Profiling Retinal Biochemistry in the MPDZ Mutant Retinal Dysplasia and Degeneration Chick: A Model of Human RP and LCA

Abstract

Purpose.: Raman microscopy, a rapid nondestructive technique that profiles the composition of biological samples, was used to characterize retinal biochemistry in the retinal dysplasia and degeneration (rdd) and wild-type (wt) chick retina during retinogenesis and at hatching.

Methods.: Embryonic day (E)13 and posthatch day (P)1 rdd and wt retinal cross-sections (n = 3 of each line at each age) were profiled using 633 helium–neon laser excitation. The biochemical composition was determined using computational analysis of the Raman spectra. In parallel histology, TUNEL and glial fibrillary acidic protein (GFAP) immunostaining were used to visualize retinal dysfunction.

Results.: Principal component (PC) analysis of the Raman spectra identified 50 major biochemical profiles, but only PCs that made significant contributions to variation within rdd and wt retina were mapped. These significant PCs were shown to arise from DNA, various fatty acids, melanin, and a number of proteins. Distinct patterns of GFAP immunostaining and a larger population of TUNEL-positive nuclei were observed in the rdd versus wt retina.

Conclusions.: This study has demonstrated that Raman microscopy can discriminate between major retinal biomolecules, thus providing an unbiased account of how their composition varies due to the impact of the MPDZ null mutation in the rdd chick relative to expression in the normal wt retina.

Apremature stop codon in the multiple PDZ domain protein (MPDZ) gene has been identified as the mutation responsible for pathology in the retinal degeneration and dysplasia (rdd) chick and analogous mutations have been identified in patients with human retinitis pigmentosa (RP) and Leber congenital amaurosis (LCA). ¹ MPDZ participates in tight junction formation with the extracellular proteins, occludin and claudin, and intracellular docking proteins, zonula occludens 1 (ZO-1) and ZO-2. Collectively, these proteins determine the permeability of tight junctions in many tissues, including the retina where MPDZ has been localized to the outer limiting membrane (OLM). ^{1 –3} Effects of the MPDZ null mutation in the rdd chick retina were first observed at embryonic day (E)18, with the onset of abnormal undulation in the outer plexiform layer (OPL) and associated disarrangement of the maturing outer nuclear layer (ONL). These retinal abnormalities are accentuated at P1 and accompanied by apparent detachment of the disorganized photoreceptor outer segments from the retinal pigmented epithelium (RPE). The rdd chicks are sighted at hatching but no vision is detectable 10 weeks after hatching. ^4,5 Proteomic analysis of retina from E12 to posthatch day (P)1 revealed that several proteins were differentially expressed during retinogenesis and with the onset of degeneration in the rdd chick; this would suggest that the MPDZ mutation may have an impact on retinal development before observable defects at E18. ⁶  

Genetic analysis has successfully identified many mutations in diverse photoreceptor proteins associated with degenerative retinopathies. ⁷ Regardless of the identified mutation, patients with RP present with acute or chronic progressive retinal degeneration. Analysis of the protein structural defects from common retinal gene mutations has provided insight into the potential mechanisms that drive retinal degeneration. ^8,9 Proteomics has begun to provide insight at the level of the effector molecules, demonstrating the broader effect that a mutant protein has on retinal protein expression. ^6,10 However, no mode of investigation to date has provided evidence of the broad biochemical impact that a specific mutation has on retinal physiology and how it becomes manifest with pathologic progression. 

Raman microscopy enables molecular level evaluation of tissue composition and potentially offers new insights of physiologic, pathologic, and pharmacologic events in situ that are not readily detectable by other modes of investigation. ¹¹ Advantages that Raman spectroscopy offers include detection of multiple native unlabeled biochemicals, thus allowing rapid, noncontact, and noninvasive analysis of discrete changes in biomolecular composition. ¹² The Raman spectrum is proportional to the relative abundance of each biochemical in a specimen, generating semiquantitative data. Raman spectroscopy has been used to analyze retinal carotenoids, ¹³ profile diabetic vitreous, ¹⁴ characterize advanced glycation end products (AGEs) in Bruch's membrane, ^{15 –18} and to map the distribution of the multiple biochemical constituents in porcine retina. ^19,20  

