Multiple Glycosylated Forms of T Cell-derived Interleukin 3 (IL-3) HETEROGENEITY OF IL-3 FROM PHYSIOLOGICAL AND NONPHYSIOLOGICAL SOURCES*

Interleukin 3 (IL-3) derived from mouse T cells was biosynthetically labeled with either [36S]methionine or [3H]mannose, affinity-purified using various anti-IL-3 antibodies, and analyzed by sodium dodecyl sulfate- polyacrylamide gel electrophoresis. Autoradiography revealed the same three major bands with M, values of 21,500-22,500,27,000-31,000, and 32,000-36,000, irrespective of whether the anti-IL-3 antibody had been directed to the N or C termini of the IL-3 poly- peptide. Bioassay of eluates from the gels confirmed that all three bands exhibited IL-3 bioactivity. IL-3 produced from two nonphysiological sources, the myelomonocytic leukemia WEHI-3B or Cos 7 cells that had been transfected with an IL-3 cDNA clone, had in each case a different pattern of microheterogeneity. Treatment with either tunicamycin or N-glycanase resulted in IL-3 running as one band with M, 16,000, corresponding to its 140-amino acid polypeptide chain. No evidence for proteolytic processing was detected. These results show that the M, IL-3 Assay-IL-3 activities in supernatants were measured in a t3H]thymidine incorporation assay using the IL-3-dependent clone, R6-X E4.8.9, derived from (C57BL/6 X DBA/P)Fl bone marrow (26). Data were analyzed using a computer program which fitted a straight line to the descending part of the titration curve, the slope of the line being matched to that obtained by parallel titration of a standard preparation of IL-3 (27). One unit of biological activity was defined as the concentration of IL-3/ml required to stimulate 50% maximal response in this assay.

interleukin 2 (IL-2) and interleukin 3 (IL-3)l (1, 2). Glycosylation is known to be a feature of many of these lymphokines and cytokines, although the functional significance of this is not well understood. For example human interferon-y has been reported to exist in two different glycosylated forms with M , 25,000 and 20,000 and a nonglycosylated 15,500 form (3), whereas two murine forms with M, 21,800 and 20,600 have been described (4).
IL-3 has been purified from medium conditioned by cultured cells by several research groups using chromatographic methods. Two groups purified IL-3 obtained from a nonphysiological source, the myelomonocytic leukemia WEHI-3B, with yields of 8.4 (5) or 4% (6). Other groups used either the T cell lymphoma LBRM-33-5A4 (7) or mitogen-stimulated spleen cells (8) as the source and obtained IL-3 with yields of 10 (7) and 4%, respectively (8). Only microgram quantities of purified material were obtained in each case, making further structural characterization difficult. Moreover, the low yields made studies of the heterogeneity of the IL-3 molecule impossible, because selective losses of different forms of IL-3 molecules could not be ruled out. It seems likely that some of the disparities for the M, of WEHI-3B-derived IL-3 reported in various studies (for example M, 41,000 (9), M, 28,000 (5, lo), M, 25,000 ( l l ) , and M, 32,000 (6)) could be due to such losses during the many biochemical purification steps that were necessary. The availability of specific antibodies directed against synthetic peptides (12,13) offers an approach to the study of the molecular heterogeneity of IL-3. Previously we showed that polyclonal (14) and monoclonal (15) antibodies to IL-3 have facilitated quantitative purification of the native molecule with yields of up to 97%.
In this paper, we describe the use of antipeptide antibodies to characterize IL-3 produced by three different sources: activated T cells, the only known physiological source of IL-3; the myelomonocytic leukemia WEHI-3B in which IL-3 is produced as a result of aberrant activation of the IL-3 gene (16); and as a model for the systems that are being used to produce IL-3 in large quantities from mammalian cells, Cos 7 cells which have been transfected with a plasmid expressing an IL-3 cDNA clone.

MATERIALS AND METHODS
Conditioned Media-Medium conditioned by high density cultures (5 X 105-106 cells/ml) of the myelomonocytic leukemia WEHI-3B containing approximately 50 units/ml IL-3 were collected by centrifugation and concentrated 10-fold using an Amicon hollow fiber system (Mr cutoff >10,000).
