Gene expression profiling reveals early cellular responses to intracellular magnetic labeling with superparamagnetic iron oxide nanoparticles

Authors

  • Dorota A. Kedziorek,

    1. Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    2. Cellular Imaging Section and Vascular Biology Program, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Naser Muja,

    1. Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    2. Cellular Imaging Section and Vascular Biology Program, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Piotr Walczak,

    1. Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    2. Cellular Imaging Section and Vascular Biology Program, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    Search for more papers by this author
  • Jesus Ruiz-Cabello,

    1. Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    2. Cellular Imaging Section and Vascular Biology Program, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    3. Ciber de enfermedades respiratorias, Instituto de Estudios Biofuncionale, Universidad Complutense de Madrid, Madrid, Spain
    Search for more papers by this author
  • Assaf A. Gilad,

    1. Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    2. Cellular Imaging Section and Vascular Biology Program, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    Search for more papers by this author
  • Chunfa C. Jie,

    1. Institute for Basic Biomedical Sciences, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    Search for more papers by this author
  • Jeff W. M. Bulte

    Corresponding author
    1. Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    2. Department of Biomedical Engineering, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    3. Department of Chemical & Biomolecular Engineering, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    4. Cellular Imaging Section and Vascular Biology Program, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    • Department of Radiology, The Johns Hopkins University School of Medicine, 217 Traylor Bldg, 720 Rutland Ave, Baltimore, MD 21205-2195
    Search for more papers by this author

Abstract

With MRI (stem) cell tracking having entered the clinic, studies on the cellular genomic response toward labeling are warranted. Gene expression profiling was applied to C17.2 neural stem cells following superparamagnetic iron oxide/PLL (poly-L-lysine) labeling over the course of 1 week. Relative to unlabeled cells, less than 1% of genes (49 total) exhibited greater than 2-fold difference in expression in response to superparamagnetic iron oxide/PLL labeling. In particular, transferrin receptor 1 (Tfrc) and heme oxygenase 1 (Hmox1) expression was downregulated early, whereas genes involved in lysosomal function (Sulf1) and detoxification (Clu, Cp, Gstm2, Mgst1) were upregulated at later time points. Relative to cells treated with PLL only, cells labeled with superparamagnetic iron oxide/PLL complexes exhibited differential expression of 1399 genes. Though these differentially expressed genes exhibited altered expression over time, the overall extent was limited. Gene ontology analysis of differentially expressed genes showed that genes encoding zinc-binding proteins are enriched after superparamagnetic iron oxide/PLL labeling relative to PLL only treatment, whereas members of the apoptosis/programmed cell death pathway did not display increased expression. Overexpression of the differentially expressed genes Rnf138 and Abcc4 were confirmed by quantitative real-time polymerase chain reaction. These results demonstrate that, although early reactions responsible for iron homeostasis are induced, overall neural stem cell gene expression remains largely unaltered following superparamagnetic iron oxide/PLL labeling. Magn Reson Med 63:1031–1043, 2010. © 2010 Wiley-Liss, Inc.

Intracellular uptake of superparamagnetic iron oxide (SPIO) nanoparticles enables noninvasive in vivo MRI of the delivery and biodistribution of cellular therapeutics. Feridex I.V.®, an injectable solution of dextran-coated ferumoxide for hepatic and splenic imaging (1), has been widely applied to label a wide variety of cells and has undergone initial phase I/II testing for dendritic cells in melanoma patients (2), neural stem cells (NSCs) in traumatic brain injury (3), and for pancreatic islet grafts in type I diabetic patients (4). Given the promise of cell therapy and the anticipated pivotal role of clinical MRI cell tracking for real-time delivery and assessment of initial biodistribution, the compatibility and safety of Feridex® labeling must be assured (5).

Several in vitro and in vivo studies have demonstrated that cell viability, differentiation, and proliferation are normal in SPIO-labeled cells (2, 6, 7). However, reduced synthesis of collagen and extracellular matrix following the induction of chondrogenesis in Feridex®- (8, 9) or Resovist- (10) labeled bone marrow–derived MSC has been reported and immunomodulatory effects have been encountered for SPIO-labeled macrophages and their interactions with T cells (11). Recently, the use of acidic medium (pH 4.5, sodium citrate chelate) modeling the pH of the lysosomal compartment has been shown to release Fe3+ from iron oxide nanoparticles (12), indicating that retention of SPIO within this compartment could elevate cytosolic Fe3+ concentration. In support of this hypothesis, coordinated changes in the expression of genes encoding iron sequestration proteins, transferrin receptor 1, and ferritin are induced in HeLa cells and human MSCs following SPIO labeling (13), indicative of posttranscriptional regulation of gene expression by iron regulatory proteins (IRPs; IRP1 and IRP2) in adaptation to increased cytosolic Fe3+ concentration (14).

In conditions where iron homeostasis is not maintained, increased intracellular Fe3+ can contribute to Fe2+-catalyzed oxidative damage to nucleic acids, proteins, and lipids (15). Because the signaling properties of iron are not accounted for by routine histochemical detection of cellular markers, complementary assays of intracellular processes, including signal transduction analysis, proteomic studies, and gene expression profiling, are essential for complete characterization of the response of a specific cell type to SPIO labeling prior to transplantation. The multipotent NSC line C17.2 has been widely applied as cell therapy in experimental models of CNS disease (16, 17), and the migratory properties of this cell line have been characterized using MRI cell tracking (18, 19). In this study, we have labeled C17.2 NSCs with Feridex® (Bayer Healthcare Pharmaceuticals; 25 μg/mL) and the transfection agent poly-L-lysine (PLL) and examined the effect of labeling on gene expression over the course of 1 week in vitro. The present results are directly relevant to current safety considerations in clinical MRI cell tracking trials.

