BR A IN RE S EA RCH 1 1 07 ( 20 0 6 ) 2 4 –41
a v a i l a b l e a t w w w. s c i e n c e d i r e c t . c o m
w w w. e l s e v i e r. c o m / l o c a t e / b r a i n r e s
Research Report
Murine central and peripheral nervous system transcriptomes: Comparative gene expression Mark S. LeDoux⁎, Lijing Xu, Jianfeng Xiao, Brett Ferrell, Daniel L. Menkes, Ramin Homayouni University of Tennessee Health Science Center, Departments of Neurology and Anatomy and Neurobiology and Center of Genomics and Bioinformatics, 855 Monroe Avenue, Link Building-Suite 415, Memphis, TN 38163, USA
A R T I C LE I N FO
AB S T R A C T
Article history:
The central and peripheral nervous systems exhibit significant embryological,
Accepted 28 May 2006
morphological, and functional differences. Moreover, the pathology of most acquired and
Available online 7 July 2006
hereditary neurological diseases preferentially targets specific components of the nervous system. In order to test the hypothesis that central and peripheral neural transcriptomes
Keywords:
show fundamental quantitative differences, Affymetrix GeneChip® expression arrays were
Dorsal root ganglion
used to compare murine lumbar spinal cord (SC) and dorsal root ganglion (DRG) gene
Spinal cord
expression. As the crucial component of a novel technique to preserve RNA integrity, mice
Microarray
were perfusion-fixed with RNAlater™ before the SC and DRG were harvested. As per
RNA
Affymetrix terminology, a total of 111 transcripts were present (P) on all DRG arrays, absent
G-protein
(A) on all SC arrays, and demonstrated at least 10-fold greater expression in DRG than in SC.
Sodium channel
Conversely, a total of 112 transcripts were present on all SC arrays, absent on all DRG arrays,
Neuropathy
and showed at least 10-fold greater expression in SC than in DRG. For a subset of transcripts, quantitative real-time RT-PCR was used to corroborate and validate microarray results. Among those genes enriched in DRG, many belonged to a few distinct functional classes: Gprotein coupled receptor–protein signaling pathways, potassium transport, sodium transport, sensory perception, and cell-surface receptor-linked signal transduction. In contrast, genes associated with synaptic transmission, organic acid transport, neurotransmitter transport, and circulation were enriched in SC. Notably, the majority of genes causally associated with hereditary neuropathies were highly enriched in DRG. These differential neural gene expression profiles provide a robust framework for future molecular and genetic studies of neuropathy and SC diseases. © 2006 Elsevier B.V. All rights reserved.
1.
Introduction
Neuroanatomical localization of neural dysfunction is the cornerstone of clinical neurology and is governed by the topological and temporal specificity of gene expression. A common tenet of molecular medicine is that the propensity of
particular organs and tissues to exhibit disease manifestations is closely related to the cell-type specificity of protein expression. Along these lines, the cell bodies, axons, and myelin sheaths of peripheral nerves are either preferentially or selectively involved in a large number of acquired and hereditary diseases of the human nervous system.
⁎ Corresponding author. Fax: +1 901 448 7440. E-mail address:
[email protected] (M.S. LeDoux). 0006-8993/$ see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.brainres.2006.05.101
BR A IN RE S E A RCH 1 1 07 ( 20 0 6 ) 2 4 –4 1
The central nervous system (CNS) consists of the brain and spinal cord whereas the peripheral nervous system (PNS) includes 31 pairs of spinal nerves and 10 pairs of cranial nerves along with their associated ganglia, (N.B., the olfactory bulb and optic nerves are CNS projections). Spinal nerves contain general afferent and general efferent fibers. General afferent fibers originate from dorsal root ganglia (DRG) cells and are divided into two subtypes: general somatic afferent (GSA) and general visceral afferent (GVA). GSA fibers transduce and transmit exteroceptive, proprioceptive, and kinesthetic information. GVA fibers transduce and transmit information from interoceptors (visceroceptors) which are related to functions of the autonomic nervous system such as excretion, circulation, digestion, and respiration. In contrast to general afferent fibers, general efferent fibers arise from neurons within the CNS. Specifically, general somatic efferent (GSE) and visceral efferent (GVE) fibers originate from motor neurons in lamina IX of the anterior horn whereas preganglionic sympathetic or parasympathetic neurons are located in the intermediolateral region of the spinal cord (SC). Therefore, to study gene expression in the afferent division of the PNS, one must focus on the DRG. Moreover, based on the functional specia-
25
lizations of general afferent fibers, the DRG can be predicted to uniquely express genes forming the molecular machinery for mechanoreception, thermoreception, and nociception. Herein, we describe a novel method to facilitate RNA acquisition from mouse DRG and the analysis of this genetic material with high density oligonucleotide arrays to study differences in gene expression between the CNS and afferent division of the PNS. The resultant dataset provides fundamental support for a range of physiological, molecular, and clinical studies related to the PNS, particularly its sensory component. Importantly, this dataset can also be used to isolate candidate genes for mutation screening within hereditary neuropathy loci.
