Expression of Dual-Specificity Phosphatase 5 Pseudogene 1 (DUSP5P1) in Tumor Cells

Sequencing of individual clones from a newly established cDNA library from the chemoresistant Hodgkin's lymphoma cell line L-1236 led to the isolation of a cDNA clone corresponding to a short sequence from chromosome 1. Reverse transcriptase-polymerase chain reaction indicated high expression of this sequence in Hodgkin's lymphoma derived cell lines but not in normal blood cells. Further characterization of this sequence and the surrounding genomic DNA revealed that this sequence is part of a human endogenous retrovirus locus. The sequence of this endogenous retrovirus is interrupted by a pseudogene of the dual specificity phosphatase 5 (DUSP5). Reverse transcriptase-polymerase chain reaction revealed high expression of this pseudogene (DUSP5P1) in HL cell lines but not in normal blood cells or Epstein-Barr virus-immortalized B cells. Cells from other tumor types (Burkitt's lymphoma, leukemia, neuroblastoma, Ewing sarcoma) also showed a higher DUSP5P1/DUSP5 ratio than normal cells. Furthermore, we observed that higher expression of DUSP5 in relation to DUSP5P1 correlated with the expression of the pro-apoptotic factor B cell leukemia/lymphoma 2-like 11 (BCL2L11) in peripheral blood cells and HL cells. Knock-down of DUSP5 in HL cells resulted in down-regulation of BCL2L11. Thus, the DUSP5/DUSP5P1 system could be responsible for regulation of BCL2L11 leading to inhibition of apoptosis in these tumor cells.


Introduction
The prognosis for patients with Hodgkin's lymphoma (HL) has been improved constantly during the last decades [1]. Nevertheless, some patients cannot be cured with currently established therapy protocols. On the other hand, toxicity of applied chemoand radiotherapy is high and this treatment is associated with a risk of late complications including secondary malignancies, infertility, breastfeeding problems or cardiovascular diseases [2][3][4][5]. Therefore, the recognition of new prognostic factors for patients with HL is desirable [6]. The characterization of factors associated with resistant disease may lead to the identification of alternative therapeutic strategies. Vice versa, the identification of such factors might also allow the reduction of toxic treatment elements for patients with favorable disease characteristics.
Using DNA microarray analysis, we characterized the gene expression profile of HL cell lines with different sensitivity for cytotoxic drugs and identified transcripts which are present only in resistant cells [7]. DNA microarrays allow the simultaneous detection of the complete transcriptome of tumor cells and control samples in a single experiment [8]. However, this method did not allow the screening for unknown transcripts and did not allow the direct functional analysis of tumor relevant transcripts. One elegant method for the identification of cDNAs corresponding to transcripts of interest is based on the transfection of cDNA libraries into eukaryotic cells and functional selection of transfected cell [9]. For the functional screening of cDNAs encoding HL derived transcripts with involvement in disease pathology we established cDNA libraries from different established HL cell lines. During initial characterization of such a cDNA library from HLcell line L-1236, we detected unusual short transcripts with homology to endogenous retroviruses. A large number of human endogenous retroviruses (ERV) have been identified [10] and activation of these ERV has been implicated in autoimmunity, neurodegenerative diseases and cancer [11][12][13]. ERV-related sequences form approximately 10% of the human genome. ERV loci are derived from exogenous retroviruses that were stably integrated during evolution in the host genome and are transmitted via the germ line. The majority of ERV have been inactivated by deletions, insertions, or other mutations. Only few ERV contain intact open reading frames for one or more viral genes. Recently, activation of ERV sequences in HL has been observed as a mechanism for aberrant gene expression [14]. Here, we describe the expression of another ERV-associated locus in HL cells. Our observation adds further evidence for the activation of endogenous retroviruses in these tumor cells. Expression of the newly identified ERV in HL cells is closely linked to expression of a pseudogene of the dual specificity phosphatase 5 (DUSP5). DUSP5 is involved in negative regulation of the extracellular signal-regulated kinase (ERK) pathway [15]. Interference of transcripts from the DUSP5 pseudogene locus with DUSP5 activity might influence this important signaling pathway.