MATERIALS AND METHODS

C17.2 NSC Culture and Labeling

The LacZ-expressing mouse NSC cell line C17.2 was cultured in Dulbecco's modified Eagle's medium containing 4.5 mg glucose/mL, 10% fetal bovine serum, 5% horse serum, 1% L-glutamine, 100 IU/mL penicillin, 100 μg/mL streptomycin, and 0.025 μg/mL amphotericin B (Invitrogen, Carlsbad, CA). For cell labeling, 25 μg Fe/mL ferumoxide formulation (Feridex I.V.®, Bayer Healthcare Pharmaceuticals) and PLL hydrobromide (Mw = 388 kDa; Sigma, St. Louis, MO) at a concentration of 375 ng/mL was incubated in culture medium at room temperature for 1 h (9). The SPIO-PLL-complexed medium was then added to C17.2 NSCs and incubated for 24 h. To control for potential effects induced by the transfection agent PLL alone (without SPIO complexation), cells cultured in parallel were incubated overnight with PLL (375 ng/mL). C17.2 NSCs were labeled with SPIO/PLL or treated with PLL only on day 1, 4, 6, or 7 after plating, and total RNA was isolated from each sample on day 8 of the experiment. After labeling, C17.2 NSCs were transferred to basal culture medium that was refreshed every 2 days. A duplicate time course experiment was performed in each series. Unlabeled C17.2 NSCs (no incubation with either SPIO/PLL or PLL) were included as a baseline reference.

Total RNA Isolation and Microarray Assay

Total RNA was extracted from C17.2 NSCs using TRI reagent (Promega, Madison, WI) and Phase Lock Gel Heavy (Eppendorf, Westbury, NY) under RNase-free conditions. RNA was purified using the RNeasy microkit (Qiagen, Valencia, CA) and quality was assessed using an Agilent 2100 Bioanalyzer microchip (Agilent, Palo Alto, CA). Microarray assays were performed on Mouse Genome 430 2.0 cDNA arrays (Affymetrix, Santa Clara, CA). Ten micrograms of biotin-labeled and fragmented cRNA was hybridized. The microarrays were hybridized, washed, and scanned according to Affymetrix standard protocols.

Gene Expression Analysis

The RNA samples were analyzed with Affymetrix GeneChip® Mouse Genome 430 2.0 Arrays containing 45,101 probes for over 39,000 transcripts, including 34,000 well-characterized mouse genes. Quality of the microarray experiment was assessed with affyPLM and Affy, two Bioconductor packages for statistical analysis of microarray data. To estimate the gene expression signals, data analysis was conducted on the chips' CEL file probe signal values at the Affymetrix probe pair (perfect match probe and mismatch probe) level, using the statistical algorithm robust multiarray expression measure (20) with Affy. This probe-level data processing includes a normalization procedure utilizing quantile normalization method (21) to reduce variation between microarrays, which might be introduced during the processes of sample preparation, manufacture, fluorescence labeling, hybridization and/or scanning. With the signal estimates, multidimensional scaling analysis was performed in R to assess sample variability and identified a batch effect attributed to the operation time. The expression signals were adjusted for the batch effect with the R package ComBat, using the nonparametric empiric Bayes method (http://statistics.byu.edu/johnson/ComBat/; (22)).

With the signal data in a log-transformed format, differential gene expression between individual conditions was assessed by statistical linear model analysis using the bioconductor package limma, in which an empiric Bayes method is used to moderate the standard errors of the estimated log-fold changes of gene expression and results in more stable inference and improved power, especially for experiments with small numbers of microarrays (23). Likewise, the raw P values provided a way to rank genes in terms of the evidence for differential gene expression in order to obtain the most likely differentially expressed genes between conditions. Analyses other than the time course analysis and pattern recognition were performed under R environment (http://www.r-project.org) (4).

Time course analyses were made with the software package EDGE (http://faculty.washington.edu/jstorey/edge/about.php) to identify gene expression changes over the temporal development course of day 0 through day 7 within and between the two treatment conditions PLL and SPIO/PLL. Specifically, gene expression curves over time were modeled flexibly on a natural cubic spline basis, and statistical significance was calculated while accounting for sources of dependence over time (24). The raw P values provided a way to rank genes in terms of the evidence for differential gene expression in order to obtain the most likely differentially expressed genes over the time course. For pattern visualization, the expression data of the top ranked differentially expressed genes were scaled by the median value within each treatment condition in a per-probe manner and displayed in Heatmap, using the software package, GeneSpring GX 7.3 (http://www.chem.agilent.com/scripts/pds.asp?lpage = 27881).

The TIBCO Spotfire software was used for data visualization and organization postanalysis.

Quantitative Real-Time PCR

Real time-PCR (RT-PCR) (iScript One-Step RT-PCR; BioRad) was performed to confirm the validity of the microarray results. The relative expression of the target genes Rnf138 and Abcc4 in C17.2 NSCs was determined. The criteria for choosing Rnf138 and Abcc4 for additional verification were based on the representation of affected gene groups. The iCycler iQ Thermocycler system (BioRad) was used with a lower detection limit of 1 pg of total RNA input. Total RNA (1 μL; 100 ngμL) was reverse transcribed and primers internal to the microarray probe features for Rnf138 and Abcc4 were used for cDNA amplification. Primer pairs were validated by analyzing PCR amplimers on a 1% agarose gel to confirm the presence of a single band of the predicted size under ultraviolet light (ethidium bromide DNA staining). In addition, melting curves were generated using the iCycler analysis software to determine whether there were any spurious amplification products. To ensure equal input conditions, RT-PCR for ActB (β-actin), a reference gene, was performed in parallel for each sample. After normalizing for primer amplification efficiency, raw concentrations of Rnf138 and Abcc4 amplimers were determined relative to ActB content within each sample, and statistical differences between the means (± standard deviation) untreated, PLL-treated, and SPIO/PLL-labeled samples were computed using the Student's t test. Data were considered statistically significant when P < 0.05.

RESULTS

Single-Time-Point Analysis of Individual Gene Expression

Bland-Altman plots were used to display the fold-difference (log2 SPIO/PLL minus log2 PLL) in individual gene expression versus the mean signal intensity at each time point; this approach highlights transcripts that are transiently altered following cell labeling. For example, the expression of some gene transcripts could be altered in immediate response to SPIO/PLL labeling, whereas other gene transcripts may show differences following the withdrawal of SPIO/PLL. Overall, the majority of genes (99%) expressed by C17.2 NSCs exhibited less than a 2-fold difference (log2 fold change less than ±1) in gene expression following SPIO/PLL labeling (Fig. 1). Importantly, these included genes encoding for members of the programmed cell death pathway, neural differentiation, and genes regulating cellular metabolism. Differences in gene expression outside the 2-fold threshold were considered most likely to be representative of biologically relevant gene responses to SPIO labeling. A total of 49 transcripts were differentially expressed by greater than 2-fold (log2 fold change greater than ±1) at any given point during the study, with more than half represented on day 4 after labeling (Table 1).

