A comprehensive atlas of Aggrecan, Versican, Neurocan and Phosphacan expression across time in wildtype retina and in retinal degeneration

As photoreceptor cells die during retinal degeneration, the surrounding microenvironment undergoes significant changes that are increasingly recognized to play a prominent role in determining the efficacy of therapeutic interventions. Chondroitin Sulphate Proteoglycans (CSPGs) are a major component of the extracellular matrix that have been shown to inhibit neuronal regrowth and regeneration in the brain and spinal cord, but comparatively little is known about their expression in retinal degeneration. Here we provide a comprehensive atlas of the expression patterns of four individual CSPGs in three models of inherited retinal degeneration and wildtype mice. In wildtype mice, Aggrecan presented a biphasic expression, while Neurocan and Phosphacan expression declined dramatically with time and Versican expression remained broadly constant. In degeneration, Aggrecan expression increased markedly in Aipl1-/- and Pde6brd1/rd1, while Versican showed regional increases in the periphery of Rho-/- mice. Conversely, Neurocan and Phosphacan broadly decrease with time in all models. Our data reveal significant heterogeneity in the expression of individual CSPGs. Moreover, there are striking differences in the expression patterns of specific CSPGs in the diseased retina, compared with those reported following injury elsewhere in the CNS. Better understanding of the distinct distributions of individual CSPGs will contribute to creating more permissive microenvironments for neuro-regeneration and repair.

www.nature.com/scientificreports/ neurodegenerative diseases. Importantly, they have been shown to reduce the efficacy of various therapeutic approaches [4][5][6] . CSPGs consist of a core protein covalently attached to long glycosaminoglycan (GAG) carbohydrate chains 7 . There are 16 different types of CSPGs, of which Neurocan, Brevican, Phosphacan, Aggrecan and Versican are most common within the nervous system 7 . CSPGs exert much of their inhibitory function via their GAG chain composition 3 , however the core proteins themselves also inhibit neurite outgrowth 8 , an effect that persists after the removal of GAGs 9 . Understanding the specific changes in expression of individual CSPGs in response to injury or disease is crucial for creating extracellular environments that are permissive for regenerative therapies 4 . Aggrecan is one of the major CSPGs expressed in the nervous system and is a constituent of perineuronal nets (PNNs) 3,10,11 . It has been studied extensively in spinal cord injury where it prevents Schwan cell migration following transplantation by inhibiting integrin-mediated cell adhesion 12 , as well as inhibiting endogenous axonal growth in injured spinal cord in vivo 13 . Neurocan is similarly elevated around acute lesions in the CNS and active multiple sclerosis plaque edges [14][15][16][17] as well as in many experimental models of acute injury in the brain and is commonly associated with inhibition of neurite outgrowth 14,18,19 .
Little is known about CSPG expression in the retina, although there are indications of unique roles of individual CSPGs: In the rodent retina, Phosphacan protein levels increase from late embryonic to late postnatal stages but are markedly reduced in the adult retina 20 . Conversely, Aggrecan increases with age 21,22 . In the normal rat retina, Neurocan, Aggrecan and Versican are differentially distributed at embryonic, postnatal and adult stages 21 . With respect to the human eye, a recent study using human embryonic stem cell-derived retinal organoids indicated that BREVICAN and VERSICAN were found throughout the retina across much of development, although a downregulation of VCAN, and a concomitant upregulation of BCAN expression was observed in late developmental stages 23 . In the human adult eye, VERSICAN and AGGRECAN are found throughout the neural retina, choroid and sclera 22 .
Much less is known about CSPG expression in the diseased retina. Using the pan-CSPG marker CS-56, we have previously shown that CSPG deposition varies markedly both across degeneration and between different murine models of photoreceptor degeneration 24 . However, the changes in expression of specific CSPGs following different types ocular injury, and particularly inherited disease, are poorly characterised. Aggrecan was shown to be upregulated in rat models of retinal dystrophy 25 . Conversely, Aggrecan expression decreased, while Brevican and Phosphacan mRNA levels were unchanged in the neural retina following induced injury (ischaemia) in the rat retina, but all three were significantly increased in the optic nerve 26 .