This study used Raman spectroscopy to map multiple native unlabeled biochemicals during normal wild-type (wt) chick retinogenesis and the onset of retinal degeneration in the rdd chick. Two distinct ages were chosen. The first was E13, when the predominant phase of retinal neurogenesis is complete and synaptogenesis has commenced. Histologically, the E13 rdd and wt are indistinguishable, yet subtle proteome variation was evident. ⁶ The second time point, P1, compared the normal functional wt retina with rdd chick retina exhibiting histologic features of the onset of retinal degeneration and exhibits modulation of the proteome, but at an age when the birds still retain functional vision. ⁶  

Materials and Methods

Experimental procedures were conducted in compliance with the United Kingdom Animals (Scientific Procedures) Act 1986 that fulfills the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and European Directive 86/609/EEC to minimize animal use for research. Fertilized White Leghorn chicken (Gallus gallus domesticus wt) and rdd eggs were incubated in a commercial egg setter and hatcher at 38°C (Roslin Institute). Intact E13 embryos were killed by decapitation and P1 chick by cervical dislocation; ocular tissues (n = 3) were immersion fixed in 4% (w/v) paraformaldehyde (4 hours, 4°C), cryoprotected in 5% (w/v) sucrose/PBS (24 hours, 4°C), and stored in 30% (w/v) sucrose/PBS containing 0.01% (w/v) sodium azide (4°C). Cryosections (10 μm) from the central retina of three rdd/wt chicks were subject to Raman examination followed by routine hematoxylin and eosin histology. 

Raman Microscopy and Data Analysis

Raman spectra were obtained as described previously. ^19,20 In brief, a Raman microspectrometer (LabRam HR800; Jobin-Yvon, Villeneuve d'Ascq, France) was used, such that a 633 (20 mW) helium–neon laser, focused on the sample with a ×100 objective in a marginally confocal arrangement (confocal hole diameter: 400 μm), with the laser focused to generate a 3-μm-diameter spot. The microspectrometer was fitted with a 0.1-μm step xyz stage and a 300 groove mm⁻¹ diffraction grating. All optical images and spectral maps were recorded and processed using commercial Raman spectroscopy software (Labspec; Jobin-Yvon). Raman spectra were acquired at 1-μm spacing intervals for the retinal cross-section (n = 3), approximately 120 μm long (the length varied slightly according to the age of retina used). 

The background was subtracted using the method of singular value decomposition–based linear interpolation background correction (Matlab; The MathWorks, Cambridge, UK), which permits baselines to be adapted for variable biochemical composition, providing a more reliable and reproducible data set. ²¹ The baseline-corrected data were analyzed by principal component analysis (PCA), without mean-centering (Unscrambler; Camo, Oslo, Norway). Because some layers of the retina are less dense than others (generating decreased signal intensity) the Raman spectra were not normalized before PCA. Since the retinal layers have distinct compositions, there is no suitable candidate constituent for normalization and so the scores were normalized to the first principal component (PC). 

Raman distribution maps of the PC scores were generated (in Matlab) by creating two-dimensional intensity maps, with the hue of the two constituents set to red and blue to ensure no visual confusion of the colors. The maps were imported into a digital picture and photo editing software program (Adobe Photoshop, San Jose, CA) and processed in parallel using the editing hue/saturation tool to alter the hue of the color images to aid differentiation of the PCs. Each PC describes the variation between two biochemical constituents, which appear as positive and negative peaks in the PC loading, so each map has two corresponding colors. The Raman spectra corresponding to these constituents were obtained by score-weighted averaging of the signals, with scores > 0 (constituents relating to positive bands in loadings) and <0 (constituents relating to negative bands). These signals were compared with reference biochemical signals to identify the potential composition of the constituents. 

