Erythroid-induced commitment of K562 cells results in clusters of differentially expressed genes enriched for specific transcription regulatory elements

Sankar Addya1, Margaret A. Keller1, Kathleen Delgrosso1, Christine M. Ponte1, Rajanikanth Vadigepalli2, Gregory E. Gonye2 and Saul Surrey1

1 The Cardeza Foundation for Hematologic Research and Division of Hematology, Department of Medicine, Jefferson Medical College
2 Daniel Baugh Institute for Functional Genomics and Computational Biology, Department of Pathology, Thomas Jefferson University, Philadelphia, Pennsylvania 19107


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
Understanding regulation of fetal and embryonic hemoglobin expression is critical, since their expression decreases clinical severity in sickle cell disease and ß-thalassemia. K562 cells, a human erythroleukemia cell line, can differentiate along erythroid or megakaryocytic lineages and serve as a model for regulation of fetal/embryonic globin expression. We used microarray expression profiling to characterize transcriptomes from K562 cells treated for various times with hemin, an inducer of erythroid commitment. Approximately 5,000 genes were expressed irrespective of treatment. Comparative expression analysis (CEA) identified 899 genes as differentially expressed; analysis by the self-organizing map (SOM) algorithm clustered 425 genes into 8 distinct expression patterns, 322 of which were shared by both analyses. Differential expression of a subset of genes was validated by real-time RT-PCR. Analysis of 5'-flanking regions from differentially expressed genes by PAINT v3.0 software showed enrichment in specific transcription regulatory elements (TREs), some localizing to different expression clusters. This finding suggests coordinate regulation of cluster members by specific TREs. Finally, our findings provide new insights into rate-limiting steps in the appearance of heme-containing hemoglobin tetramers in these cells.

gene expression profiling; hemin; cell differentiation; transcription regulatory elements


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
K562 CELLS ARE IMMORTALIZED erythroleukemia cells derived from a patient with chronic myelogenous leukemia (CML) characterized by aneuploidy and dysregulated constitutive expression of BCR-ABL due to the presence of the Philadelphia chromosome (t9:22) (29). The cells are bipotential and can be induced to express erythroid or megakaryocytic properties depending on the inducer (30). K562 cells therefore represent an important in vitro model for basic studies of leukemogenesis. K562 and other erythroleukemia cell lines can be induced to express embryonic and/or fetal hemoglobins (44) and, therefore, are useful also for gaining an understanding of regulation of embryonic and fetal globin-chain expression (19, 48). Elucidation of what regulates these genes could lead to development of therapeutic agents for treatment of sickle cell disease (SCD) and the ß-thalassemia syndromes (7), for which few drugs are currently available. The dominant therapeutic approaches have been aimed at upregulating expression of the fetal globin genes, since overexpression of fetal globins in adult life has no deleterious effects in individuals with deletional and nondeletional forms of hereditary persistence of fetal hemoglobin (HPFH) (18). Furthermore, forced expression of fetal globins, caused either by inhibition of fetal globin-gene silencing or by reactivation of expression in adult life decreases clinical severity in SCD (48).

Currently available therapeutic agents for elevating Hb F levels in patients with SCD and ß-thalassemia include 5-azacytidine and derivatives, butyrate and derivatives, and hydroxyurea. The mechanisms of action of these drugs are not at all clear but are presumed to involve changes in DNA methylation, histone acetylation, and possibly stress-induced, accelerated erythropoiesis, respectively (48). Patient response to these agents varies widely, use of these compounds can cause neutropenia, and some are thought to have oncogenic potential. Therefore, the motivations for development of new drugs aimed at elevated fetal globin in adult life include 1) targeted mechanism of action, 2) high response rate, 3) low toxicity/oncogenicity.

To increase our understanding of the molecular basis of the responses of K562 cells to the fetal/embryonic hemoglobin inducer, hemin, and to identify rate-limiting steps in hemoglobinization, we performed transcriptome analyses of these cells. K562 cells, although a neoplastic cell line, represent a tool to study globin gene regulation, as demonstrated by recent studies showing similar effects of chemical treatment on globin gene expression in K562 and primary erythroid cells (60, 41, 28). We now describe changes in steady-state mRNA profiles before and after hemin treatment during 72 h in culture and provide insights into rate-limiting steps involved in appearance of heme-containing globins in these cells. In addition, we show that differentially expressed genes can be assigned to a limited number of expression patterns. We also present analyses of their 5'-flanking regions and suggest that specific transcription regulatory elements (TREs) could effect coordinate regulation of the putative target genes in these clusters.


    MATERIALS AND METHODS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
Cell culture.
The human erythroleukemia cell line K562 (ATCC, Philadelphia, PA) was cultured in RPMI 1640 (Mediatech, Herndon, VA) supplemented with 10% (vol/vol) fetal bovine serum, 2 mM L-glutamine, 100 U/ml of penicillin, and 100 µg/ml of streptomycin (Mediatech) under 5% (vol/vol) CO2 at 37°C. Cells were seeded at a density of 2 x 105 cells/ml and cultured for 96 h in the presence or absence of the 50 µM hemin (Sigma-Aldrich, St. Louis, MO). Hemin stocks (4 mM) were prepared by dissolving 13 mg of hemin in 0.2 ml of 0.5 M NaOH followed by addition of 0.25 ml of 1 M Tris·HCl, pH 7.8, and dilution to 5 ml with distilled water. The solution was filter sterilized and stored at –20°C until use. Viable cell counts were determined using trypan blue dye exclusion.

Benzidine staining.
Phosphate-buffered saline-washed untreated and treated K562 cells were stained in a benzidine solution (43) containing 0.6% (wt/vol) benzidine base, 2% (vol/vol) hydrogen peroxide, and 12% (vol/vol) acetic acid. At least 500 cells were counted at each time point after treatment using a light microscope to assess the percent of cells that appeared blue due to the presence of heme-containing globin tetramers.

RNA isolation.
DNA-free total RNA was isolated from ~5 x 107 cells from duplicate (A, B) untreated control (0 h) and hemin-treated cultures (after 6, 12, 24, 48, and 72 h) using the RNAqueous-4PCR kit as described by the manufacturer (Ambion, Austin, Texas). RNA was treated with DNase, purified by ethanol precipitation, resuspended in DEPC-treated water, and stored at –80°C.

Microarray analysis.
Microarray analysis was accomplished using the procedure recommended by the manufacturer (Affymetrix, Santa Clara, CA). Briefly, double-stranded cDNA was synthesized using a cDNA synthesis kit (Invitrogen, Carlsbad, CA), and biotinylated cRNA prepared using BioArray High Yield T7 RNA transcript labeling kit (Enzo Diagnostics, Farmingdale, NY). The cRNA was fragmented by heat and ion-mediated hydrolysis and hybridized to the Human Genome HU133A oligonucleotide array GeneChip (Affymetrix) containing ~500,000 spots with 22,283 different probe sets representing 14,397 unique genes. Arrays were washed and stained using a GeneChip fluidics station (Affymetrix), and hybridization signals were amplified using antibody amplification with goat IgG (Sigma-Aldrich) and anti-streptavidin biotinylated antibody (Vector Laboratories, Burlingame, CA). GeneChips were scanned using a GeneArray scanner (Agilent Technologies, Palo Alto, CA). Data analysis was performed using Affymetrix MAS 5.0, Micro DB 3.0, Data Mining Tool (version 3.0), and GeneSpring 5.1 (Silicon Genetics, Redwood City, CA) software to identify expressed and differentially expressed genes. For absolute expression analysis, images were analyzed by MAS 5.0 to determine whether a gene is expressed ("present"). The target signal value used was 150 with the "all probe sets" choice for scaling to ensure consistency of the files to be compared. Expression patterns at each time point were compared with those of untreated cultures. Detection change call (i.e., increase, decrease, marginal, or no change) was determined by comparison expression analysis (CEA) using MAS 5.0 software. Four pair-wise comparisons were performed at each time point using duplicate treated and untreated cultures (i.e., 6A vs. 0A, 6B vs. 0A, 6A vs. 0B, and 6B vs. 0B). Only genes designated as differentially expressed in common from all four comparisons at each time point were selected for further analysis. The data sets discussed herein can be accessed at the Gene Expression Omnibus (GEO) web site under the submission numbers GSM16523–GSM16529, GSM16531–GSM16533, GSM16535, and GSM16536 (http://www.ncbi.nlm.nih.gov/geo/). Supplemental materials containing complete results of transcriptome analysis in untreated K562 cells (Supplemental Table S1) and results of differentially expressed genes in response to hemin (Supplemental Table S2) are available online at the Physiological Genomics web site.1

