contrast agents for retinopathy, which we used with magnetic resonance imaging (MRI). Biotinylated rabbit polyclonal to chick IgY (rIgPxcIgY) and phosphorylthioate-modified oligoDNA (sODN) with random sequence (bio-sODN-Ran) were conjugated with NeutrAvidin-activated superparamagnetic iron oxide nanoparticles (SPION). The resulting Ran-SPION-rIgPxcIgY carries chick polyclonal to microtubule-associated protein 2 (MAP2) as Ran-SPION-rIgP/cIgY-MAP2, or to rhodopsin (Rho) as anti-Rho-SPION-Ran. We examined the uptake of Ran-SPION-rIgP/cIgY-MAP2 or SPION-rIgP/cIgY-MAP2 in normal C57black6 mice (n = 3 each, 40 μg/kg, i.c.v.); we found retention of Ran-SPION-rIgP/cIgY-MAP2 using molecular contrast-enhanced MRI in vivo and validated neuronal uptake using Cy5-goat IgPxcIgY ex vivo. Applying this novel method to monitor retinopathy in a bilateral carotid artery occlusion-induced ocular ischemia, we observed pericytes (at d 2, using Gd-nestin, by eyedrop solution), significant photoreceptor degeneration (at d 20, using anti-Rho-SPION-Ran, eyedrops, P = 0.03, Student's t test), and gliosis in Müller cells (at 6 mo, using SPION–glial fibrillary acidic protein administered by intraperitoneal injection) in surviving mice (n ≥ 5). Molecular contrast-enhanced MRI results were confirmed by optical and electron microscopy. We conclude that chimera and molecular contrast-enhanced MRI provide sufficient sensitivity for monitoring retinopathy and for theranostic applications.—Ren, J., Chen, Y. I., Mackey, A. M., Liu, P. K. Imaging rhodopsin degeneration in vivo in a new model of ocular ischemia in living mice.

The pathogenesis of retinopathy is poorly understood, at least in part because of the current inability to image disease-associated changes in the composition of cell types and gene expression in vivo. The neurovascular unit (NVU) of the retina includes astrocytes and Müller cells as well as amacrine and ganglion neurons. These cells deliver oxygen and nutrients from the microvasculature and define the physical and biochemical relationships among neurons, glia, and specialized vasculature, mediating their close interdependency in the CNS for energy homeostasis and neurotransmitter regulation. The retinal NVU is similar to that of the brain (1) and thus shares common biomarkers, the exception being rhodopsin (Rho), which is found uniquely in the photoreceptors of the retina. Given the proximity of the retina to the brain and its close interaction with the rest of the CNS, we applied target-specific contrast agents (CAs) and molecular contrast-enhanced (MCE) MRI that we have developed and validated for use in the brain to identify and evaluate molecular signatures of the retina. A major challenge in this undertaking is imaging the small cell populations of the retina with sufficient sensitivity. By using specific magnetic resonance (MR) CAs to target Rho and mRNA of glial fibrillary acidic protein (GFAP), we aimed to noninvasively identify photoreceptors and Müller cells by MCE-MRI in a mouse model of ocular ischemia.

The present work builds on our extensive experience of developing gene-targeting methods to noninvasively examine the cellular and molecular mechanisms that regulate neuroplasticity in health and disease conditions. By labeling standard T2 MR-CAs to small DNAs (18–26 nt in length), we have shown that MR-CAs enter the vascular endothelia by caveolae and are then transported through the blood–brain barrier and glial end-foot, then to the rough endoplasmic reticulum of specific cells where mRNAs are located (2, 3). Binding to correct mRNA has been validated by showing targeting MR-CAs with sequences complementary to RNA are hybridized to specific biomarkers in the CNS (4). Importantly, these targeting MR-CAs are visible in vivo (by MRI) as well as ex vivo (with optical and electron microscopy), thus lending themselves to validation for targeting specificity using conventional assays. On the other hand, normal resting mouse brains take up sODN with random (Ran) sequence or superparamagnetic iron oxide nanoparticle (SPION)-Ran transiently, and it is not visible in either assay. We have quantitatively measured gene transcripts using this approach in combination with TaqMan analysis, the results of which showed excellent linear regression (r² = 0.8–1.0) in normal and disease conditions (5). This modality exhibits near cellular resolution with signal specificity and sensitivity using clinical or high-field-strength magnets. We have further demonstrated the robustness of this technique for detecting specialized cells associated with the progression of brain diseases (as shown through measures of oxidative stress and inflammatory responses) and/or recovery (as evidenced by gliogenesis, angiogenesis, and neurogenesis).

Taking advantage of the arterial anatomy—that is, the branching of the ophthalmic arteries from the common carotid artery—we induced ocular ischemia in C57black6 mice by surgically occluding the carotid arteries for 60 min (BCAO-60). Here, we report that MCE-MRI monitors active gliosis in the retina after ocular ischemia in vivo, with successful delivery to photoreceptors of Rho-specific MR-CA administered by intraperitoneal injection or eyedrops, and ex vivo histology validated the region of interest (ROI) detected by MCE-MRI. Moreover, we validated the chimera design by finding the evidence that SPION-Ran, a nontargeting MR-CA, carried immunoglobulin to cellular antigen in the complete chimera, allowing it to pass the plasma membrane. The mechanism of chimera MR-CA specificity relied on the presence of immunoglobulin to cellular protein and allowed retention in the neurons according to the concentration and location of cellular protein. This technology has great potential to dramatically reshape future approaches in many areas of neurobiology, as well as to lay the groundwork for new preclinical research and eventual clinical advances.

