1 Center for Animal Functional Genomics, Michigan State University, East Lansing, Michigan 48824
3 Animal Behavior and Welfare Group, Department of Animal Science, Michigan State University, East Lansing, Michigan 48824
2 Texas Instruments, Stafford, Texas 77477
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
pig; behavior; swine; CNS; model system
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
A domestic animal system interesting for both its applicability to agriculture and its potential use as a human model is porcine brain. Pigs are widely considered to be a useful nonprimate model for human medical research (8, 24). As early as 1972, the advantages of Sus scrofa domestica as an ideal experimental animal in human medical research was reviewed (8). Similarities between pig and human brains have also been well documented (7). In fact, gray/white matter distribution, brain morphology, and overall shape and gyral pattern of the piglet brain is similar to that of humans (7). Pig brain models have been used to study a wide range of neurological diseases, particularly diffuse brain injury and ischemia (13, 26). In addition, pig brain models have been used to study Parkinsons disease (19), Alzheimers disease in relation to brain injury (21), and the effects of hypothermia on the brain (23).
In addition to its importance as a model system, swine are also widely important in agriculture. Increased emphasis on understanding animal behavior and welfare in production agriculture has made it extremely important to examine how common agricultural practices effect cognitive development and welfare of the piglet (11, 12, 17, 27).
Sequencing of cDNA clones from libraries of specific tissue and cell types to generate expressed sequence tags (ESTs) has proved to be a rapid and efficient means of gene discovery (1). Creation and characterization of a porcine brain cDNA library (PBL) for EST discovery and microarray construction would aid in the applications presented above and allow for expansion of functional genomics to pig brain models of human conditions.
To date, only two publicly available porcine cDNA microarrays exist, both constructed from skeletal muscle cDNA libraries (28, 3). In addition, there are an increasing number of swine ESTs generated from cDNA libraries made from various tissues available in public databases (28, 10, 5, 22, 25). However, no significant number of ESTs have been generated from libraries where porcine brain tissue is highly represented.
To address this shortfall and facilitate use of the pig brain as a model system, we describe construction of a sequence-tagged, normalized PBL, generation of 965 EST sequences from this library, and classification of these sequences in a publicly available web-accessible database (http://nbfgc.msu.edu). In addition, we describe development of a first-generation cDNA microarray representing each unique EST clone derived from the normalized PBL.
![]() |
MATERIALS AND METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Construction of cDNA library.
Total RNA was purified from all tissues using TRIzol reagent essentially as recommended by the manufacturer (Invitrogen Life Technologies, Carlsbad, CA). Total RNA from each tissue was pooled and mRNA isolated using a PolyATtract Small Scale mRNA Isolation kit (Promega, Madison, WI). A commercial cDNA library synthesis system (Invitrogen Life Technologies) was used for construction of the library. First-strand cDNA synthesis was performed using the oligonucleotide 5'-GACTAGTTCTAGATCGCGAGCGGCCGCTTAAGG(T)15-3', containing a NotI site (underlined) and a 6-bp library identifier tag (TTAAGG, bold type). Second-strand cDNA synthesis was performed using first-strand cDNA as template. Directional cloning was facilitated by addition of a SalI adapter to the 5' end of cDNAs. Final cDNAs were digested with SalI and NotI and subjected to column chromatography with a cDNA size fractionation column (Invitrogen Life Technologies) in TEN buffer [10 mM Tris-HCl (pH 7.5), 0.1 mM EDTA, 25 mM NaCl] to remove residual SalI adapters, digested NotI fragments, and small cDNA products. Double-stranded cDNA fragments were cloned directionally into SalI and NotI sites of plasmid vector pSPORT1 (Invitrogen Life Technologies), and a portion of the library was transformed into Escherichia coli DH10B cells by electroporation. The library was then amplified in a semisolid medium to minimize biased growth of primary transformants (16).
Normalization of cDNA library.
Library normalization was performed using a modification of the method described in Bonaldo et al. (4). This technique is used to reduce prevalence of highly represented clones and is based on the different reassociation kinetics for cDNA molecules of different relative abundance (4). Double-stranded plasmid library DNA was converted to single-stranded circles by the nicking action of gene II protein and the digestive action of exonuclease III (Invitrogen Life Technologies). Digested samples were purified by hydroxyl apatite (HAP) chromatography to remove any remaining double-stranded plasmids. An aliquot of purified single-stranded plasmid DNA was used to generate PCR-amplified cDNA inserts using primers designed to flank insert regions of pSPORT1 plasmid. Amplified inserts were purified using QIAquick PCR purification kit (Qiagen, Valencia, CA) and ethanol precipitated. Approximately 3 µg of purified PCR products was then used to hybridize with 150 ng of single-stranded plasmids prepared earlier. Normal hybridization kinetics dictates that the most prevalent sequences will find a complementary strand more rapidly than low-abundance sequences, becoming double-stranded. Double-stranded cDNA was then removed by HAP chromatography, and single-stranded portions, representing low-abundance transcripts, were isolated. Single-stranded cDNAs were converted to double-stranded plasmids by primer extension using Sequenase (Invitrogen Life Technologies) and an oligonucleotide complementary to the single-stranded plasmid DNA. Resulting plasmids were electroporated into E. coli DH10B cells, generating the normalized PBL.
