Agric. Biol. Chem., 55 (5), 1383-1390, 1991 1383

Cloning, Sequencing, and Characterization of the Intracellular Invertase Gene from Zymomonasmobilis Hideshi Yanase,*'1" Hideaki Fukushi,** Naoki Ueda,** Yukitoshi Maeda,** Atsushi Toyoda** and Kenzo Tonomura*** * Department of Biotechnology, Faculty of Engineering, Tottori University, Tottori 680, Japan ** Department of Agricultural Chemistry, Faculty of Agriculture, University of Osaka Prefecture, Sakai 591, Japan *** Department of Food Technology, Faculty of Engineering, Fukuyama University, Fukuyama 729-02, Japan Received December 13, 1990

The structural gene for the intracellular invertase El of Zymomonasmobilis strain Z6C was cloned in a 2.25-kb DNAfragment on pUSHll, and expressed in Eschevichia coli HB101.The enzyme produced by the E. coli carrying pUSHll was purified about 1,122 fold to homogeneity with a yield of 4%. The molecular weight and substrate specificity of the enzymewere identical with those of the intracellular invertase El from Z. mobilis. The nucleotides of the cloned DNAwere sequenced; they included an open reading frame of 1,536bp, coding for a protein with a molecular weight of 58,728. The N-terminal amino acid sequence predicted was identical with the sequence of the first 20 N-terminal amino acid residues of the protein obtained by Edmandegradation. Comparison of the predicted amino acid sequence of El protein with those of the four other known^-D-fructofuranosidases from Escherichia coli, Bacillus subtilis, and Saccharomyces cerevisiae indicated a stronger homologyin the N-terminal portion than in the C-terminal portion.