MATERIALS AND METHODS

Animals and housing

All procedures were approved by the Massachusetts General Hospital Subcommittee on Research Animal Care in accordance with the Public Health Service Policy on the Humane Care and Use of Laboratory Animals. We examined adult male C57black6 mice (Taconic Farm, Germantown, NY, USA) (n ≥ 3 litters at a time), 2 to 3 mo of age (23 ± 2 g body weight). The animals were housed in cages (n = 4 per cage) in a room with controlled light cycles (12 h/12 h light/dark), and they had free access to water and standard lab chow. We conducted stratification for baseline MRI values to eliminate animals with abnormal brain anatomy before experimentation. Mice were operated on and tested in a random manner, with a masked observer performing all tests. We assigned each mouse a numerical identifier, which was given to the examiner; the treatment was revealed at the completion of testing for data analysis.

Imaging precursor

We linked NeutrAvidin (NA, a derivative of avidin) to Molday Ion (BioPal, Worcester, MA, USA) to facilitate binding of biotinylated sODN-Ran to SPION as an imaging precursor. Our SPION-NA, with a hydrodynamic diameter of 30 ± 20 nm or a core diameter of 8 ± 2 nm, can bind 30 nmol biotin per microgram of iron. After incubating SPION-NA (1 μg) to biotin-sODN-Ran (1 pmol) at 4°C for 16 h, we purified the resulting SPION-Ran by membrane filtration to remove unbound sODN, then stored it in sodium citrate buffer at a concentration of 25 mM (pH 7.8), in which SPION-NA has a shelf life of 3 mo (6). We used a designated syringe for each MR-CA to avoid cross-contamination.

Chimera MR-CA for cellular proteins of CNS

We incubated 10 μg biotin-labeled rabbit polyclonal immunoglobulin (IgP) to chick IgY (rIgPxcIgY, ab6752; Abcam, Cambridge, MA, USA), or alternatively, goat IgP to rabbit IgG (gIgPxrIgG, ab6720; Abcam) per microgram Fe in SPION-Ran at 4°C for 24 h. The shelf life of these Ran-SPION-rIgPxcIgY or Ran-SPION-gIgPxrIgG constructs was 1 mo at 4°C. One day before use, we incubated chick IgP to microtubule-associated protein 2 (MAP2) (cIgPxMAP2, ab5392) to Ran-SPION-rIgPxcIgY (5 μg/μg Fe) at 4°C for a complete chimeric MR-CA (Ran-SPION-rIgP/cIgYxMAP2). We conjugated rIgPxRho (ab3424) to Ran-SPION-gIgPxrIgG (16 h, 4°C) to make anti Rho-SPION-Ran. As a control, we prepared SPION-rIgP/cIgYxMAP2 without sODN-Ran. We delivered different doses of MR-CA, depending on the route of delivery. For the blood–retina barrier (BRB) bypass procedure, animals were provided a dose of either 40 μg/kg (for intracerebroventricular delivery) or 8 mg/kg (for intraperitoneal injection to mice with an intracerebroventricular port or intraperitoneal/intracerebroventricular delivery); the dose administered by eyedrop solution was 80 μg/kg to bilateral carotid artery occlusion (BCAO)-treated mice.

BCAO-60 to induce ischemic retinopathy

We used the BCAO model, as described in our earlier studies, for long-term examination of retinopathy (4). We randomly selected 6 mice each day and arbitrarily assigned them to 1 of 2 treatment groups: either sham operation (control) or BCAO (n = 3 each). The animals in the sham operation group underwent surgical procedures identical to those of BCAO, except that we did not occlude the arteries. Throughout the operation we maintained the animals’ body temperature at 36 ± 1°C using a heating blanket, and we monitored temperature with a rectal thermosensor. We used Doppler analysis to assess the blood flow in the middle cerebral artery in mice that underwent surgical procedures, positioning the Doppler probes to the skull during surgery (7). By the third day after BCAO-60, all of the surviving postischemic mice exhibited spontaneous activity in their home cages, were responsive to handling, and engaged in normal self-grooming behavior.

Molecular contrast enhanced MRI protocols

We performed in mice before and after BCAO following previously published procedure (34), using a 9.4 T MRI system (Bruker Avance system; Bruker Biospin MRI, Inc., Billerica, MA, USA). We measured changes in R2* values before and after MR-CAs and evaluated SPION-labeled gene expression in the R2* maps acquired using multi-echo gradient echo sequences: TR 800 ms, 6 echoes (TE = 1.94, 3.41, 4.88, 6.35, 7.82, and 9.29 ms), with spatial resolution of 0.1 × 0.1 × 0.25 mm.

