Global Gene Expression Profile of Human Cord Blood-Derived CD133 + Cells

Global Gene Expression Profile of Human Cord Blood–DerivedCD133� Cells

TAINA JAATINEN,a HEIDI HEMMORANTA,a SAMPSA HAUTANIEMI,c JARI NIEMI,d DANIEL NICORICI,d

JARMO LAINE,a OLLI YLI-HARJA,d JUKKA PARTANENa,b

aResearch and Development and bDepartment of Tissue Typing, Finnish Red Cross Blood Service, Helsinki,

Finland; cBiological Engineering Division, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA;dInstitute of Signal Processing, Tampere University of Technology, Tampere, Finland

Key Words. Human cord blood • Hematopoietic stem cells • Microarray

ABSTRACT

Human cord blood (CB)–derived CD133� cells carry char-acteristics of primitive hematopoietic cells and proffer analternative for CD34� cells in hematopoietic stem cell (HSC)transplantation. To characterize the CD133� cell populationon a genetic level, a global expression analysis of CD133�

cells was performed using oligonucleotide microarrays.CD133� cells were purified from four fresh CB units byimmunomagnetic selection. All four CD133� samplesshowed significant similarity in their gene expression pat-tern, whereas they differed clearly from the CD133� controlsamples. In all, 690 transcripts were differentially expressedbetween CD133� and CD133� cells. Of these, 393 wereincreased and 297 were decreased in CD133� cells. The

highest overexpression was noted in genes associated withmetabolism, cellular physiological processes, cell communi-cation, and development. A set of 257 transcripts expressedsolely in the CD133� cell population was identified. Colony-forming unit (CFU) assay was used to detect the clonalprogeny of precursors present in the studied cell popula-tions. The results demonstrate that CD133� cells expressprimitive markers and possess clonogenic progenitor capac-ity. This study provides a gene expression profile for humanCD133� cells. It presents a set of genes that may be used tounravel the properties of the CD133� cell population, as-sumed to be highly enriched in HSCs. STEM CELLS 2006;24:631–641

INTRODUCTIONHematopoietic stem cells (HSCs), possessing self-renewingand differentiation potential, are required for the lifelongsustenance of a functional blood system. Stem cell transplan-tation is an established procedure in the treatment of hema-tological malignancies. Recently, stem cell transplantationhas been used as a therapy for many nonhematological dis-orders, including immunodeficiency syndromes, inborn er-rors of metabolism, and autoimmune diseases [1–3]. Morespecific transplants consisting of selected HSCs are requiredfor novel indications of stem cell transplantation, especiallywhen human leukocyte antigen-identical sibling donors arenot available. The use of T-cell depletion effectively dimin-ishes graft-versus-host disease, and the depletion of B cellsmay prevent Epstein-Barr virus-associated lymphoprolifera-tive disease [4, 5]. The number of primitive cells and theirproliferation capacity are considered preferable parametersfor the engraftment potential as compared with nucleatedcellularity [6, 7]. To increase the number of cells used in

transplantation and to promote ex vivo expansion of HSCs, agreater understanding of profitable cell populations is re-quired.

Human cord blood (CB) is an excellent source of HSCs.Rapidly available CB unit serves as an alternative for pa-tients without potential bone marrow (BM) donor. Lower riskof graft-versus-host disease and cytomegalovirus infectionis associated with CB transplantation. The comparison ofthe gene expression profiles of HSCs from peripheralblood (PB), BM, and CB suggests that CB-derived HSCs alsohave the potential to differentiate into cells of nonhemato-poietic lineages [8]. HSCs from different sources displayunique characteristics in terms of key transcription factorsand genes associated with cell cycle, homing, and apoptosis[8 –10]. HSCs from CB express transcription factors not seenin HSCs from other sources. Overexpression of these tran-scription factors may inhibit differentiation and might ex-plain the higher proliferation rate observed in CB-derivedHSCs [8].

Correspondence: Jukka Partanen, Ph.D., Department of Tissue Typing, Finnish Red Cross Blood Service, Kivihaantie 7, 00310Helsinki, Finland. Telephone: �358-9-5801298; Fax: �358-9-5801495; e-mail: [email protected] Received April 22,2005; accepted for publication September 28, 2005; first published online in STEM CELLS EXPRESS October 6, 2005. ©AlphaMedPress 1066-5099/2006/$20.00/0 doi: 10.1634/stemcells.2005-0185

STEM CELL GENETICS AND GENOMICS

STEM CELLS 2006;24:631–641 www.StemCells.com

The CD34 antigen has been the most widely used marker forHSC enrichment. Although the reconstruction of the adaptiveimmune system has been demonstrated with human CB-derivedCD34� progenitor cells in mice [11], the CD34� cell fraction isheterogeneous. The CD133 antigen provides a promising selec-tion marker for HSC enrichment. CD133� cells are consideredto be highly noncommitted with the capacity to self-renew anddifferentiate. In addition, CD133� cells have been shown tohave a higher clonogenic capacity than CD34�/CD133� cells[12]. Most of the CD133� cells are CD34�-bright, whereasCD34�-dim cells are CD133�. A small population of CD34�/CD133� cells (0.2%) has been found in CB, demonstrating thatCD133 expression is not necessarily associated with CD34expression [13].

The CD133 molecule has been found on the surface ofHSCs, neuronal stem cells, and embryonic stem cells (ESCs).Moreover, the expression of CD133 is related to several solidorgan malignancies, including lung and brain cancers [14, 15].A recent study demonstrates that only CD133� cancer stemcells are capable of brain tumor initiation while they sustain theability to self-renew and proliferate [15]. In addition, CB-de-rived CD133� cells may be able to differentiate into endothelialand neuronal cells [16].

The aim of this study was to characterize CB-derivedCD133� cells on a genomic level and to provide a global geneexpression profile of CD133� cells. The clonogenic progenitorcapacity of CD133� cells was demonstrated, showing that theyare highly noncommitted and hold the potential to differentiateinto all cell types of the hematopoietic system. The expressionanalysis presented in this study focuses on transcripts that areassociated with hematopoiesis and the cell cycle. The geneexpression data bank of CD133� cells may be used to study thepathogenesis of hematological diseases deriving from HSCs.

MATERIALS AND METHODS

CellsUmbilical CB was obtained from Helsinki Maternity Hospitaland the Department of Obstetrics and Gynaecology, HelsinkiUniversity Central Hospital, Finland. All donors gave informedconsent, and the study protocol was accepted by the ethicalreview board of the Helsinki University Central Hospital andFinnish Red Cross Blood Service. CB was collected in sterilecollection bags (Cord Blood Collection system; Medsep Corpo-ration, Covina, CA, http://www.medsep.com) containing citratephosphate dextrose solution and was processed within 4–20hours. All CB units tested negative for HIV, hepatitis C virus,hepatitis B virus, human T-cell lymphotropic virus, and syphilis.Mononuclear cells (MNCs) were isolated by Ficoll-Hypaquedensity gradient (Amersham Biociences, Piscataway, NJ, http://www.amersham.com). CD133� cells were enriched throughpositive immunomagnetic selection using CD133 Cell IsolationKit and magnetic cell sorting (MACS) affinity columns (Milte-nyi Biotec GmbH, Bergisch Gladbach, Germany, http://www.miltenyibiotec.com). CD133� cells were subjected to tworounds of separation. CD133� cells from the same CB unit werecollected for control purposes. Microarray analysis was per-formed using four separate CB units. In addition, six CB unitswere processed for quantitative real-time polymerase chain re-action (qRT-PCR) analysis.