TUNEL and GFAP Immunostaining

Cryostat E13 and P1 retinal sections from three rdd/wt chicks were air-dried on gelatin-coated slides. DNA fragmentation was assessed using the manufacturer's protocol and included the addition of a positive control that used recombinant DNAse1 to induce fragmentation of nuclear DNA and the negative control, which consisted of label buffer lacking terminal transferase (in situ cell death detection kit; Dako UK Ltd., Cambridgeshire, UK). Polyclonal rabbit anti-bovine glial fibrillary acidic protein (GFAP) antisera (code Z0334; Dako UK) was used at 1:400 dilution in PBS buffer containing 0.5% (w/v) Triton X-100 and 0.5% (w/v) BSA (24 hours, 4°C), washed in PBS (10 minutes), and incubated with a fluorescein-conjugated secondary anti-rabbit immunoglobulin (1:100; Alexa Fluor 488; Invitrogen) for 1 hour and washed in PBS. Nuclei were counterstained with propidium iodide (Invitrogen) at 2.5 μg/mL for 10 minutes, washed in PBS, mounted, and viewed using an inverted microscope (Eclipse TE2000-E; Nikon Instruments). Control studies included incubation with nonimmune primary rabbit antisera, inclusion of a 1-hour 0.5 M salt wash after incubation of the primary GFAP antibody, the addition of poly-l-lysine (2 mg/mL, MW 3800) to the primary GFAP antiserum, and omission of primary antiserum. ^22,23  

Results

Assignment of Raman Constituents

PCA identified 50 spectrally significant principal components (signal-to-noise ratio > 3) during chick retinogenesis and with the onset of retinal degeneration in age-matched rdd and wt chick retina. Examples of mean raw Raman signals from the retina of E13 and P1 wt and rdd chicks are displayed in Figure 1. The variable, low-information broad non-Raman background is eliminated to allow assessment of only the Raman spectroscopic information. The resultant subtraction spectra show clear Raman spectral differences, even without further processing of the spectra (Fig. 1, bottom). Nevertheless, to derive a more complete biochemical overview, it is necessary to conduct more comprehensive multivariate analysis of these raw data. Twenty-three of the derived PCs exhibited statistically significant differences (t-test for mean sample spectra gave P < 0.05) between either wt and rdd or between pre- and posthatch, and those that differed most significantly (P < 0.001) were evaluated and are displayed in Figure 2 alongside the corresponding histologic images for rdd and wt E13 and P1 retinas. Each of these constituents is now considered. 

Figure 1.

 

Mean raw Raman signals from the retina of E13 and P1 wt and rdd chicks. The variable, low information broad non-Raman background is eliminated to allow assessment of only the Raman spectroscopic information. The subtraction spectra at the bottom show clear Raman spectral differences even without processing the spectra.

Figure 2.

 

Representative pseudocolor Raman maps for E13 and P1 rdd and wt retina depicting the distribution of +ve and −ve PCs mapped onto the corresponding histologic images. Each PC describes the variation between two biochemical constituents that appear as positive and negative peaks in the PC loading; therefore, each map has two corresponding colors. PC2, distribution of DNA (orange) and low unsaturated fatty acid lipids (sky blue); PC3, melanin (yellow); PC4, an alpha helical protein (cyan) and a random secondary structure protein (red); PC8, distribution of a mixed alpha helical/random structure protein (green) and PUFA (purple): PC10, distribution of the further protein signals (royal blue and yellow ochre). PC18 maps the distribution of transfatty acids (magenta) and partially oxidized PUFA (turquoise). The color hue coding corresponds to that used in the spectra displayed in Fig. 3.

The signal corresponding to the positive PC2 bands (orange, PC2⁺) contains significant contributions from DNA and nuclear proteins (Fig. 3, Tables 1 and 2). ²⁰ PC2⁻ displays peaks characteristic of a low unsaturated fatty ester (Fig. 3, Tables 1 and 2). The map of the Raman scores (Fig. 2) for the DNA constituent correlates with the histologically determined developing outer (DONL) and inner nuclear layers (DINL) at E13 and in the maturing P1 retina. Likewise, the low unsaturated fatty acid constituent (sky blue) correlates with the DONL and inner plexiform layer (IPL) and nerve fiber layer (NFL); this is consistent with the increasing density of lipid-rich molecules localized in the membranes of these axonal and synaptic zones. The RPE layer also exhibits an intense DNA score in both rdd and wt at E13 but only for the wt chick on hatching because the posthatch rdd chicks show much reduced DNA scores in this region. 

