Global Genomic Approaches to the Iron-Regulated Proteome

Available online at www.annclinlabsci.org 230 Annals of Clinical & Laboratory Science, vol. 35, no. 3, 2005

Overview: Global Genomic Approaches to the Iron-Regulated Proteome

Ying Liu, Zvezdana Popovich, and Douglas M. Templeton Department of Laboratory Medicine and Pathobiology, University of Toronto,Toronto, Canada

Abstract. Iron interacts with cells to regulate the proteome through complex effects on gene expression. In simple organisms such as bacteria and yeast, intra- and extra-cellular iron influences gene expression through defined signal transduction pathways. In higher organisms, effects are probably mediated at the transcriptional level through secondary effects of reactive oxygen species, while post-transcriptional effects operate through well-defined pathways involving iron-regulatory proteins. To investigate the impact of iron levels on gene expression and the proteome, approaches such as differential display and subtractive hybridization have the advantage of surveying the entire geneome. However, they are technically demanding and have given way to microarray techniques. To date, numerous microarray experiments with various organisms have not yielded any definitive picture of the role of iron. Common themes throughout such studies are that both iron excess and iron depletion influence expression of proteins related to energy metabolism, cell proliferation, matrix structure, and the metabolism of iron itself. That no consistent set of genes is involved from one study to the next probably results both from the uncertainties inherent in the technique and the biological variability of the systems under study. We briefly describe two types of iron-dependent microarray experiments from our laboratory to examine major cellular targets of iron toxicity. Using Affymetrix oligonucleotide arrays with cardiac cells, we found several hundred genes whose mRNA levels were affected by iron, including an increase in several genes responding to oxidative stress and a decrease in several kinases and phosphatases. In a simpler experiment using a human liver cell line with a limited cDNA array, we targeted 13 genes affected by iron chelation. Metabolic pathway analysis shows links of 5 of these through phorbol ester responsiveness, and additional links through prostaglandin E2. We conclude that definitive understanding of the complex iron-regulated proteome requires global gene approaches and rigorous interlaboratory standardization. (received 22 April 2005; accepted 28 April 2005)

Keywords: gene expression, microarray, genome, iron overload, iron chelation, iron-responsive elements

Introduction synthesis to maintain the blood’s oxygen-carrying capacity. On the other hand, iron’s redox properties Iron is required by living organisms for a variety of contribute to its toxicity. Through participation in purposes related to its favourable redox properties Fenton chemistry, iron produces reactive oxygen and rich coordination chemistry. Thus, in the species (ROS) that are harmful to biological human organism, it is essential for the mitochondrial molecules. And replete iron stores also increase our electron transport chain, as part of cytochromes and susceptibility to infection; many opportunistic iron-sulfur proteins, and for nucleic acid synthesis organisms rely on the host to supply precious iron. as a cofactor in ribonucleotide reductase. The body As might be expected of such a bioactive metal, iron utilizes about 20 mg of Fe per day for hemoglobin also influences cellular responses and phenotype through diverse effects on gene expression. Our Address correspondence to Douglas M. Templeton M.D., purpose here is to survey various studies that have Ph.D., Department of Laboratory Medicine and Pathobiology, taken a broad genome/proteome-based approach to Medical Sciences Building Rm 6302, University of Toronto, 1 understanding the role of iron in orchestrating King’s College Circle, Toronto M5S 1A8, Canada; tel 416 978 3972; fax 416 978 5959; e-mail [email protected]. cellular biochemistry.