2.
Results
2.1.
RNA integrity and data reliability
An average of 13.6 μg (10.8 μg, 13.2 μg, and 16.8 μg) of total RNA was obtained for the 3 independent pooled (i.e., 48 ganglia) DRG samples and an average of 35.2 μg (36.0 μg, 37.2 μg, and
Fig. 1 – Electropherograms of total RNA generated by the Agilent Bioanalyzer 2100. (A) Composite of all electropherograms including a control lane (L). The spike controls appear as a green band. Lanes 1, 2, and 3 correspond to pooled DRG1, DRG2, and DRG3, respectively. Lanes 4, 5, and 6 correspond to pooled SC1, SC2, and SC3, respectively. (B) Individual electropherograms for each sample. The ordinate for each graph is fluorescence whereas the abscissa for each graph is time (s).
26
BR A IN RE S EA RCH 1 1 07 ( 20 0 6 ) 2 4 –41
32.4 μg) of total RNA was obtained for the 3 independent pooled (i.e., 12 lumbar cord segments) SC samples. As seen in Figs. 1A and B, 28S and 18S bands and peaks, respectively, were crisp for all DRG and SC RNA specimens. Furthermore, low baseline fluorescence was obtained for all samples. Small peaks migrating at approximately 25 s represent 5S rRNA, possibly in combination with tRNA. In Fig. 1B, note that the height of the spike control relative to the heights of the 5S, 18S, and 28S peaks shows moderate variability which is due to differences in RNA concentrations among the 6 samples. Total RNA was used to generate labeled cRNA probes which were hybridized to Mouse 430 2.0 GeneChip® arrays as described in Experimental procedures. Affymetrix Quality Control data (see Supplementary Material Online) were very similar for the DRG and SC arrays. The expression values for each GeneChip® were pre-processed and normalized. To evaluate whether there were any intensity dependent biases in the dataset, we generated MA plots, where the averaged intensity ratio (M) for each probe set in the DRG and SC samples was plotted against the product of their expression intensities (A). Using this method, we did not observe any intensity dependent effects in the normalized data (Fig. 2). Scatter plots of the normalized expression values between
replicate experiments demonstrated that the surgical and gene expression array procedures used in this study were very reproducible. The lowest R2 was 0.9786 and very few outliers were seen on the six plots. Analysis of Figs. 3D and F indicates only minor discordance of highly expressed genes in the SC2 sample in comparison with the SC1 and SC3 samples. Therefore, we conclude that overall reproducibility was similar among the DRG and SC arrays. For statistical measures of data reliability, we examined the distribution of P and Q values calculated from pairwise comparisons (two-sided t-tests) of the averaged expression values for each probe set in the DRG (N = 3) and SC (N = 3) samples. The Q value is derived from the distribution of P values for the experiment and is a measure of the expected number of false positives at a specific P value cutoff. In other words, if the difference in gene expression between two conditions is random, then the distribution of P values is expected to be flat (i.e., a null distribution; Fig. 4A). In our study, both P and Q value histograms were unimodal with exponential distributions (Figs. 4A and B). The marked concentration of genes that are differentially expressed with low P and Q values indicates that the expression differences between the DRG and SC presented here are reliable. We found 12,171 probe sets showing differential expression with P < 0.05. Since the estimated Q value for a P of 0.05 is 0.0248, approximately 302 of these 12,171 probe sets may be false positives (Fig. 4C). As illustrated with a volcano plot, similar numbers of genes were either enriched in DRG or SC across a broad range of P values (Fig. 5). Those genes showing striking enrichment in either DRG or SC are presented in Tables 1 and 2. QRT-PCR results for a subset of these highly enriched genes are provided in Table 3.
2.2.
Fig. 2 – MA plots before (A) and after (B) normalization of the microarray data.
Genes enriched in the DRG
Among the filtered collection of 22,946 probe sets, 3452 probe sets were expressed greater than 2-fold in DRG compared to SC. Among these, 3329 probe sets showed a P value
