Electrophysiological and Histological Characterization of Rod-Cone Retinal Degeneration and Microglia Activation in a Mouse Model of Mucopolysaccharidosis Type IIIB

Sanfilippo syndrome Type B or Mucopolysaccharidosis IIIB (MPS IIIB) is a neurodegenerative autosomal recessive lysosomal storage disorder in which patients suffer severe vision loss from associated retinopathy. Here we sought to study the underlying retinal functional and morphological changes associated with MPS IIIB disease progression using the established model of MPS IIIB, the B6.129S6-Naglu(tm1Efn)/J mouse line. Electroretinogram (ERG) was recorded from MPS IIIB and wild-type (WT) mice at the age of 28 and 46 weeks, and retinal tissues were subsequently collected for immunohistochemistry analysis. At the 28th week, rod a- and b-wave amplitudes were significantly diminished in MPS IIIB compared to WT mice. The cone a- and b-waves of MPS IIIB mice were not significantly different from those of the control at the 28th week but were significantly diminished at the 46th week, when MPS IIIB mice showed a major loss of rods and rod bipolar cells in both central and peripheral regions and a minor loss of cones in the periphery. Activation of microglia and neovascularization were also detected in the MPS IIIB retina. The new findings that cones and rod bipolar cells also undergo degeneration, and that retinal microglia are activated, will inform future development of therapeutic strategies.

Scientific RepoRts | 5:17143 | DOI: 10.1038/srep17143 (0.77 mg/ml). Both pupils were dilated with a drop each of 1% tropicamide and 2.5% phenylephrine and the corneas anesthetized with 0.5% proparacaine hydrochloride. Mice were placed inside a Ganzfeld dome coated with highly reflective white paint (Munsell Paint, New Windsor, NY) on a heating pad maintained at 39 °C. A drop of 2.5% methylcellulose gel (Goniosoft, Ocusoft Inc, TX) was applied to the eye and a blunt platinum rod electrode placed in contact with the center of each cornea. Ground and reference platinum subdermal electrodes were gently inserted in the tail and the forehead, respectively. The mice were then allowed to remain in complete darkness for 5 minutes prior to initiating the experiment.
Half millisecond square flashes for scotopic a-wave and b-wave measurements were produced by cyan light emitting diodes of 503 nm peak wavelength, calibrated with a photometer (ILT1700 International Light, MA) and converted to the unit photoisomerizations/rod (R*/rod) 31,32 , where 1 scot cd m^2 = 581 photoisomerizations/rod/s. At the lowest intensity, 25 responses were averaged with a delay of 2 seconds between each flash. As the intensity of the flash increased, fewer responses were averaged with a longer delay between flashes. At the end of the scotopic protocol, a pair of 1500W xenon lamps (Novatron, Dallas, TX) attenuated with apertures and diffusers were used to produce two saturating light stimuli. Since rods are more sensitive than cones, the rod ERG was measured by using single flash stimuli with a strength below the operative range of cones, while the cone ERG was measured by a paired-flash protocol using the xenon flashes 33 . In the paired flashes protocol, an initial conditioning flash (4.6 × 10 6 R* per M-cone) saturates both rods and cones 2 sec before a probe flash. The ERG recorded by the probe flash (1.8 × 10 6 R* per M-cone) is attributed to responses driven by the cones because cones recover faster than rods.
Electrical signals were amplified with a Grass P122 amplifier (Grass Instruments, West Warwick, RI) and band-pass filtered from 0.1 to 1,000 Hz. Data was digitized with a National Instruments data acquisition unit (USB-6216, National Instruments, TX) at a 10 kHz sampling rate, and analyzed with custom Matlab code (MathWorks, Natick, MA). To remove oscillatory potentials before fitting, the scotopic b-wave was digitally filtered using the filtfilt function in Matlab (low-pass filter; Fc = 60 Hz). The a-wave was measured from baseline to trough of the initial negative deflection (unfiltered) and the b-wave was measured from the a-wave trough to the peak of the subsequent positive deflection (filtered). Stimulus strengths are displayed in log units where 1E + 0 = 1 photoisomerization/rod. Two tails statistical analyses were performed using unpaired t-test or Welch's t-test in SPSS Version 20 (IBM). Alpha level is set at 0.05.
