| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
|
FJ
EXPRESS SUMMARY ARTICLE The Full-length version of this article is also available, published online June 13, 2005 as doi:10.1096/fj.04-3552fje. |
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Department of Molecular Medicine, Neurocrine Biosciences Inc., San Diego, California, USA
1Correspondence: E-mail: phevezi{at}neurocrine.com
SPECIFIC AIMS
Postmortem donors are a significant source of normal human tissues. We isolated total RNA from 
80 tissues from 4 human donors to evaluate RNA quality and to generate a body-wide index of gene expression (BIX). We used both a rat duodenum model system and the human tissue RNAs to determine how RNA quality affects gene expression data generated using Affymetrix Genechips. By including expression data from 250 of the human RNAs, we generated a BIX of gene expression for normal human tissues. We evaluated the value of the BIX by 1) determining the expression of 4 potential therapeutic targets: CCL27, GPR22, GPR113, and GPR128; and 2) defining a set of tissue-specific genes for the thymus.
PRINCIPAL FINDINGS
1. Evaluation of total RNAs isolated from donor tissues
Results showed tissue-specific variation in RNA quality. The highest quality RNA was isolated from CNS tissues while samples from gastrointestinal tissues exhibited significant RNA degradation. We were able to rank the 80 tissues for RNA quality using a semiquantitative scoring system. Donor-to-donor variations were also noted when RNAs derived from the same tissue were compared between the 4 donors.
2. Affymetrix oligonucleotide arrays generated gene expression data from both intact and degraded RNA samples
For rat duodenum RNA and human RNA from 4 representative organs (cerebellum, vestibular nuclei, skeletal muscle, and tongue), more than 80% of the probe sets called present in intact RNA are also called present in degraded RNA (Fig. 1
). By comparing data from intact and degraded RNA, we identified a subset of genes whose expression data are more sensitive to degradation and which we have designated "labile."
|
3. Two approaches determined whether the BIX accurately represented gene expression in normal human tissues
First, expression patterns of known tissue-specific genes were analyzed. Figure 2
shows profiles of individual gene expression for two representative genes: kallikrein 3 (prostate-specific antigen) (Fig. 2A
) and Ghrelin precursor (Fig. 2B
). The expression profiles of both genes clearly show robust expression at the expected sites with little or no expression elsewhere. Second, we identified genes specifically expressed in a given tissue. To exemplify this approach we chose the thymus as the test tissue. We identified 129 transcripts that were greater than 8-fold more abundant in the thymus samples than the mean level in the remaining samples. Thirty-nine of the top 50 genes are known thymus-specific genes. Six of the remaining 11 encode proteins of unknown function.
|
4. Identification of novel sites of expression for genes that encode potential drug targets
CCL27 is a skin-specific chemokine that has been implicated in the etiology of skin inflammation. We detected significant expression of CCL27 mRNA in the skin but also in nipple, a novel site of expression (Fig. 2C
). As orphan G-protein-coupled receptors, GPR22, 128, and 113 are all potentially interesting therapeutic targets. In addition to the previously reported expression in various CNS regions, we detected expression of GPR22 in the heart. Both GPR113 and 128 had been identified by searching EST sequences for novel GPCRs and predicted to have expression in the testis alone or testis and colon, respectively. However, we found significant GPR113 expression both in the kidney cortex (4 of 4) and medulla (4 of 4), as well as high levels of GPR128 transcript in both the liver (5 of 5) and small intestine (3 of 3) (Fig. 2E, F
).
CONCLUSIONS AND SIGNIFICANCE
Human postmortem tissues represent the most practical source for the study of gene expression in normal tissues. However, there is always a lag period of 3 h before collection of tissues, and in this time some RNA degradation may occur. A key issue inherent in microarray experiments in general is to determine minimum level of RNA quality that can be used to generate representative gene expression data. We have conducted a comprehensive survey of RNA stability in 80 human tissues across 4 donor bodies. By scoring each preparation using a numerical scale, we were able to establish a rank order of RNA postmortem stability for all 80 tissues across the 4 donor bodies. Gut tissues, including small and large intestine, contained the poorest quality RNA that was significantly degraded in most samples. The presence of normal gut flora (and hence RNases), the increased rate of tissue turnover, or the presence of digestive enzymes in the gut may account for reduced RNA stability in these tissues. In contrast, the CNS is one of the best protected regions of the body, contains tissues that exhibit very low levels of turnover and is a poor source of digestive enzymes. RNA samples from the CNS were of high quality and the RNA largely intact.
In addition to the tissue-specific variation, we also found evidence for donor-specific variation of RNA stability. One possible source of the inconsistency may be cause of death. There are several other possible sources of donor-specific variation. These include diet (especially for gut tissues), donor’s health status, and age.