Self-organizing map clustering.
Data from all time points, using MAS 5.0 absolute analysis, were published to Micro DB 3.0. Self-organizing map (SOM) clustering (8, 49) was performed after averaging of duplicate signals of each time point by Data Mining Tool 3.0 (Affymetrix). The SOM algorithm distributes genes into an arbitrarily selected number of patterns according to their expression profile after hemin treatment. The details and description of the parameters can be found in the Affymetrix Data Mining Tools Users Guide, version 3.0.

Gene annotation.
Differentially expressed genes from microarray analysis were annotated using the NetAffx Analysis Center (27) and http://www.affymetrix.com.

Real-time RT-PCR.
A selected subset of differentially expressed genes was chosen for independent confirmation by SYBR Green-based, real-time RT-PCR using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as an internal reference gene. Primer pairs (Table 1) (Integrated DNA Technologies, Coralville, IA) for representative genes from the SOM-defined clusters were designed to span or bridge exons when possible, using Primer Express software [Applied Biosystems (ABI)]. Melting curve analysis with Dissociation software (ABI) was used to confirm generation of a single product.


View this table:
[in this window]
[in a new window]
 
Table 1. Sequences of primer pairs used for real-time RT-PCR

 
First, validation experiments were performed at varying cDNA input amounts followed by locus-specific PCR to show that the relative difference between reference and test transcripts remained constant for the two different RNA sources over the input range. The cDNAs were synthesized as follows: total RNA (2 µg) was used in a 20-µl reaction containing 500 ng of oligo-dT, 500 µM dNTPs, 10 mM DTT, and 200 U of SuperScript II (Invitrogen). Reactions were incubated at 44°C for 60 min followed by 15 min at 70°C. For this validation phase, untreated (0 h) cDNA was diluted to ~1 ng/µl and used as the working stock solution. RT-PCR was performed in 10 µl for each gene and control (GAPDH) using six different amounts of cDNA (2, 1, 0.5, 0.25, 0.125, and 0.0625 ng). Each reaction was performed in quadruplicate using the 2x SYBR Green I Master Mix (ABI). Cycle thresholds (CT) were determined in real time on the ABI 7900 real-time PCR instrument. The CT then was plotted vs. the log of input cDNA, and only samples showing a slope between –3.8 and –3.3 were analyzed to determine the relative fold change in expression of the test gene compared with the reference gene from each of the different time points after treatment using the formula 2{Delta}{Delta}CT (3).

Transcriptional regulatory network analysis (TRNA) using PAINT v3.0.
Promoter analysis was performed using PAINT v3.0 (50). PAINT v3.0 contains a database of promoters (UpstreamDB) constructed for all the annotated genes (known and putative) in the Ensembl genome database for Homo sapiens, version 16.33, containing 24,261 genes. The transcription start site (TSS) (first exon) of a gene does not necessarily correspond to the start of the open reading frame (ORF). The 5'-untranslated region (UTR) needs to be considered to identify the TSS. For ~10,000 genes, 5'-UTR sequence information is available in the Ensembl human genome database. For the remaining genes, the start of the ORF is considered as an estimate of the TSS. For each gene, 5,000 base pairs (bp) upstream (5' to the TSS) were retrieved from the genome and placed in UpstreamDB. A total of 20,291 genes contain cross-references to information in the UniGene database, thus allowing queries using the UniGene ClusterIDs for the genes of interest in this study. Because of the nature of the cross-referencing between the UniGene and genome assembly, in several cases, multiple genes correspond to a single UniGene cluster.

Promoter analysis for putative TREs, i.e., binding sites for known transcription factors (TFs) is performed by MATCH tool using TRANSFAC database version 7.2 (31). The results are transformed into a candidate interaction matrix (CIM) in which each row corresponds to a promoter, each column represents a TRE, and each element cij of the matrix contains a 1 if ith promoter has jth TRE, 0 otherwise. For any functionally related group or "cluster" of genes, PAINT v3.0 first retrieves the cognate genomic sequences, next analyzes these sequences for presence of known TREs, and finally compares results to those from random samples of genes to analytically test for enrichment of specific TREs. In addition, PAINT v3.0 can combine TRNA with results of gene expression analysis, in particular, clustering based on microarray data. Identification of statistically significant enrichment of a specific TRE within a particular expression cluster may indicate a role for the cognate transcription factor in coordinate regulation of genes in that cluster.

For each group of genes analyzed, the P values for significance of enrichment of each TRE in that group are calculated using hypergeometric distribution, by comparing the abundance of each TRE to that from a reference set of randomly selected genes (15, 50). For each TRE V$XYZ, given 1) a reference CIM of n promoters of which l promoters contain V$XYZ, and 2) a CIM of interest with m promoters of which h contain V$XYZ, the associated P value for enrichment is given as:

The P value for underrepresentation of a TRE in the observed CIM is calculated similarly with the summation in the above equation going from 1 to h. These estimates of significance are utilized in filtering for those TREs that meet a threshold (P < 0.1) to identify most likely regulators of the genes considered in the experimental context of interest.

For TRNA analysis, there are multiple choices of reference sets (whole genome, microarray or untreated, "control" expressed genes). Each comparison of promoter set to reference addresses a different scientific question, and the reference set acts as a filter for predicting significantly enriched TREs corresponding to that filter.


    RESULTS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
Effects of hemin treatment on appearance of benzidine-positive cells.
In the absence of hemin, ~3–5% of K562 cells stained with benzidine after 96 h. Addition of hemin (50 µM) to cultures resulted in the increased appearance of benzidine-positive cells (20%, 30%, and 60% after 24, 48, and 72 h, respectively), which reached ~80% after 96 h in culture.

Microarray analyses before and after varying times of hemin treatment.
As depicted in Fig. 1, transcriptome analyses were performed by expression profiling to define the number and identity of expressed genes and to monitor for changes in steady-state mRNA levels as a function of time in culture after hemin treatment. Array data reproducibility was documented by comparison of spot intensities at each register on duplicate arrays using targets derived from duplicate untreated cultures (Fig. 2A, slope = 0.9294, R2 = 0.9922). The total number of probe sets designated as a "present" (P) on arrays from duplicate untreated cultures was 10,471, which decreased to 7,282 when average intensities <100 were excluded. This represents ~32% of arrayed probe sets (7,282/22,283) and corresponds to 5,184 unique genes. The number of expressed probe sets showed minimal change comparing untreated to 72-h treated cultures (7,282 vs. 7,424, respectively). Many of the highest values were mRNAs coding for ribosomal proteins and {gamma}-globin. A complete listing of values on duplicate arrays, averages, and annotations for the 7,282 probe sets expressed in untreated cells is in Supplemental Table S1.



