Transcript Determination of the Neuroprotective Index for

ARVO Poster 3093/B6-46
Determination of the Neuroprotective Index for Neuroprotective Treatments
Based on a Mouse Model of Retinitis Pigmentosa
W.Raffelsberger1, R.Reddy1, F.Chalmel1, N.Wicker1, A.Legrand1, N.Chadderton2, P.Humphries2, J.Sahel3, T.Leveillard3 and O.Poch1
1 RetNet
Team VI : Laboratoire de Bioinformatique et Génomique Intégrative, Department of Structural Biology, IGBMC, 1 rue Laurent Fries, F-67404 Illkirch Cedex, France;
2 RetNet Team II : Trinity College,Department of Genetics, Smurfit Institute, Dublin, Ireland; 3 RetNet Team V: INSERM-U592, Institut de la Vision, Paris, France
Schematic Overview
Retinitis Pigmentosa (RP) is an inheritable degeneration of photoreceptors though two
subsequent steps characterized by i) the loss of rods in a cell autonomous manner
which is followed by ii) loss of cones in a non cell autonomous manner, leading to
complete blindness. The rd1 mouse serves as animal model for RP as both stages can
be observed during retinal degenration. We investigated the neuroprotective index of
neuroprotective substances in the mouse transcriptome during the first phase of RP in
rd1 mice.
Methods
Clustering of regulated genes
rd1 mice at PN15
+/- Neuroprotective Treatment (48h)
data used in clustering :
log Signal Intensity PN17 (untreated),
secure Gene Regulation (untreated PN15, PBS, GDNF, CNTF, Diltiazem versus PN17)
Isolation of RNA form retina
log Sign Intensity /
secure Gene Regulation
Purpose
Preparation of Biological Duplicate Samples, QC
Affymetrix Transcription Profiling
The neuroprotective agents GDNF, CNTF and Diltiazem as well as mock-controls were
injected to rd1 mice at 15 days postnatal and RNA was isolated from mouse retina 48 h
later. Duplicate samples were subjected to Affymetrix transcription profiling experiments
and that were analyzed using a novel bioinformatics protocol based on the assessment
of a quality indexes. The degree of homogeneity for the probes defining each
microarray’s probe set and the agreement of the biological duplicates were analyzed by
an automated protocol measuring each gene’s apparent quality assessment index.
These indexes were then considerated in the subsequent clustering analysis based on
the Mixture Models algorithm (combined with AIC for the determination of the number of
clusters) to characterize regulated genes representing the molecular targets of RP and
measuring the neuroprotective index.
MG U74-2 arrays (36k transcripts)
Quality Index Assessment
of transcription profiling data
Clustering of co-regulated Genes
by Mixture Models & AIC (for determination of # of clusters)
Search for Enrichment of Specific Functions
using Cluspack and RetScope/GOAnno annotation protocol
Conclusions from Clustering
◦ apr. 30 000 genes don’t change expression levels upon treatment with GDNF, CNTF or Diltiazem
◦ GDNF : many genes with weaker upregulation (148)
◦ 11 genes are stronger upregulated both by PBS and CNTF
◦ apr. 600 genes show very small changes at their expression levels
Determination of Neuroprotective Index
GDNF induces histological and functional protection of
rod photoreceptors in the rd1 mouse
Conclusions
Quality Index Assessment
The determination of a neuroprotective index across genes and neuroprotective agents
combined with a very low risk of false positives was made achieved through clustering
of transcription profiles while considering the apparent quality assessment index.
for Affymetrix Transcriptomics Data
PBS
GDNF
Homogeneity of Redundancy
within ProbeSet
Agreement of Replicates
Biological Duplicates on
independent Affymetrix Arrays
This project is supported though the European Retinal Research Training Network
‘RETNET’, MRTN-CT-2003-504003.
References:
Frasson et al., 1999
Quality Index Unbiased to Signal Intensity
Integrated & Automated Model
Results
This procedure allowed to significantly reduce the amount of false positives among the
genes reported, which is especially important for low level expressed genes as they are
typically subject to less precise measurements. These results were further examined for
enrichment of functional ontologies and signalling cascades. Finally, we defined and
determined the neuroprotective index considering for each i) treatments and ii) all
transcripts.
Improvement of duplicate agreement as revealed by quality indexduplicate :
MAS 4
dChips 1.3
improved duplicate
agreement
| norm Difference |
Frasson M, Sahel JA, Fabre M, Simonutti M, Dreyfus H, Picaud S. Retinitis pigmentosa: rod photo-receptor rescue by a
calcium-channel blocker in the rd mouse. Nat Med. 1999 Oct;5(10):1183-7
Yeung KY, Fraley C, Murua A, Raftery AE, Ruzzo WL.
Model-based clustering and data transformations for gene expression data. Bioinformatics. 2001 Oct;17(10):977-87
Raffelsberger W, Dembele D, Neubauer MG, Gottardis MM, Gronemeyer H.
Quality indicators increase the reliability of microarray data. Genomics. 2002 Oct;80(4):385-94
Wicker N, Dembele D, Raffelsberger W, Poch O. Density of points clustering, application to transcriptomic data analysis.
Nucleic Acids Res. 2002 Sep 15;30(18):3992-4000
Leveillard T, Mohand-Said S, Lorentz O, Hicks D, Fintz AC, Clerin E, Simonutti M, Forster V, Cavusoglu N, Chalmel F, DolleP,
Poch O, Lambrou G, Sahel JA. Identification and characterization of rod-derived cone viability factor. Nat Genet. 2004
Jul;36(7):755-9
Chalmel F, Chalmel F, Lardenois A, Thompson JD, Muller J, Sahel JA, Leveillard T, Poch O.
GOAnno: GO annotation based on multiple alignment. Bioinformatics. 2005 Jan 12
Dortet-Bernadet JL and Wicker N.
Model based clustering on the unit sphere of standardized gene experssion profiles. Biostatistics, submitted
Raffelsberger W, Reddy RK, Legrand A, Wicker N, Poch O. manuscript in preparation