STAT3-Activated GM-CSFRα Translocates to the Nucleus and Protects CLL Cells from Apoptosis Free

B-cell chronic lymphocytic leukemia (CLL), the most common hematologic malignancy in the Western hemisphere, is characterized by a dynamic imbalance between proliferation and apoptosis of neoplastic B lymphocytes coexpressing CD5 and CD19 antigens (1, 2). Despite recent improvements in managing this disease, CLL remains incurable. Like other lymphoid neoplasms, CLL cells usually express the CD20 antigen. Combining the anti-CD20 antibody rituximab with granulocyte-macrophage colony-stimulating factor (GM-CSF) produced higher response rates than did single-agent rituximab in relapsed follicular B-cell lymphoma (3) and in initial studies in CLL (4).

GM-CSF is produced by a variety of cells, including stromal cells and cells of hematopoietic origin, including B1a cells (5), and regulates the survival, proliferation, differentiation, and activation of hematopoietic cells (6) as well as the function of dendritic cells (7) and T cells (8). GM-CSF regulates by binding to the cell-surface GM-CSF receptor (GM-CSFR). GM-CSFR, first identified on cells of the myelomonocytic lineage by ligand-binding studies (9, 10), belongs to a structurally distinct family of colony-stimulating hematopoietic growth factor receptors that include receptors that bind to GM-CSF, M-CSF, or G-CSF (11). GM-CSFR is a heterodimer comprising GM-CSFRα (12) and GM-CSFRβ (also known as βc) subunits (13). The 80-kDa GM-CSFRα subunit (CD116) is cytokine-specific, whereas the 120-kDa CSFRβ subunit (CD131) is nonspecific and is shared with the cytokine-specific β subunits of the IL3 and IL5 receptors. GM-CSFR does not have intrinsic tyrosine kinase activity but associates with the tyrosine kinase JAK2, which is required for the initiation of signaling and biologic activity. Although the Ig-like domain of GM-CSFRα is a crucial determinant of GM-CSF binding (14), in the absence of GM-CSFRβ, the GM-CSFRα subunit binds to GM-CSF with low affinity (11). Both α and β subunits are required for GM-CSF signaling, and the cytoplasmic domains of both GM-CSFRα and β are essential for receptor activation (15, 16); however, only the β domain associates with JAK2 (17).

B-cell CLL cells express CD5, a cell-surface antigen commonly expressed on normal T lymphocytes (2). Although primarily a myeloid growth factor, GM-CSF affects T-cell function (8). Antigen-stimulated CD8⁺ T cells express GM-CSFRβ (18), and human natural killer (NK) cells, 80% of which express CD8, also express CD160, recently found expressed on CLL cells from 98% of patients (19). Because of the similarities between CLL cells and T lymphocytes, because data suggested that GM-CSF upregulates the expression of CD20 on the surface of CLL cells (20), and because GM-CSF enhanced the effect of anti-CD20 antibodies in follicular lymphoma (3), we sought to explore the effect of GM-CSF on CLL cells.

Consistent with previous reports (21), we found that GM-CSF did not activate GM-CSFR–induced signaling pathways in CLL cells. However, we detected GM-CSFRα, but not GM-CSFRβ, on the cell surface, in the cytoplasm, and in the nucleus of CLL cells. We demonstrated that signal transducer and activator of transcription (STAT)-3, constitutively activated in CLL cells (22), activates the GM-CSFRα promoter and induces GM-CSFRα production, and that GM-CSFRα protects CLL cells from apoptosis.

Peripheral blood cells were obtained from patients with CLL treated at The University of Texas MD Anderson Cancer Center Leukemia Clinic (Houston, TX). Institutional Review Board approval and patients' written informed consent were obtained. Peripheral blood was obtained from untreated patients. The clinical characteristics of the patients whose peripheral blood samples were used in this study are presented in Supplementary Table S1.

To isolate low-density cells, peripheral blood cells were fractionated using Histopaque 1077 (Sigma-Aldrich). More than 95% of the fractionated peripheral blood lymphocytes obtained from these patients were CD19⁺/CD5⁺, as assessed by flow cytometry.

Fractionated CLL cells were maintained in DMEM (Sigma-Aldrich) supplemented with 10% FBS (HyClone). The MM1 and the human embryonic kidney (HEK) 293FT cell line were obtained from the American Type Culture Collection. MM1 cells were maintained in RPMI-1640 (Sigma-Aldrich) and 293FT cells were maintained in DMEM supplemented with 10% FBS, both in a humidified atmosphere of 5% carbon dioxide at 37°C. Before analysis, MM1 cells were incubated with either recombinant human IL6 (BioSource International) or tissue culture medium.

Total RNA was extracted using the TRIzol reagent (Life Technologies) and cDNA was synthesized using the SuperScript First-Strand Synthesis System (Life Technologies). Human GM-CSFRα and GM-CSFRβ primers were obtained from Applied Biosystems (Life Technologies). The cell lines OCI/AML3, OCI-2, K562, and Jurkat were used as positive and negative controls. To amplify GM-CSFRα transcripts, a regular PCR program was performed with 30 cycles, and to amplify GM-CSFRβ transcripts a 50-cycle touchdown program was executed.

Cells were suspended in 100 μL of PBS and split into duplicate tubes. Twenty microliters of antibody or its isotype was added, and the tubes were incubated in the dark at room temperature for 30 minutes. After incubation, the cells were washed, suspended in 500 μL of PBS, and analyzed using the FACSCalibur flow cytometer (Becton Dickinson Immunocytometry Systems). The following antibodies and their corresponding isotype controls were used: mouse anti-human CD116, CD131, CD19, and CD5 (BD Biosciences).

