Bleomycin Resistance in Mammalian Cells Expressing a Genetic Suppressor Element Derived from the SRPK1 Gene¹

BLMs,³ discovered by Umezawa et al. (1) as a group of antitumor antibiotics produced by Streptomyces verticillus, are presently used clinically in combination with other drugs for the treatment of various tumors, including head and neck cancers, Hodgkin’s disease, and germ cell tumors (2, 3, 4). The major cause of BLM cytotoxicity is probably related to its ability to provoke DNA degradation. BLM (1) is a water-soluble glycopeptide that binds iron and oxygen in vivo to produce activated BLM, an Fe³⁺ hydroperoxide. When bound to DNA, with the metal binding domain located in the minor groove and the bithiazole group partially intercalated into the double helix, this intermediate induces the formation of single- and double-strand breaks at sequence-specific sites. The DNA double-strand breaks are considered as responsible for BLM toxicity (5, 6, 7). However, various studies showed that BLM may also be responsible for other cellular damage, including RNA cleavage (8) and oxidative cell wall damage (9), which might also contribute to its toxicity.

A few hundred molecules of BLM are enough to kill a cell (10). However, its effectiveness is markedly restricted because BLM does not enter cells readily. In mammalian cells, BLM was shown to bind a M_r 250,000 cell surface protein, before internalization by endocytosis (11), and cell sensitivity to BLM depends on the abundance of this protein (12). Electroporation of cultured cells was shown to enhance its potency at least 1000-fold. Extension of this technique to the treatment of human tumors led to a novel antitumor approach, termed electrochemotherapy (13, 14), which is presently developing rapidly. Finally, activity of internalized BLM is also controlled by cell-dependent factors, such as DNA lesion repair pathways (15, 16) and expression level of BLM hydrolase, a protease that inactivates BLM (17). Despite a number of studies over more than 30 years, the mechanism of action of BLM remains unclear, most likely because of the structural complexity of the molecule and the variety of potential targets and effectors.

In fact, cell response to BLM, and to any other cytotoxic agent as well, depends on the balance between two groups of genes: genes which expression inhibits BLM activity and results in a resistance phenotype, and genes which expression is required for cell killing. To identify genes mediating cell sensitivity to BLM, we selected GSEs conferring resistance to this drug. GSEs are short cDNA fragments encoding peptides acting as dominant inhibitors of protein function or antisense RNAs inhibiting gene expression (18). GSEs behave as dominant selectable markers for the phenotype associated with the repression of the gene from which they derived, thus allowing identification of this gene. For example, this strategy previously allowed the demonstration that kinesin heavy chain (19) and some members of the protein arginine methyltransferase family⁴ are involved in the control of cell response to various DNA-damaging agents.

In this study, we selected and characterized a GSE, GSE^BLM, conferring selective resistance to BLM in Chinese hamster and human cells by functional inhibition of SRPK1. SRPK1 is a protein serine kinase that regulates the activity of RS-proteins (arginine-serine-rich proteins), a group of nuclear factors controlling a variety of physiological processes including RNA processing and spliceosome assembly (20).

Human epitheloid carcinoma cells (HeLa) and Chinese hamster lung fibroblasts (DC-3F) were cultured as monolayers in standard conditions in DMEM and MEM (Life Technology, Cergy-Pontoise, France), respectively (21). Medium was supplemented with 10% FCS and antibiotics.

BLM, cisplatin, and etoposide were purchased from Laboratoires Roger Bellon (Neuilly, France), Aventis (Montrouge, France), and Pharmachemie BV (Haarlem, the Netherlands), respectively. Staurosporine, camptothecin, and paraquat were from Sigma-Aldrich (La Verpillère, France).

A normalized library of random cDNA fragments of approximately 2.5 × 10⁷ clones, inserted in the ClaI site of the retroviral plasmid pLNCX (22), was prepared from DC-3F cells as described previously (19).