To explore the potential heterogeneity in the microenvironment between different degenerative disease models, we examined four individual CSPGs, Neurocan, Aggrecan, Phosphacan and Versican. We present a comprehensive assessment of protein and mRNA expression both across time and with degeneration in three murine models of retinal degeneration of differing severity. Our findings revealed a marked heterogeneity in the expression of different CSPG core proteins and this can arise even between models with similar rates of photoreceptor loss. This data provides a valuable resource and reference for future studies but also shows the vital importance of understanding the specific microenvironment changes associated with a given disease type when assessing the suitability and timings of therapeutic interventions.

Methods
Animals. Mouse  London in individually ventilated cages and given access to nesting material and food and water ad libitum. They were kept on a standard 12/12 h light/ dark cycle and at the same light levels throughout the study and used at the ages specified in the text. Animals of both sexes were used in this study without discrimination.
Time points and degeneration models. We used the Aipl1 -/model of Lebers congenital amourosis, and the Pde6b rd1/rd1 and Rho -/models of Retinitis Pigmentosa, all of which have mutations in genes expressed by photoreceptors. We examined each model at 'early' (defined as ONL > 70% of the thickness of the wild-type retinae), 'mid' (ONL thickness of 30-70% of wildtype), and 'advanced' stage degeneration (ONL < 30%), as pre- Immunohistochemistry and histology. Animals were euthanized and a small burn was administered to the sclera overlying the orbit to provide a landmark for the superior retina. Eyes were enucleated and eye cups prepared in phosphate-buffered saline (PBS) and then carefully orientated and embedded in a standardized fashion in OCT (TissueTek) without fixation. Embedded eyecups were frozen and left at − 20 °C at least overnight before being cut as transverse sections, 18 μm thick. In order to avoid oblique cuts, all images shown are from the central most region of the eye, immediately adjacent to the optic nerve (Fig. 1f). Specific details for each antibody and protocols used for immunohistochemistry can be found in Supplementary Table S1 Supplementary Figs. S1-S4, S6, S7) retina at standardized regions in the equatorial region, and the superior retina in anterior margin (denoted as 'peripheral' in the images) immediately adjacent to the optic nerve (see schematics in figures). Images were acquired within 7 days of completing immunostaining and the same laser intensity, gain and offset settings were used across all sections for any given marker. Note that in order to allow direct visual comparison between wildtype data sets and those from the different degenerating models in different figures, images from wildtype retinae are duplicated in a number of figures. Schematics denoting regions imaged are adapted from our previous publication 24 .
RNA extraction and reverse transcription. Retinal   Data were analysed based on N > 3 individual retinae taken from different litters. Specifically, this included N = 4 for wildtype P14, all time-points in Rho -/for Ncan and Ptprz1, and Aipl1 -/-P10 for Ptprz1; N = 5 for wildtype P10 and 3 wks for Vcan; N = 7 for Aipl1 -/-P10 and P14 for Vcan; N = 8 for Aipl1 -/-3 weeks for Vcan; and N = 3 for all other measurements. Statistical tests were performed using Graph Pad Prism v9 software. Two-way ANOVA was used to assess the differences across time within models (red asterisk in the graphs), one-way ANOVA with Bonferroni's multiple comparison test for inter-group comparisons (black asterisk), and parametric t-tests (unpaired) for comparisons against age-matched wildtype controls, blue asterisk). Statistical significance is presented in the figures as follows: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns; non-significant.

Results
Changes in expression of four CSPGs commonly expressed in the CNS, Neurocan, Aggrecan, Phosphacan and Versican, were assessed across three different murine models of inherited blindness and in age-matched wildtype mice. In order to qualitatively assess the regional changes protein distribution in each model we used immunohistochemistry (IHC) and examined both the superior and inferior retina and equatorial regions, adjacent to the optic nerve (see schematics in each figure) in early, mid and advanced stages of degeneration. For quantitative analysis, real-time (RT-)qPCR was used to examine the levels of mRNA expression in whole neural retina, without RPE, over time. We first describe the patterns of expression of all four CSPGs in the normal wildtype mouse across time, from P10 through to 6 months of age, and then how these expression patterns are altered by retinal degeneration.

Aggrecan, Versican, Neurocan and Phosphacan show distinct changes in expression in later stages of maturation and across adulthood in wildtype retinae.
