Plasma Hyaluronan-binding Protein Is a Serine Protease*

CTCF is an essential factor for optimal transcription from the amyloid β-protein precursor promoter. A proteolytic activity detected in bovine, rabbit, horse, and human serum cleaves CTCF at three major sites, resulting in a modified mobility shift pattern of the fragments that retain DNA binding ability. The protease was purified to electrophoretic homogeneity, partially sequenced, and identified as the plasma hyaluronan-binding protein. The proteolytic activity was selectively abolished by various serine protease inhibitors, including the Kunitz-type protease inhibitor domain of amyloid β-protein precursor. Reduction with β-mercaptoethanol showed that the 70-kDa protein consists of two polypeptides with apparent molecular masses of 44 and 30 kDa. The serine protease domain was localized to the 30-kDa polypeptide as determined by [3H]diisopropylfluorophosphate binding.

The promoter of the amyloid ␤-protein precursor (APP) 1 gene is a necessary element in the regulation of APP transcription and it has been shown to confer cell type-specific expression in transgenic mice (1,2). The proximal APP promoter is devoid of CCAAT and TATA boxes but contains a prominent initiator element associated with the main transcriptional start site (ϩ1) (3). The integrity of the initiator element is essential for both start site selection and optimal transcriptional activity (3). In addition, transcription from the APP promoter is critically dependent on the presence of an intact nuclear factor binding site designated APB␤ (3,4). The core recognition sequence for this binding site is located between positions Ϫ82 and Ϫ93 and its elimination reduces transcriptional activity by ϳ70 -90% (3,4). The nuclear factor that activates transcription from APB␤ was identified as CTCF (5), a nuclear regulatory protein comprising 727 amino acids (6). It contains a centrally located DNA binding domain with 11 zinc finger motifs that are flanked by 267 amino acids on the N-terminal side and 151 amino acids on the C-terminal side. This protein was first identified as a factor that binds to the chicken c-myc promoter (7) and to the silencer element of the chicken lysozyme gene (8,9). A functional role for CTCF in both positive and negative transcriptional regulation has been documented (5,6,10,11).
While purifying CTCF, we observed that binding to the APB␤ sequence became increasingly unstable as the level of purity increased. However, this instability could be overcome by supplementing the incubation mixture with the zwitterionic detergent CHAPS and large amounts of nonspecific protein (5). Routinely, fetal calf serum (FCS) was used as a source of such protein. While optimizing the conditions for the binding reaction, we noticed that if crude nuclear extract was preincubated with an excess of FCS prior to the binding reaction, substantial changes in the electrophoretic mobility shift pattern occurred, resulting in binding complexes with higher electrophoretic mobilities. We have here isolated the factor responsible for this altered mobility shift and identified it as a protease activity associated with the plasma hyaluronan-binding protein PHBP (12). The cDNA for this protein was previously cloned, and sequence analysis indicated the presence of a serine protease consensus domain. However, in the original preparations, proteolytic activity of PHBP was not demonstrated (12).
Double-stranded oligonucleotide APB␤-WT (5) containing the CTCF recognition sequence was 5Ј-end-labeled with [␥-32 P]ATP (14). 10 ng of labeled oligonucleotide (50,000 -500,000 cpm) were incubated for 30 min at 25°C with 10 -20 g of protein from nuclear extract in buffer D supplemented with 2 g of poly(dI-dC), 5 g of yeast tRNA, 2.5% CHAPS, and 1 ml of FCS in a total reaction volume of 30 l. The incubation mixture was electrophoresed in 1% agarose or 6% polyacrylamide gels with 0.5ϫ Tris-borate-EDTA (14) at 180 V constant voltage for 45 min. Gels were dried and autoradiographed for 2-4 h at Ϫ80°C.
Proteolytic activity was monitored by incubating 2 l of nuclear extract aliquots prior to the binding reaction either with whole serum or with purified serum protein fractions in 5-10 l of buffer D at 25°C for 1 h. When needed, concentrated purified material was prediluted 10 -100-fold with buffer D.