Figure 1.

Bland-Altman plots of the log2 fold change in gene expression versus mean expression of each gene at single time points. Probes for reference genes (blue filled circles, Actb, Aco1, Aco2, Fth1, Gapdh, Trf, Trfr2) and genes of interest (green filled circles, Abcc4, Rnf138), and iron metabolism genes exhibiting greater than a 2-fold change in expression in at least one particular time point (red filled circles, Cp, Trfc) in C17.2 NSCs are highlighted on day 1 (a), day 2 (b), day 4 (c), and day 7 (d) after SPIO/PLL labeling. Probes for genes exhibiting less than a 2-fold difference in expression are represented by gray filled circles. Several genes (e.g., Actb, Cp, Fth1, Gapdh, Tfrc) are represented by hybridization signal from multiple probes, distributed along the 5′ to 3′ extent of each transcript, indicating the quality of the RNA isolation and hybridization process. Raw fold change values are indicated in Table 1.

Table 1. List of Genes That Were Differentially Expressed by More Than 2-Fold in C17.2 NSCs at Any Measured Time Point After SPIO/PLL Labeling*
Gene symbolGene titleDayAccessionProbe name
1247
  • *

    Raw fold change values for each measured time point are indicated, along with the NCBI gene accession identifier and the Affymetrix probe name for each gene. Several genes (e.g., Adcy7, Clu, Cp, Fabp4, Fstl1, Sulf1, Tapbp, Tfrc, and Tspan13) are represented by hybridization signal from multiple microarray probes positioned along the 5′ to 3′ length of their respective RNA transcripts.

Adcy7adenylate cyclase 7−1.25−1.042.061.21NM_0074061450065_at
Adcy7adenylate cyclase 7−1.09−1.072.141.20BB7468071456307_s_at
Angptl4angiopoietin-like 4−1.091.132.18−1.01NM_0205811417130_s_at
Btcbetacellulin, epidermal growth factor family member−2.00−1.65−1.55−1.10NM_0075681421161_at
C1rcomplement component 1, r subcomponent1.271.132.091.28NM_0231431417009_at
C1scomplement component 1, s subcomponent1.09−1.212.261.30BC0221231424041_s_at
Calm3Calmodulin 3−2.081.091.02−1.15AV0475701438825_at
Cd24aCD24a antigen−1.032.281.321.33NM_0098461416034_at
Cluclusterin1.161.142.931.19NM_0134921418626_a_at
Cluclusterin−1.081.092.531.14AV1522881437689_x_at
Cluclusterin1.141.002.431.17BB4336781454849_x_at
Col12a1collagen, type XII, alpha 1−1.281.072.481.00BB1143981434411_at
Col1a1collagen, type I, alpha 1−1.23−1.112.071.10U080201423669_at
Colec12collectin sub-family member 12−1.131.802.721.15NM_1304491419693_at
Cpceruloplasmin1.041.192.451.52BB3324491417494_a_at
Cpceruloplasmin1.35−1.042.602.40BB3324491417495_x_at
Cpceruloplasmin−1.071.153.012.93BB3324491417496_at
Cpceruloplasmin1.06−1.092.772.62BB3324491448734_at
Cpceruloplasmin−1.101.062.701.18BB3324491448735_at
Cpceruloplasmin−1.011.152.491.29BB0090371455393_at
Creld2cysteine-rich with EGF-like domains 2−1.01−1.54−2.44−1.12AK0178801452754_at
Ctgfconnective tissue growth factor−1.021.372.28−1.35NM_0102171416953_at
Cxcl12chemokine (C-X-C motif) ligand 12−1.112.011.901.09BC0066401448823_at
Dcndecorin1.181.092.39−1.11NM_0078331449368_at
Dptdermatopontin1.111.312.811.17NM_0197591418511_at
Evi2aecotropic viral integration site 2a−1.051.332.141.26NM_0101611450241_a_at
Fabp4fatty acid binding protein 4, adipocyte1.26−1.97−1.04−3.25NM_0244061417023_a_at
Fabp4fatty acid binding protein 4, adipocyte1.04−2.16−1.04−3.39BC0021481451263_a_at
Fn1fibronectin 1−1.101.042.12−1.03BM2343601437218_at
Fstl1follistatin-like 1−1.201.282.241.22BI4527271416221_at
Fstl1follistatin-like 1−1.191.212.251.22BI4527271448259_at
Gstm2glutathione S-transferase, mu 21.221.211.382.05NM_0081831416411_at
Hmox1heme oxygenase (decycling) 12.32−1.16−1.251.10NM_0104421448239_at
Hs6st2heparan sulfate 6-O-sulfotransferase 2−1.121.012.28−1.04AW5364321450047_at
Lbplipopolysaccharide binding protein1.161.143.041.42NM_0084891448550_at
LOC640441 ///       
Thbs1similar to thrombospondin 1 /// thrombospondin 1−1.452.291.981.01AI3855321421811_at
Lrrc17leucine rich repeat containing 17−1.15−1.462.021.23BB5039351429679_at
Mgpmatrix Gla protein−1.11−1.063.08−1.13NM_0085971448416_at
Mgst1microsomal glutathione S-transferase 11.011.932.391.38BI1501491415897_a_at
Mmemembrane metallo endopeptidase1.04−1.112.151.70NM_0086041422975_at
Mrc2mannose receptor, C type 21.011.442.631.33BB5284081421045_at
Myadmmyeloid-associated differentiation marker1.01−1.322.18−1.46BB5000551439389_s_at
Pdgfbplatelet derived growth factor, B polypeptide−1.022.001.301.18BC0234271450414_at
Pdia4protein disulfide isomerase associated 4−1.14−1.10−2.221.29J051861416497_at
Pthlhparathyroid hormone-like peptide−1.142.332.261.39NM_0089701422324_a_at
Robo1roundabout homolog 1 (Drosophila)1.071.013.001.25BG0652301427231_at
Sdf2l1stromal cell-derived factor 2-like 1−1.08−1.28−2.281.01NM_0223241418206_at
Sema5asema domain, seven thrombospondin repeats (type 1 and type 1-like), (semaphorin) 5A−1.101.342.581.53AV3756531437422_at
Serinc3serine incorporator 3 solute carrier family 18 (vesicular monoamine), member−1.04−1.111.262.00NM_0120321448847_at
Slc18a221.18−1.251.052.11AV3346381437079_at
Steap2six transmembrane epithelial antigen of prostate 2−1.29−1.311.502.36BB5293321446071_at
Sulf1sulfatase 1−1.371.143.611.12BB7514591436319_at
Sulf1sulfatase 1−1.301.203.901.51BB0657991438200_at
TapbpTAP binding protein−1.131.292.001.29AF0439431421812_at
TapbpTAP binding protein−1.221.532.031.31AF0439431450378_at
Tfrctransferrin receptor−2.03−1.94−1.111.21BB8104501422966_a_at
Tfrctransferrin receptor−2.23−1.95−1.091.07BB8104501422967_a_at
Tfrctransferrin receptor−1.98−1.72−1.171.01AK0115961452661_at
AFFX-AFFX-      
Tfrctransferrin receptor−2.01−1.84−1.151.16TransRecMur/X57349_3TransRecMur/X57349_3_at
AFFX-AFFX-      
Tfrctransferrin receptor−2.25−1.97−1.111.04TransRecMur/X57349_5TransRecMur/X57349_5_at
AFFX-AFFX-      
Tfrctransferrin receptor−2.18−1.86−1.161.05TransRecMur/X57349_MTransRecMur/X57349_M_at
Tiam2T-cell lymphoma invasion and metastasis 21.162.171.241.03BM2289571423186_at
Tnfrsf9tumor necrosis factor receptor superfamily, member 91.121.032.131.81BC0285071428034_a_at
Trim25tripartite motif-containing 251.042.001.28−1.01AA9601661419879_s_at
Tspan13tetraspanin 13−1.331.163.991.24BB8077071418643_at
Tspan13tetraspanin 13−1.011.062.141.04BB8077071460239_at
Vegfcvascular endothelial growth factor C−1.102.801.831.02BB0891701439766_x_at

