Mouse primitive (embryonic) erythropoiesis begins around embryonic day 7.5 (E7.5) in the visceral yolk sac (Haar and Ackerman,1971). The visceral yolk sac is composed of endoderm-derived epithelium and a mesoderm cell layer (Haar and Ackerman,1971; reviewed in Palis and Yoder,2001). The visceral endoderm participates in nutrient uptake and transport and secretes signaling factors that include Indian hedgehog and vascular endothelial growth factor A (VEGF-A), which are important for both hematopoietic and vascular development (reviewed in Baron,2003). Between E7 and E8, the mesodermal layer gives rise to mesodermal cell masses or cords of cells termed blood islands that differentiate into primitive erythroid cells. The outer layer of the blood island differentiates into endothelial cells (Palis and Yoder,2001). Mature yolk sac primitive erythroblasts are large, nucleated, and synthesize mainly embryonic globin transcripts (Baron,2003; reviewed in McGrath and Palis,2005). Later in development at approximately E11.5, primitive erythropoiesis is replaced with definitive (adult) erythropoiesis (Baron,2003). Blood cell production occurs in the fetal liver and thereafter in the bone marrow throughout adult life. Mature erythroid cells at the definitive stage are small, enucleated, and produce adult globin.
There are a small number of known regulators required for both primitive erythropoiesis and vasculogenesis. These regulators include the leukemia transcription factor SCL/TAL1, the proto-oncogene RBTN2/LMO2, the tyrosine kinase receptor FLK-1, VEGF-A which is a ligand for Flk-1, and Krüppel-like Factor 6 (KLF6; reviewed Warren et al.,1994; Orkin and Zon,1997; Sanchez et al.,2001; Martin et al.,2004; Matsumoto et al.,2006). Mouse knockouts of TAL1 and LMO2 have abnormal yolk sac erythropoiesis and die between E8.5 and E9 (Warren et al.,1994; reviewed in Orkin and Zon,1997). VEGF-A null embryos die at approximately E9 due to severe abnormalities in the yolk sac vasculature, including a reduction in the number of blood islands in the yolk sacs of these embryos (Damert et al.,2002). Homozygous embryos with a disruption of the Flk-1 gene die between E8.5 and E9.5 due to abnormal yolk sac blood island formation and have reduced numbers of hematopoietic progenitors (Shalaby et al.,1995). KLF6−/− embryos die by E12.5 and have poorly defined yolk sac vasculature, and a reduced number of erythroid cells compared with wild-type (Matsumoto et al.,2006). All of these regulators are involved in both primitive erythropoiesis and vasculogenesis, but SCL is also required for definitive erythropoiesis (Orkin and Zon,1997; Martin et al.,2004).
There are other regulators involved in primitive erythropoiesis but not necessarily vasculogenesis, including Erythroid Krüppel-like Factor (EKLF), Krüppel-like Factor 2 (KLF2), Nucleophosmin (Npm) and Bone Morphogenetic Protein 4 (BMP4). EKLF regulates adult β-globin expression and is essential for definitive erythropoiesis (reviewed in Bieker,2001). Recently, it was shown that EKLF also has a role in primitive erythroid development (Drissen et al.,2005; Hodge et al.,2006). EKLF−/− primitive erythroid cells have red cell membrane and cytoskeletal abnormalities in addition to inclusion bodies caused by precipitation of the α-globin chains (Drissen et al.,2005). Basu et al. (2005) showed that KLF2 positively regulates the mouse embryonic globin genes Ey and βh1 and is necessary for normal primitive blood cell morphology. Of interest, simultaneous ablation of both EKLF and KLF2 much more severely affects both globin gene expression and embryonic erythropoiesis (Basu et al.,2007). Nucleophosmin (Npm) is a nucleolar protein that is essential for mouse embryonic development. Genetic ablation of Npm affects both primitive and definitive hematopoiesis. E9.5 Npm−/− yolk sac hematopoietic precursors fail to differentiate toward the erythroid lineage and Npm−/− embryos die by E12.5 due to severe anemia. In addition, Npm+/− mice display features consistent with the hematological condition myelodysplasia, such as binucleated erythroid precursors (Grisendi et al.,2005). BMP4 is important for mesoderm formation and patterning. Disruption of mouse BMP4 can result in embryonic lethality between E6.5 and E9.5. BMP4−/− embryos that survive beyond E7 have less yolk sac blood islands in addition to a lower number of erythroid and endothelial cells (Winnier et al.,1995). Taken together, the data suggest that EKLF, KLF2, Npm, and BMP4 have specific roles in primitive erythropoiesis. There are likely other regulators that are essential for primitive erythroid development that have not yet been discovered.