View larger version (19K):
[in this window]
[in a new window]
 
Fig. 1. Flow diagram of gene expression analyses. Schematic showing work flow beginning with duplicate K562 cultures treated for different times with hemin followed by RNA isolation, cRNA preparation, microarray hybridization, data analyses [absolute expression analysis, self-organizing map (SOM), comparative expression analysis (CEA)], real-time RT-PCR, and Transcriptional regulatory network analysis (TRNA) by PAINT v3.0.

 


View larger version (24K):
[in this window]
[in a new window]
 
Fig. 2. Scatter plots and log scatter plots are used to define criteria for identifying differentially expressed genes. A: a scatter plot of gene expression intensities from microarrays using duplicate untreated K562 cell cultures ("self-versus-self" experiment) is shown. Signal values are graphed from all probe sets on the two arrays, each hybridized with one of two duplicate untreated sample targets. B: log2 scatter plot of self-versus-self experiment is shown. The log2 of the average of the two signal values >100 (>6.65 on x-axis) from duplicate untreated cultures in A is plotted against the log2 of the ratio of the two signals (y-axis). These data show 99.8% of all points are <2-fold up or down.

 
Identification of differentially expressed genes during hemin treatment.
Definition of criteria for designation of a spot as being "differentially expressed" (e.g., steady-state mRNA level up or down relative to untreated cultures) was assessed from "self-versus-self" experiments (28) in which no spots should be so designated [e.g., intensity ratio at each spot from duplicate arrays (I1, I2) should be unity or log2 I1/I2 = 0].

One such plot is shown in Fig. 2B underscoring the number of false positives when stringency for defining a gene as differentially expressed changes from 1.5-fold, to 1.75-fold, to 2-fold up or down. Average intensities >100 on duplicate arrays were chosen, and false positives were defined as spots located outside the limits of two lines parallel to the x-axis corresponding to each of the three different arbitrary y-axis ratio cutoffs. Those limits/lines correspond to y = log2 of ±0.5849 for 1.5-fold, y = log2 of ±0.8074 for 1.75-fold, and y = log2 of ±1.0000 for 2-fold up or down. The number of false positives increases from 11 (0.2%) to 33 (0.4%) to 162 (0.8%) when going from 2- to 1.75- to 1.5-fold as the cutoff for definition of a differentially expressed gene. If only P calls in common from both arrays are used in the analysis, then there were no false positives using a cutoff of ≥2-fold up or down. In addition, the graph shows increased scatter from the ideal line of y = log2 ratio of 0 at lower spot intensities, whereas the data at higher intensities is much tighter. Such increased scatter of the ratios from ideal at low intensities indicates a potentially higher number of false positives can be expected when ratios are derived from low-intensity comparisons, emphasizing the need for independent follow-up confirmation on differentially expressed candidate genes. Based on these self-versus-self results ≥2-fold up or down (log2 ratio greater/equal to +1 and less/equal to –1) relative to RNA from untreated cultures was used as the definition of a differentially expressed gene. Using this criteria, truly differentially expressed genes with high intensities which may change between only 1.5- to 2-fold up or down after treatment may be missed, but the expected number of false positives at lower intensities is minimized.

We next performed overlap analysis, comparing arrays from duplicate cultures from five different time points against RNA from duplicate untreated cultures using intensities with no cutoff, thus allowing detection of significant changes in low-abundance transcripts. We required that all four comparisons (e.g., 48A vs. 0A, 48B vs. 0A, 48A vs. 0B, 48B vs. 0B) at each time point showed a ≥2-fold up or down change to classify a gene as differentially expressed; this was accomplished using CEA and overlap analysis. A total of 1,209 probe sets including 899 known unique genes were classified as differentially expressed. The number of differentially expressed probe sets as a function of time after hemin treatment was 124, 202, 293, 548, and 847 at 6, 12, 24, 48, and 72 h, respectively.

Identification of common expression patterns for differentially expressed genes.
Data from all time points following hemin treatment were analyzed by SOM clustering where expression is statistically evaluated for change over time in which patterns can emerge showing similar expression profiles for the different genes. Results show 517 probe sets representing 425 genes were clustered into 8 distinct timed expression patterns (Fig. 3). Of these 425 genes generated from SOM analysis, 322 (399 probe sets) or ~76% were in common with the list of 899 genes found to be differentially expressed by CEA. A sampling of the probe sets in the eight expression clusters is shown in Table 2; the entire listing with expression ratios at different time points is presented in Supplemental Table S2. This listing represents a conservative estimate of differentially expressed genes since we required they be categorized as differentially expressed by both analysis methods.



View larger version (30K):
[in this window]
[in a new window]
 
Fig. 3. Genes differentially expressed following hemin treatment cluster into distinct expression patterns. The SOM algorithm identified 517 different probe sets demonstrating differential expression at 0, 6, 12, 24, 48, and 72 h following addition of hemin to K562 cells, which clustered into eight distinct expression patterns. Relative expression levels are shown on the y-axis, with the corresponding time points on the x-axis. Each SOM cluster represents the average gene expression pattern for genes within the cluster, with the two outer lines indicating the standard deviation of expression. The number of cluster members is indicated in parentheses next to the cluster number.

 

View this table:
[in this window]
[in a new window]
 
Table 2. Selected differentially expressed genes in K562 cells after hemin treatment

 
Real-time RT-PCR confirmation of differential expression.
Independent confirmation of differential expression was assessed by real-time RT-PCR (Fig. 4) using selected members of the SOM-derived expression patterns (depicted in Fig. 3). FCGR2A (cluster 1) showed a 65% reduction in expression at 48 h, mirroring the array data. The protooncogene and transcriptional transactivator, MYB (cluster 2), was downregulated 75% by 48 h. Expression of erythroid Kruppel-like factor (6) (KLF1, cluster 4) was low and upregulated >2-fold between 6 and 12 h followed by downregulation at later time points. HSPA5 (grp78) (cluster 5), a heat shock 70-kDa family member and chaperone protein, showed >2-fold increase at 48 and 72 h. {alpha}-Globin (HBA, cluster 6) expression peaked (~3-fold) at 72 h. Cathepsin L (CTSL, cluster 7) showed a similar induction pattern with highest expression (~17-fold) at 48 and 72 h. {zeta}-Globin (HBZ, cluster 8) was also validated as being induced by hemin in K562 cells, showing maximal expression (~12-fold) at 48 and 72 h.



View larger version (32K):
[in this window]
[in a new window]
 
Fig. 4. Validation of a subset of differentially expressed genes and assessment of fetal and embryonic globin gene expression by real-time RT-PCR. Representative differentially expressed genes were chosen from the SOM clusters shown in Fig. 2. The fold change in expression (y-axis) is shown during the time course of hemin treatment (x-axis) as determined by real-time RT-PCR (open bars) or from microarrays (solid bars) with standard deviations derived from the averages of the 4 array comparisons. Gene names are shown, with GenBank accession numbers in parentheses. Although not detected as differentially expressed by microarray analysis, real-time RT-PCR analysis using gene-specific primer pairs detected an increase in steady-state levels of mRNA for both {gamma}-globin (bottom left) and {epsilon}-globin (bottom right).