Retinal Immunohistochemistry (IHC). Eyes were enucleated under deep anesthesia, then mice were immediately euthanized by an overdose of anesthesia. The eyes were marked for orientations and carefully dissected to isolate whole retina, which was then incubated in 4% paraformaldehyde (Electron Microscopy Science, Fort Washington, PA) in phosphate buffer (DPBS, pH7.4, Invitrogen, La Jolla, CA) at room temperature for 45 minutes for fixation.
For the quantification of activated microglia in the inflammation assay, whole-mount retina was tile-scanned with a 10× objective. Further to quantify the percentage occupancy of two layers of Iba-1 positive cells separately, four regions of each flat retina were imaged using 32 um thick z-stack with 1 um interval (33 slices) with a 40× objective. Then images were quantified using imageJ (33 slices/region; 4 region/retina; 2 retina/animal and 3 animals/genotype). Average occupying Iba-1 area (%) was separately calculated in each region (GCL-IPL and OPL). For the cell profiling study, whole retina was further processed into retinal vertical sections. To control for variability from regional retinal difference, the dissection and handling procedures were designed to ensure that the same regions on the retina were being assessed consistently among samples (Fig. 1). The whole retina was first hemisected, then the temporal half of the retina was cut using a microtome (Vibratome; Leica Microsystems, Bannockburn, IL) into 40 μ m vertical sections. The samples were then blocked with 10% donkey serum (Jackson Immunoresearch, West Grove, PA) in TBS (DPBS with 0.5% Triton X-100 (Sigma) and 0.1% sodium azide (Sigma), pH 7.2) at 4 °C overnight to reduce nonspecific labeling.
An indirect antibody method was used for the IHC. The whole-mount retina or free-floating sections were incubated with primary antibodies for 4 days at 4 °C in TBS in the presence of 3% donkey serum. Controls lacking primary antibodies were processed as well. After rinsing several times, the samples were transferred and incubated in 3% normal donkey serum-TBS solution containing donkey-host secondary antibodies conjugated with Cy3 (1:200, Jackson Immunoresearch) or Alexa Fluor 488 (1:200, Molecular Probes, Eugene, OR) at 4 °C overnight. A fluorescent nuclear dye, TO-PRO3 (Molecular Probe, Eugene, OR, Cat. No. T3605; dilution 1:3000) 34 , was used to stain the nuclei in the. Rod bipolar cells were immuno-labeled with mouse antibody against PKCα (BD Transduction Lab, San Jose, CA, Cat. No. 610107; dilution 1:250) 35 . Cone cell bodies were immuno-labeled with rabbit antibody against GNAT2 (Santa Cruz Biotechnology, Santa Cruz, CA, Cat. No. SC-390; dilution 1:100) 36,37 . Activated microglial cells were immune-labeled with rabbit antibody against Iba-1 (Wako Chemicals, Richmond, VA, Cat. No. 019-19741; dilution 1:500) 38,39 . Astrocytes and Müller cells end-feet were immune-labeled with chicken anti-GFAP (Millipore, MA, Cat. No. AB5541; dilution 1:500) 39 . Alexa-488 conjugated Isolectin B4 from Griffonia simpliciflia (GS-IB4) was used to stain the retinal vasculature (Invitrogen, Grand Island, NY, Cat. No. I121411; dilution 1:200) 40 . After rinsing several times, the samples were mounted with Vectashield medium (Vector Laboratories, Burlingame, CA), coverslipped and analyzed with confocal laser scanning microscopes (LSM 510/LSM700; Zeiss, Thornwood, NY). Images of the vertical sections were acquired with Zeiss LSM software using 40× and 63× oil-immersion objectives. Two designated regions on the sections, representing central and peripheral retina, were photographed. Adobe Photoshop CS5 (Adobe Systems, San Jose, CA) was used to apply uniform brightness and contrast adjustments and to crop images. To measure the number of cone photoreceptors (GNAT2 positive cells) and rod bipolar cells (PKCα positive cells), cells with positive staining were counted in a 200 μ m segment, using the ruler and cell counter in ImageJ. The dimensions of the outer nuclear layer and the inner nuclear layer were also measured using ImageJ. Images of the whole mounted retina (Iba-1 staining) were acquired using the tile-scan feature of the Zeiss Zen software with a 10× objective. Iba-1-positive cell density and percentage area were analyzed using ImageJ.