RT-qPCR revealed striking differences in the pattern of expression of Acan, Vcan, Ncan and Ptprz1 mRNA across the later stages of postnatal development and adulthood. These trends are summarised in Fig. 1e and Table 2. Specifically, Acan showed a biphasic pattern of expression in wildtype mice (Fig. 1a); expression increased significantly between P10 and 3 weeks of age (fold change, 2.35 ± 0.49; Table 2), which was followed by a significant decrease between 3 and 6 weeks. Expression levels remained consistently low thereafter, up to 6 months of age (latest timepoint examined). While comparisons of expression levels by IHC should be assessed cautiously, a similar trend was seen in the IHC, with immunolabelling being low at P10, increasing substantially between P10 and P14 and remaining fairly constant throughout adulthood. With respect to its distribution within the retina, Aggrecan immunolabelling was particularly evident in the outer plexiform layers (OPL), inner plexiform layer (IPL) and ganglion cell layer (GCL), with relatively little staining in the outer nuclear layers (ONL) ( Levels of Vcan mRNA showed greater inter-sample variability than that of the other CSPGs. However, expression was broadly similar between P10 and 6 weeks of age, and lower at 3 and 6 months of age, although this was not a statistically significant reduction (Figs. 1b, e, g; Table 2), due to inter-sample variations. IHC showed that Versican is present in all the layers of the retina, with particularly strong labelling in the photoreceptor inner/ outer segment region (IS/OS) region and the OPL ( Fig. 1g; Supplementary Figs. S1, S2). Staining intensity did week (advanced-stage) Pde6b rd1/rd1 retinae compared to age-matched wildtype mice. (f) Acan mRNA levels decreased in adulthood in Rho -/retinae but remained higher than age-matched wildtype retinae. N.B. mRNA levels shown for wildtype are the same as in Fig. 1b www.nature.com/scientificreports/ not change markedly over time in wildtype mice, either across the later stages of maturation or during adulthood, although did appear weaker at 3 and 6 months (Fig. 1g, Supplementary Figs. S1, S2), consistent with the PCR data.
Ncan mRNA was already present at P10, but expression reduced dramatically within the first three weeks of life in wildtype mice, with a relative value (compared to P10 wildtype) of 0.15 (± 0.04) by 3 weeks and 0.01 (± 0.01) at 6 weeks of age, a level maintained up until at least 6 months of age (Fig. 1c, e; Table 2). Neurocan can undergo proteolytic cleavage of the intact molecule, resulting in smaller isoforms; the C-terminal fraction, known as Neurocan-C, and two variants of the N-terminal end, termed Neurocan-N. We therefore used two antibodies, one that detects Neurocan-C and another that detects both forms of Neurocan-N (Fig. 1g, Supplementary Figs. S1, S2). In the P10 wildtype retina, both Neurocan-C and Neurocan-N were expressed throughout all layers of the retina (Fig. 1g, Supplementary Figs. S1, S2). Both forms were particularly highly expressed in the OPL and at the apical side of the retina, particularly in the region of the photoreceptor inner segments, and at the basal side of the retina, in the GCL. This distribution pattern and level of expression was maintained until 3 weeks, but staining decreased dramatically between 3 and 6 weeks and remained low thereafter (Fig. 1g, Supplementary Figs. S1, S2).
Next, we investigated the expression levels and distribution of Phosphacan. Phosphacan is a splice variant of the receptor protein tyrosine phosphatase (also known as PTP-zeta, Ptprz1).   www.nature.com/scientificreports/ with a relative value of 0.17 (± 0.03) by 3 weeks and 0.03 (± 0.01) at 6 months of age (Fig. 1d, e, Table 2). IHC showed that Phosphacan immunolabelling was most evident in the GCL, at the apical margin of the OPL and in the photoreceptor inner segment region, with weak staining in the INL and ONL (Fig. 1g). Similar to Neurocan, immunolabelling for Phosphacan decreased markedly after the first 3 weeks, with very little staining evident from 6 weeks of age in wildtype retinae (Fig. 1g, Supplementary Figs. S1, S2). Together, these data show that the four CSPG core proteins show very different and distinct patterns of expression both with respect to their distribution in the retina and the amount of each core protein expressed over time in the normal wildtype adult retina. We next examined how expression of these four core proteins changes in three different models of inherited retinal degeneration. Each model was examined at early, mid and advanced stages of degeneration and compared against the relevant age-matched wildtype controls.