02 Liu 230-239 230 7/20/05, 1:11 PM Genomic approaches to iron regulation 231

Major steps in understanding iron metabolism TonB in E. coli. In situations of low iron, the Fec were made over the last decade [1]. A specific role proteins are transcribed following removal of Fur for iron in regulating genes involved in its own repression, but the transport system is inactive. metabolism and trafficking at a post-transcriptional When a low concentration of ferric dicitrate is level [2-4] became textbook examples of such a mode present in the cell’s vicinity, it is sensed by binding of regulation. A perhaps reawakened interest in the to FecA in the outer membrane. This leads to metal saw classical molecular biological approaches interaction of FecA with a protein complex that such as differential display and subtraction cloning includes FecR and TonB on the inner membrane, used to address more global effects of iron [5,6]. and this induces a conformational change in FecR. Recently, microarray technology has dominated such FecR then interacts with FecI in the cell, activating efforts. But iron remains somewhat enigmatic. it to bind to the FecA promoter, recruit RNA Multiple effects on gene expression have been polymerase, and drive transcription of genes of the proposed but remain hard to categorize in terms of Fec iron transport system (FecA, FecB, FecC, etc.) a unified response [7]. And, in view of its potential that are contiguously arranged on the chromosome. effects on multiple signaling mechanisms, iron This leads to increased iron uptake through FecA, overload, in cultured cells at least, often has which is thus both a regulator and a transporter. surprisingly little effect. Distinguishing acute toxic Again, this system is repressed by Fur at high iron responses from adaptive ones is a goal of such studies. concentrations. A number of other bacteria express The lack of a dramatic effect of iron overload components homologous to those of the Fec and on gene expression notwithstanding, iron does Fur systems, and signal transduction appears to be influence a variety of cellular processes. We have a general theme in iron-dependent bacterial gene recently reviewed effects that go beyond regulation regulation. Another good example is the PmrA/ of genes related to iron metabolism itself (operating PmrB system of Salmonella that responds to extra- at both transcriptional and translational levels) to cellular Fe3+ to activate a regulon involved in iron genes grouped as affecting oxidative stress responses resistance [9]. (eg, glutathione peroxidase, heme oxygenase-1, metallothionein); tissue fibrosis (eg, collagen, TGF- Transcriptional and Post-transcriptional Regulation β); energetics of metabolism (eg, aldolase, lactate dehydrogenase); and cell cycle control (eg, retino- In general, transcriptional regulation by iron in blastoma protein, p21, various cyclins) [7]. higher animals is poorly understood. Generation of ROS by iron may be a general means of gene Signaling Evoked by Iron regulation through the ROS-activated transcription factor NF-κB [10,11]. For instance, an NF-κB Studies with bacteria have linked iron with signal binding site is found in the ferritin H chain gene transduction mechanisms that terminate in promoter region [12]. ROS also lead to lipid per- regulating gene expression, and these may prove to oxidation. Products of lipid peroxidation in turn be instructive for understanding iron in higher activate transcription factors Sp1 and Sp3, account- organisms. The E. coli Fur protein has homologues ing at least in part for an increased expression of in many bacteria. It can act directly as a transcript- a1(I) collagen [13,14]. On the other hand, hypoxia- ional repressor when Fe2+ is present as a cofactor, like responses play a central role in gene regulation and turns off genes related to iron uptake. When by iron chelators [7]. Hypoxia-inducible factor iron is scarce, then, derepression alone can increase (HIF-1α) is a transcription factor that binds to a synthesis of iron transport proteins and enzymes of hypoxia-responsive element (HRE) in a number of siderophore biosynthesis [8]. However, in other target genes. The iron chelator deferoxamine (DFO) systems, positive regulation is required for synthesis mimics the effects of hypoxia on a number of genes of the iron transport systems. A case in point is [15,16] including erythropoietin [17] and the ferric citrate signaling through the Fec proteins and transferrin receptor [18]. A possible mechanism

02 Liu 230-239 231 7/20/05, 1:11 PM 232 Annals of Clinical & Laboratory Science, vol. 35, no. 3, 2005

involves the Fe(II)/2-oxoglutarate-dependent dioxy- structure with a base-paired stem of about 10 base genase, which hydroxylates critical proline and pairs in length, a central unpaired loop of sequence arginine residues in HIF-1α under normoxic and CAGUG, and a conserved C, 5 bases upstream of iron-replete conditions, targeting HIF-1α for this loop, that forms an unpaired bulge in the stem degradation by the proteasome [19,20]. and is necessary for protein binding. IREs in the 5' Post-transcriptional regulation in eukaryotes of UTR of transcripts occur in single copies, and are several genes controlling iron metabolism and energy found in ferritin, ferroportin, m-aconitase, succinate utilization is better understood. It involves iron- dehydrogenase, and δ−ALAS mRNAs. Their regulatory elements (IRE) in mRNA, and associated binding to IRPs is increased by iron deficiency, and IRE-binding iron-regulatory proteins (IRP) when they are occupied, progression of the [2,3,21,22]. Iron-related genes containing IREs translational machinery is blocked. IREs in the 3' include those coding for ferritin, transferrin receptor, UTR of TfR and DMT-1 mRNAs are not positioned divalent metal transporter-1 (DMT-1), and ferro- to block translation. Rather, they form complexes portin. In addition, IREs are found in the transcripts with IRPs to stabilize the RNA against degradation. for the control enzyme of heme synthesis, δ- The transferrin receptor mRNA contains 5 copies aminolevulinate synthase (δ-ALAS), and the of IRE [3,27], while the single IRE in DMT-1 tricarboxylic acid cycle enzymes, mitochondrial mRNA appears to participate in a more complicated aconitase and succinate dehydrogenase. regul-atory mechanism [28,29]. It is teleologically Two IRPs are known [3,4], designated IRP1 and apparent why the iron storage protein ferritin would IRP2. IRP1 contains a [4Fe-4S] iron-sulfur cluster decrease and the transferrin receptor increase in that resides in a cleft between 2 domains of the conditions of low iron. protein. It is a homologue of mitochondrial aconitase, and in the holo form is a cytosolic Gene Expression Screening aconitase that lacks RNA binding activity. The [4Fe- 4S] cluster in m-aconitase contains a labile iron atom Because of the wide variety of genes that are and is susceptible to conversion to [3Fe-4S] [23,24]. affected by iron or iron chelation, it is desirable to Loss of iron may lead to cluster disassembly, allowing use genome-wide screening of responses to derive the protein to open up into an RNA binding form. general patterns in particular experimental or Thus, IRP1 is a bifunctional protein. When the pathological circumstances. Earlier efforts have now iron-sulfur cluster is intact it exhibits aconitase largely given way to studies with DNA microarrays activity; in iron-deficient conditions, IRE binding (see below). In general, microarrays offer rapid and occurs. IRP2 is a homologue of IRP1 that lacks an simple screening of potentially thousands of genes, iron-sulfur cluster and contains an additional 73- whereas the earlier methods are more laborious and amino-acid motif [3,21]. The 73-amino-acid technically demanding. However, it should be noted sequence confers susceptibility to iron-dependent that microarrays can only provide information on oxidative damage [25], subsequently involving the genes chosen for inclusion on the gene chip, degradation by the proteasome [26]. Thus, IRP2 whereas other screening methods can potentially RNA binding activity, like that of IRP1, is increased provide information on all genes affected, whether at low iron levels; unlike IRP1 binding that requires known or unknown. only reconstitution of the iron-sulfur cluster, IRP2 Among the earlier approaches, Ye and Connor reactivation requires new protein synthesis. IRP1 [6,30] used suppression subtraction hybridization may confer a rapid response capability to the IRP/ (SSH) to generate cDNA of human astrocytoma cell IRE system, with IRP2 reflecting longer-term mRNA after exposing the cells to either conditions adaptation. of high iron loading (100 µg/ml for 48 hr, as ferric IREs are sequences of about 30 nucleotides that ammonium citrate (FAC)) or chelation with DFO. occur in the untranslated regions (either 3' or 5' In this approach [31], DNA is digested by restriction UTR) of mRNA [4,22]. They have a consensus enzymes into small fragments and a pool (the tester)

