A sequence of 91 amino acids residues, probably starting from the N-terminal of the mature protein, was determined for the 105-kDa protein of the non-haemolytic enterotoxin of Bacillus cereus. The last part of this sequence was similar to parts of the N-terminal portions of two collagenases of Clostridium histolyticum and Clostridium perfringens. Zymography, with intact collagen fibril and gelatin as substrates, showed that the 105-kDa protein had collagenolytic and gelatinolytic activity. The 105-kDa protein also showed activity against a typical collagenase substrate, azocoll, and was inhibited by EDTA and 1,10-phenanthroline. We conclude that the 105-kDa protein is a collagenase.
The diarrhoeal type of food poisoning caused by Bacillus cereus is due to protein enterotoxins . It has been suggested that haemolysin BL (Hbl), composed of the three proteins B, L1 and L2, is a diarrhoeal virulence factor of B. cereus. Another non-haemolytic enterotoxin (Nhe), composed of the 45-kDa protein (NheA), the 39-kDa protein (NheB) and the 105-kDa protein, was isolated from a strain of B. cereus which was involved in a large food poisoning outbreak in 1995 [3,4]. A gene, nheC, in the same operon as nheA and nheB, with the possible product NheC, may also be involved in Nhe . An interesting and probably unique feature of these complexes is their dependence on at least three components for maximal toxicity. The functions of the individual components of Nhe are unknown, but it has been proposed that the 105-kDa protein is responsible for binding Nhe to the target cells . The degree of cytotoxicity of the Nhe complex appears to vary from strain to strain and this variation has been suggested to be due to differences in the 105-kDa protein . Furthermore, the activity of Nhe may be reduced due to inactivation of the 105-kDa protein during purification. There is an immunoassay which detects the 45-kDa protein of Nhe, but there is no assay to detect the activity of the 105-kDa protein. Further characterisation of the 105-kDa protein of Nhe of B. cereus is therefore important.
Some bacteria produce collagenases which cleave native collagen. In contrast to vertebrate collagenases, the bacterial collagenases possess a broad substrate specificity and degrade both native collagen and gelatin . Collagenases from Clostridum histolyticum and Clostridium perfringens are regarded as virulence factors, being implicated in tissue destruction [7,8]. The genes encoding two collagenases, from C. histolyticum and C. perfringens, have been cloned and sequenced [9,10]. ColH (from C. histolyticum) has a molecular mass of 116 kDa, while the colA gene (C. perfringens) encodes a collagenase of 120 kDa. They are metalloproteases, containing zinc metalloprotease consensus sequences. A collagenase of 90 kDa from B. cereus, which resembles the C. histolyticum collagenase(s), has also been characterised . Here, we show that the amino acid sequence of a N-terminal part of the 105-kDa protein of Nhe is similar to ColH and the collagenase encoded by colA. Due to these similarities, we have investigated whether the 105-kDa protein has gelatinolytic and collagenolytic activity and found that it possesses both.
2Materials and methods
Strain 1230-88 was used throughout. This strain was responsible for an outbreak of diarrhoeal syndrome food poisoning in Norway in 1988 . The culture medium used for production of enterotoxins was a modification of the CGY medium . The cells were grown at 32°C for 5 h. When indicated, EDTA (1 mM) was added at the time of harvesting. Extracellular proteins were separated from the cells by centrifugation (10 000×g at 4°C for 20 min). The supernatant proteins were precipitated with 70% saturated (NH4)2SO4.
2.2Purification of proteins
When EDTA was added at the time of harvesting, proteins were isolated as described in . Chromatography on a DEAE-Sephacel column (Pharmacia) with Bistris-HCl buffer at pH 5.9 was followed by chromatography on a HA column (Bio-Rad), with sodium phosphate buffer at pH 6.8, and finally, a Resource Q (ReQ) column (Pharmacia) with triethanolamine buffer at pH 7.5 was used. EDTA (1 mM) was used in the dialysis buffers before the first and last purification step. This purification scheme was used for purification of the 39- and 45-kDa proteins and for the 105-kDa protein which was used for protein sequencing. Chromatography on DEAE-Sephacel and ReQ columns was also used for purification of proteins for characterisation of enzyme activity. In this case, however, the chromatography was carried out in 50 mM Tris-HCl and 5 mM CaCl2, pH 7.5, with a NaCl gradient from 0 to 0.5 M and the proteins were dialysed in 50 mM Tris-HCl and 5 mM CaCl2 before chromatography.
The 105-kDa protein (0.5 mg ml−1) was treated with thermolysin (Sigma) (100:1, w/w) for 30 min at 37°C at pH 7.8. Cleaved fragments were purified by chromatography on ReQ at pH 7.8. A fragment of approximately 50 kDa, eluting at 175 mM NaCl, was used for sequence determination.
Purified proteins and fragments were sequenced from the N-terminus by Edman degradation using an Applied Biosystems 477 A automatic sequence analyser with an on-line 120 A phenylthiohydantoin amino acid analyser.
