TTSPs [type II TMPRSSs (transmembrane serine proteases)] are a growing family of trypsin-like enzymes with, in some cases, restricted tissue distribution. To investigate the expression of TTSPs in the nervous system, we performed a PCR-based screening approach with P10 (postnatal day 10) mouse spinal cord mRNA. We detected the expression of five known TTSPs and identified a novel TTSP, which we designated neurobin. Neurobin consists of 431 amino acids. In the extracellular part, neurobin contains a single SEA (sea-urchin sperm protein, enterokinase and agrin) domain and a C-terminal serine protease domain. RT–PCR (reverse transcription–PCR) analysis indicated the expression of neurobin in spinal cord and cerebellum. Histochemical analysis of brain sections revealed distinct staining of Purkinje neurons of the cerebellum. Transiently overexpressed neurobin was autocatalytically processed and inserted into the plasma membrane. Autocatalytic activation could be suppressed by mutating Ser381 in the catalytic pocket to an alanine residue. The protease domain of neurobin, produced in Escherichia coli and refolded from inclusion bodies, cleaved chromogenic peptides with an arginine residue in position P1. Serine protease inhibitors effectively suppressed the proteolytic activity of recombinant neurobin. Ca2+ or Na+ ions did not significantly modulate the catalytic activity of the protease. Recombinant neurobin processed 17-kDa FGF-2 (fibroblast growth factor-2) at several P1 lysine and arginine positions to distinct fragments, in a heparin-inhibitable manner, but did not cleave FGF-7, laminin or fibronectin. These results indicate that neurobin is an authentic TTSP with trypsin-like activity and is able to process FGF-2 in vitro.
Extracellular proteolytic activity is an essential process for tissue development and differentiation, homoeostasis, remodelling and wound healing. The deregulation of extracellular proteolytic activity is frequently associated with a variety of pathological conditions. Recently performed extensive database analyses and experimental screens revealed the existence of a novel family of trypsin-like enzymes named TTSPs [type II TMPRSSs (transmembrane serine proteases)] [1–4]. Members of this family are composed of a usually short cytosolic N-terminus, a single transmembrane domain, a region containing a variable set of domains found in very different proteins and the C-terminal protease domain.
Physiological functions of TTSPs were increasingly uncovered in recent years. The identification of mutations in the gene coding for TMPRSS3, in patients with congenital autosomal recessive deafness, suggested a role of this TTSP in hearing . Ablation of the matriptase gene in mice demonstrated a critical role of this TTSP in the development of the epidermis, hair follicles and cellular immunity . Transgenic overexpression of matriptase caused spontaneous squamous cell carcinoma and potentiated carcinogen-induced tumour formation . Mice with a null mutation in the gene encoding the TTSP corin developed hypertension . Deletion of the hepsin gene in mice demonstrated a significant role of this protease in auditory function . Experimental overexpression of hepsin promoted prostate cancer progression and metastasis . Such findings underscore the involvement of TTSPs in a variety of vital functions.
Molecular analysis of TTSP proteolytic activity allowed the identification of a broad spectrum of possible substrates. Matriptase-2 multiply cleaves type I collagen and fibronectin . Corin mediates pro-atrial natriuretic peptide activation . TMPRSS3 activates epithelial sodium channels . TMPRSS2 activates protease-activated receptor-2 in prostate cancer cells , and hepsin cleaves and activates pro-hepatocyte growth factor and pro-urokinase [15,16].
Many TTSPs show a restricted expression pattern, suggesting tissue-specific functions. Enteropeptidase, for example, is confined to the proximal part of the small intestine and acts as an activator of trypsinogen . The TTSP DESC1 (differentially expressed in squamous cell carcinoma 1) is expressed in skin and, additionally, in the salivary gland and epididymis . HAT (human airway trypsin-like) is predominantly expressed in the trachea  and corin is mainly present in heart myocytes . In the nervous system, only spinesin was so far identified as a TTSP expressed by neuronal cells . The function of spinesin is not yet known.
In view of the complex differentiation of the nervous system, it can be assumed from the evidence available that many serine proteases, including known and possibly novel TTSPs, are expressed especially during development and differentiation. This assumption is supported by results of extensive mouse genome analyses that predict the existence of more than 220 serine protease genes or pseudogenes . Therefore we established a PCR-based screening approach to estimate the number of TTSPs expressed during nervous system differentiation. Using P10 (postnatal day 10) mouse spinal cord, we detected the expression of at least 21 non-TTSP serine proteases and five known TTSPs. Additionally, we discovered a novel TTSP, which we designated neurobin (neuronal and Ca-independent like thrombin). Neurobin was most distinctly detected in Purkinje neurons of the cerebellum. Recombinant neurobin processed FGF-2 (fibroblast growth factor-2) in vitro.
MATERIALS AND METHODS
Chemicals were obtained from either Fluka or Sigma (unless otherwise stated) and were of the highest quality. Enzymes were from Promega (unless otherwise stated).