Electroretinography (ERG).
Function of the photoreceptors─a-wave. ERG analysis of MPS IIIB mice at 28 and 46 weeks of age showed decreased a-wave amplitudes at both time points compared to age-matched WT mice (Figs 2a and 3a). The differences between groups were significant at the 4 highest stimulus levels ( Fig. 3a), indicating that rod function was adversely affected in the MPS IIIB mice. Furthermore, a-wave amplitudes in the MPS IIIB group decreased with time, while the same parameters in the WT group remained relatively stable over time. A-wave responses from the two brightest stimuli were fitted using a formula devised by Cideciyan and Jacobson 41 to determine the maximum response amplitude (R max ) for each group (detailed method is described in our previous paper 42 ). As shown in Table 1, there were significant differences in the a-wave R max values between the two groups recorded at both time points. MPS IIIB mice had mean R max values reduced by 60.5% and 86.2% in the 28 th and 46 th weeks, respectively.
Isolated cone a-waves measured using the paired flash method revealed cone dysfunction at a later time point (Fig. 3b). The difference in amplitudes of the cone a-wave between groups was significant in the 46 th week (WT: − 37.5 ± 2.9 μ V vs. MPS IIIB: − 24.7 ± 1.9 μ V, unpaired t-test, two-tailed p = 0.0008) but was not significant in the 28 th week. The finding that rod a-wave but not cone a-wave was suppressed in the 28 th week suggests that rods become dysfunctional earlier than the cones in this mouse model.
Function of the depolarizing bipolar cells─b-wave. Decline of b-wave amplitudes in MPS IIIB mice shared a similar time course with that of the a-wave. Figure 4a shows the stimulus-response curve for the ERG b-wave in the 28 th and the 46 th weeks, fitted with the Naka-Rushton equation to determine the rod-driven b-wave maximum response amplitude (B max ). In the 28 th week, MPS IIIB mice had a mean B max value 22.7% lower than that of the WT mice (Welch t-test, two tailed p = 0.0142). This difference increased to 50.3% in the 46 th week (unpaired t-test, two tailed p < 0.0001). Using a paired-flash protocol, we found that the isolated cone b-wave in the MPS IIIB mice was not affected in the 28 th week but was significantly reduced in the 46 th week (WT: 241.0 ± 8.8 μ V vs. MPS IIIB: 196.2 ± 15.6 μ V, Welch's t-test, p < 0.0053). Similar to the case of a-wave, a dysfunction in the b-wave was first detected in the rod pathway in the 28 th week. Function of the cone pathway appears to be preserved during the early phase of the disorder until a time point between 28-46 weeks. Sections containing the optic nerve head was processed with antibodies and florescent staining, and then mounted on slides. Being zoned into five zones of equal width, the second and the fourth zones from the optic nerve were respectively designated as the "central region" and "peripheral region" and were photographed by a confocal microscope for cell counting.
Scientific RepoRts | 5:17143 | DOI: 10.1038/srep17143 ERG b-to a-wave ratio. The above ERG data demonstrate progressive decline of amplitudes in both the photoreceptor-generated a-wave and the bipolar cell-generated b-wave. We sought to determine whether the b-wave was directly affected by this disorder or was secondarily affected due to decreased photoreceptor input. An indirect approach to answering the above question is to evaluate the b-wave/a-wave ratios. Compared to the WT, the MPS IIIB mice had generally higher B/A ratios ( Table 2). The difference was most obvious in the rod operative range, where the B/A ratio in the MPS IIIB mice was 46.2% higher in the 28 th week and increased to 73.5% in the 46 th week. The difference was smaller in the isolated cone-driven response, in which the B/A ratio in the MPS IIIB mice was 16.1% and 23.8% higher in the 28 th and 46 th week respectively. The data showed that, while both a-wave and b-wave were reduced in the MPS IIIB mice, the a-wave was suppressed by a higher percentage, leading to an increased B/A ratio in these mice. This indirect evidence supports the hypothesis that the primary site of neuronal dysfunction is represented by photoreceptors, particularly rods, although a loss of bipolar cell function is not ruled out.
