Many members of the nearly 200-strong homeobox gene family have been implicated in cancer, mostly following ectopic expression. In this study we analyzed homeobox gene expression in Hodgkin lymphoma (HL) cell lines. Both reverse transcription–polymerase chain reaction (RT-PCR) using degenerate primers and microarray profiling identified consistently up-regulated HOXB9 expression. Analysis of HOXB9 regulation in HL cells revealed E2F3A and BMI1 as activator and repressor, respectively. Furthermore, a constitutively active ERK5 pathway was identified in all HL cell lines analyzed as well as primary HL cells. Our data show that ERK5 probably mediates HOXB9 expression by repressing BMI1. In addition, expression analysis of the neighboring microRNA gene mir-196a1 revealed coregulation with HOXB9. Functional analysis of HOXB9 by knockdown and overexpression assays indicated their influence on both proliferation and apoptosis in HL cells. In summary, we identified up-regulation of HOXB9 in HL mediated by constitutively active ERK5 signaling which may represent novel therapeutic targets in HL.


Homeobox genes code for transcription factors, affecting developmental processes during embryogenesis and in the adult.1 The conserved 180-bp homeobox facilitates both their classification and informs evolutionary hypotheses.2 Two classes of homeobox genes have been distinguished concerning their chromosomal localization: clustered homeobox genes (HOX), arranged in 4 clusters on different human chromosomes versus the nonclustered homeobox genes which are widely dispersed. Clustered arrangement reflects their regulation in embryonal develoment.3 Repressor proteins of the polycomb group and higher order structures of the chromatin influence these processes.4,5

In humans about 200 homeobox genes have been identified.6 Many of these are implicated in cancer because of ectopic expression,7 that is, TLX1, TLX3, and NKX2-5 which are closely related members of the NK-like homeobox subfamily involved in T-cell acute lymphoblastic leukemia (T-ALL), suggesting a fundamental role for subtype specificity of tumorigenesis.8

Hodgkin lymphoma (HL) is a lymphoid malignancy characterized by the presence of rare neoplastic Hodgkin/Reed-Sternberg (H/RS) cells within a heterogeneous background of nonneoplastic inflammatory and accessory cells. A few aberrantly activated pathways have been identified in HL cells, comprising NFκB, JAK/STAT, MAPK, and PI3K pathways.913 Four homeobox genes of the nonclustered type are hitherto implicated in HL. PAX5, POU2F2/OCT2, and POU2AF1/BOB1 regulate B-cell development and expression of immunoglobulins. Their expression levels are reduced in HL and involved in an incomplete B-cell phenotype.14,15 Ectopic expression of homeobox gene HLXB9 in HL cells promotes activation of IL6, a key factor for HL pathogenesis.13

In this study, we comprehensively analyzed homeobox gene expression in well-characterized HL cell lines and identified dysregulated expression of a clustered homeobox gene, HOXB9, and a constitutively active ERK5 pathway involved in its activation.

Materials and methods

Cell lines and treatments

Cell lines used in this study are all reposited with the DSMZ (Braunschweig, Germany). Details of origin, pathobiologic properties, and cell culture conditions are given by Drexler.16 Several reagents were used for cell stimulation: genistein, NF-κB activation inhibitor, and calphostin C were obtained from Calbiochem (Bad Soden, Germany); 12-O-tetradecanoyl-phorbol-13-acetate (TPA) was from Genzyme (Rösselsheim, Germany); trichostatin A (TSA) and AG490 were from Upstate (Hamburg, Germany); anti-FGF basic antibody and recombinant fibroblastic growth factor 2 (FGF2) were from R&D Systems (Minneapolis, MN); PD98059, SB202190, and wortmannin were from Sigma (Taufkirchen, Germany).

For knockdown experiments, phosphorothioate-modificated antisense oligonucleotides obtained from MWG (Martinsried, Germany) were applied as described previously.13,17 siRNA directed against HOXB9 and Lamin A/C was obtained from Qiagen (Hilden, Germany). Using an EPI 2500 (Fischer, Heidelberg, Germany), 1 μL siRNA solution of a 20-mM stock was electroporated into 1 × 106 cells. Subsequently, cells were transferred into 1 mL fresh medium and incubated for 24 hours. Down-regulation of both Lamin A/C and HOXB9 was confirmed by reverse transcription–polymerase chain reaction (RT-PCR) analysis (data not shown).

For HOXB9 expression PCR product comprising the complete coding region of HOXB9 (Table S1, available on the Blood website; see the Supplemental Materials link at the top of the online article) was subcloned into pGEM-Teasy (Invitrogen, Karlsruhe, Germany) and subsequently cloned via EcoRI restriction into expression vector pcDNA3 (Invitrogen) generating pcHOXB9. This construct was checked by sequence analysis, and 2 μg were transferred into the cells by electroporation.