Cell pellets were lysed in ice-cold radioimmunoprecipitation assay (RIPA) buffer. The lysed cell pellets were incubated on ice for 5 minutes and the protein concentration was determined using the Micro BCA Protein Assay Reagent Kit (Thermo Scientific/Pierce). Supernatant proteins were denatured by boiling for 5 minutes in SDS loading buffer, separated by SDS-PAGE, and transferred to a nitrocellulose membrane.

Following transfer, equal loading was verified by Ponceau S staining. The membranes were blocked with 5% skimmed milk in Tris-buffered saline and incubated with one of the following antibodies: monoclonal mouse anti-human STAT3 (BD Biosciences and Cell Signaling Technology); monoclonal mouse anti-human phosphotyrosine STAT3, polyclonal rabbit anti-human phosphoserine STAT3, monoclonal mouse anti-human STAT5 and phosphotyrosine STAT5, AKT and phosphoserine AKT, ERK-1/2 and phosphotyrosine ERK-1/2, and monoclonal mouse anti-human JAK2 and phosphotyrosine JAK2 (Cell Signaling Technology); monoclonal mouse anti-human GM-CSFRα and GM-CSFRβ (BD Biosciences); monoclonal mouse anti-human β-actin (Sigma-Aldrich); or mouse anti-GFP (Millipore/Upstate) antibodies.

After binding with horseradish peroxidase–conjugated secondary antibodies, blots were visualized with an ECL detection system (GE Healthcare), and densitometry analysis was performed using the Epson Expression 1680 scanner (Epson). Densitometry results were normalized by dividing the numerical value of each sample signal by the numerical value of the signal from the corresponding loading control.

A standard immunohistochemistry method was used. Briefly, bone marrow (BM) clots and smears were subjected to antigen retrieval performed with Diva Decloaker buffer in a Decloaking Chamber pressure cooker (Biocare Medical). After the slides were blocked with 3% hydrogen peroxide and Background Sniper protein block (Biocare Medical), they were incubated for 1 hour at room temperature with GM-CSFRα antibodies (1:100; Sigma-Aldrich) and counterstained with hematoxylin (Dako). Tris-buffered saline supplemented with 0.1% Tween were used for washing between the steps.

Whole-cell lysates, cytoplasmic fractions, or nuclear fractions, were incubated with 4 μg of polyclonal rabbit anti-human JAK2 or rabbit anti-GFP antibodies (Millipore/Upstate) for 16 hours at 4°C. Protein A agarose beads (Millipore/Upstate) were added for 2 hours at 4°C. As a negative control, whole-cell lysates, cytoplasmic fractions, or nuclear fractions were incubated either with rabbit serum and protein A agarose beads or with protein A agarose beads alone. After three washes with RIPA buffer, the beads were suspended in SDS sample buffer, boiled for 5 minutes, and removed by centrifugation, and the supernatant proteins were separated by SDS-PAGE, as described above.

CLL low-density cells were cytospun on poly-l-lysine–coated slides and fixed in 3.7% formaldehyde for 15 minutes at room temperature on a shaker. The slides were washed in PBS, incubated with 1% Triton X-100 for 5 minutes, and blocked with goat serum for 1 hour. After blocking, the slides were washed and incubated overnight with phycoerythrin (PE)-conjugated mouse anti-human GM-CSFRα antibodies (BD Bioscience). The slides were then washed in PBS and incubated with the nuclear acid stain TOPRO-3 (Invitrogen). In other experiments, cells were incubated with mouse anti-S6 ribosomal protein (Calbiochem/Millipore) overnight, extensively washed in PBS, and then incubated with PE rabbit anti-mouse antibody (BD Biosciences) for 1 hour. The slides were then mounted with VECTASHIELD HardSet (Vector Laboratories) and viewed using an Olympus FluoView FV500 Confocal Laser Scanning Microscope (Olympus).

Nuclear and cytoplasmic extracts were prepared using the NE-PER Nuclear and Cytoplasmic Extraction Kit (Thermo Scientific/Pierce) as previously described (22). To confirm that pure extracts were obtained, monoclonal mouse anti-human lamin B and mouse anti-human S6 ribosomal protein antibodies were used (Millipore/Calbiochem).

A 4.0-kb fragment of the human GM-CSFRα promoter was generated by PCR using, as a template, genomic DNA isolated from peripheral blood low-density cells that were obtained from 2 patients with CLL. The DNA was purified via the Wizard DNA Purification Kit (Promega). Human GM-CSFRα promoter PCR primers were designed in accordance with the sequence of the GM-CSFRα–flanking region (Ensembl Resource; www.ensembl.org). The GM-CSFRα promoter PCR primers included a forward primer starting at bp-4012-CTTGGTGAACCGTCTGGTTGA (relative to the start of the first exon), a reverse primer starting at bp-23-CTTCTGAGTAGCTCCCTTCAG, and primers for several truncated constructs designed at bp-2517-GAATGGACTAAGACAGCTCCT and bp-496-TGTATGCCCATGTGCACTGTG. The amplified fragments were attached to BgIII and KpnI digestion sites of the pGL4-basic vector (Promega) containing a luciferase reporter gene. The pGL4 vectors harboring ligation products were introduced into One Shot TOP10 chemically competent Escherichia coli (Invitrogen), spread onto agar plates containing ampicillin (Invitrogen), and incubated overnight at 37°C. Single colonies were microaspirated and analyzed following restriction enzyme digestion. The sequences of all constructs were verified by automated sequencing (SeqWright).