Retinal Immunohistochemistry. Cell profiling-vertical section. Four mice from each genotype were sacrificed after ERG had been recorded in the 46 th week. One eye from each animal was microtome-sectioned and processed for immunohistochemistry analysis. Three retinal sections from each eye were imaged and had their cells counted and dimensions measured, resulting in a sample size equal to 12 for each group. As shown in the representative micrographs (Fig. 5), there were degenerative changes in both the outer nuclear layer (ONL) and inner nuclear layer (INL). In particular, the mean ONL thicknesses of MPS IIIB mice was 46.7% and 54.8% thinner than that of WT mice in the central and peripheral regions, respectively (Fig. 6a-c). In terms of photoreceptor somas in the ONL, the MPS IIIB mice had lost 48.1% and 42.6% in the central and peripheral regions, respectively. Since the TO-PRO3 dye stained both rod and cone somas, we utilized a cone-specific anti-GNAT2 antibody to determine if there was a differential loss in cones versus rods. The data showed that there was no significant difference between WT and MPS IIIB mice in the number of cone somas present in the ONL in the central region, and that MPS IIIB mice had 9.8% fewer cone somas in the peripheral region. Together, the large reduction in the number of photoreceptors and the minor reduction in the number of cones suggest that the thinning of ONL was primarily caused by a loss of rods in the MPS IIIB mice. We also found that the INL thickness of the MPS IIIB mice was 39.2% and 48.4% thinner compared to WT mice in the central and peripheral regions, respectively (Fig. 6d,e). In an attempt to identify the retinal cell type underlying the loss in the INL, we used antibodies for protein kinase C alpha (PKCα ), a marker for rod bipolar cells (BC R s). Our data showed that the BC R population had decreased by 60.2% and 54.8%, respectively, in the central and peripheral regions.
Inflammatory cell makers and neovascularization-flat-mount. Four WT and three MPS IIIB mice from a different cohort were sacrificed in the 46 th week. Whole-mount retinas from both eyes were processed for immunohistochemistry and cell counting, giving sample sizes of 8 and 6 for the WT and MPS IIIB groups, respectively. There were marked differences between MPS IIIB and WT mice in Iba1-positive cells morphologically (Fig. 7a-f). After examined the tile-scanned whole retina image, we found that the retinal area occupied by the Iba-1-positive cells was much higher in the MPS IIIB mice (Fig. 7g).
Under imageJ particle analysis, the cells occupied 13.3% and 5.4% of the retinal area in MPS IIIB and   WT mice, respectively (Welch-t test, two-tailed p = 0.026). Further analysis of sectioned MPS IIIB retina revealed that Iba-1 positive cells are located at the OPL and the ganglion cell layer-inner plexiform layer (GCL-IPL). Therefore the relative change of Iba-1 positive cells in the two layers were further analyzed separately, by processing z-stack images on representative regions of the flat-mounted retinas. The results showed that the percentage area occupied by Iba-1 positive cells were specifically higher at the GCL-IPL (Fig. 7h,i).
Retinal vasculature labelled with GS-IB4 identified the presence of neo-vascular tufts in MPSIIIB retina but not in WT retina (Fig. 8a,b).

Table 2. Ratios of b-wave/a-wave. Values for rod-driven ERG were calculated from scotopic ERG data measured using stimulus strength of 112.2 isomerizations/rod, which was below the threshold for cones.
Values for cone-driven ERG were calculated from paired-flash data, in which rod activity was effectively suppressed. There were no significant differences between WT and MPSIIIB retina in GFAP reactivity on reactive astrocytes and Müller cells (Fig. 8e-g). The morphology of astrocytes in MPSIIIB retina was similar to those in WT. No obvious loss of secondary cell processes was observed (Fig. 8c,d). Also, GFAP-reactive astrocyte/Müller cell processes seemed to be limited in GCL layer in both WT and MPSIIIB (Fig. 8g).

