Human proteinase-3 expression is regulated by PU.1 in conjunction with a cytidine-rich element.

Human proteinase-3 is one of three serine proteinases present in the azurophil granules of polymorphonuclear leukocytes along with elastase and cathepsin G. Proteinase-3 gene expression is confined to the promyelocytic stage of polymorphonuclear leukocyte maturation. The present investigation identifies elements responsible for this highly controlled tissue- and developmental-specific expression of proteinase-3. Within the first 200 base pairs of the proteinase-3 promoter, two elements were identified as important for expression, these elements at −101 and −190 confer the majority of the activity. The element at −101 has a PU.1 consensus. It binds a myeloid nuclear protein of approximately 45 kDa that “supershifts” with PU.1 antibody and is competed by the CD11b PU.1 element. The element at −190 has a core sequence of CCCCGCCC (CG element). The cytidines but not the guanidine are essential for promoter activity. The CG element binds a second nuclear protein with a molecular mass of approximately 40 kDa that is found in cells of myeloid lineage as well as non-myeloid HeLa cells. However, the proteinase-3 promoter is not active in HeLa cells which suggests that the CG element alone is not sufficient for proteinase-3 gene expression. Maturation of promyelocytic cells results in an inhibition of proteinase-3 gene expression and a reduction in nuclear protein binding to the PU.1 and CG elements. Similar elements occur in the elastase and cathepsin G promoters. Using the elastase and cathepsin G PU.1 and CG-like elements as probes results in identical band-shift patterns to that obtained with proteinase-3 PU.1 and CG elements. These data suggest that there is cooperative interaction between a PU.1 and a CG element with a consensus of CCCCXCCC and that they are important control elements for tissue- and developmental-specific expression of azurophil serine proteinases of polymorphonuclear leukocytes.

Proteinase-3 (PR-3, 1 synonyms: myeloblastin and azurophil granule protein 7; AGP-7) a neutral serine proteinase with a broad spectrum of proteolytic activity is stored in the azurophil granules of polymorphonuclear leukocytes (PMNL) (1,2). PR-3 is similar to but distinct from two other azurophil serine proteinases, human leukocyte elastase (HLE) and cathepsin G (Cat G). Physiologically, PR-3 may assist in the killing (3) and digestion (4) of bacteria and in the movement of PMNL through the basement membrane at sites of inflammation (5). Pathologically, since PR-3 degrades connective tissue proteins (6 -8), it is potentially an important factor in destructive inflammatory diseases such as emphysema (1), the adult respiratory distress syndrome, rheumatoid arthritis, and glomerulonephritis (9). PR-3 cleaves the nuclear factor-B subunit (10) and also degrades the 28-kDa mammalian heat shock protein (11), previously linked to the differentiation of normal and neoplastic cells. As myeloblastin, PR-3 was implicated in the growth and differentiation of leukemic cells (12,13). Perhaps most importantly, PR-3 is the antigen recognized by cytoplasmic staining by anti-neutrophil cytoplasmic antibodies present in patients with Wegener's granulomatosis (14 -16).
PR-3 is found in cells of myeloid lineage, and its mRNA synthesis is confined to the promyelocytic stage of maturation. Upon differentiation beyond the promyelocytic stage, PR-3 is coordinately down-regulated along with HLE and Cat G.
The human gene for PR-3 has been characterized (17) and localized to chromosome 19 p13.3 (17), within 3 kb of HLE and 8 kb of azurocidin (an homologous protein with bactericidal but no proteolytic activity) forming a second serine proteinase gene cluster (18). A Cat G family of serine proteinases is located on chromosome 14 q11-12 (19,20).
Because of its important physiologic and pathologic roles, understanding the mechanisms of PR-3 gene expression is fundamental to the control of PR-3 expression and to the development of new therapeutic strategies for Wegener's granulomatosis or emphysema. In this report we examined the human PR-3 promoter and identified two elements, a PU.1 binding site and a cytidine-rich site, important in PR-3 gene expression. A PU.1 site and a cytidine-rich site are also present in the promoters of HLE (21) and Cat G (22). These two elements appear to constitute major regulatory regions for the lineage and stage-specific expression of azurophil serine proteinases.
Sequencing of the Human PR-3 5Ј-UTR-The human PR-3 clone was a 40-kb human genomic DNA insert isolated by hybridization with human PR-3 cDNA probes. The cosmid library was constructed with DNA from normal peripheral blood lymphocytes and was a generous gift from Dr. R. Lemons. This clone contained 680 bp of the 5Ј-upstream region of human PR-3. Sequence analysis of both strands was performed with the dideoxy chain termination method (23). Transcription factor binding sites were identified by searching the Gosh data base of previously identified sites (24).
Cell Culture-U937, HL-60, PLB 985, and K562 cell lines were grown in RPMI 1640 supplemented with sodium pyruvate, non-essential amino acids, 10% fetal calf serum, and antibiotics to a density of 10 6 cells per ml. Differentiation of U937, HL-60, and PLB 985 toward monocytic cells was obtained by adding 100 ng/ml phorbol myristate acetate (PMA) for 72 h. This dose of PMA was not toxic. The cells were grown at high densities (1-5 ϫ 10 6 cell/ml). Differentiation of HL-60 and PLB 985 cells toward granulocytes was achieved by growing the cells in retinoic acid (10 Ϫ6 M) for 24 h and then supplementing with dimethylformamide (DMF, 60 mM) for an additional 48 h (25).