Protease Purification from Human Serum-Human serum (male) derived from whole clotted blood was purchased from Sigma. Solid ammonium sulfate was added to 500 ml of serum to 25% saturation. After centrifugation at 10,000 ϫ g for 15 min, the supernatant was collected, and proteins were further precipitated by increasing the ammonium sulfate concentration to 50% saturation. The centrifugation was repeated, and the resulting pellet containing the bulk of the proteolytic activity was resuspended in 100 ml of buffer T (20 mM Tris-HCl, pH 7.5, 2 mM MgSO 4 , 0.1 mM EDTA, and 200 mM KCl). The solution was dialyzed overnight against 2 liters of buffer T with one change of buffer.
The dialyzed material was loaded on a 10-ml DEAE Sepharose Fast Flow (Amersham Pharmacia Biotech) column preequilibrated with buffer T. The column was subsequently washed with three 20-ml portions of buffer T containing 300, 350, and 400 mM KCl. The bulk of the proteolytic activity was eluted with 30 ml of buffer T containing 700 mM KCl. Proteins were precipitated in ammonium sulfate at 60% saturation. After centrifugation, the pellet was resuspended in 5 ml of buffer D and dialyzed against 500 ml of the same buffer. The material was then loaded on a 1-ml HiTrap heparin-agarose column (Amersham Pharmacia Biotech) preequilibrated with buffer D. The column was washed with 10 ml of buffer D, and the proteins were eluted with 18 ml of a linear KCl gradient (100 -700 mM). Fractions of 1 ml were collected, and the proteolytic activity was monitored as described above.
Separation and Identification of CTCF Proteolytic Fragments-One milliliter of HeLa cell nuclear extract was incubated with 3 l of heparin-agarose-purified protease for 1 h at 25°C. The reaction was stopped by the addition of 15 g of a peptide containing the KPI domain of APP (15). The KCl concentration in the reaction mixture was adjusted to 200 mM, and the reaction products were loaded on a 1-ml HiTrap SP Sepharose column preequilibrated with buffer D containing 200 mM KCl. The column was washed with 5 ml of buffer D containing 200 mM KCl and 4 ml of the same buffer containing 300 mM KCl. Proteins were eluted with 20 ml of a 300 -700 mM linear KCl concentration gradient. Fractions of 1 ml were collected and analyzed by mobility shift electrophoresis.
An Amersham Pharmacia Biotech Superose 6HR 10/30 gel filtration column was equilibrated with buffer D containing 500 mM KCl, 2.5% CHAPS and calibrated with a set of globular proteins (Amersham Pharmacia Biotech HMW Calibration kit). SP Sepharose chromatography fractions 19 -23 containing mobility shift activity were combined, supplemented with 2.5% CHAPS, concentrated on a Centricon-10 device (Amicon), and loaded on the gel filtration column. The gel filtration was performed at a 0.4 ml/min flow rate. Fractions of 0.5 ml were collected and analyzed by either mobility shift electrophoresis or Western blotting. Antibodies against the N-and C-terminal CTCF sequences were described elsewhere (5). They were affinity purified with a Pierce Sulfo-link kit and used as primary antibodies in the Western blotting ECL procedure (Amersham Pharmacia Biotech), which was carried out according to manufacturer's instructions.
Gel Electrophoresis, Extraction, Renaturation, and Sequencing of Proteins-SDS-PAGE of proteins was performed with the Laemmli Tris-glycine system (14). For some applications, 4 -20% precast gradient Ready Gels (Bio-Rad) were used. When indicated, reducing agent (␤-mercaptoethanol) was omitted in the sample buffer. Gels were stained with Coomassie Brilliant Blue R-250 and photographed.
Proteins were extracted from gels and renatured as described (16) with some modifications. After electrophoresis under nonreducing conditions, the gel was stained with cold 0.25M KCl, and the protein bands were excised. The gel slices were weighed, rinsed with nonreducing sample buffer, minced, and incubated with an equal volume of nonreducing sample buffer at 75°C for 20 min. Extracted protein samples were collected with an Amicon Micropure .22 spin device. The gel extraction was repeated, and the samples were combined. Aliquots of the extracted proteins were analyzed by SDS-PAGE. For renaturing purposes, the remaining extracted samples were supplemented with 9 volumes of 50 mM Tris-HCl, pH 7.6, 0.1 mM EDTA, 0.15 M NaCl, and 0.1% SDS. To each sample was added as a carrier 5 l of DEAE chromatography (see above) flow-through fraction. This fraction exhibited no protelytic activity and was thus preferred over bovine serum albumin, which could carry traces of endogenous serum-derived proteolytic activity. The proteins were precipitated with 5 volumes of acetone at Ϫ20°C overnight.