Flow CytometryImmunomagnetically selected cells were labeled with phyco-erythrin (PE)– and fluorescein isothiocyanate (FITC)–conju-gated monoclonal antibodies (mAbs) to evaluate the purity ofcell fractions. Labeling was carried out using CD133/2-PE(clone 293C3; Miltenyi Biotec) and CD45-FITC (clone 2D1;Becton, Dickinson and Company, Franklin Lakes, NJ, http://www.bd.com) in 50 �l of phosphate-buffered saline (PBS) atroom temperature for 20 minutes. Isotype-identical mAbsIgG2b-PE and IgG1-FITC (Becton, Dickinson and Company)served as controls. Flow cytometry analysis was performed onFACSCalibur (Becton, Dickinson and Company) with a 488-nmblue argon laser. Fluorescence was measured using 530/30-nm(FITC) and 585/42-nm (PE) bandpass filters. Data were ana-lyzed using the CellQuest software (BD Biosciences, San Jose,CA, http://www.bdbiosciences.com) and Windows MultipleDocument Interface for Flow Cytometry, WinMDI version 2.8(http://facs.scripps.edu/software.html).

Colony-Forming Unit AssayColony-forming unit (CFU) assay was performed using meth-ylcellulose, MethoCult GF H4434 with recombinant cytokines,and erythropoietin (StemCell Technologies, Vancouver, BC,Canada, http://www.stemcell.com). A total of 2 � 103 CD133�

cells, 1 � 105 CD133� cells, or 1 � 105 MNCs were plated induplicate and cultured for 14 days at 37°C with 5% carbondioxide in a humidified atmosphere. Colonies were countedaccording to their morphological characteristics.

RNA IsolationTotal RNA from up to 2 � 107 pelleted cells was purified withRNeasy Mini Kit (Qiagen GmbH, Hilden, Germany, http://www1.qiagen.com) according to the manufacturer’s instruc-tions. Yield and quality of the RNA were measured by spectro-photometric analysis. Each sample was assessed for the integrityof RNA by discrimination of 18S and 28S ribosomal RNA on1% agarose using ethidium bromide for visualization.

Microarray AnalysisTotal RNA from each sample was used to prepare biotinylatedtarget RNA, with minor modifications from the manufacturer’srecommendations (http://www.affymetrix.com/support/techni-cal/manual/expression_manual.affx). In brief, first-strandcDNA was generated from 100 ng of total RNA using a T7-linked oligo(dT) primer. After the first cDNA synthesis cycle, invitro transcription was performed with unlabeled ribonucleo-tides. A second round of cDNA synthesis was then performedfollowed by in vitro transcription with biotinylated UTP andCTP (Enzo Biochem, Inc., Farmingdale, NY, http://www.enzo.com). Cleanup of double-stranded cDNA was performed usingPellet Paint Co-Precipitant (Novagen, Madison, WI, http://www.emdbiosciences.com/html/NVG/home.html) instead of Glyco-gen. Standard Affymetrix hybridization cocktail was added to15 �g fragmented cRNA. After overnight hybridization usingAffymetrix GeneChip Instrument System (Affymetrix, SantaClara, CA, http://www.affymetrix.com), arrays were washed,stained with streptavidin-phycoerythrin, and scanned on an Af-fymetrix GeneChip Scanner 3000. All experiments were per-formed using Affymetrix Human Genome U133 Plus 2.0 oligo-nucleotide arrays (http://www.affymetrix.com/products/arrays/

632 Gene Expression Profile of CD133� Cells

specific/hgu133plus.affx). The replicate results of hybridizationdata for CD133� and CD133� cells were obtained from fourindividual CB units. Sample labeling and hybridization werecarried out at the Finnish DNA Microarray Centre at TurkuCentre for Biotechnology, Turku, Finland.

Statistical AnalysisPearson correlation coefficient (m � 8, n � 54,612) was cal-culated for each sample pair using original signals values ob-tained from Operating Software detection algorithm. Pearsoncorrelation was also calculated for fold-change values of mi-croarray and qRT-PCR. The Pearson correlation coefficient, rik,between ith and kth samples, that is {y1i, y2i, . . . , yni} and {y1k,y2k, . . . , ynk}, respectively, is defined by

rik �

�j�1

n

� yji � yi�� yjk � yk�

�n � 1�sisk, for i � k and 1 � i, k � m,

(1)

in which

yk � �j�1

n

yjk/n and sk � ��j�1

n

�yjk � yk�/�n � 1� (2)

are the mean and SD of the kth sample, respectively.

Preprocessing and Filtering of Microarray DataThe Affymetrix GeneChip Operating Software detection algo-rithm was used to determine the presence or absence of expres-sion for each transcript. A transcript with either the detectioncall present or marginal was considered expressed. The com-plete gene expression data are available at http://qp01.novogroup.com/vpu. GeneChip Operating Software change al-gorithm was used to compare the CD133� data against theCD133� data to detect and quantify changes in gene expression.The transcripts assigned with change call increased, decreased,marginally increased, or marginally decreased were considereddifferentially expressed. The direction of change (increased ordecreased) was to be the same in all CD133� samples, and thefold-change cutoff value was set to 3.

Clustering and AnnotationTo identify and visualize the differences between the CD133�

and CD133� samples, two clustering algorithms, hierarchicalclustering and self-organizing map (SOM) with the componentplane representation, were applied [17, 18]. In hierarchical clus-tering, average and correlation were used as linkage and dis-tance metric, respectively. Hierarchical clustering was per-formed for all eight CD133� and CD133� samples. The SOMalgorithm clusters transcripts having a similar expression profilein the same neuron of the component plane. Accordingly, tran-scripts clustered close to each other are similar whereas topo-logically distant transcripts have dissimilar expression pattern.The component plane representation also includes a unified-matrix (U-matrix) representation, which can be used to identifyrobust clusters consisting of several neurons [18]. For the SOM,the mean expression across four CD133� and four CD133�

samples resulting in two component planes was used. The SOMtoolbox with Euclidean distance function, Gaussian neighbor-hood function, sheet SOM map with 15 � 9 neurons, and batchlearning algorithm was applied for the SOM analysis [19].Affymetrix GO Ontology Mining tool was employed to obtainmolecular functions, biological processes, and cellular compo-nents for the transcripts in the clusters. The statistically signif-icant hits were defined by �2 test and the associated p value withthe significance level at 5% (p � .05).