To generate a protease domain-enriched cDNA library, degenerate primers were designed. A hidden Markov model-based alignment of murine trypsin-like serine proteases (http://supfam.org/SUPERFAMILY/index.html, version 1.61) was used to determine the most frequent sequences around the catalytic histidine and serine residues (boxed in Figure 2). The primers+degH (5′-CCGGAATTCTKBRTINTIWCIGCIGCNCAYTG-3′) and −degS (5′-CCGGAATTCGGICCICCISWRTCNCC-3′) were generated that theoretically allow the amplification of more than 80 different protease sequences from the alignment. mRNA was isolated from the spinal cord of three male and three female P10 C57Bl/6 mice using RNeasy and Oligotex mRNA systems (both from Qiagen). cDNA was synthesized from 2 μg of mRNA using PowerScript™ reverse transcriptase (BD Biosciences) with oligo(dT)20 primers (Invitrogen) and 20 units of RNasin. In the subsequent PCR reaction with degenerate primers, 100 ng of cDNA template and the Expand High Fidelity PCR System (Roche) with 1.75 mM Mg2+ were used. The PCR was as follows: 3 min at 94 °C, 5 cycles of 25 s at 94 °C, 40 s at 45 °C, 45 s at 72 °C and 30 cycles of 15 s at 94 °C, 30 s at 55 °C, 45 s at 72 °C and finally 5 min at 72 °C. PCR products in the range 400–600 bp were extracted from agarose gels and cloned into EcoRI-cut pBSSK+. Escherichia coli strain DH5α (Invitrogen) was used for transformation. Transformants were plated on to LB (Luria–Bertani)-agar plates with 100 μg/ml ampicillin (for method comparison, see [21,23]). More than 85% of the clones in the resulting library contained protease sequences.
The protease domain-enriched cDNA library was screened in an iterative two-step manner. First, 20 clones were randomly selected and submitted to DNA sequencing by a CEQ 2000XL DNA Analysis System (Beckman Coulter). Secondly, sequences occurring frequently in this selection were used to generate digoxigenin-labelled DNA probes to deselect identical clones by colony hybridization following ‘The DIG System User's Guide for Filter Hybridization’ (Boehringer/Roche). The necessary components were obtained from Roche. In total six probes [tPA (tissue plasminogen activator), neurotrypsin, hepsin, neurosin, neuropsin and a non-protease sequence] were used and allowed the deselection of more than 90% of the clones. From the remaining library, clones were again selected randomly and analysed by DNA sequencing.
The BD SMART™ RACE (rapid amplification of cDNA ends) cDNA amplification kit was used. In the first-strand cDNA synthesis for 5′-RACE, the primer −atS (5′-CCGGAATTCAACACGTCCCTTCAGGA-3′) was used. In addition to the supplied primers, four gene-specific primers were used in PCR amplifications [primers −GSP1 (5′-GACAGCTCGTGGTGCTTGTGGTTTA-3′) and −GSP2 (5′-CCACCTTCCCTTTGGCTTTGCTTACT-3′) for 5′-RACE; primers +GSP3 (5′-GCTGTTGTGCGTCTGTCTTCGCCAG-3′) and +GSP4 (5′-CCCACCCAACTCAGATGTAGT-3′) for 3′-RACE]. The sequence information from 5′- and 3′-RACE was used to generate full-length neurobin using primers +5′N (5′-GGAGGATGACTGTGCCACCCAGCA-3′) and −3′N (5′-ATTTGCGGCCGCAGATATGCAACCACTGAGCTTGA-3′) in a hot-start PCR reaction as follows: 1 min at 94 °C, 42 cycles of 15 s at 94 °C, 30 s at 68 °C, 1.5 min (+2 s/cycle) at 72 °C, 7 min at 72 °C. To insert the catalytically inactivating S381A mutation, the Transformer Site-Directed Mutagenesis kit (BD Biosciences) was used [primer S>A (5′-CCAATGGTCCACCAGCATCACCCTGGCAGGC-3′)].
Eukaryotic expression constructs
Full-length neurobin cDNA was cloned into pcDNA3.1 vector (Invitrogen) with either the endogenous stop codon to generate untagged neurobin [primers +5′N and −NstopHindIII (5′-CCCAAGCTTCTAGAGACCAGTTTTGGATGTGATCCA-3′)] or in frame with the C-terminal Myc–His tag from pcDNA3.1myc-His(–)A [primers +5′N and −NHindIII (5′-CCCAAGCTTGAGACCAGTTTTGGATGTGATCCA-3′)]. The construct bearing the S381A mutation and the Myc–His tag was termed m1 and was the basis for the generation of mutants m2–m5 by megaprimer-based site-directed mutagenesis.
RT–PCR (reverse transcription–PCR) analysis
Liver, lung, spleen, kidney, heart cortex, cerebellum and spinal cord were dissected from C57Bl/6 mice. The shock-frozen tissues were homogenized by either a rotor–stator or by shaking with stainless steel beads. Total RNA was prepared from 50 mg of tissue. Of each RNA preparation, 1 μg was used for cDNA synthesis using the ThermoScript™ RT–PCR System (Invitrogen) with oligo(dT)20 primers. A portion (10%) of the cDNA synthesis reaction was used for PCR with the Expand High Fidelity PCR System (Roche). The PCR cycles were as follows: for GAPDH (glyceraldehyde-3-phosphate dehydrogenase), 22 cycles of 15 s at 94 °C, 30 s at 58 °C, 30 s at 72 °C and finally 5 min at 72 °C. GAPDH PCR was restricted to 22 cycles to assess the exponential phase of product formation. No product was formed in the PCR even after 42 cycles, when the reverse transcriptase was omitted. For neurobin [primers +atH (5′-CCGGAATTCTGTTTCATACGGGCTGCA-3′) and −atS (5′-CCGGAATTCAACACGTCCCTTCAGGA-3′)], the PCR cycles were as follows: 22 cycles of 15 s at 94 °C, 1 min at 60 °C, 30 s at 72 °C, 20 cycles of 15 s at 94 °C, 30 s at 60 °C, 1 min at 72 °C and finally 7 min at 72 °C. The neurobin PCR product was cut with EcoRI, cloned into pBSSK+ and verified by sequencing.