02 Liu 230-239 232 7/20/05, 1:11 PM Genomic approaches to iron regulation 233

is divided in half and tagged with 2 different adapters A tail of a transcript to be reverse-transcribed. at the 5' end. Each pool is hybridized with an excess Following reverse transcription with additional of cDNA (the driver). Based on hybridization upstream primers of arbitrary sequences, mRNA kinetics, more abundant species in the tester are sub-populations are visualized by denaturing PAGE. enriched. Then the hybridization products are This allows direct side-by-side comparison of mixed for a second round of hybridization, followed mRNAs between or among related cells. As in the by 2 rounds of PCR with primers to the adapters. SSH approach, differentially expressed PCR The PCR products are ligated into vectors and products are selected, eluted, cloned, sequenced, and cloned in E. coli. Genes expressed at higher levels analyzed with reverse Northern blotting. HepG2 in the tester are recovered at higher frequency. To human hepatocarcinoma cells were treated for 3 to identify down-regulated genes, the experiment is 7 days with 40 µg/ml iron, again as FAC. Time- repeated with driver and tester reversed. Clearly the dependent decreases in apolipoprotein B100 mRNA approach is time consuming and technicaly and increases in aldose reductase and semaphorin demanding. In a DFO experiment with astrocytoma transcripts were demonstrated, but the significance cells [30], following elimination of repeat or of this is unclear and the method seems rather blunt. hybridizationally similar sequences and reverse A high-throughput method for genomic Northern blot analysis, forward subtraction yielded screening of iron-responsive genes in Salmonella was cDNA fragments increased by DFO, and backward reported by Bjarnason et al [33]. Digested genomic subtraction yielded those decreased. Twelve mRNA Salmonella DNA was ligated into a plasmid with species were increased by DFO (5 confirmed by the luminescence luxCDABE operon reporter gene, Northern blotting) and 29 were decreased (most and this random promoter library was reintroduced confirmed by Northern blots). Transcripts known into Salmonella. Luminescence was observed in low to be regulated by iron, such as ferritin and Tfr, were and high iron media, and clones with differential not identified by this approach. However, a parallel expression of 3-fold or more were selected for further SSH study with iron-treated astrocytes performed study. This approach follows expression from any by the same authors [6] revealed the expected promoter, whether of known or unknown genes, if increase in ferritin mRNA. DFO down-regulated they are represented in the constructed library. The some genes involved in energy production, such as use of a luminescent reporter allows monitoring cytochrome c oxidase and NADH:ubiquinone changes of expression from a given promoter (clone) oxidoreductase, and some related to protease in real time. The authors identified 182 promoters function. The authors noted disconnection between affected by increased iron exposure and 298 by iron gene expression and protein function, as iron depletion. They concluded that 7% of the overload has been reported to decrease activity of Salmonella genome may be regulated directly or cytochrome c oxidase. However, it is also possible indirectly by iron, and both Fur-regulated and Fur- that optimal iron levels are needed for expression of independent genes are involved. These results are a given gene, and both increase or decrease of iron biologically significant, but compared to microarray may suppress expression. A case in point is TGF-β methods the approach is laborious, requiring in cardiac cells [32], which is decreased by both iron construction of libraries, screening, and DNA loading and chelation. With iron overloaded sequencing. With resort to commercial micro- astrocytes, SSH identified 18 transcripts that were arrays, the synthesis and hopefully the confirmation increased by iron loading and 19 that were decreased. by sequencing of the targets have already been done Overall, no general pattern was obvious, but some by the supplier of the array. genes involved in energy production, such as α- enolase, were actually decreased by iron exposure. Microarray Methods–Plants and Bacteria An approach taken by Barisani et al [5] was conventional differential display. Anchored primers Microarray methods have been exploited well are designed to bind to the 5' boundary of the poly- in studies of lower organisms, as selected examples