Gene PrioritizationTo order the genes according to their discriminatory power, astepwise gene selection algorithm was used [20]. Briefly, thealgorithm computes mean and SD across CD133� samples(��, ��) and across CD133� samples (��, � �). The weightfor the ith gene is computed using signal-to-noise ratio [21]:

wi ���i� � �i��

�i� � �i�

If a gene has a large magnitude weight, then the gene is stronglydifferentially expressed between CD133� and CD133� sam-ples, and variation in CD133� and CD133� is low.

Quantitative qRT-PCR AnalysisTo confirm the information obtained from the microarray data,10 genes (CD133, CD34, KIT, SPINK2, NOTCH1, SOX4, TIE,CD2, CD14, and CD45) were subjected to qRT-PCR analysisusing pools with three samples in each. Analysis was performedon two biological replicates. Total RNA was DNase-treated withDNA-free Kit (Ambion, Inc., Austin, TX, http://www.ambion.com), and reverse transcription was performed using High-Capacity cDNA Archive Kit with RNase Inhibitor Mix (AppliedBioSystems, Foster City, CA, http://www.appliedbiosystems.com) in a final volume of 100 �l. Thermal cycling conditionsfor reverse transcription were 25°C for 10 minutes and 37° for120 minutes on GeneAmp PCR System 9700 (Applied BioSys-tems).

For PCR, the template was added to PCR mix consisting of12.5 �l TaqMan Universal PCR Master Mix containing Ura-cil N-glycosylase for PCR carry-over prevention, 1.25 �l ofTaqMan Gene Expression Assays probe (Hs00156373_m1,Hs00195682_m1, Hs00174029_m1, Hs00221653_m1, Hs00413187_m1, Hs00268388_s1, Hs00178500_m1, Hs01040181_m1,Hs0069122_g1, Hs00365634_g1, Hs99999905_m1; AppliedBiosystems), and diethyl pyrocarbonate–treated water (Ambion,Inc.). Samples were assayed in triplicate in a total volume of 25�l. The qRT-PCR thermal cycling conditions were as follows:an initial step at 50°C for 2 minutes for Uracil N-glycosylaseactivation; 95°C for 10 minutes; and 40 cycles of 15 seconds at95°C and 1 minute at 60°C.

A standard curve for serial dilutions of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) template was simi-larly constructed. GAPDH was chosen as the internal controlbecause its expression levels had no variance between thesamples in the microarray analysis. Changes in fluorescencewere monitored using the ABI PRISM 7000 Sequence De-tection System (Applied BioSystems), and raw data wereanalyzed by Sequence Detection System 1.1 Software (Ap-plied BioSystems). The relative standard curve method was

633Jaatinen, Hemmoranta, Hautaniemi et al.

www.StemCells.com

used to balance the variation in the amount of cDNA and tocompensate for different levels of inhibition during reversetranscription and PCR.

RESULTS

Quality AssessmentTo ensure validity of the samples and preprocessed microarraydata, several methods were used for quality assessment. Thepurity of positively selected CD133� cells was more than 90%by flow cytometry, and the CD133� cell population was nearly100% pure (Fig. 1). Generally, 105–106 CD133� cells wererecovered from a CB unit and the viability of selected cells wasat least 99%. The integrity of total RNA was confirmed byspectrophotometry and agarose gel electrophoresis.

To ensure the uniformity and comparability of the biologicalreplicates, their pair-wise relationships were defined by Pearsoncorrelation coefficients. The Pearson correlation coefficientswere calculated for all the data points, excluding Affymetrixcontrol samples, thus 54,609 transcripts per array were com-pared. The consistency in all cases was high, but the correlationwithin CD133� samples was stronger than the correlation be-tween CD133� and CD133� samples. The correlation coeffi-cients between CD133� replicates had a mean of 0.98 (range,0.95–1.00). The correlation coefficients indicated significantsimilarity of the CD133� samples. The correlation coefficientsbetween CD133� and CD133� samples reached an average of0.78. In hierarchical clustering, the CD133� and CD133� sam-ples clustered at the opposite ends of the dendrogram. Theseresults demonstrate that the CD133� cells are much more sim-ilar to one another than to the CD133� cells from the sameindividual.

The microarray result for 10 selected genes was confirmedby qRT-PCR analysis. The average fold change was calculatedfor each gene and compared with the result from microarrayanalysis (supplemental online Table 1). The results correlatedstrongly (Pearson correlation coefficient, 0.95).

The Expression Profile of CD133� CellsThe comparison of CD133� and CD133� data sets resulted in690 transcripts that were differentially expressed at least three-

fold (supplemental online Table 2). In CD133� cells, 393 of thetranscripts were upregulated and 297 were downregulated. An-notation was found for 227 (58%) overexpressed transcripts,which encode molecules involved in biological processes rang-ing from metabolism to development (Fig. 2). A functional rolewas found for 221 (74%) of the underexpressed transcripts, theprotein products of which participate in cell communication,immune response, organogenesis, apoptosis, and chemotaxis(Fig. 2).

Two different clustering methods were applied to the set of690 transcripts passing the initial screening filter. Hierarchicalcluster analysis showed moderate variation in expression withina transcript between replicates. The expression of genes encod-ing CD133, CD34, and other transmembrane proteins, such asFLT3, LAPTM4B, EBPL, and CRIM1, had minor variance inall four CD133� samples. Other very similarly expressed tran-scripts were ANKRD28, several members of the HOX genefamily, and transcripts encoding hypothetical proteins. More-over, DKC1, BAALC, and JUP had minimal variation withinCD133� replicates. In contrast, slightly more variation wasobserved in the expression of KIT, a known stem cell marker.

The SOM was constructed using mean values of 690 differ-entially expressed genes between the CD133� and CD133�

samples (Fig. 3). The mean value was used to determine thesimilar expression behavior common to all CD133� samples.The SOM revealed four prominent clusters of genes distinguish-ing CD133� and CD133� cell populations. The clusters areillustrated by the U-matrix.

SOM clusters 1 and 2 represented upregulated genes, andclusters 3 and 4 comprised downregulated genes. In cluster 1,the association to a biological process was attained for 88 (57%)of the transcripts. The significantly represented biological pro-cesses were primary metabolism, cell proliferation, and regula-tion of transcription. In cluster 2, a functional role was foundfor 69 (59%) of the transcripts. The most significant func-tional categories were transcription and development. Cluster 3

Figure 1. Purity assessment of CD133� and CD133� cell fractions byflow cytometry. CD133� and CD133� cell populations were defined byfirst gating on forward and side scatter properties, excluding plateletsand debris. Subsequent gates were set to exclude �99% of control cellslabeled with isotype-specific antibody. Percentages indicating the purityof isolated cell fractions are shown for both plots. Abbreviations: PE,phycoerythrin; SSC, side scatter.

Figure 2. Biological processes represented by the differentially ex-pressed genes in CD133� cells.

634 Gene Expression Profile of CD133� Cells

contained a group of downregulated genes associated with cellcommunication and immune response. Annotation was foundfor 86 (76%) of the genes in cluster 3. In addition, cluster 4contained a number of genes whose protein products participatein signal transduction and response to stimulus. Moreover, thephosphorylation– and protein modification–related genes weredownregulated. Cluster 4 contained 64 (70%) transcripts withknown biological function.