We went on to measure gene expression in the RNA samples using Affymetrix microarrays. We found that, for this platform, we obtained comparable gene expression data from either intact or degraded RNA. In the rat duodenum model we used to further investigate this finding in a controlled system, we found 90% of the genes called present in intact RNA were also called present in the degraded sample. A similar analysis of human tissues gave a similar result (Fig. 1)
. An explanation for the tolerance of this platform to degraded RNA samples may lie in the nature of the Affymetrix GeneChip design, which is 3'-biased. For a transcript to be detected on the array, two features are required: the last 600 residues of each transcript must be present as it is these sequences that hybridize to the gene-specific oligonucleotides on the array, and these sequences must be contiguous with a short poly A tail (for labeling of transcripts). Thus, transcripts will be detected regardless of loss of upstream (5') or poly A sequences. However, as RNA degrades, there is a loss of detection especially for transcripts that are not abundant. The percent of positive calls for transcripts drops significantly for probe sets with an intensity (in intact RNA) below 50. Because many of the gut samples were significantly degraded, we supplemented the BIX with samples from commercial sources that contain high quality RNA. The majority of RNA samples obtained from other parts of the donor bodies were of sufficient quality to be used in GeneChip studies and constitute the bulk of the samples used to generate the BIX.
We evaluated the BIX in two ways. First, we checked the expression profiles for a number of genes with known expression patterns. Our findings are exemplified by two examples shown in Fig. 2
: kallikrein 3 and ghrelin precursor. Kallikrein 3 (prostate-specific antigen) is a prostate-specific protease that is used to detect and monitor prostate cancer growth. The expression profile clearly shows robust expression in the 4 prostate samples with little or no expression elsewhere (Fig. 2A
). We then sought to query genes that are expressed in the gastrointestinal tract because of its faster RNA degradation characteristics (Fig. 1A
). Ghrelin is a small peptide secreted by the fundus of the stomach that is believed to play a role in food intake control. As shown in Fig. 2B
, ghrelin precursor RNA was detected in all 8 samples from the cardiac and fundus regions of the stomach. As reported previously, we did not detect ghrelin RNA in the pyloric region of the stomach (0 of 4 samples).
We are able to detect novel sites of expression data for genes that encode potential drug targets: CCL27 (Fig. 2D-F
). We have confirmed the profiles of these genes using real-time quantitative PCR (data not shown).
An alternate approach was to focus on a given organ and identify genes that are specific to or strongly expressed in that tissue compared with other tissues. To exemplify this approach, we searched for genes with strong thymus expression compared with the rest of the BIX. Many previously characterized thymus-specific genes were identified. The chemokine CCL25 and its receptor CCR9 are both strongly expressed in the thymus. CCL25 is also expressed in the small intestine but at a much lower level. While these genes are well characterized, if we did not know that they constituted a ligand-receptor pair, the BIX data would have suggested a link. We are currently querying the body index for similar organ-specific relationships, especially in the CNS. We conclude that our body index is a powerful tool to uncover genes strongly associated with the physiology of different organs and cells.
In summary, we have ranked postmortem human tissues for RNA stability and identified a donor-specific component of RNA stability. We applied a powerful and commonly used molecular biological method, microarray analysis of gene expression, to generate a database that has many applications, including identifying tissue-specific genes and novel drug target discovery. Our findings have a direct effect on molecular medicine since a human gene expression index will allow us to better understand, at the molecular level, the makeup of normal human tissues. By comparing these data with expression profiles obtained from diseased tissues, we can identify genes involved in various pathophysiological processes.
|
FOOTNOTES
To read the full text of this article, go to http://www.fasebj.org/cgi/doi/10.1096/fj.04-3552fje;
This article has been cited by other articles:
|
|
 |
|
  J. C. Sluimer, N. Kisters, K. B. Cleutjens, O. L. Volger, A. J. Horrevoets, L. H. van den Akker, A.-P. J. Bijnens, and M. J. Daemen Dead or alive: gene expression profiles of advanced atherosclerotic plaques from autopsy and surgery Physiol Genomics, August 20, 2007; 30(3): 335 - 341. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Hever, R. B. Roth, P. Hevezi, M. E. Marin, J. A. Acosta, H. Acosta, J. Rojas, R. Herrera, D. Grigoriadis, E. White, et al. Human endometriosis is associated with plasma cells and overexpression of B lymphocyte stimulator PNAS, July 24, 2007; 104(30): 12451 - 12456. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Turchin, C. Z. Guo, G. K. Adler, V. Ricchiuti, I. S. Kohane, and G. H. Williams Effect of Acute Aldosterone Administration on Gene Expression Profile in the Heart Endocrinology, July 1, 2006; 147(7): 3183 - 3189. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