Production, refolding and activation of recombinant neurobin
PCR using primers +NepPD (5′-CCCGAGCTCGACGATGACGATAAAGTAGCAGGAGGCCAGGATGCT-3′) and −NHindIII resulted in a product encoding Val200-Leu431 preceded by the enteropeptidase cleavage motif. Cloning into pET-28a(+) (Novagen) in frame with the N-terminal His tag (via introduced SacI and HindIII sites) resulted in the protease domain expression construct (HisPD). After overexpression in the E. coli strain BL21 CodonPlus(DE3)-RIL (Novagen), HisPD was purified from washed and solubilized inclusion bodies using HisSelect nickel-affinity resin (Sigma). Purified HisPD (600 μg/ml) in 8 M urea, 0.1 M Tris/HCl (pH 7.5) and 4 mM dithiothreitol was slowly diluted into 10 vol. of refolding buffer (0.1 M Tris/HCl, pH 9, 0.5 M arginine, 150 mM NaCl, 3 mM GSH, 0.7 mM GSSG and 1.3 mM CaCl2) at 4 °C by using a peristaltic pump. The solution was stirred for at least 8 h and then dialysed three times against 100 vol. of 20 mM Tris/HCl (pH 7.5), 150 mM NaCl, 1 mM CaCl2 and 0.1% Tween 20. The refolded HisPD was either activated directly or first concentrated using a Vivaspin 20 concentrator (molecular-mass cutoff 50 kDa; Vivascience). For activation, 2 mg of HisPD was incubated overnight at room temperature (22 °C) with 1 unit of EKMax™ (bovine enteropeptidase; Invitrogen). For the peptide cleavage assay (Figure 7A), directly activated neurobin was used. For the subsequent experiments, neurobin, activated after concentration and additionally purified by cation-exchange chromatography [MonoS PC1.6/5 in an Ettan (Amersham Biosciences), pH 7.0, linear gradient 0.03–0.5 M NaCl], to remove enteropeptidase, was used. Neurobin was stored in 50% (v/v) glycerol at −20 °C for several months without detectable loss of activity. The yield was 200 μg of active protease domain from 1 litre of E. coli culture.
Two rabbits (strain Hsdllf:NZW) were each immunized with eukaryotically expressed and purified SEA (sea-urchin sperm protein, enterokinase and agrin) domain of neurobin. IgGs were purified from the rabbit sera using Protein A 4FF chromatography (Amersham Biosciences). The IgG fraction was affinity-purified on prokaryotically expressed SEA domain immobilized on NHS (N-hydroxysuccinimide)-activated Sepharose 4FF (Amersham Biosciences). The affinity-purified antibody preparations from both rabbits gave virtually identical results both in Western blotting and in indirect immunofluorescence.
Peptide cleavage assays
Chromogenic peptide substrates were from Sigma, Fluka, Chromogenix and Jerini respectively. Trypsin was from hog pancreas (Fluka). Absorptions were measured from 100 μl reaction volumes in 96-well plates using a 1420 Multilabel Counter VICTOR3 (PerkinElmer) at 405 nm. Assay buffer in Figure 7(A) was 20 mM Tris/HCl (pH 7.5), 150 mM NaCl, 1 mM CaCl2 and 0.1% Tween 20. Calcium was omitted in all subsequent assays, unless otherwise stated. For Figure 7(A) reaction volumes contained 4.5 μl of activation buffer with refolded neurobin (HisPD) activated with enteropeptidase (EKMax™), or 150 nM trypsin. In control reactions, no enzyme, HisPD or EKMax™ was used respectively. For Figures 7(B)–7(D), in each case 250 μM GHR-pNA (Gly-His-Arg-pNA; where pNA is p-nitroanilide) was incubated with 30 nM of activated MonoS-purified neurobin protease domain. All reactions were monitored by 5-min interval absorbance (A) measurements. Trypsin control reactions rapidly reached completion, whereas reactions containing activated neurobin did not reach completion within the 2 h measurement period.
Proteins at a concentration of 100 ng/μl in PBS (FGF-7, 140-amino-acid from Amgen; a gift from Professor Sabine Werner, Institute of Cell Biology, ETH-Hönggerberg, Zurich, Switzerland), TBS (Tris-buffered saline; bovine fibronectin from Sigma), TBS-T (Tris-buffered saline containing Tween 20; laminin from Invitrogen) or 10 mM Tris/HCl (pH 7.6) (human FGF-2 from Invitrogen) were added to the enzyme in 50 mM Tris/HCl (pH 7.5) with 150 mM NaCl. The resulting concentrations were 50 ng/μl protein substrate and 0.5 ng/μl enzyme. After 4 h at 37 °C, the reaction was stopped by the addition of loading buffer.