 
Relative steady-state levels for embryonic and fetal globin mRNAs not identified as differentially expressed by microarray analysis were examined by real-time RT-PCR before and after hemin treatment. Interestingly, both {epsilon}- and {gamma}-globin mRNAs are upregulated >2-fold by 48 h with {epsilon}-globin showing 9-fold induction at 48 and 72 h (Fig. 4). When the microarrays were examined, {epsilon}-globin, with a low level of expression in untreated cells, was differentially expressed in some but not all of the four pair-wise comparisons, and so did not meet our criteria for differential expression. In contrast, the {gamma}-globin mRNA signal had the highest array intensity of all globin messages in untreated cultures, with <2-fold change in ratio (e.g., largest change at any time point was 1.8-fold up at 48 h), while RT-PCR analysis shows ~4-fold induction at 48 h. The level of ß-globin expression in untreated cells as assessed by microarray was low with a maximum of 2.3-fold induction at 48 h (Supplemental Table S2).

Transcriptional regulatory network analysis.
The nonredundant set of 322 differentially expressed genes (based on GenBank accession number) was computationally mapped to 311 unique UniGene clusters (DIFF311). TRNA was performed for the genes corresponding to these clusters using PAINT v3.0 ("promoter analysis and interaction network generation tool") (54) to identify biologically relevant transcription factor binding sites, referred to herein as transcription regulatory elements or "TREs," found in the regulatory regions of these genes.

As listed in Supplemental Table S1, 7,282 probe sets representing 5,184 unique genes were expressed in untreated cells (designated U5184). We found that 227 genes of DIFF311 were also detectable in untreated cells (DIFF227). The remaining 84 differentially expressed genes were not expressed in untreated cells (DIFF84). Therefore, TRNA was done as follows: 1) DIFF227 was compared with U5184, and 2) DIFF84 was compared with the entire microarray (A12432). Comparison of the differentially expressed gene set (DIFF227) to those expressed in untreated cells (U5184) affords a filter, since only 32% of arrayed probe sets are represented in these cells. This comparison will most likely identify TREs involved in erythroid commitment in these cells following hemin treatment. The remaining 84 differentially expressed genes (DIFF84) were not expressed in untreated cultures and therefore could not be similarly analyzed and were compared with the total array (A12432).

A total of 4,862 promoter entries were found for the U5814 set. TRNA identified a total of 213 TREs as present, at least once, in promoters from DIFF227 out of 456 motifs tested. TREs identified as significantly enriched in DIFF227 compared with the U5184 set (P < 0.1) and occurring in at least 5% of promoters are listed in Table 3A, with a graphical representation of sample TRE distributions in Fig. 5. The DIFF84 set was represented by 85 promoter entries for 81 genes. Significantly enriched TREs in the 84 genes turned on by hemin treatment are listed in Table 3B. Only TREs with both a P value <0.1 and an occurrence rate >5% are presented in the DISCUSSION.


View this table:
[in this window]
[in a new window]
 
Table 3. TREs significantly enriched in the differentially expressed gene sets

 


View larger version (32K):
[in this window]
[in a new window]
 
Fig. 5. Probability distributions of transcription regulatory element (TRE) frequency. The P value for enrichment of a TRE is the area under the curve to the right of the corresponding TRE frequency (solid vertical line). The dashed vertical line represents the TRE frequency corresponding to the P value threshold of 0.1. In each of the representative plots, if the frequency of a TRE observed in L227 (dashed line) is to the right of the solid line (as in AC), then its P value for enrichment will meet the threshold of <0.1 and be considered significantly enriched. If the frequency of the TRE is to the left of the solid line (as in D), then its frequency falls within the expected distribution for that TRE, and it is not significantly enriched.

 
We were interested to determine whether there was enrichment of TREs unique to each of the eight expression clusters resulting from our SOM analysis. PAINT v3.0 is capable of utilizing "class membership" information as part of its analyses. Groupings of the most significantly enriched TREs by cluster are shown in Table 4, where highlighted P values represent statistically significant enrichments in a cluster relative to the appropriate reference (U5184 and A12423, in the case of DIFF227 and DIFF84, respectively), suggesting coordinate regulation by a factor or factors that interact with that TRE. A number of statistically significant TREs were identified by this analysis. Several of interest are V$OCT1_Q6 (P value = 0.0191, cluster 7), V$GATA3_03 (P value = 0.0016, cluster 5; and P value = 0.0257, cluster 7), V$AML1_Q6 (P value = 0.0018, cluster 8), V$BACH2_01 (P value = 0.0486, cluster 6), V$MAZ_Q6 (P = 0.0196, cluster 5; and P = 0.0199, cluster 6), and V$EVI1_04 (P value = 0.048, cluster 3). Interestingly, all of the 29 5'-flanking regions in cluster 8 contain V$AML1_Q6, a TRE for AML-1. A graphical representation is shown in Fig. 6, where the negative log of the P value is plotted across the eight clusters with the fraction of 5'-flanking regions in each cluster containing the TRE indicated above the bar.


View this table:
[in this window]
[in a new window]
 
Table 4. Significantly enriched TREs in different expression pattern clusters

 


View larger version (19K):
[in this window]
[in a new window]
 
Fig. 6. Expression pattern clusters contain overrepresented TREs. Shown are six significant TREs generated with PAINT v3.0 by comparing 5'-flanking regions from DIFF227 in each of the eight expression clusters to those genes expressed in untreated K562 cells (U5184). The negative log of the P value is shown (y-axis) plotted against cluster number (x-axis) with fraction of promoters containing the TRE shown above the bar, when significant (P < 0.1).

 

    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
Microarray expression patterns for selected genes after hemin treatment.
Differentiation toward the erythroid lineage in hemin-treated K562 cells was evidenced by upregulation in mRNAs characteristic of erythroid cells [erythrocyte pyruvate kinase (PKLR) induced >3-fold at 48 h; erythrocyte glycophorin A (GYPA) upregulated 3-fold at 24 h] coupled with the downregulation of transcripts characteristic of megakaryocytes [glycoprotein 1b (GP1BA) and Fc{gamma}RIIa (FCGR2A) downregulated 3-fold at 72 h and >2-fold at 48 h, respectively].

In addition, genes involved in leukemogenesis and proliferation were affected, including the apoptosis inhibitor survivin (BIRC5) (49) (downregulated >2-fold at 72 h) and the apoptosis inducer TAT interactive protein-2 (HTATIP2) (57) (upregulated 11-fold at 72 h). Cathepsin H (CTSH) and protein tyrosine phosphatase PTP4A3, both of which are overexpressed in leukemic cell lines (35, 43), were downregulated 3-fold at 72 h. mRNAs for cathepsins B, D, and L are upregulated (see CTSL, Fig. 4), consistent with increased enzymatic activity for cathepsins B and D in K562 cells following hemin treatment (35). Also, IL-8 (cluster 6, Table 2), a recently described substrate of cathepsin L (39), is dramatically upregulated more than 50-fold at 48 h. Interestingly, suppressor of cytokine signaling 2 (SOCS2), which is overexpressed in CML (48), was downregulated upon hemin treatment (>2-fold at 72 h). LM02, the cysteine-rich, LIM domain-containing protein rhombotin 2, which is critical to early hematopoiesis (58) and previously found to be differentially expressed in K562 cells treated with sodium butyrate (44), is downregulated (>2-fold at 72 h) in our study.