Transient Transfection of the PR-3 5Ј-UTR-pGL3 luciferase reporter vectors were used for assay of promoter activity. Fragments were generated using PCR and cloned into the BglII site of the pGL3 basic vector that contains no promoter or enhancer. All constructs were sequenced to determine correct insertion and proper orientation for functional promoter activity and to ensure no PCR-introduced errors. Specific mutations were constructed using PCR to insert primers containing the desired mutation. The PCR-derived fragments were cloned into pGL3 basic vector and sequenced to ensure that the mutation was correct and that no PCR-derived errors had occurred. Plasmid DNA for transfection was prepared using Qiagen tip-500 columns and frozen at Ϫ80°C in 20-g aliquots to prevent repeated freezing and thawing and to minimize handling. The DNA was not linearized prior to transfection. Three separate plasmid preparations were used for each construct to ensure that any change in efficiency was not a result of variation in the quality of DNA. The pGL3 control vector with promoter and enhancer activity was used to optimize transfection conditions for each cell line and as a positive control in each experiment. Using the pGL3 control vector, the most efficient conditions for transfection of K562 cells were 10 7 cells/ml and 240 V at 1180 F using a Cell-Porator and 0.4-cm cuvettes. Preparation of the cells and electroporation was as described by Baum et al. (26). The pSV ␤-galactosidase plasmid (10 g) was co-transfected with 20 g of the experimental pGL3 construct. Twentyfour hours after transfection the cells were lysed in 1 ϫ lysis buffer, and luciferase activity was measured. ␤-Galactosidase enzyme activity and the protein content of each sample were determined to correct for transfection efficiency and cell number. At least three independent transfections in triplicate were carried out to determine the reporter activity of the different constructs.
Preparation of Nuclear Extracts-Nuclear extracts were prepared from U937, HL-60, PLB 985, K562, and HeLa cells. For each extract 10 8 cells/ml were harvested by centrifugation and washed twice with icecold phosphate-buffered saline. Nuclear proteins were obtained as described by Digman et al. (27) with the addition of proteinase inhibitors as follows: 1 mM phenylmethylsulfonyl fluoride, 1 g/ml pepstatin A, 0.5 g/ml chymostatin, 1 g/ml antipain, 1 g/ml leupeptin, and 4 g/ml aprotinin. Protein concentrations of the extracts were determined spectroscopically using the Bio-Rad protein reagent. Aliquots of the nuclear extracts were frozen at Ϫ80°C to prevent repeated freezing and thawing. Electrophoretic Mobility Shift Assay (EMSA)-The following probes were generated. Initially, consecutive fragments (60 -80 bp) of the entire 680-bp 5Ј-upstream sequence of human PR-3 gene were generated using specific primers and PCR and analyzed by EMSA. Coupled with deletion analysis using reporter gene constructs, "hot-spots" were identified. Wild type and mutant constructs of these hot-spots were analyzed using duplex oligonucleotides as probes. Sense and antisense probes were annealed immediately prior to use by heating to 95°C for 5 min and returning to room temperature over 30 min. After annealing, the duplex was labeled with [␥-32 P]ATP using T4 polynucleotide kinase (Promega). Free radionucleotide was removed using a Sephadex G-25 column.
Each probe (0.5 ng) was incubated with 10 g of nuclear extract in 20 l containing a final concentration of 10 mM Hepes, pH 7.5, 50 mM KCl, 5 mM MgCl 2 , 1 mM dithiothreitol, 1 mM EDTA, and 5% glycerol plus 0.5 or 5 g of poly(dI-dC) to reduce nonspecific binding. In some experiments unlabeled competitor probes were added to the nuclear extracts immediately prior to the addition of the radionucleotide probe. Incubations were carried out for 20 min at room temperature. Reactions were electrophoresed at 14 V/cm for 1 1 ⁄2-2 h on a 6% nondenaturing polyacrylamide gel cast in 0.5 ϫ TBE (45 mM Tris borate, 45 mM boric acid, 1 mM EDTA) at 4°C. Antibody supershifts were performed by adding 1 l of a PU.1 rabbit antiserum against an amino-terminal peptide (a generous gift from Dr. R. Maki) or 1 l of control rabbit serum to the nuclear extract and preincubating on ice for 15 min. The radiolabeled probe was then added, and the reaction was incubated for a further 15 min on ice. The complex was electrophoresed as above.
Molecular Weight Assessment of Transacting Proteins-The method was essentially that of Wu et al. (28). Labeled hot-spot probes were incubated with nuclear protein extracts as above and then run on a 1% low-melting agarose gel until the bromphenol dye front has migrated about 5 cm. The gel was wrapped in Saran Wrap, placed upside-down on a UV transilluminator, and covered with ice to prevent overheating. The gel was irradiated for 7 min with UV light (300 nM) and then autoradiographed. The radiolabeled bands were excised, 10 l of 2 ϫ Laemmli buffer was added to each excised band, and the samples were boiled for 2 min and then while warm directly loaded onto an SDSpolyacrylamide gel electrophoresis gel (Ref. 29; 7.5% running gel and 4% stacking gel) prior to adding the electrophoresis buffer. After elec- trophoresis the gel was fixed (40% methanol, 10% acetic acid), dried, and autoradiographed.
mRNA Expression of Azurophil Serine Proteinases-Total RNA was prepared from myeloid cells by the method of Chomczynski and Sacchi (30). The RNA concentration was determined spectrophotometrically, and 5 g were used for reverse transcription employing a standard protocol with Moloney murine leukemia virus reverse transcriptase. Excess RNA was digested with 2 g of DNase-free RNase and incubated for 5 min at 37°C. The reaction was extracted with phenol/chloroform and precipitated with ethanol at Ϫ20°C overnight. The cDNA concentration was spectrophotometrically determined. Semi-quantitative PCR was performed by using a known amount of cDNA per reaction and analyzing the radioactive product on a polyacrylamide gel. Optimal cDNA amplification and number of cycles for amplification were determined by titration from 1 to 500 ng of cDNA and from 18 to 40 cycles. Optimal parameters were determined to be 200 ng of cDNA for 20 cycles. PCR buffer containing Mg 2ϩ (Perkin-Elmer) and dNTP concentrations of 100 M were used plus 0.25 Ci of [ 32 P]dCTP. For consistency of samples, a master mix for each set of primers was prepared. Reactions of 25 l were amplified, and the PCR conditions were as follows: denaturation at 94°C for 15 s, annealing at 59°C for 15 s, and elongation at 72°C for 30 s. Following PCR an aliquot was added to an equal volume of DNA sample buffer, heated to 95°C for 5 min, and electrophoresed in a 6% acrylamide gel. Bands were detected by autoradiographic exposure and compared with each other and against amplified ␤-actin as an internal control.