Stratification before MCE-MRI

Before administering MR-CAs to the animals, we performed stratification procedures by acquiring baseline frequency of signal reduction (R2*) and constructed R2* maps of all mice scanned on different dates. One week before MR-CA delivery, we prepared an intracerebroventricular port by transiently inserting and removing a 28-gauge needle (8). On the day of delivery, we measured background R2* values again. Mice with background R2* values greater than 2 sds from the baseline R2* maps were eliminated from further study. After satisfying the stratification, we delivered complete chimera, unlinked control (SPION-rIgGXcIgY plus biotin-free sODN-Ran), or naive control (SPION-Ran or SPION-NA without Ig) to mice at a dose of 8 mg Fe/kg (n = 3 each group, or as determined by power analysis) intraperitoneal/intracerebroventricular delivery. We acquired MCE-MRI at 3 time points within 24 h after MR-CA delivery; we constructed R2* maps in ROIs in the contralateral striatum and cortex for statistical analysis.

Data acquisition and theoretical considerations of quantitatively measuring endogenous protein antigens by MRI

Before acquiring animal data on each imaging day, we performed automatic high-order shimming to reduce magnetic field inhomogeneity and to determine the ability to reproduce the MRI data acquired on different dates (9, 10). Because R2* values above the baseline are positively proportional to iron concentration, and because only photoreceptors specifically express Rho antigen, we compared R2* maps in all ROI in the retina, where R2* is the rate of signal reduction (R2* = 1/T2*, s⁻¹). We acquired background MRI before SPION delivery, then compared the background R2* values to baseline R2* values by Student's t test. If the background R2* measured at these ROIs in any one mouse was significantly different from the overall average baseline R2* values, then we concluded that the images from that animal contained significant B₀-related noise. In such cases, the data were eliminated from further analysis. If the background R2* values were not different from those of the overall baseline, we delivered the MR-CA as described. This protocol also minimizes off-target effects as shown in histology. To determine the pharmacodynamic profile of MR-CA uptake (n = 6 each protocol), we repeated the protocol to collect data representing an average from at least 2 mice at each time point.

Data analysis

Contrast-to-noise ratio is defined as the ratio of the difference between 2 imaging signals and the square root of the sd of the background noise (5). For our purposes, R2* maps before SPION delivery (baseline) showing endogenous iron levels served as background, and their sds are the noise-to-R2* maps of brains containing SPION-sODN. Therefore, we defined the representative SPION-sODN uptake in each ROI at any given time point as the change in contrast—that is, ΔR2* (R2*_SPION-sODN – R2*_baseline) divided by noise (the square root of the sd of R2* within the same ROI in baseline brains).

To determine the optimal dose, we compared R2* values in the ROI contralateral to the puncture site. We analyzed the data to determine the optimal dose for peak retention and the proper clearance time (generally, within 24 h) in animals that were administered the CA. We used GraphPad Prism software (La Jolla, CA, USA) to perform statistical analysis of R2* values.

Examinations of within- and between-litter differences

We delivered MR-CAs to at least 4 but no more than 6 mice for each MRI acquisition, and on each day we examined a control (sham operation or nontargeting MR-CA) group along with the experimental (BCAO or targeting MR-CA) group. We acquired at least 5 brain slices (0.5 mm each) from each mouse (11). When we compared the data from mice in the control group to assess potential variations, we found that the mean and sd for each ROI in the 5 brain slices were statistically the same (Student's t test, P > 0.05). Furthermore, the mean and standard error of the mean (sem) of the 2 controls were not statistically different from those of previous control groups that received similar treatment on different days. If they were, the data from those mice were eliminated from further analysis. By analyzing the accumulated data from control mice scanned on different days, as well as from different litters, we examined within- and between-litter differences as well as potential spikes of inhomogeneity related to B₀.

Power analysis for number of animals required

To determine the overall number of mice required for each experiment, we performed a power analysis using the means and sem from the experimental group and the control group scanned each day. By using power analysis to calculate sample size, we could minimize type II error. Once we had completed MRI scans on the minimum number of mice, we performed statistical analysis. The mean R2* values and sem were calculated from the average values in each group. We aligned T2 MRI before constructing the R2* maps, and the percentage increase above predelivery baseline R2* maps was computed and shown as ΔR2*: (R2*_postdelivery − R2*_baseline)/R2*_baseline × 100%. We used a post hoc power analysis (P = 80% at α = 0.05) to compute the sample size required to avoid a type II error.

Validation of delivery

We verified CA delivery by comparing 9.4 T MRI data acquired before and after MR-CA infusion (scanning time = 0.5 h each); altogether, we examined 5 time intervals after delivery. We obtained additional serial T2*-weighted gradient echo images in live animals, and using the data from these scans, we generated R2* maps for quantitative analysis of SPION retention. Scans before SPION delivery served as baseline. Subtraction R2* (ΔR2*) maps (R2* values at final time point minus R2* values of baseline) allowed us to determine noise due to brain CA infusion.

Immunohistochemistry

After MR acquisition, the mice (n ≥ 2, each group) were placed under general anesthesia and retrograde perfused with ice-cold saline. We removed whole eyeballs from the animals and froze them in n-butanol on dry ice; frozen eye tissue sections (thickness, 20 μm) were prepared and stored at −70°C before staining. The eye samples were fixed in 4% paraformaldehyde (50 ml) in 0.1 M phosphate buffer (0.1 M Na₂HPO₄/NaH₂PO₄ pH 7.4) for 10 min, then washed in buffer. For specific uptake of IgY-labeled chimera, the samples were stained overnight with Cy5-labeled goat IgG anti-cIgY (ab6872) all at 1/1000 dilution at 4°C.