Several genes known to play a role in response to oxidative stress or heavy metals were differentially expressed. These include the genes for metallothionein 1G and 2A and NAD(P)H dehydrogenase, quinone 1 (upregulated 2.4-fold at 6 h, 2.7-fold at 12 h, and 7.4-fold at 48 h, respectively). Several genes involved in heme biosynthesis showed differential expression upon hemin treatment, including the genes for ferrochetalase (upregulated >2-fold at 24 h) and coproporphyrinogen oxidase (downregulated >2-fold at 48 h). In agreement with other studies of hemin treatment of K562 cells (44), the gene for ferritin was differentially expressed (heavy chain upregulated >4-fold at 48 h). The gene for chaperone protein HSPA5/grp78, implicated in the control of apoptosis through interaction with caspase 7 (45) was upregulated >3-fold at 72 h.

Differential expression of transcriptional regulators.
Expression of transcriptional regulators was also affected by hemin treatment, including several molecules that play a role in globin-gene regulation. The genes for CCAAT/enhancer binding proteins C/EBP-ß and -{gamma} (59) were induced (>4-fold and >2-fold at 72 h, respectively). Previous studies showed moderate overexpression of C/EBP-{gamma} in the erythroid lineage in mice containing the human ß-like globin cluster transgene led to increased expression of {gamma}-globin relative to ß-globin (59). V-maf, homolog G, which binds maf recognition elements (MAREs) such as those present in the ß-globin locus control region HS2 enhancer (21), was upregulated early, peaking at >3-fold at 72 h. Interestingly, maf "activity" is increased by heme binding to the transcriptional repressor Bach1 (38). Heme binding inhibits formation of hetero-oligomers of Bach1 and maf-family proteins, thereby effectively increasing levels of maf-family proteins for dimerization with transcriptional activators like those possibly required for increased embryonic/fetal globin expression. Of note, Bach1 is expressed in hematopoietic cells and has MARE-binding activity in MEL cells (23). Paradoxically, nuclear transcription factor-Y {alpha}, another protein known to bind the CCAAT box of {gamma}-globin (13), was downregulated upon hemin treatment (>2-fold downregulated at 72 h). Gfi-1B, the zinc finger-containing transcriptional repressor recently found to be involved in regulation of erythroblast proliferation (38), is downregulated by hemin (>2-fold at 72 h). Also of interest is the finding that MCM5, a protein recently shown to play a role in transcriptional repression (14), is upregulated eightfold in K562 cells after 72 h of hemin treatment.

Comparison of microarray and real-time RT-PCR analysis data.
In general, these results comparing microarray vs. real-time RT-PCR data are in agreement in the relative expression patterns of those genes tested. It is not unusual to see differences in absolute fold change in expression comparing microarray and RT-PCR data; however, the overall expression pattern (i.e., up or down over time) was similar. The discrepancy between array and RT-PCR data for {epsilon}- and {gamma}-globin transcripts highlights the limitation of Affymetrix microarray analysis for studies of gene families. Since these microarrays use the average intensities at multiple spots with distinct, not necessarily gene-specific oligonucleotides derived from the mRNA, gene expression studies of members of highly homologous gene families (e.g., globin) require use of other methods such as gene-specific real-time RT-PCR.

Regulatory regions of differentially expressed genes provide insight into potential coordinate regulation.
PAINT analysis identified binding sites for transcription factors of known importance in hematopoiesis, erythropoiesis, and megakaryopoiesis (GATA-1, AML-1, GATA-2, Lmo2, Evi-1, NF-E2) (27, 56). The DIFF84 TRE list includes transcription factors identified as differentially expressed genes (i.e., GATA-2), as well as several known to be involved in hematopoiesis, such as NF-{kappa}B (12) and CREB (34). Interpretation of TRNA data is complicated by several factors. First, P values indicate enrichment relative to the large reference set. Thus a TRE can be significantly enriched yet still rare within a cluster. Second, some TREs are found at high frequency in the genome, making enrichment difficult to discern. Finally, sample sizes are small (<10) for some clusters, further weakening inference. Thus the desired result would be a TRE significantly enriched with a high frequency in a given cluster. An example of such a result is AML1, a TRE present in all 5'-flanking regions of cluster 8 members. Our microarray data show low expression of AML1 at all time points. AML1 is expressed in K562 cells and implicated in erythroid and megakaryocytic differentiation (15, 37); however, it is not known if AML1 activity is regulated transcriptionally or posttranscriptionally. Some care must be taken in interpretation of the enrichment of the AML1 TRE in this analysis, since the TRE for AML1 is short, consisting of only six nucleotides. Although AML1 sites would be predicted to be common in the genome, cluster 8 still shows an enrichment greater than what would be found by chance, as is depicted in Fig. 5. A complete graphical representation is available (Supplemental Fig. S1).

The finding that V$BACH2_01, the TRE for the transcriptional repressor Bach2, is significant in cluster 5 is intriguing, since this cluster behaves differently than others in that gene members are upregulated late. The TREs for Bach1 and Bach2 share a core of 7 of 11 bases, suggesting overlapping functions. Bach2 is known to interact with small Maf transcription factors and with MAZR to alter gene expression (26), and interestingly, Bach2 expression is regulated by BCR-ABL (55). Recent studies show that Bach2 expression is induced by oxidative stress, and Bach2 protein accumulates in the nucleus and is associated with inhibition of proliferation and apoptosis (36). Our findings that hemin induction enriches for genes with Bach-binding elements in their 5'-flank and that Bach factors are not differentially expressed following hemin treatment support a model in which Bach1/2 is a transcriptional repressor whose activity is diminished upon hemin induction. This could occur by direct heme binding to Bach1/2, inhibiting its ability to dimerize with maf-like proteins (36) as well as by enrichment for genes with Bach1/2 binding sites. Interestingly, recent studies show that heme positively regulates ß-globin gene expression in MEL cells by blocking Bach1/MARE interactions in the locus-control region (52). Furthermore, changing the dimerization partner of MafK from Bach1 to NF-E2p45 is critical for globin expression in MEL cells (7).

It is noteworthy that TREs for several candidate factors not known to be expressed in hematopoietic cells were found to be significantly enriched in this analysis (see Table 4). Examples include HNF1 and HNF4 (clusters 3 and 2, respectively) expressed in liver (10), Hand1/E47 (cluster 1), involved in heart and neural crest development (18), and Pax-4 (clusters 1 and 3), expressed in pancreatic B cells (24). There are several possible explanations for these findings: 1) the result represents a false positive; 2) these genes may play a role completely unrelated to hematopoiesis that is regulated using these TREs; or, 3) there are as yet unidentified factors expressed in hematopoietic cells that utilize these TREs.

Our findings using the PAINT v3.0 software to examine 5'-flanking regions of differentially expressed genes suggest that factors binding to the identified TREs may be influential in the expression pattern of a particular cluster. Our findings of distinct clusters for the differentially expressed genes may be useful in investigations probing for functional relationships between genes in each cluster, genes in different clusters, and their timing of expression. Further studies of promoter occupancy and transcription factor perturbation are now required to functionally validate these sites and the factors that bind to them.

Hemin-induced hemoglobinization in K562 cells.
In this report, changing steady-state mRNA profiles are characterized in K562 cells before and after varying times of hemin treatment. Several questions emerge as to the effects of hemin on globin induction in these cells. For example, our microarray analyses agree with real-time RT-PCR data of others (50) showing that steady-state {gamma}-globin mRNA levels are high and do not appear to increase >2-fold with hemin treatment of K562 cells. Our real-time RT-PCR results, however, show an ~4-fold increase in {gamma}-globin mRNA at 48 h (Fig. 4).