FGF-2 cleavage analysis
For Edman degradation, FGF-2 incubated with neurobin was separated on a 4–12% NuPAGE gel (Invitrogen) using Mes buffer and transferred on to a PVDF membrane. Bands were cut out and sequenced on a Procise 492 cLC sequencer (Applied Biosystems) at the Functional Genomics Center Zurich. Additionally, FGF-2 incubated with neurobin was directly submitted to MALDI–TOF-MS (matrix-assisted laser-desorption ionization–time-of-flight MS) analysis. The analysis was performed on an Ultraflex II (Bruker) at the Functional Genomics Center Zurich.
Cell culture and transfection
HEK-293T cells [HEK-293 cells (human embryonic kidney cells) expressing the large T-antigen of SV40 (simian virus 40)] were grown on 3 cm plates with 2 ml of Dulbecco's modified Eagle's medium without sodium pyruvate (Invitrogen), 10% (v/v) FCS (foetal calf serum) and 2 mM L-glutamine at 37 °C in 10% CO2. For calcium phosphate transfection, 2 μg of DNA was used. In co-transfection experiments, up to 5 μg of each plasmid was used. COS-7 cells were transfected with Lipofectamine™ 2000 (Invitrogen) as suggested in the manual. The growth media were exchanged 5 h after transfection. Cells were further processed 24 or 48 h post-transfection.
SDS/PAGE and Western blotting
For SDS/PAGE, 12.5 or 15% (w/v) polyacrylamide gels were used. With purified protein, concentration was determined by UV absorption measurement at 280 nm using the calculated molar absorption coefficient (ϵ). Protein concentration in extracts was determined with the BCA (bicinchoninic acid) system (Pierce). Whole cell extracts were prepared by lysing the cells in lysis buffer [20 mM Tris/HCl, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100 and 1×protease inhibitor cocktail ‘Complete Mini, EDTA free’ (Roche)]. Before loading, samples were boiled for 5 min in reducing Laemmli buffer . Each lane was loaded with 25 μg of protein extract. Proteins were transferred on to nitrocellulose (Schleicher and Schuell) or PVDF (Millipore) membranes. Transfer quality was controlled with Ponceau S staining. The membranes were blocked in TBS-T containing 10% Western Blocking Reagent (Roche Diagnostics, Mannheim, Germany) overnight at 4 °C. The membranes were incubated for 2 h at room temperature with TBS-T containing 1 μg/ml of affinity-purified polyclonal anti-SEA domain antibody or 0.1 μg/ml anti-Tetra His monoclonal antibody (Qiagen). Membranes were washed three times for 5 min in TBS-T. Membranes were incubated for 45 min in TBS-T containing either 25 ng/ml peroxidase-coupled goat anti-rabbit IgG (Chemicon) or 0.2 μg/ml peroxidase-coupled goat anti-mouse IgG (KPL). Membranes were washed again three times for 10 min in TBS-T. After incubation in CHEMIGLOW™ substrate (Alpha Innotech), chemiluminescence was detected by a CCD camera (charge-coupled-device camera; Alpha Innotech).
COS-7 cells grown on glass coverslips were fixed at 1 day post-transfection in 4% (w/v) PFA (paraformaldehyde) in PBS for 10 min. Then the cells were incubated for 1 h in blocking buffer (PBS, 10% FCS, 3% BSA and 0.5% glycine) and incubated overnight at 4 °C in blocking buffer containing 3 μg/ml affinity-purified anti-SEA IgG or 2 μg/ml Tetra His monoclonal antibody (Qiagen). Cells were rinsed and washed three times for 10 min in TBS-T and then incubated for 45 min at room temperature in blocking buffer containing 1.3 μg/ml Cy3-labelled donkey anti-rabbit IgG (Jackson Immunoresearch Laboratories) or 2.5 μg/ml FITC-labelled donkey anti-mouse IgG (Jackson Immunoresearch Laboratories). Washing was performed as before. In the case of the analysis of permeabilized cells, 0.1% saponin was included in all antibody incubation steps. The coverslips were mounted upside down on to glass slips using DakoCytomation mounting medium (DakoCytomation) and stored at 4 °C. Analysis was performed on a Leica Leitz DM RXE microscope with a PL Fluotar objective. Pictures were taken with a Leica DFC350FX camera.
Anaesthetized P10 mice were perfused with PBS containing 4% PFA and the isolated organs were incubated in an increasing sucrose gradient. Cryosections of Tissue-Tek (Sakura) embedded brain (20 μm) sections on glass slips were processed with a similar protocol as cells in immunofluorescence (omission of the Tetra His and the corresponding secondary antibody, inclusion of 0.1% saponin during antibody incubation). Analysis was performed on an Olympus IX81F-2 microscope. Pictures were taken with an F-ViewII camera.