Transfection of Cos 7 Cells-Cos 7 cells were grown to 50% confluence and then transfected with plasmid DNA using a calcium phosphate transfection method (18 ZL-3 Peptides and Anti-IL-3 Antibodies-Peptides corresponding to murine IL-3 residues 1-29 and 1-6 and full-length IL-3 (1-140) were synthesized by solid phase methods and were gifts from Dr. Ian Clark-Lewis (20,21).
Polyclonal antisera to IL-3 were generated in rabbits immunized with synthetic IL-3 peptides corresponding to amino acid residues IL-3 (1-29) and IL-3 (1-6) coupled to keyhole limpet hemocyanin (14). Antibodies were affinity-purified on peptides coupled to Sepharose. The antibodies were then coupled to CNBr-activated Sepharose beads (10 mg of antibody/ml of gel) to form affinity columns recognizing specific peptide sequences within IL-3 (14). Monoclonal antibodies (mAbs) specific for native IL-3 were raised by the fusion of myeloma cells with spleen cells from mice immunized with a mixture of IL-3 peptides coupled to keyhole limpet hemocyanin, followed by a boost with synthetic IL-3 as described fully elsewhere (15). Monoclonal anti-IL-3 antibodies were partially purified from ascites fluid by differential NH4SOl precipitation and were coupled to CNBractivated Sepharose beads.
Affinity Purification of IL-3-Conditioned media from the different cell sources were applied to the immunosorbent columns made with polyclonal (14) or monoclonal (15) anti-IL-3 antibodies. These were then washed with at least 10 times their bed volumes of phosphatebuffered saline. IL-3 was eluted into siliconized Corex tubes (Du Pont) with 5 times the bed volume of 0.1 M glycine (pH 2.5) and subsequently neutralized with a 25% volume of 1 M Tris (pH 8.0).
Zodination of IL-3-IL-3 derived from ConA-stimulated T cells or from the myeloid leukemia WEHI-3B was purified once using an anti-IL-3 (1-29) antibody-Sepharose column (14). The column eluate was iodinated by the chloramine-T method (22) to a specific activity of approximately 60 pCi/mmol. The iodinated IL-3 samples were then further affinity-purified using the anti-IL-3 monoclonal antibody 2 E l l (15 (23). Additional N-glycanase (0.3 unit) was added after 12 and 24 h of treatment in order to obtain a more complete digestion. 30 pl of sample buffer was added before loading onto a 13% polyacrylamide gel. Polyacrylamide Gel Electrophoresis-Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) was performed using 13% polyacrylamide gels and the buffer system of Laemmli (24) as described by Goding (25). The molecular weight standards (Bio-Rad) were: lysozyme, M, 14,400; soybean trypsin inhibitor, M. 21 66,200; phosphorylase b, M, 92,500. Samples were concentrated by methanol precipitation as follows: 9 volumes of methanol were added to 1 volume of the IL-3 sample that was in the elution buffer used either in affinity or HPLC columns, or the buffer used in eluting material from SDS gels (see below) and held overnight at -20 "C. Bovine serum albumin (50 pg) was added to the samples obtained from the affinity or HPLC columns, prior to the addition of methanol in order to obtain a complete precipitation of IL-3. Precipitates were collected by centrifugation at 16,000 X g for 30 min, resuspended in SDS-sample buffer, boiled for 5 min, and then loaded onto the gel. Gels were routinely stained with Coomassie Blue. The autoradiographic detection of 35S was enhanced by soaking the gels in Amplify (Amersham Corp.) for 30 min before drying and exposing to Kodak XAR-5 or Amersham Hyperfilm MP for between 1 and 28 days.
Elution of ZL-3 from SDS-PAGE Gels-After appropriate autoradiographic exposure, gel tracks were cut into 5 X 15-mm slices, and each slice was homogenized in 0.5 M NaCl containing 1 mg/ml of bovine serum albumin and 0.1% SDS as described (6). The homogenate was agitated for 24 h at room temperature, spun for 5 min at 15,000 X g, and the supernatant precipitated in methanol as described above. Precipitated samples were resuspended in medium and assayed for IL-3 bioactivity at a starting dilution of 1:2.
