A Truncated Isoform of the Human β Chain Common to the Receptors for Granulocyte-Macrophage Colony-Stimulating Factor, Interleukin-3 (IL-3), and IL-5 With Increased mRNA Expression in Some Patients With Acute Leukemia

THE PROLIFERATION, differentiation, and cellular functions of hematopoietic cells are regulated by the interaction of a number of different cytokines with the receptors expressed on the cell surface.1 Heterogeneity in the cytokine receptor family is provided by the ligand binding–induced oligomerization of their component chains. The erythropoietin receptor (EpoR) and granulocyte colony-stimulating factor (G-CSF) receptor (G-CSFR) are probably homodimers,2,3 the receptors for granulocyte-macrophage colony-stimulating factor (GM-CSF) (GMR), interleukin-3 (IL-3R), and IL-5 (IL-5R) are oligomers of α and β chains,4 and the IL-2-R involves three different chains.5 A common characteristic appears to be the sharing of one “β-like” receptor chain with several different “α-like” chains, for example, the specific α chains for GM-CSF, IL-3, and IL-5 all bind ligand with low affinity and complex to form high-affinity receptors with the common β chain (β_C), which itself does not bind ligand.4Similarly, gp130 can bind to the IL-6, IL-11, ciliary neurotrophic factor, leukemia inhibitory factor, and oncostatin M receptor chains, and the IL-2Rγ chain binds to the IL-4, IL-7, IL-9, and IL-15 receptor chains.6 

Another form of diversity reported for a number of the receptors is the demonstration of different isoforms for a single chain. Several isoforms of the human GMRα chain have been reported from either cell lines or normal tissues, mostly affecting the 3′ end of the molecule. Some are soluble receptors lacking the transmembrane domain,7-9 another chain has an altered cytoplasmic domain,10 and one isoform has an insertion just 5′ of the transmembrane domain.11 Three isoforms have also been described affecting the 5′ end of the chain. Two of them have changes in the 5′ untranslated region and would not alter the mature protein, although they may affect translational efficiency,12 and the other isoform completely removes the signal peptide sequence.13 The genomic structure of the GMRα chain indicates that most of these different isoforms arise from alternative splicing of mRNA.14 These isoforms may have different functional consequences, as suggested by an isoform of the EpoR with a truncated cytoplasmic region that was found to be the predominant form in early-stage but not late-stage erythroid progenitor cells.15 It was unable to transduce a mitogenic signal and had a dominant-negative effect over the wild-type receptor.16 However, only one form of the full-length human GM-CSF/IL-3/IL-5 receptor β_C chain has been reported.17 

The β_C chain is not only required for high-affinity binding of ligands, but is also crucial for the signaling of downstream pathways through its intracytoplasmic tail.18 Although there are no known kinase consensus sequences within the 432 amino acid residues of the intracytoplasmic tail, a series of truncated mutants have shown that it contains distinct functional domains.18-20 In common with other receptors, the membrane-proximal region of β_C appears to be both essential and sufficient for proliferation.18,21 This region is known to bind the family of JAK kinases, and activation of JAK2 by tyrosine phosphorylation correlates closely with the induction of mitogenesis.22,23 It is also essential for the induction of c-myc, pim-1, and cis.^19,24Membrane-distal domains have been identified that are responsible for the major tyrosine phosphorylation of proteins, induction of c-fos and c-jun transcription, activation of theRas pathway, and prevention of apoptosis.18,19,25In addition, the membrane-distal region has been associated with the induction of differentiation20 and negative regulation of receptor signals.26 

We describe here an isoform of the β_C chain with deletion of a 104-bp exon just 3′ of the transmembrane region. This resulted in a truncated receptor chain with an intracytoplasmic tail containing 23 amino acids in common with β_C and then an altered C-terminal tail of a further 23 amino acids. JAK2 bound to this isoform and was tyrosine-phosphorylated by stimulation with GM-CSF; however, it did not transduce a mitogenic signal when transfected into CTLL-2 cells. In primary myeloid cells and myeloid cell lines, transcript and protein levels comprised approximately 10% to 25% of the total β chain. Increased transcript levels were detected in blast cells from approximately 75% of patients with acute myeloid leukemia (AML), suggesting that this isoform may play a role in the pathogenesis of the disease.