Cloning of mouse neurobin
To estimate the number of TTSPs expressed in differentiating mouse spinal cord, we screened approx. 5000 individual clones of a serine protease domain-enriched library. Enrichment for serine proteases encoding DNA was achieved by using degenerate primers encoding the highly conserved amino acid motifs around the catalytic histidine and serine residues respectively in the RT–PCR reaction. We identified 26 known serine proteases with approx. 50% of these not previously identified in the nervous system. Since this screen was not exhaustive, we suppose that an even larger number of serine proteases are simultaneously expressed in P10 spinal cord. We found five known TTSPs: spinesin, TMPRSS3, hepsin, MAT (mouse airway trypsin-like) and matriptase. Expression of the latter three TTSPs in the nervous system has not been reported before. Additionally, we isolated a clone encoding a hitherto uncharacterized serine protease domain. Full-length cloning of this cDNA by RACE indicated that the clone encoded a new serine protease belonging to the TTSP family. We designated this TTSP ‘neurobin’.
Structure of neurobin deduced from the cloned cDNA
The longest cDNA clone contained a 1293 bp ORF (open reading frame) encoding a protein with 431 amino acids with a calculated mass of 48 kDa (Figure 1). Hydropathy analysis indicated the presence of a single hydrophobic segment between residues 34 and 56 of the deduced amino acid sequence. This region was followed by a predicted SEA domain (residues 58–183), a short zymogen activation segment (residues 184–199) and a serine protease domain (residues 200–431). Within the protease domain the three residues histidine, asparagine and serine, forming the catalytic triad of serine proteases, were found at positions 240, 285 and 381 respectively, flanked by amino acids highly conserved in trypsin-like serine proteases. Three consensus motifs for N-linked glycosylation were identified in the putative extracellular part, one in the SEA domain and two in the protease domain. These structural characteristics, together with the absence of a hydrophobic signal peptide segment at the N-terminus, provide strong evidence that neurobin is a TTSP.
Neurobin cDNA and protein sequence and neurobin domain composition
Six highly conserved cysteine residues were identified in the protease domain (Figure 2). These cysteine residues usually form three disulfide bonds in serine protease domains (Cys42-Cys58, Cys168-Cys182, Cys191-Cys220, chymotrypsin numbering ). A seventh cysteine (Cys122) could theoretically form a disulfide bond to one of the two cysteine residues present in the zymogen activation domain, thereby linking the stem region and protease domain after activation cleavage. A homology search with the protease domain of neurobin uncovered highest sequence identities to the protease domains of other TTSPs such as DESC1 (58%), HAT (53%) and enteropeptidase (41%). Sequence identity to secreted serine protease-like trypsin, tPA, thrombin and Factor Xa was 34–37% (Figure 2).
Amino acid sequence alignment of the protease domain of neurobin with selected trypsin-like serine protease domains
Expression of neurobin in mouse tissue
For a first analysis of the tissue expression of neurobin in P10 mice, we performed RT–PCR analysis. As shown in Figure 3, neurobin was expressed in cerebellum and spinal cord. No signal was detected in liver, lung, spleen, kidney, heart and brain cortex, suggesting that neurobin, at this stage, has restricted tissue distribution. Expression in cerebellum and spinal cord was also detected in adult mice (results not shown).
Tissue expression of neurobin in P10 mice
Localization of neurobin in the nervous system and subcellular distribution
To identify neurobin in vivo, we used the affinity-purified antibody against the SEA domain. Histochemical analysis of brain sections revealed specific immunoreactivity in the cerebellum (Figure 4A), consistent with the RT–PCR analysis. Higher magnification indicated that this reactivity was confined to Purkinje neurons, in particular to the cell body and to the dendritic tree of Purkinje neurons (Figure 4B). Pre-immune serum showed no reactivity (Figure 4C). The restricted localization of neurobin is consistent with our inability to unambiguously detect neurobin in spinal cord and cerebellar protein extracts with our available antibodies (results not shown), suggesting that the overall amount of neurobin in vivo is fairly low.
Histo- and cyto-chemical detection of neurobin in mouse brain and in COS-7 cells
The distribution of neurobin was further studied by indirect immunofluorescence analysis in transiently transfected COS-7 cells. Using Myc–His-tagged neurobin, double staining of permeabilized cells with the anti-SEA domain (Figure 4D) and the anti-His antibody (Figure 4E) respectively revealed ER (endoplasmic reticulum)-like intracellular immunoreactivity (Figure 4E). With untagged neurobin, the intracellular pattern was more Golgi-like (Figure 4G) and immunoreactivity on the surface of non-permeabilized cells was much stronger (Figure 4I). These findings indicate that neurobin is present at the plasma membrane. The accumulation of tagged neurobin in the ER may, therefore, be caused by the presence of the C-terminal His tag.
Expression and activation of neurobin in transfected HEK-293T cells
To study neurobin protein production and zymogen activation, we transiently transfected HEK-293T cells with neurobin-encoding cDNA vectors. Protein production was tested by Western-blot analysis. With overexpression of wild-type neurobin and the anti-SEA domain antibody for analysis, we predominantly detected a product of 23 kDa (Figure 5A), a mass corresponding to neurobin without the protease domain. Using neurobin S381A, a mutant with the catalytic serine residue altered to alanine, a doublet signal at 50 kDa was obtained (Figure 5A), reflecting full-length neurobin. Transfection with empty vector gave no signal (Figure 5A). Using Myc–His-tagged neurobin, we obtained similar results, although processing of catalytically competent neurobin was less extensive (results not shown). These findings suggest that neurobin is subject to proteolytic processing, presumably autocatalytic zymogen activation, in HEK-293T cells.