Human PR-3 5Ј-Upstream Sequence and Putative Regulatory
Elements-We sequenced 680 bp upstream from the initiation ATG of human PR-3 (Fig. 1). Analysis of the sequence revealed the following putative elements. There is a TATA box at position Ϫ44 but no CAAT box. A PU.1 regulatory element is at position Ϫ101. This member of the ets family has been shown to be important in myeloid cell development (31) and is contained in a 30-bp regulatory fragment of the HLE promoter identified by Srikanth and Rado (32). The core element of the ets family, GGAA (33), also occurs at Ϫ366 and Ϫ418 on the complementary strand. The immediate 5Ј-region contains, in addition to the PU.1 binding site (Ϫ101) and the TATA box (Ϫ44), a potential binding site for the CAAT binding (C/EBP Ϫ82) and Myb (Ϫ56) transcription factors. Other recognized elements include a NF-B element at Ϫ606 on the antisense strand (35) and five ␤-globin-specific elements at Ϫ361, Ϫ539, Ϫ565, and Ϫ622. Three of the five ␤-globin elements (Ϫ361, Ϫ539, and Ϫ565) are preceded by a T forming the retinoic acid regulatory element, TCACC (36).
Identification of Cis-acting Elements in the PR-3 Promoter-To analyze the 5Ј-upstream region of the PR-3 gene for promoter activity, we cloned step-deletion constructs into pGL3-luciferase (Promega) reporter plasmids and transiently transfected them into myeloid (K562 and U937) or non-myeloid (HeLa) cells. K562 cells were used initially in the experiments because of their greater efficiency of transfection. The results were confirmed with U937 cells. K562 cells like U937 cells are derived from promyelocytic cells from a patient with chronic myelogenous leukemia (37). The data presented ( Fig. 2A) is from one experiment with four determinations of each fragment. The experiment was repeated six times using three different plasmid preparations. Luciferase activity was consistently greatest in fragment Ϫ212 to Ϫ0, with fragments Ϫ174 to Ϫ0 and Ϫ91 to Ϫ0 showing the second and third highest activ- ity levels, respectively. In the fragments from Ϫ284 to 0 to Ϫ680 to 0 the promoter activity was reduced by at least 75% compared with fragment Ϫ212 to 0 ( Fig. 2A) suggesting the presence of a suppressor element between Ϫ484 and Ϫ212. Fragment Ϫ174 to 0 contains a PU.1, a C/EBP, and a Myb element as well as a TATA box; Ϫ91 to 0 contains the C/EBP and MYB sites plus the TATA box but not the PU.1 element. These results suggest that in addition to the Ϫ91 to 0 fragment that contains putative binding sites for C/EBP, Myb, and TATA elements, a PU.1 element and an unidentified element between Ϫ212 and Ϫ174 are required for maximum promoter activity.
To determine whether the response was specific to cells of myeloid origin, HeLa cells were transiently transfected by electroporation with the same constructs. In these non-myeloid cells neither the PU.1 containing fragment (Ϫ174 to 0) nor the Ϫ212 to Ϫ0 fragment demonstrated promoter activity (Fig. 2B) implying that HeLa cells lack some or all of the trans-activating factors that are present in myeloid cells and necessary for PR-3 promoter activity.
In order to localize the unidentified sequence within the Ϫ212 to Ϫ174 fragment that conferred promoter activity when transiently transfected into myeloid cells, constructs were made that were sequentially reduced by 6 bp at the 5Ј-end. The result presented in Fig. 3A is representative of four separate experiments where four determinations of each fragment were made. The promoter activity of fragment Ϫ188 to 0 was 50% less than the Ϫ212 to 0 fragment but was equal to the level of promoter activity of the PU.1 containing fragment (Ϫ174 to 0). Moreover, the 50% reduction in promoter activity was sharply delineated and occurred between base pairs Ϫ194 and Ϫ188 (data not shown) which implied that an important element was present within these six base pairs. Compared with the Ϫ194 to 0 fragment, the Ϫ188 to 0 fragment removed two 5Ј cytidines of a four-cytidine sequence suggesting that all four cytidines were essential. These cytidines formed part of the sequence CCCCGCCC that was similar to the sequence CCCCACCC present in the promoters of HLE and Cat G (21,22). We therefore concluded that this CG sequence was likely to be the essential element for the additional promoter activity. The PU.1 element and the CG element require the additional pres-

FIG. 3. PR-3 promoter activity lies within the first 212 bp of the 5-UTR.
A, localization of the sequence conferring additional promoter activity within fragment Ϫ212 to Ϫ174. K562 cells were transiently transfected with pGL3 luciferase constructs (20 g) that were sequentially reduced by 6 bp at the 5Ј-end. All constructs terminated at 0 at the 3Ј-end. pSV ␤-galactosidase (10 g) was co-transfected as an internal control for transfection efficiency. Luciferase activity was determined 24 h after transfection. The corrected activity and standard error of the mean of four determinations for each fragment is reported as a percentage of the maximal activity seen with Ϫ212 luciferase construct; 2029 Ϯ 123 luciferase units/5 ϫ 10 6 cells. B, the presence of the Ϫ91 to 0 fragment is essential for the promoter activity. pGL3 luciferase constructs (20 g) containing PU.1 and CG elements with or without the C/EBP, Myb, and TATA elements were transiently transfected in K562 cells along with pSV ␤-galactosidase (10 g). The mean of the corrected luciferase activity from four determinations is reported as a percentage of the maximal activity seen with Ϫ212 to 0 construct that contains at least five elements; 2519 Ϯ 59 luciferase units/5 ϫ 10 6 cells. ence of the elements within the Ϫ91 to 0 fragment (Fig. 3B).