The objective of this study was to generate a profile of mouse embryonic yolk sac erythroid cells and identify novel regulatory genes differentially expressed in erythroid compared with nonerythroid (epithelial) cells. Identifying these genes will contribute to a greater understanding of how the primitive erythroid program is controlled. Laser capture microdissection (LCM) was used to isolate homogeneous populations of mouse E9.5 erythroid precursors and epithelial cells. RNA isolated from the erythroid and epithelial samples was amplified, labeled, and hybridized to mouse Affymetrix GeneChip Mouse Genome 430A 2.0 arrays. Ninety-one genes were expressed higher in erythroid than in epithelial cells. Of these, 67 are newly discovered erythroid-enriched genes and are candidates for positively regulating primitive erythropoiesis.
Genetic Profiles of Microdissected E9.5 Embryonic Yolk Sac Erythroid and Epithelial Cells
We previously developed a protocol for isolating E10.5 erythroid cells and performing microarray assays (Redmond et al.,2006). To identify the genes expressed in primitive erythroid precursor cells, we used the LCM/microarray strategy to generate an expression profile of E9.5 yolk sac cells. The example in Figure 1A shows erythroid cells within E9.5 yolk sac blood islands and neighboring epithelial cells before LCM. As shown in Figure 1B, multiple erythroid cells can be captured by LCM. Epithelial cells were captured from the same slides. For each of the four erythroid and four epithelial samples obtained, capillary electrophoresis was performed with the Agilent Bioanalyzer using an aliquot of pooled total RNA from at least 500 E9.5 yolk sac erythroid or epithelial cells. The pooled RNA was considered to be of high quality, displaying distinctive 18S and the 28S ribosomal RNA peaks (Supplementary Figure S1, which can be viewed at http://www.interscience.wiley.com/jpages/1058-8388/suppmat).
From 10 to 39 ng of total erythroid and epithelial RNA were amplified using a two-round linear amplification method and labeled for microarray analyses. The cRNA products were subjected to capillary electrophoresis, and the electropherogram profiles indicated that high-quality products were obtained from both erythroid and epithelial cells, with a median size of 1,000 bp (Table 1). Each of the four erythroid and four epithelial samples were hybridized to an Affymetrix GeneChip Mouse Genome 430A 2.0 array. Results from the microarray experiments indicated that background and noise levels for all four pairs of arrays were within acceptable ranges and the scale factors were similar, making comparisons valid. In addition, 41–49% of the genes on the chip were considered present for all arrays, ensuring a successful hybridization (Dumur et al.,2004; Table 1). An RNA degradation plot analysis was performed as a qualitative assessment of RNA quality and amplification. Generally, if RNA is severely degraded, the plots will have an increasingly positive slope. However, a linear agreement between the plots is more important than the slope values (Bolstad et al.,2005). In Figure 2A, the RNA degradation plots show similar linear relationships and slope values for the erythroid and epithelial microarrays, indicating that all of the amplified samples have equivalent RNA quality. Unsupervised hierarchical clustering analysis was performed to compare the erythroid and epithelial arrays. The dendrogram in Figure 2B shows that the four erythroid and four epithelial arrays cluster with their respective class. This finding indicates that the microarray results are robust and can discriminate between the cell types. There is some subclustering within the erythroid and within the epithelial classes of microarrays, which is likely due to differences in either sample processing or hybridization. However, this is not of consequence because each pair of erythroid and epithelial samples was processed in parallel.