In the SOM component plane, the most prominent findingwas that known HSC markers CD133, CD34, and KIT hadsimilar expression patterns and clustered into the same neuron.Interestingly, this neuron also contained the gene for SPINK2,expressed by 77-fold in microarray analysis. The markedly highexpression of SPINK2 was confirmed by qRT-PCR (fold change196). The role of SPINK2 is poorly understood but its expres-sion is seen in human BM CD34� cells and testicle tissues(http://genome.ucsc.edu/cgi-bin/hgNear).

CD133� Cell–Enriched GenesAltogether, 22,764 (42%) of the 54,675 transcripts on the arrayswere expressed in one or more of the CD133� samples. On eachCD133� array, a similar number of transcripts was expressedwith maximum variance of 0.8%. Upregulation was seen in6178 (11%) transcripts in at least one CD133� sample. Eachindividual CD133� sample had a similar number of unique geneexpressions. The common expression pattern for all fourCD133� samples encompassed 2285 upregulated transcripts. Ofthese, 2034 (89%) transcripts were overexpressed at least two-fold. The 2285 transcripts common for all CD133� samplesincluded genes whose protein products participate in cell com-munication, development, response to endogenous stimulus,chromosome organization, and biogenesis. Also, genes associ-ated with RNA processing and mRNA metabolism were signif-

icantly overexpressed. Annotated biological process was foundfor 1399 (61%) of the transcripts.

The expression of 257 transcripts was seen in CD133�

samples only (Fig. 4A; supplemental online Table 3). Thesetranscripts were absent in CD133� samples. Annotation wasfound for 155 (60%) transcripts. The most significantly repre-sented biological processes among this set were DNA metabo-lism, cell proliferation, and regulation of transcription (Fig. 4B;Table 1). The transcripts expressed in CD133� cells containedonly 32 genes encoding potential integral membrane proteinsthat may serve as markers for HSCs (supplemental online Table4). In addition, the 257 transcripts common for CD133� sampleswere ranked using a gene prioritization method. The gene codingfor transmembrane protein LAPTM4B, overexpressed by 26-fold,got the highest weight value in prioritization.

Cell CycleThe expression data were surveyed to establish the cell cyclestate of CD133� cells. The expression of GATA2 (fold change,7.0) and N-MYC (fold change, 15) that keep the HSCs inundifferentiated state was significantly elevated in CD133�

cells [22, 23]. The downregulation of these genes would initiatethe cell cycle. DST (fold change, 5.3) and PLAGL1 (fold change,9.1), which support cell cycle arrest, were upregulated as well.A cell cycle inhibitor and negative regulator of proliferation,NME1, was overexpressed in CD133� cells by 3.7-fold.

Figure 3. Classification of CD133� and CD133� samples by meanself-organizing map (SOM) analysis. (A): The four clusters determinedby unified matrix (U-matrix). (B): Mean SOM component planes forCD133� and CD133� samples.

Figure 4. Common transcripts expressed in CD133� cells. (A): Sche-matic representation of intersections and differences in CD133� cells.Only transcripts expressed in CD133� cells but absent in CD133� cellswere included. (B): Categorization of common genes expressed inCD133� cells based on Gene Ontology annotation.

635Jaatinen, Hemmoranta, Hautaniemi et al.

www.StemCells.com

Table 1. The genes representing the most significant biological processes in CD133� cells

Symbol Name Unigene ID

Signal transductionDPYSL3 Dihydropyrimidinase-like 3 519659ALCAM Activated leukocyte cell adhesion molecule 150693SOCS2 Suppressor of cytokine signaling 2 485572PLCB4 Phospholipase C, 4 472101RBM14 RNA binding motif protein 14 11170TGFBRAP1 Transforming growth factor, -receptor associated protein 1 446350CRHBP Corticotropin-releasing hormone-binding protein 115617ITGA9 Integrin, 9 113157HRMT1L2 HMT1 hnRNP methyltransferase-like 2 (Saccharomyces cerevisiae) 20521FLT3 Fms-related tyrosine kinase 3 507590PDE1A Phosphodiesterase 1A, calmodulin-dependent 416061DDAH1 Dimethylarginine dimethylaminohydrolase 1 379858GPR125 G protein-coupled receptor 125 99195AKT3 V-akt murine thymoma viral oncogene homolog 3 (protein kinase B, �) 498292RHOBTB1 Rho-related BTB domain-containing 1 148670PTPRD Protein tyrosine phosphatase, receptor type, D 446083TCF7L2 Transcription factor 7-like 2 (T-cell specific, HMG-box) 501080MAP3K4 Mitogen-activated protein kinase kinase kinase 4 390428ITPR1 Inositol 1,4,5-triphosphate receptor, type 1 374613TNFRSF21 Tumor necrosis factor receptor superfamily, member 21 443577CYTL1 Cytokine-like 1 13872PILRB Paired immunoglobin-like type 2 receptor 530084LRP6 Low-density lipoprotein receptor-related protein 6 210343C12orf2 Chromosome 12 open reading frame 2 269941MAGI1 Membrane associated guanylate kinase interacting protein-like 1 16064SOCS6 Suppressor of cytokine signaling 6 44439DST Dystonin 485616ERG V-ets erythroblastosis virus E26 oncogene-like (avian) 473819

DNA metabolismPOLD2 Polymerase (DNA-directed), � 2, regulatory subunit 50 kDa 306791RUVBL2 RuvB-like 2 (Escherichia coli) 515846MCM5 MCM5 minichromosome maintenance deficient 5, cell division cycle 46

(S. cerevisiae)517582

BAT8 HLA-B–associated transcript 8 520038SMARCA1 SWI/SNF-related, matrix-associated, actin-dependent regulator of

chromatin, subfamily a, member 1152292

POLA Polymerase (DNA-directed), 495880RAD52 RAD52 homolog (S. cerevisiae) 525220ATR Ataxia telangiectasia and Rad3-related 271791MSH5 MutS homolog 5 (E. coli) 371225MCM7 MCM7 minichromosome maintenance deficient 7 (S. cerevisiae) 438720CBX5 Chromobox homolog 5 (HP1 homolog, Drosophila) 349283SMARCAL1 SWI/SNF-related, matrix-associated, actin-dependent regulator of

chromatin, subfamily a-like 1516674

DNMT3A DNA (cytosine-5-)-methyltransferase 3 515840CBX2 Chromobox homolog 2 (Pc class homolog, Drosophila) 368410

Response to stimulusHSPCB Heat shock 90-kDa protein 1, 509736CHML Choroideremia-like (Rab escort protein 2) 170129SERPING1 Serine (or cysteine) proteinase inhibitor, clade G (C1 inhibitor), member 1

(angioedema, hereditary)384598

SEPP1 Selenoprotein P, plasma, 1 275775HSPB1 Heat shock 27-kDa protein 1 520973TCEA2 Transcription elongation factor A (SII), 2 505004IGLL1 Immunoglobulin -like polypeptide 1 348935D2S448 Melanoma-associated gene 332197

Cell proliferationAREG Amphiregulin (schwannoma-derived growth factor) 270833IGFBP7 Insulin-like growth factor binding protein 7 479808EMP1 Epithelial membrane protein 1 436298

(continued)

636 Gene Expression Profile of CD133� Cells

Most of the CB-derived HSCs have been shown to be inG0 [24]. However, factors promoting the G1 phase, such asCDK6 (fold change, 10) and BCAT1 (fold change, 19), wereoverexpressed along with CDK4 (fold change, 3.9), which

acts in the G1/S transition. The negative regulator of CDK4and CDK6, p18, was underexpressed by 5.1-fold. Moreover,the overexpression of BMI-1 was observed by 2.8-fold.BMI-1 enhances the cell cycle by inhibiting p16, the negativeregulator of the cell cycle. As expected, p16 was not ex-pressed in CD133� cells.

