Characterization of the Stat5 Protease*

Immature myeloid cells have been shown to transduce signals through a carboxyl-terminally truncated isoform of Stat5. This functionally distinct signal transducer and activator of transcription isoform is generated through a unique protein-processing event. Evaluation of numerous cell lines has determined that there is a direct correlation between the expression of truncated Stat5 and protease activity. Moreover, protease activity is found only in the myeloid and not in lymphoid progenitors. To further characterize the protease small quantities have been purified to near homogeneity. Studies on this purified material indicate that the protease has an apparent molecular mass of ∼25 kDa and is active over a wide range of pH values. The protease will also cleave both activated (i.e. tyrosine-phosphorylated) and inactivate Stat5. Although this activity is sensitive to phenylmethylsulfonyl fluoride, it is notably not sensitive to several other serine protease inhibitors. Additional studies have led to the identification of the unique site where the protease cleaves Stat5. Mutagenesis of this site renders Stat5 resistant to cleavage. Consistent with the model that Stat5 cleavage is important for early myeloid development, introduction of a “non-cleavable” isoform of Stat5 into FDC-P1 cells (a myeloid progenitor line) leads to significant phenotypic changes.

Characterization of the ability of IFNs 1 to induce genes rapidly has led to the identification of the JAK-STAT pathway (1)(2)(3). In this signaling paradigm, JAKs are receptor-associated tyrosine kinases, and STATs (signal transducers and activators of transcription) are the cytoplasmic transcription factors they activate. Once activated, STATs dimerize, translocate to the nucleus, and bind to enhancer elements, culminating in gene induction. Subsequent studies have determined that all members of the cytokine family transduce signals through one or more of the seven members of the STAT family (4,5). These STATs share several functionally conserved domains including an amino-terminal coiled-coil domain, a DNA binding domain, a linker domain, an SH2 domain, a tyrosine activation domain, and a divergent carboxyl-terminal transcriptional activation domain (6,7).
Interleukin (IL)-3 is a member of a subfamily of functionally related cytokines (i.e. IL-3, IL-5, and GM-CSF) that all signal through a common receptor chain (8,9). Consistent with this, all three ligands play an important role in the maturation, proliferation, and activation of myeloid lineages (10). Moreover, a number of studies have determined that these ligands transduce signals through several isoforms of Stat5. In most cell types, full-length Stat5a (96 kDa) and Stat5b (94 kDa) are activated in response to stimulation with IL-3 or other members of this family (1)(2)(3). This leads to the induction of several known Stat5 target genes (11)(12)(13)(14)(15)(16)(17)(18)(19). In the absence of both Stat5a and Stat5b, there are significant defects in the development of CFU-Mix, CFU-Eos, and CFU-GM colonies, as well as defects in the induction of target genes (20).
In myeloid progenitors, IL-3 stimulates the induction of carboxyl-terminally truncated isoforms of both Stat5a (i.e. 77 kDa) and Stat5b (i.e. 80 kDa). These isoforms, which are missing their transcriptional activation domain, are functionally distinct and fail to promote the induction of Stat5 target genes (19,21,22). In contrast to other STATs (23)(24)(25), the truncated isoforms of Stat5 are generated through a unique proteinprocessing event. Previous studies have indicated that this protease is specific for Stat5 and can only be found in several immature cell lines (19).
In this work, we extend these studies and demonstrate that there is a direct correlation between the expression of truncated Stat5 isoforms and protease activity. Moreover, this activity is only found in myeloid progenitors, supporting our hypothesis that the Stat5 protease plays an important regulatory role during myelopoiesis. We find no evidence for this during lymphoid development. Biochemical characterization of this protease has revealed that it has an apparent molecular mass of ϳ25 kDa and exhibits a unique pattern of sensitivity to serine protease inhibitors. Functional evidence is provided for a role of this protease in myeloid development.