MM1 cells were transfected by electroporation with the Gene Pulser Xcell Electroporation System (Bio-Rad) as previously described (23). A total of 5 × 10⁶ cells, suspended in 0.2 mL of serum-free OPTI-MEM medium (Invitrogen), were transfected with 2 μg of each of the specific reporter constructs. To assess luciferase activity, we used the Dual-Luciferase Reporter Assay System (Promega) and the SIRIUS luminometer V3.1 (Berthold Detection Systems) 24 hours after transfection. We determined the luciferase activity by calculating the constructs' luciferase activity relative to the activity of the Renilla reniformis luciferase produced by the pRL-SV40 control vector. The primers used in these experiments are listed in Supplementary Table S2.

RNA was isolated using an RNeasy purification procedure (Qiagen) and its quality and concentration were analyzed with a NanoDrop spectrophotometer (ND-1000; Thermo Scientific/NanoDrop Technologies). We used 500 ng of total RNA in One-Step qRT-PCR (Applied Biosystems). The sequence detection system ABI Prism 7700 (Applied Biosystems) used the TaqMan gene expression assay for ROR1, STAT3, Bcl2, Bcl-X_L, c-Myc, p21 (WAF1), cyclin D1, and 18S according to the manufacturer's instructions. We ran the samples in triplicate and used the comparative C_T method for relative quantification (22).

We used a SimpleChIP Enzymatic Chromatin IP Kit (Cell Signaling Technology) in accordance with the manufacturer's instructions. Briefly, cells were cross-linked with 1% formaldehyde for 10 minutes at room temperature and then harvested and incubated on ice for 10 minutes in lysis buffer. Nuclei were pelleted and digested by micrococcal nuclease. Following sonication and centrifugation, sheared chromatin was incubated with anti-STAT3 or rabbit serum (negative control) overnight at 4°C. Then, protein-G beads were added and the chromatin was incubated for 2 hours in rotation. An aliquot of chromatin that was not incubated with an antibody was used as the “input” control sample. Antibody-bound protein–DNA complexes were eluted and subjected to PCR analysis. The primer sets used to amplify GM-CSFRα promoter sites are listed in Supplementary Table S2. The primers to amplify the human STAT3 promoter were forward, 5′-CCGAACGAGCTGGCCTTTCAT-3 and reverse, 5′-GGATTGGCTGAAGGGGCTGTA-3, which generated an 86-bp product; primers to amplify the WAF/p21 promoter were forward, 5′-TTGTGCCACTGCTGACTTTGTC-3 and reverse, 5′-CCTCACATCCTCCTTCTTCAGGCT-3, which generated a 303-bp product; and primers to amplify the c-Myc promoter were forward, 5′-TGAGTATAAAAGCCGGTTTTC-3 and reverse, 5′-AGTAATTCCAGCGAGAGGCAG-3, which generated a 63-bp product. The human RPL30 gene primers were provided by Cell Signaling Technology.

Nuclear extracts were prepared using an NE-PER Nuclear and Cytoplasmic Extraction kit (Thermo Scientific/Pierce). Two micrograms of nuclear protein extracts was incubated with biotin-labeled GM-CSFRα promoter DNA probes in binding buffer for 30 minutes on ice. All probes were synthesized by Sigma-Genosys. The sequences of the probes used are listed in Supplementary Table S2. Following incubation, the samples were separated on a 5% polyacrylamide gel in Tris-borate EDTA, transferred onto a nylon membrane, and fixed on the membrane by UV cross-linking. The biotin-labeled probe was detected with streptavidin horseradish peroxidase (EMSA Gel-Shift Kit; Affymetrix/Panomics). A 7-fold excess of unlabeled cold probes combined with biotin-labeled probes was used for competition control. To determine the effect of antibodies on protein–DNA binding, 1 μg of monoclonal mouse anti-human STAT3 (BD Biosciences; Cell Signaling Technology) or mouse anti-human phosphoserine STAT3 (Cell Signaling Technology) antibody was incubated with the nuclear extracts for 30 minutes on ice before adding the biotin-labeled DNA probe. The isotypic controls consisted of mouse IgG1 (BD Biosciences).

MM1 and CLL cells were transfected by electroporation as described above. Knockdown of endogenous STAT3 was performed using the predesigned siRNA and scrambled siRNA from Invitrogen/Ambion. The siRNA sequences used to target exons 14 and 15 of the human STAT3 gene were antisense 5′-GGGAAGCAUCACAAUUGGCTC-3′ and sense 5′-GCCAAUUGUGAUGCUUCCCTT-3′. We mixed 30 nmol/L of siRNA with siPORT NeoFX Transfection Agent (Invitrogen/Ambion) and transfected them by electroporation into CLL cells. Transfected cells were maintained in coculture with mesenchymal stroma cells, and qRT-PCR was performed 48 hours after transfection.