The hematopoietic cell lines TF-1,27 HL60,28U937,29 and CTLL-230 were grown in RPMI 1640 supplemented with 10% fetal calf serum (FCS), plus 5 ng/mL GM-CSF for TF-1 cells and 30 U/mL IL-2 for CTLL cells. Neutrophils from normal healthy volunteers, mononuclear cells from normal human bone marrow, and leukemic blast cells from patients at presentation with AML and acute lymphoblastic leukemia (ALL) were all prepared by density-gradient centrifugation (Nycomed, Oslo, Norway). CD34⁺ cells were prepared using a Ceprate SC column (CellPro, Bothell, WA) and were 90% pure. Blast cells (50 × 10⁶) from AML patients were cultured for 9 or 10 days in Iscove's medium supplemented with 20% FCS, 10 ng/mL IL-3, 10 ng/mL GM-CSF, and 100 ng/mL G-CSF. Cytospins were prepared before and after culture, and morphology was examined using May-Grünwald-Giemsa staining.

Total cellular RNA was extracted using a standard method of guanidinium isothiocyanate lysis and ultracentrifugation through cesium chloride.31 

One microgram of total RNA was reverse-transcribed using 250 ng oligo-dT as a primer (Promega, Madison, WI) in a total volume of 20 μL containing 1× Taq polymerase buffer, 5.25 mmol/L MgCl₂, 1 mmol/L each dNTP, 20 U RNAse inhibitor, and 3 to 5 U AMV reverse transcriptase (RT; Promega). Reactions were incubated at 42°C for 1 hour and then at 95°C for 5 minutes. A 4-μL RT reaction was used for polymerase chain reaction (PCR) in a total volume of 20 μL containing 1× Taq polymerase buffer, 2.25 mmol/L MgCl₂, and 40 ng of each primer designed to amplify a fragment of 339 bp between nucleotides 1281 and 1620 flanking the transmembrane region of the β_C chain (Table 1, primers 1 and 2). This mixture was heated at 95°C for 5 minutes and held at 85°C while 1 U Taq polymerase was added, and then 35 cycles of 95°C for 30 seconds, 64°C for 30 seconds, and 72°C for 1 minute were performed. The final extension step was 72°C for 5 minutes.

For longer-length RT-PCR, 1 to 2 μg total RNA from TF-1 cells was heated at 65°C for 5 minutes with 250 ng oligo-dT, cooled at room temperature for 10 minutes, and then reverse-transcribed in a total volume of 20 μL containing 2 μL 10× Stratascript buffer, 10 mmol/L DTT, 1 mmol/L of each dNTP, 20 U RNAse inhibitor, and 50 U Stratascript RNAse H minus reverse transcriptase (Stratagene, La Jolla, CA). Reactions were incubated at 37°C for 1 hour and then at 90°C for 5 minutes. PCR was performed on a 4-μL RT reaction as described earlier using primers 2 and 10 (Table 1), which cover the extracellular portion, transmembrane region, and membrane-proximal intracytoplasmic tail of the β chain, and the extension time was 2 minutes at 72°C for each cycle. The PCR product was electrophoresed through 1% low–melting-point agarose, the fragment around 1.5 to 1.6 kb was cut out and melted, and 5 μL was placed in a fresh 50-μL PCR using primers 1 and 2 (Table 1).

Semiquantitative RT-PCR using primers 1 and 2 was performed as described earlier except that an [α-³²P] end-labeled primer was incorporated and the number of cycles was reduced to 25. PCR products were separated through 6% polyacrylamide (19:1 acrylamide:bis-acrylamide) in 1× TBE, and the gels were dried and exposed to Hyperfilm-MP (Amersham, Bucks, UK). Signals were quantified using a Hoefer densitometer (Hoefer Scientific Instruments, San Francisco, CA).

One microgram of genomic DNA prepared from TF-1 cells was used as template in a 100-μL hot-start PCR as described earlier except that the extension time at 72°C was lengthened to 3 minutes and combinations of primers 1 to 9 (Table 1) were used. Products were sequenced using either the chain-termination method with modified T7 DNA polymerase (Sequenase Version 2.0; Amersham, Bucks, UK) or thefmol DNA sequencing kit with kinase-labeled primers (Promega).