Protein precipitates were centrifuged at 14,000 ϫ g for 10 min, air dried, resuspended in 20 l of 6 M guanidine hydrochloride, and incubated at 25°C for 1 h. Subsequently, the concentration of guanidine hydrochloride in the samples was reduced to ϳ5, 4, 3, 2, and 1.2 M by the addition of aliquots of buffer D at 10-min intervals. Residual guanidine hydrochloride was removed from the samples by desalting on Amersham Pharmacia Biotech G-25 spin columns preequilibrated with buffer D. Proteolytic activity was analyzed in the renatured samples as described above.
For protein sequencing, samples were separated by SDS-PAGE in the presence of ␤-mercaptoethanol, transferred to polyvinylidene difluoride membrane, and stained with amido black. Bands corresponding to 44-and 30-kDa proteins were excised. , dried, and exposed to x-ray film at Ϫ80°C for 72 h.
The protease was purified to electrophoretic homogeneity by cation exchange chromatography in the presence of 6 M urea. Specifically, heparin-agarose-purified protease was diluted with 4 volumes of buffer D containing 6 M urea and was concentrated with a Centriplus-10 device (Amicon) to the original volume. The resulting material was diluted with an additional three volumes of buffer D containing 6 M urea, and the concentration procedure was repeated. The final concentrate was loaded on an UnoS 0.12-ml polishing column (Bio-Rad) preequilibrated with buffer D containing 6 M urea. Proteins were eluted with 5 ml of a 100 -600 mM linear KCl gradient in buffer D containing 6 M urea. 200-l fractions were collected, aliquoted, and immediately frozen at Ϫ80°C.
Inhibition of the PHBP Protease-0.05 l of heparin-agarose-purified protease was incubated in 4 l of buffer D with the indicated amounts of the protease inhibitors ( Fig. 7) for 10 min at 25°C. Subsequently, the protease assay was performed as described above. A recombinant KPI domain peptide comprising amino acids 285-345 of the APP-751 protein (15) and a mutated version of the recombinant peptide with a single substitution of arginine with isoleucine at position 301 were kindly provided by W. Van Nostrand.

Treatment of Nuclear Extract with Serum Changes the Mobility Shift Pattern of the APB␤ Binding
Complex-When nuclear extract from HeLa cells was incubated with an 80-mer oligonucleotide containing the APB␤ domain of the APP promoter, a characteristic mobility shift complex b was formed (Fig. 1, lane 2). This complex is the result of transcription factor CTCF binding to the APB␤ recognition sequence (5). However, upon incubating the nuclear extract for 1 h at 25°C with FCS before assembling the binding reaction, a substantial change in the electrophoretic mobility shift pattern occurred (Fig. 1, lanes [3][4][5][6]. With increasing amounts of FCS in the preincubation reaction, the original mobility shift complex b was gradually eliminated as two new complexes, b 1 and b 2, with higher electrophoretic mobilities emerged. Moreover, at the highest concentration of FCS, complex b 1 was also eliminated and only b 2 , the complex with the highest mobility, remained. A similar change in the mobility shift pattern was observed with adult bovine, horse, or rabbit serum (not shown), as well as with an ϳ5-10-fold lower concentration of human serum (Fig. 1, lanes  7-10). They also occurred when crude nuclear extract was replaced with purified CTCF. However, significantly lower amounts of serum were required to achieve comparable results (not shown). The most plausible explanation for this alteration in mobility shift behavior is a proteolytic cleavage of the CTCF protein by a serum enzyme.
A Purified Fraction from Human Serum Cleaves the CTCF Transcription Factor at Several Sites-To isolate the putative proteolytic activity, human serum was initially fractionated by ammonium sulfate precipitation, and the proteolytic activity was recovered in fractions that precipitated between 30 and 50% saturation. The precipitated material was solubilized and loaded on a DEAE-Sepharose column. The active fraction, which eluted at a 700 mM KCl concentration, was again precipitated with ammonium sulfate, solubilized, and further purified by affinity chromatography on heparin-agarose. The proteolytic activity eluted over a broad range of KCl concentration from 250 to 500 mM (Fig. 2B), which coincided with the concentration profile of eluted protein as measured by A 280 (Fig. 2A). In addition to numerous other proteins distributed over a wide molecular mass range, the elution fractions contained a major protein with an approximate molecular mass of 70 kDa (Fig. 2C). Fractions 12-18 that contained the highest level of activity ( Fig. 2) were combined and used for further experiments.