HEK 293FT cells were cotransfected with GFP-lentivirus STAT3-shRNA or with GFP-lentivirus empty vector and the packaging vectors pCMVδR8.2 and pMDG (generously provided by Dr. G. Inghirami, University of Torino, Torino, Italy), using the superfect transfection reagent (Qiagen). 293FT cell culture medium was filtered through a 45-μm syringe, the lentivirus was concentrated by filtration through an Amicon Ultra centrifugal filter device (Millipore), and the concentrated supernatant was used to infect CLL cells. CLL cells (5 × 10⁶/mL) were incubated in 6-well plates (Becton Dickinson) in 2 mL of DMEM supplemented with 10% FBS and transfected with 100 μL of viral supernatant. Polybrene (10 ng/mL) was added to the viral supernatant at a ratio of 1:1,000 (v/v). Transfection efficiency was measured after 48 hours and ranged between 45% and 60% [calculated on the basis of the ratio of propidium iodide (PI)–negative/GFP-positive cells]. RNA was isolated and prepared for qRT-PCR analysis as previously described (22).

GM-CSF was found to upregulate the expression of the cell-surface receptor CD20 in CLL cells (20), and administering GM-CSF to patients with CLL appeared to improve the therapeutic efficacy of the anti-CD20 antibody rituximab (4). Therefore, we wondered whether CLL cells express the GM-CSFR and whether GM-CSF induces signal transduction in these cells. To explore whether GM-CSF activates GM-CSFR–induced signaling pathways in CLL cells (24), we incubated peripheral blood CLL cells with GM-CSF and used Western immunoblotting to assess the levels of unphosphorylated and phosphorylated STAT-3, STAT5, AKT (also known as protein kinase B), and ERK-1/2. As shown in Supplementary Fig. S1A, GM-CSF did not induce phosphorylation of STAT3 or STAT5, nor did it change the phosphorylation levels of AKT or ERK, both of which were reported to be constitutively activated in CLL cells of some, but not all, patients with CLL (25). Furthermore, unlike the process in myeloid cells, in which GM-CSF binds to the GM-CSFR and activates JAK2 (24), GM-CSF did not induce JAK2 phosphorylation in CLL cells (Supplementary Fig. S1B), and GM-CSF- or GM-CSFRα-neutralizing antibodies did not affect CLL cell-surface CD20 levels as assessed by flow cytometry (Supplementary Fig. S1C).

Because activation of the GM-CSFR depends on the interaction between the GM-CSFRα and GM-CSFRβ subunits (24), we sought to determine whether both subunits are present in CLL cells. Using flow cytometry, we found that GM-CSFRα (CD116), but not GM-CSFRβ (CD131), is present on the surface of peripheral blood CD19⁺ CLL cells (Fig. 1A, left). The myeloid cell lines OCI/AML3 and HMC860 were used as positive controls, and the T-cell Jurkat cell line was used as a negative control (Fig. 1A, right). To confirm these data, we performed PCR and a Western blot analysis of cellular extracts of CLL cells from 6 and 7 patients and of normal B cells from 2 healthy donors. GM-CSFRα transcripts were readily detected, whereas weak signals of GM-CSFRβ were detected in all CLL samples by 50-cycle PCR. GM-CSFRα, but not GM-CSFRβ, was detected in cell lysates of all CLL peripheral blood samples, whereas neither GM-CSFRα nor GM-CSFRβ was detected in normal B cells (Fig. 1B). To confirm these results, we analyzed peripheral blood low-density cells from another patient with CLL by flow cytometry. As shown in Fig. 1C, 58.7% of the cells coexpressed CD19, CD5, and CD116, suggesting that CLL cells, typically coexpressing CD19 and CD5 antigens, also express GM-CSFRα. We repeated this analysis on peripheral blood low-density cells from 5 additional, randomly selected patients with CLL. Coexpression of CD19, CD5, and CD116 was detected in 16.5%, 24.5%, 62.7%, 39.1%, and 28.9% of the 5 patients' CLL cells. Although CD116 expression was likely underestimated because of quenching (26), these data demonstrate that GM-CSFRα is expressed on the surface of CLL cells. Thus, although CLL cells express GM-CSFRα, GM-CSF does not activate the GM-CSFR, likely because the GM-CSFRβ subunit is not present in CLL cells.

To confirm these data, we stained CLL BM smears with anti-GM-CSFRα antibodies. Similar to granulocytes, several mononuclear cells (some of which exhibited typical CLL cell morphology) expressed GM-CSFRα, and GM-CSFRα–positive speckles were detected both in the cytoplasm and nucleus of these cells (Fig. 1D). To further delineate these findings, we performed confocal microscopy of peripheral blood CLL cells using GFP-conjugated mouse anti-human GM-CSFRα antibodies. Surprisingly, as in the CLL BM cells, GM-CSFRα was detected on the surface, to a lesser extent in the cytoplasm, and in the nucleus of CLL cells (Fig. 2A). To confirm this observation, we prepared cytoplasmic and nuclear protein extracts of peripheral blood low-density CLL cells. S6 ribosomal protein was used as a positive control for the cytoplasmic fraction, and lamin B1 served as a positive control for the nuclear fraction. As shown in Fig. 2C, GM-CSFRα was detected mainly in the nucleus of CLL cells obtained from 2 different patients.