02 Liu 230-239 233 7/20/05, 1:11 PM 234 Annals of Clinical & Laboratory Science, vol. 35, no. 3, 2005

will illustrate. Arabidopsis has become a prototype negative animal pathogen Pasteurella multocida in studies of plant genetics. An array of about 6,000 demonstrated that genes involved in energy Arabidopsis cDNAs was used to study the response metabolism and electron transport were generally of roots and shoots to growth in iron-deficient decreased, and those of iron-binding and iron- defined medium [34]. Generally, about 100 genes transport were increased during 2 hr of iron were depressed and a similar number induced in one deprivation [37]. The authors noted that 27% of day by both tissues. The numbers in both categories the genes affected by iron deficiency in this whole- rose at later times. Genes of many classes were genome study had no known function, underlining represented, but the main conclusions were that the power of the microarray approach to explore new respiration increases in response to a need to increase territory when a complete genome is available. In iron import, but the demand is exceeded in roots, another example with an animal pathogen, 66 and oxidative phosphorylation is supplemented by putative open reading frames of the virulence increased carbon import and anaerobic metabolism. plasmid of the immunopathogen Rhodococcus equi Bacterial gene expression studies promise were examined on a DNA microarray [38]. One important new insight into the biology of human group of genes, including those associated with the pathogens. The transcriptome of Neisseria meningit- virulence-associated protein (vap) family, was idis was studied to understand the dependence of induced by iron restriction, whereas a second group this important human pathogen on iron availability of magnesium-regulated genes was down-regulated. [35]. Cultures were depleted of iron with DFO; then 100 µM ferric nitrate was added for 5 hr to see Microarray Methods–Higher Animals which suppressed genes were reactivated. One hundred and fifty-three genes were up-regulated by Among the initial microarray studies of iron- addition of iron, and 80 were down-regulated. Many responsive genes was that of human HL-60 pro- of these genes were Fur-regulated, including some monocytes by Alcantara et al [39]. The cells were previously unknown to be Fur-dependent but tested following stimulation to differentiate with confirmed to be so by gel shift assays. One phorbol ester. Using an early commercial array that conclusion was that such gene expression profiling tested only 43 genes, mostly relevant to apoptosis might be useful to identify new iron-governed or cell cycle regulation, the authors found that 11 regulons in bacteria. In a different approach with genes were suppressed by DFO. Each was confirmed the intestinal pathogen Campylobacter jejuni, early by RT-PCR, Northern, and/or Western analysis, and response to iron-repletion was compared to a significant general conclusion could be drawn: iron subsequent adaptation [36]. When iron-deficient supports expression of genes involved in regulating organisms were supplied with iron, 647 genes were cell cycle progression and apoptosis, and thus is affected within 15 min, whereas only 208 genes were fundamentally involved in the differentiation process differentially expressed when iron-rich and iron- of a hematopoietic cell line. Despite the small deficient conditions were compared at steady state. number of genes tested, this remains one of the most In general, genes involved in iron acquisition and biologically instructive studies to address general oxidative stress response were down-regulated in genomic regulation by iron. both scenarios, while expression of genes associated Muckenthaler et al [40] developed the “Iron- with energy metabolism was increased. Changes in Chip” specifically to probe patterns of human gene protein glycosylation were also revealed in this study, regulation by iron. They initially selected about 115 which may explain differences in virulence genes based on “literature searches … microarray dependent upon intestinal iron availability. Further, experiments performed on filters that contain use of a Fur mutant identified 53 Fur-regulated approximately 20,000 human nonredundant genes, including some that were not previously expressed sequence tags … and gene lists from known to be so [36]. published microarray studies that address metabolic Whole-genome microarray analysis of the gram- pathways of interest” [40]. From this and other