Figure 3.

 

Color-coded Raman spectra for the positive and negative PCs depicted in Fig. 2. The Raman spectra corresponding to each constituent were obtained by score-weighted averaging of the signals, with scores greater than zero relating to positive bands in the loadings and those less than zero relating to negative bands. These signals were compared with reference biochemical signals to describe the biochemical composition of the tissue. The hues used to color the spectra are consistent with those used to represent the constituents in the maps of Fig. 2.

Table 1.

 

Assignment of Selected Bands and Regions in the Fingerprint Region of the Raman Spectra of Retina Tissue

Band Position		Assignment
Low	High	Biochemical Class	Mode
1730	1750	Fatty acid ester	ν(C═O) ester carbonyl
1665	1680	Protein—Amide I	β-sheet
1660	1665	Protein—Amide I	Random structure
1655	1665	Fatty acid	ν(C═O) unsaturated bond
1645	1660	Protein—Amide I	α-helix
1590		Melanin	Melanin D mode
1580		DNA/RNA	A, G
1550	1630	Protein	His, Tyr + Phe, Trp, Pro
1545	1625	Heme, Cytochrome C	Resonance-enhanced porphyrin ring
1490		Heme, DNA/RNA	A, G
1400	1490	Protein, DNA, lipid	CH2 scissor + Trp His Phe Asp, Glu
1375		DNA/RNA	A
1200	1370	Protein	CH2 twists and rocks, Trp, His, Tyr
1345		DNA/RNA	A,G
1320		Melanin	Melanin G mode
1312	1400	Heme	Cytochrome C
1305		Protein—Amide III	α-helix
1295	1305	Fatty acid	δ(CH2) twist, saturated carbon
1299		DNA/RNA	A
1265		Protein—Amide III	α-helix
1260		Fatty acid	δ(C—H) bend, unsaturated carbon
1250	1255	Protein—Amide III	Random structure
1240	1250	DNA/RNA	U,C,A
1225	1232	Protein—Amide III	β-sheet
1095		DNA/RNA	Phosphate
1130	1230	Heme, Cytochrome C	Resonance-enhanced porphyrin ring
1100	1140	Fatty acid	ν(C—C)ip, straight chain
1080	1090	Fatty acid	ν(C—C), bent chain
1062		Fatty acid	ν(C—C)op, straight chain
1020	1130	Protein	ν(C—O,C,N)
1003		Protein	Phe

 Nucleotide abbreviations: A, G, U, and C. Assignments derived from Terada et al., ²⁴ Deng et al., ²⁶ Tu, ²⁷ Filipek et al., ²⁸ Beattie et al., ²⁹ and Wood and McNaughton. ³⁰

Table 2.

 

Assignments for Extracted Constituent Raman Signals

Constituent	Assignment	Key Features
PC2+	Nuclei	DNA mode 1578 cm−1, R 2 = 0.89 vs. porcine ONL
PC2−	Fatty acid esters	Carbonyl ester mode 1732 cm−1, unsaturated mode 1659 cm−1, R 2 = 0.96 vs. bovine orbital adipose tissue
PC3+	Sucrose	No Raman modes above 1500 cm−1 indicates the absence of any unsaturation, esters, acid, or aromatic groups
PC3−	Melanin	Broad double bands at 1320 and 1587 cm−1, characteristic of massive extended conjugated double bond systems
PC4+	Alpha helical protein	Amide I and III modes are located at 1307 and 1655 cm−1
PC4−	Random coil protein	Amide I and III modes are located at 1253 and 1660 cm−1
PC8+	Unassigned protein	Unique features include a broad band at 1578 cm−1 and a sharp peak at 1279 cm−1
PC8−	PUFA	Carbonyl ester mode at 1736 cm−1. Very strong contribution from unsaturated bonds at 1655 cm−1
PC10+	Unassigned protein	Mixture of alpha helix and random coil
PC10−	Unassigned protein	Alpha helical with unique bands at 1607 and 1280 cm−1
PC18+	Trans fatty acid	Unsaturated bond mode at 1670 cm−1, pattern of bands Characteristic of solid fats (1062 and 1100 cm−1)
PC18−	Oxidized PUFA	Intense band at 1655 cm−1 indicates high unsaturation, bands at 1578 and 1680 cm−1 typical of oxidized PUFA, shoulder at 1640 cm−1 due to conjugation of double bonds