Table 1. Quality Assessment Parameters
Number of captured cells
cRNA bp median size
% genes present
Ery, erythroid; Epi, epithelial
Identification and Functional Classifications of E9.5 Erythroid-Enriched Genes
Probe set expression summaries were calculated using the MAS5 algorithm for the erythroid and epithelial data sets. The S-score statistical method was used to compare the four paired sets of erythroid and epithelial GeneChips with respect to gene expression differences. Three hundred forty-four probe sets had S-score values greater than 2, indicating significant differences between the erythroid and epithelial samples. Using a Bonferroni correction method, 91 unique genes are expressed higher in erythroid than in epithelial cells (Table 2) and 205 unique genes are expressed higher in epithelial than erythroid cells (Supplementary Table S1).
Table 2. Genes Expressed Higher in Erythroid Than in Epithelial Cells
aClassifications: OF, other factors; ES, known hematopoietic/erythroid-specific; UF, unknown factors; CS, cell signaling; TF, transcription; CCF, Cell Cycle Factors; CSF, Cell Surface Receptors; CM, Chromatin Remodeling/Assembly; RP, RNA Processing. Bold font indicates genes verified by quantitative real-time polymerase chain reaction.
K+ intermediate/small conductance Ca2+-activated channel, subfamily N, member 4
Protein tyrosine phosphatase, non-receptor type 12
SRY-box containing gene 4
Armadillo repeat containing, X-linked 2
Actin, alpha 2, smooth muscle, aorta
Ubiquitin specific petidase 45
Coagulation factor II (thrombin) receptor
Thymosin, beta 10
Protein tyrosine phosphatase-like (proline instead of catalytic arginine), member a
Suppressor of Ty 4 homolog 1(S. cerevisiae)
Heat shock 22kDa protein 8
Guanosine monophosphate reductase
Ankyrin repeat and SOCS box-containing protein 17
Isopentenyl-diphosphate delta isomerase 1
DNA segment, human D4S114 (Chromosome 5 open reading frame 13)
Thymosin, beta 4, X chromosome
Regulator of G-protein signalling 10
H3 histone, family 3A
SH3-binding domain glutamic acid-rich protein like
Immediate early response 5
ATP-binding cassette, sub-family B (MDR/TAP), member 10
Melanocortin 2 receptor accessory protein
3-oxoacid CoA transferase 1
Speckle-type POZ protein
The list of erythroid-enriched genes was arranged into various functional categories using GO available through DAVID, and the available literature. These functional categories are represented in the pie chart in Figure 3. They include transcription factor, cell signaling, cell cycle, cell surface receptor, chromatin modeling, known erythroid/hematopoietic specific, RNA processing, and other genes, and also genes of unknown function. Several known erythroid/hematopoietic-specific transcripts such as cell membrane proteins (glycophorin A, transferrin receptor 1, CD24a Antigen), enzymes involved in heme synthesis (aminolevulinic acid synthase 2, ferrochelatase), and transcription factors (GATA1 and EKLF/KLF1) are expressed significantly higher in erythroid than in epithelial cells, as expected. This finding confirms that an erythroid-enriched sample was obtained. LMO2, TAL1, and EKLF, three of the small number of genes known to be involved in primitive erythropoiesis, are up-regulated in erythroid compared with epithelial cells. Of the 91 erythroid-enriched genes, 67 genes have not previously been characterized as erythroid-enriched. Twenty-nine of these novel erythroid-enriched genes are categorized as other factors, having roles in alcohol synthesis, protein, carbohydrate, and lipid metabolism. Cell signaling molecules (15 total) comprise the third highest functional category and include genes like reelin (Reln), thrombospondin-1 (Thbs1), and platelet factor 4 (PF4). The nine erythroid-enriched genes of unknown function include two expressed sequence tags containing either a ChaC-like or a TLC-like domain, suggesting that they are involved in cation transport. Genes such as muscleblind-like 1 (Mbnl1) and nucleosome binding protein 1 (NSBP1) are classified as transcription factors (5 total). Several transcription factors and cell signaling genes are expressed significantly higher in epithelial than in erythroid cells, and some of these genes are listed in Table 3.