It is unclear whether {gamma}-globin mRNA is translated maximally in untreated cells. If translated, then do {gamma}4 tetramers form? And, if so, why is there minimal benzidine staining in untreated cells? One possibility is that heme may be rate limiting; and, a specific translational block may be in effect for globin synthesis thereby protecting cells against accumulation of unstable and nonfunctional apoglobin tetramers (22). In support of this, our results show the gene for heat shock 70-kDa protein HSPA1B (cluster 7), involved in protein stability, is induced >10-fold at 72 h. HSPA1A is known to interact with the gene for NAD(P)H dehydrogenase, quinone 1 (2), and expression of both peaks at 48 h. In addition, translational (5) and posttranscriptional (25) controls relating to butyrate-, hydroxyurea-, and 5-azacytidine-induced appearance of hemoglobins in K562 cells have been proposed.

A second possibility for lack of appreciable benzidine staining in untreated K562 cells is that {gamma}-globin mRNA is translated but degraded in the absence of free heme and/or its normal hetero-chain partners {alpha}- and/or {zeta}-globin. Alternatively, apo {gamma}4 tetramers indeed might form in untreated cells, but in the absence of heme, do not stain with benzidine. Accumulation of these heme-lacking tetramers in high concentrations might be unstable and precipitate. However, if apoglobin tetramers were present, then the timing for appearance of benzidine-positive cells might be expected to occur sooner than the observed 48 h, since heme derived directly from hemin can be inserted in vitro into apoglobins (1). Interestingly, expression of ferrochelatase, which catalyzes chelation of ferrous iron and protoporphyrin to form heme (22), is upregulated during hemin treatment in our studies (Table 2). Taken together, these results suggest that heme may be rate limiting for benzidine staining in untreated K562 cells. Interestingly, previous studies indicate heme biosynthesis is rate limiting in early stages of differentiation of MEL cells (17). Others have shown that heme-containing {gamma}4 tetramers and/or embryonic and fetal globin heterotetramers are detected by benzidine staining upon treatment with hemin (5, 47). Our results show steady-state {zeta}-globin mRNA and to a lesser extent {alpha}-globin mRNA are upregulated after treatment (Fig. 4). Previous radiolabeled chain-synthesis studies demonstrated synthesis of radiolabeled homo- and heterotetramers in untreated K562 cells (42), suggesting that lack of appreciable benzidine staining prior to treatment may be in part due to absence of heme-containing chains/tetramers; or alternatively, absolute amounts of tetramers may be rate limiting, thereby precluding their detection by benzidine staining. Further studies are required to identify rate-limiting steps for detection of benzidine-positive globins in untreated compared with treated K562 cells, such as detection of apoglobin chains and tetramers in untreated cells, as well as studies on determining whether globin mRNAs are maximally translated in these cells.

Our results also show that the total number of different probe sets expressed as assessed by microarray monitoring of the steady-state mRNA population remains about the same comparing untreated to 72-h treated cultures. In addition, we found that the majority of genes are expressed at low levels and that the percentage of expressed probe sets in different arbitrarily binned intensity classes does not change appreciably between untreated and 72-h treated cultures. This implies that hemin treatment of these cells leads to minimal overall complexity change in the total expressed mRNA population. Earlier estimates indicate that 85% of the total mRNA mass in a cell is contained within the complex mRNA class (e.g., mRNAs present in 1–5 copies per cell) (11). Deciphering the molecular control effecting changes in steady-state mRNA abundance and the relationship between transcriptomes and proteosomes are central to our understanding of development and differentiation.

These studies address one aspect of the information defining overall cell phenotype and response to inducer (e.g., steady-state mRNA levels). Translational and posttranslational events have not been evaluated here. Protein studies are needed to evaluate the relationship between changes in the transcriptome and the cell’s proteome. Notwithstanding these caveats, characterization of K562 transcriptomes with and without hemin should facilitate studies on defining the molecular basis of leukemogenesis. In addition, transcriptome information will aid in our understanding of the lineage commitment from a bipotential cell to an erythroid cell and help define the molecular basis of increased hemoglobin expression in these cells. This information will be important for the identification of rate-limiting changes required for upregulation of fetal and embryonic globins in these cells and may identify new targets for therapeutic intervention in SCD and ß-thalassemia. Selection of such targets might involve study of the differentially expressed genes we identified involved in signaling, chromatin remodeling, and transcriptional activation or repression, followed by studies of their overexpression or knockdown in erythroid progenitors while monitoring effects on fetal globin expression. The differentially expressed genes identified with hemin-induced commitment now can be compared with those from other inducers to delineate common or distinct mechanisms for upregulating fetal globin expression.


    GRANTS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
This work was supported in part by National Heart, Lung, and Blood Institute Grant 1-RO1-HL-69256-01 (to S. Surrey), by the Cardeza Foundation for Hematologic Research (to S. Surrey and M. A. Keller), DARPA BioComp Initiative, contract number F30602-01-2-0578 (to G. E. Gonye and R. Vadigepalli), and by National Institutes for General Medical Sciences Grant P20-MH-64459-01 (to G. E. Gonye and R. Vadigepalli).


    ACKNOWLEDGMENTS
 
We thank Paolo Fortina (Genomics Core Facility, Center for Translational Medicine, Jefferson Medical College) for contributions to the microarray experiments. We thank Drew Likens (Cardeza) for assistance with figure preparation, Praveen Chakravarthula for computational expertise, and Steven McKenzie (Jefferson Medical College) for input, encouragement, and support of this work.


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

Address for reprint requests and other correspondence: S. Surrey, Cardeza Foundation for Hematologic Research and Division of Hematology, Dept. of Medicine, Jefferson Medical College, Thomas Jefferson Univ., Curtis Bldg., Rm. 703, 1015 Walnut St., Philadelphia PA 19107 (E-mail: saul.surrey{at}jefferson.edu).

1 The Supplementary Material for this article (Supplemental Tables S1 and S2 and Supplemental Fig. S1) is available online at http://physiolgenomics.physiology.org/cgi/content/full/00028.2004/DC1. Back