The library was packaged by transfecting 9 × 10⁷ BOSC 23 cells (1.5 × 10⁶/60 mm plate) with 80 μg of library DNA (23). The viral suspension, collected 24, 36, and 48 h later, was used to infect DC-3F/cl23 cells (1 × 10⁶/plate, 10 100-mm plates), after treatment with DEAE dextran (20 μg/ml) for 20 min. DC-3F/cl23 cells are DC-3F cells made sensitive to retrovirus infection by transfection with the pJET plasmid, which carries an ecotropic retrovirus receptor gene (24). Infection was repeated three times at 12-h intervals. Under these conditions, ∼80% of the cells were infected as determined by PCR amplification of proviral inserts from genomic DNA of isolated clones (see below). Genomic DNA was prepared using the QIAamp Blood Mini kit (Qiagen, Courtabeuf, France). Forty-eight h after the last infection, cells were replated and grown for 24 h.

One million cells were plated in 100-mm plates (five plates) and treated 24 h later with BLM (30 μm; 72 h). After drug exposure, surviving cells were rinsed three times with PBS and replated in five 100-mm dishes. Culture medium was renewed daily, and surviving clones were allowed to grow for 6–7 days. After trypsinization, cells were replated again in five 100-mm dishes and grown for 24 h, before a second BLM treatment in the same conditions as above. Twenty isolated clones were picked and grown first in 24-wells plates and then in 60-mm dishes. PCR amplification (see below) revealed that 17 of them contained at least one insert. After fractionation by electrophoresis in 1.2% agarose gel, 10 fragments were purified using the QIAquick Gel Extraction kit (Qiagen). Purified fragments were sequenced by Cybergen ESGS (Evry, France).

The following oligonucleotides were used as PCR primers: 5′-GCCCCAAGCTTTGTTAAC-AACGATGGATG-3′, 5′-CTCCGCGGCCCC-AAGCTTTGTTTACATCGAT-3′ (sense for pLNCX with inserts and empty vector, respectively), and 5′-ATGGCGTTACTTAAGCTAGCTTGCCAAACCTAC-3′ (antisense). The sequence of the sense primer was designed to eliminate the ClaI site. Inserts were amplified from 50 ng of genomic DNA through 30 cycles of PCR (80 s at 58°C, 120 s at 72°C, 70 s at 94°C, and 10 min at 72°C).

PCR products, first digested by HindIII and ClaI, were then ligated in the corresponding sites of the pLN12 vector, a pLNCX vector in which the G418 resistance gene is replaced with the HR5 selection gene. HR5 selection is based on a point mutation of the Na⁺/H⁺ exchanger isoform 1 (NHE1), which confers resistance to lethal intracellular acidification in the presence of amiloride (25). In pLN12, inserts were kept in the same orientation as in the original pLNCX. Each insert was individually tested for its ability to confer BLM resistance after retroviral transduction in DC-3F/cl23 cells. More than 70% of the cells surviving HR5 selection contained inserts detected by PCR amplification. BLM resistance was evaluated following two different protocols: treatment with 30 μm BLM for 72 h, or cell electropermeabilization by eight electric pulses of 1300 V/cm and 100 μs at the frequency of 1 Hz, in the presence of 5 nm BLM (26). Surviving clones were stained with crystal violet either for determination of the absorbance at 600 nm (23) or individual clone counting. For all drug resistance studies, the number of plated cells was systematically controlled by measuring the cloning efficiency from 500 untreated cells plated in triplicate in 60-mm dishes.

These procedures have been described previously (27).

DC-3F or HeLa cells, stably transduced with pLN12-GSE^BLM, were plated at 8 × 10⁵ cells/plate in 100-mm plates. Twenty-four h later, the cells were treated with the different drugs at the indicated concentrations. After washing with PBS and trypsinization, the cells were returned to drug-free medium and allowed to form colonies for 8–10 days. Controls were cells infected with insert-free vector.