Because the excess of serum distorts the bands on polyacrylamide gels during mobility shift electrophoresis, we were restricted to agarose gels in our initial experiments (Figs. 1 and 2). Employing heparin-agarose-purified material instead of whole serum for nuclear extract preincubation allowed us to switch to polyacrylamide gels in mobility shift assays. This enabled us to study the mobility shift patterns at a higher level of resolution (Fig. 3A). In addition to the three major complexes b, b 1 , and b 2 observed in agarose gels (Fig. 2), two additional complexes b 0 and b 10 could be resolved (Fig. 3A). As with whole serum, increasing amounts of partially purified active material in the preincubation reaction lead to a gradual increase in the amount of complex b 2 at the expense of all other bands, which were progressively eliminated from the mobility shift pattern. At very high concentrations of material purified from human serum there was also a gradual decrease in complex b 2 (Fig. 3A).
To demonstrate that the changes in the mobility shift patterns were indeed because of proteolytic digestion of the CTCF protein, we proceeded with separation of the treated nuclear extract and analysis of the resulting fractions by mobility shift electrophoresis and Western blotting. Initially, nuclear extract in preparative quantities was treated with heparin-agarosepurified active material. This resulted in a mobility shift pattern where the prominent complexes b, b 1 , and b 2 , as well as minor complexes b 0 and b 10 , were represented (Fig. 3B, lane 1). The treated extract was loaded on a cation exchange column, and all proteins generating the mobility shift complexes eluted in fractions 19 -24 (Fig. 3B, lanes 3-8). This purification step considerably reduced total protein concentration in the preparation, and the combined fractions 19 -23 (Fig. 3B, lanes 3-7) were used as starting material (Fig. 4A, lane 1) for subsequent gel filtration.
Fast protein liquid chromatography gel filtration was performed using a Amersham Pharmacia Biotech Superose 6 column. Mobility shift analysis of the loading material (Fig. 4A, lane 1) revealed the presence of all binding complexes that were observed prior to the cation exchange chromatography (Fig. 3B,  lane 1). Meanwhile, Western blot analysis of the cation exchange-purified material with antibodies against either the N- (Fig. 4B) or C-terminal (Fig. 4C) end of CTCF recognized a protein with an apparent molecular mass of 140 kDa (band p) corresponding to native CTCF (Fig. 4, B and C, lanes 1 and 2). Antibodies against the N-terminal part of the protein also reacted with a 120-kDa band [p 0 ], as well as with a much less pronounced band at 130 kDa [p 01 ] (Fig. 4B, lane 2). In contrast, antibodies against the C-terminal end of CTCF recognized a new 70-kDa (p 1 ) immunoreactive band (Fig. 4C, lane 2). The emergence of these lower molecular weight bands that are differentially recognized by antibodies either against the Cterminal or the N-terminal end proves that CTCF indeed undergoes proteolytic cleavage during incubation with the purified fraction of serum.
Gel filtration of the fragments allowed an estimation of the number and relative positions of the cleavage sites. Both antibodies recognized the peak of intact CTCF that eluted in fractions 13 and 14 (Fig. 4, B and C, lanes 4 and 5). This represented an apparent molecular mass of 400 kDa as determined by calibration with globular protein standards. The corresponding binding complex b was observed in the same fractions by mobility shift electrophoresis (Fig. 4A, lanes 2 and 3). Fragment p 0 eluted in fractions 15 and 16 (Fig. 4B, lanes 6 and 7), generating the corresponding binding complex b 0 (Fig. 4A,  lanes 4 and 5). Furthermore, fragment p 1 eluted in fractions 17 and 18 (Fig. 4C, lanes 8 and 9) where the matching complex b 1 was observed (Fig. 4A, lanes 6 and 7). The CTCF fragments corresponding to the complexes b 10 and b 2 , which eluted in fractions 19 -21, could not be recognized by either antibody (Fig. 4A, lanes 8 -10; B and C, lanes 9 -11). According to both SDS-PAGE and gel filtration, the CTCF fragment producing the faint band p 01 migrated to a position between the fulllength CTCF and fragment p 0 (Fig. 4B, lanes 5 and 6). However, no binding complex has been identified that could be assigned to this fragment, presumably because its low prevalence does not allow detection. Alternatively, the hypothetical binding complex formed by p 01 may not be separable from binding complexes b and b 0 under the applied mobility shift electrophoresis conditions.