High Performance Gel Permeation Chromatography-P41.1-derived ?+labeled IL-3 (6400 units) that had been purified twice on a 2 E l l mAb affinity column was applied to a TSK-3000 SW HPLC column (Beckman Instruments) equilibrated with 10 mM sodium phosphate, pH 7.0, containing 0.02% Tween 20 and operated with a Waters HPLC system (Waters Associates, Milford, MA). A flow rate of 0.2 ml/min was maintained, and 400-pl fractions were collected. The bioactivity of each fraction was measured in the standard IL-3 bioassay with a starting dilution of 1:60 and six duplicate 3-fold dilutions. The radioactivities of 10-p1 samples were measured in duplicate in a Packard TriCarb 2000 liquid scintillation counter.
IL-3 Assay-IL-3 activities in supernatants were measured in a t3H]thymidine incorporation assay using the IL-3-dependent clone, R6-X E4.8.9, derived from (C57BL/6 X DBA/P)Fl bone marrow (26). Data were analyzed using a computer program which fitted a straight line to the descending part of the titration curve, the slope of the line being matched to that obtained by parallel titration of a standard preparation of IL-3 (27). One unit of biological activity was defined as the concentration of IL-3/ml required to stimulate 50% maximal response in this assay.

SDS-PAGE Analysis of Affinity-purified T Cell-derived
IL-3-P41.1, a BALB/c-derived L3T4 positive T cell clone, that consistently produced high levels of IL-3 after stimulation with ConA, was chosen for the biosynthetic labeling of IL-3 using [Y3]methionine. Labeled IL-3 was affinity-purified by two passages over a polyclonal anti-IL-3 (1-29) affinity column (14). The yield was about 95% based on biological activity. Samples of this material containing approximately 10,000-40,000 cpm were further analyzed on a 13% polyacrylamide gel run under reducing conditions. In order to confirm the specificity of the polyclonal anti-IL-3 (1-29) antibody column, IL-3 was also purified on immobilized monoclonal anti-IL-3 antibodies (15). %-Labeled IL-3 from the P41.1 T cell clone was purified twice using either mAb 2 E l l or mAb 1A3 (15). The yields were approximately 90% (15). Once again three major bands were seen after SDS-PAGE and autoradiography (Fig. 1, lanes C (mAb 1A3) and D (mAb 2Ell)). None of these bands corresponded to any major band in the starting conditioned medium (lane A ) nor did passage of T cell material over the affinity columns result in a detectable loss of a major band (lane B ) . An aliquot of [35S]IL-3 was also run over a control column composed of CNBr-activated Sepharose coupled with glycine in order to 145,000 cpm) were applied to a TSK 3000 SW HPLC column and run at low ionic strength. The IL-3 bioactivity eluted as a broad peak with a trailing shoulder that coincided with the profile of radioactivity (Fig. 2). Total recovery of radioactivity was 95%, the peak fraction being at an elution volume of 16.6 ml. Fractions were pooled as indicated in Fig. 2 and analyzed by SDS-PAGE and autoradiography. Comparison of the radioactivity showed patterns that fractions 3, 6, and 9-13 contained only forms corresponding to apparent Mr values of 21,500-22,500, 27,000-31,000, and 32,000-36,000, respectively. Each of the fractions contained significant amounts of bioactivity.
The second approach involved the elution of bioactivity directly from SDS-polyacrylamide gels that were run under nonreducing conditions to avoid the adverse effect of reduction on the recovery of bioactivity (6). clearly contained IL-3 bioactivity.
Analysis of IL-3 with Anti-1-6 Antibody-Two forms of IL-3 which differ at the N terminus have been identified (2). An antibody specific for the first 6 amino acids can distinguish and divide these forms into binding and nonbinding fractions. The fraction of IL-3 binding to the anti-IL-3 (1-6) antibody varies between 70 and 98%, depending on the cellular source and conditions of preparation. Repeated application of nonbinding material to the affinity column ruled out a saturation or equilibrium mechanism (2). The molecular basis for the failure of a fraction of molecules to bind to the IL-3 (1-6) antibody is not known. To help clarify this issue, the binding and nonbinding material was analyzed for the various Mr forms by SDS-PAGE. T cell-derived [35S]IL-3 was purified  (Fig. 5, lanes A and C). Tunicamycin treatment reduced the M , of most of the T cell-derived IL-3 to approximately 16,000,

FIG. 4. SDS-PAGE does not detect differences between forms of IL-3 that bind to anti-IL-3 (1-6) antibodies and those that do not. 35S-Labeled IL-3 affinity-purified twice on mAb 1A3
was applied to an affinity column recognizing the first 6 amino acids of IL-3. The breakthrough representing 5% of the bioactivity ( l a n e A ) and the glycine eluate representing 95% of the bioactivity ( l a n e E ) were then analyzed by SDS-PAGE run under reducing conditions.