Western-blot analysis of neurobin overexpressed in HEK-293T cells
To identify the activation cleavage site in the zymogen region, we created four new mutants (m2–m5) of neurobin–Myc–His S381A (designated m1 in Figure 5B). These mutants were overexpressed in HEK-293T cells without or with wild-type neurobin for activation. No protease domain was detected upon co-transfection with the empty pcDNA3.1 vector (Figure 5C, lanes p). Co-expression with wild-type neurobin (Figure 5C, lanes w) resulted in zymogen activation, indicated by the appearance of the Myc–His-tagged protease domain (Figure 5C, PD) in neurobin–Myc–His S381A (m1). With K199A (the P1 site) (m2), cleavage was strongly reduced, although a faint signal was still present. In the double-mutant m3 (R196A and K199A), processing was almost completely suppressed and, clearly, no protease domain signal was detected in the mutant m4, which contained no basic residues in the zymogen activation region. Mutant m5, a form carrying the activation site, but containing all other mutations, was still cleaved, indicating that the mutations themselves did not prevent neurobin from activating its zymogen form. The presence of autocatalytically activated wild-type neurobin was confirmed with the anti-SEA domain antibody (Figure 5C, lower panel). These results indicate that Lys199 is likely to be the sequence-predicted zymogen-activation cleavage site. The observation that mutant m3 was needed to efficiently suppress the release of the protease domain is an indication that possibly in vitro in the mutant forms neurobin can use a second, very similar motif for cleavage. Whether this cleavage can occur in vivo is unknown, but it is not very likely, given the observation that predominantly the hydrophobic residues isoleucine or valine contribute to further protease domain stabilization . Alternatively, a strong binding of neurobin residues Gly197 and His198 in the corresponding subsites might still enable peptide bond hydrolysis after Ala199.
Production of recombinant neurobin and enzymatic activity
To investigate the catalytic potential of neurobin, we used a construct encoding an N-terminal His tag, an enteropeptidase cleavage motif and the entire protease domain of neurobin beginning with Val200 and ending with Leu431. The protein was expressed in E. coli, solubilized from inclusion bodies and purified via nickel-affinity chromatography. After refolding, silver gel analysis of the purified protease domain revealed a predominant band of 30 kDa, the non-activated protease domain (Figure 6A). Activation of purified neurobin with enteropeptidase resulted in a new band of 27 kDa. Two additional faint bands were detected at the bottom of the gel (Figure 6A). Under non-reducing conditions, the apparent mass of activated neurobin was reduced to 24 kDa and the low-molecular-mass bands had disappeared, suggesting that these bands are disulfide-bridge-linked fragments of recombinant activated neurobin. To remove enteropeptidase after neurobin activation, MonoS cation-exchange chromatography was applied, resulting in homogeneously pure neurobin (Figure 6B). Size-exclusion chromatography indicated that activated neurobin is a monomeric protein (results not shown).
Production of recombinant neurobin
Activated neurobin was used to cleave chromogenic peptide substrates. Neurobin effectively cleaved peptides with an arginine residue in position P1 and uncharged residues in positions P2 and P3 (Figure 7A). The P1 acceptance of the arginine residue is consistent with the presence of an aspartic acid residue in position 375 (position 189 in chymotrypsin numbering), forming the bottom of the S1 subsite . In contrast with trypsin, which hydrolysed all our peptide substrates, neurobin did not accept a single arginine residue for cleavage and poor cleavage was measured with a proline residue in position P2. Hardly any cleavage was observed with acidic amino acids in positions P2 or P3. We obtained the best cleavage with the peptide Val-Gly-Arg-pNA. The peptide GHR-pNA, resembling the sequence of the zymogen-activation cleavage site (Gly-His-Lys, see Figure 1), was used for the experiments shown in Figures 7(B)–7(D). Neurobin had a pH optimum of approximately pH 8 (Figure 7B), and it was insensitive to a broad range of NaCl concentrations (Figure 7C). Peptide cleavage by neurobin was inhibited by a variety of serine protease inhibitors, but not by E-64 [trans-epoxysuccinyl-L-leucylamido-(4-guanidino)butane], a cysteine protease inhibitor, and pepstatin, an aspartic acid protease inhibitor. Heparin, added at two concentrations, did not suppress neurobin activity against peptide substrates (Figure 7D), and neither the removal of Ca2+, or other bivalent cations, by EDTA, nor the extra addition of Ca2+ ion, modulated neurobin activity (Figure 7D).