Protease Assays-Stat5 cleaving activity was evaluated either through immunoblotting or electrophoretic mobility shift assay (EMSA). Briefly, 1 l of recombinant Stat5b (or Stat5a) substrate (prepared by transient transfection; see above) was incubated with 1-5 l of crude or partially purified protease for 30 -60 min at 37°C. The reactants were then evaluated either by SDS-PAGE or EMSA with an IRF1-GAS probe (gatcGATTTCCCCGAAAT; Oligos Etc.), as described previously (37,38). For pH optima studies, either 1 M Hepes or 1 M Tris, at the appropriate pH, was added to each cleavage assay to achieve a final concentration of ϳ330 mM (i.e. 1 l of buffer ϩ 1 l of protease ϩ 1 l of rStat5b*).
Protein Analysis-Proteins were fractionated by SDS-PAGE as described previously (1,19). For immunoblotting, proteins were transferred to nitrocellulose membranes (Schleicher & Schuell) and probed with antibodies at a 1:2000-fold dilution or as recommended by the manufacturer (38). Pan-Stat5 and antiphosphotyrosine (4G10) monoclonal antibodies were purchased from Transduction Laboratories and Upstate Biotechnology Inc., respectively. The mass of peptides was determined by MALDI mass spectrometry (Voyager-DE RP with a 337-nm pulsed nitrogen laser; Perspective Biosystems) at the Columbia University Protein Core facility. Briefly, 200 -300 pmol of biotinylated peptides, biotin-GSGATYMDQAPS and biotin-GASATYMDQAPS (prepared at Amgen, Boulder, CO), were incubated with 5-8 l of crude or partially purified protease in CHAPS buffer in a final volume of 10 l at 37°C for 60 min. The reactants were dried and redissolved in 10 l of matrix solution (10 mg/ml 4-hydroxy-␣-cyanocinnamic acid in 50% acetonitrile, 0.1% trifluoroacetic acid ϩ bradykinin as an internal standard).
Protein Purification-Crude protease was prepared by gentle resuspension of a pellet of washed cells (ϳ2 ϫ 10 6 cells) in 2 volumes of cold (4°C) CHAPS buffer adjusted to 150 mM NaCl (CHAPS, 150 mM). This material was then homogenized by 30 strokes in a 7-ml glass Dounce with a "type A" pestle (Kontes). After a 1-h incubation at 4°C, the lysate was cleared by centrifugation (12000 ϫ g for 20 min at 4°C). For purification, 10 ml of extracts were prepared from 500 ml of cultured FdTrk cells at one time and then snap-frozen. 20 ml of accumulated extract was fractionated on a DEAE-Sephacel (Amersham Pharmacia Biotech) column (2.5 ϫ 10 cm), equilibrated in 1ϫ CHAPS, 150 mM. Proteins were eluted with a linear 0.15-1.2 M NaCl gradient (in 1ϫ CHAPS buffer), and protease activity was determined by EMSA. Peak fractions, eluting between 500 and 650 mM NaCl (radiometer conduc-tivity meter), were pooled, diluted to 150 mM NaCl, and applied to a heparin-agarose (Sigma) column (1.5 ϫ 12 cm), also equilibrated in 1ϫ CHAPS, 150 mM. Bound proteins were again eluted with a linear 0.15-1.2 M NaCl gradient (in 1ϫ CHAPS buffer). Peak fractions, which eluted between 600 and 800 mM NaCl, were diluted to 150 mM NaCl and applied to a Econo-Pac CM Cartridge (2 tandem 5-ml columns; Bio-Rad), equilibrated as described above. Peak activity eluted between 280 and 380 mM NaCl. Next, 1-2 ml of this peak activity was either applied to a 50-ml size exclusion column (2 ϫ 27 cm; SE100/40, Bio-Rad) or a second 3-ml DEAE-MacroPrep column (Bio-Rad), equilibrated to pH 8.4. A concentrated sample (ϳ4-fold; Centricon-3, Millipore) was applied to the exclusion column both before and after calibration with Sigma molecular weight standards. These included blue dextran (200,000 kDa), bovine serum albumin (66 kDa), ovalbumin (45 kDa), carbonic anhydrase (29 kDa), lysozyme (14.3 kDa), and cytochrome c (12.4 kDa). Conditions for the second DEAE fractionation were similar to the first, except binding and elution were carried out at pH 8.5. In both of these final chromatographic steps, fractions with peak activity were concentrated 3-5-fold by Centricon 10 membranes (Millipore) or 50 -60-fold by lyophilization and then fractionated on a 12% SDS-PAGE.