Characterization of Cis and Trans Factors Important in PR-3 Gene Expression-To further characterize the cis-acting ele-ments and to identify the trans-acting factors that bind these elements, electrophoretic mobility shift assays (EMSAs) were performed. Crude nuclear protein extracts were prepared from HL-60, U937, PLB 985, K562, and HeLa cells. Using crude nuclear protein extracts from HL-60 cells, strong shifts occurred with fragments Ϫ212 to Ϫ91 and Ϫ174 to Ϫ91. Both fragments contain a PU.1 element, but the larger fragment possesses both a PU.1 element and a CG element identified above, and two shifts can be observed with this larger fragment. The remaining 5Ј-fragments upstream of the CG element showed only nonspecific shifts. Nuclear protein extracts from U937 and PLB 985 cells gave identical shifts to HL-60 extracts with fragments Ϫ212 to Ϫ91 and Ϫ174 to Ϫ91 (data not shown).
To identify the trans-acting factors binding to PR-3 PU.1 and CG elements, we used radiolabeled double-stranded oligonucleotide probes of these sites in EMSAs (PR-3 PU.1 element: 5Ј-AGGCAAAAGGAGGAAGTGGGGACCCAGCCT; PR-3 CG element 5Ј-CAGCCAGCCTCCCCGCCCCCACAAAGGTGG). The PR-3 PU.1 element gave a two-band (A and A * ) shift when incubated with myeloid nuclear protein extracts (Fig. 4). This is typical of PU.1 when incubated with nuclear extracts from cells with high proteolytic content, A * being a proteolytic cleavage product of band A (38). In this experiment, with low (0.5 g) poly(dI-dC), the PR-3 CG element gave an intense band (band B) with a faster electrophoretic mobility than PU.1, suggesting that this element binds a protein of lower molecular weight. Under low stringency (0.5 g of poly(dI-dC)) the PR-3 CG element also gave a shift similar to that obtained with the PR-3 PU.1 element, but the shift was completely abolished by increasing the poly(dI-dC) to 5 g per reaction. Band B occurred in the presence of high poly(dI-dC) indicating a strong binding reaction, whereas the binding affinity of PU.1-like protein to FIG. 4. Identification of myeloid factors that bind PR-3 PU.1 and CG elements. Nuclear protein extracts (10 g) from K562 (lanes 1-6) and PLB 985 cells (lanes 7-12) were used in an EMSA to identify myeloid proteins that bind these elements. A 30-bp oligonucleotide from bp Ϫ110 to Ϫ80 of the PR-3 5Ј-UTR that includes the PU.1 element was used as the radiolabeled probe in lanes 1-3 and 7-9, and a 30-bp oligonucleotide from bp Ϫ200 to Ϫ170 of the PR-3 5Ј-UTR that includes the CG element was used as the probe in lanes 4-6 and 10-12. Lanes 1, 4, 7, and 10 are controls containing probe and no nuclear extract or unlabeled competitor. Lanes 2, 5, 8, and 11 contain labeled probe and 10 g of nuclear protein extract. Lanes 3, 6, 9, and 12 contain in addition to labeled probe and nuclear extract a 100-fold excess of the same probe unlabeled as competitor. Bands that represent specific protein binding to the probes are indicated with arrows and labeled A, A*, and band B.  (Ϫ110 to Ϫ80 bp, lane 2). Lane 1 is a control and only contains radiolabeled probe. Lane 3 contains in addition to the radiolabeled PR-3 PU.1 probe and U937 nuclear extract a 100-fold excess of unlabeled double-stranded oligonucleotide containing a PU.1 binding site from CD11b (38). Lane 4 is the same as lane 3 except the unlabeled competitor is the PU.1 binding site from the SV40 enhancer (39). Free probe is indicated at the bottom of the gels. the PR-3 CG element was weaker. The PR-3 CG oligonucleotide used in the experiments contained the sequence AAAGGTGG at the 5Ј-end which is similar to the PU.1 consensus. We, therefore, designed a new CG oligonucleotide without this potential PU.1 binding site (GTGGGTGACAGCCAGCCTC-CCCGCCCCCAC). This modified PR-3 CG element under low stringency also gave a shift similar to that obtained with the PR-3 PU.1 element as well as the specific band B shift, and like the original PR-3 CG element this shift was abolished by increasing the poly(dI-dC) to 5 g per reaction (data not shown).
The binding of labeled PR-3 PU.1 or PR-3 CG probes to their specific proteins was totally inhibited by addition of a hundredfold excess of unlabeled probe (Fig. 4, lanes 3 and 6). The data presented in Fig. 4 were obtained with K562 nuclear protein extracts (lanes 1-6) and with PLB 985 nuclear protein extracts (lanes 7-12), and the results were identical. U937 and HL-60 nuclear protein extracts also exhibited identical shifts with the PU.1 and CG sites (data not shown).