Ethics Statement
Peripheral blood mononuclear cells (PBMC) from anonymous healthy donors and patients with HL were isolated with written informed consent and approval by the ethics committee of the Medical Faculty of the Martin Luther University Halle-Wittenberg.
Generation of a cDNA library from L-1236 cells Total RNA was isolated from cell line L-1236 by using Trizol reagent (Invitrogen, Karlsruhe, Germany) following the manufacturer's protocol. Thereafter, mRNA was enriched starting from 650 mg of total RNA by using the mMACS mRNA isolation kit (Miltenyi, Bergisch-Gladbach, Germany). The cDNA library was generated by using the pCMV-Script XR cDNA library construction kit (Agilent Technologies, La Jolla, CA, USA) with slight modifications. In short, 4.5 mg mRNA was subjected to first strand cDNA synthesis in a total volume of 45 ml for 1 hour at 42uC. After addition of 20 ml 106 second strand buffer, 6 ml second strand dNTP mixture, 116 ml sterile water, 2 ml RNase H (1.5 U/ml), and 11 ml DNA polymerase I (9 U/ml), second strand synthesis was performed at 16uC for 2.5 hours. Blunt ending and adapter ligation was performed according to manufacturer's instructions. After digestion with EcoR1, phosphorylation of EcoR1 ends and digestion with XhoI, cDNA was fractionated by sepharose CL-2B gel filtration. Fractions with high cDNA concentration and the expected cDNA size were ligated into the vector pCMV-Script. After ligation, vectors were transformed into XL10-Gold ultracompetent cells and the primary library was amplified at 30uC. After plating on agarose plates, random clones were analyzed by restriction digest with NotI.

Sequence analysis
Sequencing of pCMV-Script vector inserts was performed as described [22] by using the following pCMV-Script specific primers: 59-aat taa ccc tca cta aag gg-39 and 59-taa tac gac tca cta tag gg-39. The sequence of the ERVK_1q42.13 insert was submitted to GenBank (accession number JZ534323). Sequence analysis was performed with BLAST [24] and RetroSearch [25]. For multiple sequence alignments, CLUSTALW [26] was used. In silico promoter analysis and protein structure prediction was performed with NNPP2.2 [27] and SWISS-MODEL [28], respectively. For secondary structure prediction the structure of dual-specificity phosphatese 6 (DUSP6) [29] was used as template.

Knock-down of DUSP5 and microarray analysis
Vector-based knock-down of DUSP5 in L428-cells was performed by using the BLOCK-iT POL II miR RNAi expression vector kit with EmGFP (Invitrogen, Karlsruhe, Germany) according to manufacturer's instructions. For this end, the two oligonucleotides 59-TGC TGA TGG TAG GCA CTT CCA AGG GTA GTT TTG GCC ACT GAC TGA CTA CCT TGG GTG CCTA CCA T-39 (top strand) and 59-CCT GAT GGT AGG CAC CCA AGG TAG TCA GTC AGT GGC CAA AAC TAC CTT GGA AGT GCC TAC CAT C-39 (bottom strand) were annealed and cloned into the vector pcDNA6.2-GW/ EmGFP-miR. After transfection of L-428 cells with this vector or the empty control vector, global gene expression was analyzed by using Affymetrix HG_133A microarrays essentially as described [7]. Cel files were processed with Affymetrix Expression Console using the MicroarraySuite 5.0 algorithm and scaled to the same target intensity of 500. DNA microarray data have been submitted to the Gene Expression Omnibus (GEO) data base (accession number GSE52831).