Unlike normal B lymphocytes (24), CLL cells produce GM-CSFRα, so we sought to determine what induces GM-CSFRα production in CLL cells. Recently, we found that CLL STAT3 is constitutively phosphorylated on serine 727 residues and activates a variety of genes (22). Typically, phosphorylated STAT3 binds to the γ-interferon (IFN) activation sequence (GAS)–like element, also referred to as the sis-inducible element, located in the promoter region of various genes (27). Therefore, we performed a sequence analysis of the GM-CSFRα promoter and identified GAS-like elements, representing putative STAT3-binding sites (Supplementary Fig. S2). We used MM1 as an in vitro model to test whether STAT3 induces GM-CSFRα (23). Unstimulated MM1 cells expressed low to undetectable levels of phosphotyrosine-STAT3, and exposure of MM1 cells to IL6 induced STAT3 phosphorylation (28). We incubated MM1 cells with IL6 and, using Western immunoblotting, established that IL6 induced upregulation of both pSTAT3 and GM-CSFRα in a dose- and time-dependent manner (Fig. 3A).

Then, using the MM1-inducible system, we cloned the proximal region of the human GM-CSFRα promoter (4.0 kb). We generated a series of GM-CSFRα–truncated constructs (Fig. 3B) and inserted them into a plasmid carrying a luciferase reporter. The location of the 5′ regions of these constructs is depicted in Supplementary Fig. S2, and the primers used to amplify them are listed in Supplementary Table S2. All constructs were transfected into MM1 cells (transfection efficiency, 45%–50%; data not shown); following incubation of the constructs either with 20 ng/mL of IL6 or with culture medium, we used a luciferase reporter assay and the Renilla reniformis system to assess their luciferase activity. Compared with results in control samples, exposure of MM1 cells to IL6 significantly increased the luciferase activity of the truncated GM-CSFRα −4,000 (spanning from −4,012 to +23 bp), GM-CSFRα −3,000 (spanning from −2,965 to +23 bp), GM-CSFRα −2,500 (spanning from −2,517 to +23 bp), and to the same extent GM-CSFRα −500 (spanning from −496 to +23 bp). IL6 did not affect the luciferase activity of PGL4.17 (Fig. 3B). Among the serially truncated clones, GM-CSFRα −4,000 exerted the highest, and GM-CSFRα −2,500 and −500, the lowest IL6-induced promoter activity (Fig. 3B).

Because these data suggested that STAT3 activates the GM-CSFRα gene, we used chromatin immunoprecipitation (ChIP) to determine whether STAT3 indeed binds to GAS-like elements within the GM-CSFRα promoter. We obtained nuclear extracts of unstimulated and IL6-stimulated MM1 cells and used ChIP and primers deduced from the luciferase activity data (Supplementary Table S2) to assess the binding of STAT3 to five putative STAT3-binding sites (Supplementary Fig. S3). We found that primers 1, 3, and 5, but not primers 2 or 4, corresponding to the putative STAT3-binding sites depicted in Supplementary Fig. S3, amplified DNA fragments that coimmunoprecipitated with anti-STAT3 antibodies (Fig. 3C, left). These results suggest that STAT3 binds to sites 1, 3, and 5 of the GM-CSFRα promoter, shown at the bottom of Fig. 3C. In addition, we found that anti-STAT3 antibodies coimmunoprecipitated DNA of the STAT3-regulated genes STAT3, c-Myc, and Waf1/P21, but not of the control gene RPL30 (Fig. 3C, right), as previously reported (23). Then, to validate these findings, we performed electoromobility shift assay (EMSA). As shown in Fig. 3D, MM1 nuclear protein bound the biotinylated GM-CSFRα promoter DNA fragments of regions 1, 3, and 5 (Supplementary Table S2), and excess cold probes or anti-STAT3 antibodies attenuated this binding. In contrast, the binding of MM1 nuclear protein to the biotinylated mutant DNA fragments of regions 1, 3, or 5 (Supplementary Table S2) was diminished or significantly reduced, suggesting that activated STAT3 binds to the binding sites identified by our previous experiments.

Next, we sought to confirm that STAT3 activates the GM-CSFRα gene. We transfected MM1 cells with STAT3-siRNA, stimulated the cells with IL6, and quantitated mRNA levels of GM-CSFRα, STAT3, and various STAT3-regulated genes by using relative qRT-PCR. As shown in Fig. 3E, STAT3-siRNA downregulated mRNA levels of GM-CSFRα and the STAT3-regulated genes STAT3, Bcl-2, Bcl-X_L, cyclin D1, c-Myc, and Waf1/p21. Furthermore, STAT3-siRNA, but not scrambled siRNA or GAPDH, downregulated STAT3 protein levels by 60% and GM-CSFRα protein levels by 70%, suggesting that STAT3 activates the GM-CSFRα gene and induces production of GM-CSFRα protein by MM1 cells (Fig. 3F).

We used IL6 to induce STAT3 phosphorylation in MM1 cells. However, in CLL cells, STAT3 is constitutively phosphorylated (22). We used ChIP to determine whether STAT3 activates GM-CSFRα in CLL cells. As shown in the left panel of Fig. 4A, primers 1, 3, and 5, but not primers 2 or 4 (Supplementary Table S2), bound to CLL cell DNA fragments that were coimmunoprecipitated with anti-STAT3 antibodies (Supplementary Table S2 and Supplementary Fig. S3), suggesting that, as in MM1 cells, in CLL cells STAT3 bound the STAT3-binding sites 1, 3, and 5. In addition to coimmunoprecipitating DNA of GM-CSFRα, STAT3 coimmunoprecipitated DNA of the STAT3-regulated genes STAT3, c-Myc, p21/Waf1, ROR1, and VEGF, but not the control gene RPL30 (Fig. 4A, right).