Assuming that CTCF binds to DNA as a monomer (10), the results of the gel filtration suggest that there are three prominent proteolytic cleavage sites on the CTCF molecule. One site  (lanes 2-10). Arrows indicate binding complexes, and a bracket indicates the free oligonucleotide. B, aliquots of 20 l from elution fractions 12-21 (as described in A) were separated by SDS-PAGE and analyzed by Western blotting using antibodies against the N-terminal part of full-length CTCF (lanes 3-12). The elution fractions were compared with the loading material (lane 2) and purified CTCF (lane 1). The positions of full-length CTCF (p), as well as proteolytic fragments p 01 and p 0 , are indicated by arrows. C, same as in B except that antibodies against the C-terminal part of CTCF were used. The positions of full-length CTCF (p) and fragment p 1 are indicated by arrows. Arrowheads point to minor reactive fragments in fractions 17 and 18. D, schematic diagram representing the CTCF protein and its proteolytic fragments. The positions of the zinc finger DNA binding domain (Zn), and sequences recognized by antibodies against the N-(N) and C-terminal (C) domains are indicated. Arrows mark the approximate relative positions of the protease cleavage sites. Additional arrowheads indicate the position of a minor protease cleavage site, and a bracket delineates the three potential fragments produced by cleavage at that site. is located between the N terminus and the zinc finger DNA binding domain, and two are located between the zinc finger domain and the C terminus (Fig. 4D, arrows). Incomplete cleavage at these sites would produce numerous protein fragments. Among them, the fragments p 0 , p 01 , p 1 , p 10 , and p 2 , which are schematically shown in Fig. 4D, would contain the zinc finger domain and produce the corresponding binding complexes. Fragments p 0 and p 01 would be recognized by the Nterminal but not the C-terminal antibody. Similarly, protein p 1 would be only recognized by the C-terminal antibody. Finally, neither antibody would recognize the proteins p 10 and p 2 that produce binding complexes b 10 and b 2 .
Depending on the conditions of the cleavage reaction, additional minor binding complexes could be observed on the mobility shift gel between the b 1 and b 2 bands. An arrowhead in Fig. 4A (fraction 18) indicates an example of such a weak binding complex, designated b 3 . Incidentally, an exceedingly weak band (p 31 ) that reacted with the C-terminal antibody was detected migrating slightly ahead of fragment p 1 in fractions 17 and 18 (Fig. 4C, arrowheads). This suggests the existence of an additional cleavage site on the CTCF molecule in close proximity to the major site that cleaves off the N terminus generating fragment p 1 (Fig. 4D, arrowhead). Cleavage at that site would generate three additional hypothetical CTCF protein fragments containing the zinc finger DNA binding domain (Fig. 4D, fragments p 31 , p 32 , and p 33 , indicated by a bracket). Fragment p 31 would thus retain an intact C-terminal sequence of CTCF and therefore react with antibodies against the C terminus. The same fragment could conceivably account for the appearance of the minor binding complex b 3 observed in fraction 18 on the mobility shift gel. Because of the lower prevalence of CTCF protein cleavage at this site, it is possible for example that it only becomes accessible after cleavage at the N-terminal p 1 site. However, we consider it to be a marginal cleavage site for the protease, and we therefore disregard it in the further discussion of the results.
The Protease Activity Is Identified as PHBP-Initial attempts to further purify the protease under nondenaturing conditions were unsuccessful. Employing a variety of separation techniques, we observed the same major proteins were co-purified. Preliminary results indicated that the protease probably exists in serum as part of a high molecular weight multiprotein complex (data not shown). Therefore, we proceeded with the protease identification using preparative SDS-PAGE.
The extracted proteins were renatured and incubated with nuclear extract, followed by mobility shift electrophoresis (Fig. 5B, lanes 1-4). Proteolytic activity was detected only in the sample containing the pp2 protein (Fig. 5B, lane 2). As a control, the total heparin-agarose-purified protease fraction was denatured under reducing, as well as nonreducing, conditions and then renatured while omitting electrophoretic separation. Proteolytic activity could only be restored from the nonreduced sample (Fig. 5B, lanes 5 and 6). From these exper-iments we conclude that the proteolytic activity is associated with the 70-kDa protein pp2. This protein comprises two polypeptide chains with apparent molecular masses of 44 and 30 kDa, which are connected via disulphide bonds. Disruption of the bonds irreversibly abolishes activity.