The eukaryotic expression vector pCDNA3-SRPK1, containing the human SRPK1 cDNA, was kindly provided by Dr. Xiang-Dong Fu (University of California San Diego, San Diego, CA). After a first transfection with pLN12 or pLN12-GSE^BLM, HR5 resistant HeLa cells were again transfected with either pcDNA3 or pcDNA-SRPK1 using LipofectaminePLUS (Life Technology, Cergy-Pontoise, France). Transfected cells were selected in the presence of G418 (Prolabo, Gradignan, France) at 800 μg/ml, and BLM resistance was tested as above on the entire resistant population.

SRPK1 expression in transfected cells was determined by Western blot analysis, using an anti-SRPK1 monoclonal antibody (PharMingen, Le Pont de Claix, France). Preparation of cell extracts, protein fractionation by SDS-PAGE (50 μg/well), and immunoblotting conditions have been described previously (28). Transfer to Millipore Immobilon-P membranes (Merck-Eurolab Polylabo, Strasbourg, France) was carried out as described by the manufacturer. Horseradish peroxidase-conjugated mouse antibody was used as secondary antibody (Amersham Pharmacia Biotech, Orsay, France). For control of gel loading, the membranes were also probed with anti-α-tubulin mouse monoclonal antibody (Sigma-Aldrich, Saint Quentin-Fallavier, France). After enhanced chemiluminescence detection, band intensities were quantified in volume using the variable mode imager Typhoon 8600 and the Image Quant software from Molecular Dynamics (Amersham Pharmacia Biotech, Orsay, France).

Cell extracts, protein fractionation by SDS-PAGE, and immunoblotting conditions were as described above. A polyclonal antibody against human bleomycin hydrolase, kindly provided by Dr. P. A. O’Farrell (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY), was used as a primary antibody. Bound antibody was visualized with horseradish peroxidase-conjugated goat antirabbit IgG and enhanced chemiluminescence detection (Amersham-Pharmacia Biotech, Saclay, France). Recombinant bleomycin hydrolase, used as a control, was also kindly provided by Dr. P. A. O’Farrell.

Fig. 1 shows a schematic representation of the GSE selection procedure. The plasmid library was transfected into packaging BOSC 23 cells, and the virus released in the supernatant medium was used to infect DC-3F/cl23 cells. Virus-infected cells were then submitted to BLM selection under conditions allowing the growth of ∼1 in 10,000 cells (30 μm; 72 h) The number of library-transduced cells surviving BLM treatment at this stage was approximately twice higher than that of the control (DC-3F cells transduced with insert-free pLNCX). BLM-surviving cells were then replated in an equal number of dishes and grown for 24 h, before a second BLM selection in the same conditions as above. At this stage, the number of surviving clones in the experiment was ∼3-fold higher than that in the control, indicating that the cell population was enriched for cells containing biologically active GSEs.

Twenty BLM-resistant clones were picked and individually analyzed. PCR analysis of the integrated proviral inserts showed that 17 of them contained at least one fragment, with sizes ranging approximately from 200 to 1000 bp (Fig. 2). Ten GSEs were purified and sequenced. The longest one (∼1000 bp) was a combination of three other fragments. This result was consistent with previous data showing that fragments longer than ∼500 bp are usually artifactual composite sequences⁴. Nine fragments of ∼300 bp displayed different sequences. For five of them, homology analysis in sequence databases did not allow unambiguous identification of the corresponding genes. In contrast, four others, with strong homologies with known nucleic acid and protein sequences, were individually tested for their capacity to confer resistance to bleomycin. For this purpose, each of them was inserted in the pLN12 retroviral vector, in the same orientation as in the original vector, and transduced into DC-3F/cl23 cells. After HR5 selection, >70% of the surviving cells were found to carry the transduced GSE. Only one of these fragments was identified as a functional GSE, able to confer resistance to BLM (30 μm; 72 h). This GSE, designated as GSE^BLM, was picked for further analysis.