 Constituent nomenclature is based on the principal component rank (PC#), with superior “+” and “−” indicating that the signal was reconstructed from the positive or negative portion of the component, respectively. R ², correlation of the intensities at each Raman shift of the reference and test signals. PUFA, polyunsaturated fatty acid (content).

The PC3⁻ signal (yellow) displays the characteristic features of melanin ²⁴ (Fig. 3, Tables 1 and 2), which was localized to the RPE layer in the E13 rdd and wt and P1 wt profiles. This contrasts with the reduced melanin signal detected in the P1 rdd retina. The PC3⁺ signal (navy blue) was identified as sucrose, the tissue cryoprotectant (Fig. 2). 

The positive constituent signal for PC4 displays features characteristic of an α-helical protein (cyan blue) relative to the negative constituent that exhibits features of a random coiled protein (red) (Fig. 3, Tables 1 and 2). At E13 the α-helical protein constituent of PC4 exhibits a limited distribution in the RPE, as well as a weak profile in the ganglion cell layer (GCL). At P1 PC4⁺ in the wt was evident in an area straddling the ONL margin and the OPL and also in the GCL. Discrete PC4⁺ clusters are dispersed in the ONL, OPL, and ONL. PC4⁻ was dominant in the E13 rdd and wt retina, exhibiting widespread intense expression. The intensity of the PC4⁻ constituent was maintained in the ONL in the P1 wt retina with less intense expression in the OPL and GCL region. This contrasts with the P1 rdd, which exhibited significantly decreased signal intensity throughout the retina (Fig. 2). 

The positive PC8⁺ constituent (green) corresponds to a protein signal exhibiting a mixture of α-helical and random structures, whereas the negative PC8⁻ constituent (purple) exhibits characteristic bands for a fatty acid ester along with α-helical protein (Fig. 3, Tables 1 and 2). This PC8⁻ lipid contrasted with the lipid in PC2⁻ in having a significantly enhanced contribution from unsaturated C═C bonds, indicating a high polyunsaturated fatty acid content (PUFA). PC8⁻ was also very prominent in the RPE layer in the E13 rdd and wt and P1wt retina. A diffuse PC8⁻ signal is evident throughout the P1 rdd retina with moderate accumulation in the GCL and NFL, with a weak, diffuse signal detected in the ONL of the P1 wt retina. An intense PC8⁺ band was evident in the NFL of the E13 rdd retina, with diffuse features dispersed throughout the E13 and P1 rdd and wt retina (Fig. 2). 

A difference in the expression of the positive (royal blue) and negative (yellow ochre) constituents in the embryonic and posthatch retina is evident in PC10. The PC10⁺ signal displays many typical protein bands, including modes indicative of a mixed α-helical and random secondary structure. Moderately intense PC10⁺ was evident in the RPE of E13 and P1 rdd and wt retina with a weak diffuse distribution in the ONL of E13 rdd and P1 retina. The PC10⁻ signal is also characteristic of a protein, although more predominantly α-helical compared with the mixed signal in PC10⁺ (Fig. 3, Tables 1 and 2). In the P1 rdd retina, a moderately strong PC10⁺ signal extended from the RPE to GCL; this contrasted markedly with the weak, diffuse bands detected in the P1 wt retina. An intense PC10⁻ band was evident at E13 GCL in the rdd retina, with discrete areas also apparent in the DONL at E13 for both rdd and wt. At P1 a weak PC10⁻ is evident in GCL and NFL in both rdd and wt (Fig. 2). 