To validate these findings, we performed EMSA. As shown in Fig. 4B, the biotinylated DNA fragments of probes 1, 3, and 5 bound nuclear protein obtained from peripheral blood low-density CLL cells of 2 patients with CLL, and excess cold probes or anti-STAT3 antibodies attenuated protein–DNA binding. Conversely, the binding of CLL nuclear protein to the mutated DNA biotinylated fragments (Supplementary Table S2) was diminished or significantly reduced, suggesting that, as in MM1 cells, activated STAT3 binds to the GM-CSFRα promoter–binding sites 1, 3, and 5.

Finally, to confirm that STAT3 activates the GM-CSFRα promoter, peripheral blood low-density CLL cells were infected with lentiviral STAT3-shRNA (infection efficiency, 40%), and STAT3-regulated gene mRNA levels were quantitated by relative qRT-PCR. As shown in Fig. 4C, infection with STAT3-shRNA reduced mRNA levels of GM-CSFRα and the STAT3-regulated genes STAT3, Bcl-2, Bcl-X_L, cyclin D1, c-Myc, and Waf1/p21. Furthermore, unlike the empty vector that reduced STAT3 levels by 10% and GM-CSFRα levels by 40%, STAT3-shRNA downregulated the protein levels of STAT3 by 80% and of GM-CSFRα by 90% (Fig. 4D).

The GM-CSFRα subunit is largely extracellular but contains a 54-amino acid intracytoplasmic tail (Fig. 5A). We detected GM-CSFRα on the surface, in the cytosol, and in the nucleus of CLL cells. However, we have not established whether our GM-CSFRα antibodies detected the extracellular domain exclusively or also detected the intracellular domain. Removal of the intracytoplasmic tail of the GM-CSFRα subunit was reported to deactivate the GM-CSF–induced activity of GM-CSFR (29). To reveal the biologic activity of GM-CSFRα in unstimulated cells lacking the GM-CSFRβ subunit, it was crucial to determine whether part of or the entire GM-CSFRα subunit was present in the cytosol and/or the nucleus. Therefore, GFP-tagged full-length GM-CSFRα (Supplementary Table S2; amino acids 1–400) and its extracellular (amino acids 1–323) and intracellular (amino acids 347–400) regions (Fig. 5A) were transfected into 293FT cells. The protein of transfected 293FT cells was immunoprecipitated with anti-GFP antibodies and, as shown in Fig. 5B–D, full-length GM-CSFRα and its extracellular and intracellular domains were detected in total cell extract as well as in the cytosol and the nucleus. However, these anti-GM-CSFRα antibodies bound the extracellular, not the intracellular, GM-CSFRα domain (Fig. 5E), and therefore detected both the full-length GM-CSFRα receptor and its extracellular domain in the cytosol and the nucleus (Fig. 5F) of GM-CSFRα–transfected 293FT cells. Taken together, these data suggest that our anti-GM-CSFRα antibodies detected the extracellular domain of GM-CSFRα and that the entire GM-CSFRα subunit migrated to the cytosol and the nucleus.

The function and biologic effects of GM-CSF–stimulated GM-CSFR following ligand binding and dimerization of its GM-CSFRα and GM-CSFRβ subunits have been well established (24). However, little is known about the function of GM-CSFRα in the absence of the GM-CSFRβ subunit. To explore the role of GM-CSFRα, we transfected GFP or the GFP-tagged GM-CSFRα gene into 293FT cells, and, following overnight incubation, cell extract was immunoprecipitated with anti-GFP antibody-coated agarose beads. Both RNA and protein were extracted from the immunoprecipitate and analyzed.

The one-way ANOVA program identified 7,220 RNA transcripts that coimmunoprecipitated with GFP-tagged GM-CSFRα and 1,593 RNA transcripts that coimmunoprecipitated with GFP-tagged GM-CSFRα, but not with GFP; the top 30 of these transcripts are depicted in the heat map presented in Fig. 6. The GM-CSFRα–attached RNA transcripts at the top of the list are likely involved in active translation of processed GM-CSFRα mRNA. At the top of the list were also the SKI and MAFA oncogenes, two serine/threonine kinases, and three microRNAs (Table 1). Other top-list RNA transcripts included DEFT1P2 and DEFP1P, which encode proteins that belong to members of the death effector domains (DED) family, known to have a role in the control of programmed cell death and in the regulation of a variety of signal transduction pathways (30, 31). Pathways enriched with transcripts that coimmunoprecipitated with GFP-tagged GM-CSFRα, but not with GFP, included the JAK–STAT signaling and hematopoietic lineage pathways (Table 2).

Then, we used mass spectrometry to identify GM-CSFRα–interacting proteins. In addition to various keratins, heat shock proteins, and GM-CSFRα, Krüppel-associated protein [or KRAB zinc finger-associated protein 1 (KAP1)], also known as transcription intermediary factor (TIF)-1β or tripartite motif-containing protein (TRIM; refs. 28, 32) and IFN-stimulated gene (ISG)-15 (also known as TRIM25) were pulled down together with GM-CSFRα (Supplementary Fig. S4).

Because KAP1-siRNA was found to upregulate STAT3 transcription and induce accumulation of phosphoserine STAT3 in the nucleus (33) and because STAT3 is constitutively phosphorylated on serine 727 residues in CLL cells (22), we sought to explore how GM-CSFRα affects the transcription of STAT3. Using immunoprecipitation with anti-GM-CSFRα antibodies, we confirmed that KAP1 coimmunoprecipitated with GM-CSFRα (Fig. 7A). Then, to determine how GM-CSFRα affects STAT3–DNA binding, we performed EMSA. As shown in Fig. 7B, we found that transfection of CLL cells with GM-CSFRα-siRNA reduced CLL nuclear protein–DNA binding, whereas anti-STAT3 antibodies completely prevented the binding, as previously reported (22). Taken together, these data suggest that GM-CSFRα binds to the negative regulator of STAT3 transcription KAP1, and as a result, enhances the transcriptional activity of STAT3.