For activation of CD30 pathway 10 μg agonistic anti-CD30 antibody, clone BerH8 (BD Pharmingen, San Diego, CA) was added to 1 × 106 cells suspended in 1 mL medium and incubated overnight.

Expression analysis

RNA extraction, cDNA synthesis, and RT-PCR were performed as described previously.17 cDNA for RT-PCR analysis of mir-196a1 was prepared using Quantitect-kit from Qiagen. Oligonucleotides designed to amplify mir-196a1 cDNA detect the precursor RNA (pri-miRNA). Oligonucleotide sequences are listed in Table S1. Quantitative RT-PCR (RQ-PCR) was performed as described previously.18 For expression profiling we used U133A Affymetrix DNA chips (Affymetrix, Santa Clara, CA), performed as described previously.19 Samples were normalized using Affymetrix GCOS software to a median expression level of 500 (in GCOS arbitrary units). All the data were log2-transformed for display and analysis. Absent signals were floored to a common baseline (50 units). From the selected probe set a t test was performed. Figure 1 A shows the highest ranked probe sets (P < .1) analyzed using CLUSTER and TREEVIEW software (Michael Eisen, Berkeley, CA).

For detection of FGF2 protein in the supernatant of cell cultures, we used an enzyme-linked immunoabsorbent assay (ELISA) kit human basic FGF obtained from R&D Systems. Thus, 1 × 106 PBS-washed cells were seeded out in 1 mL medium and incubated overnight. On the next day an aliquot of supernatant was frozen until ELISA detection.


Fluorescence in situ hybridization (FISH) analysis was performed as described previously.20 BACs used as probes were RP11-456D7 and RP11-501C14, obtained from the Sanger Centre (Cambridge, England).

Construction of BMI1 siRNA

Six DNA oligonucleotides corresponding to positions 14 to 32, 209 to 227, 244 to 262, 423 to 441, 465 to 483, and 620 to 638 of the human BMI1 gene sequence (Gene bank accession no. L13689) were subjected to BLAST homology search and thereafter chemically synthesized, including 5′-BglII and a 3′-SalI restriction site cloning overhangs (BioSpring, Frankfurt, Germany). The numbering of the first nucleotide of the small hairpin RNAs (shRNAs) refers to the ATG start codon. The oligonucleotide sequences of the most effective shRNAs used in this study were as follows: FP465bmi-1, 5′-GATCCCCGGAGGAGGTGAATGATAAATTCAAGAGATTTATCATTCACCTCCTCCTTTTTTGGAAG-3′; RP465bmi-1, 5′-TCGACTTCCAAAAAAGGAGGAGGTGAATGATAAATCTCTTGAATTTATCATTCACCTCCTCCGGG-3′; 3′-FP620bmi-1, 5′-GATCCCCTGGACATTGCCTACATTTATTCAAGAGATAAATGTAGGCAATGTCCATTTTTTGGAAG-3′; RP620bmi-1, 5′-TCGACTTCCAAAAAATGGACATTGCCTACATTTATCTCTTGAATAAATGTAGGCAATGTCCAGGG-3′. The anti-BMI1 shRNA 620 has a point mutation (G-U wobble base pair) to the human BMI1 mRNA at position 624 without functional consequence.

The oligonucleotides share 19 nt identical with the target gene (italics). The noncomplementary 9-nt loop sequences are underlined, and each sense oligonucleotide harbors a stretch of T as a pol III transcription termination signal. The oligonucleotides were annealed and inserted 3′ of the H1-RNA promoter into the BglII/SalI-digested pBlueScript-derived pH1-plasmid to generate pH1-BMI1-465 and pH1-BMI1-620 as described previously.21 The isolated clones were verified by DNA sequencing. The plasmid pH1-GL4 (control) has been described earlier.21

Construction of lentiviral vectors

pdc-SR (pdc indicates plasmids resulting in double-copy proviruses) were used to generate lentiviral transgenic plasmids containing H1-siRNA expression cassettes located in the U3 region of the Δ3′-LTR.21 To generate the lentiviral plasmids pdcH1-bmi1-465-SR and pdcH1-bmi1-620-SR, the plasmids pH1-bmi1-465 and pH1-bmi-620 were digested with SmaI and HincII, and the resulting DNA fragments (360 nt) were blunt-end ligated into the SnaB I site of the pdc-SR. The lentiviral plasmids encode RFPEXPRESS as reporter gene.