Several genes responsible for iron homeostasis were differentially expressed in response to SPIO/PLL labeling. Under conditions where intracellular iron is high, iron-binding proteins bind to iron responsive elements and thus regulate messenger RNA (mRNA) stability and translational activity of Tfrc and FtH, respectively. Accordingly, transferrin receptor 1 expression was strongly reduced 24-48 h after SPIO/PLL labeling (−2.25-fold) and exhibited a gradual recovery in expression over the time course (Table 1). Similarly, the expression of ferritin heavy chain was weakly upregulated on day 1 and day 2. Finally, ceruloplasmin (also called ferroxidase) expression was strongly elevated on day 4 and day 7 after SPIO/PLL labeling. Other genes encoding for proteins important for iron homeostasis, such as aconitase 1 and aconitase 2 (Aco1 and Aco2), transferrin (Tfr), and the transferrin receptor isoform 2 (Tfr2) were not affected by SPIO/PLL labeling (Fig. 1). The expression of Aco1 and Aco2 is weakly regulated by binding of IRPs to iron responsive elements in the 5′ untranslated region and both Trf2 and Tfr lack iron responsive elements (25). Overall, the pattern of expression exhibited by genes involved in iron regulation indicates that cytosolic iron is elevated following SPIO/PLL labeling and that the activity of IRPs at genes containing iron responsive elements may be increased. These array data corroborate a previous report that applied PCR and immunoblotting detect a decrease in Trfc and an increase in Fth1 in human MSCs and human ESCs that were labeled with SPIO (14).

Rather than characterizing genes independently, microarray analysis probes the expression pattern of 39,000 transcripts, providing a broad, unbiased overview of the genetic response to treatment (e.g., SPIO/PLL labeling) at a given point in time. Other than genes directly related to iron metabolism, genes exhibiting greater than a 2-fold change in expression can be functionally grouped into lipid binding (Fabp4, Lbp), receptor signaling (Adcy7, Angptl4, Calm3, Cxcl12, Pthlh, Sema5a, Sdf2l1, Tnfrsf9, Vegfc), and cell stress (Clu, Cp, Gstm2, Hmox1, Mgst1, Sluf1, Tnfrsf9) categories. Among these, clusterin (2.93-fold), ceruloplasmin (3.01-fold), sulfatase 1 (3.90-fold), and tetraspanin 13 (3.99-fold) were the most strongly upregulated genes. Fatty acid binding protein 4 (Fabp4) was the most strongly downregulated (−3.39-fold) gene (Table 1).

Time Course of Differential Gene Expression Profiles

Raw P values from false discovery rate calculations were used to rank genes in order to obtain the most likely differentially expressed genes over the 1-week time course. Of such ranked genes exhibiting the greatest difference in expression between SPIO/PLL and PLL conditions, 2695 probe sets, representing 1399 genes, were selected for further analysis. Gene ontology analysis was performed in order to determine whether the 1399 differentially expressed genes were enriched for genes related to specific cell properties (e.g., iron binding, oxidative stress, apoptosis). Gene ontology analysis segregates genes into two major categories: (1) molecular functions of the cell and (2) biologic processes. Genes within these two categories are further parsed into gene families with related function, and the probability that a particular gene family is enriched is determined. As genes often encode for proteins that subserve multiple cellular functions, a given gene may be represented within multiple categories.

In the group of genes regulating molecular functions of the cell (1309 genes), which describe activities such as catalytic reactions or binding that occur at the molecular level, 1001 genes are involved in catalytic activity (P = 0.000274), structural molecule activity (P = 0.0147), and transcription regulator activity (P = 0.000332) and binding (P = 1.95e-14), with the largest subpopulation of genes (196, P = 0.00195) coding for genes involved in zinc ion binding, with a large subpopulation of genes coding for ring finger proteins.

In the group of genes involved in biologic processes, series of events accomplished by one or more ordered assemblies of molecular functions (1198 genes from the group of chosen 2695 probe sets), based on a P value below 0.05 (the P value describes the significance level in the sense that the annotations do not occur by chance), 299 genes encode regulate gene expression (P = 1.5e-8), as well as cellular (970 genes, P = 0.0138), developmental (312 genes, P = 3.73e-5), and metabolic (685 genes, P = 1.36e-12) processes and biologic regulation (431 genes, P = 1.01e-8). The other groups of genes affected differently by SPIO/PLL than by PLL alone belonging to biologic processes and representing P ≤ 0.05, are secretion, transport, locomotion, reproductive processes, and establishment of localization.