To confirm that the shift of the PR-3 PU.1 probe was due to the binding of PU.1 protein, we used an antibody that specifically recognizes the amino-terminal end of PU.1. The binding complex formed when PR-3 PU.1 probe was incubated with myeloid nuclear protein extracts "supershifted" upon addition of the antibody (Fig. 5A). Moreover, the binding of PR-3 PU.1 probe to PU.1 protein from myeloid nuclear extracts could be blocked with a double-stranded oligonucleotide containing a PU.1 binding site from CD 11b (39) and from the SV40 enhancer ( Fig. 5B; 40). These results confirm that the protein binding to PR-3 PU.1 probe was PU.1.
The specific shift obtained when the CG probe was incubated with myeloid nuclear protein extract (band B) did not "supershift" upon addition of the PU.1 antibody, indicating that a second protein was responsible for this shift. UV-cross-linking of the specific proteins that bind to the PR-3 PU.1 and CG elements, followed by separation on an SDS-polyacrylamide electrophoresis gel, gave sizes of approximately 45 and 40 kDa, respectively (Fig. 6). Because the sequence CCCCGCCC is complementary to an Sp1 binding element, it is possible that the protein complexing with PR-3 CG element is Sp1. However, cold competition with a 100-fold excess Sp1 consensus sequence (Promega) did not inhibit the formation of band B when PR-3 CG element was incubated with myeloid nuclear extracts. Moreover, the addition of Sp1 protein (Promega) did not result in a shift when incubated with the PR-3 CG element. Finally, Sp1 protein has a molecular mass between 95 and 105 kDa depending on glycosylation (41) which is much larger than the 40-kDa protein that bound to the CG element determined by UV-cross-linking. Thus the complex formed when the PR-3 CG element was incubated with myeloid nuclear extracts was not due to Sp1 binding. It did not appear to be myeloid-specific as EMSA studies with PR-3 CG element and nuclear extracts from HeLa cells gave results identical to those with myeloid nuclear extracts, compared with no shift with the PR-3 PU.1 element incubated with HeLa nuclear extracts (data not shown).
The CG Element Consensus Is CCCCXCCC-Other myeloid genes contain sequences similar to the PR-3 CG element (CCCCGCCC). For example, the sequence CCCCACCC is found in the promoters of HLE (21), Cat G (22), and myeloperoxidase (42). In other myeloid genes the central A is substituted by a T (42)(43)(44)(45). This implies that the central non-cytidine nucleotide can either be an A or a T. To investigate the relevance of a G instead of A or T and to confirm the importance of the flanking cytidines, we constructed three mutations of the PR-3 CG element: CG mut 1 was identical to wild type PR-3 CG element with the exception of a 1-bp substitution of the G at position 15 to an A (and A to G mutation is conservative). CG mut 2 substitutes the 2 Cs at positions 11 and 12 for 2 As which disrupted the 5Ј four Cs of the core element. CG mut 3: substituted the 2 Cs at positions 18 and 19 which disrupted the 3Ј three Cs of the core element.
The three mutants and wild type constructs were analyzed by reporter gene expression and EMSA. In transient transfection studies CG mut 1 had the same activity as wild type, but in CG mut 2 and 3 the activity was reduced to the level of the constructs that contain only PU.1 and TATA elements, thus abolishing completely the increased activity due to the PR-3 CG element (Fig. 7A). The effect of the mutations on protein binding is shown in Fig. 7B. Lane 4 confirms that the substitution of G for an A did not have an effect on the binding of the protein that constitutes band B compared with wild type. Mutation of Cs at positions 11 and 12 to As totally abolished binding (Fig. 7B, lane 6), whereas mutation of Cs at positions 18 and 19 to As considerably reduced band B and increased the binding of PU.1 protein to this probe (Fig. 7B, lane 8). Additional experiments showed that CG mut 3 was also able to completely abolish band B. The use of CG mut 1, 2, and 3 as cold competitors (100-fold excess) for the CG wild type probe confirmed that substitution of G to A was unimportant as CG mut 1 competed with CG wild type, whereas CG mut 2 and mut 3 did not (data not shown). Thus, the cytidines are essential whereas the central non-cytidine nucleotide can be an A, T, or a G. The consensus sequence for this element is, therefore, CCCCXCCC.
Expression of the protein that binds to the CG element is not confined to myeloid cells since EMSA studies using nuclear extracts from HeLa cells showed an identical shift to that obtained with myeloid cells with the CG element. There was no shift of the PR-3 PU.1 element with nuclear extracts from HeLa cells (data not shown). (18). Therefore transcriptional factors involved in regulating the PR-3 gene may show a similar stage-specific expression. Differentiation of HL-60, U937, and PLB 985 cells to monocytes with PMA and of HL-60 and PLB 985 cells to granulocytes with retinoic acid plus DMF inhibited PR-3 mRNA expression and protein production. Retinoic acid plus DMF resulted in more efficient granulocytic maturation than either factor alone (25), and reverse transcriptase-PCR analysis indicated that PR-3 gene expression was reduced by more than 80% following this treatment (data not shown).

Relevance of PU.1 and CG Elements to PR-3 Gene Expression-PR-3 gene expression is stage-specific being confined to the promyelocytic stage of granulocyte maturation
In order to determine whether the PU.1 and CG transcriptional elements are similarly reduced upon myeloid differentiation, we prepared nuclear protein extracts from myeloid leukemic cells stimulated to differentiate with PMA or retinoic acid plus DMF and compared their ability to shift PR-3 PU.1 and PR-3 CG probes with extracts from untreated cells. Differentiation of U937, HL-60, and PLB 985 to monocytes with PMA reduced the degree of binding to both the PU.1 and CG probes (Fig. 8, A and B). In Fig. 8A PMA failed to reduce the binding of U937 nuclear proteins to the CG element. Further experiments with PMA-treated U937 cells did show a reduction in binding to the CG element (Fig. 8). A similar reduction in binding was obtained when PLB 985 cells were treated with retinoic acid and DMF. Conversely the treatment of K562 cells with PMA had no effect (Fig. 8A). In these experiments nuclear protein extracts of all the samples were prepared at the same time to minimize experimental variation. Ice-cold buffers containing a mixture of proteinase inhibitors (see "Experimental Procedures") were used to reduce degradation.