Preparation of recombinant lentiviral supernatants and lentiviral transduction

VSV.G-pseudotyped lentiviral particles were generated by calcium phosphate cotransfection of 293T cells, and viral supernatants were concentrated by low-speed centrifugation. dcH1-shRNA-SR lentiviral preparations were titered in triplicate by serial dilutions of the concentrated vector stocks on 1 × 105 K-562 cells in 24-well plates. The number of RFP-positive cells was analyzed 72 hours after transduction by flow cytometry analysis (FACS-Calibur; Becton Dickinson, Heidelberg, Germany), and the titers were averaged and typically ranged between 1 and 5 × 108 IU/mL. Lentiviral supernatants were used to transduce HDLM-2 and KM-H2 cells with an MOI between 2 and 4 as described previously.22

Cell-cycle analysis

Cell-cycle analysis was performed by flow cytometry as described previously.13

Immunostaining of cell lines and primary lymph nodes

Cultured cells were centrifuged onto microscope slides and fixed with methanol for 90 seconds. Primary lymph nodes derived from 12 patients with HL (classical HL of the mixed cellularity type) were selected from the archive of the Institute of Pathology, University of Wörzburg, Germany. The study was approved by the local ethics committee of the University of Wörzburg. Paraffin-embedded cuts (5 μm) of primary lymph nodes were incubated in xylol for 10 minutes, dehydrated in an alcohol series, and cooked in citrate buffer (10 mM, pH 5.5) for 3 × 5 minutes. The following steps were identical for both cultured and primary cells. The slides were blocked with 5% BSA dissolved in PBS for 30 minutes, incubated with the first antibody dissolved in 5% BSA/PBS for 30 minutes, washed for 3 × 5 minutes in PBS, incubated with the second antibody dissolved in 5% BSA/PBS for 30 minutes, washed for 3 × 5 minutes in PBS, dehydrated in an alcohol series, air dried, and covered with DAPI/Vectashield (Vector Laboratories, Burlingame, CA). We used polyclonal anti–phospho-ERK5 and fluorescein isothiocyanate (FITC)–labeled antirabbit antibodies, both obtained from Santa Cruz Biotechnology (Heidelberg, Germany). Images were visualized using a Zeiss Axioscope 2 microscope equipped with a Neofluar 100×/1.3 numerical aperture oil objective (Zeiss, Göttingen, Germany). Images were acquired and processed using Smart Capture 2 software (Applied Imaging, Newcastle, United Kingdom).


Analysis of homeobox gene expression in HL cell lines

Expression of homeobox genes was analyzed by 2 strategies, RT-PCR and profiling. First, we performed RT-PCR using degenerate primers to identify the strongly expressed homeobox genes in HL cell lines (Table S1). Primers were designed to amplify the conserved homeobox of most of the clustered and many of the nonclustered homeobox genes. To validate this strategy, we analyzed 4 clones obtained from a T-ALL cell line (PEER) which contains t(5;14)(q35;q32) affecting ectopic NKX2-5activation.8 Sequence analysis by BLAST identified 3 of 4 NKX2-5clones, indicating overexpression of this homeobox gene (Table S2). Thirty-three clones of RT-PCR products obtained from HL cell lines HDLM-2, KM-H2, L-1236, L-428, and L-540 were analyzed in the same way. Clone frequencies highlighted HOXB9 as the most prominently expressed homeobox gene in HL cell lines, with HDLM-2 and L-540 expressing the highest levels (Table S2).

Second, for a comprehensive analysis of homeobox gene expression in 6 well-characterized HL cell lines (HDLM-2, KM-H2, L-1236, L-428, L-540, SUP-HD1), we performed gene expression profiling using Affymetrix DNA chips (U133A). As controls we analyzed 3 non-HL (NHL) cell lines. SC-1 is derived from a follicular lymphoma (FL) and both RI-1 and RC-K8 are from diffuse large B-cell lymphomas (DLCLs). All data sets for homeobox genes are based on Ensembl homeobox domain view (, comprising 188 genes analyzed using CLUSTER and TREEVIEW software. To identify homeobox genes specific for HL we looked for those genes predominantly expressed in either HL or NHL cell lines.