Discussion
In the present study, we have characterized the functional and morphological changes that occur in the retina of the Sanfilippo syndrome Type B (MPS IIIB) mouse model. Our findings are in agreement with, and significantly extend, a previous screening study of this model 28 . The decline of the ERG rod a-wave by 60% and 86% in the 28 th and the 46 th postnatal weeks, respectively, were most remarkable. In the 46 th week, 43-48% of photoreceptor somas were lost (rods to a greater extent than cones). Furthermore, we found that the cone pathway began to show signs of dysfunction late in the disease course. Both the ERG cone a-and b-waves were significantly smaller in the KOs. A roughly 10% reduction of cones photoreceptors was found in the peripheral retina of the KOs. This defect in the cone pathway was not detected in the previous study of this mouse model, probably because cone degeneration is not evident until a late stage of the disease (> 40 postnatal weeks). It was also surprising to find that rod bipolar cells (BC R s), the secondary neurons postsynaptic to the rods, had also degenerated and lost more than 50% of cells by the 46 th week. Analysis of the B/A wave ratio indicated that the a-wave declined to a greater extent than the b-wave, thus providing indirect support that the primary site of degeneration was the rods rather than the rod bipolar cells. Another novel finding is that the density of activated microglia in MPS IIIB mice was approximately triple compared with that in WT mice at the 46 th week time point, indicating that retinal inflammation might be involved in the degeneration. In addition, neo-vascularization was found to be present in the MPSIII retina as well.
The most prominent retinal change in the MPS IIIB mouse model was the dysfunction and degeneration of photoreceptors. While the exact mechanism remains elusive, it is likely to be linked to the lysosomal inclusion and accumulation of HS at the retinal pigment epithelium (RPE) as a result of enzymatic deficiency. The RPE is known to be crucial for sustaining the photoreceptors by serving a number of functions including the regulation of ion and metabolite transport 43 , retinoid transport 44 , and the phagocytosis of outer segments 45 . GAGs including HS are synthesized and secreted by the RPE under physiological conditions 46,47 . In MPS III patients, lysosome inclusion or lysosomal storage vacuoles are present in large numbers in the RPE 23,24 . Similar melanosome-like structures were also found in the RPE of 30-week-old MPS IIIB mice 28 . Though there is no specific experimental evidence for the role of RPE in photoreceptor degeneration in MPS III, previous studies using cultured MPS VII RPE cells have shown that the accumulation of GAGs in RPE cells was associated with the deficiency of the corresponding enzyme (β -glucuronidase) for their degradation 48 , and that genetically restoring the enzymatic activity of RPE cells using retrovirus reduced accumulation of GAGs in vitro 48 . Furthermore, supplementing the missing β-glucuronidase activity in RPE by transferring the correct gene to retinal cells through intravitreally injected adeno-associated virus improved the ERG in vivo 49 .
A noteworthy feature of the MPS IIIB mouse model is that rod degeneration was evident much earlier than that of the cones. This finding is consistent with the typical clinical signs and symptoms of MPS III. These patients often report their first visual symptom as night blindness, while some of them have near normal visual acuity under photopic conditions 25 . Their rod-driven ERG is also reduced to a greater extent than the cone-driven ERG 50 . A possible explanation for the differential impact of the disease on rods versus cones could be that rods are more sensitive to the disrupted supportive function of RPE induced by HS. It is also possible that a bystander effect from rod degeneration may contribute to the degeneration of cones. Numerous studies have shown that degeneration of rods induced by rod-specific genetic defects may spread to neighboring cones through an unclear mechanism 51,52 .
The above rod-cone sequence of loss in MPS III is common among patients with MPS I and MPS II as well 25 , but different trends were found in a MPS VII animal model 53,54 . One possible explanation for this discrepancy could be difference in the composition of the primary accumulated GAGs in the different MPS types. MPS VII, or Sly syndrome, is known to have chondroitin sulfate as an undegraded GAG in addition to HS 53 . It is yet to be determined whether chondroitin sulfate is particularly deleterious for cones.