After treatment with BLM in the above conditions, the cloning efficiency of DC-3F/cl23 cells containing the GSE^BLM was 1.7 (± 0.09)-fold higher than that of control cells. We also examined the capacity of GSE^BLM to confer BLM resistance to DC-3F/cl23 cells electropermeabilized (eight pulses of 100 ms, 1300 V/cm, 1 Hz frequency) in the presence of 5 nm BLM. In these conditions, the survival of GSE^BLM-containing cells was 2.5-3–fold higher than that of the control (Fig. 3). This experiment indicated that BLM resistance associated with GSE^BLM expression did not result from an altered cellular drug uptake.

GSE^BLM efficiency was also tested in human cells. HeLa cells were transfected with the pLN12 vector containing or not the GSE. After HR5 selection, treatment of the transfected cells with BLM (30 μm) was reduced to 24 h because, in preliminary testing, HeLa cells were found to be more sensitive to this drug than DC-3F cells. After BLM treatment, the cloning efficiency of GSE^BLM-transfected HeLa cells was 8 (± 2)-fold higher than that of control cells transfected with insert-free vectors. These results showed that the effect of GSE^BLM is not species specific.

Search in sequence databases showed that GSE^BLM sequence shares a remarkable identity with a sequence located in the functional domain of the human SRPK1 (94% in nucleotides and 98% in amino acids). GSE^BLM was then used to probe a λZAPIIcDNA library made from DC-3F cells. Five clones, carrying inserts with sizes ranging from 2000 to 4200 bp, were selected. Sequence analysis showed that the longest cDNA nucleotide sequence had 76.7 and 89% identity to human and mouse SRPK1 nucleotide sequences, respectively. By comparison with the human sequence, an open reading frame of 4071 bp⁵ was identified in the Chinese hamster cDNA. The corresponding amino acid sequence is 92.6 and 89.7% homologous to the human and mouse SRPK1 sequences, respectively, but is missing the first three amino acid sequence (Met, Glu, and Arg).

Sequence analysis also showed that GSE^BLM is sense oriented and encodes a peptide extending from amino acids 94–153. To verify that GSE^BLM is indeed acting as a peptide, a modified GSE was PCR amplified using the following sense primer: 5′-GCCCCAAGCTTTGTTAACATCGATGGATGGATGGGTAGTTATC-3′. This primer contains a stop codon (underlined) preventing GSE translation. When the modified GSE, inserted between the ClaI and HindIII sites of pLN12, was transduced into HeLa cells, the HR5 selected cells did not show any detectable resistance to BLM (Fig. 4). These results showed that GSE^BLM translation is required for induction of BLM resistance and that it is most likely active as a peptide inhibiting SRPK1 activity.

To further demonstrate that GSE^BLM induced inhibition of SRPK1 activity is associated with resistance to BLM, HeLa cells, stably transduced with pLN12-GSE^BLM, were transfected with pcDNA3-SRPK1, a plasmid carrying the human SRPK1 cDNA and the G418 resistance marker. To avoid potential problems of clonal variability, BLM resistance was determined on whole G418-resistant cell populations. Fig. 5,A shows that an average 1.5-fold increased expression of SRPK1 was detected in cells transfected with pcDNA3-SRPK1 (Lanes 2 and 4) compared with cells transfected with insert-free pcDNA3 (Lanes 1 and 3). Fig. 5,A also shows that SRPK1 expression was not modified in cells expressing GSE^BLM (Lanes 1 and 3). Increased expression of SRPK1 had no effect on cell sensitivity to BLM (Fig. 5,B, Lanes 1 and 2), whereas it completely abolished an 8-fold resistance induced by GSE^BLM expression (Fig. 5 B, Lanes 3 and 4). These experiments confirmed that a decreased SRPK1 activity in GSE^BLM-expressing cells is associated with resistance to BLM.