Data showed HOXC6, SHOX2, HOXB8, HOXB9, HOXB4, SATB1, and HOP to be significantly overexpressed in HL cell lines, and POU2F2/OCT2 and PAX5 in NHL cell lines (Figure 1A). By RT-PCR analysis high expression was confirmed for HOXB9, HOXB4, and HOP in HL cells, and for PAX5 and POU2F2/OCT2 in NHL cells (Figure 1B). For HOXC6, SHOX2, HOXB8, and SATB1 no preferential expression was detected in HL cell lines (data not shown). In HL cells, underexpression of both PAX5 and POU2F2/OCT2 has been described previously, further validating our screening strategy.14,15

Figure 1

Gene expression analysis. (A) Expression profiling data were converted into a heatmap using CLUSTER and TREEVIEW software (red indicates high; green, low; black, medium expression level). Data show homeobox genes differentially expressed in HL and NHL cell lines. Genes showing significant up-regulation/down-regulation in HL (HOXC6, SHOX2, HOXB8, HOXB9, HOXB4, SATB1, HOP/PAX5, POU2F2), shown in red/green, respectively, were subsequently analyzed by RT-PCR, depicted in panel B together with HLXB9, genes located within the HOXB cluster (HOXB13, PRAC, PRAC2, mir-196a1) and involved in HOXB9 transcription (FGF2). Expression of upstream binding factor (UBF) was used as positive control for cDNA quality. NTC indicates no template control.

The HLXB9 homeobox gene, previously described in HL,13 was not exclusively expressed in HL cell lines in our data set, prompting our survey of homeobox gene specificity in different hematopoietic tumors. Hence, we analyzed expression of HOP, HOXB9, HOXB4, and additionally HLXB9 by RT-PCR in a comprehensive panel of 84 hematopoietic cell lines (Table 1). Expression of HOP was detected in 12 (17%) of 69, HLXB9 in 21 (29%) of 72, HOXB9 in 26 (32%) of 81, and HOXB4 in 45 (63%) of 72 cell lines. This analysis indicated that no homeobox gene tested was absolutely HL specific. Although HOP exhibited the greatest specificity, its expression was detected in 4 (67%) of 6 HL cell lines only (Figure 1B). Qualitative analysis of HOXB4 expression indicated no specificity for any hematopoietic malignancy. However, consideration of quantitative aspects of PCR product intensities (Table 1) may indicate low HOXB4 expression levels in T-cell lines in contrast to B-cell and myeloid cell lines, as previously described in primary cells.23 Both HOXB9 and HLXB9 were expressed in about 30% of hematopoietic cell lines and displayed overlapping profiles. Whereas HOXB9 expression is centered on HL, anaplastic large cell lymphoma (ALCL) and mediastinal large B-cell lymphoma (MLBCL), HLXB9 expression was additionally detected in DLBCL/FL cell lines (Table 1). Interestingly, expression patterns of these 2 homeobox genes may reflect overlapping biologic features of HL, ALCL, and MLBCL.24,25 Although no single homeobox gene displayed strict HL specificity, their combined expression patterns may indicate a unique homeobox signature for HL.

Table 1

Gene expression analysis in cell lines by RT-PCR

HOXB9 expression has been described during embryonal neural tube development26 but hitherto not in hematopoietic cells or tissues indicating dysregulated expression in these cell lines as described herein. Given the prominent status of HOXB9, combining high expression with a restricted profile, we focused our work on both regulation and potential function of this homeobox gene in HL cell lines.

Regulation of HOXB9 expression in HL cell lines

Chromosomal aberrations combined with ectopic oncogene activation are common abnormalities in both leukemia and lymphoma. To check whether HOXB9 expression in HL cell lines is coupled with chromosomal mutations we performed FISH analysis in HDLM-2, KM-H2, L-1236, L-428, and L-540 cells. All analyzed cell lines showed wild-type configuration of the HOXB locus at 17q21 (data not shown), discounting any chromosomal basis for dysregulated HOXB9 expression.

Transcription factor binding site analysis of the HOXB9 gene using data from UCSC genome bioinformatics ( indicated potential sites for NFκB, MEF2, and E2F within the promoter region which may contribute to dysregulated expression. NFκB is constitutively active in both primary HL cells and cell lines.27 However, pharmacologic inhibition of NFκB activity showed no effect on HOXB9 expression in HDLM-2, KM-H2 (Figure 2A), and L-428 cells (data not shown). The pre–B-ALL cell line TS-2 expresses an active MEF2D fusion protein resulting from t(1;19)(q23;p13) but lacks HOXB9 mRNA (Table 1).28 Furthermore, array data show that neither described target genes NUR77 nor LKLF29,30 were expressed in HL cell lines, indicating that MEF2 is not active and, therefore, not involved in HOXB9 activation. In contrast, KM-H2 and SUP-HD1 cells treated with antisense oligonucleotide directed against E2F3A showed reduced expression levels of HOXB9 (Figure 2A), indicating an activating input for this factor.

