Development of a porcine brain cDNA library, EST database, and microarray resource

William Nobis1, Xiaoning Ren1, Steven P. Suchyta1, Thomas R. Suchyta2, Adroaldo J. Zanella3 and Paul M. Coussens1

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
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Recent developments in expressed sequence tag (EST) and cDNA microarray technology have had a dramatic impact on the ability of scientists to study responses of thousands of genes to internal and external stimuli. In neurobiology, studies of the human brain have been expanding rapidly by use of functional genomics techniques. To enhance these studies and allow use of a porcine brain model, a normalized porcine brain cDNA library (PBL) has been generated and used as a base for EST discovery and microarray generation. In this report, we discuss initial sequence analysis of 965 clones from this resource. Our data revealed that library normalization successfully reduced the number of clones representing highly abundant cDNA species and overall clone redundancy. Cluster analysis revealed over 800 unique cDNA species representing a redundancy rate for the normalized library of 6.9% compared with 29.4% before normalization. Sequence information, BLAST results, and TIGR cluster matches for these ESTs are publicly available via a web-accessible database (http://nbfgc.msu.edu). A cDNA microarray was created using 877 unique porcine brain EST amplicons spotted in triplicate on glass slides. This microarray was assessed by performing a series of experiments designed to test hybridization efficiency and false-positive rate. Our results indicate that the PBL cDNA microarray is a robust tool for studies of brain gene expression using swine as a model system.

pig; behavior; swine; CNS; model system


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
MICROARRAY AND GENE EXPRESSION profiling has become an important tool to illuminate complex biological processes, such as those comprising responses in the central nervous system (CNS). Studies of brain function and learning are expanding rapidly using microarray technologies, functional genomics, and functional molecular imaging methods (14, 15). However, these resources have been largely limited to human and rodent systems, where it may not be possible or ethical to conduct many studies related to disease, behavior, learning and memory, and brain function, among others.

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 Parkinson’s disease (19), Alzheimer’s 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
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Tissue collection.
Brain and CNS tissues were collected from six pigs (commercial breed Yorkshire and Landrace cross) for use in cDNA library construction. The frontal cortex, hippocampus, hypothalamus, parietal cortex, amygdala, cerebellum, and spinal cord were collected from pigs at 10 days of age, a mature boar, a postpubertal gilt, and a lactating sow. Additionally, the CNS of a fetal pig, the eye of a 5-wk-old pig, and the nucleus accumbens of a lactating sow were collected. All tissue was snap-frozen in liquid nitrogen and stored at -80°C until use.

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

where logCy3* and logCy5* are normalized log intensities. Here, M* = M - M^ and represents each of the normalized M values, with M^ being the LOESS predicted value for each spot.

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
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Evaluation of porcine brain cDNA library normalization.
The original PBL was normalized as described in MATERIALS AND METHODS. To assess library quality, in terms of average insert size, restriction enzyme analysis was performed on 88 randomly selected clones with Mlu1 enzyme. Digested cDNA clones were electrophoresed on agarose gels, and insert size for each clone was estimated by reference to a 1-kb ladder (Invitrogen Life Technologies) run on the same gel. Results of these analyses indicated that all 88 randomly picked clones had inserts, with an average insert size of 370 bp and a range of 150 bp to 1.2 kb.

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.


View this table:
[in this window]
[in a new window]
 
Table 1. Comparison of abundant cDNA clones present in the porcine brain library both before and after normalization based on the BLAST results of the 5' EST sequences generated

 
The efficiency of normalization was further assessed by Southern blot analysis comparing the relative abundance of clone inserts representing two highly expressed transcripts and one rarely expressed transcript (18) in the original and normalized libraries. As shown in Fig. 1, the relative abundance of clone inserts representing GAPDH and ß-actin were dramatically decreased in the normalized library compared with the original library, whereas relative abundance of clone inserts representing the rare transcript BDNF was enhanced in the normalized library compared with the original library. These data indicate that the normalization process successfully reduced the relative number of clones representing highly abundant cDNA species (GAPDH and ß-actin) while not affecting or enhancing the relative abundance of clones representing rare species.



View larger version (59K):
[in this window]
[in a new window]
 
Fig. 1. Southern blot analysis of the relative frequencies of cDNA species in the porcine brain library (PBL). Plasmid DNA from the original library (lane 1) and the normalized library (lane 2) was digested with Mlu1, separated on a 1% agarose gel, and blotted on a nylon membrane. PCR-amplified cDNAs for highly abundant transcripts ß-actin and GAPDH and lower expressed transcript brain-derived neurotrophic factor (BDNF) were labeled with 32P and hybridized to the membranes overnight. The vector sequences in the GAPDH and ß-actin probes are responsible for the intense vector signal in these autoradiograms.