The S phase was demonstrated by high expression of genesencoding minichromosome maintenance proteins crucial inDNA replication. Known S-phase inducers MCM2 (fold change,3.1), MCM5 (fold change, 4.2), MCM6 (fold change 2.5), andMCM7 (fold change, 2.8) were upregulated. Interestingly,CDK2AP1, a suppressor of DNA replication, was overexpressedby fourfold and CDKN2D, needed in S phase, was underex-pressed by 20-fold. However, the low expression of CDKN2Drefers to G1 phase [25]. No known transcripts encoding mole-cules acting in G2 phase or G2/M transition were seen. Severaltranscripts coding for molecules with ubiquitin-protein ligaseactivity, such as SH3MD2, UHRF1, ZNRF1, EDD, and TIF1,were overexpressed more than threefold. Many cell cycle

Figure 5. Clonogenic progenitor cell capacity of CD133�, CD133�,and MNC populations. Abbreviations: CFU, colony-forming unit;MNC, mononuclear cell.

Table 1. Continued

Symbol Name Unigene ID

PAWR PRKC, apoptosis, WT1, regulator 406074LDOC1 Leucine zipper, downregulated in cancer 1 45231MPHOSPH9 M-phase phosphoprotein 9 507175NDN Necdin homolog (mouse) 50130DKC1 Dyskeratosis congenita 1, dyskerin 4747SKB1 SKB1 homolog (Schizosaccharomyces pombe) 367854CCNL2 Cyclin L2 515704ANAPC7 Anaphase promoting complex subunit 7 529280CDK6 Cyclin-dependent kinase 6 119882

TransportNUP93 Nucleoporin 93 kDa 276878SV2A Synaptic vesicle glycoprotein 2A 516153CPT1A Carnitine palmitoyltransferase 1A (liver) 503043ICA1 Islet cell autoantigen 1, 69 kDa 487561KPNB1 Karyopherin (importin) 1 532793OSBPL1A Oxysterol binding protein-like 1A 370725ATP9A ATPase, class II, type 9A 368002COL5A1 Collagen, type V, 1 210283FLVCR Feline leukemia virus subgroup C cellular receptor 7055SYTL4 Synaptotagmin-like 4 (granuphilin-a) 522054SFXN1 Sideroflexin 1 369440SLC25A27 Solute carrier family 25, member 27 40510SLC16A14 Solute carrier family 16 (monocarboxylic acid transporters), member 14 504317UNQ9438 TIMM9 534663MIPEP Mitochondrial intermediate peptidase 507498

DevelopmentMAP7 Microtubule-associated protein 7 486548HLF Hepatic leukemia factor 196952ADAM28 A disintegrin and metalloproteinase domain 28 528304WHSC1 Wolf-Hirschhorn syndrome candidate 1 113876HOXA9 Homeo box A9 127428TRO Trophinin 434971HOXA10 Homeo box A10 110637GCNT2 Glucosaminyl (N-acetyl) transferase 2, I-branching enzyme 519884LMO2 LIM domain only 2 (rhombotin-like 1) 34560

Abbreviations: HMG, high mobility group; HMT, heavy metal tolerance; HP, haptoglobin; LIM, an acronym derived from three geneproducts, Lin-11, Is1-1, and Mec-3; MCM, minichromosome maintenance deficient; PRKC, protein kinase C; SKB, SKH1 kinasebinding protein; SWI/SNF, mating type switching/sucrose nonfermenting; TIMM, translocase of inner mitochondrial membrane; WT,Wilms tumor.

637Jaatinen, Hemmoranta, Hautaniemi et al.

www.StemCells.com

regulatory molecules are controlled by ubiquitin-mediated pro-teolysis to regulate the number of cells in each phase of the cellcycle [26]. Genes associated with mitosis, such as SKB1,STAG1, ANAPC7, and MPHOSPH9, were overexpressed by2.6-fold, 1.6-fold, 2.6-fold, and 3.1-fold, respectively. Thesedata suggest that a portion of CD133� cells are cycling.

HematopoiesisThe expression of genes associated with self-renewal and dif-ferentiation was studied to unravel the hematopoietic state ofCD133� cells. Several HSC-associated genes were overex-pressed: CD133 by 60-fold, CD34 by 13-fold, KIT by 26-fold,TIE by 3.2-fold, SCA-1 by 2.1-fold, MEIS1 by 10-fold, and

ANGPT1 by 12-fold. Genes supporting self-renewal, such asGATA2, MPLV, STAT5A, and TCF7L2, were upregulated by7.0-fold, 12-fold, 1.9-fold, and 3.3-fold, respectively. Hoxgenes, thought to be involved in HSC regulation, were alsohighly upregulated. The expression of HOXA9 (fold change,130) induces stem cell expansion, and HOXA5 (fold change, 10)and HOXA10 (fold change, 3.7) are specific to the long-termrepopulating population of HSCs [27, 28]. Upregulation ofGATA2 and other transcription factors supporting self-renewalmay account for the differentiation arrest and support the moreprimitive nature of CB-derived HSCs [8]. The previously re-ported early markers for hematopoeitic progenitors, BAALC andC17 [29–31], were expressed in CD133� cells only. TheBAALC gene was overexpressed by 33-fold. The expression ofBAALC has been shown in brain tissue, yet its functional role isunknown [30]. The overexpression of C17, a gene coding for anextracellular molecule with signal transduction activity, was15-fold.

AML1, overexpressed by 2.5-fold in CD133� cells, mayalso support HSC self-renewal although it has been character-ized as an early differentiation marker of the myeloid lineage.The other early myeloid differentiation gene, PU.1, was absent.GATA1, which affects erythropoiesis, and PAX5, which pro-motes B-precursor development, were both absent. No changein expression of GFI1 leading to T-lymphoid differentiation wasdetected. NFE2, required for HSCs determination to mega-karyocyte and erythrocyte lineage, was downregulated.

The expression of lineage-determination markers glyco-phorin-A, CD38, CD7, CD33, CD56, CD16, CD3, or CD2 wasundetected in CD133� cells. The expression of CD45 was seenin CD133� cells, but it was downregulated. The CD45 antigenis abundant in lymphoid cells, covering approximately 10% ofthe cell surface. The gene expression results suggest a naivestate for the CD133� cell population, containing long-term andshort-term repopulating HSCs as well as early progenitors withmyeloid and lymphoid lineage potential.