Figure 2

RT-PCR analysis of HOXB9 after stimulation of HL cell lines. (A) HDLM-2 and KM-H2 were treated with 200 μM genistein and 14 mM NFκB inhibitor for 12 hours. In contrast to HDLM-2, genistein treatment reduced HOXB9 expression in KM-H2 cells, whereas NFκB inhibitor showed no effect. Antisense oligonucleotide (AS) treatment directed against E2F3A in SUP-HD1 and KM-H2 cells affected HOXB9 reduction relative to control oligonucleotide, indicating an activatory influence for this protein. (B) HDLM-2 and L-428 cells were treated with the following inhibitors of signaling pathways for 12 hours: 1 μM wortmannin (PI3K), 100 μM AG490 (JAK/STAT), 50 μM PD98059, and 100 μM SB202190 (ERK1/2 MAPK). HOXB9 showed reduction in HDLM-2 cells only, following MAPK pathway inhibition, indicating stimulation by this pathway. (C) KM-H2 cells were treated with 10 μg/mL inhibitory antibody directed against FGF2. HDLM-2 and L-540 cells were stimulated with 10 ng/mL recombinant FGF2 protein. HOXB9 expression decreased in KM-H2 and increased in HDLM-2 and L-540 cells. Data indicate an activatory influence of FGF2 on HOXB9 expression. (D) Influence of PKC on HOXB9 expression was analyzed by treating HDLM-2 and KM-H2 cells with a PKC activator TPA (100 nM) or PKC inhibitor calphostin C. Data show the absence of PKC activity in both cell lines. PKC activation reduced HOXB9 expression in KM-H2 but not in HDLM-2. NTC indicates no template control. For cDNA control the expression of ets variant gene 6 (TEL) or UBF was analyzed.

Repressor proteins of the polycomb group have been shown to act as key HOX gene regulators.31 Two polycomb group-related complexes (PRC1 and PRC2) have been isolated, consisting of several distinct proteins. In contrast to PRC1, PRC2 recruits histone deacetylase (HDAC) mediating its repressive action.32 We have treated a HOXB9-positive HL cell line (HDLM-2) and a HOXB9-negative control cell line (PEER) with HDAC inhibitor TSA and subsequently analyzed HOXB9 expression by RT-PCR. No changes in expression levels were observed (data not shown), indicating the absence of regulation by PRC2. BMI1 is the most prominent component of PRC1.31 Using lentiviral-mediated transfer of RNA(i)nterference construct directed against BMI1, we were able to knockdown BMI1 expression in both HDLM-2 and KM-H2 cells as analyzed by RQ-PCR. HOXB9 expression was concomitantly strengthened, confirming a repressive mode of action for BMI1/PRC1 (Figure 3).

Figure 3

Quantitative RT-PCR analysis after knockdown of BMI1. HDLM-2 and KM-H2 cells were lentivirally transfected with RNAi construct directed against BMI1. RQ-PCR analysis indicated an about 2-fold reduction in BMI1 expression in comparison to GAPDH. Correspondingly, expression of both HOXB9 and HOXB13 increased about 9-fold and 3-fold in HDLM-2 and KM-H2, respectively. Experiments were performed twice with similar results; error bars indicate SD.

Recently, BMI1 was described as a downstream target of mitogen-activated protein kinase (MAPK) pathways abrogating repression.33 To check for such an influence on HOXB9 expression, we analyzed MAPK ERK1/2 in addition to PI3K and JAK/STAT pathways by pharmacologic inhibition in HDLM-2, L-1236, and L-428 cells. In contrast to inhibition of PI3K and JAK/STAT, inhibition of ERK1/2 reduced HOXB9 expression in HDLM-2 cells but not in L-428 (Figure 2B) or in L-1236 cells (data not shown). Combination of 2 inhibitors for MAPK ERK1/2 abolished HOXB9 expression in HDLM-2 but showed no effect in L-428 cells (Figure 2B). Consistently, ERK1/2 MAPK pathway has been shown to be highly activated in HDLM-2 in comparison to other HL cell lines.11

Specific pharmacologic inhibition of MAPK ERK5 pathway is currently not possible. ERK5 gets activated by phosphorylation, resulting in translocation into the nucleus.34 Therefore, the activity of this pathway was analyzed by immunofluorescence using phospho-ERK5 antibody. In contrast to SC-1 cells, all HL cell lines showed activated phospho-ERK5 in their nuclei (Figure 4A), suggesting a possible influence of this pathway on HOXB9 expression.