02 Liu 230-239 234 7/20/05, 1:11 PM Genomic approaches to iron regulation 235

published reports [41,42], it is difficult to evaluate cDNA arrays, including higher specificity, more the selection process. However, this specialized chip uniform melting temperatures, and more uniformity has been used with some interesting results, at least of spots. Affymetrix GeneChips consist of up to 16 in the hands of its developers, and more genes have probes representing each transcript, each a 25-mer been added in later versions. Initial implementation unit, that together span the length of the gene. of the chip used the human HeLa cell line exposed Perfect match sequences and mismatch (1-base to various stresses including peroxide, nitric oxide, change) probe pairs are used by Affymetrix to hemin, 100 µM FAC, and DFO. Genes increased minimize non-specific cross hybridization. The by iron loading with hemin or FAC included heme major advantage of Affymetrix GeneChips over non- oxygenase-1 and several heat shock proteins. In synthetic types of microarrays is the minimized chip- addition to the IRP-regulated transferrin receptor to-chip variation resulting from the use of synthetic gene, a decrease with FAC was seen in metallothion- targets. On the other hand, synthetic oligonucleo- ein-2. Lysyl oxidase and c-jun were increased more tides are expensive compared to direct spotting on than 3-fold by DFO, and HIF-1α was unchanged cDNA arrays, and this limits the number of by either treatment. Interestingly, similar profiles experiments that most laboratories will conduct. were seen with DFO and HeLa cells over-expressing Several head-to-head comparisons of different the hemochromatosis gene, Hfe [40], suggesting a platforms have been performed, but there is no role of the protein in decreasing an intracellular iron consensus about a preference for cDNA or pool involved in signaling, perhaps the so-called oligonucleotide arrays [43-45]. labile iron pool. The IronChip was used to study We used the rat genome U34A chip from gene regulation more extensively in Hfe-deficient Affymetrix to analyze approximately 7,000 full- mice [41]. The hepatic regulator of iron metabolism, length sequences and 1,000 expressed sequence tag hepcidin, was decreased in Hfe-/- mice and its (EST) clusters in myocytes isolated from ventricles regulation by iron was altered. This was accompanied of 1- to 3-day-old rat pups. The cells were treated by dysregulation of the intestinal redox iron with 20 µg/ml iron, as FAC, for 72 hr. One hundred transporters Ireg1 and DcytB, indicating additional and seventy genes or ESTs were up-regulated and regulatory defects involved in hemochromatosis. 380 genes or ESTs were down-regulated, by at least Profiling gene expression in human epithelial cells, 2-fold. Genes increased by iron loading included the IronChip revealed that infection with Neisseria oxidative stress response genes such as glutathione menigitidis produces changes similar to those in peroxidase, glutathione S-transferase A, catalase, uninfected cells in iron-deficient conditions [42]. metallothioneins I and II, and heat shock protein 60s. Surprisingly, heat shock protein 70s was Original Studies with Liver and Heart Cells unchanged. Genes decreased by iron loading included those coding transferrin receptor, HIF-1α, Three basically different microarray platforms are glyceraldehyde phosphate dehydrogenase (GAPD), in current use, where the spotted target sequence is cyclin B1, collagen type 1, and fibronectin. Down- full length or partial cDNA, oligonucleotides regulation of the IRE-regulated transferrin receptor (typically about 60-mer sequences) derived from served as a positive control for these experiments, libraries of digested cDNA, or shorter synthetic and changes in the other gene transcripts were oligonucleotides typified by the Affymetrix consistent with data published previously by us [32] GeneChips. cDNA arrays are preferentially used and others [40,46]. An interesting observation from with a 2-color protocol (for example, control and our gene list is that several protein phosphatases and treatment cDNA samples are labeled with distin- kinases were down-regulated by iron, such as guishable fluorophores and hybridized together on phosphatases 2A, 2B, 2C, protein tyrosine one array), whereas in the Affymetrix format a single phosphatase, phosphoglycerate kinase 1, phos- sample is hybridized to each array. In principle, photidylinositol 4-kinase, serine threonine kinases oligonucleotide arrays have several advantages over 3 and 39, mitogen activated protein kinase kinase

02 Liu 230-239 235 7/20/05, 1:11 PM 236 Annals of Clinical & Laboratory Science, vol. 35, no. 3, 2005

1, and the stress-related mitogen activated protein 2-fold. These genes are listed in Table 1. In this kinase 14 (p38α). Down-regulation of both kinases limited survey, 3 genes (ie, AKR1C2, ALDH1A1, and phosphatases is consistent with the idea that and GAPD) changed in the opposite direction upon many phosphatases are induced by the kinases they iron loading in the Affymetrix experiment with regulate [47], achieving a negative feedback loop. cardiac myocytes. This lends credence to the iron- Another observation is that a number of mRNAs dependence of these genes. for growth factors were decreased by iron loading, Critical issues not discussed here are how the including transforming growth factor β3, several data should be normalized [we used Lowess forms of vascular endothelial growth factor, platelet- normalization in the OCI experiments and the GC- derived growth factor A chain, and heparin binding RMA algorithm from ArrayAssist software EGF-like growth factor. Insulin-like growth factors (Stratagene, La Jolla, CA) with the Affymetrix data], IGF-I and IGF-II were increased, and the potentially and how they should be combined from multiple inhibitory IGF binding protein-3 was decreased. experiments to test statistical significance. There is In a different approach, we used a 2-color a well-established means of reporting such data experiment with the Ontario Cancer Institute’s adopted by many journals (Minimum Information (OCI) 1.7 k human cDNA microarray, containing About a Microarray Experiment (MIAME) [48]) 1,700 human genes and (or) ESTs, to study the that acknowledges the biological variation among effects of iron chelation on liver cells. Each gene cell types and subtle changes in conditions, and is was spotted in duplicate, and the experiment was intended to facilitate interlaboratory comparisons. repeated 4 times. Illustrative of the problems of MIAME requires details of (a) experimental design, using semi-commercial approaches, gene lists were (b) array design, (c) sample selection, (d) changed between experiments, many genes were not hybridization protocol, (e) image analysis, and (f) verified, and technical backup was lacking. Never- normalization and controls for comparison. It still theless, when HepG2 (human hepatocarcinoma- seems optimistic to think that this information will derived) cells were treated with 1 mM DFO for 24 be sufficient to consolidate data from many hr, the expression of 13 genes was changed at least laboratories. Here, we describe briefly a further stage in the analysis that the richness of data from microarray experiments facilitates. Table 1. Gene expression altered by iron chelation in HepG2 Using the minimal set of 13 genes suggested to human hepatocarcinoma cells. Confluent cell cultures were treated with 1 mM deferoxamine (DFO) for 24 hr. Genes change in the OCI cDNA experiments, we used showing 2-fold changes av eraged over 4 independent PathwayAssist software (Stratagene) to uncover links experiments with 1.7 K OCI microarrays are listed. Gene among the genes differentially expressed. Many symbols are shown in parentheses. different approaches can be explored in this package. Increased by DFO We only present, as an example, an analysis of Glyceraldehyde-3-phosphate dehydrogenase (GAPD) shortest-path connections of the 13 genes altered in Insulin-like growth factor binding protein 1 (IGFBP1) Table 1, based on small-molecule regulators and Hypothetical protein FLJ20967 minimization of connections (Fig. 1). Ten of the Glutaredoxin (thioltransferase) (GLRX) 13 gene products interconnect by one step through Retinoblastoma binding protein 1 (RBBP1) Kelch-like 4 (KLHL4) known small molecule regulators. The role of PGE2 Spermidine/spermine N1-acetyltransferase (SAT) in the generated scheme deserves comment. It has Enolase 2 (ENO2) been found that iron down-regulates prostaglandin E2 (PGE2) expression [49] and DFO increases Decreased by DFO Aldo-keto reductase family 1 (AKR1C2) prostaglandin synthesis by increasing cyclooxy- Aldehyde dehydrogenase 1 (ALDH1A1) genase-2 expression [50]. PGE2 in turn has been Ferritin L chain (FTL) shown to increase an insulin-like growth factor Stathmin 1/oncoprotein (STMN1) binding protein (IGFBP4) in chondrocytes [51] Methionine adenosyltransferase 1 (MAT1A) (compare IGFBP1 in Table 1) and is linked to