Aggrecan expression is markedly increased in degeneration in Pde6b rd1/rd1 mice. Among the CSPGs we examined, Aggrecan expression showed the most distinct changes in expression both compared to wildtype and between models. In Aipl1 -/and Pde6b rd1/rd1 mice, Acan mRNA increased between P10 and 3 weeks, following a similar pattern to wildtype. Notably, at middle stages of degeneration in Aipl1 -/and advanced stage in Pde6b rd1/rd1 mice, Acan expression was significantly higher when compared to age-matched wildtype (Fig. 2d, e). The increase in Acan expression was most evident in Pde6b rd1/rd1 mice, with a relative value of 5.34 (± 0.08) at advanced stage (6 weeks) compared to 0.87 (± 0.06) at early stage (P10) degeneration in the same model (Fig. 2e, Table 2). This also yielded a highly significant increase in expression compared to age-matched (6 week) wildtype controls (5.34 ± 0.08 and 0.99 ± 0.16, respectively; Fig. 2e, Table 2). The overall distribution patterns of Aggrecan immunolabelling in Aipl1 -/and Pde6b rd1/rd1 were similar to that of wildtype, being most evident in the OPL, inner nuclear layer (INL), IPL and GCL, although increases in labelling was most noticeable in the IPL (Fig. 2a,  b, Supplementary Fig. S3).
Surprisingly, Rho -/mice, which exhibit a slower rate of degeneration compared to either Aipl1 -/or Pde6b rd1/rd1 , presented a different pattern of Aggrecan expression. Acan mRNA levels followed the same pattern as seen in development, reducing between 4 and 6 weeks of age, and lower at 6 months of age (Figs. 1b, 2f, Supplementary  Fig. S8a). However, similar to Aipl1 -/and Pde6b rd1/rd1 mice, expression was elevated at mid and advanced stages of degeneration, compared to age-matched wildtype mice (Figs. 1b, 2f). IHC revealed a pattern of immunolabelling very similar to that of wildtype and no degeneration-associated changes in the localisation of expression were observed (Fig. 2c, Supplementary Fig. S3).

Versican expression in degeneration is similar to the pattern of expression in wildtype retina across time but demarcates Müller glial processes in Rho -/mice.
In the disease models, Vcan mRNA expression remained largely unchanged across the respective time points examined for each model (Fig. 3d-f, Supplementary Fig. S8b). There were weak trends for an increase in Vcan expression between P10 and 3 weeks in Aipl1 -/- (Fig. 3d) and a peak in expression at 3 weeks (compared to P10 and 6 weeks) in Pde-6b rd1/rd1 (Fig. 3e), reflecting the relevant IHC data, but these were not statistically significant due to large intersample variation. In Rho -/retinae, Vcan mRNA expression showed a trend for reducing expression with time (Fig. 3f). Notably, Vcan expression was significantly higher in more advanced stages of degeneration in Rho -/-, but was similar to, or even lower than, age-matched wildtype between P10 and 6 weeks. Versican maintained a similar distribution as seen in wildtype across time, even in the later stages of degeneration in both Aipl1 -/and Pde6b rd1/rd1 , with expression predominantly located in the two plexiform layers, INL and GCL (Fig. 3a, b, Sup-  Fig. S4). Conversely, in Rho -/the immunolabelling for Versican was stronger in 3 months old Rho -/retinae, compared both to earlier time points and to age-matched wildtype controls; labelling was most evident in the ONL and IS/OS at this stage (Fig. 3c, Supplementary Fig. S4). Versican was also more noticeable in the INL and IPL at middle time-points in Pde6b rd1/rd1 mice and at later time-points in Rho -/mice, something not seen in age-matched wildtype mice (Fig. 3b, c, Supplementary Fig. S4). As noted above, main figures show representative images of the central, superior retina only. However, we also examined the entire retina in all cases for other spatially restricted changes. In the main, there were no obvious region-specific differences, but we did observe a disease-and region-specific change in Versican expression. Versican was found associated with Müller glial-like processes in the peripheral retina of Rho -/mice from 6 weeks onwards (Fig. 4a). Co-staining with Müller specific markers confirmed the identity of the Versicanpositive processes (Fig. 4b). This observation was specific to Versican and Rho -/and was not observed for any other model or CPSG we examined (see Supplementary Figs. S1-S7). Given this apparently glia-specific pattern of Versican expression, we also examined whether Versican was expressed by Iba1-posiitive microglia cells (Supplementary Fig. S5). Co-staining was not possible due to incompatible labelling protocols; however, comparison of the patterns of Iba1 and Versican immuno-labelling does not indicate colocalization, at least within the ONL, OPL, INL and IPL. However, since some microglia somata may be located in the GCL and nerve fiber layer, we cannot rule out the possibility that these may express Versican in the later stages of degeneration in the periphery of Rho -/mice.