For RNA probes, a 339-bp PCR product was prepared from the KH97 GMRβ chain cDNA clone kindly provided by Dr Miyajima17 using primers 1 and 2 (Table 1), ie, nucleotides 1281 to 1620, and subcloned into the pGEM-T vector (Promega). Rsa I digestion of a clone with the insert in the reverse orientation produced a 1370-bp fragment containing the T7 RNA polymerase promoter and GMRβ sequence from nucleotides 1429 to 1620. Antisense probe was synthesized in a total volume of 20 μL containing 500 ng template, 1× transcription buffer (Promega), 10 mmol/L DTT, 500 μmol/L each for ATP, GTP, and TTP, 50 μCi [α-³²P]CTP (>400 Ci/mmol; Amersham), 20 to 40 U RNAse inhibitor, and 10 to 20 U T7 RNA polymerase (Promega). The mixture was incubated at 37°C for 90 minutes, and then 5 U RQ DNAse 1 (Promega) was added and incubation continued for 30 minutes. The reaction was phenol/chloroform-extracted, ethanol-precipitated, and dissolved in 100 μL TES (10 mmol/L Tris, pH 7.5, 1 μmol/L EDTA, and 0.1% sodium dodecyl sulfate [SDS]). Sense probe was similarly transcribed using T7 polymerase from a 1,465-bp fragment obtained byDde I digestion of a clone containing the 339-bp insert in the forward orientation.

For the protection assays, between 10 and 25 μg total cellular RNA was freeze-dried and resuspended in 25 μL hybridization buffer (80% deionized formamide, 40 mmol/L PIPES, pH 6.4, 1 mmol/L EDTA, 400 mmol/L NaCl, and 0.1% SDS). Labeled probe (0.5 to 1 μL) was added, and the mixture was heated at 90°C for 10 minutes and incubated at 50°C overnight. Samples were nucleased in 10 mmol/L Tris, pH 7.5, 5 mmol/L EDTA, 300 mmol/L NaCl, 14 μg RNAse A, and 3,000 U RNAse T1 (Boehringer Mannheim, Lewes, UK) for 60 minutes at 30°C and then incubated with 100 μg proteinase K (Boehringer) and 10 μL 20% SDS for 30 minutes at 37°C. The reactions were extracted with phenol/chloroform, and the RNA was precipitated with ethanol, washed with 70% ethanol, and resuspended in loading buffer. Protected fragments were resolved on 6% denaturing polyacrylamide gels (19:1 acrylamide:bis-acrylamide, 7 mmol/L urea) in 0.5× TBE. The gels were dried and exposed either to preflashed XAR-5 film (Kodak, Rochester, NY) at −80°C for 1 to 5 days or to phosphorimager plates and evaluated on a Fujimax bas 1000 phosphorimager (Fuji Photo Film Co, Tokyo, Japan). The signals were quantified using Millipore Whole Band Analyzer software (Millipore, Watford, UK) on a Sun Sparc workstation and corrected for GTP content to obtain the relative percentage of the different isoforms.

One microgram of RNA from TF-1 cells was reverse-transcribed as described earlier, and PCR was performed using primers 1 and 9 (Table1) to amplify fragments of 529 bp (from the β_C chain) and 425 bp (β_IT) covering nucleotides 1281 to 1810 of the published GMRβ sequence.17 DNA from the KH97 GMRβ clone and the 425-bp PCR fragment were both digested with BssHII andBglII, restriction enzymes that cut uniquely at nucleotides 1319 and 1705 of the GMRβ_C, respectively, producing fragments of 5,600 plus 386 bp for the clone and 282 bp for the PCR product. The 5,600-bp and 282-bp fragments were extracted from low–melting-point agarose, ligated using T4 DNA ligase (Promega), electroporated into JM109 bacteria, and selected on ampicillin. Positive clones were sequenced across the entire region inserted from the PCR fragment to check for possible Taq errors.