CFU assay was used to identify primitive hematopoieticcells from CD133�, CD133�, and MNC fractions by stimulat-ing them to express their developmental potential (supplemen-tary online Table 5). Total CFU (CFU-TOT) number was de-termined as the sum of granulocyte-erythroid-macrophage-megakaryocyte (CFU-GEMM), granulocyte-macrophage (CFU-GM), erythroid (CFU-E), and burst-forming erythroid (BFU-E)colonies (Fig. 5). CFU-TOT counts were 80, 0.58, and 1.09 per1000 cells for CD133�, CD133�, and MNC populations, re-spectively. The highest proportion of CD133� cells formedCFU-GM colonies (58%) and CFU-GEMM colonies (38%).BFU-E represented 4.2% of the colonies, yet CFU-E colonieswere not observed. Taken together, CD133� is a valid selectionmarker for HSC enrichment. The clonogenic progenitor capacityof CD133� cells demonstrates that they are highly noncommit-ted and hold the potential to differentiate into all cells in thehematopoietic system.

DISCUSSIONThe gene expression profile of human HSCs, especially CD34�

cells, has been reported from various sources [8–10, 27, 32–34].This study characterizes the gene expression profile of CB-derived cells selected using CD133, a marker thought to bespecific for HSCs. Altogether, 42% of the transcripts on the

Table 2. Common genes between the present study andpublished data on cord blood–derived hematopoietic stem cells

Genesymbol

Presentstudy

He et al.(2005)37

Georgantaset al.

(2004) 9

Wagneret al.

(2004)32

CD34 � �KIT � �MEIS1 � �HOXA9 � �GATA2 � �CRIM1 � �DAPK1 � �ALDH1A1 � �AKR1C3 � �ME3 � �CRHBP � �JUP � �KPNB1 � �GNA15 � �IQGAP2 � �MEST � �NME1 � �VAV3 � �TCF4 � �TIF1 � �HHEX � �CD133 � � �FHL1 � � �RBPMS � � � �FLJ14054 � � � �SPINK2 � � �NRIP1 � � �KIAA0125 � � �SOCS2 � �GUCY1A3 � �HLF � �HOXA3 � �KIAA1102 � �MLLT3 � �NPR3 � �PLS3 � �TFPI � �ERG � �D2S448 � �PLCB1 � �SEPP1 � �

�, present.

638 Gene Expression Profile of CD133� Cells

arrays were expressed in one or more of the CD133� samples.The great number of expressed transcripts in CD133� cells maybe due to the open chromatin structure of HSCs [9, 35, 36]. Inall, 690 transcripts were found to be differentially expressedbetween CD133� and CD133� cells. Among these were manygenes encoding known stem cell markers and genes coding forhematopoietic regulators. The genes encoding mature hemato-poietic markers were not expressed in CD133� cells, whereastheir expression was detected in CD133� cells.

Hierarchical cluster analysis presented a set of 537 tran-scripts with differential expression between CD133� andCD133� cells. The expression pattern of these transcripts wassimilar within all CD133� and CD133� samples, and the levelof expression was uniform. Some transcripts showed variationin their expression level between biological replicates eventhough the direction of change was the same. The variance ofexpression level in CB-derived HSCs is known to be higher thanin HSCs from other sources [8]. The higher individual variancemay be explained by the unique birth event in each case.

SOM clustering demonstrated that the biological processesassociated with upregulated or downregulated genes were diver-gent. SOM clustering segregated genes into separate neurons, pro-viding sets of genes with a similar gene expression pattern. Thegenes associated with cell growth and maintenance, transcriptionalactivity, and cell cycle were significantly overexpressed inCD133� cells. The emphasized activity of these biological pro-cesses is known to be representative of hematopoietic progenitorcells [10, 34, 37]. In contrast, the CD133� cell fraction displayeda significantly elevated number of genes whose protein productsparticipate in immune response and reaction to stimulus, corre-sponding to the expression pattern of mature blood cells.

SOM analysis was performed on the 690 differentially ex-pressed genes. It revealed that SPINK2 had a similar expressionpattern with known HSC markers CD133, CD34, and KIT. Theincreased expression of SPINK2 has recently been described inCB-derived CD34�CD133� cells [37]. The decreased expressionof SPINK2 in testis has been shown to be associated with infertility[38]. Similarly, CD133 has been suggested to take part in theformation of spermatozoa and thereby have a significant role inmale fertility [39]. CD133 expression is assumed to affect theformation of lamellipodia, enabling HSC migration [32].

In this study, the main focus of the expression analysis was ongenes related to cell cycle and hematopoiesis. A number of differ-entially expressed genes involved in these processes were identifiedin CD133� cells. According to the literature, most of the CD133�

cells reside in the G0/G1 state of the cell cycle [40, 41]. However,certain enhancers of cell cycle and S-phase inducers were upregu-lated in CD133� cells, suggesting that a portion of these cells maybe cycling. Genes supporting self-renewal and differentiation arrestwere highly expressed in CD133� cells. The expression pattern ofCD133� cells alludes to proliferation activity.

Furthermore, the expression of genes encoding cell adhesionmolecules related to functionally important processes in HSCmigration and homing was examined. Among the 690 differen-tially expressed genes, 11 that encode adhesion molecules wereupregulated in CD133� cells. The overexpression of these genes(CD34, IL-18, JUP, DST, COL5A1, TRO, DSG2, ITGA9,SEPP1, PKD2, and VAV3) is also associated with cell cyclearrest and response to external stress. The 16 downregulatedgenes associated with cell adhesion encoded known mature cell

markers, such as CD2 and CD36. Several genes encoding che-mokines and integrins were downregulated. The low or unde-tectable expression of genes associated with migration probablyrelates to CB as the source of the CD133� cells, as the CBmicroenvironment differs from that of BM. The engraftmentpotential of CB-derived HSCs is known to be delayed comparedwith other sources of HSCs [42]. The gene coding for VLA-4,needed for HSC homing, was upregulated. The upregulation ofVLA-4 has been shown to be crucial to HSC engraftment in mice[43]. CB-derived HSCs have higher long-term engraftment ca-pacity, and their engraftment potential is significantly greater ascompared with BM and PB [8, 44].

A set of 257 transcripts, expressed solely in CD133� cells,was found. This set encompassed several genes coding forputative integral membrane proteins. The expression and local-ization of these proteins cannot be deduced from the presentdata and are a subject of further investigations. Of the commongenes expressed in CD133� cells, LAPTM4B got the highestweight value in gene prioritization. The overexpression ofLAPTM4B has been detected in mouse and human ESCs, HSCs,and neuronal stem cells by several independent studies [27, 45,46]. LAPTM4B has no known biological function, but someobservations link its upregulation to certain cancer cell lines andpoor differentiation of human hepatocellular carcinoma tissues[47]. For 125 of the 257 transcripts, a biological function couldnot be found. These novel genes may serve as the basis forfurther studies on HSC regulation.