Figure 4

Immunostaining of phospho-ERK5 in HL cell lines and lymph nodes. (A) Cytospins of HL cell lines (HDLM-2, L-428, L-540) and of NHL cell line (SC-1) were stained by immunofluorescence with phospho-ERK5 antibody (magnification, × 630). In contrast to SC-1, HL cells show intensive nuclear staining. DAPI counterstaining indicates the nuclei. (B) The HL cell line KM-H2 was treated with 10 μg/mL anti-FGF2 antibody or with 200 μM tyrosine kinase inhibitor genistein and subsequently stained with anti–phospho-ERK5 (magnification, × 630). In contrast to control cells, treated cells show weak staining of nuclei indicating that (1) FGF2 signaling is mediated by ERK5 and (2) genistein inhibits this pathway. (C) Lymph nodes obtained from 3 patients with HL (HL1, HL8, HL10) stained with anti–phospho-ERK5 (magnification, × 630) show positive giant cells, indicating a constitutively active ERK5 pathway in H/RS cells.

FGF2, also implicated in HL pathogenesis,35,36 has been described to activate the ERK5 pathway.37 To investigate a potential autocrine activation mechanism we first analyzed both FGF2 mRNA and protein expression in HL cell lines by RT-PCR and ELISA, respectively. All HL cell lines in addition to the NHL cell line RC-K8 were positive at the mRNA level (Figure 1B), but high amounts of FGF2 protein were detected in the supernatants of KM-H2 and RC-K8 cells only (Table 2). The effect of FGF2 was investigated in KM-H2 cells treated with an antibody inhibitory against FGF2 protein and with genistein, an inhibitor of tyrosine kinase known to inhibit growth factor receptor signaling.38 Subsequent analysis of these treatments after 12 hours showed reduction of both phospho-ERK5 signal in the nucleus (Figure 4B) and HOXB9 mRNA expression (Figure 2A,C), confirming the signaling of FGF2 by ERK5 to HOXB9. Stimulation of HDLM-2 and L-540 cells with FGF2 increased HOXB9 expression, demonstrating the potential to react on FGF2 stimulation in other HL cell lines (Figure 2C). FGF2 signaling was inhibited by protein kinase C (PKC) activation using TPA.39 To analyze the regulation of HOXB9 expression by PKC, HDLM-2, KM-H2, and L-540 cells were treated with TPA or calphostin C to activate or inhibit this kinase, respectively. Although TPA affected down-regulation of HOXB9 expression in KM-H2, calphostin C showed no effect in any cell line (Figure 2D; data for L-540 not shown), indicating that PKC is inactive but inhibits FGF2 signaling if activated.

Table 2

ELISA analysis of FGF2 protein

To analyze whether the ERK5 pathway is also active in primary HL cells, we performed immunostaining of lymph nodes obtained from patients with HL. Accordingly, 8 of 12 patients clearly showed phospho-ERK5 signals in giant cells in contrast to surrounding lymphocytes, indicating constitutively active ERK5 pathway in H/RS cells (Figure 4C).

Expression analysis of the posterior HOXB cluster

Increased expression of 3 additional genes located within the posterior part of the HOXB cluster (HOXB13, PRAC, PRAC2) was indicated by array data (not shown) and confirmed by RT-PCR in HL cell lines KM-H2, L-428, and SUP-HD1 (Figure 1B). The presence of active neighboring genes may indicate regulation by a chromatin-dependent mechanism.5 To check whether BMI1 is also involved in regulation of HOXB13, expression levels were analyzed after RNAi-mediated down-regulation of BMI1 in both HDLM-2 and KM-H2 cells. Data clearly showed HOXB13 up-regulation, comparable to that of HOXB9(Figure 3), supporting the view of a regionally acting regulation mechanism.

The microRNA gene mir-196a1 is located directly upstream of HOXB9 within the HOXB cluster. RT-PCR analysis of mir-196a1 confirmed expression in 6 HL cell lines in contrast to controls (Figure 1B; Table 1). Extended RT-PCR analysis of mir-196a1 was performed in 27 hematopoietic cell lines, 11 of which were positive and 16 negative. In 25 (93%) of these cell lines expression of mir-196a1 corresponded to that of HOXB9, strongly supporting the existence of a mechanism for their coregulation (Table 1).