Approximately 8 × 10⁶ COS-7 monkey cells grown to semiconfluence in Dulbecco's modified medium with sodium pyruvate, 1,000 mg/L glucose, and 10% FCS were electroporated with either water, 10 μg GMRα + 10 μg GMRβ_C DNA (αβ_C), or 10 μg GMRα + 10 μg GMRβ_IT DNA (αβ_IT) using a Gene Pulser (BioRad, Hercules, CA) with capacitance 500 μFD, 0.4 kV, and then cultured for 72 hours in Costar plates or petri dishes. The GMRα construct was obtained by RT-PCR of TF-1 RNA using experimental conditions as described earlier with primers designed to amplify the full-length cDNA sequence32(5′-GTAGAACCCTGTACGTGCTT-3′ and 5′-AGAAAACAGTTCCCCCGTGT-3′) and an annealing temperature of 62°C. The PCR product was ligated into the vector pCR1000 (Invitrogen, San Diego, CA) and then subcloned into the expression vector pSVL (Pharmacia, St Albans, UK) by blunt-ending the HindIII/SacII fragment with T4 DNA polymerase.

For transfection into CTLL cells, the GMRα, β_C, and β_IT chains were subcloned into the vector pcDNA3 (Invitrogen) and linearized using Sca I digestion. Approximately 8 × 10⁶ CTLL cells were transfected with either 25 μg vector DNA (V), 25 μg GMRα DNA (α), 25 μg GMRα + 25 μg GMRβ_C DNA (αβ_C ), or 25 μg GMRα + 25 μg GMRβ_IT DNA (αβ_IT) by electroporation at 960 μFD, 0.35 kV and then selected in 1 mg/mL G418 (GIBCO-BRL, Paisley, Scotland). Expression of transcripts from the transfected GMR constructs was checked by RT-PCR. Individual clones were plucked by plating 250 transfectants/mL in Methocult (Terry Fox Laboratories, Metachem, UK) supplemented with G418 and IL-2.

[¹²⁵I]GM-CSF (human recombinant Escherichia coliproduct) with specific activity of 850 to 1,200 Ci/mmol was obtained from Amersham, and the specific activity was confirmed by analysis of maximal binding capacity and self-displacement assay. Binding was performed in situ in 12- or 24-well Costar plates. Two to 3 × 10⁵ cells were incubated with varying concentrations of¹²⁵I-GM-CSF in binding buffer (RPMI 1640, 25 mmol/L HEPES, and 2% FCS, pH 7.4) for 2 hours at 37°C. Parallel samples were incubated with greater than a 100-fold molar excess of unlabeled GM-CSF to control for nonspecific binding. After incubation, samples were washed four times in ice-cold binding buffer and then lysed with 1 mmol/L NaOH and counted. Equilibrium binding data were analyzed using the LIGAND program (Biosoft, Cambridge, UK).

Aliquots of 1 × 10⁶ CTLL cells in 50 μL binding buffer were incubated in round-bottom 96-well plates with monoclonal mouse anti-human antibodies to the GMRα chain (S20; Santa Cruz Biotechnology, Santa Cruz, CA) or extracellular GMRβ chain (DC-9; kindly provided by Dr J. Tavernier, Gent, Belgium) at a final concentration of 2.5 μg/mL in 3% BSA and 0.1% sodium azide for 1 hour on ice. The cells were then washed three times in binding buffer at 4°C and incubated with 50 μL [¹²⁵I]-labeled sheep anti-mouse F(ab′)₂ fragments (0.3 μg/mL in binding buffer; specific activity, 500 to 2,000 Ci/mmol; Amersham) for 30 minutes on ice. Cell-associated radioactivity was separated from unbound antibody by centrifuging through chilled FCS, and then these samples were snap-frozen and counted.