To further characterize and identify the protease, the heparin-agarose-purified material was subjected to SDS-PAGE in the presence of ␤-mercaptoethanol. Proteins were subsequently transferred to polyvinylidene difluoride membrane and visualized by Amido Black staining. Both the 44-and 30-kDa proteins were excised, and the N-terminal sequences of the proteins were determined. The 44-and 30-kDa proteins contained the sequences SLLESLDPDTP and IYGGFKSTAGAKHP, respectively, and they displayed a perfect match with the sequence of  6 -10). The starting material contained the four major proteins pp1-pp4 (lanes 1 and 6). These proteins were individually extracted from the nonreducing gel and separately analyzed under reducing and nonreducing conditions: pp1 (lanes 2 and 7), pp2 (lanes 3 and 8), pp3 (lanes 4 and 9), and pp4 ( lanes 5 and 10). B, after extraction from the gel, proteins were renatured. The renatured proteins pp1-pp4 (lanes 1-4) were incubated with nuclear extract and analyzed by acrylamide mobility shift electrophoresis with radiolabeled APB␤ oligonucleotide. In addition, the total heparin-agarose-purified material was denatured and renatured, without prior separation by SDS-PAGE, followed by mobility shift electrophoresis. Denaturation was carried out either under nonreducing (lane 5) or reducing (lane 6) conditions. Binding complexes b and b 2 are indicated by arrowheads, free oligonucleotides (f) are indicated by a bracket. C, autoradiograph of the [ 3 H]DFP-labeled protease. The heparin-agarose-purified protease was incubated with [ 3 H]DFP and subjected to SDS-PAGE both under nonreducing (lane 1) and reducing (lane 2) conditions. the human PHBP protein described by Choi-Miura et al. (12).
PHBP cDNA sequence data (12) suggested that the protein contained a putative serine protease domain. In such a case the active center of the protease might form a covalent bond with DFP. To provide additional evidence that the protease is PHBP, we performed [ 3 H]DFP labeling of the protease. We incubated the heparin-agarose-purified material with tritiumlabeled DFP and analyzed the reaction products by SDS-PAGE both under reducing and nonreducing conditions (Fig. 5C). Autoradiography of the gel yielded a single labeled 70-kDa protein in the preparation without reducing agent (Fig. 5C,  lane 1). Under reducing conditions a single labeled fragment with a molecular mass of 30-kDa band was observed, whereas the 44-kDa fragment remained unlabeled (Fig. 5C, lane 2). Moreover the labeled bands on the autoradiograph exactly overlapped with the corresponding Coomassie Blue-stained protein bands of PHBP. Thus the 30-kDa subunit of the PHBP protein indeed contains an active serine protease domain.
The PHBP Protease Is Purified to Electrophoretic Homogeneity-As indicated above, we initially could not achieve chromatographic separation of the heparin-agarose-purified proteins under nondenaturing conditions. However, we later found that denaturing the protease with SDS or urea is reversible. Therefore, as the final step of purification we employed cation exchange chromatography in the presence of 6 M urea. Heparinagarose-purified material was subjected to ultrafiltration to concentrate the proteins and exchange the buffer. The new buffer contained 100 mM KCl and 6 M urea. The material was loaded on a Bio-Rad UnoS Polishing column, and the proteolytic activity was assayed in the elution fractions (Fig. 6A). Low levels of proteolytic activity was recovered in elution fractions 14 -19 (Fig. 6A, lanes 1-6). However, a peak of activity was observed in fractions 20 and 21 (Fig. 6A, lanes 7 and 8) corresponding to a 250 mM KCl concentration. The 70-kDa PHBP protein band was present at low concentrations in elution fractions 14 -19 as determined by SDS-PAGE (Fig. 6B, lanes 1-6). The same fractions also contained traces of a 140-kDa protein.