Our immunohistochemistry finding that rod bipolar cells (BC R s) had undergone significant degeneration was unexpected because it is commonly believed that MPS III-associated retinopathy originates in (a-d) were cone somas (Some of those cells are identified by arrows in panel (a) as examples). There was no obvious change in the number of cone somas in the MPS IIIB retina. The green channel was enhanced to help visualizing the cones, leading to some non-specific reactivity in the inner nuclear layer (INL). (e-h) Rod bipolar cells were immuno-reactive to antibody for PKCα . In panel E, white arrows indicate the soma and yellow arrows indicate axonal terminals of rod bipolar cells as examples. Compared to samples from WT mice, MPSIIIB samples showed a general loss of rod bipolar cells in both regions. Green arrows denote non-specific labeling of blood vessels (BV) in G and H. Representative confocal micrographs comparing showing only nuclear staining (TO-PRO3) of WT and MPSIIIB mice at the 46 th week (i-l).
Scientific RepoRts | 5:17143 | DOI: 10.1038/srep17143 at the RPE 48,49,53 . However, this finding can be interpreted by taking into consideration that BC R s degeneration also takes place in aged patients of non-MPS retinitis pigmentosa (RP), as well as in mouse strains that have retinal degeneration caused by sole mutations in rods 55 . A cell loss of 20-60% in the inner nuclear layer has been observed in RP patients, whereas greater loss was found in extramacular region 56,57 . Reduced complexity of BC R s dendrites 58 , and a 20% loss of BC R s in period of 2 months were also observed in mouse RP models 55 . Thus, BC R degeneration appeared to be independent of the primary insult deriving from the underlying genetic defect and to rather be a result of the secondary effects of photoreceptor death. One hypothesis is that BC R requires a normal glutamatergic synaptic input or some unknown factors secreted from the rods to maintain its validity, and the deprivation of those factors causes the degeneration.
An alternative hypothesis specific for the BC R degeneration in MPSIII is that the activation of microglia and the subsequent inflammatory response exacerbate the death of BC R s. This notion is supported by our immunohistochemistry data, which showed obvious activation of retinal microglia. Additional support to this notion comes from the aberrant lysosome storage found in non-neuronal cells in the inner nuclear layer of MPS IIIB mouse retina 28 . Indeed, microglial over-activation was suggested to be involved in the pathogenesis of MPS III 8,9 , as well as in some other disorders of the central nervous system such as Alzheimer's disease and Parkinson's disease [59][60][61] . In the eye, microglial activation is associated with, and sometimes precedes, photoreceptor apoptosis and retinal degeneration 62,63 . Neo-vascularization previously reported in the heart valve and cartilage of MPSVII dogs 64,65 was also found on MPSIIIB retina. Its link with microglia activation is yet to be determined. In contrast, morphological changes in astrocytes and increased GFAP reactivity in Müller cells, previously reported to accompany microglial activation during retinal injuries (such as ocular hypertension and light damage 39,66 ), was not observed in our MPSIIIB model.
In the present study, we have performed detailed analyses of the retinal electrophysiological and morphological changes in a mouse model of MPS IIIB. The slow progression of retinal degeneration from normal to mild dysfunction and eventually massive cell loss provides a large time window and numerous measurable parameters for experimental manipulations. We have also shown that the ERG is a reliable, non-invasive and quantitative test to longitudinally monitor the progression of the retinopathy, and it has the potential to determine the effectiveness of future ocular/systemic therapy. Finally, the current interpretation of MPS IIIB retinopathy is based on a model in which rod degeneration following accumulation of GAGs at the RPE drives pathogenesis and degeneration. However, our findings that degeneration of rod bipolar cells and cones photoreceptor also takes place indicate that the RPE may not be the only primary site of the disease and imply that additional measure (e.g. anti-inflammation, neurotropic factors) may be needed to protect vision in MPSIII patients when treatments are being developed in the future. The fact that retinal microglia activation is involved in the degeneration implies that modulating retinal inflammation may be beneficial to delay the retinal degeneration, which warrants further investigation.