PC18 discriminated between different compositions of fatty acid–based lipids, with PC18⁺ (magenta) PUFA exhibiting features characteristic of trans-unsaturated lipid and of a higher solid fat content and crystalline order. PC18⁻ (turquoise) demonstrated features characteristic of highly unsaturated fatty acids, with additional modes present that correspond with lipid oxidation products such as conjugated dienes and cyclical structures (Fig. 3, Tables 1 and 2). The PC18⁺ component (magenta) exhibits moderate features throughout the retina at E13 in the rdd and similar widespread but less intense expression in the E13 and P1 wt retina. This contrasted with the P1 retina, which exhibited a diminished PC18⁺ contribution. PC18⁻ (turquoise) exhibited minimal expression in the rdd E13 retina relative to the wt E13 retina. By P1 PC18⁻ was the predominant component in the rdd retina, but localized to RPE/ONL and OPL/GCL interfaces (Fig. 2). 

TUNEL-positive nuclei were detected at E13 in both the rdd and wt retinas (Fig. 4c, 4f). Larger numbers of TUNEL-positive nuclei were observed in the rdd retina, primarily in the INL, which also contained a consistent population of TUNEL-positive nuclei (Fig. 4c). At P1 intense TUNEL-positive nuclei were observed primarily in the photoreceptor layer (PRL) (Fig. 4i). A small population of weak TUNEL-positive nuclei was evident in the INL in the P1 wt retina. Control studies were used to validate positive (Figs. 4a, 4d, 4g, 4j) and negative (Figs. 4b, 4e, 4h, 4k) TUNEL staining patterns. 

Figure 4.

 

Representative TUNEL staining (green) in E13 rdd and wt retina (a–f) and P1 rdd and wt retina (g–l) using propidium iodide (red) to counterstain nuclei. A larger number of TUNEL-positive cells were evident in the E13 rdd (c) relative to the wt (f) retina; TUNEL staining (arrowhead) was also detected in a population of cells in the E13 rdd retina (c). At P1 TUNEL-positive cells were primarily evident in the ONL (i) and a limited number of weak TUNEL-positive cells in the wt INL (l). TUNEL-positive control E13 and P1 rdd (a, g) and wt (d, j) and TUNEL-negative control E13 and P1 rdd (b, h) and wt (e, k), respectively. Scale bar: 20 μm.

The rdd and wt E13 retinas were GFAP negative since GFAP is not expressed at this early developmental time point. ²⁵ GFAP immunostaining in fibers in the wt retina at P1 are indicative of radial Müller cell processes at the IPL–INL interface, with an additional zone of GFAP immunopositive fibers linking with and traversing the central IPL. Horizontal fibers were observed in the NFL that may be derived from either Müller or astrocyte cells (Fig. 5b). GFAP immunostaining was evident in fibers ramifying throughout the IPL with fibers in the INL and disrupted OPL, ONL, and PRL. An extensive network of intense, moderate, and weak GFAP-positive fibers was observed in the NFL (Fig. 5a). Control studies used to assess specificity did not alter or diminish the pattern of immunostaining in either rdd or wt retina at P1. 

Figure 5.

 

Confocal microscopy analysis of GFAP immunostaining (green) with propidium iodide (red) counterstain in the P1 rdd (a) and wt (b) retina. Representative confocal micrographs (Z-series stack, 10 × 1 μm) demonstrating GFAP immunoreactivity in a large number of fibers in the NFL, ramifying throughout the IPL, INL, and in the disorganized ONL. The dashed line delineates the deformation of the OPL (a). GFAP immunostaining in a discrete number of fibers in the NFL and localized to a defined zone in the IPL of the P1 wt retina (b). Scale bar: 20 μm.