Additional experiments were carried out to confirm that PMA treatment inhibited PR-3 gene expression under the same conditions that reduced the degree of nuclear protein binding to PU.1 and CG probes. Immediately prior to nuclear protein extraction, an aliquot of each sample was taken and analyzed for PR-3 gene expression using reverse transcriptase-PCR. In an experiment using U937 cells treated with 1 and 10 ng/ml PMA for 48 h, only the higher concentration reduced nuclear protein binding to the PR-3 PU.1 probe and switched-off PR-3 gene expression. In K562 cells where PMA does not affect nuclear protein binding to the PU.1 and CG probes (Fig. 8A), there is also no switch-off in PR-3 gene expression. These data suggest that both the PU.1 and CG sites in the PR-3 promoter are intimately involved in PR-3 expression.
Relevance of PU.1 and CG Elements to the Expression of Other Azurophil Serine Proteinase Genes-PR-3 is one of three serine proteinases present in the azurophil granules of polymorphonuclear leukocytes along with HLE and Cat G. We determined by PCR analysis coordinate mRNA expression of PR-3, HLE, and Cat G genes. The specific primers were designed to generate fragments of non-homologous regions to ensure no cross-reactivity among the related genes. Differentiation of U937 cells with PMA resulted in a loss of 80% within 8 h of all three serine proteinase transcripts (Fig. 9).
Because PR-3, HLE, and Cat G genes are coordinately expressed, we hypothesized that elements similar to PU.1 and CG play a role in regulating all three genes. Analysis of the promoters of HLE and Cat G compared with PR-3 reveals that the three genes have a CG-like element (CCCCXCCC ; Table I) and a PU.1 element within the first 300 bp of the 5Ј-upstream sequence; moreover, the CG-like element is always 5Ј of the PU.1 site with at the most 60 intervening base pairs.
We synthesized duplex oligonucleotides of the 30 bp flanking HLE and Cat G PU.1 and CG-like elements and compared their EMSA with PR-3 PU.1 and CG elements. The three PU.1 sites (Fig. 10A) and the three CG-like elements (Fig. 10B) showed

FIG. 7. Mutation of the CG element significantly reduces PR-3 promoter activity and the specific protein binding to form band B.
A shows the effect of mutations within the CG element on luciferase activity. Transient transfection of PR-3 promoter fragment containing five (CG, PU.1, C/EBP, Myb, and TATA), four (PU.1, C/EBP, Myb, and TATA), and three (C/EBP, Myb, and TATA) promoter elements (upper four histograms) were compared with fragments that contained mutations of the CG element with wild type PU.1 and TATA sites (bottom three histograms). Ϫ194 mut 1 is a single base substitution of G to A at Ϫ186. Ϫ194 mut 2 is a 2-bp substitution of cytidines (Ϫ190 and Ϫ189) for two adenosines. Ϫ194 mut 3 is a 2-bp substitution of cytidines Ϫ183 and Ϫ182 for two adenosines. Luciferase activity was measured 24 h after transfection into K562 cells. The corrected activity and standard errors of the mean of four transfections are expressed as a percentage of activity of fragment Ϫ212 to O; 4 030 Ϯ 180 luciferase units/5 ϫ 10 6 cells (top histogram). B shows the effect of mutations within the CG element on the specific binding of a nuclear protein to form band B. Wild type PR-3 CG oligonucleotide probe (Ϫ200 to 170) and three mutations were incubated with U937 nuclear protein extracts (10 g) and analyzed using EMSA. The specific binding to form band B is indicated by an arrow as is the free probe at the bottom of the gel. Lanes 1, 3, 5, and 7 are the controls for each probe and contain no nuclear extracts. In lanes 1 and 2 the probe is the wild type PR-3 CG oligonucleotide probe. In lanes 3 and 4 the probe is PR-3 CG mut 1 that is identical to the wild type except for the substitution of the G at position 15 for an A. In lanes 5 and 6 the probe is PR-3 CG mut 2 that is identical to wild type except for the subtitutions of two Cs at positions 11 and 12 for two As. In lanes 7 and 8 the probe is PR-3 mut 3 that is identical to wild type except for subtitutions of two Cs at position 18 and 19 for two As.
identical shifts when incubated with nuclear protein extracts from U937 cells. The Cat G PU.1 element consistently gave an identical pattern of binding, but the degree of binding was always less than that with either the PR-3 or HLE PU.1 element. These results indicate that both PU.1 protein and the unknown protein binding to CG-like elements to form band B are likely common control factors for gene regulation of the three azurophil neutral serine proteinases. DISCUSSION In the present investigation analysis of 680 bp of human PR-3 promoter indicates that the first 200 bp is sufficient to give maximal and myeloid-specific expression of a reporter gene. Seventy-five percent of the activity is conferred by a CG element (Ϫ190) plus a PU.1 element (Ϫ101). This activity, however, is dependent on the presense of the Ϫ91 to 0 fragment that contains potential sites for C/EBP and Myb in addition to a TATA site. Interestingly a TATA site is present in all neutrophil azurophil granule serine proteinase promoters (17,18,20,21,42) but is not present in almost all the other myeloid genes including CD11b (38), CD18 (43), CD11a (44), and the CSF receptor genes (62).