The covalent binding of ISG15 to various proteins, known as ISGylation, negatively controls the NF-κB pathway (34); because NF-κB is constitutively activated in CLL (35) and because ISG15 binds to and coimmunoprecipitates with GM-CSFRα (Fig. 7C), we wondered how GM-CSFRα affects the activation of NF-κB in CLL cells. We transfected CLL cells with GM-CSFRα-siRNA and, by using Western immunoblotting, found that GM-CSFRα-siRNA downregulated the levels of IκB (Fig. 7C). Because ubiquitination of IκB enables NF-κB to shuttle to the nucleus and bind to DNA, we performed EMSA. We found that transfection of CLL cells with GM-CSFRα-siRNA enhances NF-κB–DNA binding (Fig. 7D), suggesting that, as in other proteins (34), GM-CSFRα ISGylation inhibits the NF-κB pathway.

As described above, we found that GM-CSFRα inversely affects two transcription factors known to be major regulators of cell survival (36) by enhancing the activity of STAT3 and contributing to the inhibition of NF-κB. To decipher the net effect of GM-CSFRα, we transfected peripheral blood low-density CLL cells with GM-CSFRα-siRNA and assessed the cellular apoptosis rate using PI and Annexin V. At a transfection rate of 33.5% to 55% (Fig. 7E), GM-CSFRα-siRNA significantly downregulated GM-CSFRα mRNA and protein levels (Fig. 7F) and induced CLL cell apoptosis (Fig. 7G), suggesting that GM-CSFRα exerts an overall antiapoptotic effect. CLL cells from four randomly selected patients were transfected with GM-CSFRα-siRNA and the rate of cellular apoptosis was assessed by Annexin V/PI staining using flow cytometry. The percentage of apoptosis in GM-CSFRα-siRNA–transfected cells was higher than that of GAPDH-transfected cells by 20.4%, 44.6%, 18.7%, and 12.3%, respectively (P = 0.036, paired t test). Figure 7F depicts results of patients 1 and 2 while results of patients 3 and 4 are not shown. To confirm that GM-CSFRα provides CLL cells with a survival advantage, we transfected MM1 cells with GFP or GM-CSFRα and incubated them for 48 hours with or without 20 μmol/L of dexamethasone, an agent known to induce apoptosis in MM1 cells (37). As shown in Fig. 7H, dexamethasone induced apoptosis at a significantly higher rate in GFP-transfected cells than in GM-CSFRα–transfected cells, suggesting that GM-CSFRα protected MM1 cells from apoptosis.

In this study we show that unlike normal B lymphocytes in CLL cells, constitutively activated STAT3 activates the GM-CSFRα promoter, and that GM-CSFRα, present on the cell surface, in the cytosol, and in the nucleus, provides the cells with survival advantage in a ligand-independent manner.

Several investigators have previously studied the expression GM-CSFR by normal and neoplastic B lymphocytes. Using sensitive triple-layer immunophenotypic techniques, Till and colleagues (21) found that both the α- and β-chains of the GM-CSFR are expressed on hairy cells and myelomatous plasma cells but not on CLL or prolymphocytic leukemia (PLL) lymphocytes. In that study, the GM-CSFR was demonstrable on normal plasma cells in tonsils but not on either activated or resting tonsillar B cells or on circulating normal B lymphocytes. Our data confirm that GM-CSFR protein is not present in normal B lymphocytes; however, we found that CLL cells express the α, but not the β, subunit of GM-CSFR. Like other investigators (38), we found that CLL cells express GM-CSFRα mRNA. We also detected GM-CSFRα, but not GM-CSFRβ, protein on the cell surface, in the cytosol, and in the nucleus of CLL cells. B1a cells were recently found to produce GM-CSF (5); however, GM-CSF does not affect any subtype of normal B lymphocytes that lack the GM-CSFR. Although CLL cells are thought to be B1 cells, we found that GM-CSF did not elicit signal transduction in B-cell CLL cells, probably because CLL cells lack the β subunit of the GM-CSFR that associates with JAK2 to initiate signal transduction (17).

We detected the full-length GM-CSFRα in the cytosol and nucleus of CLL cells. Nuclear localization of cell-surface receptors has been reported by several investigators. Direct nuclear translocation of full-length growth factor receptors or fragments of them has been described in other cell types (39, 40), and importin-β and the nuclear pore protein Nup358 (40) were found to be involved in the nuclear transport process. Whether the karyopherin or another transport system is involved in shuttling GM-CSFRα to the nuclei of CLL cells, remains to be determined.

GM-CSFRα has been detected in various tissues, and ligation of GM-CSFRα in the absence of the β subunit (41) was reported to elicit effects in other nonhematopoietic cell lineages. Prostate carcinoma cells express GM-CSFRα (42), and GM-CSFRα is distributed throughout the CNS, predominantly in neuronal cells (43). A reduction in GM-CSFRα protein levels in the nuclei of the hippocampus, cortex, thalamus, and brainstem is thought to contribute to neurodegeneration in Alzheimer's disease (44). However, the factors responsible for GM-CSFRα expression in these cells have not been identified.