Correlation between the Expression of Truncated Stat5 and Stat5
Protease Activity-Stimulation of cells with members of the IL-3 family of cytokines leads to the induction of two distinct types of Stat5 DNA binding complexes (1,27). The faster migrating complex, consisting of two carboxyl-terminally truncated isoforms of Stat5a (i.e. p77) and Stat5b (i.e. 80), is found in nuclear extracts prepared from immature cell types (1,22). A slower migrating complex, consisting of the full-length isoforms of Stat5a (i.e. p96) and Stat5b (i.e. p94), is found in most other cells (1,2). The faster migrating complex has been well characterized in DA-3 and FDC-P1 cells, which represent myeloid progenitors (1, 19, 39 -41). This complex has also been reported in other early myeloid cell lines, including 32Dc1 (27). But a more differentiated subline of 32Dc1 cells, which exhibits early erythroid features (42), yields only the full-length complex after stimulation with IL-3 or erythropoietin (27).
Recent studies have determined that a protease is responsible for the generation of the truncated isoforms of Stat5 in DA-3 and FDC-P1 cells (19,41). To extend these initial observations, extracts were prepared from several additional cell lines and evaluated for Stat5 protease activity ( Fig. 1). In the absence of protease activity, the full-length DNA binding activity (i.e. an "Intact" p94 homodimer) was recovered. If protease activity was present, then the full-length recombinant activated Stat5b (i.e. rStat5b) was cleaved into the faster migrating "truncated" DNA binding activity (i.e. the "Cleaved" p80 homodimer). Extracts prepared from FdTrk cells served as the positive control for protease activity ((19) see Fig. 1, lane 2). In several cases, an additional intermediate complex (i.e. a p94:p80 heterodimer; data not shown), which correlated with less potent protease activity (e.g. see Fig. 3A), was recovered (see Fig. 3A, lanes [3][4][5]. Similar results were obtained with the slower migrating (19), recombinant, activated Stat5a (rStat5a; see Fig. 1, lanes 10 and 11).
Stat5 protease activity was recovered from each of four cell lines where truncated Stat5 had previously been identified (i.e. FdTrk, WEHI-3b, DA-3, and 32Dc1 (1,19,27)). Yet, in four cell lines, where only full-length Stat5 had previously been identified (i.e. 293, BaF/3, HeLa, and 3T3 cells (1,19,27,43)), rStat5b was not cleaved (see Fig. 1, lanes 6 -9). As noted previously (1,19,22,27), the truncated species of Stat5 exhibits a more robust level of DNA binding activity than the full-length isoform. An important control for these studies, evaluation of the intrinsic DNA binding activity recovered from each of the CHAPs extracts (Fig. 1, lanes 12-19), demonstrated that there was an insufficient level of endogenous Stat5 DNA binding activity to confound the results. These studies demonstrate that there is a direct correlation between the activation of the carboxyl-terminally truncated isoforms of Stat5 and the presence of Stat5 protease activity.
Signaling in Lymphocytes-Analogous to myelocytes, the development of lymphocytes is a multistep process that is carefully regulated. Two cytokines that are critical to lymphoid development, IL-2 and IL-7, transduce their signals through Stat5 (20, 32, 44 -48). To determine whether the truncated species of Stat5 may also be differentially activated in developing lymphocytes, extracts were prepared from cell lines representing several stages in lymphoid development (e.g. 22D6, 1881, S49, AKR and P3X (49 -52)). In each case, stimulation with IL-7 led to the formation of the full-length DNA binding complexes (data not shown), indicating that the activation of a carboxyl-terminally truncated Stat5 isoforms is not important for these cells. Reasoning that the truncated isoform of Stat5 may be activated at an earlier developmental stage than those represented by these cell lines, lymphocytes from RAG-1 null mice were evaluated. As RAG-1 is important in immunoglobulin and T-cell receptor chain rearrangements, these mice exhibit an earlier block in T-cell and B-cell maturation (53).