For Rho detection, the samples were stained overnight with Cy3-labeled antibodies against Rho (ab6434 or ab3424; Abcam), diluted to 1/1000, at 4°C. We stained the vascular endothelia with Cy3–Griffonia simplicifolia lectin I (1/100 dilution) and the nucleic acids with DAPI (1:500 dilution). All histologic images were acquired using the same exposure time and gain using a cellSens Dimension device (Olympus America Corp., Nashua, NH, USA).

Statistical analysis

We used power analysis to calculate the minimum number of animals required in each group as stated previously, or a minimum of n ≥ 2 viable mice in each comparison. We computed the mean and sem from the average values in each group of animals, then compared these values by 1-tailed Student’s t test.

RESULTS

Cellular protein-guided chimera MR-CA

The rationale for the chimera design was based on several important factors. First, we can compensate for the fact that living cells do not take up and retain immunoglobulins. Second, normal resting mouse brains take up sODN-Ran or SPION-Ran transiently and exclude it rapidly as soon it is delivered (Fig. 1, n = 3, P = 0.2 vs. baseline, Student's t test). The R2* values of SPION-Ran faded gradually but rapidly over time to a low intensity at 4 h. We hypothesized that a chimera of immunoglobulins against cellular protein and SPION-Ran would facilitate SPION-Ran retention. This chimera design, as depicted in Fig. 2A, does not include an immunoglobulin to membrane receptor for membrane receptor–assisted uptake; the purpose of modifying the design away from receptor-assisted design was to avoid the risk of associated noise elevation in the MR images.

Figure 1.

R2* elevation was transient after SPION-Ran. SPION-Ran and timeline for MRI (A) and R2* maps of SPION-Ran at 0.5, 2, 4, and 6 h after delivery (B); 4 slices of MRI from same mouse (n = 3) are shown.

Figure 2.

Design of complete chimera for cellular protein (A) and timeline for MRI (B). We delivered complete chimera (Ran-SPION-rIgP/cIgYxMAP2) to mice at low (C) and high (D) doses, along with control (unlinked mixture of SPION-rIgP/cIgYxMAP2 and biotin-free sODN-Ran) at high dose, E) (n = 3 each group). R2* maps are shown with scale bar to right. Circles represent ROI in striatum, where we obtained data for uptake/retention curve shown in Fig. 3A.

Before developing and applying a Rho-targeting chimera for the retina, we aimed to demonstrate the uptake, significant retention, and semiquantitation of chimera using neuron-specific MAP2 in the CNS. We longitudinally monitored the complete chimera (Ran-SPION-rIgP/cIgYxMAP2, intracerebroventricular injection, Fig. 2B) evaluating the retention of 2 doses (8 and 40 μg Fe/kg, i.c.v., Fig. 2C, D). We found that animals provided the higher dose exhibited significant R2* elevation, without apparent noise, in the striatum at 5 to 7 h after chimera delivery (Fig. 3A). Animals in the control group (which received either unlinked SPION-rIgP/cIgYxMAP2 or biotin-free sODN-Ran, n = 3) exhibited no significant uptake compared to the baseline R2* when measured at 7 h (Fig. 2E). To evaluate the retention of chimeric MR-CA in mouse brain, we aimed to find the location of cIgP, mostly cIgYxMAP2 in Ran-SPION-rIgP/cIgYxMAP2 or SPION-rIgP/cIgYxMAP2, as evidence of uptake using Cy5-gIgGxcIgY. We examined necropsy samples obtained after MRI acquisition and found cIgYxMAP2 present in the neuronal formation of the dentate gyrus and the cerebellum (Fig. 3B1). We observed no cIgYxMAP2 in the same ROI of control mice (Fig. 3B2).

Figure 3.

Constructed curve (A, data from Fig. 2D) represents uptake and retention of MR-CA in striatal ROI (circles in Fig. 2D). We validated presence or absence of cIgYxMAP2 in chimera and control using Cy5-gIgGxcIgY (shown as blue pseudocolor).

BCAO-60 induced ischemic retinopathy

We performed traditional ex vivo histology to examine for anatomic abnormalities at 1, 2, 3, and 8 wk after BCAO-60 (n > 2 each). When compared to the normal group (Fig. 4A, C), our histologic data showed no apparent retinal abnormality. Compared to normal or sham operation (control) mice, the retinas of BCAO-60 animals examined 8 wk after surgery (Fig. 4B) exhibited no visible deterioration in the photoreceptor and retinal pigment epithelium (RPE) layers or in the ganglion cell layer (GCL). However, we did note possible thickening of the retina after BCAO, likely brought on by an outgrowth of cells from the outer nuclear layer (ONL) to the inner segment (IS) of the photoreceptor cells (arrows, Fig. 4D). To determine the feasibility of imaging the mouse retina using an MR-CA and MCE-MRI in vivo and to examine whether BCAO-60 induces retinopathy, we measured the thickness of the retina in living mice using MnCl₂ (a T1 MR-CA) and T1-weighted MRI at 9.4 T using a previously reported protocol (12–14). Although MnCl₂ is not a target-specific CA, the strong T1-weighted signal seen in Fig. 5A signifies neuronal retention of MnCl₂ in the retina (dashed arrow) and in the brain (solid arrow). Measured in vivo, the mouse retina is 0.20 to 0.25 mm thick. The 0.12 mm per pixel resolution of the R2* values acquired using this 9.4 T MR system is thus suitable for detecting changes in the retina.

Figure 4.