Functional analysis of HOXB9

Both BTG1/2 and Geminin interact physically with HOXB9 protein and are involved in regulating proliferation.40,41 Additionally, BTG1 and BTG2 were shown to regulate apoptosis.42 Therefore, we analyzed the potential effect of HOXB9 expression on cell growth by treating HL cell lines HDLM-2 and KM-H2, and NHL cell line SC-1 (not expressing HOXB9) with antisense oligonucleotide directed against HOXB9. In comparison to nonspecific control oligonucleotide the number of cells decreased in both HL cell lines after 3 days, in contrast to SC-1 cells in which no differences were observed (Figure 5A). The same effect of growth reduction was visible using siRNA-mediated knockdown of HOXB9 in KM-H2 (Figure 5B). However, subsequent analysis of treated cell lines by flow cytometry revealed no specific cell-cycle arrest (data not shown). Furthermore, the effect of cell number reduction was reinforced in the absence of fetal bovine serum (FBS), probably indicating the antiapoptotic function of HOXB9 (Figure 5B). Overexpression of HOXB9 in KM-H2 and HELA (HOXB9 negative) resulted in increased cell number in the absence of FBS (Figure 5C), supporting an antiapoptotic mechanism. However, KM-H2 cells showed an increase in cell number in the presence of FBS, too, indicating both proliferative effects in HL cells and cell-type–specific differences.

Figure 5

Functional analysis of HOXB9 expression in HL cell lines. (A) HL cells (HDLM-2 and KM-H2) and NHL cells (SC-1) were treated with antisense oligonucleotide (AS) directed against HOXB9. In contrast to SC-1, HL cell lines showed reduced growth after 3 days of incubation with AS-HOXB9. (B) KM-H2 cells were electroporated with siRNA directed against HOXB9 or control siRNA and were subsequently incubated in the presence/absence of fetal bovine serum (FBS) or with TPA. Cell counts indicate 60% reduction of viability in the absence of FBS and about 25% in the other settings. Cell counts are related to control siRNA treatment in the presence of FBS. (C) In KM-H2 and HELA cells HOXB9 was overexpressed using expression construct pcHOXB9 in the presence or absence of FBS, respectively. Cell counts of treatments with control vector pcDNA3 in the presence of FBS were set at 100%. In the absence of FBS, both cell lines showed enhanced growth/survival following HOXB9 overexpression. In the presence of FBS only KM-H2 cells showed increased cell counts, indicating cell-type–specific differences. All experiments were performed twice with similar results; error bars indicate SD.


Analysis of homeobox gene transcription in HL cells by both expression profiling and RT-PCR identified high levels of HOXB9. Hitherto, physiologic expression of HOXB9 has been only reported in posterior neural tube cells during embryogenesis,26,4345 highlighting the dysregulated expression in HL cells addressed in this study. Although none of the remaining candidate homeobox genes considered, including HOXB9, HOP, HOXB4, and HLXB9, showed strict HL specificity, the combined expression patterns of HOXB9, HOP, and HLXB9 may represent a unique homeobox gene signature for HL. Furthermore, expression patterns of HOXB9 and HLXB9 reflected biologic features of HL overlapping other lymphoma subtypes (eg, ALCL and MLBCL).24,25 This indicates a potentially significant role of homeobox genes in the pathogenesis of lymphoma subtypes. HOP expression has been detected in the heart and neural tube of mouse embryos and in the heart, lung, and liver of adult mice, indicating the ectopic nature of its expression in positive hematopoietic cell lines.46,47 The role of HOP in HL cell lines is currently undergoing investigation in our laboratory. HOXB4 is a focus of research because of its role in the expansion of hematopoietic stem cells accounting for its wide expression in hematopoietic cell lines.23,48 Our approach confirmed down-regulation of PAX5 and POU2F2/OCT2 in HL cells. Both genes are involved in B-cell differentiation and their absence may contribute to the attenuated B-cell phenotype of HL cells.14,15

Chromosomal abnormalities were excluded for HOXB9 activation by FISH analysis. However, we identified activator E2F3 and repressor BMI1/PRC1-mediating expression of HOXB9 in HL cell lines. During embryogenesis in Xenopus, E2F3 has been identified as an activator of HOXB9 expression within the neural tube.49 Interestingly, repressive E2F factors such as E2F6 interact with PRC1 and may function as a sequence-specific anchor for the complex,50 and knockout mice for BMI1 and E2F6, respectively, showed similar skeletal transformations confirming their functional interaction.51 Although E2F6 and PRC1 protein RING1 are coexpressed in HL cells, no colocalization has been detected.52 However, strong expression of BMI1 has been detected in H/RS cells.53,54 Taken together, these data suggest dysregulated crosstalk between activating and repressive E2F factors, recruiting BMI1/PRC1, and regulating HOXB9 expression in HL cell lines.