In contrast, fractions 20 and 21 displayed much higher concentrations of the 70-kDa PHBP protein (Fig. 6B, lanes 7 and 8), whereas no other protein band could be detected in these fractions. Similarly, the corresponding concentration peak of the 44-and the 30-kDa polypeptides eluted in the same fractions under reducing conditions (Fig. 6C, lanes 7 and 8). The protease purified to homogeneity is very unstable. Any prolonged handling of the nonfrozen material lead to degradation and reduction in activity (data not shown). The 30-kDa polypeptide chain is especially labile. Thus, two independent experiments, [ 3 H]DFP labeling and chromatographic purification of the protease to homogeneity, confirmed our assignment of the proteolytic activity to the PHBP protein.
The PHBP Protease Is Specifically Inhibited by the KPI Domain of APP-Several common protease inhibitors were examined for their effect on the protease activity. Heparin-agarosepurified PHBP was premixed with commonly used maximal concentrations of either phenylmethylsulfonyl fluoride, aprotinin, pepstatin, or leupeptin and subsequently incubated with HeLa cell nuclear extract followed by mobility shift electrophoresis (Fig. 7, lanes 1-6). The protease was inhibited by high concentrations of the serine protease inhibitors phenylmethylsulfonyl fluoride, aprotinin, and leupeptin, whereas the aspartic protease inhibitor pepstatin did not significantly affect the protease activity (Fig. 7, lanes 3-6). Furthermore, a peptide containing the KPI domain of APP completely inhibited the protease at a concentration of 12 g/ml, whereas the inhibitory effect was still significant at 1.2 g/ml concentration (Fig. 7,  lanes 7-9). In contrast, a mutated version of the KPI domain peptide did not affect the PHBP protease activity even at a concentration of 800 g/ml (Fig. 7, lane 10). This mutation in which the arginine at position 301 was substituted with isoleucine, leads to a shift in the inhibition specificity of the mutant peptide from trypsin-like to elastase-like proteases. 2 These re-2 W. Van Nostrand, personal communication. sults are consistent with the notion that PHBP is a serine protease as was predicted from its primary structure (12). In addition, the capacity of the KPI domain to inhibit the PHBP serum protease might represent a novel function for APP. DISCUSSION PHBP was originally purified from human plasma by affinity chromatography on hyaluronic acid-agarose columns (12). The protein migrated on SDS-polyacrylamide gels as a single band corresponding to a molecular mass of 70 kDa under nonreducing conditions. Under reducing conditions two protein bands were observed at 50 and 17 kDa. The 50-kDa protein yielded two overlapping N-terminal sequences: NH 2 -SLLESLDPDWT-PDQY (major) and NH 2 -FSLMSLLESLDPDWT (minor). Two overlapping N-terminal sequences were also determined for the 17-kDa polypeptide: NH 2 -IYGGFKSTAGAKHPWQ (minor) and NH 2 -STAGAKHPWQASLQSS (major). The cloning of the cDNA encoding the PHBP protein and the deduced sequence of the full-length protein were reported in the same publication (12). The authors concluded that the PHBP protein in serum underwent posttranscriptional processing, which removed the putative signal peptide from the N terminus of the protein and introduced an internal cleavage on the N-terminal side of isoleucine at position 314. The protein deduced from the cDNA sequence contained a serine protease domain that was localized to the C-terminal part of PHBP. The authors mentioned a weak proteolytic activity associated with the protein. However, no experimental data were provided in support of this observation. Moreover, the cleavage at position 314 should produce an ϳ25-kDa C-terminal protein fragment instead of the 17-kDa fragment observed by the authors. They concluded that the 25-kDa C-terminal polypeptide underwent further proteolytic degradation that resulted in the destruction of the putative serine protease domain (12). In a further development, the same group reported the cloning of the human PHBP gene, which was designated as HABP2 (17).
In the process of our studies on the binding of transcription factor CTCF (7) to the APB␤ element of the APP promoter, we serendipitously discovered that preincubation with serum resulted in proteolytic digest of CTCF. The proteolytic activity was purified and identified as PHBP. At the final stage of PHBP purification, SDS-PAGE revealed several protein bands in the fractions containing the protease activity. Proteolytic activity was recovered only under nonreducing conditions from the protein band with an apparent molecular mass of 70 kDa.