02 Liu 230-239 236 7/20/05, 1:11 PM Genomic approaches to iron regulation 237

Fig. 1. Stratagene PathwayAssist analysis of the 13 genes listed in Table 1. Small molecule effects with 1-step linkage were chosen for this example. Ten of the genes were identified by the software as having known regulation paths in this analysis. Solid line, expression or synthesis; dashed line, regulation; RA, retinoic acid; PMA, phorbol myristoyl acid. The gene symbols are defined in Table 1.

increased polyamine synthesis and spermidine/ genome-wide screening of cells treated with iron or spermine N1-acetyltransferase activity in iron chelators, and many reports have focused more lymphocytes [52]. Additionally, SAT, GAPD, on methods than results. Such approaches are often IGFBP1, and ENO2 are all increased by DFO validated by demonstrating regulation of genes (Table 1). The pathway analysis reveals a linkage already known to be iron-dependent from classical through a phorbol ester (PMA)-dependent (eg, biochemical methods. Otherwise, individual genes protein kinase C) mechanism, though stathmin revealed in specific studies require validation by (STMN1) is regulated contrary to expectation. Such additional approaches such as real-time PCR and analyses are highly speculative, but illustrate the Northern blotting, and significance must be richness of data that can be generated for closer demonstrated at the protein level. Novel genes, ie, scrutiny by global microarray experiments. The those not already known to be iron-dependent, may consistency of these observations with the cited have idiosyncratic responses dependent on specific reports supports the veracity of the array data. cell types and under conditions of culture and treatment. This underscores the need for reporting- Concluding comments protocols, such as MIAME. Pre-microarray methods have not generally proven to be very responsive to In general, no clear or consistent patterns of iron- detecting genes known or strongly suspected to be dependent gene expression have been uncovered by influenced by iron, and microarrays yield large