Discussion

Degenerative retinal disease such as RP affects 1 in 2000 in the population and exhibits a high degree of genetic heterogeneity with >172 gene mutations identified to date (RetNet; http://www.sph.uth.tmc.edu/retnet/). Consequently, RP is a complex disease but, regardless of the identified mutation patients, generally display a common phenotype of night blindness, depletion of visual fields leading to tunnel vision, and often complete blindness. Therefore, given the genetic heterogeneity, a range of disease models have been proposed. ³¹ Nonetheless, the manner by which the diverse range of mutations generates a common degenerative photoreceptor phenotype remains a matter of debate. 

Presently, a range of techniques has been successfully used to study aspects of degenerative retinopathies, although no mode of investigation can provide a comprehensive overview of the complex pathology. This study used Raman microscopy to map biochemical constituents in the developing retina at E13 and P1 in the rdd chick that has a premature stop codon in the MPDZ gene, which drives retinal degeneration. ¹ MPDZ has been detected in the retina at the adherens junction in the OLM, a specialized interactive zone between the photoreceptor inner segments and the innermost process of the radial Müller glia. ¹ In the wt retina this structural relationship maintains the orientation between the photoreceptor inner segment and outer nuclear layer. The earliest morphologically visible impact of the absence of functional MPDZ is at E18 in the rdd retina, which exhibits undulation at the OPL–INL interface and disorganization in the ONL and aberrant photoreceptor outer segments and by P1 the rdd retina exhibits evidence of retinal degeneration. ^1,6  

Raman microscopy has generated a unique data set cataloging previously undetected biochemical modulation in the rdd versus wt chick. Although the earliest observable histologic event related to the lack of MPDZ expression occurs at E18 in the rdd, the Raman data demonstrated that retinal biochemistry was already deviating from age-matched wt. Changes in the PC2⁺ assigned to nuclear constituents matched changes in the distribution of the retinal nuclear layers in the rdd and wt histologic images, whereas the low unsaturated fatty acid constituent (PC2⁻) was characteristically localized in axonal and synaptic zones, consistent with previous Raman studies. ^19,20  

In addition to the low unsaturated fatty acid in PC2⁻, other fatty acid signals were identified in PC8⁻ and in PC18⁻, for which elevated intensity ca. 1659 cm⁻¹ indicated PUFA. The PC8⁻ constituent was highly localized to the photoreceptor layer in the E13 wt/rdd and P1 wt, suggesting that this signal was attributed to the fatty acids in the PRL, of which docosahexaenoic acid (six unsaturated bonds per fatty acid chain) is a major constituent. The PUFA signal was not detected in the P1 rdd PRL, demonstrating alteration of its biochemical composition. The PC2 map indicated that some fatty acid remains in the PRL, but the sighted nature of the posthatch chicks indicates that the remaining lipid is sufficient for some visual function. The P1 rdd chick also exhibited weak accumulation of the PUFA constituent in the NFL, mirroring the increase in GFAP immunostaining, relative to the P1 wt NFL (Fig. 4). PUFA is a major constituent of myelinated nerve fibers; therefore the increased accumulation of PUFA in the NFL of the P1 rdd chick retina may reflect the major changes highlighted by the increased detection of GFAP immunopositive fibers. 

In the P1 rdd chick the dominant PUFA signal was PC18⁻, which was distributed much more widely throughout the retina than the PC8⁻ PUFA. Furthermore, this signal incorporated features typical of PUFA oxidation. ³² In the P1 wt retina this partially oxidized PUFA is mainly concentrated in the PRL and in the IPL adjacent to the GCL. 

PC18⁺ corresponded to a trans fatty acid that is characterized by the position of the C═C stretching mode at 1670 cm⁻¹ combined with the absence of the cis-unsaturated mode at 1260 cm⁻¹. Trans fat was found throughout the retina in E13 wt/rdd and P1 wt, but significantly reduced in the P1 rdd. The significance of this observation remains unknown and warrants further analysis. 