Rabbit polyclonal antipeptide antibodies were raised against the novel C-terminal tail of the β_IT chain and purified by caprylic acid precipitation and absorption on a peptide-specific column.33 

For cell lysates, transfected COS-7 cells were harvested with a rubber policeman in 0.5 mL lysis buffer (137 mmol/L NaC1, 20 mmol/L Tris, pH 8, 1 mmol/L MgCl₂, 1 mmol/L CaCl₂, 1% NP-40, 10% glycerol, 1 mmol/L sodium orthovanadate, 1 mmol/L β-glycerophosphate, 2 mmol/L EDTA, 1 mmol/L PMSF, and 10 μg/mL each of aprotinin, leupeptin, and pepstatin), and then an equal volume of 2× Laemmli sample buffer was added to the supernatant and the sample was boiled for 10 minutes. For immunoprecipitates, transfected COS-7 cells were incubated with or without 100 ng/mL GM-CSF (Behringwerke, Marburg, Germany) for 5 minutes at 37°C, lysed in situ with 0.5 to 1 mL ice-cold lysis buffer, and then harvested and incubated for 30 minutes on ice. After centrifugation at 12,000g for 10 minutes, the supernatant was incubated with either JAK2 polyclonal antiserum (UBI, Lake Placid, NY), the anti-β antibody DC-9, or the anti-β_IT antibody. After 2 to 14 hours, 60 μL protein A-agarose (50% in lysis buffer; Repligen, Cambridge, MA) was added, incubated end-over-end for 2 hours, washed four times in lysis buffer, and then resuspended in 50 μL 2× Laemmli sample buffer and boiled for 10 minutes. For immunoprecipitates of cell lines or primary myeloid cells, aliquots of cells were suspended in ice-cold lysis buffer at 50 × 10⁶ mL and processed as already described. Proteins were electrophoresed through SDS-7% polyacrylamide gels and transferred to nitrocellulose membrane (Hybond-C Extra; Amersham). The blots were incubated in PBS, pH 7.4, containing 3% BSA (Fraction V; Sigma, St Louis, MO) to block nonspecific binding sites and then incubated with the appropriate antibody, JAK2 (1 in 1,000), 4G10 antiphosphotyrosine antibody (1 μg/mL; UBI), DC-9 (2 μg/mL), or anti-β_IT, for 2 hours at room temperature. After washing four times in PBS/0.05% Tween-20, pH 7.4, blots were incubated with peroxidase-conjugated secondary antibody (Dakopatts, High Wycombe, UK) for 2 hours at room temperature, washed four times in PBS/0.05% Tween-20, pH 7.4, and then developed by enhanced chemiluminescence (Amersham) and analyzed by autoradiography.

Aliquots of CTLL clones were washed four times and then incubated with IL-2 or varying concentrations of GM-CSF in triplicate at 5 × 10⁴ cells/well for 72 hours. [³H]thymidine (0.5 μCi/well, 26 Ci/mmol; Amersham) was added and the incubation continued for a further 4 hours before the cells were harvested on fiberglass filters and counted.

RT-PCR of RNA from the TF-1 cell line using primers 1 and 2 (Table 1) consistently yielded two products, one of 339 bp, as expected from the published sequence for the GMRβ chain,17 and another of approximately 240 bp. The relative proportion of this smaller fragment varied with the sample analyzed: for TF-1, HL60, and U937 cell lines and primary neutrophils, it comprised approximately 10% to 25% of the total product as estimated visually, but was increased to greater than 50% of total in RNA from leukemic blasts of some patients with AML and ALL (Fig 1). Preparative PCR, purification, and sequencing of the fragment showed that there was a deletion of 104 bp between nucleotides 1491 or 1492 and 1595 or 1596. This is just 69 bp 3′ of the end of the transmembrane region in the membrane-proximal region of the intracytoplasmic tail. The deletion causes a frameshift alteration and introduces a premature stop codon, which if translated into protein would produce a β chain with a truncated intracytoplasmic tail of just 46 amino acids instead of 432, hence the designation β_IT for β intracytoplasmic truncated (Fig2A). Twenty-three amino acids of the β_IT tail would be identical to the β_C chain followed by a new sequence of 23 amino acids (Fig 2B). Other bands were sometimes seen after PCR amplification (Fig 1), but they were always minor components of the total product and were not investigated further.