The Ϫ91 to 0 fragment confers 24% promoter activity compared with 100% for the Ϫ212 to 0 fragment. The Ϫ91 to 0 fragment contains a potential binding element for C/EBP (Ϫ82), Myb (Ϫ56), and TATA (Ϫ44). All three sites are present in the immediate 5Ј-promoters of HLE, mouse neutrophil elastase, and azurocidin (18,34). There is no C/EBP or Myb site, only a TATA site in the immediate 5Ј-region of the Cat G promoter (20). Thus C/EBP and Myb sites are not ubiquitous to all azurophil serine proteinase gene promoters, and consequently these sites are unlikely to be major elements in controlling coordinate expression of azurophil serine proteinase genes ( Table I).
The element at Ϫ101 is a binding site for the PU.
plays a role in regulation of their expression in a myeloidspecific manner (38,(51)(52)(53)(54)(55)(56)(57)(58)(59). Previously a PU.1 transcription factor was reported to be essential for the myeloid-specific expression of HLE (32). A PU.1 binding sequence also is present within the first 200 bp of human Cat G (22) and human myeloperoxidase promoters (42). The results of the present investigation confirm and extend the previous studies by demonstrating PU.1 involvement in the expression of all neutral serine proteinases contained within the azurophil granules.
PU.1 is involved in multiple steps of myeloid development. In addition to its involvement in the expression of azurophil serine proteinase genes, PU.1 is functional in the promoters of M-CSF receptor (60), the GM-CSF receptor ␣ (61), and the G-CSF receptor (58) suggesting a role for PU.1 in early myelopoiesis as well as azurophil serine proteinase development. In the CSF receptor genes as in the azurophil serine proteinase genes, PU.1 is present within a 100 bp of the transcription start site (62). Thus the presence of PU.1 within azurophil serine proteinase promoters constitutes a second group of myeloid genes in which PU.1 plays a central and perhaps unique role.
In the present investigation, we identified a cis-acting element with a core sequence of CCCCGCCC that was essential for maximal expression of the human PR-3 gene. By substitution analysis we found that both clusters of cytidines were necessary for binding and promoter activity, whereas the central guanidine was not. Studying the promoter of CD11a Shelley et al. (44) identified a similar sequence (CCCCTCCC) that bound a specific protein termed MS-2. These authors demonstrated, in contrast to our data, that the 3Ј cytidines were non-essential and only the 5Ј cytidines complexed with nuclear proteins (44). The molecular mass of MS-2 is unknown and therefore cannot be compared with the CG-binding protein that has a molecular mass of approximately 40 kDa.
The sequence CCCCACCC is found in the promoters of several myeloid genes. It is present within 300 bp of the initiation site of azurophil myeloperoxidase (42), HLE (21), Cat G (22), and CD43 (44) genes. The sequence CCCCTCCC which is also similar to the CG element is present in the promoters of CD18 (43), CD11a and 11b (44), and aminopeptidase N (CD13; 45) genes suggesting that substitution of the central adenosine to a thymidine does not alter protein binding (Table II). The site in PR-3 is unique in substituting the non-cytidine residue to a guanidine. We therefore suggest that a consensus element CCCCXCCC is important in myeloid gene regulation.
The trans-acting factor that binds the PR-3 CG element is unknown. It is not PU.1 as the complex that has greater electrophoretic mobility than the PU.1 complex, and the CG binding protein is smaller than the PU.1 protein. It could, however, be Sp1 because the CG site is complementary to the Sp1 binding site. Despite being universally expressed (63)(64)(65), Sp1 is essential for myeloid-specific promoter activity in the CD11b gene (66). However, neither Sp1 protein nor a known Sp1 binding site (Promega) competed with the complex formed upon incubation of the PR-3 CG element with myeloid cell nuclear proteins. Moreover, Sp1 (95-105 kDa) is too large to be the CG binding protein. Another possible candidate for the factor that binds PR-3 CG element is a C2H2 zinc finger gene MZF-1 (67). The gene is preferentially expressed in myeloid cells and binds two core sequences one of which has the sequence 5Ј-CGGGnGAGGGGGAA-3Ј and is thus similar to a complementary sequence to the CG consensus. However, it is likely that MZF-1 with an approximate molecular mass of 48 kDa is too large to be the putative CG binding protein.  1 and 2), this sequence randomized (lanes 3 and 4), the HLE CG element (lanes 5 and 6), and the Cat G CG element (lanes 7 and 8). The specific binding is marked by an arrow and labeled band B. Free probe ran at the bottom of the gel.
The CG-like element in azurophil neutral serine proteinase genes is always upstream and within close proximity (within 60 bp) of the PU.1 element. EMSA studies with the CG-like element from all three azurophil serine proteinase genes indicated that this element binds a second protein resulting in a PU.1like complex, which was inhibited by increasing the concentration of poly(dI-dC) in the binding reaction, whereas the faster migrating complex with the 40-kDa protein was not inhibited by the high concentration of poly(dI-dC). The PU.1-like complex is competed by the addition of cold CD11b or cold PR-3 PU.1 elements, suggesting that the CG-like element is binding PU.1 or a related ets protein in addition to the 40-kDa protein.
PR-3, HLE, and Cat G gene promoters all contain a CG-like, a PU.1, and a TATA element organized identically within the first 300 bp of sequence implying that these three elements and the factors that bind them associate to form an important regulatory complex of azurophil neutral serine proteinase gene expression (Table I). Among the neutral serine proteinases of immune/inflammatory cells, gene regulation via PU.1 and a CG-like element appears to be specific to enzymes contained in azurophil granules of PMNL as neither site is present in the promoters of the closely related granzyme serine proteinase family present in T lymphocytes nor in mast cell proteases. Thus, the presence and close proximity of both PU.1 and CGlike elements appears to be unique to the azurophil serine proteinases PR-3, HLE, and Cat G.