When comparing the expression data of CD133� cells withpublished data on HSCs from CB, the highest similarity wasseen with CD34�CD133� cells [37]. Among the differentiallyexpressed genes found in CD34�CD133� cells, 28 were over-expressed also in the CD133� cell population. In addition, 14common genes between CD34�CD38�Lin� and CD133� cellpopulations were found [9]. CD34�CD38� cells showed simi-larity to CD133� cells, having 10 genes in common [32]. Thecommon genes are listed in Table 2. RBPMS and FLJ14054were identified in all four studies. The expression of RBPMS hasbeen shown in heart, prostate, intestine, and ovary, but itsexpression level is low in skeletal muscle, spleen, thymus, brain,and peripheral leukocytes [48]. RBPMS has been suggested tohave tissue-specific alternative splicing and may play a role inRNA metabolism [48]. A few ESC-related stem cell markers,such as DNMT3B, DNMT3A, and DPPA4, were overexpressedin CD133� cells as well [46, 49–53]. Transcriptional evidenceof ESC-related genes is a sign of the primitive nature of CB-derived CD133� cells. CB-derived CD133� cells have beenshown to have nonhematopoietic differentiation potential withthe capacity to develop into endothelial and neuronal cells [16].Comparison of different cell populations, based on publisheddata, is troublesome due to differences in the cell populations,platforms, and preprocessing methods used. Furthermore, thenomenclature of genes is inconstant. To find true overlap be-tween different data sets, unprocessed data should be used.

This study provides a global gene expression profile for humanCB-derived CD133� cells. The clonogenic progenitor activity ofCD133� cells was demonstrated, showing that the CD133� cellfraction is an excellent source of HSCs. The gene expressionprofile of CD133� cells may be used to study the pathogenesis ofhematological disorders and development of malignancies. Animproved understanding of CD133� cells furthers their potential in

639Jaatinen, Hemmoranta, Hautaniemi et al.

www.StemCells.com

therapeutic applications. This study provides additional informa-tion regarding previous HSC gene expression analyses. Combiningall published data would bring the scientific community closer tounraveling the riddle of HSCs.

ACKNOWLEDGMENTSWe thank the staff of the Finnish Red Cross Blood Service CordBlood Bank. We also acknowledge Miina Miller for technical help

with microarray analysis and Sirkka Mannelin for help with CFUassay. Tuija Kekarainen is acknowledged for help with flow cy-tometry analysis and for valuable comments on the manuscript.

DISCLOSURESThe authors indicate no potential conflicts of interest.

REFERENCES

1 Bielorai B, Trakhtenbrot L, Amariglio N et al. Multilineage hematopoi-etic engraftment after allogeneic peripheral blood stem cell transplanta-tion without conditioning in SCID patients. Bone Marrow Transplant2004;34:317–320.

2 Gratwohl A, Passweg J, Bocelli-Tyndall C et al. Autologous hematopoi-etic stem cell transplantation for autoimmune diseases. Bone MarrowTransplant 2005;35:869–879.

3 Grigull L, Beilken A, Schrappe M et al. Transplantation of allogeneicCD34-selected stem cells after fludarabine-based conditioning regimenfor children with mucopolysaccharidosis 1H (M. Hurler). Bone MarrowTransplant 2005;35:265–269.

4 Lang P, Handgretinger R, Niethammer D et al. Transplantation of highlypurified CD34� progenitor cells from unrelated donors in pediatricleukemia. Blood 2003;101:1630–1636.

5 Handgretinger R, Schumm M, Lang P et al. Transplantation of mega-doses of purified haploidentical stem cells. Ann N Y Acad Sci 1999;872:351–361.

6 Mavroudis D, Read E, Cottler-Fox M et al. CD34� cell dose predictssurvival, posttransplant morbidity, and rate of hematologic recovery afterallogeneic marrow transplants for hematologic malignancies. Blood1996;88:3223–3229.

7 Migliaccio AR, Adamson JW, Stevens CE et al. Cell dose and speed ofengraftment in placental/umbilical cord blood transplantation: graft pro-genitor cell content is a better predictor than nucleated cell quantity.Blood 2000;96:2717–2722.

8 Ng YY, van Kessel B, Lokhorst HM et al. Gene-expression profiling ofCD34� cells from various hematopoietic stem-cell sources reveals func-tional differences in stem-cell activity. J Leukoc Biol 2004;75:314–323.

9 Georgantas RW III, Tanadve V, Malehorn M et al. Microarray and serialanalysis of gene expression analyses identify known and novel tran-scripts overexpressed in hematopoietic stem cells. Cancer Res 2004;64:4434–4441.

10 Steidl U, Kronenwett R, Rohr UP et al. Gene expression profilingidentifies significant differences between the molecular phenotypes ofbone marrow-derived and circulating human CD34� hematopoietic stemcells. Blood 2002;99:2037–2044.

11 Traggiai E, Chicha L, Mazzucchelli L et al. Development of a humanadaptive immune system in cord blood cell-transplanted mice. Science2004;304:104–107.

12 de Wynter EA, Buck D, Hart C et al. CD34�AC133� cells isolated fromcord blood are highly enriched in long-term culture-initiating cells,NOD/SCID-repopulating cells and dendritic cell progenitors. STEMCELLS 1998;16:387–396.

13 Gallacher L, Murdoch B, Wu DM et al. Isolation and characterization ofhuman CD34(�)Lin(�) and CD34(�)Lin(�) hematopoietic stem cellsusing cell surface markers AC133 and CD7. Blood 2000;95:2813–2820.

14 Hilbe W, Dirnhofer S, Oberwasserlechner F et al. CD133 positiveendothelial progenitor cells contribute to the tumour vasculature innon-small cell lung cancer. J Clin Pathol 2004;57:965–969.

15 Singh SK, Hawkins C, Clarke ID et al. Identification of human braintumour initiating cells. Nature 2004;432:396–401.

16 Baal N, Reisinger K, Jahr H et al. Expression of transcription factorOct-4 and other embryonic genes in CD133 positive cells from humanumbilical cord blood. Thromb Haemost 2004;92:767–775.

17 Eisen MB, Spellman PT, Brown PO et al. Cluster analysis and display ofgenome-wide expression patterns. Proc Natl Acad Sci U S A 1998;95:14863–14868.

18 Hautaniemi S, Yli-Harja O, Astola J et al. Analysis and visualization ofgene expression microarray data in human cancer using self-organizingmaps. Machine Learning 2003;52:45–66.

19 Vesanto J, Himberg J, Alhoniemi E et al. SOM Toolbox for Matlab 5,Technical Report A57. Helsinki, Finland: Helsinki University of Tech-nology; 2000.

20 Kauraniemi P, Hautaniemi S, Autio R et al. Effects of Herceptin treat-ment on global gene expression patterns in HER2-amplified and nonam-plified breast cancer cell lines. Oncogene 2004;23:1010–1013.