Regulation of HOX genes by polycomb repressor proteins is well known,4 although differences between particular HOX genes have been observed. Knockdown of BMI1 in mouse embryonic fibroblasts showed no effect on HOXB9 expression in contrast to many other HOX genes, either activated or repressed.55 However, whether the influence of PRC1 on HOX genes is altered in cancer cells remains unknown. Recently, MAPK pathways have been shown to regulate both activity and intranuclear distribution of BMI1 by phosphorylation.33 Accordingly, we identified the influence of MAPK pathways involving ERK1/2 and ERK5 on HOXB9 expression in HL cell lines. Constitutively active ERK5 was identified in all HL cell lines analyzed in addition to primary H/RS cells. ERK5 can now be added to the list of constitutively active signaling pathways in HL cells comprising ERK1/2, JAK/STAT, and PI3K.1013 Many activators for ERK5 pathway have been described, including FGF2,37,56 for which we confirmed autocrine activation of both ERK5 and HOXB9 expression in KM-H2 cells. FGF2 is a pathologic factor for HL.35,36 Interestingly, both FGF226,4345 and ERK557 mediate HOXB9 expression and neuronal differentiation, respectively, in the neural tube, indicating a similar regulation mechanism aberrantly activated in HL cells. Alternatively, other growth factor receptors shown to be activated in HL may contribute to ERK5 activation.58 In multiple myeloma, IL6 was found to mediate ERK5 activation.59 Additionally, ERK5 together with ERK2 has been shown to regulate NFκB,60 an important pathologic factor in HL.61 Taken together, we speculate that constitutively active MAPK pathways, including the ERK5 pathway identified in the present study, mediate HOXB9 expression by inactivating polycomb repressor protein BMI1.

Expression of HOXB9 in cell lines originated from HL, ALCL, and MLBCL may indicate a connection to the CD30 pathway present in these lymphoma subtypes.62 We confirmed active ERK5 pathway in 2 ALCL cell lines as tested by immunostaining (data not shown), but activation of CD30 by agonistic antibody treatment did not change HOXB9 expression either in HL or in ALCL cell lines tested (data not shown).

The posterior HOXB cluster comprises HOXB8, HOXB9, mir-196a1, PRAC, PRAC1, and HOXB13. Expression of HOXB13 is regulated by BMI1 in HL cells just like HOXB9. However, absence of further specific activators may be responsible for a low HOXB13 expression level. Interestingly, expression of HOXB9-neighboring mir-196a1 parallels that of HOXB9 strongly indicating combined regulation. However, in mouse embryogenesis posterior expression boundaries of both HOXB9 and mir-196a1 are not identical, in contrast to those of neighboring genes HOXB4 and mir-10a. These data indicated coregulation of HOXB4 and mir-10a,63 unlike HOXB9 and mir-196a1, contrasting with the picture in HL cells. HOXB8 was identified as a downstream target of mir-196a1 regulated at the transcriptional level.63,64 Therefore, the lack of HOXB8 expression in HL cell lines (data not shown) may be connected with the high-expression levels of mir-196a1. Dysregulated expression of microRNA genes play important roles in cancer.65 Additional hitherto unknown targets of mir-196a1 may also contribute to the pathogenesis of HL. For the HOXB cluster a looping-out model has been described explaining the regulation of clustered HOXB genes during embryogenesis.5 Taken together, dysregulated chromatin, including polycomb group proteins, and higher order structures may be responsible for aberrant expression of genes located in the posterior part of the HOXB cluster in HL cells.

Functional analysis of HOXB9 expression by knockdown and overexpression experiments indicates a role for both proliferation and antiapoptosis in HL cell lines. HOXB9 was described to interact with BTG1/2 and Geminin.40,41 However, if growth and survival are mediated by these proteins awaits further investigation.

In summary, we identified aberrant HOXB9 expression and constitutively active ERK5 pathway in HL cells, both of which may represent potential targets for therapeutic intervention in lymphoma.


Contribution: S.N. designed the research, performed RT-PCR, and wrote the paper; C.B. performed expression profiling analysis; L.V. cloned the RNAi constructs against BMI1; M.S. designed RNAi constructs and performed real-time PCR analysis; H.Q. performed cell-cycle analysis; C.M. performed cell-culture work, RT-PCR analysis, and immunostaining; A.R. supported expression profiling analysis and contributed patient material; H.G.D. contributed cell lines and supported labwork; and R.A.F.M. performed cytogenetic analysis and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Stefan Nagel, Human and Animal Cell Cultures, DSMZ, Inhoffenstr. 7B, 38124 Braunschweig, Germany; e-mail: sna{at}


This work was supported by the José Carreras Leukemia Foundation, Germany (SP 04/02).


  • The online version of this article contains a data supplement.

  • The publication costs of this article were defrayed in part by page charge payment. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

  • Submitted August 30, 2006.
  • Accepted November 18, 2006.


View Abstract