Cluster Analysis of Gene Expression Profiles

Clustering algorithms sort different objects into groups such that the degree of association between two objects is maximal if they belong to the same group (e.g., genes with similar patterns of expression over time) and minimal otherwise. Cluster analysis simply discovers structures/relationships in data without explaining their biologic relevance. For each time point, there are three possibilities for gene expression: increased, decreased, or unchanged. Therefore, for the four time points in this study, there is a total of 81 (e.g., 34) unique gene expression profiles across time. Applying 81 clusters for K-means analysis accounted for 87% of explained variance (inertia ratio) in gene expression over time (Fig. 2). K-means clustering was done by correlation with data centroid-based search and yielded different clusters of the greatest possible distinction for the 2695 probe sets. The 81 clusters of transcripts show unique patterns of gene expression over the 7-day time course. Figure 3 demonstrates the expression patterns of 81 unique gene clusters. A list of the transcripts of interest, chosen from the list of most differentially regulated genes between SPIO/PLL- and PLL-treated cells, is provided in Table 2.

Figure 2.

Graph demonstrating 81 clusters explaining 87% of the explained variance (inertia ratio). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Figure 3.

K-means clustering of 2695 gene probes assembled into 81 clusters for SPIO/PLL relative to PLL labeling, illustrated in log2 scale of fold change with blinded dimensions to conditioning variables. Genes exhibiting potentially important expression profiles over time are indicated in Table 2.

Table 2. List of Selected Genes From the Set of 2695 Probes That Exhibited the Greatest Difference in Expression (SPIO/PLL Labeling Relative to PLL Treatment) Over the 1-Week Time Course Study*
Gene symbolGene titleClusterAccessionProbe number
  • *

    Genes were clustered into 81 independent groups according to the overall temporal pattern of expression.

Tfrctransferrin receptor76NM_0116381422966_a_at
Tfrctransferrin receptor76NM_0116381422967_a_at
Tfrctransferrin receptor76NM_0116381452661_at
Tfrctransferrin receptor76NM_011638AFFX-TransRecMur/X57349_3_at
Tfrctransferrin receptor76NM_011638AFFX-TransRecMur/X57349_5_at
Tfrctransferrin receptor76NM_011638AFFX TransRecMur/X57349_M_at
Npas3neuronal PAS domain protein 37NM_0137801450287_at
Nptx2neuronal pentraxin 259NM_0167891420720_at
Dbhdopamine beta hydroxylase41NM_1389421447592_at
Drd3dopamine receptor 333NM_0078771422278_at
Ldb2LIM domain binding 229NM_0010773981421101_a_at
Limd1LIM domains containing 151NM_0138601422731_at
Lhx1LIM homeobox protein 133NM_0084981450428_at
Lhx2LIM homeobox protein 266NM_0107101418317_at
Lhx9LIM homeobox protein 970NM_0010255651419324_at
Lhx9LIM homeobox protein 940NM_0010255651431598_a_at
Limk2LIM motif-containing protein kinase 231NM_0010340301439896_at
Apba1amyloid beta (A4) precursor protein binding, family A, member 146NM_1770341459605_at
Abcc4ATP-binding cassette, sub-family C (CFTR/MRP), member 451NM_0010333361443870_at
Abcc5ATP-binding cassette, sub-family C (CFTR/MRP), member 557NM_0137901447384_at
Abcd2ATP-binding cassette, sub-family D (ALD), member 265NM_0119941439835_x_at
Abcf1ATP-binding cassette, sub-family F (GCN20), member 151NM_0138541452236_at
Mrpl1mitochondrial ribosomal protein L160NM_0010390841460508_at
Mrpl15mitochondrial ribosomal protein L1552NM_0253001460490_at
Mrpl3mitochondrial ribosomal protein L317NM_0531591422463_a_at
Mrpl3mitochondrial ribosomal protein L317NM_0531591422464_at
Mrpl37mitochondrial ribosomal protein L3751NM_0255001423764_s_at
Mrpl38mitochondrial ribosomal protein L3851NM_0241771447961_s_at
Mrps25mitochondrial ribosomal protein S2566NM_0255781418716_at
Mrps9mitochondrial ribosomal protein S914NM_0235141435843_x_at
Rpl14ribosomal protein L1421NM_0259741426793_a_at
Rpl17ribosomal protein L1715NM_0010022391423855_x_at
Rpl22l1ribosomal protein L22 like 155NM_0265171417126_a_at
Rpl27ribosomal protein L2714NM_0011103391448217_a_at
Rpl27aribosomal protein L27a7NM_0119751426661_at
Rpl31ribosomal protein L3144NM_0532571441304_at
Rpl35ribosomal protein L3515NM_0255921436840_x_at
Rpl35ribosomal protein L358NM_0255921454856_x_at
Rpl37ribosomal protein L3761NM_0260691434872_x_at
Rpl39ribosomal protein L3935NM_0260551423032_at
Rpl4ribosomal protein L470NM_0242121425183_a_at
Rpl41ribosomal protein L4133NM_0188601446726_at
Rpl7ribosomal protein L753NM_0112911415979_x_at
Rpl7ribosomal protein L743NM_0112911426162_a_at
Rpl8ribosomal protein L87NM_0120531417762_a_at
Rps12ribosomal protein S1241NM_0112951447205_x_at
Rps13ribosomal protein S1315NM_0265331438794_x_at
Rps15aribosomal protein S15a55NM_1706691453467_s_at
Rps20ribosomal protein S2047NM_0261471456436_x_at
Rps24ribosomal protein S2430NM_0112971436064_x_at
Rps24ribosomal protein S2447NM_0112971456628_x_at
Rps25ribosomal protein S2556NM_0242661451068_s_at
Rps3ribosomal protein S344NM_0120521435151_a_at
Rps3ribosomal protein S374NM_0120521447563_at
Rps6ribosomal protein S666NM_0090961453466_at
Rps6ribosomal protein S621NM-0090961454620_x_at
Rps6ribosomal protein S643NM_0090961434377_x_at
Rps6ribosomal protein S660NM_009061435817_x_at
Rps8ribosomal protein S849NM_0090981436760_a_at
Rps9ribosomal protein S947NM_0297671433689_s_at
Rps9ribosomal protein S921NM_0297671434624_x_at
Ubap2lubiquitin associated protein 2-like64NM_0284751454643_at
Ube3aubiquitin protein ligase E3A74NM_0010339621431224_at
Usp24ubiquitin specific peptidase 2473XM_0014812811441018_at
Usp26ubiquitin specific peptidase 2663NM_0313881421502_at
Usp27xubiquitin specific peptidase 27, X chromosome77NM_0194611427606_at
Usp3ubiquitin specific peptidase 372NM_1449371441056_at
Usp36ubiquitin specific peptidase 3672XM_1267721458311_at
Usp50ubiquitin specific peptidase 5047NM_0291631430760_a_at
Usp53ubiquitin specific peptidase 539NM_1338571452385_at
Usp7ubiquitin specific peptidase 746NM_0010039181419920_s_at
Usp9xubiquitin specific peptidase 9, X chromosome46NM_0094811428193_at
Usp9yubiquitin specific peptidase 9, Y chromosome65NM_1489431452509_at
Ube2v1ubiquitin-conjugating enzyme E2 variant 151NM_0232301415755_a_at
Ube2v1ubiquitin-conjugating enzyme E2 variant 117NM_0232301444523_s_at
Ube2fubiquitin-conjugating enzyme E2F (putative)7NM_0264541429568_x_at
Ube2fubiquitin-conjugating enzyme E2F (putative)7NM_0264541451272_a_at
Ube2subiquitin-conjugating enzyme E2S21NM_1337771430962_at
Ufm1ubiquitin-fold modifier 121NM_0264351418899_at
Ubl4bubiquitin-like 4B65NM_0262611460272_at
Utyubiquitously transcribed tetratricopeptide repeat gene, Y chromosome72NM_0094841459565_at
Ndufa12NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 1264NM_0255511425919_at
Ndufb9NADH dehydrogenase (ubiquinone) 1 beta subcomplex, 946NM_0231721436803_a_at
Ndufs7NADH dehydrogenase (ubiquinone) Fe-S protein 751NM_0292721424313_a_at
Ndor1NADPH dependent diflavin oxidoreductase 172NM_0010824761428292_at
Noxo1NADPH oxidase organizer 157NM_0279881425151_a_at
Cybasc3cytochrome b, ascorbate dependent 365NM_2013511454895_at
Cox6a1cytochrome c oxidase, subunit VI a, polypeptide 146NM_0077481417417_a_at
Cox6a1cytochrome c oxidase, subunit VI a, polypeptide 164NM_0077481417418_s_at
Cyp2c37cytochrome P450, family 2. subfamily c, polypeptide 3755NM_0100011419094_at
Cyp2c44cytochrome P450, family 2, subfamily c, polypeptide 4440NM_0010014461424576_s_at
Cyp26a1cytochrome P450, family 26, subfamily a, polypeptide 178NM_0078111419430_at
Cyp27a1cytochrome P450, family 27, subfamily a, polypeptide 16NM_0242641457665_x_at
Cyb5r2cytochrome b5 reductase 215NM_1772161459448_at
Cyb5r3cytochrome b5 reductase 366NM_0297871422186_s_at
Cyb5r3cytochrome b5 reductase 361NM_0297871430734_at
Uqcrqubiquinol-cytochrome c reductase, complex III subunit VII20NM_0253521427880_at
Rc3h2ring finger and CCCH-type zinc finger domains 272NM_0011005911426925_at
Rnf11ring finger protein 1146NM_0138761452058_a_at
Rnf12ring finger protein 1278NM_0112761440850_at
Rnf138ring finger protein 13817NM_0197061419368_a_at
Rnf144bring finger protein 144B8NM_1460421443252_at
Rnf166ring finger protein 16664NM_0010331421448079_at
Rnf170ring finger protein 17069NM_0299651434956_at
Rnf180ring finger protein 18071NM_0279341453857_at
Rnf180ring finger protein 18081NM_0279341438306_at
Rnf185Ring finger protein 18533NM_1453551459877_x_at
Rnf187ring finger protein 18733XM_9057661423896_a_at
Rnf213Ring finger protein 21375NM_0010400051460018_at
Rnf4ring finger protein 417NM_0112781423654_a_at
Rnf41ring finger protein 4139NM_0262591432003_a_at
Rnf168ring fnger protein 16873NM_0273551455586_at