Ex vivo histology of retina before and after BCAO shows normal group (A, C) and BCAO group (B, D) exhibiting minimal anatomic abnormality in RPE and GCL when examined under phase contrast (A, B). Immunohistologic staining (C, D) shows nuclei (blue, DAPI) at ONL and inner nuclear layer (INL) as well as Rho⁺ cells (Cy2-IgG-Rho; ab3424). Rho⁺ signal is located between RPE and ONL, where photoreceptors are located without overlap with RPE, ONL, or INL; no visible difference was evident (normal mice n = 3; BCAO n = 2, scale bars, 100 μm).

Figure 5.

We measured thickness of mouse retina in vivo at 0.20–0.25 mm (A, broken arrow) using T1-weighted (T1w) MRI and MnCl₂. We demonstrated retinopathy using T1w MRI of retina using Gd-NA-nestin within 1 wk after BCAO-60 in mouse model of ocular ischemia (B). MR-CA (Gd-NA-nestin, 2 μg, eyedrop to left eye) contains sODN antisense to nestin mRNA expressed in pericytes of NVU. Signal remained visible 41 h after delivery. We detected elevation of GFAP mRNA expression or gliogenesis by SPION-gfap (8 mg/kg, i.p.) in retina (broken arrows) and optic track (solid arrow) 6 mo later (C; same mouse as in B), but not in control group (D, intraperitoneal/intracerebroventricular delivery). Numbers in scale show R2* values increase with brightness in R2* maps.

BCAO-60 is known to stimulate the growth of new vasculature (neovascularization) by recruiting pericytes (which express nestin) and expansion of GFAP-expressing glial cells active in gliogenesis in the CNS (2, 8). To determine whether BCAO-60 similarly recruits progenitor cells to the retina as a sign of ischemic retinopathy, we examined the retinal NVU in mice with and without BCAO-60 (n = 2 each) using a nestin-targeting MR-CA to identify the presence of pericytes and GFAP-targeting MR-CA to detect gliogenesis. We used gadolinium-diethylenetriamine penta-acetic acid (Gd-DTPA; a T1 MR-CA) as an alternative MR-CA for MCE-MRI. We also modified Gd-DTPA with human serum albumin (hSAB) (BioPal) to NA-hSAB-Gd and conjugated to it biotin-sODN-nestin, yielding Gd-nestin, which targets nestin mRNA.

We delivered Gd-nestin to one eye (the other eye of the same mouse was examined as the control) of a mouse that had undergone BCAO-60 (n = 2). Figure 5B shows that the T1-weighted MR-CA reaches the ipsilateral retina and brain within 3 h, with a smaller amount of the CA also reaching the contralateral retina. Gd-nestin was still detectable in the retina of both eyes at 41 h, thus allowing us to conclude that Gd-nestin was retained in the retina. Although this T1 agent satisfactorily allowed MCE-MRI of the retina, retention of the CA beyond 36 h does introduce a potential risk of intracellular toxicity. We thus opted to investigate biocompatible Fe as an alternative CA.

Six months after BCAO-60, we identified gliogenesis in the retina using SPION-gfap (Fig. 5C) and validated our findings with ex vivo histology an additional 2 mo later—that is, 8 mo after BCAO-60. In the controls, results were below the detectable range (Fig. 5D and Fig. 6A). The ex vivo examination using anti-GFAP antibodies (Fig. 6B) validated the GFAP elevation observed in the same mouse (Fig. 5B, C). Together, these data demonstrate that BCAO induces proliferative retinopathy as ocular ischemia (15, 16); furthermore, these findings show that our method provides specificity for identifying pericytes and retinal astroglia using Gd-nestin and SPION-gfap, respectively, and MCE-MRI can be used for longitudinal study of retinopathy in living mice. To reduce possible binding of RNA-targeting MR-CA to untranslated RNA in non–photoreceptor cells within the retina, we decided to apply chimera MR-CA for cellular proteins. The chimera we designed was anti-Rho-SPION-Ran with antibodies to Rho and SPION-Ran (Fig. 2A).

Figure 6.

Ex vivo histology of Müller (gfap⁺) cells before (A) and 8 mo (B) after BCAO show pronounced gliosis in GCL and slight but visible gliosis of layer between RPE and ONL after BCAO-60 using IgG against GFAP (Cy3-ab4674) in same mouse (Fig. 5B, C). Normal retina (with light path, arrow, C) represents relative location of different layers of retina in A and B.

Given that the composition of NVU is the same in the CNS and the retina (1) and that our previous studies demonstrated that BCAO compromises the BRB (7), we delivered chimeric anti-Rho-SPION-Ran (a T2 MR-CA) through an eyedrop solution administered to the left eye (2 μg Fe in 1.5 μl) of C57black6 mice (n = 3) 1 d after BCAO (2). We acquired MR images of the eyes at 6 time points (each mouse received sequential MRI twice, with an interval of 1.5 h awakeness) and analyzed the R2* values using the right eye of the same mouse as the negative control. We repeated the acquisition and analysis in another 3 mice. Figure 7A shows that the ΔR2* values (average of 2 mice) were highly elevated at 3.5 h but returned toward baseline 4 h after eyedrop delivery. At peak uptake time, we examined the uptake of iron oxide in anti-Rho-SPION-Ran as electron-dense nanoparticles (EDNs) in the retina and optic nerve. Figure 7B shows free-floating EDNs in the capillary of the optic nerve track near the retina; at high magnification (Fig. 7B1), we observed EDNs in the bloodstream (arrows) near a red blood cell (R); although the EDNs entered primarily through the endothelia (E), some EDNs also entered peripheral cells (P, or pericytes identified by location). The average diameter of these EDNs was 8 nm. These results suggest that the MR-CA distributes to the vasculature by crossing the BRB, thus providing evidence that anti-Rho-SPION-Ran is robustly transported, taken up, and distributed, with little aggregation, when delivered through eyedrops (17).