Identification of endogenous retrovirus transcripts in Hodgkin's lymphoma cell lines
We established a cDNA library from HL cell line L-1236. After ligation of cDNA fractions with the cloning vector, individual ligation reactions were transformed in E. coli and plated on agar plates for determination of ligation efficiency. Individual E. coli colonies were arbitrarily chosen for further characterization of the transformed vectors. Vectors were isolated, digested (linearized) with NotI, and analyzed on agarose gels. As shown in Figure 1A, individual vectors carried different cDNA inserts as revealed by different sizes of the linearized vectors. Individual vectors were sequenced by using primers flanking the cloning site. We found cDNA inserts of varying length in all vectors. A very short insert of 192 base pairs (bp) with no long open reading frame was found in one of the smallest vectors ( Figure 1A). We designed primers with specificity for this insert and tested the presence of the corresponding transcripts in HL cell lines. We detected expression of this sequence in all HL cell lines tested but not in normal PBMCs ( Figure 1B). A BLAST analysis indicated that this sequence was derived from chromosome 1 in close proximity to the 59 end of a locus encoding a dual specificity phosphatase 5 pseudogene (DUSP5P1). A RetroSearch analysis indicated that the cloned sequence had high similarity to endogenous retroviruses (ERV). The highest similarity was found to an ERV from chromosome 4 (RetroSearch ID: 16168). We used the complete sequence of this ERV (MER65I) as query in a BLAST search. The result revealed that the ERV from the DUSP5P1 locus on chromosome 1 is split into multiple segments from which two segments surround the DUSP5P1 locus (Figure 2A and B). Based on the chromosomal location and similarity with other ERV sequences from ERV family K, in the following sections we use the name ERVK_1q42.13 for this ERV.

Expression of DUSP5P1 and ERVK_1q42.13 in HL cells
We asked whether the DUSP5P1 gene and additional segments of the ERV were also expressed in HL cells. For this end, we designed additional primers corresponding to segments 2 and 3 of ERVK_1q42.13 (see Figure 2). As shown in Figure 3A, in all HL cell lines investigated we found expression of ERVK_1q42.13. Normal PBMCs showed no signals. Reverse transcription with different reverse transcriptase enzymes led to identical results but we obtained no signals in HL cell lines without reverse transcription ( Figure 3B), excluding contamination with genomic DNA. Expression of ERVK_1q42.13 was independent of the B cell phenotype of the HL cells as indicated by the expression of ERVK_1q42.13 in HL cells with a T cell phenotype (L-540, HDLM-2). In addition, expression of ERVK_1q42.13 was not detectable in Epstein-Barr virus-immortalized B cells (LCL, Figure 3C). In contrast to DUSP5P1, DUSP5 was detectable by conventional RT-PCR ( Figure 4A) and quantitative RT-PCR ( Figure 4B) both in HL cell lines and normal PBMCs. In all samples the molar concentrations of DUSP5 exceeded the concentrations of DUSP5P1 ( Figure S2). High signals for DUSP5P1 were also detected in DNA microarray data from HL  cell lines and microdissected HL cells (microarray data from the Gene Expression Omnibus (GEO) data base; Figure S3).
Using the DUSP5 RNA sequence as query for a BLAST search, we identified 3 additional processed pseudogenes in the human genome ( Figure S4), suggesting that DUSP5-like sequences were repeatedly involved in gene rearrangements during evolution. In silico promoter prediction indicated that transcription of DUSP5P1 most likely starts 44 base pairs up-stream of the sequence with high homology to DUSP5. Such transcripts allow the translation of a polypeptide corresponding in part to the substrate binding domain of DUSP5 surrounding the putative substrate binding site ( Figure  S5). In DUSP5 this binding site is characterized by two arginine residues and is highly conserved in vertebrates. Interestingly, these and the following amino acids are mutated in DUSP5P1 ( Figure  S5). Figure 5 shows results from homology modeling of this peptide using the structure of the mitogen-activated protein kinase 1-binding domain of DUSP6 [29] as template.