Iron affects genes of many different classes, but we first analyzed the genes involved in iron metabolism. A pronounced difference between SPIO/PLL and PLL treatments was observed for the transferrin receptor (Tfrc), where six probes representing this gene were assigned to cluster number 76, showing pronounced downregulation of Tfrc in SPIO/PLL-labeled cells in comparison to PLL (Fig. 3). In contrast, other genes encoding involved in iron-dependent regulation of iron metabolism, like genes for ferritin and ferroportin/Ireg1, were not classified to the group of 1399 genes.

Iron overload can also impact a number of genes associated with metabolism and energy production. For example, NADH: dehydrogenase (ubiquinone) 1 beta subcomplex, 9 (Ndufb9, cluster 46), NADH: dehydrogenase (ubiquinone) Fe-S protein 7 (Ndufs7, cluster 51), and NADH: dehydrogenase (ubiquinone) 1 alpha subcomplex, 12 (Ndufa12, cluster 61) are three genes that compose the NADH ubiquinone oxidoreductase complex and contain iron in their structure. Moreover, many genes encoding for cytochromes b, c, and P450 were also affected in a different temporal pattern by SPIO/PLL treatment rather than by PLL treatment alone.

Since the present experiment was performed on neural progenitors, it is critical to evaluate the impact of SPIO/PLL on neural genes. The time-course group of 1399 contains few genes that were affected distinctly by SPIO/PLL labeling relative to PLL. These genes include neuronal PAS domain protein 3 (Npas3, cluster 7), a factor that is involved in regulation of neurogenesis, dopamine beta hydroxylase (Dbh, cluster 41), which encodes for protein converting norepinephrine from dopamine, and dopamine receptor 3 (Drd3, cluster 33). SPIO/PLL had also unique impact on LIM homeobox proteins: Lhx1, (cluster 33), Lhx2, (cluster 66) and Lhx9 (clusters 40 and 70), which are transcriptional regulators of neuronal differentiation. Finally, another gene related to neural cells was amyloid beta (A4) precursor protein binding, family A, member 1 (Apba1, cluster 46), which encodes for neuronal adaptor protein that interacts with the Alzheimer's disease amyloid precursor protein. It stabilizes amyloid precursor protein and inhibits production of proteolytic amyloid precursor protein fragments, including the A beta peptide that is deposited in the brains of Alzheimer's disease patients.