To show that full-length transcripts of the truncated β chain isoform could be obtained, TF-1 RNA was reverse-transcribed using oligo-dT as a primer with Stratascript RNAse H minus reverse transcriptase (Stratagene), and PCR was performed with a primer located at the 5′ end of the extracellular portion of the published sequence (Table 1, primer 10) and primer 2, which is just 3′ of the deletion. A band of approximately 1,600 bp, as expected, for the full-length β_C was observed plus a slightly smaller band. The bands were cut out from a low–melting-point agarose gel as one fragment and used in a new PCR performed with primers 1 and 2 (Table 1). Both the β_C band (339 bp) and β_IT band (235 bp) were seen in approximately the same proportion as originally observed, indicating that full-length transcripts of the β_IT chain were present in the RNA.

PCR of genomic DNA from TF-1 cells and direct sequencing of the products showed that the 104 bp deleted in the β_ITisoform was a complete exon, with an intron of approximately 850 bp between nucleotides 1492/1493 and another intron of approximately 750 bp between nucleotides 1596/1597 (Table 2).Two other introns in the transmembrane region could also be demonstrated between nucleotides 1343/1344 (∼1,400 bp) and 1434/1435 (187 bp) (Table 2). Further PCR and Southern blot analysis of genomic DNA showed that the remainder of the intracytoplasmic tail, nucleotides 1597 to at least the Xba I restriction site at 2991, was contained within one exon (data not shown).

The relative levels of β_IT and β_Ctranscripts present in hematopoietic cell lines and primary cells were quantified using two methods. To show that the deleted fragment was not an artifact produced by secondary structure of the RNA, RNAse protections were first performed on total cellular RNA from cell lines, primary neutrophils, normal bone marrow, purified CD34⁺cells, and leukemic blasts. The RNA probe of 248 bp was designed to protect a 191-bp fragment from the β_C chain (nucleotides 1429 to 1620) and two fragments of 64 bp (nucleotides 1429 to 1493) and 24-bp (nucleotides 1596 to 1620) from the β_IT chain. Typical results obtained are shown in Fig 3. Secondly, semiquantitative RT-PCR using an end-labeled primer was performed on some RNA samples. There was good agreement between results obtained using the two methods. Quantification of β_ITshowed that it comprised between 10% and 25% of the total β chain message for U937 and TF-1 cells, similar to the levels obtained for primary myeloid cells, normal bone marrow (14.6% ± 6.9%, n = 7), purified neutrophils (13.8% ± 4.5%, n = 6), and CD34⁺ cells (16.6% ± 8.4%, n = 4). However, β_IT levels varied from 10% to 90% in blast cells from patients with AML and ALL and was greater than 80% of the total β message in 24 of 32 (75%) AML patients and two of 11 (18%) ALL patients (Fig 4). The one ALL patient with a β_IT level of 91% had an immunophenotype consistent with common ALL (CD10 68%, CD19 97%, peroxidase-negative, TdT-positive, CD33 2%, CD13 14%). There was no correlation between the β_IT expression detected and arrest of leukemic cells at different stages of differentiation as defined by the French-American-British classification.

To investigate whether the level of β_IT transcripts changed with differentiation of myeloid cells, leukemic blasts from six AML patients were cultured in liquid suspension in a cocktail containing IL-3, GM-CSF, and G-CSF. RNA was prepared at days 0 and 9 or 10, and semiquantitative RT-PCR was performed. Morphologic differentiation was observed in each patient, and the relative proportion of β_IT message decreased by approximately 10% to 30% (Table 3). The difference in β_IT expression before and after differentiation was highly significant (P = .0009, paired ttest).

To analyze GM-CSF binding to the truncated chain, β_IT DNA was transfected with GMRα DNA into COS-7 cells. Two classes of receptor affinity were obtained with K_D values of 2.5 nmol/L, consistent with reported values for the GMRα chain,32 and 200 pmol/L (Fig 5). This indicated that the β_IT could be expressed as protein and was able to convert low-affinity α chains to a higher-affinity αβ complex with binding affinity similar to that reported for the full-length β_C chain.17 

Previous studies have shown that the tyrosine kinase JAK2 binds to the β_C chain, and its activation requires the membrane-proximal region.23 To examine whether JAK2 could associate with β_IT, COS-7 cells were transfected with GMRα and either β_C or β_IT and grown to semiconfluence. The cells were then stimulated with or without GM-CSF, harvested, and immunoprecipitated with an anti-JAK2 antibody. Immunoblotting with the anti-β antibody DC-9 demonstrated that the β_IT chains coimmunoprecipitated with JAK2 (Fig 6A). Blotting with the phosphotyrosine antibody 4G10 showed that JAK2 associated with β_IT could be phosphorylated on tyrosine residues by stimulation of the cells with GM-CSF (Fig 6B).