2.5Cloning of PCR products and DNA sequencing
A PCR product, obtained with degenerate primers deduced from the amino acid sequences of fragments of the 105-kDa protein, was cloned using the pMOSBlue T-vector kit (Amersham) following the manufacturer's instructions. Sequencing of the cloned PCR product was performed on a Perkin Elmer ABI Prism 377 automatic sequencer. A Dye Terminator Cycle Sequencing Ready Reaction kit was used following the manufacturer's instructions.
Sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) was carried out using a Bio-Rad Mini-Protean II Dual Slab Cell. The gels (10% acrylamide) were stained using the Bio-Rad Silver Stain Plus kit and the molecular mass of the proteins was estimated using Bio-Rad SDS-PAGE standards.
Gelatin zymography was carried out essentially as described in . The samples were applied to a SDS-10% polyacrylamide gel, which was co-polymerised with gelatin (0.1%). After electrophoresis, SDS was removed by washing the gel with 2.5% Triton X-100 in 100 mM NaCl and then, the gel was incubated overnight at 37°C in 50 mM Tris-HCl (pH 7.5) containing 10 mM CaCl2 and finally stained with Coomassie brilliant blue R-250. The method of Birkedal-Hansen and Taylor  was used for zymography with collagen fibril as substrate, using acid-soluble type I collagen from calf skin (Sigma). After electrophoresis, SDS was removed by washing with 2.5% Triton X-100 and the gel was rinsed in phosphate-buffered saline (pH 7.4) containing 1 mM CaCl2 and 1 μM ZnCl2. The gel was then incubated on top of a collagen fibril film overnight at 37°C. The collagen fibril film was stained with Coomassie brilliant blue R-250 to visualise unstained bands.
2.8Assay for collagenase
Collagenase activity was assayed with azocoll by the method of Kameyama and Akama  with modifications as described by Matsushita et al. . Azocoll (Sigma) powder was washed with and resuspended in 0.2 M boric acid/borate buffer (pH 7.2), containing 0.15 M NaCl to give a final concentration of 5 mg ml−1 azocoll. Enzyme samples (8–20 μl) were added to 0.5 ml of this assay mixture and incubated with shaking at 37°C for 1 h. In some experiments, to the purified 105-kDa protein in 50 mM Tris-HCl, 5 mM CaCl2 (pH 7.5) was EDTA added at a final concentration of 10 mM. This mixture was either assayed with azocoll directly or passed through a PD-10 column (Pharmacia) to remove metal-chelator complexes and excess EDTA. Divalent metal ions at different concentrations were then added to the 105-kDa protein before it was assayed with azocoll as substrate.
2.9Vero cell assay
Toxicity was determined using the inhibition of protein synthesis in Vero cells .
2.10Amino acid sequence similarity search
A database search for similar protein sequences was carried out by using the BLAST mail server  at the National Centre for Biotechnology Information in the National Library of Medicine, National Institutes of Health (Bethesda, MD, USA). The amino acid sequence of the 105-kDa protein was used as query sequence and this was compared with the non-redundant protein databases, including GenBank CDS translations, PDB, SwissProt, SPupdate and PIR.
3Results and discussion
We have previously described the purification of the 105-kDa protein of B. cereus Nhe and a degradation product of the protein . These two proteins have now been purified from strain 1230-88 and the amino acid sequences of their N-terminal ends were determined (Fig. 1). The analysis gave 40 amino acids from the largest fragment and 23 amino acids from the smallest, without overlapping sequences. Degenerated DNA primers were made from the sequences and used for PCR. A PCR product of approximately 150 bp was cloned and sequenced. The deduced amino acid sequence overlapped with both amino acid sequences derived from N-terminals and a total sequence of 72 amino acids was established (Fig. 1). It appeared that the slightly degraded form lacked the 49 first amino acids of the intact protein. A fragment of 50 kDa, resulting from digestion of the 105-kDa protein with thermolysin, was also purified. The N-terminal sequence of this fragment overlapped with the original sequence and a sequence of 91 amino acids was obtained (Fig. 1). This sequence was compared to known protein sequences, using the BLAST algorithm . The sequence with the highest similarity to the 105-kDa protein was that of a putative collagenase precursor from another strain of B. cereus (accession number Y11141). A new search with the putative collagenase sequence, using BLAST, revealed a high similarity with collagenases from C. histolyticum and C. perfringens. Comparison of the 105-kDa protein and the collagenases revealed that the sequence from amino acid 65 to 91 of the 105-kDa protein was similar to a N-terminal part of the mature collagenases (Fig. 2).