DNA sequencing and EST analysis.
Transformed cells from both the pre- and postnormalized libraries were grown on Luria broth (LB) agar plates containing 50 µg/ml ampicillin. Colonies were randomly picked, transferred to 96-well plates, and grown overnight in LB containing 50 µg/ml ampicillin. Sequencing was performed from rolling circle amplification (RCA) products via a commercial sequencing company (Genome Therapeutics, Ann Arbor, MI) using an ABI 3700 DNA analyzer with M13 primers designed to flank the 5' end of plasmid inserts. Resulting sequences were screened for quality using "phred" (9), and vector sequences were removed using "cross_ match" (9), with the recommended parameters for vector screening (minmatch = 10, minscore = 20). Sequences larger than 150 bp were screened from the resulting list using a custom Perl script. A bioinformatics server running on a UNIX platform is provided for web-accessible storage of EST data (http://nbfgc.msu.edu). The 965 EST sequences from the normalized library were submitted in batch to GenBank dbEST (accession numbers CD571919 to CD572883). Since there is no publicly available tool for simple batch EST submission to GenBank, a Perl script was developed to accomplish this task and is available at our web site (http://nbfgc.msu.edu).
Construction of porcine brain cDNA microarray.
A set of unique clones from the PBL (877) were reracked into a new set of 96-well plates and inserts amplified by PCR, using 1 µl of RCA product as template in PCR master mix [20 µM each dNTP, 0.5 µM forward primer, 0.5 µM reverse primer, 200 mM Tris-HCl (pH 8.4), 500 mM KCl, and 2 U of Taq DNA polymerase]. Resulting PCR amplicons were purified using a Millipore Multiscreen PCR 96-well purification system (Millipore, Bedford, MA) on a Te-Vac vacuum manifold using a Tecan Genesis 150 liquid handling robot (Tecan US, Research Triangle Park, NC). A vacuum of 700 mBar was applied for 10 min to remove buffer and bind amplicon DNA. Final insert amplicons were suspended in 50% dimethyl sulfoxide (DMSO) for microarray printing. Approximately 2 µl of each purified insert amplicon was separated on Invitrogen E-gel 96 1% agarose gels (Invitrogen Life Technologies) to ensure that amplicons representing each clone were represented on the microarray. Purified amplicons were then transferred to 384-well source plates prior to microarray spotting.
cDNA microarray spotting was performed using a GeneTAC G3 arraying robot (Genomic Solutions, Ann Arbor, MI) equipped with a 48-pin head, with each pin having a nominal end diameter of 200 µm. The microarray was designed for a 9 x 9 spot pattern, with each clone insert spotted in triplicate. A synthetic lambda Q-gene was spotted in the corners (4 per patch) of each patch to serve as a positive hybridization control. Forty-eight glyceraldehyde-3-phosphate dehydrogenase (GAPDH) control spots and 48 ß-actin control spots were also placed randomly in the patches as positive controls, and numerous DMSO spots (average 5 per patch) and blanks (4 per patch) were included to serve as negative controls.
Southern blot analysis.
To determine the affect of normalization on the relative frequency of highly abundant and rarely expressed transcripts, a Southern blot was performed. Approximately 0.5-µg aliquots of plasmid DNA from both the original and normalized library were separately digested with Mlu1, separated on 1% agarose gels, and transferred to a Zeta-probe membrane (Bio-Rad, Hercules, CA). The membranes were separately hybridized at 68°C in PerfectHyb solution (Sigma Chemical, St. Louis, MO) with 32P-labeled PCR-amplified cDNAs, labeled using a commercial random primers labeling system (Invitrogen Life Technologies). Probes for the highly abundant transcripts GAPDH and ß-actin were amplified from cDNA clones using primers recognizing vector sequences flanking the inserts in pSPORT1. The probe for brain-derived neurotrophic factor (BDNF), representing a low-abundance transcript, was PCR amplified [20 µM each dNTP, 0.5 µM forward primer, 0.5 µM reverse primer, 200 mM Tris-HCl (pH 8.4), 500 mM KCl, and 2 U of Taq DNA polymerase] from 500 ng of cDNA template prepared from a reverse transcriptase reaction with whole porcine brain mRNA. The specific primers BDNF-forward (5'-ACATGTATACGTCCCGAGTC-3') and BDNF-reverse (5'-ACATACGACTGGGTAGTTCG-3') were used in this reaction to produce a probe with a size of 382 bp. Membranes were washed twice at 65°C in 2x saline sodium citrate (SSC)/0.1% SDS and four times at 65°C in 0.2x SSC/0.1% SDS following 24-h hybridization. Washed membranes were sealed in plastic film and exposed to Hyperfilm MP (Amersham Pharmacia Biotech, Piscataway, NJ) at -80°C to visualize bands.