3. Th&e was-some material &king at a lower M , than the major band in both the tunicamycin-treated and synthetic preparations; however, this may have resulted from incomplete reduction. There were also minor bands in the tunicamycin-treated preparation that ran at slightly higher M , (  C) and removed all the 3H from the IL-3 labeled with [3H] mannose (Fig. 6, lane D). In both cases IL-3 bioactivity was eluted from gel slices corresponding to the 35S-labeled M , 16,000 band (data not shown).
Analysis of the M, Forms of WEHI-3B-derived IL-3-Unlike affinity-purified T cell-derived IL-3 (Fig. 3), chromatographically purified IL-3, derived from WEHI-3B cells, gave a broad band around M, 32,000 when analyzed by SDS-PAGE (6). WEHI-3B-derived material was examined to determine whether the difference in M , patterns reflected differences in the cellular sources or the techniques used in their purification. We radioiodinated the WEHI-3B-derived material because the low amounts of IL-3 bioactivity in medium conditioned by WEHI-3B cells (50 units/ml compared to 10,000 units/ml for the T cell clones P41.1) made 35S labeling very difficult. Concentrated (10 X) supernatant from WEHI-3B, containing about 500 units/ml IL-3, was purified in parallel with an unlabeled P41.1 supernatant on the anti-IL-3 (1-29) antibody column. The two column eluates were then iodinated to a high specific activity by the chloramine-T method and repurified on the 2Ell mAb column. SDS-PAGE analysis under nonreducing conditions revealed that the WEHI-3Bderived material had a similar electrophoretic pattern to the chromatographically purified IL-3 (6), but a strikingly different pattern to that of T cell-derived IL-3 (Fig. 7). Because iodination by this method destroyed over 95% of the bioactivity of the IL-3, a preparation of unlabeled WEHI-3B-derived IL-3 was run in parallel on the same gel, and the bioactivity profile was determined by elution of the unlabeled track. Samples of labeled IL-3 were run under reducing conditions (data not shown) to determine the M , values given below. The bioactivity was contained in a broad band between M, 29,000 and 45,000 as well as in a very faint band migrating at exactly the position of the smallest (21,500-22,500) T cellderived band. Two other iodinated bands were apparent at M, 27,000 and 17,500 but had no detectable bioactivity associated  8, lane A ) . Elution from a gel run under nonreducing conditions (Fig. 8, lane B ) established that the bands obtained after autoradiography contained IL-3 bioactivity. COS 7 cell-derived [35S]IL-3 was also treated with N-glycanase in order to remove N-linked carbohydrate. Comparison of N-glycanase-treated [35S]IL-3 with untreated material (Fig.  6, lane E versus lane F ) established that approximately 50% of the IL-3 is converted to the M , 16,000 form, whereas the other 50% of the 35S-labeled material appears as a smear between M, 28,000 and 16,000 (as determined by scanning densitometry of the autoradiograph, data not shown).

DISCUSSION
In this study we describe the use of polyclonal and monoclonal antibodies to characterize murine IL-3 obtained from three different sources: T cell clones, the myelomonocytic leukemia WEHI-3B, and Cos 7 cells expressing an IL-3 cDNA. Affinity columns prepared from these antibodies bound 90-98% of IL-3 bioactivity (14, 15) and were used for purification and study of the heterogeneity of the molecules.
SDS-PAGE analysis of affinity-purified IL-3, derived from T cell clones, revealed the presence of three broad bands, with apparent M , of 32,000-36,000, 27,000-31,000, and 21,500-22,500. The three M , forms were present in preparations purified using antibodies recognizing either N-terminal sequences (corresponding to amino acid residues 1-6, 1-29) or C-terminal sequences (amino acid residues 130-135) (Fig. 1). All three bands isolated using two independent separation techniques, gel permeation HPLC and preparative SDS-PAGE, contained bioactive IL-3.