Table 3. Selected Genes Expressed Higher in Epithelial Than in Erythroid Cells
Identification of Relevant Biological Networks for E9.5 Erythroid-Enriched Genes
To determine how the differentially expressed genes might interact in erythroid cells, an Ingenuity Pathway Analysis (IPA) was performed using the Ingenuity Systems program at www.ingenuity.com. IPA uses a knowledge base program to generate relevant biological networks. Thirteen major networks were discovered in the gene list and five of the networks have a significant P value of 1.0E-21 or less, indicating that it is highly unlikely that these networks were detected by chance. Three of the five highest scoring networks are shown as examples in Figure 4. The first network (Fig. 4A) includes the globin genes, EKLF/KLF1 and GATA1 but, interestingly, it also includes LMO2, TAL1, thrombospondin-1 (Thbs1), and platelet factor 4 (PF4), which the microarray data indicates are up-regulated in the primitive erythroid fraction. This network is important for hematological disease, cell death, and hematological system development and function. The second network (data not shown; P value 1.0E-25) includes carbonic anhydrase 2 and CD24a antigen, and is necessary for cellular development, amino acid metabolism, and posttranslational modification. The third network (Fig. 4B) contains CD24a antigen, retinoic acid, muscleblind-like 1 (Mbnl1), and ferrochelatase and is important for tissue morphology, cell-to-cell signaling and interaction, and cell death. The fourth network (data not shown; P value 1.0E-21) contains KLF2, which is a known regulator of primitive erythropoiesis, vimentin, and transferrin receptor. This network is essential for cell growth and proliferation, cell death, and cancer. The fifth network (Fig. 4C) includes glycophorin A and reelin (Reln) and is important for tissue development, cell-to-cell signaling and interaction, and cell assembly. The data suggest that novel differentially expressed genes that cluster in networks containing known erythroid-specific genes may be important for primitive erythropoiesis.
Prioritization and Validation of Erythroid-Enriched Candidate Genes
The novel erythroid-enriched candidate genes were prioritized with a scoring system using a combination of statistical and biological criteria. Information derived from the microarray and IPA analyses and gene ontology was used to generate the scoring criteria in Table 4. PF4, Thbs1, Mbnl1, and Reln were the highest scoring genes with 30, 31, 32, and 39 of 40 possible points, respectively (Table 4). Regulatory genes that are essential during development, differentiation and/or proliferation were of great interest. PF4, Thbs1, Mbnll, and Reln have established roles in development or differentiation in other systems (reviewed in Bornstein,1995; Gewirtz et al.,1995; Vacca et al.,1999; reviewed in Lawler,2000; Adams,2001; Eslin et al.,2004; reviewed in Slungaard,2005; Strieter et al.,2005; Forester et al.,2006; reviewed in Pascual et al.,2006; Zhao et al.,2007) and, therefore, may have an important role in embryonic erythroid development. There were 22 additional novel erythroid-enriched candidates within the 20 to 26 scoring range. The highest scoring genes in this range are Nudix-type motif 4 (Nudt4), Elongation factor RNA Polymerase II (Ell2), X-linked myotubular myopathy gene 1/Myotubularin (Mtm1), and BMP2 inducible kinase 2 (BMP2K; Supplementary Table S2).
Table 4. Scoring System for Prioritization of E9.5 Erythroid-Enriched Genes
Denotes rank-ordered P values (more points allowed for lower P values).
Erythroid-enriched expression (10 points)
Biological networks with other candidates (5 points)
Established role in differentiation/development (7 points)
The four genes that scored 30 or more points were selected for quantitative real-time polymerase chain reaction (qRT-PCR) verification of differential expression. For qRT-PCR verification, additional LCM experiments were performed using new samples to avoid any possible bias in the microarray data. Approximately 1,500 erythroid and 2,600 epithelial cells were collected from three individual E9.5 yolk sacs. qRT-PCR using SYBR green chemistry was then performed to verify the microarray expression data for PF4, Thbs1, Mbnl1 and Reln. LMO2, and TAL1 were also selected for qRT-PCR verification because these two genes are important for primitive erythropoiesis. Cyclophilin A was used as an internal standard to normalize the expression data. In Figure 5, LMO2, TAL1, PF4, Thbs1, Mbnl1, and Reln mRNAs are at least fourfold more abundant in erythroid than in the epithelial cells. For the microarray assays, the amounts of Thbs1, Reln, and Mbnl1 mRNA are approximately 3-fold higher and PF4 is 4.8-fold higher in the erythroid than in the epithelial cells. In addition, LMO2 mRNA is approximately 3.7-fold higher and TAL1 mRNA is 4.6-fold higher in erythroid than in epithelial cells. There was a positive association between the microarray and qRT-PCR data, thus confirming the accuracy of the microarray assays in identifying differentially expressed genes.