    REFERENCES
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 

  1. Adachi K, Zhao Y, and Surrey S. Effects of heme addition on formation of stable human globin chains and hemoglobin subunit assembly in a cell-free system. Arch Biochem Biophys 413: 99–106, 2003.[CrossRef][ISI][Medline]
  2. Anwar A, Siegel D, Kepa JK, and Ross D. Interaction of the molecular chaperone Hsp70 with human NAD(P)H:quinone oxidoreductase 1. J Biol Chem 277: 14060–14067, 2002.[Abstract/Free Full Text]
  3. Applied Biosystems. Relative Quantitation of Gene Expression. User bulletin no. 2, ABI PRISM 7700 Sequence Detection System. Applied Biosystems, 1997.
  4. Atweh GF, DeSimone J, Saunthararajah Y, Fathallah H, Weinberg RS, Nagel RL, Fabry ME, and Adams RJ. Hemoglobinopathies. Hematology (Am Soc Hematol Educ Program): 14–39, 2003.
  5. Benz EJ Jr, Murnane MJ, Tonkonow BL, Berman BW, Mazur EM, Cavallesco C, Jenko T, Snyder EL, Forget BG, and Hoffman R. Embryonic-fetal erythroid characteristics of a human leukemic cell line. Proc Natl Acad Sci USA 77: 3509–3513, 1980.[Abstract]
  6. Bieker JJ. Isolation, genomic structure, and expression of human erythroid Kruppel-like factor (EKLF). DNA Cell Biol 15: 347–352, 1996.[ISI][Medline]
  7. Brand M, Ranish JA, Kummer NT, Hamilton J, Igarashi K, Francastel C, Chi TH, Crabtree GR, Aebersold R, and Groudine M. Dynamic changes in transcription factor complexes during erythroid differentiation revealed by quantitative proteomics. Nat Struct Mol Biol 1: 73–80, 2004.
  8. Bunn HF and Forget BG. Hemoglobin: Molecular, Genetic and Clinical Aspects. Philadelphia: Saunders, 1986.
  9. Burton GR, Guan Y, Nagarajan R, and McGehee RE Jr. Microarray analysis of gene expression during early adipocyte differentiation. Gene 293: 21–31, 2002.[CrossRef][ISI][Medline]
  10. Costa RH, Kalinichenko VV, Holterman AX, and Wang X. Transcription factors in liver development, differentiation, and regeneration. Hepatology 38: 1331–1347, 2003.[CrossRef][ISI][Medline]
  11. Davidson E. Gene activity in early development. In: RNA Sequence Complexity and Structural Gene Transcription in Early Embryos (2nd ed.). New York: Academic, 1976, p. 189–236.
  12. Denk A, Wirth T, and Baumann B. NF-kappaB transcription factors: critical regulators of hematopoiesis and neuronal survival. Cytokine Growth Factor Rev 11: 303–320, 2000.[CrossRef][ISI][Medline]
  13. Duan Z, Stamatoyannopoulos G, and Li Q. Role of NF-Y in in vivo regulation of the {gamma}-globin gene. Mol Cell Biol 21: 3083–3095, 2001.[Abstract/Free Full Text]
  14. Dziak R. Evidence for a role of MCM (mini-chromosome maintenance)5 in transcriptional repression of sub-telomeric and Ty-proximal genes in Saccharomyces cerevisiae. J Biol Chem 278: 27372–27381, 2003.[Abstract/Free Full Text]
  15. Elagib KE, Racke FK, Mogass M, Khetawat R, Delehanty LL, and Goldfarb AN. RUNX1 and GATA-1 coexpression and cooperation in megakaryocytic differentiation. Blood 101: 4333–4341, 2003.[Abstract/Free Full Text]
  16. Elkon R, Linhart C, Sharan R, Shamir R, and Shiloh Y. Genome-wide in silico identification of transcriptional regulators controlling the cell cycle in human cells. Genome Res 13: 773–780, 2003.[Abstract/Free Full Text]
  17. Fadigan A and Dailey HA. Inhibition of ferrochelatase during differentiation of murine erythroleukaemia cells. Biochem J 243: 419–424, 1987.[ISI][Medline]
  18. Firulli AB. A HANDful of questions: the molecular biology of the heart and neural crest derivatives (HAND)-subclass of basic helix-loop-helix transcription factors. Gene 312: 27–40, 2003.[CrossRef][ISI][Medline]
  19. Forget BG. Molecular basis of hereditary persistence of fetal hemoglobin. Ann NY Acad Sci 850: 38–44, 1998.[Abstract/Free Full Text]
  20. Gianni AM, Presta M, Polli E, Peschle C, Lettieri F, Saglio G, Comi P, Giglioni B, and Ottolenghi S. Preferential induction of fetal versus embryonic globin chains in human leukemic cell lines. Leuk Res 6: 155–163, 1982.[CrossRef][ISI][Medline]
  21. Gui CY and Dean A. A major role for the TATA box in recruitment of chromatin modifying complexes to a globin gene promoter. Proc Natl Acad Sci USA 100: 7009–7014, 2003.[Abstract/Free Full Text]
  22. Han AP, Yu C, Lu L, Fujiwara Y, Browne C, Chin G, Fleming M, Leboulch P, Orkin SH, and Chen JJ. Heme-regulated eIF2alpha kinase (HRI) is required for translational regulation and survival of erythroid precursors in iron deficiency. EMBO J 20: 6909–6918, 2001.[Abstract/Free Full Text]
  23. Igarashi K, Hoshino H, Muto A, Suwabe N, Nishikawa S, Nakauchi H, and Yamamoto M. Multivalent DNA binding complex generated by small Maf and Bach1 as a possible biochemical basis for beta-globin locus control region complex. J Biol Chem 273: 11783–11790, 1998.[Abstract/Free Full Text]
  24. Kalousova A, Benes V, Paces J, Paces V, and Kozmik Z. DNA binding and transactivating properties of the paired and homeobox protein Pax4. Biochem Biophys Res Commun 259: 510–518, 1999.[CrossRef][ISI][Medline]
  25. Keefer JR, Schneidereith TA, Mays AC, Purvis SH, Dover GJ. Differential regulation of Hb F induction in K562 cells and human CD34+ erythrocyte cultures: the possible role of cyclic nucleotide dependent signaling pathways (Abstract). Blood 102: 267, 2003.
  26. Kobayashi A, Yamagiwa H, Hoshino H, Muto A, Sato K, Morita M, Hayashi N, Yamamoto M, and Igarashi K. A combinatorial code for gene expression generated by transcription factor Bach2 and MAZR (MAZ-related factor) through the BTB/POZ domain. Mol Cell Biol 20: 1733–1746, 2000.[Abstract/Free Full Text]
  27. Koury MJ, Sawyer ST, and Brandt SJ. New insights into erythropoiesis. Curr Opin Hematol 9: 93–100, 2002.[CrossRef][ISI][Medline]
  28. Lampronti I, Bianchi N, Borgatti M, Fibach E, Prus E, and Gambari R. Accumulation of gamma-globin mRNA in human erythroid cells treated with angelicin. Eur J Haematol 3:189–95, 2003.[CrossRef]
  29. Liu G, Loraine AE, Shigeta R, Cline M, Cheng J, Valmeekam V, Sun S, Kulp D, and Siani-Rose MA. NetAffx: Affymetrix probesets and annotations. Nucleic Acids Res 31: 82–86, 2003.[Abstract/Free Full Text]
  30. Lockhart DJ, Dong H, Byrne MC, Follettie MT, Gallo MV, Chee MS, Mittmann M, Wang C, Kobayashi M, Horton H, and Brown EL. Expression monitoring by hybridization to high-density oligonucleotide arrays. Nat Biotechnol 14: 1675–1680, 1996.[ISI][Medline]
  31. Lozzio CB and Lozzio BB. Human chronic myelogenous leukemia cell-line with positive Philadelphia chromosome. Blood 45: 321–334, 1975.[Abstract]
  32. Lumelsky NL and Forget BG. Negative regulation of globin gene expression during megakaryocytic differentiation of a human erythroleukemic cell line. Mol Cell Biol 11: 3528–3536, 1991.[ISI][Medline]