Major proteolytic cleavage can be excluded as a cause of the multiple molecular weight forms, since IL-3 that was purified first on mAb 1A3 (recognizing amino acid residues 130-135) followed by purification on an antibody specific for the 6 N-terminal amino acids contained all three M, forms (Fig. 4). Interestingly, the IL-3 that failed to bind to the polyclonal anti-IL-3 (1-6) antibody column also had the same pattern of bands on SDS-PAGE. Thus, no major molecular differences between the binding and nonbinding forms could be detected by this method. A possible minor modification close to the N terminus that would result in the loss of the 1-6 epitope (for example, loss of amino acids or 0-linked glycosylation) could not be identified by our experiments.
IL-3 has been described as a glycoprotein (5,28) and has four potential sites for N-linked glycosylation (29,30). Our experiments using tunicamycin and N-glycanase (Figs. 5 and 6) clearly show that the M, heterogeneity of T cell-derived IL-3 is principally due to N-linked glycosylation. No evidence for 0-linked glycosylation was detected, although the present data cannot completely exclude a minor level of 0-linked modification.
The pattern of WEHI-3B-derived IL-3 (Fig. 7) differed strikingly from that of the T cell clone (Fig. 3) in that distinct bands were not apparent, but rather a broad smear was present between M , 29,000 and 45,000. As IL-3 is the product of a single gene and the major source of M, heterogeneity of IL-3 is N-linked glycosylation (this study), it is likely that WEHI-3B glycosylates IL-3 differently from T cells, leading to the different patterns observed. IL-3 produced by transient expression in Cos 7 cells was again different from the T cell or WEHI-3B-derived material. The M, forms of IL-3 were larger than the T cell-derived material with strong bands of IL-3 bioactivity at M , 26,000 and 30,000-31,000 and a broad smear up to M , 45,000. In the case of Cos 7 cell-derived material only about 50% of IL-3 was converted to the M , 16,000 form by N-glycanase. The remaining material was converted to a smear running between M , 28,000 and 16,000 (Fig. 6). Neuraminidase treatment reduced the mean apparent M, of this material further, indicat-ing the presence of residual carbohydrate (data not shown). It is not clear whether this smear is due to incomplete removal of N-linked sugar or whether a portion of IL-3 from these cells undergoes partial 0-linked glycosylation or other modification.
There are many precedents for variations in the glycosylation in different cell types. It is well documented that patterns of glycosylation may differ not only between different tissues (31,32) and at different stages of development within the same cell lineage, but also between normal and tumorigenic cells of the same type (33,34). In the case of murine interferon-?, one T cell lymphoma has been shown to make two species with M, 16,800 and 17,800 rather than the normal forms of M, 20,600 and 21,800 (4). The human "acute T cell lymphoma" line HUT 102B2 expresses IL-2 receptors from a normal gene but with an aberrant glycosylation pattern (35). However, there are also many cases in which glycosylation patterns are the same in normal and tumorigenic cells (36,37).
The fact that most cell-surface or secreted proteins are glycosylated suggests that glycosylation has an important, although poorly understood, function. Deglycosylated forms of lymphokines such as IL-3, granulocyte-macrophage colonystimulating factor, and interferon-? can mediate biological effects of native molecules (21,38,39). However, it is not always clear whether the potencies of some recombinant lymphokines are as high as those of the native glycoproteins. Deglycosylation of granulocyte-macrophage colony-stimulating factor, produced by recombinant techniques in yeast or Chinese hamster ovary cells, has been reported to lead to an increase in biological activity (40), suggesting that abnormal glycosylation in these nonphysiological host cell types resulted in molecules with impaired activity. Clearly attention must be paid to glycosylation when secreted proteins like IL-3 or granulocyte-macrophage colony-stimulating factor are produced under unusual conditions or by nonphysiological cell types. As demonstrated here, affinity-purified and biosynthetically labeled lymphokines should prove a very useful system for further study of the effect of glycosylation on these factors.

Multiple Glycosylated
Forms of IL-3 14517