Microarray experiments: RNA extraction, labeling, and hybridization.
Total RNA was extracted from 0.3 g of homogenized whole brain tissue from a 23-day old female pig using TRIzol (Invitrogen Life Technologies). Quantity and quality of extracted total RNA was estimated by UV spectrophotometry and electrophoresis on 1% native agarose gels. For hybridization to the microarray, 18 µg of total RNA was used as a template for cDNA production in a reverse transcription reaction incorporating a modified dUTP into the cDNA (BD Atlas PowerScript Fluorescent labeling kit; BD Biosciences, Alameda, CA). Oligo(dT)18 was used as a primer, and 0.6 ng of synthetic lambda Q-gene RNA containing an engineered poly-A tail was spiked into each reaction as a positive control for cDNA synthesis and hybridization. After first-strand cDNA synthesis, the sample was split into two aliquots and labeled according to instructions for the BD Atlas PowerScript Fluorescent labeling kit (BD Biosciences) in a two-step process, with one aliquot labeled with Cy3 and the other with Cy5 (Amersham Biosciences, Piscataway, NJ). Dye-labeled products were purified using QIAquick spin columns (Qiagen), recombined, and concentrated to 10 µl using Microcon 30 spin concentrators (Millipore, Bedford, MA). SlideHyb buffer no. 3 (Ambion, Austin, TX) was added to the probe to a final volume of 110 µl, and probes were incubated at 70°C in a water bath for 5 min. Microarray hybridizations were performed using an 18-h step-down procedure (3 h at 60°C, then 3 h at 55°C, and finally 12 h at 50°C) in a GeneTAC HybStation (Genomic Solutions). Following hybridizations, the station applies three washes, two with medium stringency buffer and one with high-stringency buffer (Genomic Solutions). Slides were finally rinsed briefly at room temperature in 2x SSC and once in double-distilled H2O. Washed microarrays were dried by centrifugation at 1,000 g for 5 min in a cushioned 50-ml conical centrifugation tube. Final microarrays were scanned using a GeneTAC LS IV microarray scanner and software (Genomic Solutions). GeneTAC Integrator 4.0 software (Genomic Solutions) was then used to process images and create spot intensity reports. Final intensity reports were retrieved as raw spot intensities in comma-separated value files, compatible with Microsoft Excel and SAS analysis programs. Data for microarray experiments used in this report can be found at the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO), with series accession number GSE456.
Microarray data analysis.
Microarray data was normalized using the scatter plot smoother LOESS (6) from the statistical package SAS (PROC LOESS) according to the method of Yang et al. (29). Briefly, the normalization process is as follows, M-A plots are constructed for each slide, where log-intensity ratios M = log(Cy3/Cy5) = [logCy3 - logCy5] are plotted against the mean log-intensity A = [(logCy3 + logCy5)/2] as described by Yang et al. (29). Normalization is then performed considering a robust local regression technique (Cleveland and Gross), using the LOESS procedure of SAS (SAS Institute, Cary, IN). Efficiency of LOESS (also known as "LOWESS") normalization is assessed by monitoring M-A plots before and after normalization. The normalized data is then back transformed prior to further statistical analysis using the formula
![]() |
LOESS normalized array data was imported into Microsoft Excel for further analysis. After back-transforming normalized intensity values as above, the median negative intensity for each dye on a microarray was subtracted from the respective normalized spot intensity values. Residual intensity values were loge transformed prior to further analysis. The loge-transformed and negative subtracted values were used to calculate a mean loge expression difference (Cy3 - Cy5) for each gene. Since each gene is represented in triplicate, a standard error, a t-statistic, and t-distribution (P value) can be calculated for each gene represented on the array. This P value yields a confidence interval that refers only to measurement error for any given gene on the array. The antilog of the mean loge expression difference for an individual gene on the array yields the approximate fold change in expression for this gene.