Laser capture microdissection and microarray technology was used to generate a gene expression profile of mouse E9.5 primitive erythroid development. High-quality gene expression profiles of yolk sac primitive erythroid and epithelial cells were compared to identify factors that may be involved in embryonic erythropoiesis. There were 91 genes with higher expression in erythroid than in epithelial cells, and 67 of these were not previously known to be erythroid-enriched. Ingenuity Pathway Analyses indicated that 45 of these 67 novel erythroid-enriched genes, which are candidates for the embryonic erythroid program, clustered in highly significant networks containing known erythroid-specific genes.
We were particularly interested in novel erythroid-enriched genes that are known to be important for development, differentiation, and proliferation in other systems. Of the 67 candidate genes, 6 are required in organ development. These are platelet factor 4 (PF4), thrombospondin-1 (Thbs1), reelin (Reln), BMP2 inducible kinase (Bmpk2), which are cell signaling factors; Myotubularin 1 (Mtm1), which is classified as an other factor; and the transcription factor muscleblind-like 1 (Mbnl1). Differential expression of PF4, Thbs1, Reln, and Mbnl1 mRNA was verified by qRT-PCR, and the functional significance of these genes in embryonic erythroid cells needs to be further determined.
Platelet factor 4, the founding member of the chemokine CXC family, is important for thrombosis, megakaryopoiesis, and angiogenesis (Gewirtz et al.,1995; Eslin et al.,2004; Slungaard,2005; Strieter et al.,2005). PF4 is expressed in megakaryocytes and to some extent in mature platelets (Gewirtz et al.,1995; Slungaard,2005). Tober et al. (2007) recently showed that the megakaryocytes are very rare in the E9.5 yolk sac. It is, therefore, likely that PF4 mRNA is expressed in primitive erythroid precursors and not detected due to megakaryocyte contamination. Zhang et al. (2004) showed that PF4 enhances the adhesive properties in normal and leukemia hematopoietic stem/progenitor cells. Other studies have reported that PF4 increases the viability of normal hematopoietic cells (Aidoudi et al.,1996; Han et al.,1997). It is possible that PF4 is necessary for adhesion of embryonic erythroid precursor cells to the extracellular matrix before terminal erythroid differentiation in the yolk sac.
Thrombospondin-1 belongs to a family of extracellular matrix proteins and is best known for its antiangiogenic properties in vivo (Bornstein,1995; Lawler,2000). Thbs1 binds and interacts with many extracellular matrix molecules, proteases, and cytokines (Bornstein,1995; reviewed in Adams,2001). Thbs1 is expressed in multiple tissues during embryonic development (Adams,2001). Iruela-Arispe et al. (1993) showed that Thbs1 mRNA is expressed in mouse heart beginning around E10, but is not detectable in the blood vessels until E16. It is, therefore, likely that Thbs1 mRNA is expressed higher in E9.5 erythroid precursors in our study, and not in contaminating endothelial cells. Thbs1 null mice are fertile and develop normally with no significant abnormalities. However, by 1 month of age, Thbs1−/− mice develop pneumonia and have increased numbers of hemosiderin alveolar macrophages as a result of hemolysis (Lawler et al.,1997). Multiple studies have reported that Thbs1 is involved in diverse functions, including adhesion, differentiation, tissue remodeling, and migration (Adams and Lawler,1994; Vacca et al.,1999; Tucker et al.,1999; Narizhneva et al.,2005). Thrombospondin-1 could regulate cell-to-cell interactions between erythroid and other cells during erythroid development.
Of interest, E9.5 erythroid cells express reelin; this factor is associated with development of the mouse nervous system. Reelin is a secreted glycoprotein that is expressed in Cajal-Retzius cells of the cerebral cortex. Several studies have shown that reelin is required for radial neuronal morphology, cortical neuronal migration, and proliferation (Hartfuss et al.,2003; Forester et al.,2006; Zhao et al.,2007). Ikeda and Terashima (1997) reported that reelin is expressed as early as E8.5 in the somites, yolk sac, and foregut in the developing mouse. Recently, Zhao et al. (2007) showed that neurogenesis is significantly reduced in the dentate gyrus of reelin mutant mice. Because reelin is expressed in E9.5 yolk sac erythroid cells, it may be required for red cell migration or proliferation.