Labeling with SPIO/PLL also appeared to have a significant impact on the expression of several genes that encode for ATP binding cassette proteins, two of which belong to subfamily C (CFTR/MRP) (Abcc), one to subfamily D (ALD), and one to subfamily F (GCN20). Member 4 of subfamily C (Abcc4) and member 1 of subfamily F (Abcf1) were ordered to cluster 51, whereas member 5 of subfamily C (Abcc5) was assigned to cluster 57, and member 2 of subfamily D (Abcd2), to cluster 65. Finally, several genes encoding ribosomal proteins and proteins involved in ubiquitination, a process of protein delivery to the proteasome for degradation, were also altered in response to SPIO/PLL labeling. These genes, along with their ordered clusters, are indicated in Table 2. Of interest, phosphorylation of ribosomal protein S6 closely corresponds to transferrin uptake, probably under the control of the PtdIns(3, 4, 5)P3-mTOR signaling pathway (26), and several ribosomal proteins are known to contain iron in their structures in the form of iron-sulfur clusters.

Validation of Microarray Analyses Using Quantitative Real-Time RT-PCR

To confirm the time course and single-time-point analyses, Rnf138 and Abcc4 were chosen for further characterization using quantitative real-time RT-PCR. Both of these genes are present within the 2695 differentially expressed genes ranked according to false discovery rate. Rnf138, also known as NARF (Nemo-like kinase associated ring finger), is an E3 ubiquitin-protein ligase that contains a conserved N-terminal C3HC4 (Cys3-His-Cys4) RING domain that functions to coordinate two zinc ions (5, 27). RING domains are found in a variety of functionally distinct proteins known to be involved in protein-DNA and protein-protein interactions. Because gene expression of RING domain family members was similarly regulated over the time course of SPIO/PLL treatment, we considered that free iron might affect the DNA binding activity via competition with zinc ion. Therefore, Rnf138 was selected for validation using real-time RT-PCR analysis. Similarly, Abcc4 is a member of another family of coordinately regulated genes that may be relevant for active extracellular transport of iron. Abcc4 (ATP binding cassette, subfamily C (CFTR/MRP), member 4), is a multidrug resistance–associated protein that binds and hydrolyzes ATP to drive the transport of a variety of endogenous and xenobiotic organic anions compounds, including cyclic nucleotides, conjugated steroid hormones, and eicosanoids, across the cell membrane against a diffusion gradient (28). As detected in time course and in single-time-point analyses, quantitative real-time RT-PCR revealed that both Rnf138 and Abcc4 exhibited a significant increase in gene expression in response to SPIO/PLL on day 2 postlabeling (P < 0.05; Figs. 4 and 5).

Figure 4.

Quantitative real-time RT-PCR analysis of Rnf138 mRNA expression in C17.2 NSCs. The expression of Rnf138 mRNA in PLL-treated versus SPIO/PLL-labeled C17.2 NSCs was analyzed using mRNA from samples isolated over the time course of the microarray study. The result values are expressed as the mean of quadruplicate real-time RT-PCR assays. Error bars denote standard deviations. After normalizing for β-actin mRNA content, statistically significant differences between untreated, PLL-treated, and SLIO/PLL-labeled C17.2 NSC were determined. *P < 0.05.

Figure 5.

Quantitative real-time RT-PCR analysis of Abcc4 mRNA expression in C17.2 NSCs. The expression of Abcc4 mRNA in PLL-treated versus SPIO/PLL-labeled C17.2 NSCs was analyzed using mRNA from samples isolated on day 2 of the microarray study. The result values are expressed as the mean of quadruplicate real-time RT-PCR assays. Error bars denote standard deviations. After normalizing for β-actin mRNA content, statistically significant differences between untreated, PLL-treated, and SLIO/PLL-labeled C17.2 NSC were determined. *P < 0.05.

DISCUSSION

Cellular labeling for serial MRI studies of nonphagocytic cells is most commonly achieved by coupling negatively charged SPIO nanoparticles to a cationic transfection agent, such as PLL or protamine sulfate, through electrostatic interactions (29, 30). SPIO/transfection agent complexes are then added to cell culture medium, where they are endocytosed by cells. Previous studies show that increasing amounts of iron used for cell labeling result in higher cellular concentration of this particle (18, 29). These works also report that increasing intracellular iron content does not have an effect on in vitro neuronal differentiation of C17.2 cells (18). To determine the more complex effect of magnetic nanoparticle labeling on cell function, we applied microarray technology and clustering algorithms to analyze overall gene expression by C17.2 NSCs in response to either SPIO/PLL or PLL transfection agent over a course of 1 week in vitro.

Analysis of relative fold differences in gene expression for each time point revealed that genes responsible for iron homeostasis, lysosomal function, detoxification, signal transduction, growth factors, and lipid peroxidation are differentially expressed by more than 2-fold following SPIO/PLL labeling. Relative to cells treated with PLL transfection agent, C17.2 NSCs labeled with SPIO/PLL complexes exhibited differential (either increased or decreased) expression of genes throughout the time course from which, based on the maximum difference in gene expression between SPIO/PLL- and PLL-treated cells, the top 2695 gene probes representing 1399 genes were further analyzed. Among the 1399 genes, genes encoding zinc-binding proteins were highly enriched. In comparison, genes encoding for proteins involved in apoptosis/programmed cell death did not reveal differences in expression.

Iron is an essential cofactor for cell proliferation, mitochondrial function, and oxygen transport (heme). However, an excess accumulation of intracellular free iron can oxidize and damage the protein, lipid, and nucleic acid components of cells. The expression of proteins important for iron transport (transferrin receptor) and iron storage (ferritin) is coordinately and reciprocally regulated at the posttranscriptional level by IRPs (IRP-1 and IRP-2), which act as sensors of cytosolic iron concentration (14). For example, when iron is abundant, IRP-1 lacks binding activity at iron-responsive elements found in the mRNA of target transcripts and IRP-2 is targeted for degradation by the ubiquitin-proteosomal system. In the absence of IRP binding, the translation of ferritin mRNA is no longer repressed and the half-life of transferrin receptor mRNA is reduced, resulting in sequestration of excess cytosolic iron within ferritin. The overall gene expression profile of C17.2 NSCs following SPIO/PLL labeling observed herein is indicative of such a cellular response to an elevation in intracellular free iron concentration.