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The effectiveness of library normalization was initially evaluated by DNA sequence analysis. After quality screening and vector trimming the sequences, BLASTN results of ESTs generated from the original (111) and normalized libraries (867) were used to estimate the number of abundant genes represented before and after normalization. As depicted in Table 1, 24.2% (26/111) of the original library consisted of clones considered to represent abundant transcripts, whereas 4.8% (42/867) of clones for the normalized library represented abundant transcripts. These data indicate that frequency of abundant clones had been successfully reduced in the normalized library.
|
|
Analysis of expressed sequence tags.
To date, a total of 867 high-quality EST sequences have been successfully generated from the 5' end of isolated cDNA clones from the normalized PBL. These sequences were added to 111 EST sequences from the original library, and cluster analysis was performed on all sequences, creating 71 clusters and 736 singletons for a total of 807 unique sequences represented by clones isolated to date. All 867 EST sequences from the normalized library and the 27 unique sequences from the original library are available in our publicly available database under the PBL library designation. The database system also includes BLASTN and BLASTX searches on each of the ESTs against the public GenBank (nonredundant, nr) database.
BLASTN results were analyzed and 558 of the 807 unique ESTs (69.1%) have significant similarity (BLASTN matches with E value <10-14) to known genes in the public GenBank (nr) database. BLASTN and BLASTX hits for each EST were extracted using a Perl script and used to generate putative identifications for each clone. The most significant BLASTX or nongenomic BLASTN hit was derived for each clone using a series of custom Perl scripts. This information was then processed by GeneSpring software (Silicon Genetics, Redwood City, CA), using the Build Simplified Ontology function, which is based on the Gene Ontology Consortium classifications (2). Ontology classifications for 206 of the 558 (36.9%) ESTs with significant similarity to known genes were extrapolated (Fig. 2). Furthermore, the web interface of our database allows online BLAST homology to be searched directly by keywords. Using this function, an additional 48 clones were placed into CNS ontology groups (Table 2 and Fig. 3) including neurotransmitter receptors, neurotransmitter transporters, hormones and hormone receptors, Na+/K+ pumps, Na+, K+, and Ca2+ channels, neuron morphology proteins, synapse-related proteins, and other neuron-specific proteins. This information provides a better outline of the PBL library as a suitable source of brain-specific transcripts and gene discovery for the porcine CNS.
|
|
|
A cDNA microarray was then printed with 877 amplicons representing unique clones, spotted in triplicate with several positive and negative controls, as outlined in MATERIALS AND METHODS.
Prior to use of the PBL cDNA microarray in assessing expression patterns of genes represented in porcine brain, it was important to first determine the potential for false-positive gene expression changes. To accomplish this, a series of two microarrays using differentially labeled aliquots of cDNA derived from RNA representing the same homogenized piglet brain (23 day old female) were performed. Data were analyzed as described in MATERIALS AND METHODS and used to establish expected PBL microarray backgrounds (false-positive rate) under various selection criteria (Fig. 4). Our results demonstrate that without addition of a fold-change cutoff, false-positive rates follow statistical expectations (i.e., 5% of genes represented on the microarray at P < 0.05). However, addition of various fold-change criteria dramatically reduced the observed rate of false positives. Inclusion of a minimum fold change criteria of 2.0 essentially eliminated all false positives at confidence intervals below 0.05 (Fig. 4). This result demonstrated that design of the PBL microarray, combined with LOESS normalization and negative subtraction, yields a low rate of false positives, defined as observed significant gene expression changes when none are actually present.
|
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
A cDNA microarray resource for studies of porcine brain and CNS physiology was developed with 877 unique clones from the PBL. The PBL cDNA microarray was tested by a screen for false positives. Results presented in this report indicate that inclusion of a minimum fold change criteria of 2.0 essentially eliminated all false positives at confidence intervals below 0.05.
The porcine model should help further neuroscientific research, as its applicability as a human model system, as well as its agricultural significance, is well documented (7, 1113, 17, 19, 21, 23, 26, 27). Plans to use the PBL as a model system to study behavior, learning and memory, and neurological diseases are ongoing. Our hopes are that spatial gene expression imaging of the brain in a model organism will give better comprehension of the logic of the genome and an important starting point for future research.
![]() |
ACKNOWLEDGMENTS |
---|
Generous financial support for this work was provided by the Michigan Agriculture Experiment Station and the Michigan State University Office of the Vice President for Research and Graduate Studies. W. Nobis is a recipient of the Seevers Award from the Michigan State University Honors College and College of Agriculture and Natural Resources and also the Howard Hughes Undergraduate Research Program and thanks both programs for support.
![]() |
FOOTNOTES |
---|
Address for reprint requests and other correspondence: P. M. Coussens, 1205H Anthony Hall, Center for Animal Functional Genomics, Michigan State Univ., East Lansing, MI 48824 (E-mail: coussens{at}msu.edu).
10.1152/physiolgenomics.00099.2003.
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|