Muscleblind-like 1 is a C3H zinc finger transcription factor that belongs to a family of alternative splicing regulators (Pascual et al.,2006). Disruption of muscleblind-like 1 in mice causes myotonia, cataracts, and aberrant RNA splicing, characteristics associated with myotonic dystrophy (Kanadia et al.,2003; Lin et al.,2006). Muscleblind-like 1 is essential for muscle differentiation (Pascual et al.,2006), and, therefore, may be required for erythroid differentiation.
In summary, our data provide a novel catalog of genes with erythroid-enriched expression in the E9.5 mouse embryonic yolk sac. Testable hypotheses about how PF4, Thbs1, Reln, and Mbnl1 might positively regulate embryonic erythropoiesis can now be made. Future studies will be designed to determine the specific roles of these genes during erythroid development. In the future, this work could lead to treatments for β-thalassemia and sickle cell anemia, which are ameliorated by elevating the expression of the embryonic β-like globin genes.
Embryo Dissections, Sample Preparation, and Tissue Sections
Timed-pregnant FVB/N mice were anesthetized and killed. The uterine horns were removed and placed in 1× PBS. E9.5 yolk sacs were dissected from the embryo and cryoprotected in 20% sucrose in PBS at 4°C for 25–30 min. The tissues were rinsed once in 1:1 20% sucrose PBS:Optimal Cutting Temperature freezing medium (OCT, Tissue-Tek) and then in OCT alone. The E9.5 yolk sacs were frozen as previously described (Redmond et al.,2006) and 8-micron sections were cut using the Shandon Cryotome Cryostat.
Staining of Frozen Tissue Sections and Laser Capture Microdissection
Frozen sections were stained using the HistoGene LCM frozen section staining kit (Molecular Devices, Mountain View, CA). Using the PixCell II Laser Capture Microdissection System (Arcturus Bioscience), approximately one erythroid cell was collected for every two to three laser pulses, and one epithelial cell was collected per pulse on LCM HS Capsure caps. Epithelial cells were collected from the same microscope slide immediately after procurement of the erythroid cells and were used as the nonerythroid control. For the four erythroid and four epithelial microarray samples, erythroid and epithelial cells were collected from 12 to 50 microscope slides, using two to four yolk sacs from two different litters.
RNA Extraction, Quality Assessment, Linear Amplification, and Labeling
RNA extractions from the four erythroid and four epithelial samples, and quality assessment by capillary electrophoresis were performed as previously described (Redmond et al.,2006). For the microarrays, 10 to 39 ng of erythroid or epithelial RNA was used to generate double-stranded cDNA with the T7-oligo (dT) primer according to the GeneChip Two cycle target labeling protocol (Affymetrix Inc., Santa Clara, CA). Labeled cRNA yield and purity was determined as described (Redmond et al.,2006). RNA degradation was assessed using the RNA digestion plot function implemented in the affy package (Gautier et al.,2004) available through Bioconductor project (www.bioconductor.org) and the R statistical programming environment (R Development Core Team,2005).
Fragmentation and Microarray Hybridization
Fifteen micrograms of labeled erythroid or epithelial cRNA was fragmented and 10 μg was hybridized to Affymetrix GeneChip Mouse Genome 430A 2.0 arrays for 18 hr (Affymetrix Inc.). The eight microarrays were washed and stained with streptavidin-phycoerythrin as described (Redmond et al.,2006). Image analysis was performed on the E9.5 yolk sac data set using the GeneChip Operating System version 1.4 (GCOS1.1, Affymetrix Inc.). The image of the scanned array was stored as a DAT file using the Affymetrix software. Quality control parameters such as scaling factors used to normalize the chips, average background and noise were also analyzed. The raw intensities for each probe set were stored in electronic formats by the GCOS software. Probe set expression summaries were calculated with the Microarray Suite version 5.0 (MAS5) algorithm (Hubbell et al.,2002). MAS5 expression summaries were obtained using the mas5 function in the Affymetrix package.