Specificity of recombinant neurobin against peptide substrate
Catalytic activity of recombinant neurobin towards protein substrates
Several previously identified TTSP substrates were recruited among extracellular matrix components, cell surface molecules and growth factors. Therefore, to identify potential protein substrates of neurobin, we incubated fibronectin, laminin, FGF-7 and 17 kDa FGF-2 (also referred to as 18 kDa FGF or low-molecular-mass FGF) with trypsin and recombinant neurobin respectively. All substrate proteins were degraded by trypsin, whereas neurobin did not cleave the large extracellular matrix proteins fibronectin and laminin, or the growth factor FGF-7 (Figure 8A). Neurobin, however, selectively processed FGF-2. This cleavage resulted in the formation of a 13 kDa fragment and further fragments (Figures 8A and 8B). Since FGF-2 has strong heparin binding properties , we tested whether the presence of heparin, at concentrations used during the peptide cleavage assays, modulated FGF-2 processing by neurobin. In the presence of heparin, FGF-2 cleavage by neurobin was suppressed in a dose-dependent manner (Figure 8A). A more detailed analysis of the FGF-2 cleavage showed that neurobin processed FGF-2 to at least three major fragments. Cleavage occurred after lysine or arginine residues, as deduced from sequence analysis of the cleavage products (Figure 8B). Fragments 1a, b and 4a begin with the N-terminus of recombinant human FGF-2. All other fragments start after a basic amino acid preceding the cleavage site. Cleaved FGF-2 was also analysed by MALDI–TOF-MS. Average masses of 12794.0, 8581.8 and 4348.8 Da were detected. These masses reflect the theoretical masses of Ala2-Arg116 (12792, Figure 8B, fragment 1a), Ile43-Arg116 (8517.6, Figure 8B, fragment 2) and Ser117-Ser154 (4349.1, Figure 8B, fragment 5a). Together with the peptide substrate assays, these enzyme activity experiments demonstrate that neurobin is an authentic serine protease with amino acid preference for an arginine or lysine residue in the P1 position and, possibly, very selective protein substrate specificity.
Activity of recombinant neurobin towards protein substrates
We identified a new TTSP, designated neurobin/TMPRSS11c. Neurobin has a short cytosolic N-terminus, a single transmembrane domain, an SEA domain and the C-terminal protease domain. With this domain organization, neurobin most resembles the members of the DESC/HAT group of TTSPs. In vivo, neurobin was found in Purkinje neurons of the cerebellum. Recombinant neurobin processes 17-kDa FGF-2 to several distinct fragments in a heparin-inhibitable manner.
The ORF of the isolated cDNA codes for a protein of 431 amino acids. It is very likely that neurobin indeed starts with the predicted methionine residue, since two in-frame stop codons, 45 and 30 bp upstream of the initiation codon respectively, are present in the 5′-non-translated part. Thus neurobin has a cytosolic domain of approx. 35 amino acids. Sequence analysis by NetPhos2.0 predicts that Ser9 in this domain is embedded in a motif suitable for serine/threonine phosphorylation. However, we do not know whether neurobin is indeed a kinase substrate in vivo.
Using neurobin cDNA sequences, our gene bank homology searches indicated that the neurobin gene lies on mouse chromosome 5. This region was previously described by Hobson et al.  as a locus containing seven DESC and HAT-like TTSP genes in a cluster . These authors also predicted the existence of neurobin (referred to as HAT-like 3). However, the protein sequence predicted from genome analysis was not fully identical with our cDNA-derived sequence (predicted length 418 amino acids, resulting from the absence of 13 amino acids forming the end of the SEA domain and the zymogen activation part). Gene predictions indicated that these 13 amino acids are present in horse and short-tailed opossum, but absent in rat, dog and cow. The relevance of this difference is, however, not known. Five of these DESC and HAT-like genes could be identified on human chromosome 4 . Our gene bank homology analysis and own sequencing of human genomic DNA indicated that the predicted last exon in the genomic region, possibly encoding a human orthologue, contained an 8 bp insertion. Additionally, the SEA domain-encoding exons could not be unambiguously predicted. These findings make expression of an authentic human neurobin from this gene locus very unlikely.
A possible absence of neurobin in the human does not exclude the absence of its function. The high sequence identity of the neurobin protease domain to the protease domain of the members of the highly related DESC/HAT group could have enabled one of these enzymes to take on proteolytic specificity and function in humans similar to those of neurobin in mouse. Such a concept is compatible with the finding that DESC2  and DESC4  are expressed in human brain tissue. Human DESC1 can cleave fibronectin , a protein not cleaved by neurobin. This points to not exactly neurobin-like specificity, despite the fact that DESC1 has the highest homology to neurobin. Alternatively, a human protease with neurobin-like function might have remote homology only, as previously observed with the identification of the murine metalloprotease Mcol-1, a possible counterpart of human MMP-1 (matrix metalloproteinase-1) . Finally, genomic analyses indicate that the human and mouse share some 166 serine protease genes, whereas nine are human-specific and 43 are mouse-specific . Hence, molecules present in one species only may be instrumental in highly species-specific functions.
Catalytic activation of neurobin can occur in an autocatalytic manner at the sequence-predicted site Gly-His-Lys199, as demonstrated by site-directed mutagenesis. Autoactivation has also been described for other TTSPs [11,13,31,32]. It remains to be investigated whether neurobin can also be activated by other proteases and is subject to further processing. Further analysis of autocatalytically activated wild-type neurobin under non-reducing conditions suggested that the protease domain and the membrane-anchored N-terminal part were not covalently linked (results not shown). In the absence of a covalent linkage, which could also result from a secondary cleavage just N-terminal to the activation domain, the protease domain could be released from the plasma membrane. In fact, shedding of TTSPs has been reported previously [33,34]. However, using our available antibodies, we did not detect neurobin immunoreactivity in the culture medium of neurobin-transfected HEK-293T or COS-7 cells (results not shown). This observation suggests that neurobin remains predominantly membrane-associated in these cell lines and is consistent with our immunofluorescence finding. Based on the determined crystal structure of the DESC1 protease domain, a strong interaction of the DESC1 protease domain back side with the preceding SEA domain was inferred . DESC1 protease domain residues implicated in this interaction are conserved in neurobin, suggesting that the neurobin protease domain could be involved in a tight interaction with its SEA domain. Such an interaction would be compatible with our experimental findings and could concentrate the catalytic site of neurobin at the cell surface.