To evaluate Stat5 activation in these early lymphocytes, thymic T-cells were collected from both RAG-1 knock-out mice (predominantly CD44 ϩ /CD25 ϩ /CD4 Ϫ /CD8 Ϫ ) and control C57Bl/6J mice (predominantly CD44 Ϫ /CD25 Ϫ /CD4 ϩ /CD8 ϩ ) and then stimulated with IL-7. DNA binding activity was evaluated by EMSA (see Fig. 2). Stat5 DNA binding activity for IL-3-stimulated BaF/3 cells and FdTrk cells served as controls for the slower (i.e. intact) and faster (i.e. cleaved) migrating forms of Stat5. In each case, IL-7 induced a Stat5 DNA binding complex that comigrated with the triplet of bands representing full-length Stat5 (i.e. p94:p94, p94: p96, and p96:p96 dimers (1, 2)). Similar results were obtained when these extracts were evaluated by immunoblotting with a pan-Stat5 reactive antibody (data not shown). Therefore, in contrast to studies in myelocytes, there is no evidence that truncated isoforms of Stat5 play an important role in transducing signals during the examined stages of lymphoid development.
Purification of the Stat5 Protease-Crude cell lysates contain a complicated mixture of proteases that are likely to confound evaluation of the Stat5 protease. To avoid this, Stat5 protease activity was purified by sequential chromatography on DEAE-Sephacel, heparin-agarose, and CM affinity matrixes. This material was subsequently loaded onto an SE100/40 size exclusion column that had been calibrated with several standards. Of these standards, ovalbumin (ϳ45 kDa) eluted at fraction 20; carbonic anhydrase (ϳ29 kDa) eluted at fraction 22; lysozyme (ϳ14 kDa) eluted at fraction 36; and cytochrome c (ϳ12 kDa) eluted at fraction 28 (data not shown). Peak Stat5 protease activity was recovered in fraction 31 (i.e. fractions 30 -32; see Fig. 3A), suggesting that the protease is smaller than 40 kDa but larger than 10 kDa. Based on these observations and studies of other small proteases (see "Discussion"), it is reasonable to assume that the Stat5 protease is greater than 20 kDa but smaller than 40 kDa.
Next, peak fractions from several chromatographic steps, including the SE100/40 size exclusion column and a second DEAE column, were evaluated by silver staining. As shown in Fig. 3B, these steps achieved a substantial purification of the FIG. 1. Immature myeloid cell lines express Stat5 protease activity. Recombinant activated (*) Stat5b (rSt5b*) or Stat5a (rSt5a*), prepared from 293 cells, was incubated with CHAPS extracts prepared from each of the indicated cells lines (lanes 2-11) at 37°C for 1 h and then evaluated by EMSA with an IRF-1 GAS probe. The same extracts were also evaluated by EMSA without incubation with recombinant Stat5 (Extracts Alone, lanes [12][13][14][15][16][17][18][19]. The mobilities of the DNA binding complex representing Intact (i.e. p94:p94) and Cleaved (i.e. p80:p80) are indicated in the left margin. p80-p94 represents an intermediate complex. protease. In the most active fraction from the size exclusion column (i.e. fraction 31), 1-2 faint bands of ϳ25 kDa not found in inactive fractions (e.g. fraction 24) could be visualized, especially after concentration. Consistent with this, a single major ϳ25-kDa band was identified when the most active fraction (A) from the second DEAE-chromatography step was concentrated. This band was not evident in inactive fractions (I). These data indicate that the protease is likely to be a ϳ25-kDa protein.
The Stat5 Protease Is Active in a Wide Range of Physiological pH Values-Optimizing several of the chromatographic steps required understanding whether the Stat5 protease was sensitive to fluctuations in pH. To determine this, the pH of protease reactions was adjusted with a 30-fold molar excess of Hepes or Tris buffer. To increase the sensitivity of this assay, smaller volumes of protease and shorter incubations were employed. As shown in Fig. 4, the protease was able to cleave Stat5 under pH values ranging from 6.0 to 8.6. Although the protease has potent activity at each of these pH values, cleaving activity was modestly diminished in the pH 6.0 (Hepes) or pH 6.7 (Tris) samples. These studies suggest that the Stat5 protease is active in a wide range of physiological pH values.