Figure 7.

Pharmacodynamics of anti-Rho-SPION-Ran uptake (A) show profile of ΔR2* elevation between left (L, ipsilateral) and right (R, contralateral control) retina of same mice (n = 4). Both retinas exhibited same R2* values during first 90 min after delivery; R2* increased in ipsilateral retina (L minus R) 3.5 h after eyedrop administration. ΔR2* decreased at 4 h. These data show uptake and exclusion by retina. (B) Lead-free TEM shows that EDNs are stable in capillary of nerve track (arrow); enlarged image in box shows EDNs pass endothelia (E) and BRB after intraperitoneal injection to mice fit with an intracerebroventricular port. One of 2 EDNs is shown outside capillary in peripheral cells (P). Currently this pathway occurs in mice after BCAO.

To determine whether anti-Rho-SPION-Ran targets Rho antigen, we examined rods and cones of photoreceptors for evidence of EDNs. We delivered the same dose of anti-Rho-SPION-Ran via eyedrop to mice with BCAO-60 of 1 or 20 d (n ≥ 2 each); the following day, we prepared retina samples for lead-free transmission electron microscopic (TEM) examination. Figure 8A shows dark EDNs as Rho stain in the discs of cones and rods 1 d after BCAO-60; no such staining was visible in either the IS or the cilium. In mice examined 20 d after BCAO, we found that some cone and rod of photoreceptor cells were entirely void of EDN stain; the lower-than-normal Rho antigen levels were similar to those levels seen in the IS, cilium, and mitochondria, where Rho antigen is expected to be null (Fig. 8B).

Figure 8.

Anti-Rho-SPION-Ran detects Rho⁺ cell degradation as early as 20 d after BCAO (B); no apparent Rho⁺ cell degradation was evident 1 d after BCAO (A). TEM shows uniform Rho⁺ staining in retina (A); IS, cilium (*), and mitochondria (#) were Rho negative, confirming that Rho-SPION-Ran is specific to Rho in discs of photoreceptor cells. (B) Null stain in some of rod and cone cells 20 d after BCAO.

MRI of Rho⁺ neurons in vivo

We aimed to demonstrate that anti-Rho-SPION-Ran identifies neurons that express Rho in vivo. We prepared C57black6 mice for BRB bypass by introducing an intracerebroventricular port, inserting a 28-gauge needle through the dura and cortex to the ventricle, and immediately removing it and sealing the skull with bone wax. We obtained background MR data for all mice (stratification) 1 wk later, before chimera MR-CA delivery. We eliminated from future analysis any mice in which background R2* values were not similar to the normal baseline. We then delivered saline (100 µl, control) or anti-Rho-SPION-Ran (8 mg/kg, i.p.) to 2 groups of mice; a third group received anti-Rho-SPION-Ran (8 mg/kg, i.p.) without intracerebroventricular port. We acquired MRI 16 h later; the profile in Fig. 7A provided the basis for determining the timing for image acquisition. Figure 9A shows retinal R2* values, or the frequency of signal drop, in the group with anti-Rho-SPION-Ran; the elevation was significantly greater than in the no-targeting-agent group (P < 0.01, Student's t test, 1 tailed). The difference in R2* was 17 values per second. Lead-free TEM magnification of necropsy samples taken from the eye immediately after MRI shows few visible discs, mostly without EDNs of Rho stain, in the control group (Fig. 9B, C). We observed EDNs in Rho⁺ cells from mice that received the chimera with BRB bypass (Fig. 9D, D1). The delay of 16 h between CA delivery and image acquisition appears to allow an exclusion of excess and nonspecifically bound MR-CA after anti-Rho-SPION-Ran has reached its target.

Figure 9.

Evidence of delivery of anti-Rho-SPION-Ran to Rho⁺ cells. We delivered anti-Rho-SPION-Ran (8 mg/kg, i.p.) or saline to mice with an intraventricular port and acquired MRI 24 h later. Rho⁺ signal in normal retina in mice with anti-Rho-SPION-Ran (A) was significantly higher than in control groups in vivo. Lead-free TEM (B–D) shows dark EDNs of anti-Rho-SPION-Ran in rod and cone of photoreceptor cells (D, D1); controls received saline (B) or chimera without blood–brain barrier bypass (C) and showed few EDNs.