Anti-β immunoprecipitation and Western blotting of transfected COS-7 cells yielded a protein with molecular mass of approximately 70 kD for β_IT, in comparison to 130 kD for β_C (Fig7A). A degradation product of variable intensity was consistently observed at approximately 80 kD in COS-7 cells transfected with β_C. Similarly, anti-β immunoprecipitation and Western blotting of TF-1 cells yielded a protein band at the expected position for β_IT, but it was always present at a relatively low level and was often obscured by the β_C degradation product (Fig 7C). Polyclonal antibodies were therefore raised to the novel C-terminus of the β_ITchain. Using cell lysates from transfected COS-7 cells, the antibody was shown to be specific for β_IT chains (Fig 7B). It did not cross-react with β_C and could be competed out with specific peptide but not an unrelated peptide derived from the retinoblastoma protein. Anti-β_IT immunoprecipitates of TF-1 cells and purified CD34⁺ cells cultured for 8 days with stem cell factor, IL-3, and IL-6 were probed with the anti-β antibody and showed that β_IT was expressed as protein in myeloid cell lines and primary myeloid cells but was absent from the murine T-cell line, CTLL (Fig 7C). At least two bands were always observed for the β_IT chain using either anti-β or anti-β_IT antibodies. These were thought to be due to variable glycosylation of the chain, as only one smaller band was observed in transfected COS-7 cells incubated with tunicamycin (data not shown).

To examine the ability of β_IT to support a proliferative signal, CTLL-2 cells were stably transfected with either vector (V), GMRα (α), GMRα + GMRβ_C (αβ_C), or GMRα + GMRβ_IT (αβ_IT). CTLL-2 cells were chosen because, unlike the more commonly used BA/F3 cells, they do not express endogenous β_C chains that could interfere with the signal obtained and were dependent on IL-2, not IL-3. Individual clones were selected, and expression of the different chains was confirmed by both surface-antibody binding and Western blotting of immunoprecipitates. Proliferation in response to varying concentrations of GM-CSF was examined using [³H]thymidine uptake and compared with the response to IL-2. Two αβ_C and αβ_IT clones were examined, and the assays were repeated at least twice. The clones containing vector, GMRα, or αβ_IT did not show any proliferation in the presence of GM-CSF, whereas the two αβ_C clones showed a proliferative dose-response effect (Fig 8).

The human β_C chain common to the GM-CSF/IL-3/IL-5 receptors is located at chromosome 22q12.2 → q13.1,34 and unlike the mouse, where two distinct but highly homologous genes have been found for the β subunits, AIC2A and AIC2B, ³⁵ only one human gene has been found.17 A number of different cDNAs with insertions and/or deletions were isolated in the original cloning of β_C from TF-1 cells17 and were thought to be created by alternative splicing rather than encoded by a distinct gene, since their alterations were found at sites corresponding to the exon-intron junctions of the mouse AIC2 genes.35 We have found that full-length transcripts containing a deletion identified in one of these alternative forms, clone KH85, can be detected as approximately 10% to 25% of the total β chain message in hematopoietic cell lines and primary hematopoietic cells and up to 90% of the β mRNA in blast cells from patients with acute leukemia (Fig 4). We have called this isoform β intracytoplasmic truncated (β_IT) because the 104-bp deletion just 3′ of the intracytoplasmic tail causes a frameshift alteration with a premature stop codon, potentially producing a protein with a truncated intracytoplasmic tail of just 46 amino acids instead of 432. The finding that β_IT decreased when AML blasts were cultured with IL-3, GM-CSF, and G-CSF (Table 3) raises the possibility that its expression is differentiation stage–specific in the leukemic cells. However, expression of β_IT was not related to AML FAB type, with high levels found in all three cases of M3 leukemia examined (Fig 4). Furthermore, increased levels of β_IT expression were not found in purified normal CD34⁺ cells (mean, 16.6% ± 8.4%, n = 4). It is possible that a primitive subpopulation of CD34⁺ cells express high levels of β_IT mRNA, but even if this is the case, the presence of high levels in leukemic samples showing myeloid differentiation is indicative of an aberrant phenotype. The finding of a very high β_IT level in one ALL patient sample indicates that this is not a myeloid-specific phenomenon.