We then examined the effect of GSE^BLM on the cell sensitivity to various anticancer drugs with different mechanisms of action. We first tested the following agents on DC-3F and HeLa cells, expressing or not GSE^BLM: cisplatin (DNA-damaging agent), camptothecin (topoisomerase I inhibitor), etoposide (topoisomerase II inhibitor), and staurosporine (protein kinase inhibitor). Fig. 6 shows that in both cell lines GSE^BLM only conferred resistance to BLM. We then tested two additional agents that kill cells through mechanisms of action sharing some common traits with that of BLM: paraquat, an oxidative agent, and ionizing radiation, which provokes the formation of DNA double-strand breaks. We did not detect any resistance to either one of these agents in GSE^BLM-expressing cells.

The previous experiments indicated that expression of GSE^BLM in DC-3F and HeLa cells is selectively associated with resistance to BLM. We then examined the possibility that alteration of SRPK1 activity might affect proteins specifically involved in BLM toxicity. Two such proteins are presently known. One is the membrane BLM binding protein that transports BLM through cell membranes. However, because GSE^BLM confers drug resistance to electropermeabilized cells, this protein should not play any part in the GSE mechanism of action. The other protein is BLM hydrolase, a protease that inactivates BLM. BLM does not carry the sequence motifs phosphorylated by SRPK1. However, alteration of the phosphorylation of one or several SRPK1 substrates might result in changes in the expression of BLM hydrolase. We measured the expression of this enzyme in HeLa cells, in the presence or absence of GSE^BLM, by immunoblotting using a polyclonal antibody against human bleomycin hydrolase. Fig. 7 shows that GSE^BLM had no effect on the expression of this enzyme in HeLa cells, where we observed the highest level of drug resistance.

Selection of GSEs conferring resistance to a cytotoxic agent is a very powerful technique to identify drug sensitivity genes. From a Chinese hamster fibroblast cDNA fragment library, we selected GSEs conferring resistance to bleomycin and further characterized one of them, designated GSE^BLM. The nucleotide sequence of this GSE was 94% identical to that of a region from the human SRPK1 cDNA, extending from nucleotides 804–980. Using GSE^BLM as a probe, we cloned and sequenced the Chinese hamster SRPK1 cDNA. The overall nucleotide sequence of this cDNA was 76.7% identical to that of human SRPK1 cDNA. GSE^BLM, which is active in both Chinese hamster and human cells, targets a highly conserved sequence in SRPK1. Translation of the sense-oriented GSE^BLM is required for induction of drug resistance, as demonstrated by the absence of resistance in cells transfected with a GSE containing a stop codon in its sequence. Because it was also shown that SRPK1 expression was not modified in cells expressing GSE^BLM, these data support the conclusion that GSE^BLM is coding a peptide acting as a dominant-negative inhibitor of SRPK1.

Identification of the SRPK1 gene as the gene from which GSE^BLM was derived is based on three arguments: presence in the human gene of a nucleotide sequence 94% identical to that of GSE^BLM; cloning of the SRPK1 cDNA by probing a Chinese hamster cDNA library with the GSE; and reversion of GSE^BLM-induced BLM resistance in cells transfected with SRPK1 cDNA. However, mammalian cells contain several SRPK enzymes. Alternative splicing of SRPK1 transcripts can give rise to two enzyme forms. Although both isoforms have the same substrate specificity and the same subcellular localization, they might perform different functions; SRPK1a differs from SRPK1 by insertion of 171 amino acids at its NH₂-terminal end and specifically interacts with scaffold attachment factor-B. Because the sequence corresponding to GSE^BLM is common to both proteins, this GSE is likely to inhibit both of them. Another enzyme, SRPK2, is highly related to SRPK1 in sequence, enzyme activity, and substrate specificity. These two kinases differ by a different expression in various human tissues and by the presence of a proline-rich sequence at the SRPK2 NH₂ terminus that may contribute to in vivo specific function and/or regulation. In their functional domains, SRPK1 and SRPK2 amino acid sequences are 77% identical. Although SRPK2 homology with GSE^BLM amino acid sequence is reduced to 81.3%, a possible effect of GSE^BLM on SRPK2 activity cannot be excluded.