21 Golub TR, Slonim DK, Tamayo P et al. Molecular classification ofcancer: class discovery and class prediction by gene expression moni-toring. Science 1999;286:531–537.

22 Briegel K, Lim KC, Plank C et al. Ectopic expression of a conditionalGATA-2/estrogen receptor chimera arrests erythroid differentiation in ahormone-dependent manner. Genes Dev 1993;7:1097–1109.

23 Stanton BR, Parada LF. The N-myc proto-oncogene: developmentalexpression and in vivo site-directed mutagenesis. Brain Pathol 1992;2:71–83.

24 Lucotti C, Malabarba L, Rosti V et al. Cell cycle distribution of cordblood-derived haematopoietic progenitor cells and their recruitment intothe S-phase of the cell cycle. Br J Haematol 2000;108:621–628.

25 Okuda T, Hirai H, Valentine VA et al. Molecular cloning, expressionpattern, and chromosomal localization of human CDKN2D/INK4d, aninhibitor of cyclin D-dependent kinases. Genomics 1995;29:623–630.

26 Hershko A, Ganoth D, Pehrson J et al. Methylated ubiquitin inhibitscyclin degradation in clam embryo extracts. J Biol Chem 1991;266:16376–16379.

27 Ivanova NB, Dimos JT, Schaniel C et al. A stem cell molecular signature.Science 2002;298:601–604.

28 Thorsteinsdottir U, Mamo A, Kroon E et al. Overexpression of themyeloid leukemia-associated Hoxa9 gene in bone marrow cells inducesstem cell expansion. Blood 2002;99:121–129.

29 Baldus CD, Tanner SM, Kusewitt DF et al. BAALC, a novel marker ofhuman hematopoietic progenitor cells. Exp Hematol 2003;31:1051–1056.

30 Tanner SM, Austin JL, Leone G et al. BAALC, the human member of anovel mammalian neuroectoderm gene lineage, is implicated in hemato-poiesis and acute leukemia. Proc Natl Acad Sci U S A 2001;98:13901–13906.

31 Liu X, Rapp N, Deans R et al. Molecular cloning and chromosomalmapping of a candidate cytokine gene selectively expressed in humanCD34� cells. Genomics 2000;65:283–292.

32 Wagner W, Ansorge A, Wirkner U et al. Molecular evidence for stemcell function of the slow-dividing fraction among human hematopoieticprogenitor cells by genome-wide analysis. Blood 2004;104:675–686.

33 Zhou G, Chen J, Lee S et al. The pattern of gene expression in humanCD34(�) stem/progenitor cells. Proc Natl Acad Sci U S A 2001;98:13966–13971.

34 Shojaei F, Gallacher L, Bhatia M. Differential gene expression of humanstem progenitor cells derived from early stages of in utero humanhematopoiesis. Blood 2004;103:2530–2540.

640 Gene Expression Profile of CD133� Cells

35 Akashi K, He X, Chen J et al. Transcriptional accessibility for genes ofmultiple tissues and hematopoietic lineages is hierarchically controlledduring early hematopoiesis. Blood 2003;101:383–389.

36 Terskikh AV, Miyamoto T, Chang C et al. Gene expression analysis ofpurified hematopoietic stem cells and committed progenitors. Blood2003;102:94–101.

37 He X, Gonzalez V, Tsang A et al. Differential gene expression profilingof CD34� CD133� umbilical cord blood hematopoietic stem progenitorcells. Stem Cells Dev 2005;14:188–198.

38 Rockett JC, Patrizio P, Schmid JE et al. Gene expression patternsassociated with infertility in humans and rodent models. Mutat Res2004;549:225–240.

39 Fargeas CA, Joester A, Missol-Kolka E et al. Identification of novelProminin-1/CD133 splice variants with alternative C-termini and theirexpression in epididymis and testis. J Cell Sci 2004;117:4301–4311.

40 Grskovic B, Ruzicka K, Karimi A et al. Cell cycle analysis of theCD133� and CD133� cells isolated from umbilical cord blood. ClinChim Acta 2004;343:173–178.

41 Summers YJ, Heyworth CM, de Wynter EA et al. AC133� G0 cells fromcord blood show a high incidence of long-term culture-initiating cells anda capacity for more than 100 million-fold amplification of colony-forming cells in vitro. STEM CELLS 2004;22:704–715.

42 Rocha V, Wagner JE Jr, Sobocinski KA et al. Graft-versus-host diseasein children who have received a cord-blood or bone marrow transplantfrom an HLA-identical sibling. Eurocord and International Bone MarrowTransplant Registry Working Committee on Alternative Donor and StemCell Sources. N Engl J Med 2000;342:1846–1854.

43 Peled A, Kollet O, Ponomaryov T et al. The chemokine SDF-1 activatesthe integrins LFA-1, VLA-4, and VLA-5 on immature human CD34(�)cells: role in transendothelial/stromal migration and engraftment ofNOD/SCID mice. Blood 2000;95:3289–3296.

44 Frassoni F, Podesta M, Maccario R et al. Cord blood transplantationprovides better reconstitution of hematopoietic reservoir compared withbone marrow transplantation. Blood 2003;102:1138–1141.

45 Abeyta MJ, Clark AT, Rodriguez RT et al. Unique gene expressionsignatures of independently-derived human embryonic stem cell lines.Hum Mol Genet 2004;13:601–608.

46 Ramalho-Santos M, Yoon S, Matsuzaki Y et al. “Stemness”: Transcrip-tional profiling of embryonic and adult stem cells. Science 2002;298:597–600.

47 Liu XR, Zhou RL, Zhang QY et al. Structure analysis and expressions ofa novel tetratransmembrane protein, lysosoma-associated protein trans-membrane 4 beta associated with hepatocellular carcinoma. World JGastroenterol 2004;10:1555–1559.

48 Shimamoto A, Kitao S, Ichikawa K et al. A unique human gene thatspans over 230 kb in the human chromosome 8p11–12 and codesmultiple family proteins sharing RNA-binding motifs. Proc Natl AcadSci U S A 1996;93:10913–10917.

49 Bhattacharya B, Miura T, Brandenberger R et al. Gene expression inhuman embryonic stem cell lines: Unique molecular signature. Blood2004;103:2956–2964.

50 Richards M, Tan SP, Tan JH et al. The transcriptome profile of humanembryonic stem cells as defined by SAGE. STEM CELLS 2004;22:51–64.

51 Sperger JM, Chen X, Draper JS et al. Gene expression patterns in humanembryonic stem cells and human pluripotent germ cell tumors. Proc NatlAcad Sci U S A 2003;100:13350–13355.

52 Sato N, Sanjuan IM, Heke M et al. Molecular signature of humanembryonic stem cells and its comparison with the mouse. Dev Biol2003;260:404–413.

53 Tanaka TS, Kunath T, Kimber WL et al. Gene expression profiling ofembryo-derived stem cells reveals candidate genes associated with plu-ripotency and lineage specificity. Genome Res 2002;12:1921–1928.

See www.StemCells.com for supplemental material available online.

641Jaatinen, Hemmoranta, Hautaniemi et al.

www.StemCells.com