02 Liu 230-239 237 7/20/05, 1:11 PM 238 Annals of Clinical & Laboratory Science, vol. 35, no. 3, 2005

numbers of candidates with the suspicion of large 15. Wenger RH, Gassmann M. Oxygen(es) and the hypoxia- numbers of false positives. Nevertheless, iron is a inducible factor-1. Biol Chem 1997;378:609-616. 16. Semenza GL. Hypoxia-inducible factor 1 and the molec- metal that clearly affects numerous biological ular physiology of oxygen homeostasis. J Lab Clin Med processes, and as methods for genome-wide screen- 1998;131:207-214. ing become more refined, reproducible, and robust, 17. Ponka P. Iron utilization in erythrocyte formation and patterns of iron-dependent gene regulation with hemoglobin synthesis. In: Molecular and Cellular Iron general applicability to many cell types may emerge. Transport (Templeton DM, Ed) Dekker, New York, 2002; pp 643-677. 18. Lok CN, Ponka P. Identification of a hypoxia response References element in the transferrin receptor gene. J Biol Chem 1999;274:24147-24152. 1. Templeton DM (Ed). Molecular and Cellular Iron 19. Jaakkola P, Mole DR, Tan Y-M, Wilson MI, Gielbert J, Transport. Dekker, New York, 2002; 827 pp. Gaskell SJ, von Kriegsheim A, Hebestreit HF, Mukherji 2. Cairo G, Pietrangelo A. Iron regulatory proteins in patho- M, Schofield C, Maxwell PH, Pugh CW, Ratcliffe PJ. biology. Biochem J 2000;352:241-250. Targeting of HIF-[alpha] to the von Hippel-Lindau 3. Eisenstein RS. Iron regulatory proteins and the molecular ubiquitylation complex by O2-regulated prolyl control of mammalian iron metabolism. Annu Rev Nutr hydroxylation. Science 2001;292:468-472. 2000;20:627-662. 20. Lando D, Peet DJ, Whelan DA, Gorman JJ, Whitelaw 4. Theil EC, Eisenstein RS. Combinatorial mRNA regul- ML. Asparagine hydroxylation of the HIF transactivation ation: Iron regulatory proteins and iso-iron responsive domain: A hypoxic switch. Science 2002;292:858-861. elements (iso-IREs). J Biol Chem 2000;275:40659- 21. Hanson ES, Leibold E. Regulation of iron homeostasis 40662. by iron regulatory proteins 1 and 2. In: Molecular and 5. Barisani D, Meneveri R, Ginelli E, Cassani C, Conte D. Cellular Iron Transport (Templeton DM, Ed), Dekker, Iron overload and gene expression in HepG2 cells: analysis New York, 2002; pp 201-235. by differential display. FEBS Lett 2000;469:208-212. 22. Johansson HE, Theil EC. Iron-response element (IRE) 6. Ye Z, Connor JR. Identification of iron responsive genes structure and combinatorial RNA regulation. In: by screening cDNA libraries from suppression subtractive Molecular and Cellular Iron Transport (Templeton DM, hybridization with antisense probes from three iron Ed), Dekker, New York, 2002; pp 237-253. conditions. Nucleic Acids Res 2000;28:1802-1807. 23. Beinert H, Kiley P. Redox control of gene expression 7. Templeton DM, Liu Y. Genetic regulation of cell function involving iron-sulfur proteins. Change of oxidation-state in response to iron overload or chelation. Biochim Biophys or assembly disassembly of Fe-S clusters? FEBS Lett 1996; Acta 2003;1619:113-124. 382:218-219. 8. Crosa JH. Signal transduction and transcriptional and 24. Beinert H, Holm RH, Münck E. Iron-sulfur clusters: post-transcriptional control of iron-regulated genes in Nature’s modular, multipurpose structures. Science 1997; bacteria. Microbiol Mol Biol Rev 1997;61:319-336. 277:653-659. 9. Wösten MM, Kox LF, Chamnongpol S, Soncini FC, 25. Iwai K, Klausner RD, Rouault TA. Requirements for iron- Groisman EA. A signal transduction system that responds regulated degradation of the RNA binding protein, iron to extracellular iron. Cell 2000;103:113-125. regulatory protein 2. EMBO J 1995;14:5350-5357. 10. Li NX, Karin M. Is NF-kB the sensor of oxidative stress? 26. Guo B, Phillips JD, Yu Y, Leibold EA. Iron regulates the FASEB J 1999;13:1137-1143. intracellular degradation of iron regulatory protein 2 by 11. Bonizzi G, Piette J, Merville M-P, Bours V. Cell type- the proteasome. J Biol Chem 1995;270:21645-21651. specific role for for reactive oxygen species in nuclear 27. Klausner RD, Rouault TA, Harford JB. Regulating the factor-kappa B activation by interleukin-1. Biochem fate of mRNA: The control of cellular iron metabolism. Pharmacol 2000;59:7-11. Cell 1993;72:19-28. 12. Kwak EL, Larochelle DA, Beaumont C, Torti SV, Torti 28. Tchernitchko D, Bourgeois M, Martin ME, Beaumont FM. Role for NF-kB in the regulation of ferritin H by C. Expression of the two mRNA isoforms of the iron tumor necrosis factor-α. J Biol Chem 1995;270:15286- transporter Nrmap2/DMTI in mice and function of the 15293. iron responsive element. Biochem J 2002;363:449-455. 13. Chojkier M, Houglum K, Soli-Herruzo JA, Brenner DA. 29. Hubert N, Hentze MW. Previously uncharacterized Stimulation of collagen gene expression by ascorbic acid isoforms of divalent metal transporter (DMT)-1: Implic- in cultured human fibroblasts: A role for lipid peroxid- ations for regulation and cellular function. PNAS USA ation? J Biol Chem 1989;264:16957-16962. 2002;99:12345-12350. 14. Ruiz IG, De la Torre P, Diaz T, Esteban E, Morillas JD, 30. Ye Z, Connor JR. Screening of transcriptionally regulated Muñoz-Yagüe T, Solís-Herruzo JA. Sp family of transcrip- genes following iron chelation in human astrocytoma cells. tion factors is involved in iron-induced collagen alpha1(I) Biochem Biophys Res Commun 1999;264:709-713. gene expression. DNA Cell Biol 2000;19:167-178. 31. Diatchenko L, Lau Y-FC, Campbell AP, Chenchik A,