SPIO nanoparticles consist of an iron oxide nanocrystal encased within a shell of crosslinked dextran. Following intravenous injection into the blood pool for medical diagnostic purposes, SPIO nanoparticles accumulate within the liver and spleen, where they are metabolized and molecular iron is incorporated into hemoglobin (31). Typically, between 5 and 20 pg of iron per cell, or between 50 and 200 times the normal endogenous cell level, is introduced by SPIO labeling, excluding erythrocytes, which contain ∼30 pg of iron (32). However, very little is known about the intracellular fate of SPIO nanoparticles following endocytosis. It has been proposed that the acidic milieu of the lysosomal compartment (∼pH 5.5) may disrupt the dextran shell and liberate Fe3+ from iron oxide nanoparticles (12). While mammalian cells lack dextranases, sulfatases are lysosomal enzymes that cleave a range of sulfated carbohydrates. Though it is not known whether the dextran shell of SPIO nanoparticles is sulfated by sulfotransferases within the cell, Sulf1 expression was strongly upregulated on day 4 (3.9-fold) after SPIO/PLL labeling. As shown for cellular iron overload (33), coordinated and reciprocal changes in transferrin receptor 1 and ferritin mRNA and protein are induced following SPIO labeling (13), suggesting that the integrity of the dextran shell is compromised and cytosolic free iron is increased; this may lead to increase of reactive oxygen species. However, previous studies demonstrated that the generation of reactive oxygen species is not induced after SPIO labeling (7, 13). While heme oxygenase 1 was transiently upregulated on day 1, a limited role for reactive oxygen species signaling after SPIO labeling is supported by the upregulation of ceruloplasmin expression at later time points, a protein that is responsible for the conversion of Fe2+ to the less oxidative form, Fe3+. Also, lipid peroxidation, another metabolic pathway influenced by changes in iron level, was not affected.

Gene expression profiling of liver tissue obtained from knockout mice lacking the gene that encodes for hereditary hemochromatosis protein or Hfe (34), which, when mutated, is the one of the major factors precipitating iron overload (35), revealed upregulation of genes involved in antioxidant defense and downregulation of genes involved in cholesterol metabolism. In comparison, gene profiling of skeletal and heart muscle from mice fed a high-iron diet (2% carbonyl iron) for 6 weeks exhibited limited changes in the expression of genes involved in cell stress responses and lipid metabolism despite a greater than 10-fold increase in hepatic iron content (36). These studies highlight tissue specific responses to iron overload and the capacity of certain tissues to sustain relatively normal gene expression patterns despite persistently elevated iron. Iron is particularly important for the development of oligodendrocytes in the central nervous system (37) and chronically or abnormally elevated iron is implicated in the pathogenesis of several central nervous system diseases (38). However, acute overnight exposure of C17.2 NSCs to SPIO for cell labeling appears to be well tolerated, as indicated by the early (e.g., day 1 and day 2) change in the expression of genes important for iron metabolism followed by the late (e.g., day 4 and day 7) induction of several genes essential for the control of cell stress (Cp, Hmox1, Gstm2).

By increasing the sampling density in microarray studies, it may be possible to further define immediate (e.g., 6 h or 12 h) and very long-term (e.g., 2 weeks and beyond) cellular responses to SPIO/PLL labeling. The time points selected for this study provide an optimal overview of the genomic response of C17.2 NSCs to SPIO/PLL labeling for a typical cell-tracking study. Other than the early induction of a homeostatic response in iron metabolism and the late changes in cell signaling and oxidative stress pathways, the pattern of gene expression found in C17.2 NSCs after overnight exposure to SPIO/PLL is highly similar to that of C17.2 NSC treated with PLL. The observed changes were not very pronounced, specifically, genes encoding for factors regulating programmed cell death, oxidative stress, neuronal cell function, and cell metabolism exhibited no remarkable change in expression, with few exceptions. In previous studies, it was shown that deferoxamine, an effective iron chelator, causes a decrease in expression of few mitochondrial genes involved in metabolism, between them NADH: ubiquinone oxidoreductase, known as complex 1 of respiratory electron transport chain (39). Though it is not known whether the effects of iron chelators are attributable to iron removal from target proteins or to indirect effects on intracellular iron stores, the present study revealed that Ndufs7, Ndufa12, and Ndufb9, each of which are parts of NADH ubiquinone oxidoreductase that contain iron in their structure, were differently affected in cells labeled with SPIO/PLL in comparison to treatment with PLL transfection agent. Also, a recent report about increased expression of Abcc4 in the presence of high iron accumulation in the cells of hfe −/− mice (40) is consistent with the results of SPIO/PLL NSC labeling.

Taken together, the genes that were differentially expressed in response to SPIO/PLL are indicative of a coordinated early homeostatic cellular response to increased intracellular iron concentration. However, while common themes are emerging from independent studies of cellular responses to acute (e.g., cell labeling) and chronic (e.g., genetic deficiency or dietary excess) iron overload, it is important to emphasize that the pattern of gene expression by C17.2 NSC in response to SPIO/PLL may differ from that of other cell types. Overall, intracellular labeling using SPIO/PLL in preparation for MR cell tracking studies poses no obvious adverse effects on cell viability, function, and gene expression, indicating that initiating clinical studies appears to be justified (5).

Acknowledgements

The C17.2 NSC line was obtained as a courtesy of Drs. E. Y. Snyder and J. H. Wolfe. The authors thank C. Connover Talbot Jr. (JHMI Microarray Core Facility) and Dr. Joe Nielsen (LDN, NINDS, National Institutes of Health) for assisting with analysis of gene expression data. We also thank Dr. Gregg Semenza (Vascular Biology Program, ICE, JHSOM) for allowing access to their Bio-Rad iCycler. This study was supported by 2RO1 NS045062 and NMSS RG3630. Dr. Jeff Bulte is a paid consultant for Surgivision Inc., a medical device company. This arrangement has been approved by The Johns Hopkins University in accordance with its Conflict of Interest policies.

Ancillary