A characteristically strong melanin signal was localized to the RPE in the E13 rdd and wt and P1 wt profiles, with a reduced signal detected in the P1 rdd retina. This observation is consistent with reports that RPE cells in the rdd chick exhibit abnormal pigmentation that may be related to the observed detachment from the photoreceptor outer segments. ^4,6  

The positive and negative constituents detected in PC4 and PC10 as well as in PC8⁺ suggested features characteristic of a range of proteins. Given the strength of these signals, they are likely to represent common proteins that exhibit variable expression in the developing and degenerating retina. To confidently assign the various individual protein signals it will be necessary to first identify potential candidate proteins through comprehensive biochemical assessment of the protein composition of the retina. This task will be performed with the assistance of additional proteome analysis. ⁶ A database of Raman signals will be constructed containing Raman spectra for all potential candidate proteins, which will be compared with in situ constituent retinal Raman signals. 

The PC10⁺ and PC18⁻ Raman constituents were of particular interest since the intensity of these signals was highly elevated in the P1 rdd retina, in the region that was undergoing retinal degeneration and remodeling. TUNEL analysis verified the extent of the ongoing degeneration since a population of intense positive cells was observed in the P1 rdd retina. An additional population of weakly positive TUNEL cells was consistently observed in the midzone of the developing INL, although the relevance of this was unclear and may reflect the fact that TUNEL staining can detect both apoptotic and necrotic or autolytic cells or that weak TUNEL-positive cells are accumulating DNA damage and entering apoptosis. ³³  

GFAP immunostaining was used as a marker of retinal injury and disease. Intense GFAP immunostaining was evident in fibers distributed in the IPL and NFL in wt retinas, which is consistent with expression in chick Müller cells. ³⁴ In contrast, an extensive network of GFAP-positive fibers was observed throughout the rdd P1 retina and a similar pattern of GFAP has been detected in the rdd chick and in several mammalian RP models. ^{34 –36} The deformation of the ONL/OPL/INL and aberrant detection of GFAP fibers in the P1 rdd chick is comparable to the pattern of retinal degeneration observed in the Crb1-mouse model ³⁷ ; this observation is not unexpected given the localization of Crb1 and MPDZ to the OLM. ^1,37  

Our previous proteomic ⁶ study and the present Raman investigation both demonstrate that the null MPDZ rdd retina ¹ is deviating from normal retinogenesis before the first observable morphologic event in the E18 rdd retina. Analysis of the E13 rdd retina, when the phase of retinal neurogenesis is largely complete and synaptogenesis has commenced, revealed some subtle differences, with elevated DNA fragmentation in the INL and a significant elevation in expression of unidentified proteins (PC8⁺ and PC10⁻) in the NFL, which is expressed in much lower levels in the P1 rdd retina. The identity of PC8⁺ and PC10⁻ will be assessed by comparing the Raman spectra of candidate proteins. ¹⁷ The P1 rdd retina exhibited a dramatic decrease in melanin content: loss of PUFA from the PRL with accumulation in the NFL. In addition, an increase in PUFA oxidation was evident throughout the retina and significant changes were observed in a number of proteins. 

This study has successfully provided proof-of-principle that Raman microscopy can detect previously unknown biochemical modulation at the onset of retinal degeneration in the MPDZ mutant rdd chick. Several distinct Raman profiles represent strong candidate biomarkers that characterize the process of degeneration, and ongoing analyses aim to identify and validate their potential using other RP animal models and human retinal samples. The concept of developing in vivo patient-based Raman profiling as a diagnostic and prognostic technique will enable direct analysis of the progression of degeneration, predict which patients are likely to respond to a specific therapy, and monitor the response to personalized therapy. 

Footnotes

 Supported in part by the Department for Employment and Learning, Research, and Development Office of Northern Ireland Grant SPI/2384/03, McCauley Endowment, UK Medical Research Council (MRC) Grant G0600053, Leverhume Trust Grant EM/2006/0049, Biotechnology and Biological Sciences Research Council (BBSRC) Grant JREI 18471, BBSRC core support to The Roslin Institute, and MRC Grant G0501050.

Footnotes

 Disclosure: J.R. Beattie, None; S. Finnegan, None; R.W. Hamilton, None; M. Ali, None; C.F. Inglehearn, None; A.W. Stitt, None; J.J. McGarvey, None; P.M. Hocking, None; W.J. Curry, None

The authors thank Graeme Robertson for technical assistance.