In the presence of ␤-mercaptoethanol this protein yielded two bands with molecular masses of 44 and 30 kDa. The N-terminal sequence that we determined here for the 44-kDa protein perfectly matched the major N-terminal sequence of the 50-kDa PHBP fragment. There was also a perfect match between the N-terminal sequence that we determined for the 30-kDa protein and the minor N-terminal sequence of the 17-kDa PHBP polypeptide identified by Choi-Miura et al. (12). Moreover, the circumstance that we observed a 30-kDa fragment instead of a 17 kDa fragment as the C-terminal PHBP polypeptide suggests that in our preparation the serine protease domain of the protein remained intact. This conclusion was further confirmed by [ 3 H]DFP labeling of the purified protease, which identified the 30-kDa fragment as the carrier of the active serine protease domain of PHBP. This difference in molecular mass of the C-terminal polypeptide generated by the two independent purification procedures can be explained by either alternative processing of the PHBP molecule or by nonspecific proteolytic degradation. Indeed, in our experience the PHBP protein purified to homogeneity is exceedingly unstable. A possible explanation for this instability is that PHBP is capable of selfinactivation by digesting its own protease domain. This might explain why Choi-Miura et al. (12) observed neither the 30-kDa fragment nor a proteolytic activity in their preparations of the PHBP protein.
Whereas PHBP is prominently present in serum, there is currently no evidence that it functions in the processing of nuclear transcription factor CTCF. However, PHBP can be useful in investigating some functional properties of CTCF. For example, It was established from the cDNA sequence that the actual molecular mass of CTCF is only 82 kDa. However, during SDS-PAGE, CTCF migrates slower than expected, resulting in an apparent molecular mass of ϳ140 kDa. The source of this abnormal rate of migration was traced to the N-terminal end of the protein, presumably resulting from the unusual shape of the molecule (18). Furthermore, sedimentation velocity experiments suggested that unbound CTCF exists in a monomeric form (10). However, based on calibration with established molecular weight standards, whole CTCF eluted from the gel filtration column as a protein with the surprisingly high apparent molecular mass of 400 kDa (Fig. 4). This value was also confirmed in independent experiments where CTCF was not treated with protease prior to separation by gel filtration (data not shown). Such an apparent deviation from the actual 82-kDa molecular mass suggested that additional evaluation of the monomeric state of CTCF was desirable.
Our assignment of the number and relative position of the proteolytic cleavage sites on the CTCF molecule was based on the assumption that CTCF binds to the APB␤ element in the form of a monomeric protein (10). The gel filtration results could thus be explained by the presence of three major proteolytic cleavage sites on the CTCF molecule (Fig. 4D). According to this interpretation one of the sites would be located between the N terminus and the zinc finger DNA binding domain and the other two between the zinc finger domain and the C terminus. However, if the hypothetical possibility that CTCF binds to the APB␤ element as a dimer is considered, a similar mobil- ity shift pattern could be generated by an incomplete digestion of the dimer assuming that each subunit contains a single cleavage site. In addition, because complexes b 0 and b 01 are much less pronounced than complexes b, b 1 , and b 2 , they would have to be disregarded under this model as resulting from nonspecific cleavage. Complex b would then be attributed to the intact homodimer, complex b 1 to the heterodimer with one subunit cleaved and another intact, and complex b 2 to the homodimer comprising two cleaved subunits. The heterodimer forming complex b 1 would then be found in the gel filtration fractions 17 and 18 (Fig. 4A). However, in such a case an equal amount of the cleaved fragment and the intact protein (p) should be observed by Western blotting in fractions 17 and 18. In contrast, all the residual intact CTCF was located in fractions 13-15 (Fig. 4, B and C). Thus, the Western blot analysis further conforms to the interpretation that CTCF contains three major cleavage sites and binds to the APB␤ site as a monomer.
The finding that PHBP exhibits proteolytic activity allows for the elucidation of its function and target preference. Although it has not been established whether CTCF is a natural target for processing by PHBP within the cell, it could indeed be used as a substrate for some aspects of this investigation, including the determination of the proteolytic target sequence. In the original PHBP study (12) the authors hypothesized that PHBP could participate in the proteolysis of the extracellular matrix. They specifically proposed inter-␣-trypsin inhibitor as a putative target. Our preliminary results suggest that in serum PHBP exists in complex with accessory factors that might regulate its proteolytic activity. Once PHBP is released from the complex, its proteolytic activity deteriorates rapidly. This suggests that in some cases PHBP proteolysis might have a transient character in a specific location. For example, high levels of hyaluronic acid are accumulated locally during wound healing (19), which could attract PHBP to the affected area.
This localization, as well as the potentially transient character of the proteolysis, suggests a possible function for PHBP in the wound healing process.