The Stat5 Protease Is a Member of the Serine Family of Proteases-Previous studies, which had determined that the Stat5 protease was sensitive to PMSF, suggested it was a member of the serine family of proteases. To evaluate this more carefully, digestions were carried out in the presence of several additional protease inhibitors. Of these, only PMSF (170 g/ml) significantly inhibited Stat5 protease activity (see Fig. 5). PMSF is a serine protease inhibitor with broad specificity. It also has some activity against several cysteine proteases (e.g. papain). In contrast to the potent activity of PMSF, leupeptin (0.5 g/ml), another serine/cysteine protease inhibitor (also active against papain), had no effect on Stat5 cleaving activity. Likewise, aprotinin (2 g/ml), a serine-specific protease inhibitor, and pepstatin (0.7 g/ml), an aspartic protease inhibitor, did not block Stat5 protease activity. However, partial inhibition was obtained with DFP, another potent serine protease inhibitor (data not shown (54)). Several additional protease inhibitors, including benzamidine (a serine-specific inhibitor), E-64 (a cysteine-specific inhibitor), and EDTA (a metalloprotease inhibitor), were also ineffective at blocking protease activity (data not shown). This pattern of inhibition indicates that  3-7) or Tris (lanes 8 -11) buffers equilibrated to the indicated pH. After digestion, for 1 h at 37°C, they were evaluated by EMSA as outlined in Fig. 1. the Stat5 protease is an unusual member of the serine family of proteases.
The Stat5 Protease Cleaves Activated and Inactivated Isoforms of Stat5 Equivalently-The cleavage assays employed to date only examined the ability of the Stat5 protease to cleave activated (i.e. tyrosine-phosphorylated) Stat5. However, as this protease can be isolated from unstimulated cells (see Fig. 1), it was of interest to determine whether it could cleave nonphosphorylated Stat5 as well. To test this, recombinant Stat5b was prepared both before and after stimulation of a cotransfected receptor (31). As anticipated, activated Stat5b bound DNA and was detected by an antiphosphotyrosine-specific antibody. In contrast, Stat5b overexpressed in unstimulated cells failed to bind DNA (see below (19)) and was not recognized by an antiphosphotyrosine-specific antibody (data not shown). When cleavage was evaluated by an immunoblotting assay, it was evident that both native and tyrosine-phosphorylated Stat5b were cleaved equally well by partially purified preparations of protease (Fig. 6). These observations indicate that the protease may both cleave Stat5 constitutively (i.e. in unstimulated cells) and after activation (i.e. in stimulated cells).
The Stat5 Protease Cleavage Site-Through both sequence analysis of purified p77 and p80 peptides (1), and subsequent mapping studies (19), the cleavage site in Stat5 has been mapped distal to amino acid 720 (i.e. in Stat5b). To map more carefully the cleavage site, biotinylated peptides spanning this region of Stat5a and Stat5b (i.e. amino acids 719 -730 and 724 -735, respectively) were generated (see Fig. 7A). The mass of the peptides was determined, before or after cleavage, by MALDI mass spectroscopy. Before cleavage, the mass of the biotinylated peptides was 1539.41 and 1524.7 m/z. After cleavage with the Stat5 protease a new major species, corresponding to a biotinylated amino-terminal fragment, was detected in both samples. This again demonstrated that phosphorylation is not required for cleavage. For Stat5a, the amino-terminal (i.e. biotinylated peptide) peak had a mass of 909.085 m/z, unequivocally placing the cleavage site between amino acids 719 (tyrosine) and 720 (methionine). Similar results were obtained for the Stat5b peptide, placing the cleavage site between amino acids 724 (tyrosine) and 725 (methionine). The ability of the protease to cleave peptides of 12 amino acids indicated that they contained all of the residues required for cleavage speci-ficity. Since only 8 of these 12 amino acids are conserved between Stat5a and Stat5b (including the human Stat5s; see Fig. 7A), it is possible to delimit further the sequence required for cleavage to ATYMDQAP.