Monitoring retinopathy using the BCAO model

To compare Rho⁺ cells in vivo with MCE-MRI, we delivered anti-Rho-SPION-Ran (8 mg Fe per kg, i.p.) to both normal mice (fit with an intracerebroventrical port) and BCAO-60 mice; this procedure was carried out 20 d after the surgical procedure. We acquired MRI 16 h after anti-Rho-SPION-Ran delivery, then performed necropsy for lead-free TEM. In this case, we did not use the eyedrop delivery method because eyedrops do not reach the brain in mice with an intracerebroventrical port. Figure 10 shows that R2* measurements in the retina were significantly higher (P = 0.01, Student's t test, 1 tailed) in normal mice than in the BCAO-60 mice. TEM revealed Rho⁺ EDNs in the outer segment of the photoreceptor cells in both normal and BCAO mice, but the number of Rho⁺ EDNs was greater in normal mice than in BCAO mice. These data provide evidence that the retina undergoes apoptotic degeneration after ocular ischemia (18–20). These data also show that MCE-MRI with anti Rho-SPION-Ran is sensitive to changes in Rho⁺-specific cells. The chimera design promises delivery of antibodies to cellular proteins, with the potential for quantitative measurement of functional photoreceptors in vivo.

Figure 10.

We compared Rho⁺ cells in 2 groups of C57black6 mice (normal and BCAO-induced ocular ischemia) using anti-Rho-SPION-Ran and MRI. MRI data showed significantly higher levels of Rho⁺ signal in normal retinas compared to retinas of BCAO mice (A). Ex vivo validation shows dark stain as EDNs in rods and cones of photoreceptor cells (B), with fewer EDNs in BCAO retina (C).

DISCUSSION

Monitoring retinal gene expression in vivo, considering the very small size of the targets of interest, represents a significant challenge. We have met this challenge by developing a novel Rho-targeting chimera (anti-Rho-SPION-Ran) and tracking retinopathy in vivo in a preclinical mouse model of global cerebral ischemia by MCE-MRI. Our method has advantage of early detection without biopsy and monitoring several biomarkers over time. These conclusions are based on our observation that anti-Rho-SPION-Ran binds to Rho and thus can detect changes in the level of cellular Rho antigen in C57black6 mice. Because it contains iron oxide, anti-Rho-SPION-Ran was visible in ex vivo lead-free TEM as EDNs in the discs of cone and rod of photoreceptor cells. Importantly, this combination method allows detection at an earlier time point than fluorescent immunohistochemistry measurement. Moreover, we have validated the specificity of this Rho-targeting modality using a new model of ischemic retinopathy, in which we observed the appearance of pericytes, proliferative degeneration of Rho, and active gliosis after BCAO-60. The chimera MR-CA we have designed to target cellular antigens is different from receptor-bound CAs, as receptor-based delivery requires that the CAs bind to the cellular membrane to enter into the cell. When used to target cellular proteins, such receptor-bound CAs may be associated with noise of unknown origins.

We improved our MR-CA design by combining sODN-Ran and antibodies. This chimera design facilitates delivery (sODN-Ran) and specific retention (IgG to Rho) while avoiding cells that may contain untranslated target mRNA, thus enhancing specificity for the target cell. We have demonstrated that uptake and retention of the chimera relies on the presence of sODN with Ran sequence (∼9 ± 3 kDa). The retention of SPION-Ran is transient but can be extended in time when SPION-Ran is conjugated with a IgG of cellular proteins (Figs. 1–3). The small sODN-Ran readily carries large nanoparticles (30 ± 20 nm) of SPION and as many as two IgG (>100 kDa) and can pass across the cellular membrane (Fig. 8B1). Retention depends on the presence of intracellular antigen and less noise due to receptor binding. Our observation of the absence of EDNs in the cilium, mitochondria, and the outer membrane of cone and rod of photoreceptor cells provides further support for the superior specificity of our chimera MR-CA.

We have carried out several successful studies demonstrating eyedrop or intraperitoneal delivery of target-guided MR-CA to the CNS of living mice and have validated the effectiveness of this method for examining neuronal death (21), gliogenesis (8), and angiogenesis (2, 22). Here we further demonstrate that BCAO causes global damage to the retina and that our anti-Rho-SPION-Ran CA can reliably detect changes in Rho antigen. Under normal conditions, when the BRB is healthy and intact, intravitreal injection is one possible route of delivery to the retina (17). As we previously demonstrated, direct delivery of MR-CA to the brain is possible via the blood–brain barrier bypass using an intracerebroventricular port (4, 23), a procedure similar to intravitreal injection, except the insertion is at some distance away. We propose that this combined method may also be used to bypass the BRB to deliver genes to normal retina. As the data presented here demonstrate intraperitoneal injection allows delivery of chimera MR-CA to the normal retina by bypassing the BRB (Fig. 9).

The outer retina consists of photoreceptors and Müller glia, which are metabolically coupled to support the generation of electrochemical impulses in response to light stimulation. Nutrients and oxygen diffuse from the choroidal vessels through the RPE layer, which forms the tight intercellular junctions known as the BRB. We chose to focus our studies on Rho antigen, also commonly known as visual purple, because it is a specific light-sensitive photoreceptor protein belonging to the G-protein-coupled cascade. Rho is made up of 2 components: opsin protein and the cofactor retinal embedded in the lipid bilayer (discs) of the retina. Retinal, a photoreactive chromophore, is produced from vitamin A or dietary β-carotene in the photoreceptor cells of the retina. This biologic pigment is responsible for the first events in the perception of light; photoisomerization of 11-cis-retinal into all-trans-retinal by light induces a conformational change (photobleaching) in opsin, which interacts with the G protein transducin. The Rho molecule is used only once in the vertebrate visual system. After photobleaching in response to light exposure, it takes about 45 min to fully regenerate Rho in humans. A degenerating RPE layer after BCAO-60 may delay the active turnover of Rho. In addition, the methods we have developed and describe here may have important applications for global tracking of microbial opsin used in optogenetics. Importantly, this approach allows longitudinal monitoring of opsin-related protein expression without the use of invasive tissue biopsy. This design applies to therapy if SPION-Ran carries antibodies with therapeutic properties.