Expression of DUSP5P1 and DUSP5 in tumor cells
Quantitative RT-PCR indicated high expression of DUSP5P1 not only in HL cells but also in other tumor cells from hematopoietic and non-hematopoietic malignancies ( Figure 6A). In contrast, DUSP5 expression was lower in these cells compared to non-malignant cells ( Figure 6A). Again, in most samples the molar concentrations of DUSP5 exceeded the concentrations of DUSP5P1 ( Figure 6B and Figure 6C). Only in some of the tumor cell lines this ratio was inverted (samples with values below zero in Figure 6B). We asked whether transcripts corresponding to DUSP5 and DUSP5P1 were detectable in the blood of patients with HL. We analyzed blood samples from two patients with fatal course of HL with quantitative RT-PCR. As shown in Figure 7, both patients showed persistence of a high DUSP5P1 expression over the course of the disease after relapsing.
The ratio of DUSP5P1 and DUSP5 correlates with expression of B cell leukemia/lymphoma 2-like 11 (BCL2L11) Down-regulation of the pro-apoptotic factor BCL2L11 is a key event after activation of the extracellular signal-regulated kinase (ERK) pathway which is inhibited by members of the dual specificity phosphatase family. We asked whether the expression of DUSP5 and subsequent inhibition of the ERK pathway has an influence on expression of BCL2L11 in HL cells. According to DNA microarray data, HL cells have relatively low signal intensities for BCL2L11 compared to normal cells ( Figure S3). Using quantitative RT-PCR, we observed a correlation between high expression of DUSP5 and high expression of BCL2L11 ( Figure 8A). Expression of BCL2L11 was high in all tested PBMC samples. HL samples showed lower expression of BCL2L11. In one cell line (L-540) expression of BCL2L11 was completely absent. The expression of DUSP5P1 correlated inversely with the expression of BCL2L11 ( Figure 8B). The ratio of DUSP5P1 and DUSP5 was inversely correlated with expression of BCL2L11 ( Figure 8C). After knock-down of DUSP5 in HL cell lines, we   Figure 9B). We tested whether over-expression of the short cloned transcript from the cDNA library was able to modulate expression of BCL2L11 in HL cells. As shown in Figure 10, over-expression of this transcript did not suppress expression of DUSP5 or BCL2L11.