To confirm the identity of the cleavage site, tyrosine 724 (Tyr-724) and methionine 725 (Met-725) of Stat5b were mutated to serine and alanine respectively (i.e. Y724S2M725A; Stat5b m/m ; see Fig. 7B). When this protein was expressed and activated in 293 cells, it exhibited the anticipated wild type DNA binding activity (see Fig. 8A, compare lanes 1 and 4). However, in contrast to wild type Stat5b (Stat5b wt ), it was highly resistant to cleavage. Again, cleaving activity was sensitive to PMSF. Similar results were obtained when cleavage was evaluated by SDS-PAGE (see Fig. 8B), but this assay also revealed some novel Stat5b m/m cleavage products. These products were more evident in purified protease preparations, suggesting that they may represent overdigestion or perhaps a small loss in specificity. However, as these species are only seen with the Stat5b m/m substrate, they are likely to represent secondary cleavage sites. Once again, this cleaving activity is sensitive to PMSF.
To define the cleavage site more carefully, additional mutants were generated (see Fig. 7B). First, the amino acids flanking the cleavage site, Tyr-724 and Met-725, were individually mutated. Surprisingly, mutant Stat5b m1 (Y724A) exhibited poor DNA binding activity (Fig. 9A, lanes 3 and 4) and could not be evaluated by this assay. In contrast, mutant Stat5b m2 (M725L) exhibited good DNA binding activity (Fig.  9A, lanes 5 and 6) and was readily cleaved but not as well as control Stat5b. The next mutant, Stat5b m3 (T723V), also failed to bind DNA and likewise could not be evaluated by this assay. The final mutant, Stat5b m4 (A722S), exhibited good DNA binding and was readily cleaved. A more complete picture emerged when these cleavage reactions were evaluated by immunoblotting after fractionation on an SDS-PAGE (Fig. 9B). Although Stat5b m1 was less robustly expressed, it was clearly resistant to cleavage (lanes 3 and 4). In contrast, cleavage of each of the other three mutants was comparable to that of rStat5b wt . These studies indicate that Tyr-724 in Stat5b (and by extension Tyr-719 in Stat5a) is critical for cleavage.
Expression of Stat5b m/m (Y724V2M725S) in FDC-P1 (cl.19) Cells-The preceding studies, demonstrating that a lineagespecific protease cleaves Stat5b after residue 724, support our hypothesis that the cleavage of Stat5 is important in myelogenesis. To test this, an expression plasmid encoding a noncleavable isoform of Stat5 (Stat5b m/m ) was introduced into FDC-P1 (cl.19) cells. These cells represent myeloid progenitors, which proliferate (i.e. "self-renew") when grown in IL-3, but differentiate toward monocytes when cultured in GM-CSF (29,  1 and 2), wild type Stat5b cDNA (rSt5b wt , lanes 11 and 12), or the indicated Stat5b point mutants (rSt5b m1 -rSt5b m4 , lanes 3-10; see Fig. 7B for details) were assayed by EMSA as outlined in Fig. 1. DNA binding activity was evaluated either before (Ϫ) or after (ϩ) a 1-h digestion with partially purified preparations of protease at 37°C. The mobility of p94:p94 homodimer (Intact) and p80:p80 homodimer (Cleaved) and the intermediate complex, p80:p94, are indicated in the margins. B, extracts detailed in A were evaluated by immunoblotting with a panStat5 antibody (Signal Transduction Laboratory) after fractionation on a 7% SDS-PAGE. Molecular weight markers and the mobility of p94 and p80 are Stat5b products are indicated in the left and right margins, respectively. 30). The expression of full-length Stat5 in these cells could be anticipated to promote the expression of more "mature genes," thereby altering the phenotype of these cells. Briefly, neomycin-resistant clones were selected from FDC-P1 (cl.19) cells transfected with either Stat5b m/m , Stat5b wt , or "vector alone" and then screened by an immunoblotting assay (see Fig. 10). To increase the sensitivity of this assay, nuclear extracts were probed with a Stat5 phosphotyrosine-specific antibody (Fig.  10A). Note that low levels of Stat5, representing cytoplasmic contamination, could be detected in unstimulated extracts (Fig.  10B). But these species were never activated. Four clones were evaluated. Two of the neomycin-resistant clones were recovered from the Stat5b m/m transfectants (i.e. A1 and A2) and expressed activated full-length Stat5b in nuclear extracts. This contrasted the control vector alone (i.e. D3 [Neo R ]) and Stat5b wt (B1) transfectant lines, where as anticipated only truncated Stat5 was recovered from the nuclei of stimulated cells (see Fig.  10, lanes 4 and 10). In the A1 clone (Stat5b m/m ; lane 6), both full-length and truncated Stat5 were present in stimulated nuclear extracts, suggesting that both activated endogenous and mutant Stat5 had translocated to the nucleus. However, in the A2 clone (Stat5b m/m ; lane 8), only full-length Stat5 was identified in stimulated nuclear extracts.