  33. Matys V, Fricke E, Geffers R, Gossling E, Haubrock M, Hehl R, Hornischer K, Karas D, Kel AE, Kel-Margoulis OV, Kloos DU, Land S, Lewicki-Potapov B, Michael H, Munch R, Reuter I, Rotert S, Saxel H, Scheer M, Thiele S, and Wingender E. TRANSFAC: transcriptional regulation, from patterns to profiles. Nucleic Acids Res 31: 374–378, 2003.[Abstract/Free Full Text]
  34. Mora-Garcia P, Cheng J, Crans-Vargas HN, Countouriotis A, Shankar D, and Sakamoto KM. Transcriptional regulators and myelopoiesis: the role of serum response factor and CREB as targets of cytokine signaling. Stem Cells 21: 123–130, 2003.[Abstract/Free Full Text]
  35. Murnane MJ, Phipps PM, and Denekamp DS. Expression and induction of cathepsins B and D in K562 cells. Exp Hematol 16: 33–37, 1988.[ISI][Medline]
  36. Muto A, Tashiro S, Tsuchiya H, Kume A, Kanno M, Ito E, Yamamoto M, and Igarashi K. Activation of Maf/AP-1 repressor Bach2 by oxidative stress promotes apoptosis and its interaction with promyelocytic leukemia nuclear bodies. J Biol Chem 277: 20724–20733, 2002.[Abstract/Free Full Text]
  37. Niitsu N, Yamamoto-Yamaguchi Y, Miyoshi H, Shimizu K, Ohki M, Umeda M, and Honma Y. AML1a but not AML1b inhibits erythroid differentiation induced by sodium butyrate and enhances the megakaryocytic differentiation of K562 leukemia cells. Cell Growth Differ 8: 319–326, 1997.[Abstract]
  38. Ogawa K, Sun J, Taketani S, Nakajima O, Nishitani C, Sassa S, Hayashi N, Yamamoto M, Shibahara S, Fujita H, and Igarashi K. Heme mediates derepression of Maf recognition element through direct binding to transcription repressor Bach1. EMBO J 20: 2835–2843, 2001.[Abstract/Free Full Text]
  39. Ohashi K, Naruto M, Nakaki T, and Sano E. Identification of interleukin-8 converting enzyme as cathepsin L. Biochim Biophys Acta 1649: 30–39, 2003.[ISI][Medline]
  40. Osawa M, Yamaguchi T, Nakamura Y, Kaneko S, Onodera M, Sawada K, Jegalian A, Wu H, Nakauchi H, and Iwama A. Erythroid expansion mediated by the Gfi-1B zinc finger protein: role in normal hematopoiesis. Blood 100: 2769–2777, 2002.[Abstract/Free Full Text]
  41. Pace BS, Qian XH, Sangerman J, Ofori-Acquah SF, Baliga BS, Han J, and Critz SD. p38 MAP kinase activation mediates gamma-globin gene induction in erythroid progenitors. Exp Hematol 11:1089–1096, 2003.[CrossRef]
  42. Papayannopoulou T, Nakamoto B, Kurachi S, and Nelson R. Analysis of the erythroid phenotype of HEL cells: clonal variation and the effect of inducers. Blood 70: 1764–1772, 1987.[Abstract]
  43. Pettiford SM and Herbst R. The protein tyrosine phosphatase HePTP regulates nuclear translocation of ERK2 and can modulate megakaryocytic differentiation of K562 cells. Leukemia 17: 366–378, 2003.[CrossRef][ISI][Medline]
  44. Plonczynski M, Hardy CL, Safaya S, Harrell A, McCoy L, Brinson A, Agwarangbo L, and Steinberg MH. Induction of globin synthesis in K562 cells is associated with differential expression of transcription factor genes. Blood Cells Mol Dis 25: 156–165, 1999.[CrossRef][ISI][Medline]
  45. Reddy RK, Mao C, Baumeister P, Austin RC, Kaufman RJ, and Lee AS. Endoplasmic reticulum chaperone protein GRP78 protects cells from apoptosis induced by topoisomerase inhibitors: role of ATP binding site in suppression of caspase-7 activation. J Biol Chem 278: 20915–20924, 2003.[Abstract/Free Full Text]
  46. Richon V, Rifkind R, and Marks P. Differentiation of murine erythroleukemia cells (friend cells). In: Cell Biology: A Laboratory Handbook, edited by Celis JC. London: Academic, 1994, p. 213–217.
  47. Rutherford TR, Clegg JB, and Weatherall DJ. K562 human leukaemic cells synthesise embryonic haemoglobin in response to haemin. Nature 280: 164–165, 1979.[ISI][Medline]
  48. Schultheis B, Carapeti-Marootian M, Hochhaus A, Weisser A, Goldman JM, and Melo JV. Overexpression of SOCS-2 in advanced stages of chronic myeloid leukemia: possible inadequacy of a negative feedback mechanism. Blood 99: 1766–1775, 2002.[Abstract/Free Full Text]
  49. Shinozawa I, Inokuchi K, Wakabayashi I, and Dan K. Disturbed expression of the anti-apoptosis gene, survivin, and EPR-1 in hematological malignancies. Leuk Res 24: 965–970, 2000.[CrossRef][ISI][Medline]
  50. Smith RD, Malley JD, and Schechter AN. Quantitative analysis of globin gene induction in single human erythroleukemic cells. Nucleic Acids Res 28: 4998–5004, 2000.[Abstract/Free Full Text]
  51. Stamatoyannopoulos G and Nienhuis AW. Hemoglobin switching. In: The Molecular Basis of Blood Disease, edited by Stamatoyannopoulos G, Nienhuis AW, Majerus PW, and Varmus H. Philadelphia: Saunders, 1994, p. 107.
  52. Tahara T, Sun J, Nakanishi K, Yamamoto M, Mori H, Saito T, Fujita H, Igarashi K, and Taketani S. Heme positively regulates the expression of beta-globin at the locus control region via the transcriptional factor Bach1 in erythroid cells. J Biol Chem 7: 5480–54877, 2004.
  53. Tamayo P, Slonim D, Mesirov J, Zhu Q, Kitareewan S, Dmitrovsky E, Lander ES, and Golub TR. Interpreting patterns of gene expression with self-organizing maps: methods and application to hematopoietic differentiation. Proc Natl Acad Sci USA 96: 2907–2912, 1999.[Abstract/Free Full Text]
  54. Vadigepalli R, Chakravarthula P, Zak DE, Schwaber JS, and Gonye GE. PAINT: a promoter analysis and interaction network generation tool for gene regulatory network identification. OMICS 7: 235–252, 2003.[CrossRef][Medline]
  55. Vieira SA, Deininger MW, Sorour A, Sinclair P, Foroni L, Goldman JM, and Melo JV. Transcription factor BACH2 is transcriptionally regulated by the BCR/ABL oncogene. Genes Chromosomes Cancer 32: 353–363, 2001.[CrossRef][ISI][Medline]
  56. Warren LA and Rothenberg EV. Regulatory coding of lymphoid lineage choice by hematopoietic transcription factors. Curr Opin Immunol 15: 166–175, 2003.[CrossRef][ISI][Medline]
  57. Xiao H, Palhan V, Yang Y, and Roeder R. TIP30 has an intrinsic kinase activity required for up-regulation of a subset of apoptotic genes. EMBO J 19: 956–963, 2000.[Abstract/Free Full Text]
  58. Yamada Y, Warren AJ, Dobson C, Forster A, Pannell R, and Rabbitts TH. The T cell leukemia LIM protein Lmo2 is necessary for adult mouse hematopoiesis. Proc Natl Acad Sci USA 95: 3890–3895, 1998.[Abstract/Free Full Text]
  59. Zafarana G, Rottier R, Grosveld F, and Philipsen S. Erythroid overexpression of C/EBP{gamma} in transgenic mice affects {gamma}-globin expression and fetal liver erythropoiesis. EMBO J 19: 5856–5863, 2000.[Abstract/Free Full Text]
  60. Zhou W, Clouston DR, Wang X, Cerruti L, Cunningham JM, and Jane SM. Induction of human fetal globin gene expression by a novel erythroid factor, NF-E4. Mol Cell Biol 20: 7662–7672, 2000.[Abstract/Free Full Text]