SRPK family members represent an unusual class of constitutively active protein kinases and are characterized by the presence of a large spacer (250–300 residues), which divides the kinase domain into two halves, and a long NH₂-terminal tail (29). GSE^BLM homologous sequence (amino acids 95 to 153) is located in the NH₂-terminal half of human SRPK1 kinase domain. This region is highly conserved in Sky1p, the only SRPK enzyme in Saccharomyces cerevisiae. The three-dimensional structure of the yeast protein was determined on crystals of a fully active truncated form of the protein, in which the NH₂-terminal tail and the spacer were deleted (30). From these studies, we can deduce that the GSE^BLM corresponding region contains several structural elements, including helices αC and αC′, which are essential to stabilize the active form of the enzyme through interaction with other domains of the protein. How GSE^BLM peptide can inhibit SRPK1 activity remains hypothetical. One possibility would be that this peptide would interact with the enzyme cofactor or substrate, thus preventing their binding to the enzyme. Alternatively, GSE^BLM peptide might inhibit peptide interactions within the enzyme molecule, such as the packing of the helix αC against helix αE mediated by helix αC′ and required for locking the activation loop in the active conformation (30).

RS proteins (arginine-serine-rich proteins) are essential RNA processing factors and are characterized by the presence of at least one RNA recognition motif and a COOH-terminal domain rich in Arg-Ser (RS) dipeptide repeats (20). RS domains are known to participate in protein-protein and protein-RNA interactions during spliceosome assembly and also to function as nuclear localization signals. SRPKs specifically phosphorylate most (if not all) Ser in the RS repeats and thus regulate the functions of RS proteins (31). GSE^BLM expression only induced resistance to BLM, whereas it had no effect on the sensitivity of DC-3F and HeLa cells to a variety of other cytotoxic agents with different mechanisms of action. Inhibition of SRPK1 activity in GSE^BLM-expressing cells or increased expression of the enzyme in cells transfected with the SRPK1 cDNA did not change the cell growth rates. Therefore, GSE^BLM-induced BLM resistance did not result from a modification of cell growth properties, which would also be expected to have consequences on the sensitivity to other compounds. A selective resistance to BLM might result from an effect of SRPK1 inhibition on the activity and/or the expression of protein(s) specifically controlling BLM toxicity. Because GSE^BLM induces BLM resistance in permeabilized cells, this excludes any protein involved in the transport of the drug across cell membranes. Bleomycin hydrolase is a very unlikely substrate of SRPK1, because it does not contain any RS domain, and we have shown that its expression is not modified in GSE-expressing cells. Another possibility would be that expression and/or cellular localization of protein(s) involved in the repair of BLM-induced DNA lesions would be controlled by SRPK1. Inhibition of SRPK1 in GSE^BLM-expressing cells might then alter repair pathways of these lesions and thus induce cell resistance to this drug.

Very recently, it was reported that Sky1p mediates cisplatin toxicity in Saccharomyces cerevisiae (32) . This report also showed that down-regulation of SRPK1 in human ovarian carcinoma A2780 cells with antisense oligodeoxyribonucleotides resulted in a decreased sensitivity to cisplatin. The sensitivity of these cells to other drugs was not reported. It should be pointed out that our studies were carried out with different cell lines, and that GSE^BLM did not induce a down-regulation of SRPK1 but was rather acting as a dominant-negative inhibitor. Whether these differences explain the absence of resistance to cisplatin in our cell lines remains to be studied. Nevertheless, both studies support the conclusion that SRPK enzymes are involved in the control of cell sensitivity to anticancer agents. Another important point is that both studies show that alterations of SRPK1 activity are associated with low levels of drug resistance, which are generally believed to be more relevant to clinical situations than high resistance processes frequently analyzed in experimental models. Detection and characterization of such new mechanisms, controlling cell sensitivity to anticancer agents, may lead to a better understanding of the determinants of the clinical response to these agents.