Next, several of these clones were evaluated histochemically. When the D3 clone was cultured in IL-3 (Fig. 11A), it exhibited the characteristic features of immature self-renewing FDC-P1 cells (29,30). When the D3 clone was cultured in GM-CSF, it acquired characteristic features of differentiation (see Fig.  11B), including increased size, increased (cytoplasmic) granularity, increased nuclear and cytoplasmic volume, and in-creased membrane ruffling (29,30). The A1 and A2 clones exhibited a more intermediate phenotype when grown in IL-3 (i.e. conditions of "self-renewal"), including an increase in cytoplasmic volume, nuclear volume, and cytoplasmic granularity. This was most striking for the A2 clone, which only activates and translocates the mutant full-length Stat5 into the nucleus (see Fig. 11C). The morphology of these cells did not change after culture in GM-CSF. Similar observations have been made with clones expressing an epitope-tagged version of Stat5 m/m . 2 These observations provide further evidence that the activity of the Stat5 protease is important for the normal growth of myeloid progenitors. DISCUSSION Stat5 is encoded for by two highly homologous genes (Stat5a and Stat5b) and transduces signals for the IL-3 family of ligands (1,2). The proteins encoded by these genes differ only in their carboxyl-terminal transcriptional activation domain (1)(2)(3). Stat5a and Stat5b isoforms that are missing this carboxyl terminus have also been identified in several immature myeloid lineages (1, 19, 27, 39 -41). Moreover, several groups have reported that a loss in the expression and activation of these isoforms correlates with maturation (1, 39 -41). These studies have also determined that truncated Stat5 is generated through a unique protein-processing event (19,41). In addition, the Stat5 protease has been found to be constitutively activated and associated with the nucleus (19,41). Consistent with our most recent observations, others have determined that this protease is sensitive to PMSF and cleaves both inactive (i.e. unphosphorylated) and active (i.e. tyrosine-phosphorylated) isoforms of Stat5 (41). These observations suggest that imma-2 F. Piazza, J. Valens, and C. Schindler, unpublished observations. FIG. 10. FDC-P1 subclones expressing Stat5 mutants. A, phosphotyrosine-Stat5 immunoblot of FDC-P1 sublines expressing either wild type or non-cleavable isoforms of Stat5. Extracts were prepared from cells before (Ϫ, i.e. starved) or after (؉) stimulation with IL-3, fractionated on 8% SDS-PAGE, and immunoblotted with a Stat5 phosphotyrosine-specific antibody (Upstate Biotechnology Inc.). Lines A1 and A2 represent clones expressing Stat5b m/m (lanes 5-9). Line B1 expresses a control Stat5b wt (lanes 9 and 10). Line D3 represents a neomycin-resistant vector alone control (lanes 3 and 4). Controls for truncated (FdTrk) and full-length (rStat5b) activated Stat5b are provided in lanes 1 and 2. The mobility of p94, p80 (cleaved Stat5b), and p77 (cleaved Stat5a) are indicated in the margins. B, extracts detailed in A were evaluated by immunoblotting with a panStat5 antibody (Signal Transduction Laboratory). ture myelocytes express a protease that alters the biological response to Stat5-activating ligands. Concordant with this hypothesis, immature myeloid cells exhibit a distinct response to the IL-3 family of ligands (i.e. proliferation and maturation versus the terminal differentiation and activation found in mature cells). Although Stat5 also transduces signals for members of the IL-2 family that are important for lymphoid maturation and activation, there is no evidence that the truncated Stat5 isoforms play a role in this lineage.