 
ESTs from the original and normalized library were further characterized by cluster analysis using stackPACK software (Electric Genetics, Cape Town, South Africa) (20) to calculate redundancy rate. The redundancy rate of the original library was calculated to be 29.4%, compared with 6.9% for the normalized library, providing further evidence that library normalization was successful. This interpretation is subject to the caveat that actual redundancy rates could be higher if some nonoverlapping ESTs actually represent different regions of the same transcript, although this is unlikely since all ESTs were sequences from the 5' end.

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.



View larger version (23K):
[in this window]
[in a new window]
 
Fig. 2. A pie chart representing the distribution of functional classifications of 206 PBL expressed sequence tags (ESTs) based on the Gene Ontology assignments generated by GeneSpring software.

 

View this table:
[in this window]
[in a new window]
 
Table 2. ESTs homologous to gene products related to the central nervous system, including morphology, development, and signaling

 


View larger version (17K):
[in this window]
[in a new window]
 
Fig. 3. A pie chart representing the distribution of functional classifications of 48 PBL ESTs, specifically grouped into central nervous system-specific classifications.

 
Pig brain cDNA microarray.
In addition to 807 unique clones isolated from the PBL, brain/CNS-specific clones identified by 3'-end-specific tag sequences from a pooled pig cDNA library also created at the Center for Animal Functional Genomics at Michigan State University [mixed pig library (MPL), dbEST Library ID 12217] were electronically identified for possible addition to the PBL cDNA microarray. A Perl script was designed to scan for the brain/CNS-specific tissue tag (TTAAGG) followed by a poly-A tail from the MPL EST sequences. Once those sequences were identified, a cluster analysis was performed with stackPACK software (20) to determine unique sequences in the MPL, not represented in clones already isolated from the PBL. In total, 71 additional clones determined to be unique were added to the PBL library, and the corresponding sequences were added to the PBL database. This process yielded a total of 877 unique EST sequences for application to a PBL cDNA microarray.

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.



View larger version (15K):
[in this window]
[in a new window]
 
Fig. 4. This graph represents the number of significant genes (y-axis) at different fold change levels (x-axis). Different P values are represented by the different lines.

 

    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
A high-quality normalized cDNA library (PBL) representing transcripts from porcine brain and CNS has been developed. Cluster analysis of ESTs from this library reveal a low redundancy rate of 6.9%. All 965 EST sequences discussed in this report, including 877 unique EST clones, have been deposited in GenBank (accession numbers CD571919 to CD572883). The sequences are also available on a custom web-accessible resource (http://nbfgc.msu.edu), which includes BLASTN and BLASTX results, TIGR cluster number if available, and original clone sequence data. Further analysis of PBL ESTs indicated that 558 have strong similarity to known genes, 206 of which have been defined ontologically (Fig. 2). Another subset of ESTs were grouped into CNS classifications, with 48 brain-specific genes identified (Table 2).

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
 
We acknowledge the superb technical assistance of Dr. Jianbo Yao, Sue S. Sipkovsky, Brian Tooker, Chris Colvin, Matthew Coussens, and Rachael Kruska of the Center for Animal Functional Genomics. We also acknowledge the excellent support of Peter Saama, in maintaining the Center for Animal Functional Genomics web site and databases, and Dr. Catherine Ernst, Adriana Silveira de Souza, and Jarno Jansen, for assistance with porcine tissue collection.

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
 
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).

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
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 