Discussion
HL is a lymphoproliferative disease with unclear pathogenesis. At the molecular level, activation of the nuclear factor of kappa light polypeptide gene enhancer in B-cells (NFKB) signaling pathway has been identified as important mechanism for HL pathogenesis [30]. The human herpesvirus Epstein-Barr virus (EBV) is one factor which can lead to activation of this pathway. EBV is involved in the pathogenesis of several lymphoproliferative diseases [31] and can be found in the tumor cells of a high percentage of HL patients [32]. In addition to EBV, other viruses have been implicated in the pathogenesis of HL. These viruses include measles virus [33], cytomegalovirus [34], torque teno virus (TTV) [35], and human herpesvirus 6 (HHV) [36]. The prevalence of these viruses and EBV in the healthy population is high, indicating that virus infection alone cannot explain lymphoma development. In addition to the mentioned viruses, retroviruses have been implicated in HL pathogenesis [37]. Interestingly, links between putative HL-associated viruses (EBV, HHV6, TTV) and endogenous retroviruses have been established [38][39][40][41][42]. EBV and HHV6 can induce expression of endogenous retroviruses [37,38]. High prevalence of these viruses or immunoreactivity against viral antigens in patients with autoimmune diseases including multiple sclerosis (MS) has been observed [40][41][42]. Epidemiologic similarities between MS and HL have been described and endogenous retroviruses are considered as putative pathogens in MS [13]. Gene expression profiling and immuno-logical analysis of brains from superantigen treated animals further support the concept of the involvement of endogenous retroviruses in the pathogenesis of this disease [43].
Recently, activation of the colony stimulating factor 1 receptor (CSF1R) by activation of an endogenous retrovirus in HL cells has been described [14]. CSF1R is a proto-oncogene and aberrant  expression of this receptor might be directly involved in pathogenesis of HL. The observation of expression of ERVK_1q42.13 and DUSP5P1 transcripts in HL cells is another example of linked expression of ERV related sequences and neighboring genes in HL. In contrast to CSF1R, the potential function of DUSP5P1 is unknown. DUSP5P1 is considered to be a pseudogene and no functions of DUSP5P1 transcripts have been described. DUSP5P1 transcripts have been detected in retinoic acid treated NT2 teratocarcinoma cells [44]. Expression of endogenous retroviruses in NT2 cells and other testicular germ cell tumor cell lines is a well-known phenomenon [45]. The presence of ERV sequences in human genes has been shown to confer cell type-specific gene expression [46]. The simultaneous presence of DUSP5P1 transcripts and ERV transcripts in testicular germ cell tumor cells and HL cells suggests that in both cell types transcription is deregulated in a similar way. Together with the epidemiological similarities between HL and testicular cancer [47] this observation might indicate common pathogenetic mechanisms for both diseases.
One possible function of expressed pseudogenes is interference with the activity of the genes from which the pseudogenes are derived. DUSP5P1 is not the only DUSP5 pseudogene in the human genome, suggesting that DUSP5 was frequently involved in genetic rearrangements during human evolution. DUSP5 is involved in T-cell development [48] and alterations in DUSP5 have been observed in lymphoid malignancies [49,50]. The exact function of DUSP5 in lymphoid cells has to be established. Overexpression of DUSP5 in transgenic mice results in autoimmunity but also in reduced T-cell proliferation [48]. Because DUSP5 is regulated by its substrate mitogen-activated protein kinase 1/ extracellular signal-regulated kinase 2 (MAPK1/ERK2) [51], a complex feedback regulation of this pathway might result in signals promoting or inhibiting cell growth and survival. The ERK pathway plays an important role in HL pathogenesis [52][53][54] and inhibition of the negative regulator DUSP5 might be important for proliferation and survival of tumor cells as indicated by the low expression of DUSP5 in the majority of tumor cells. Inhibition of DUSP5 might result in enhanced activation of the ERK pathway and subsequent down-regulation of pro-apoptotic BCL2L11 [55]. DUSP5P1 polypeptides may be able to interfere with the function of DUSP5. Whether DUSP5P1 transcripts were translated into such polypeptides in vivo has to be determined. The predicted  polypeptides consist only of the regulatory domain of a typical DUSP family member and a catalytic domain will not be present. Only for this domain of DUSP5 the structure has been determined [56]. However, the high sequence homology between DUSP5 and DUSP6 together with the high conservation of amino acids considered to be involved in substrate specificity allows the meaningful modeling of such polypeptide on the basis of DUSP6. It seems unlikely that such DUSP5P1 polypeptides can bind the same substrates as DUSP5, because the amino acids involved in substrate specificity are changed. On the other side, the high homology of the 59 part of DUSP5P1 and DUSP5 RNAs might allow nonsense-mediated transcriptional gene silencing or activation of other nonsense-mediated decay mechanisms [57]. Other mechanisms include the sequestering of RNA binding molecules which was recently described for non-coding RNA species in patients with amyothrophic lateral sclerosis [58]. In all cases, reduced activity of DUSP5 will result in increased ERK activity and subsequent inhibition of pro-apoptotic BCL2L11. In the future, it might be possible to develop new targeted treatment strategies for HL on the basis of the elucidation of this pathway. The short clone from the cDNA library has no BCL2L11suppressing activity. This speaks against a simple model in which expression of this sequence inhibit DUSP5 and finally regulate BCL2L11. Relaxed gene regulation in tumor cells can result in generation of transcripts without specific function. On the other hand, we cannot exclude the possibility that other transcripts from the DUSP5P1 locus interfere with expression of function of DUSP5. It seems possible that expression of the ERV sequence is not directly involved in gene regulation but is only a surrogate marker for high transcriptional activity of the complete DUSP5P1 locus in HL cells. The complete spectrum of transcripts form the DUSP5P1/ERVK_1q42.13 locus in HL cells and other tumor cells has not been determined. EBV infection leads to downregulation of BCL2L11 [59][60][61][62]. Transgenic over-expression of BCL2L11 isoforms in EBV infected B cells or nasopharyngeal carcinoma cells enhance apoptosis of these cells [59,60]. BCL2L11 is regulated by the EBV encoded nuclear antigens (EBNA)3A and 3C at multiple levels. EBNA3C can bind directly to the BCL2L11 promoter and epigenetically down-regulates BCL2L11 in conjunction with the polycomb repressive complex 2 [63]. EBNA3C is not expressed in EBV-positive HL and alternative pathways (probably involving DUSP5P1) are required for down-regulation of BCL2L11.
Taken together, we found high expression of DUSP5P1 and low expression of the DUSP5 target BCL2L11 in HL cells. Whether there is a functional connection between these two observations or whether this is only a coincidence requires further investigations. The specificity of DUSP5P1 and ERV transcripts for HL cells might allow the development of methods for detection of minimal residual disease. Circulating HL cells can be present in the peripheral blood of patients with high tumor burden [16]. Our data indicate that DUSP5P1 transcripts are detectable in the peripheral blood of HL patients with active disease. Whether the detection of such HL specific transcripts can be used for the early detection of patients with high risk for relapse should be investigated.