Sequencing of genomic DNA demonstrated that the β_IT chain had arisen from the splicing out of a complete exon, nucleotides 1493 to 1596 of the original cDNA sequence,17 with the appropriate donor and acceptor splice sites present (Table 2). The genomic structure in this region is similar to that reported for the mouse β subunit genes AIC2A and B.35 With the exception of the first 5′ nucleotide in the human gene, G, the transmembrane domain is encoded in one exon of 91 bp, exon 11 in the mouse genes. This is followed by two small exons of 58 and 104 bp, respectively, and then a large exon that includes most of the intracytoplasmic tail sequence. This genomic organization fits a common pattern found in several cytokine receptor chains, indicating that there may be a common evolution of growth factor receptor chains, possibly a shared ancestral gene.14 36 

Although the protein could be expressed in COS-7 cells and was able to convert low-affinity binding GMRα chains to a higher-affinity GM-CSF binding protein, we were unable to evaluate protein levels in hematopoietic cells by immunoprecipitation and blotting with the anti-β antibody because of an 80- to 90-kD degradation product from the β_C chain (Fig 7). In primary cells, this was consistently observed as a broad band, probably due to variable glycosylation of the chain, which interfered with detection of the β_IT band. However, using specific anti-β_ITantibodies, we were able to demonstrate that full-length β_IT chains are present both in the TF-1 erythroleukemia cell line and in primary myeloid cells. We were consistently unable to detect either β_C or β_IT in blast cells from patients with AML using the anti-β or -β_IT antibodies, presumably because these cells often express less than 100 high-affinity GMR/cell,37,38 in comparison to 2,000 ± 450 and 1,100 ± 200 for TF-1 cells and neutrophils, respectively.39 

The first 23 amino acids of the intracytoplasmic tail of the β_IT chain that it has in common with the published β_C include the sequence known as box 1, a highly conserved sequence of amino acids including the proline-X-proline found in human gp130, G-CSFR, IL-2Rβ, EpoR, IL-7R, and IL-4R chains.21 This sequence is necessary for binding and subsequent phosphorylation of the tyrosine kinase JAK2.40,41 Consistent with this, JAK2 could bind to the β_IT isoform and was phosphorylated on tyrosine residues by GM-CSF stimulation (Fig 6). However, although box 1 is essential for induction of a mitogenic signal,42 JAK2 phosphorylation alone was insufficient to support a proliferative signal, as indicated by the lack of response to GM-CSF in CTLL clones transfected with GMRα and β_IT (Fig 8). A second region of homology called box 2, which is found in a number of receptors including human β_C and is known to be important for signal transduction in gp130, is not present in β_IT.21 This region is not required for JAK2 association and phosphorylation, although it may play a role in enhancing proliferation.40-42 The new sequence of 23 amino acids created by the frameshift contained two prolines, two cysteines, a serine, and a threonine (Fig 2B). There are no known consensus binding sequences within this new sequence.

Although the β_IT chain does not support proliferation, it could modulate β_C function, particularly if the β chains dimerize in the receptor complex.43,44 The effect would be especially noticeable in leukemic blast cells, where it may be the major species of receptor β chain. It might act as a dominant-negative for proliferation, as reported for a truncation mutant of the EpoR,16 or for differentiation, as suggested for truncation mutants of the G-CSFR.45 46 It is noteworthy that β_IT expression does not prevent in vitro differentiation of myeloid leukemic blast cells, although subtle modifications of differentiation would not be detected in such systems. It remains possible that high levels of β_IT expression contribute to the differentiation arrest seen in AML.

The authors thank Dr S. Devereux for the GMRα construct, and S. Langabeer for some of the RNA samples.