Biochemical characterization of the Stat5 protease has highlighted a number of unique features of this protein. First, purification through four sequential steps has yielded an active fraction with a single protein of ϳ25 kDa. This is consistent with results from our size exclusion chromatography studies, indicating the protease is between 10 and 40 kDa. Moreover, a faint ϳ25-kDa band was recovered from the most active fractions of this column. Although 25 kDa is small for a protease, a number of small proteases have been identified. This is especially the case in viruses where their small genomes restrict the size of gene products. The smallest viral protease, ϳ100 amino acids, functions as a dimer, setting a lower limit of ϳ200 amino acids (55)(56)(57). Consistent with this, the smallest cellular serine proteases usually just exceed 200 amino acids (e.g. 25-30 kDa (58 -60)). Additionally, sequence comparison between related proteases often identify a conserved region of 200 -250 amino acids (58). A second feature, gleaned from purified preparations of the Stat5 protease, is that it is active over a wide range of physiological pH values. A third feature is that the protease exhibits a highly restricted and unique pattern of sensitivity to protease inhibitors. This pattern indicates that the protease is an unusual member of the serine family of proteases. A fourth feature is that the protease cleaves both inactive and active isoforms of Stat5, suggesting it may function constitutively, i.e. cleave Stat5 in unstimulated cells. Whereas the identification of both full-length and truncated isoforms in unstimulated FdTrk and DA-3 cells supports this observation (19), recently published studies suggest that the protease is associated with the nucleus (41). Such a sub-localization could effectively restrict access of the protease to activated Stat5. These possibilities are currently under investigation.
Purified preparations of the protease were also employed to map the site at which Stat5 is cleaved. Mass spectroscopy studies clearly placed the cleavage site between Tyr-719 and Met-720 in Stat5a and Tyr-724 and Met-725 in Stat5b. This was confirmed by mutating the two residues flanking the cleavage site and demonstrating that this Stat5b is resistant to cleavage. The biochemical features of this mutant (i.e. its ability to become activated, dimerize, and bind DNA and its comigration with wild type Stat5b) suggested that it did not represent a significant derangement of STAT structure. This was further supported by the observation that a more limited mutant (i.e. Stat5b m1 ; Y724A) was similarly resistant to cleavage. Additional mutations failed to provide any further insight in the specificity of the cleavage site. Yet it is unlikely that a single tyrosine defines the cleavage site. After all, wild type Stat5 is not cleaved at any other tyrosine. Perhaps the "nonspecific" partial cleavage products observed with Stat5b m/m (Y724S2M725A; see Fig. 8B) represent secondary or default sites that are defined by the presence of the "other tyrosines." Consistent with this possibility, this nonspecific pattern of cleavage appears to require both loss of the primary cleavage site and more purified preparations of protease. Although it is possible that these purified preparations may also have lost some substrate specificity, we do not believe this to be the case.
Biochemical studies have led to the identification and characterization of Stat5 protease activity that is active in myeloid progenitors, supporting a hypothesis that this activity is required for normal myeloid development. Consistent with this hypothesis, we have previously shown that Stat5 target genes, identified in more mature cells (i.e. where the full-length isoform transduces signals), fail to be induced in immature myeloid lines (19). In the current study we demonstrate that expression of a non-cleavable isoform of Stat5 in FDC-P1 (cl. 19) cells leads to the translocation of activated full-length Stat5 to the nucleus. This corresponds with an important morphological change in these transfectants, that is cells expressing significant levels of non-cleavable Stat5 exhibit a partially differentiated phenotype. These studies indicate that regulation of Stat5 activity is important for normal myeloid development.