The Significance score (S-score) method was used to compare four sets of paired GeneChips to determine significant changes in gene expression between erythroid and epithelial cells (Kennedy et al.,2006). It uses an error-based model to determine the variances for probe pair signals and follows a normal standard distribution. S-score values were obtained using the SScore function in the sscore package in Bioconductor. Probe sets with absolute values of the S-score greater or equal to 2.00 in all four pairwise comparisons were considered to correspond to significant differences between the two types of samples. Probe sets that had a significant S-score among all four paired comparisons were retained for further study. P values were calculated from these S-score values and were combined using Fisher's procedure for combining P values. Using a Bonferroni correction method for multiple comparisons, probe sets with P values < 2.2E-06 indicated a significant difference and were further analyzed. In a separate analysis, average linkage hierarchical clustering using a Pearson (centered) correlation was performed with the BRB-ArrayTools v3.02 package. Probe sets above a background of 60 and those not affected by the “Batch Effect” were used in the cluster analysis.
IPA, Gene Classification, and Prioritization of Erythroid-Enriched Genes
IPA (Ingenuity Systems, Mountain View, CA; http://www.ingenuity.com); was used to determine how differentially expressed genes interact in erythroid cells. The 91 erythroid-enriched genes were mapped to their corresponding gene object/symbol in the IPA biological knowledge database. These focus genes were used as the basis to generate significant biological networks.
Functional gene categories were determined with GO (Gene Ontology) using DAVID (Database Annotation Visualization and Integrated Discovery) (http://david.abcc.ncifcrf.gov/). For the analyses, only the “biological process” and “molecular function” categories were used. Genes with erythroid-enriched expression compared with epithelial samples were prioritized with a scoring system using a combination of statistical and biological criteria. The biological criteria included known role in differentiation/development or proliferation, and identity as a regulatory gene. Chromatin regulatory, cell surface receptor, transcription factor, and cell signaling genes were considered as regulatory genes. The statistical requirement included rank ordered P values, and genes with the lowest P values received the maximum points. The highest scoring genes were selected for qRT-PCR verification.
cDNA Synthesis and qRT-PCR
cDNA was prepared from total RNA from approximately 1,500 to 2,900 microdissected erythroid or 2,600 to 3,400 epithelial cells as described (Redmond et al.,2006). Primers were designed using PrimerExpress software (Applied Biosystems), and the sequences of these oligonucleotides are indicated in Table 5. The NCBI database (http://www.ncbi.nih.gov) was used to establish that the primers were gene specific. qRT-PCR experiments were performed using the ABI-Prism 7300 system (Applied Biosystems, Foster City, CA) following the cycle parameters described previously (Basu et al.,2005). A dissociation curve was used for all qRT-PCR experiments with SYBR Green Chemistry, and this indicated that only a single product was amplified. A pre-designed Taqman probe and primer set was used for qRT-PCR for cyclophilin A mRNA (Applied Biosystems), which was the internal standard to normalize the expression data. qRT-PCR was performed in duplicate using three biological replicates.
Table 5. Oligonucleotides Used for Real-Time Polymerase Chain Reaction
5′ ATGTCCTCGGCCATCGAAAG 3′
5′ CGGTCCCCTATGTTCTGCTG 3′
5′ CACTAGGCAGTGGGTTCT TTG 3′
5′ GGTGTGAGGACCATCAGAAATCT 3′
5′ CCAGCCTGGAGGTGATCAA 3′
5′ GGCAAATTTTCCTCCCATTCT 3′
5′ TGTGACAGAAAATCAAGTTTGCAA 3′
5′ CTTGGCACCAGCAAAGCA 3′
5′ GGCCCAGCAAATGC AGTTA 3′
5′ CTAAGCTTGGTGCAACTGAAAACA 3′
5′ CAAGAACAATACCGCTGATTGG 3′
5′ GATGTGGATGACTGTGCTCACA 3′
Microarray Accession Number
The microarray data for this study are available online at the NCBI Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo/) under the accession number GSE10002.
We thank Mohua Basu, Sean Fox, and Tina Lung for excellent technical assistance. We thank Dr. Gordon D. Ginder for his valuable suggestions and critical review of this work. L.C.R. and J.A.L. were funded by the NIH. The authors declare no competing financial interests.