Retinopathy is a common microvascular complication of many conditions, including diabetes (24, 25), hypertension (26), retinal vessel occlusion (stroke or ischemia) (27, 28), premature birth, radiation exposure, excessive sprouting of new blood vessels (angiogenesis), or genetic abnormality (retinal ciliopathy). Retinopathy can lead to severe vision loss or impairment and blindness. Anatomic and functional MRI detects microangiography based on changes in the oxygenation of blood flow of the retina in vivo (29–32), and ophthalmologists routinely use optical coherent tomography to diagnose retinopathy based on changes in backscattered, coherent infrared light (33). The ability to quantify gene expression and to identify changes in the composition of cells would complement current clinical modalities, thus advancing our understanding of retinal degeneration and supporting the development and evaluation of new therapies. Evaluation of the progressive pathogenesis of retinopathy would also be substantially improved by quantitative measurement of regenerative cells using current clinical platforms.

In our retinopathy model, the membranes of rod and cone photoreceptors appeared to be intact. In order to understand the pathway of photoreceptor degeneration, we will develop a new chimera to test for antigens that may be associated with apoptosis. The chimera MR-CA we developed to target cellular antigens allowed semiquantitative monitoring of cellular protein. We show here a reduction in Rho antigen 8 wk after ocular ischemia induced by BCAO-60. We first demonstrated the specificity of Rho monitoring using a chimera of anti-Rho IgG in SPION, which binds to Rho antigen in the discs of the rod and cone photoreceptors in normal mice. The iron oxide in SPION is detectable as elevated R2* values in MRI and EDNs in lead-free TEM (Fig. 9D1). Because there is no receptor-mediated targeting, the R2* map avoids noise from unspecific binding.

Moreover, we have delivered SPION-sODN through eyedrops to living mouse brains after BCAO-60-induced global ischemia in the CNS of C57black6 mice. Because the ophthalmic arteries are branches of the carotid arteries, and because the posterior communicating arteries are defective in C57black6 mice, we proposed that the BCAO procedure would also induce retinal damage by ocular ischemia and in turn trigger degeneration of the BRB such that target-guided MR-CA could be delivered through eyedrops (34). Using an ocular ischemia syndrome model in normal C57black6 mice, we have demonstrated the translation of our cell imaging approach to examine the retina in vivo. Imaging retinal cells by MRI represents a highly innovative and transformative advance that has dramatic potential to enhance our understanding of the damage and repair processes in retinopathy that occurs as a result of ocular ischemia, diabetes, and other disorders, and may therefore increase our ability to develop and evaluate effective treatment options. This approach thus represents molecular, anatomic, and functional advances for quantitative measurement of retinopathy in the inaccessible retina.

Last, MCE-MRI provides superior sensitivity and depth of penetration for functional imaging of the retina in addition to MR-CA for in vivo imaging of gene expression in the brain (30, 35). We have demonstrated the robustness of this approach for detecting specialized cells associated with the progression of disease in the brain (as evident through measurable oxidative stress, inflammatory responses), neuronal death, and/or recovery (e.g., through processes of gliogenesis, angiogenesis, and neurogenesis) with sufficient signal specificity and sensitivity, as well as excellent resolution of 100 to 120 μm at 9.4 T (5). This MR-based imaging approach should quantitatively evaluate and identify protein in the retina of living mouse. Our results support a proof of concept for monitoring early degeneration of photoreceptors, although the exact route of exclusion from the CNS is not yet totally understood. We anticipate that new evidence will be obtained using the cerebral circulation model (36). In conclusion, our results support that MR-CA migrates across the BRB after ocular ischemia; that anti-Rho-SPION-Ran targets and allows detection of Rho⁺ photoreceptor cells; that MCE-MRI is sensitive to changes in Rho antigen; and that this modality should allow repeat longitudinal evaluation of retinal pathogenesis using mRNA- and protein-targeting CA, especially after gene therapy, negating the need for biopsy. Together, these results demonstrate a workable approach by which to expand the application of our technique to detect photoreceptor degeneration and gliogenesis in the optic track of the retina. The potential implications of this approach are significant and will enable noninvasive monitoring of disease progression and therapy efficacy.

Glossary

BCAO

bilateral carotid artery occlusion

BRB

blood–retina barrier

CA

contrast agent

EDN

electron-dense nanoparticle

GCL

ganglion cell layer

gfap

glial fibrillary acidic protein

hSAB

human serum albumin

i.c.v.

intracerebroventricular

IgP

immunoglobulin polyclonal

IS

inner segment

MAP2

microtubule-associated protein 2

MCE-MRI

molecular contrast-enhanced MRI

MR

magnetic resonance

NA

NeutrAvidin

NVU

neurovascular unit

ONL

outer nuclear layer

Ran

random

Rho

rhodopsin

ROI

region of interest

RPE

retinal pigment epithelium

sODN

phosphorylthioate-modified oligoDNA

SPION

superparamagnetic iron oxide nanoparticle

TEM

transmission electron microscopy