Unlike in normal B lymphocytes, STAT3 is constitutively phosphorylated on serine 727 residues in CLL cells (22, 45) and is biologically active (22). Therefore, we sought to determine whether STAT3 affects the expression and protein production of GM-CSFRα in CLL cells. We performed a sequence analysis of the GM-CSFRα promoter and identified GAS-like elements, suggesting that STAT3 binds to the GM-CSFRα gene promoter in CLL cells. In MM1 cells, IL6 induced both STAT3 and GM-CSFRα protein expression. Therefore, we used a luciferase assay to clone the GM-CSFRα promoter and used ChIP and EMSA to identify STAT3-binding sites in both MM1 and CLL cells. Transfection of MM1 and primary CLL cells with STAT3-siRNA or STAT3-shRNA downregulated GM-CSFRα mRNA and protein levels, confirming that STAT3 binds to the GM-CSFRα promoter and activates the GM-CSFRα gene in CLL cells.

In human embryos, GM-CSF was found to regulate cell viability through a mechanism independent of the β subunit (46). However, the role of GM-CSFRα in CLL cells is unknown. GM-CSFRα does not have a DNA-binding site, and the function GM-CSFRα fulfills in the nucleus and/or cytosol is unknown. Several investigators demonstrated that mRNAs are coregulated by one or more sequence-specific RNA-binding proteins that orchestrate their splicing, export, stability, localization, and translation and that posttranscriptional processes depend largely on the functions of RNA-binding proteins (reviewed in ref. 47). RNA transcripts that coimmunoprecipitated with GM-CSFRα are those involved in GM-CSFRα translation of oncogenes (SKI and MAFA), of serine/threonine kinases, of microRNAs, and of members of the DED family (30, 31), suggesting that GM-CSFRα is involved in regulating cell survival pathways.

Indeed, the proteins that coimmunoprecipitated with GM-CSFRα, KAP1 and ISG15, regulate cellular viability. KAP1 is thought to regulate the dynamic organization of chromatin structure by modifying epigenetic patterns and chromatin compaction (32); KAP1 is a transcriptional regulator of the IL6/STAT3 signaling pathway, as KAP1 siRNA enhanced IL6-induced STAT3-dependent transcription and gene expression (33). Furthermore, reduction of KAP1 expression resulted in the marked accumulation of phosphoserine STAT3 in the nucleus, a modification that regulates its transcriptional activation (33). We found that GM-CSFRα binds to KAP1 and enhances the transcriptional activity of STAT3. In a previous report, we showed that STAT3 provides CLL cells with survival advantage (22). Our current data suggest that GM-CSFRα enhances this effect.

IFNs promote the activation of a genetic program that controls the expression of hundreds of genes named ISGs. To the ISGs family belong all the components of the molecular machinery that modifies proteins by the addition of the ubiquitin-like protein ISG15. This process of covalent conjugation to different proteins is known as ISGylation. Expression of the ISGylation system suppressed NF-κB activation via TRAF6/TAK1 and reduced the level of polyubiquitinated TRAF6, suggesting that ISGylation negatively controls the NF-κB pathway (34). However, ISG15 upregulation was also associated with telomere shortening. Apparently, this regulation is unrelated to IFNs, TP53, and the DNA damage/senescence response. Overall, the ISGylation system and ISG15 in particular could provide some advantages to cancer cells, because a negative selection was rarely observed (48). We found that similar to other proteins (34), GM-CSFRα ISGylation inhibited the NF-κB pathway. The NF-κB pathway provides CLL cells with prosurvival signals (49). Nevertheless, overexpression of GM-CSFRα protected MM1 cells from dexamethasone-induced apoptosis, whereas transfection of CLL cells with GM-CSFRα-siRNA induced CLL cell apoptosis, suggesting that the STAT3-induced prosurvival effects overcome reduced ubiquitination of IκB. We have previously reported that in CLL cells unphosphorylated STAT3 activates NF-κB in an IκB-independent manner (35). Therefore, as in neuronal (43) and embryonic (46) cells, GM-CSFRα provides CLL cells with a survival advantage.

Taken together, our data suggest that the α subunit of the GM-CSFR possesses GM-CSF–unrelated biologic activities that protect CLL cells from apoptosis. Whether selective targeting of GM-CSFRα might be a useful therapeutic strategy in CLL remains to be determined.

No potential conflicts of interest were disclosed.

Conception and design: A. Ferrajoli, Y. Wang, C. Bueso-Ramos, I. Hazan-Halevy, J. Burger, Z. Estrov

Development of methodology: P. Li, D. Harris, Z. Liu, A. Ferrajoli, Y. Wang, C. Bueso-Ramos, S. Grgurevic, Z. Estrov

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P. Li, D. Harris, A. Ferrajoli, Y. Wang, W. Wierda, S. Faderl, M. Keating

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): U. Rozovski, A. Ferrajoli, Y. Wang, S. O'Brien, Z. Estrov

Writing, review, and/or revision of the manuscript: D. Harris, U. Rozovski, A. Ferrajoli, W. Wierda, J. Burger, S. O'Brien, S. Faderl, M. Keating, Z. Estrov

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P. Li, D. Harris, Y. Wang, S. O'Brien, M. Keating

Study supervision: Z. Estrov

The authors thank Jill Delsigne and Diane S. Hackett for editing our article.

This study was supported by a grant from the CLL Global Research Foundation and the NIH through MD Anderson's Cancer Center Support Grant CA016672.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.