A novel biological function for CD44 in axon growth of retinal ganglion cells identified by a bioinformatics approach

Abstract

The failure of CNS regeneration and subsequent motor and sensory loss remain major unsolved questions despite massive accumulation of experimental observations and results. The sheer volume of data and the variety of resources from which these data are generated make it difficult to integrate prior work to build new hypotheses. To address these challenges we developed a prototypic suite of computer programs to extract protein names from relevant publications and databases and associated each of them with several general categories of biological functions in nerve regeneration. To illustrate the usefulness of our data mining approach, we utilized the program output to generate a hypothesis for a biological function of CD44 interaction with osteopontin (OPN) and laminin in axon outgrowth of CNS neurons. We identified CD44 expression in retinal ganglion cells and when these neurons were plated on poly-l-lysine 3% of them initiated axon growth, on OPN 15%, on laminin-111 (1x) 41%, on laminin-111 (0.5x) 56%, and on a mixture of OPN and laminin (1x) 67% of neurons generated axon growth. With the aid of a deoxyribozyme (DNA enzyme) to CD44 that digests the target mRNA, we demonstrated that a reduction of CD44 expression led to reduced axon initiation of retinal ganglion cells on all substrates. We suggest that such an integrative, applied systems biology approach to CNS trauma will be critical to understand and ultimately overcome the failure of CNS regeneration.

Fig. 1

Classification of the 942 proteins identified by our prototypic program suite using 20 Gene Ontology (GO-) terms. The diagram demonstrates the number of proteins present for each category. Additionally, categories including proteins which are using less common GO terms or demonstrate less esoteric characteristics and proteins for which no classification is available. As a protein may belong to several categories, the total protein sum in this diagram is 1491.

Fig. 2

RGCs plated on different substrates and characterized respectively by their axon length on these substrates. (a) Confocal images of RGCs plated on poly-l-lysine (PLL), osteopontin (OPN), laminin (1×), OPN/laminin (5×), OPN/laminin (1×), and laminin (0.5×) mixtures. (b) The percentage of RGCs that extended axons in vitro was quantified on the aforementioned substrates (5 wells/condition). RGCs adhered to all substrates; *statistically significant, p < 0.05.

Fig. 3

CD44 deoxyribozyme specifically down-regulates CD44 expression in RGCs. (a–c) Confocal images of RGCs growing for 24 h on different substrates as marked were treated with biotin-labeled deoxyribozyme to CD44 in a concentration of 8 µmol/L. Red shows the labeled deoxyribozyme localized in the cytoplasm; green is β-tubulin immunostaining. In each image, cross-sections (lower or right-hand sides of the image) demonstrate the accumulation of the deoxyribozyme in the cytoplasm (yellow/orange). (d) RT-PCR for CD44 (40 cycles) with RNA from RGCs after 24 h in culture. Reduction of CD44 mRNA after deoxyribozyme treatment against CD44 (CD44AS), as compared to control deoxyribozyme treatment (CD44MB), or to untreated cultures, was observed. (e) RT-PCR as in (d) for β-actin (25 cycles); bands demonstrate similar RNA template sample concentrations, and specificity of CD44 down-regulation from (d); MWM, molecular weight marker. (f–h) Immunostainings of RGCs to CD44 (red channel) and β-tubulin (green channel) after 24 h in culture demonstrate a reduction of CD44 after deoxyribozyme treatment to CD44 (f), compared to control deoxyribozyme treatment (g) and untreated (h) cultures. Note the strong positive CD44 staining at the short but multiple neurites of (g) and at the end-feet of the axonal growth cone in (h). (i and j) Immunoprecipitation of CD44 protein (i) or β-tubulin (j) from RGCs after treatment with 8 µmol/L of a deoxyribozyme targeting CD44 mRNA (CD33AS), 8 µmol/L control deoxyribozyme (CD44MB), or left untreated. U937 is a human leukemia monocyte lymphoma cell line, which was used as positive control, based on its known high expression of CD44. (k) Shows a Coomassie stained 4–15% SDS gel, which was loaded with crude protein extract from RGCs treated with 8 µmol/L of deoxyribozyme to CD44 (CD44AS), control deoxyribozyme (CD44MB), left untreated or as a positive control U937 cell extract. Note that the overall protein bands are equal in each lane which demonstrates that deoxyribozyme treatment has no general effect on the RGCs.

Fig. 4

CD44 down-regulation decreases RGC axon initiation response. (a) Confocal images of RGCs plated on osteopontin (OPN), laminin-111, or mixture of OPN/laminin-111 (0.5×) treated with a deoxyribozyme against CD44 (as, first column), control deoxyribozyme (mb, second column), or left untreated (third column) for 1 day and immunostained for β-tubulin. (b) Quantification of the percent of RGCs that initiated axons on each substrate with deoxyribozyme or control treatments. The use of percentages were justified by increased linearly (no-intercept model) between the number of axonal cells compared to total number of cells counted. *p < 0.05 by the General Linear Model procedure (SAS Software, Cary, NC, USA), and by subsequent pair-wise tests using Fisher’s LSD-test.

Fig. 5

(a) Measurement of axon length from RGCs growing on OPN, laminin-111, and a mixture of osteopontin (OPN)/laminin-111 (0.5×) after 24 h treated with the CD44 (CD44as) and control (CD44mb) deoxyribozymes (8 µmol/L) or left untreated. (b) Demonstrates the overall survival rate of RGCs under any of our experimental conditions (substrate or treatment). No statistical difference was observed between these various conditions.

Fig. 6

Immunostainings of osteopontin (OPN) and CD44 in the premature E15 RGC layer of 90 µm-thick vibratome sections. (a) Confocal image of 40× magnified RGC layer immunostained for OPN (green) and β-tubulin (red). (b) 100× magnification of the premature E15 ganglion cell layer (OPN, green; β-tubulin, red). Premature ganglion cells are marked by white arrows. (b′) ‘Profile’ from the laser scanning microscope (LSM) software along the red line shows OPN protein co-expressed with β-tubulin in one RGC. ‘Profile’ uses one image from a stack of images for analysis. (c) Confocal image of 40× magnified RGC layer immunostained for CD44 (green) and β-tubulin (red) in retinas of E15 embryos. (d) Higher magnification (100×) of the premature E15 ganglion cell layer (CD44, green; β-tubulin, red) marked by white arrows. (d′) ‘Profile’ from the LSM software package shows along the red line a clustered CD44 staining in areas of β-tubulin immunostaining in one RGC.

Fig. 7

Confocal images of sections from E15 eye and optic nerve, from E20 as well as P15 optic nerves stained with antibodies to CD44, osteopontin (OPN), and β-tubulin. (a) E15 retina and optic nerve stained with OPN (green) and β-tubulin (red). (a′) Is a higher magnification of the marked area in the optic nerve of (a). (a″) Demonstrates a co-existence of OPN and β-tubulin in the optic nerve by using the ortho-projection of the confocal imaging data coinciding with the blue line shown in (a). (b) E15 retina and optic nerve sections stained with CD44 (green) and β-tubulin (red). (b′) Is a higher magnification of the entry site of the retinal ganglion axons into the optic nerve, in which CD44 and β-tubulin co-exist. (b″) Shows the ortho-projection of the confocal imaging data from the blue line shown in (b), demonstrating that CD44 and β-tubulin are expressed throughout the optic nerve. (c–d) E20 and (e–f) P15 optic nerve section stained with β-tubulin (green) and OPN (red, c and e) or CD44 (red, d and f), respectively. (c′–f′) Are magnifications to demonstrate OPN and are not expressed from β-tubulin positive cells in the optic nerve at both ages.