Neurocan expression in degeneration is similar to the pattern of expression in wildtype retina across time.
Perhaps surprisingly in view of the reported upregulation of Neurocan in other regions of the brain and the optic nerve following injury 14,18,19,30 , qRT-PCR analysis revealed no indication of an increase in Ncan mRNA levels in any of the three models of progressive degeneration, compared to age-matched wildtype retinae. Both Aipl1 -/and Pde6b rd1/rd1 mice showed the same profile of expression across age, with highest levels seen at P10 and decreasing with age, in line with wildtype mice (Fig. 5d, e; Table 2). However, the relative levels of Ncan at P10 were lower than that of age-matched wildtype mice in both Pde6b rd1/rd1 and Rho -/mice and expression dropped further after P10 (Fig. 5e, f, Supplementary Fig. S8c). This initial reduction was most apparent in Rho -/- (Fig. 5f) although Ncan mRNA levels then remained stable between 3-4 weeks and 3 months of age in Rho -/mice. This differs from wildtype mice where the levels continued to drop after 6 weeks of age, yielding a significant difference between Rho -/and age-matched wildtype mice (Fig. 5f).
The spatial distribution of both Neurocan-C and Neurocan-N is similar in the different models of degeneration as in the wildtype eye, with expression predominantly locating to the OPL and outer margin of the retina in the segment region (Fig. 5a, b, c, Supplementary Figs. S1, S6). Neurocan-N levels were broadly similar in all disease models, although immunolabelling was consistently stronger in the Aipl1 -/mice than that observed in age-matched wildtype mice (Fig. 5a, Supplementary Figs. S6). Conversely, immunolabelling for Neurocan-C was typically weaker in mid to advanced stages of degeneration in both Aipl1 -/and Pde6b rd1/rd1 eyes, compared to wildtype (Fig. 5a, b, Supplementary Figs. S1, S6). In Rho -/-, Neurocan-N remained fairly consistent between 3-4 weeks and 3 months of age (earlier time points not examined), while immunostaining intensity increased over the same period for Neurocan-C (Fig. 5c, Supplementary Fig. S6). Both Neurocan-C and Neurocan-N were distributed in a pattern similar to that seen in the wildtype retina, throughout all the layers of retina, with increased deposition of Neurocan-C at mid to advanced stages in GCL, IPL, INL and OPL (Fig. 5c, Supplementary Fig. S6). No consistent differences between the superior or inferior retina in the patterns of Neurocan (C or N) staining were observed in any of the models (Figs. 5a, b, c, Supplementary Figs. S1, S6).

Phosphacan expression in degeneration is similar to the pattern of expression in wildtype retina across time.
In all disease models, Ptprz1 mRNA levels followed the same expression profile as wildtype ( Fig. 1d), with expression of Ptprz1 decreasing significantly over time (Fig. 6d, e, f). Of note, and similar to Ncan, mRNA levels of Ptprz1 in Pde6b rd1/rd1 at P10 was statistically lower than that of the same age in wildtype mice (Fig. 6e). The same tendency was also observed in Aipl1 -/and Rho -/mice but was not statistically significant (Fig. 6d, f, Supplementary Fig. S8d). In adult Rho -/mice, Ptprz1 expression remained low and unchanged over time and degeneration, like age-matched adult wildtype retina (Fig. 6f). Aipl1 -/and Pde6b rd1/rd1 mice showed a similar distribution of Phosphacan protein as that seen in wildtype mice, although labelling was more intense in the OPL and RPE, and comparatively low in the GCL (Fig. 6a, b, Supplementary Fig. S7). Of note, Rho -/mice showed little or no change in immunolabelling for Phosphacan and the distribution remained similar to agematched wildtype retinae throughout (Fig. 6c, Supplementary Fig. S7).