Crude extracellular proteins (supernatant) produced by B. cereus strain 1230-88 (which previously has been used for studies of Nhe) were analysed for gelatinolytic activity by zymography. Proteins giving a double band at approximately 105 kDa contained the main gelatinolytic activity, but several faint bands were also visible, the lowest at approximately 80 kDa (Fig. 3A, lane 2). The supernatant proteins precipitated with ammonium sulfate (lane 1) gave a similar pattern. The intact 105-kDa protein and the degraded form lacking the first 49 amino acids were also gelatinolytic (lane 6) and corresponded to the main gelatinolytic bands of the supernatant. Proteins to be characterised for collagenase (and gelatinase) activity were purified by chromatography, using buffers containing 5 mM CaCl2 to stabilise collagenases as recommended by Matsushita et al. . The main gelatinolytic activity was eluted from the DEAE column at 225 mM NaCl (lane 3) and corresponded to the protein in the lower of the main gelatinolytic bands of the supernatant (lane 2) (the slightly degraded form of the 105-kDa protein). However, this fraction contained an additional gelatinolytic band, corresponding to the lowest, faint band of the supernatant proteins (80 kDa). The protein of 80 kDa also co-purified with the 105-kDa protein on ReQ (lane 5). The specific gelatinolytic activity of the smaller protein was probably much higher than that of the 105-kDa protein, judged from comparison of the silver staining (no band at 80 kDa was visible) (Fig. 3B, lane 2) and zymography (Fig. 3A, lane 5). The main gelatinolytic protein (105-kDa protein) and that of 80 kDa were partly separated by HA chromatography and pure 105-kDa protein was obtained by the use of HA chromatography before a final step on the ReQ column (Fig. 3A, lane 4). Silver staining showed only the main gelatinolytic protein (Fig. 3B). When the proteins in a corresponding SDS-gel were incubated overnight with EDTA instead of CaCl2, no gelatinolytic activity was visible. The zymography with gelatin as substrate showed that the intact 105-kDa protein of Nhe and the slightly degraded form (lacking 49 amino acids in the N-terminal end) were responsible for the main gelatinolytic activity of the extracellular proteins of B. cereus strain 1230-88.
Zymography with acid-soluble type I collagen fibril as substrate showed that the slightly degraded form of the 105-kDa protein was able to degrade native collagen (Fig. 4, lanes 3–5), while no collagenolytic activity was observed from the 80-kDa gelatinase fragment. The ammonium sulfate-precipitated proteins (Fig. 4, lane 1) gave only one faint band, which corresponded to the slightly degraded form of the 105-kDa protein. The supernatant proteins (lane 2) gave only a very faint band in the same position.
The 105-kDa protein was assayed with azocoll as substrate. The specific activity of the enzyme was 1.2 U mg−1 when assayed in borate buffer. When EDTA was added to the reaction (final concentration 10 mM), the activity was reduced to 5% of the original. After removal of EDTA, divalent metal ions were added to see if the activity against azocoll could be restored. Addition of ZnCl2 (final concentration 30 μM) completely restored the activity, but higher concentrations (1 mM) of ZnCl2 restored the activity to only 55% of the original. When CaCl2 (30 μM and 1 mM) and MgCl2 (30 μM and 1 mM) were added, the activity was restored to 61–67% of the original. Addition of 1,10-phenanthroline (final concentration 1 mM) to the assay buffer inhibited the activity of the enzyme completely. The collagenase encoded by colA has also been assayed with azocoll as substrate, giving similar results . Together, our results indicate that the 105-kDa protein of Nhe is a collagenase similar to the clostridial collagenases.
An important question about the activity of Nhe is its dependence on the enzymatic function of the 105-kDa protein. Some preliminary tests were performed to find out if Nhe activity was influenced by divalent cations, as was the case for the enzymatic activity of the 105-kDa protein. Vero cells were treated with 105-kDa protein, 105-kDa protein in EDTA (1 and 5 mM), 105-kDa protein in CaCl2 (1 mM) and 105-kDa protein in CaCl2 and ZnCl2 (1 mM and 30 μM). The 39- and 45-kDa proteins were also added in each case. No significant difference in cytotoxicity was observed, indicating that Nhe is not dependent on the enzymatic activity of the 105-kDa protein. It should, however, be kept in mind that the medium used for the toxicity tests contained CaCl2 and that some reactivation of enzyme activity may have taken place even when EDTA was added to the toxin at the beginning of the test.
We have previously indicated that the intact 105-kDa protein is more cytotoxic than the form lacking the first 49 amino acids . The gelatinolytic activity of these two forms, however, seems to be similar (Fig. 3A, lane 6). It could therefore be that these 49 amino acids, which show no similarity in sequence to the other collagenases, are necessary for the activity of the 105-kDa protein in Nhe, but not for the enzymatic activity. Small amounts of other proteins were found together with the 105-kDa protein in some fractions with toxic activity during the purification. Some of these proteins may be of importance for Nhe as additional factors.
In addition to food poisoning, B. cereus can cause other diseases such as post-traumatic and metastatic endophthalmitis. It has been suggested that the ocular virulence is multifunctional and that Hbl contributes to virulence . For endophthalmitis, in general, there an association has been noted between the severity of the illness and the ability of the infecting organism to produce exotoxins and proteases have been noted . Due to the occurrence of collagen structures in the eye, it will be of great interest to investigate if the 105-kDa protein and Nhe contribute to endophthalmitis virulence. For the same reason, it will also be of interest to investigate if the 105-kDa protein of Nhe contributes to dermonecrosis caused by B. cereus.
The authors thank Kristin O'Sullivan for technical assistance and The Research Council of Norway and TINE for financial support through Grant 119303/112.