  1. Adams MD, Kelley JM, Gocayne JD, Dubnick M, Polmeropoulos MH, Xia H, Merril CR, Wu A, Olde B, and Moreno RF. Complementary DNA sequencing: expressed sequence tags and human genome project. Science 252: 1651–1656, 1991.[ISI][Medline]
  2. Ashburner M and Lews S. On ontologies for biologists: the Gene Ontology—untangling the web. Novartis Found Symp 247: 66–80, 84–90, 244–253, 2002.[ISI]
  3. Bai Q, McGillivray C, Costa N, Dornan S, Evans G, Stear MJ, and Chang KC. Development of a porcine skeletal muscle cDNA microarray: analysis of differential transcript expression in phenotypically distinct muscles. BMC Genomics 4: 8, 2003.[CrossRef][Medline]
  4. Bonaldo MF, Lennon G, and Soares MB. Normalization and subtraction: two approaches to facilitate gene discovery. Genome Res 6: 791–806, 1996.[Abstract]
  5. Caetano AR, Johnson RK, and Pomp D. Generation and sequence characterization of a normalized cDNA library from swine ovarian follicles. Mamm Genome 14: 65–70, 2003.[CrossRef][ISI][Medline]
  6. Cleveland WS and Gross E. Computational methods for local regression. Stat Comput 1: 47–62, 1991.
  7. Dickerson JWT and Dobbing J. Prenatal and postnatal growth and development of the central nervous system of the pig. Proc R Soc Lond 166: 384–395, 1967.[ISI]
  8. Douglas WR. Of pigs and men in research: a review of applications and analogies of the pig, Sus scrofa, in human medical research. Space Life Sci 3: 226–234, 1972.[ISI][Medline]
  9. Ewing B, Hillier L, Wendl MC, and Green P. Base-calling of automated sequencer traces using phred. I. Accuracy assessment. Genome Res 8: 175–185, 1998.[Abstract/Free Full Text]
  10. Fahrenkrug SC, Smith TPL, Freking BA, Cho J, White J, Vallet J, Wise T, Rohrer G, Pertea G, Sultana R, Quakenbush J, and Keele JW. Porcine gene discovery by normalized cDNA-library sequencing and EST cluster assembly. Mamm Genome 13: 475–478, 2002.[CrossRef][ISI][Medline]
  11. Gaines AM, Carroll JA, Yi GF, Allee GL, and Zannelli ME. Effect of menhaden fish oil supplementation and lipopolysaccharide exposure on nursery pigs. Domest Anim Endocrinol 24: 353–365, 2003.[CrossRef][ISI][Medline]
  12. Gonyou HW, Beltranena E, Whittington DL, and Patinence JF. The behaviour of pigs weaned at 12 and 21 days of age from weaning to market. Can J Anim Sci 78: 517–523, 1998.[ISI]
  13. Grate LL, Golden JA, Hoopes PJ, Hunter JV, and Dunhaime AC. Traumatic brain injury in piglets of different ages: techniques for lesion analysis using histology and magnetic resonance imaging. J Neurosci 123: 201–206, 2003.
  14. Hariri AR and Weinberger DR. Imaging genomics. Br Med Bull 65: 259–270, 2003.[Abstract/Free Full Text]
  15. Hasenkamp W and Hemby SE. Functional genomics and psychiatric illness. Prog Brain Res 138: 375–393, 2002.[ISI][Medline]
  16. Kriegler M. Gene Transfer and Expression: A Laboratory Manual. New York: Stockton, 1990.
  17. Laughlin K and Zanella AJ. Stress-induced deficits in pig spatial learning: influence of weaning age. Physiol Behav In press.
  18. Leibrock J, Lottspeich F, Hohn A, Hofer M, Hengerer B, Masiaskowski P, Thoenen H, and Barde YA. Molecular cloning and expression of brain-derived neurotrophic factor. Nature 341: 149–152, 1989.[CrossRef][ISI][Medline]
  19. Mikkelsen M, Moller A, Jensen LH, Pedersen A, Harajehi JB, and Pakkenberg H. MPTP-induced parkinsonism in minipigs: a behavioral, biochemical, and histological study. Neurotoxicol Teratol 21: 169–175, 1999.[CrossRef][ISI][Medline]
  20. Miller RT, Christoffels AG, Gopalakrishnan C, Burke J, Ptitsyn AA, Broveak TR, and Hide WA. A comprehensive approach to clustering of expressed human gene sequence: the sequence tag alignment and consensus knowledge base. Genome Res 9: 1143–1155, 1999.[Abstract/Free Full Text]
  21. Rhee P, Talon E, Eifert S, Anderson D, Stanton K, Koustova E, Ling G, Burris D, Kaufmann C, Mongan P, Rich NM, Taylor M, and Sun L. Induced hypothermia during emergency department thoracotomy: an animal model. J Trauma 48: 439–450, 2000.[ISI][Medline]
  22. Rink A, Santschi EM, and Beattie CW. Normalized cDNA libraries from a porcine model of orthopedic implant-associated infection. Mamm Genome 13: 198–205, 2002.[CrossRef][ISI][Medline]
  23. Smith DH, Chen XH, Nonaka M, Trojanowski JQ, Lee VMY, Saatman KE, Leoni MJ, Xu BN, Wolf JA, and Meaney DF. Accumulation of amyloid beta and tau and the formation of neurofilament inclusions following diffuse brain injury in the pig. J Neuropathol Exp Neurol 58: 982–992, 1999.[ISI][Medline]
  24. Swindle MM, Smith AC, and Hepburn BJ. Swine as models in experimental surgery. J Invest Surg 1: 65–79, 1988.[Medline]
  25. Tosser-Klopp G, Benne F, Bonnet A, Mulsant P, Gasser F, and Hatey F. A first catalog of genes involved in pig ovarian follicular differentiation. Mamm Genome 8: 250–254, 1997.[CrossRef][ISI][Medline]
  26. Traystman RJ. Animal models of focal and global cerebral ischemia. ILAR J 44: 85–95, 2003.[Medline]
  27. Worobec EK, Duncan IJH, and Widowski TM. The effects of weaning 7, 14, and 28 days on piglet behaviour. Appl Anim Behav Sci 62: 173–182, 1999.[CrossRef][ISI]
  28. Yao J, Coussens PM, Saama P, Suchyta S, and Ernst CW. Generation of expressed sequence tags from a normalized porcine skeletal muscle cDNA library. Anim Biotech 13: 211–222, 2002.[CrossRef][Medline]
  29. Yang YH, Dudoit S, Luu P, Lin DM, Peng V, Ngai J, and Speed TP. Normalization for cDNA microarray data: a robust composite method addressing single and multiple slide systematic variation. Nucleic Acids Res 30: e15, 2002.[Abstract/Free Full Text]