Discussion
CSPGs are widely reported to be upregulated following neuronal injury. This study demonstrates that this is an oversimplification and that there is significant heterogeneity in response to progressive neuronal degeneration, with different CSPG core proteins exhibiting very distinct patterns of expression, even between models with similar rates of neurodegeneration. Previously, we examined global changes in CSPG protein deposition using the CS-56 antibody 24 and found that deposition broadly increased over time in wildtype and Rho -/and decreased in Pde6b rd1/rd1 . However, such an approach may mask different changes in individual CSPGs. Our present data examining mRNA levels for four CSPG core proteins showed decreasing expression of Ncan and Ptprz1 in all models, contrasted by a significant increase of Acan in Pde6b rd1/rd1 and Aipl1 -/retinae, and no significant change of Vcan in any model, including wildtype. These differences highlight the importance of studying specific CSPGs, as well as global changes. This study is the first comprehensive assessment of specific CSPGs in murine models www.nature.com/scientificreports/ of progressive retinal degeneration. It also provides the first comparative assessment of multiple CSPGs across time in the late postnatal and adult wildtype retina. Many factors affect the relative levels of a given CSPG core protein, including synthesis and degradation. CSPG-degrading enzymes such as MMP (matrix metalloproteinases) and ADAMTS (A Disintegrin and Metalloproteinase with Thrombospondin Motifs) [31][32][33] degrade different CSPGs with different efficiencies. The expression of these enzymes can change significantly, both during development and in ocular diseases, such as diabetic retinopathy, myopic chorioretinal atrophy, cataract and retinitis pigmentosa [34][35][36][37][38] , and alongside Müller glia proliferation 39 . Other CSPGs that were not assessed in the current study may also contribute to the global changes in CSPG deposition, as assessed with CS-56.
Understanding how CSPGs exert their actions on neurite outgrowth is essential if we are to manipulate the diseased microenvironment in order to make it more permissive for regeneration. In our present work, Aggrecan and Versican were most prevalent, particularly in degeneration, and with aging, while Neurocan and Phosphacan were much less evident. Aggrecan and Versican each contain more GAG chains (~ 100 and 5-23, respectively) than either Neurocan or Phosphacan (< 5) 40 , suggesting that these CSPGs may present an inhibitory chemical barrier to synaptic plasticity during degeneration and/or regeneration. Enzymatic digestion of GAG chains can promote axonal regeneration and plasticity following CNS damage 7 . Consistent with this, treatment of recipient retinas with Chondroitinase ABC improved donor-host interactions following cell transplantation 1,[41][42][43][44] , while viral diffusion and transduction efficiency were improved by co-application of ChABC with viral vector delivery [45][46][47] . Enzymatic degradation is not without its limitations, however 48,49 and significant inhibitory function is retained by the core protein, even after GAG digestion 9 .
Of the CSPGs examined, Aggrecan demonstrated the greatest heterogeneity across disease models. In the normal adult eye, it is thought to play a role in the maintenance of the laminar structure of the retina [50][51][52] . In mice, refinement of synaptic sub-laminae is not complete until P21 53 . Aggrecan expression is seen from embryonic stages and increases with age; it is widespread, even in the normal adult retina, particularly in the synaptic layers 21,22 . Consistent with this, we found Acan expression to increase significantly between P10 and 3 weeks of age and remain at a similar level thereafter in the wildtype retina; immunolablling for Aggrecan increased markedly between P10 and P14 and was most evident in the inner and outer plexiform layers. Together, these changes indicate a role for Aggrecan in the establishment and maintainance of laminar structure. Previous reports examining the developing rat and mouse retina have reported the expression of both Neurocan and Phosphacan from E13-17, peaking around P7 20,21,54,55 . Given that the primary focus of this study was to examine expression across progressive degeneration, we did not look earlier than P10. However, consistent with these reports, we found that the expression of Neurocan and Phosphacan was highest in all models at P10 and declined thereafter.
Aggrecan's upregulation following acute injuries, such as spinal cord crush 13,37 , is well reported. It impedes the regeneration of both endogenous and transplanted neurons 12,13 . Its expression in progressive degeneration is less well characterised but it is upregulated in rats with retinal dystrophy 52 and induced retinal ischemia 26 . Similarly, we found that Aggrecan expression was dramatically upregulated in the Aipl1 -/and Pde6b rd1/rd1 models of rapid photoreceptor degeneration, but, surprisingly, not in the more slowly degenerating Rho -/-, which undergoes a slower rate of degeneration. These differences are intriguing and further exploration of Aggrecan in other models of slow-to-moderate degeneration would be of interest to establish whether rate of degeneration is a determining factor. Some ADAMTS genes, including the aggrecanase ADAMTS5 (also known as ADAMTS11), are reported to be upregulated in AMD and are thought to influence retinal pathology by proteolytic modification of the retinal extracellular matrix 56 . Of note, studies in the barrel cortex have shown that of PNNs are lost during sensory deprivation and that this is mediated by a reduction in the expression of Aggrecan 57,58 . Hence, targeted reduction of Aggrecan may be beneficial for therapeutic strategies that require the formation of new synapses.