02 Liu 230-239 238 7/20/05, 1:11 PM Genomic approaches to iron regulation 239

Moqadam F, Huang B, Lukyanov S, Lukyanov K, meningitidis. Cell Microbiol 2004;6:473-484. Gurskaya N, Sverdlov ED, Siebert PD. Suppression 43. Yuen T, Wurmbach E, Pfeffer RL, Ebersole BJ, Sealfon subtractive hybridization: A method for generating SC. Accuracy and calibration of commercial oligo- differentially regulated or tissue-specific cDNA probes and nucleotide and custom cDNA microarrays. Nucleic Acids libraries. PNAS USA 1996;93:6025-6030. Res 2002;30:e48. 32. Parkes JG, Liu Y, Sirna JB, Templeton DM. Changes in 44. Tan PK, Downey TJ, Spitznagel EL, Jr., Xu P, Fu D, gene expression with iron loading and chelation in cardiac Dimitrov DS, Lempicki RA, Raaka BM, Cam MC. myocytes and non-myocytic fibroblasts. J Mol Cell Evaluation of gene expression measurements from Cardiol 2000;32:233-246. commercial microarray platforms. Nucleic Acids Res 33. Bjarnason J, Southward CM, Surette MG. Genomic 2003;31:5676-5684. profiling of iron-responsive genes in Salmonella enterica 45. Woo Y, Affourtit J, Daigle S, Viale A, Johnson K, Naggert serovar typhimurium by high-throughput screening of a J, Churchill G. A comparison of cDNA, oligonucleotide, random promoter library. J Bacteriol 2003;185:4973- and Affymetrix GeneChip gene expression microarray 4982. platforms. J Biomol Tech 2004;15:276-284. 34. Thimm O, Essigmann B, Kloska S, Altmann T, Buckhout 46. Le NTV, Richardson DR. The role of iron in cell cycle TJ. Response of Arabidopsis to iron deficiency stress as progression and the proliferation of neoplastic cells. revealed by microarray analysis. Plant Physiol Biochim Biophys Acta 2002;1603:31-46. 2001;127:1030-1043. 47. Zhao Y, Xiao W, Templeton DM. Suppression of mitogen- 35. Grifantini R, Sebastian S, Frigimelica E, Draghi M, activated protein kinase phosphatase-1 (MKP-1) by Bartolini E, Muzzi A, Rappuoli R, Grandi G, Genco CA. heparin in vascular smooth muscle cells. Biochem Identification of iron-activated and -repressed Fur- Pharmacol 2003;66:769-776. dependent genes by transcriptome analysis of Neisseria 48. Brazma A, Hingamp P, Quackenbush J, Sherlock G, meningitidis group B. PNAS USA 2003;100:9542-9547. Spellman P, Stoeckert C, Aach J, Ansorge W, Ball CA, 36. Palyada K, Threadgill D, Stintzi A. Iron acquisition and Causton HC, Gaasterland T, Glenisson P, Holstege FC, regulation in Campylobacter jejuni. J Bacteriol 2004;186: Kim IF, Markowitz V, Matese JC, Parkinson H, Robinson 4714-4729. A, Sarkans U, Schulze-Kremer S, Stewart J, Taylor R, Vilo 37. Paustian ML, May BJ, Kapur V. Pasturella multocida gene J, Vingron M. Minimum information about a microarray expression in response to iron limitation. Infection Immun experiment (MIAME)-toward standards for microarray 2001;69:4109-4115. data. Nature Genet 2001;29:365-371. 38. Ren J, Prescott JF. Analysis of virulence plasmid gene 49. Hisakawa N, Nishiya K, Tahara K, Matsumori A, expression of intra-macrophage and in vitro grown Hashimoto K. Down regulation by iron of prostaglandin Rhodococcus equi ATCC 33701. Vet Microbiol 2003;94: E2 production by human synovial fibroblasts. Ann 167-182. Rheumatic Dis 1998;57:742-746. 39. Alcantara O, Kalidas M, Baltathakis I, Boldt DH. 50. Tanji K, Imaizumi T, Matsumiya T, Itaya H, Fujimoto K, Expression of multiple genes regulating cell cycle and Cui X, Toki T, Ito E, Yoshida H, Wakabayashi K, Satoh apoptosis in differentiating hematopoietic cells is K. Desferrioxamine, an iron chelator, upregulates cyclo- dependent on iron. Exp Hematol 2001;29:1060-1069. oxygenase-2 expression and prostaglandin production in 40. Muckenthaler M, Richter A, Gunkel N, Riedel D, a human macrophage cell line. Biochim Biophys Acta Polycarpou-Schwarz M, Hentze S, Falkenhahn M, 2001;1530:227-235. Stremmel W, Ansorge W, Hentze MW. Relationships and 51. Di Battista JA, Dore S, Morin N, He Y, Pelletier JP, distinctions in iron-regulatory networks responding to Martel-Pelletier J. Prostaglandin E2 stimulates insulin- interrelated signals. Blood 2003;101:3690-3698. like growth factor binding protein-4 expression and 41. Muckenthaler M, Roy CN, Custodio AO, Minana B, synthesis in cultured human articular chondrocytes: 2+ deGraaf J, Montross LK, Andrews NC, Hentze MW. possible mediation by Ca -calmodulin regulated Regulatory defects in liver and intestine implicate processes. J Cell Biochem 1997;65:408-419. abnormal hepcidin and Cybrd1 expression in mouse 52. Venza I, Ruggeri P, Valenti A, Denaro L, Severino A, Teti hemochromatosis. Nature Genet 2003;34:102-107. D. Modified responses to PGE2 in polyamine biosynthesis 42. Bonnah RA, Muckenthaler MU, Carlson H, Minana B, by T lymphocytes of gastric- and conjunctiva basal cell- Enns CA, Hentze MW, So M. Expression of epithelial carcinoma patients. Immunopharmacol Immunotoxicol cell iron-related genes upon infection by Neisseria 2001;23:541-554.

02 Liu 230-239 239 7/20/05, 1:11 PM