Neurobin activity was not modulated by Ca2+ ions. As recently reviewed , allosteric regulation of serine proteases is mediated via Ca2+ complexation by acidic amino acids in positions 70 and 80 (chymotrypsin numbering). In neurobin, phenylalanine and leucine residues respectively are present, consistent with the insensitivity of neurobin towards calcium concentration changes. Using a broad range of Na+ concentrations for the peptide cleavage assays, we measured no significant differences in the catalytic activity of neurobin, suggesting that this enzyme is not allosterically regulated by Na+ ions. With respect to allosteric regulation, site-directed mutagenesis with thrombin revealed that residue 225 (either tyrosine or proline residue in serine proteases) is implicated in Na+-induced regulation of enzyme activity, because thrombin Y225P (chymotrypsin numbering) lost its ability to bind Na+ . In neurobin, Pro412, i.e. the residue compatible with Na+-insensitivity, occupies the homologous position. Pro412 is embedded in the sequence Asn-Lys-Pro-Gly. The entire motif is conserved in all DESC/HAT-like enzymes, a further point suggesting a relationship among these enzymes.
Neurobin processed 17-kDa FGF-2 to defined products of 13, 8.5, 6.4 and 4.3 kDa and N-terminal sequence analysis confirmed neurobin preference for lysine or arginine residue in the P1 position. The available data do not yet allow us, however, to define any sequence specificity of neurobin in the positions P2–P4. Nevertheless, that neurobin is likely to be a selective protease can be deduced from (i) the fact that the proteins fibronectin, laminin and FGF-7 were not processed, and (ii) the inability of neurobin to cleave the chromogenic substrate Arg-pNA, which is effectively hydrolysed by trypsin. Therefore, in the absence of an obvious substrate specificity strongly driven by P2–P4 positions, a possibly specific interaction of neurobin and FGF-2 might instead be mediated by neurobin exosites.
In addition to 17-kDa FGF-2, there are N-terminally elongated forms of FGF-2, with masses of 22, 22.5 and 24 kDa respectively . These forms, but not 17-kDa FGF-2, were cleaved by thrombin in capillary endothelial cells, and the cleaved forms stimulated endothelial migration and proliferation . Proteolytic processing of 17-kDa FGF-2 was previously reported to be carried out by the hyaluronan-binding protease from human plasma . Similar to our finding, proteolytic processing was inhibited by heparin. These findings suggest that FGF-2 molecules are subject to possibly isoform-specific proteolytic processing in vivo. Degradation of FGF-2 was first reported to occur during incubation with plasmin .
FGF-2 is found in many compartments including the extracellular space, but the protein is not secreted through the conventional ER pathway (for a review, see ). For the secretion of FGF-2, direct interactions with heparan sulfate proteoglycans are essential and might contribute to the accumulation of FGF-2 at the cell surface . Thus neurobin and FGF-2 might indeed co-localize at the cell surface in vivo. Given our finding that heparin suppresses FGF-2 cleavage by neurobin in vitro, heparan sulfates might furthermore play a role in the control of a possible cleavage of FGF-2 in vivo.
Interestingly, histochemical localization of FGF-2 in the cerebellum revealed strong expression in Purkinje cells with a distribution very similar to the distribution of neurobin , indicating that FGF-2 could indeed be an in vivo substrate of neurobin. Since FGF-2 plays a critical role in neuronal development [45,46], cleavage of FGF-2 by neurobin may modulate FGF-2-dependent functions during cerebellar development.
In summary, our work enabled the biochemical characterization of a hitherto uncharacterized TTSP of the nervous system, designated neurobin, with highest structural similarity to DESC/HAT-like TTSPs. The protease has a restricted tissue distribution with prominent presence in Purkinje neurons. In vitro, neurobin is able to process 17 kDa FGF-2 after several basic residues, but it does not cleave extracellular matrix molecules such as laminin and fibronectin. Further analysis is needed to assess the in vivo relevance.
We thank Dr Beat Kunz for expert assistance with antibody production, Professor Sabine Werner for the gift of FGF-7, Dr Peter Hunziker for committed help with the analysis of the FGF-2 cleavage products, Dr Birgit Dreier, Dr Peter Lindner and Dr Ned Mantei for a critical review of this paper prior to submission, and Dr Peter Sonderegger for continuous support. This work was supported by grants from the Swiss National Foundation (3100-065100.01, to S. M. G.), the Hartmann Müller Foundation (project 947, to S. M. G.), the EMDO Foundation (to S. M. G.) and the Julius Müller Foundation (to S. M. G.).
differentially expressed in squamous cell carcinoma
foetal calf serum
fibroblast growth factor-2
human airway trypsin-like
- HEK-293 cells
human embryonic kidney cells
- HEK-293T cells
HEK-293 cells expressing the large T-antigen of SV40 (simian virus 40)
matrix-assisted laser-desorption ionization–time-of-flight MS
open reading frame
postnatal day 10
rapid amplification of cDNA ends
sea-urchin sperm protein, enterokinase and agrin
transmembrane serine protease
tissue plasminogen activator
type II TMPRSS