Although Versican expression was broadly similar in wildtype, Aipl1 -/and Pde6b rd1/rd1 retinae, we found notable region-specific changes in expression in the Rho -/mouse. In our study, Versican + /Gfap + Müller glial processes were observed, but only in the peripheral retina of Rho -/mice and at later stages of degeneration. Iba1 + microglia in the same region did not appear to co-label with Versican. The differential distribution of Versican, both with respect to a sub-population of Müller glia within a given retina and also between models is worthy of further investigation. Note that the rod photoreceptors in the Rho -/model are non-functional from their genesis onwards so it is conceivable that some of the model-specific changes observed in this study may reflect a developmental alteration in response to the absence of rod function. Cone function is, however, normal in these mice. In order to assess all isoforms of Versican, we used an antibody detecting the G3 domain to assess regional changes and chose PCR primers corresponding to the same sequence used to raise the antibody (exon [14][15]. The G3 domain has been reported to play a key role in the secretion of Versican from the cell 59 and binds to several molecules including the EGF receptor, tenascin, fibronectin and integrin β1 60,61 . The G3 domain has been shown to promote neurite outgrowth and cell attachment in hippocampal neurons via an EGF receptor-mediated pathway 62 , potentially indicating a localised pro-regenerative impact on the environment in the periphery. The specific association of G3 domain with peripheral Müller glial processes, which in lower vertebrates retain a more immature, progenitor/stem cell-like phenotype, indicates local microenvironmental differences within the degenerating retina.
Neurocan is upregulated in many experimental models of acute injury in the brain 14,18,19 , optic nerve crush 30 and in the ischaemic rat retina 63 . In contrast, we did not observe any upregulation of Neurocan with progressive retinal degeneration. Indeed, mRNA levels of Ncan and Ptprz1 at P10 were actually significantly lower in Pde-6b rd1/rd1 and Rho -/-, compared to age-matched wildtype mice. There was an accompanying reduction in Neurocan-C staining at the OPL/ONL interface as degeneration progresses in Aipl1 -/and Pde6b rd1/rd1 , while labelling for the Neurocan-N fraction remained fairly constant in all disease models. Neurocan, and its cleavage, demonstrates dynamic changes during development, injury and remodelling in the CNS; full length Neurocan was expressed in juvenile rat brain 65 and following acute brain injury 14 www.nature.com/scientificreports/ rat adult brain, and more interestingly, only N-and not C-terminal fragments were observed as a component of PNNs in adult rat cerebellum 64 . A potential role for cleaved Neurocan was indicated in a recent study showing Neurocan-N fractions are essential in Semaphorin 3F-induced dendric spine remodelling 66 . In our data, the labelling of both C-and N-terminal fractions were observed in early stages, indicating that full length fractions are likely present in the juvenile retina. Conversely, neither of the fractions were seen in significant amounts from 3 weeks of age in wildtype mice. Both changes are similar to those described for the brain 64 . In contrast, Neurocan-N and -C fractions predominated in Aipl1 -/and Pde6b rd1/rd1 , respectively. Similarly, Phosphacan labelling was seen at the outer margin of the OPL between P10 and 3 weeks in wildtype, but not in Aipl1 -/and Pde6b rd1/rd1 , retinae. Observationally, the separation of photoreceptors in the ONL and the OPL was also less well defined in both Aipl1 -/and Pde6b rd1/rd1 mice, compared to wildtype. While the reduced levels may simply reflect a reduction in the number of photoreceptor neurons that may produce these proteins, the different patterns of expression seen in wildtype and diseased retina, together with previous studies examining Neurocan cleavage, suggest that Neurocan and Phosphacan may play a role in the establishment and maintenance of retinal lamination. In summary, our study provides the first comprehensive characterisation of individual CSPGs and how their expression changes both with age and with progressive retinal degeneration. It also highlights an important heterogeneity in the retinal environments that arise in degeneration due to different genetic causes and the importance for researchers to characterize the specific models they use. Understanding the distinct changes in expression and distribution of individual CSPGs may help us to modulate specific diseased microenvironments, affecting synaptic plasticity, regeneration and repair.