The requirement for cooperative interaction of the CG and PU.1 elements for myeloid-specific expression is suggested by studies using non-myeloid HeLa cells. Transient transfection studies showed that HeLa cells do not transcribe the PR-3 promoter construct that contains PU.1, CG, and TATA sites. HeLa cells are known to lack endogenous PU.1 activity (38,40,68), but EMSA studies indicate that they do contain a 40-kDa protein that binds the PR-3 CG element (data not shown). Whether this protein is identical to that present in myeloid cells remains to be determined.
In contrast to the present study, a previous study did not identify the CG-like element in the HLE promoter as an important regulatory element, whereas a PU.1 element was found to be essential (32). We found that PR-3 PU.1 and CG elements have different conditions for optimal binding of their respective proteins, so it is possible that conditions that allow footprinting of the PU.1 element failed to footprint the CG element. However, in the previous study, the elastase CG-like element was not recognized by deletion analysis (32). In our studies the increase in luciferase activity due to the CG element was 200fold above control compared with 100-fold above control when the construct contained only PU.1 and TATA elements. Perhaps the contribution made to the promoter activity by the CG-like element in HLE was missed because the less sensitive CAT reporter system was used in the previous study (32).
Other cis-acting elements have been reported to be relevant in myeloid gene expression. Two sites were identified in the murine myeloperoxidase promoter (69). The protein binding to the second site with a core sequence of 5Ј-AACCACA was identified as polyomavirus enhancer binding protein 2/corebinding factor (PEBP2/CBF) (34). A similar sequence (GGC-CACA) occurs in murine elastase (34), and this sequence is present at Ϫ319 and Ϫ250 on the negative strand of human PR-3 promoter. These sites appear to be inactive as neither EMSA nor deletion analysis identified them. The element CCCTTCC is present in CD11b (70), myeloperoxidase (42), Cat G (22), c-fes (71), and CD13 (45) but is not present in PR-3. A myeloid regulatory element was located in the functional promoter region of human interleukin-5 receptor ␣ subunit gene (72). No sequence similar to this was found in PR-3. A pyrimidine-rich sequence was described as a myeloid activating transcriptional element on the FcyR1 gene (73), and similar sequences occur in CD11b (70), Cat G (22), and HLE (21). The sequence on the negative strand around PU.1 in PR-3 and HLE is similar to myeloid activating transcriptional element. Therefore it is likely that myeloid activating transcriptional element is an inverted PU.1 site. A potential C/EBP site (Ϫ82) and a potential Myb site (Ϫ56) may contribute to the promoter activity conferred by fragment Ϫ91 to 0 that also contains a TATA site. As discussed the C/EBP and Myb sites are conserved in HLE and azurocodin as well as PR-3 (18) and functional in murine neutrophil elastase (34). They are absent from the immediate 5Ј Cat G promoter. The Cat G gene is located on chromosome 14 at q11.2 (20) unlike the other azurophil serine proteinases PR-3, HLE, and azurocidin that are clustered on chromosome 19 p13.3. Because Cat G appears to be coordinately expressed with PR-3 and HLE during azurophil development, it is unlikely that C/EBP and Myb transcription factors play a significant role in developmental expression of these three genes. However, as PR-3, HLE, and Cat G do contain both a PU.1 and CG element with similar organization in all three genes, the transcription factors that bind to these sites may play a central role in the developmental expression of these genes.
Expression of the PR-3, HLE, and Cat G genes is restricted to the promyelocytic stage of differentiation. A PU.1 site is present in the promoter of many myeloid genes, so its role in differentiation-specific expression is likely to be complex. We found that differentiation of myeloid cells to either monocytes or granulocytes reduced the ability to shift both the PU.1-and CG-like elements. Moreover, increasing the percentage of cells undergoing differentiation by using retinoic acid in conjunction with DMF further decreased the intensity of the shifts. This result is in accordance with Hromas et al. (74) who found that PU.1 was absent in differentiated bone marrow myeloid cells. Conversely, Chen et al. (47) reported that peripheral blood PMNL have the highest levels of PU.1 compared with other myeloid cells and that differentiation of HL-60 and U937 cells did not significantly change PU.1 expression. While it is difficult to reconcile these data, they were obtained using widely varying techniques. Our data indicate that PMA treatments of HL-60 and U937 cells that are effective in switching-off PR-3 gene expression also reduce nuclear protein binding to the PR-3 PU.1 element. This does not necessarily imply a reduction in PU.1 protein under these conditions. A recent report by Carey et al. (75) suggests that PMA treatment of U937 cells alters the phosphorylation state of PU.1 which in turn affects the binding pattern on EMSA. We examined PU.1 gene expression using reverse transcriptase-PCR in myeloid cells after treatment with PMA and found no change in expression (data not shown) confirming the result of Chen et al. (47). One explanation of the consistent reduction in the formation of band A and A* with the PR-3 PU.1 element upon PMA treatment is a change in the phosphorylation state of the PU.1 protein. The mechanisms by which PU.1 affects myeloid differentiation are complex and largely unknown. Azurophil neutral serine proteinase genes are switched-off upon myeloid differentiation. The possibility that the CG binding protein cooperates with PU.1 in the regulation of these genes is heightened by the fact that there is a reduction in the specific binding to the CG element upon myeloid differentiation. This further excludes the possibility that the unidentified protein binding to the PR-3 CG element is MS-2 that binds a similar site in CD11a (44). MS-2 has an opposite pattern of expression being absent in undifferentiated and present in differentiated myeloid cells (44).
In the present study we have identified two common transcriptional control elements (PU.1 and CG) that are essential for the expression of PR-3. That PU.1 has a role appears certain but to confer specificity it requires cooperation of other transcriptional factors, such as the protein binding to the PR-3 CG element. HLE and CAT G promoters, but not other granuleassociated serine proteinases of immune and inflammatory cells, possess both a PU.1 and CG-like element with a similar organization and a similar pattern of shift on EMSA to PR-3. This suggests a unique function for these elements in azurophil serine proteinases.