Zymomonasmobilis is a Gram-negative its nucleotide sequence was determined and bacterium capable of producing almost the characterized by comparing with genes encod- theoretical yield of ethanol from glucose, and ing for invertase from E. coli,6) B. subtilis,1^ has higher productivity than the yeasts now and S. cerevisiae.9) used industrially. However, this bacterium can ferment only glucose, fructose, and sucrose. In addition, there is little information on the early Materials and Methods metabolism of sucrose, although it has been Plasmids and bacteria used. The bacteria used in this known that sucrose is hydrolyzed to glucose study were E. coli HB101 (F~, hsdS20, recA13, ara-14, and fructose by invertase that is produced proA2, lacYl, galK2, rpsL20, xyl-5, mtl-1, supE44, k~), inducibly,1} and the liberated fructose is and Z. mobilis Z6C, a spontaneous mutant of Z. mobilis converted to sorbitol2'3) and Ievan4'5) in part. subsp. mobilis IFO13756 (=ATCC29191) capable of Preliminary investigation suggested that there producing an extracellular levansucrase (E2) and invertase (E3), constitutively. 1} The plasmids used were pZA22 (Cmr, were three kinds of sucrose-hydrolyzing Tcr, 6.7kb), a shuttle vector between Z. mobilis and E. enzymesin Zymomonas,intracellular invertase coli,10) and pZL7 (lac'Z+, Y+, Cmr, 16.2kb), a lacY+ (that was namedEl), extracellular levansucrase plasmid that can coexist with pUC118or pUC119in E. (E2), and extracellular invertase (E3). There- coli HBIOI.1" fore, we have attempted to clone genes encoding for these enzymes and characterize Cultivation. Z. mobilis was cultivated statically at 30°C in RMmedium (2% glucose, 1% yeast extract, and 0.2% them. This paper describes that the El gene was KH2PO4, pH 6.0) and E. coli in the LB medium (1% cloned from Zymomonasgenomic library, and bacto-tryptone, 0.5% bacto-yeast extract, and 1 %NaCl, To whomreprint request should be addressed. 1384 H. Yanase et al. pH 7.0) at 37°C with shaking. Whennecessary, ampicillin the E. coli HB101carrying pUSHll as follows. The (25 Mg/ml), tetracycline (1 5 fig/ml), and/or chloramphenicol transformant cells harvested from 7 1 of the culture medium (100 /ig/ml for Z. mobilis, 25 /jg/ml for E. coli) were added were washed with 20mMphosphate buffer (pH 7.0), and to the medium. Cells of each organism were harvested by suspended in 500 ml of the same buffer, then disrupted by centrifugation and washed with 20mMphosphate buffer sonication. Thesupernatant solution was fractionated with (pH 7.0). (NH4)2SO4 precipitation between 30 to 50% saturation. The resulting precipitate was dissolved in the above buffer DNAmanipulation. Standard methods were used for the and dialyzed. The enzyme solution was put onto a purification of DNA, plasmid constructions, and transfor- DEAE-Toyopearl 650M column (2.3 x 30 cm) equilibrated mation ofE. coli.12) The plasmids were introduced into Z. with the buffer. The enzyme was recovered from un- mobilis by the transformation method using partial absorbed fractions, and dialyzed against 20mM acetate spheroplasts. 1 3) buffer, pH 5.0. The dialyzed solution was put onto a CM-Sephadex C-50 column (1.2 x 30 cm) equilibrated with Construction of Z. mobilis gene library. Z. mobilis the dialyzed buffer. After the column was washed chromosomal DNAwas isolated by the method of Saito thoroughly with the same buffer, the enzymewas eluted and Miura14) with a partial modification. It was partially with a linear gradient from 0.1 to 1.0m NaCl. Fractions digested with BamHl, and the resulting fragments having showing the enzyme activity were pooled and concentrated a size of about 8kb were ligated into the BamHlsite of in a CENTRICUTU-20 (Kurabo Biomedical P/J). The pZA22, and then transformed E. co// HB101.Recombinant enzyme solution was fractionated with (NH4)2SO4 of 0 to clones were selected for sucrose-hydrolyzing activity in 10%saturation. The resultant precipitate was dissolved in toluenized cells. 20mMacetate buffer, pH 5.0. The El enzyme from Z. mobilis Z6C cells was partially Deletion analysis. To identify the coding region of the purified as follows. Cells harvested from 7 1 of the culture El gene on pZSHl, the 7.85-kb 5amHI-fragment was medium were suspended in 700ml of 20mMphosphate subcloned into pUC118 or pUC119 using E. coli HB101 buffer, pH 7.0, and incubated at 30°C for lOmin with carrying pZL7. E. coli carrying pZL7 could take up shaking to release cell-bound levansucrase (E2) and raffinose by the lactose permease encoded by lacY on invertase (E3) from cells. The washed cells were disrupted pZL7.15) If the gene encoding for the Z. mobilis by sonication, and centrifuged to remove intact cells and sucrose-hydrolyzing enzymewas cloned in the strain, the debris. The supernatant solution was put onto a recombinant clone would be expected to use the fructose DEAE-Toyopearl 650M column (2.3 x 30 cm) equilibrated moiety ofraffinose. This clone, however, could not use the with 20mM phosphate buffer, pH 7.0, and unabsorbed galactose moiety of raffinose or melibiose, because the fractions were pooled. The enzyme was precipitated by strain is galK~. A clone capable of using raffinose was adding (NH4)2SO4 to 50% saturation, and the resultant selected by a red colony formed on a 2,3,5-triphenyl precipitate was dissolved in a small volume of 20mM tetrazolium chloride (TTC)-indicator plate, which con- phosphate buffer, pH 7.0, containing 0.2m NaCl. The tained 2% raffinose, 0.7% KH2PO4, 0.3% K2HPO4, enzymesolution was put onto a SephadexG-200column 0.01% MgSO4, 0.2% proteose peptone, 0.00025% TTC, (2.3 x 100cm) equilibrated with 20mMphosphate buffer, 0.0025% ampicillin, and 1.5% agar.16) pH 7.0, containing 0.2m NaCl, and chromatographed. Fractions having the enzyme activity were pooled and Enzyme and protein assay. The sucrose-hydrolyzing concentrated in a CENTRICUTU-20. activity was assayed at 30°C. The standard assay mixture contained 0.15 m sucrose, 0.1 macetate buffer, pH 5.0, and Molecular weight measurement. The molecular weight in an enzymesolution or toluenized cells in total volumeof a native form was estimated by gel filtration through a 1 ml. After incubated at 30°C, reducing sugar formed in Sephadex G-200 column (1.0x50cm) with 20mM the reaction mixture was measuredby the method of phosphate buffer, pH 7.0 containing 0.2m NaCl. Somogyiand Nelson, and one unit of the activity was Cytochrome c, chymotrypsinogen A, ovalbumin, BSA, defined as the amount of enzymethat produced reducing aldolase, and catalase were used as marker proteins. The sugar equivalent to 1 /miol of glucose per min. The amounts molecular weight of the subunit was measured by the of glucose and fructose liberated were measured with an method of Laemmli.19) Marker proteins used were: F-Kit for Glucose-Fructose (Boehringer Mannheim, a-lactalbumin, soybean trypsin inhibitor, carbonic anhy- GmBH).Protein concentration was assayed by the method drase, ovalbumin, BSA, and phosphorylase b. of Lowry et aL,17) and also by absorption at 280nm. Polyacrylamide gel electrophoresis was done by the Analysis of N-terminal amino acid sequence. The El procedure of Davis.18) enzyme obtained in the final step of purification from E. coli HB101 carrying pUSHll was blotted onto a Purification of the sucrose-hydrolyzing enzymes. The El polyvinylidene difluoride membrane (Millipore), and protein produced by the cloned El gene was purified from degraded sequentially with a protein analyzer, Applied Z. mobilis Intracellular Invertase Gene 1385

Fig. 1. Restriction Maps ofpZSHl and pUSHll. Symbols: H, Z. mobilis chromosomal DNA; =, pZA22; , pUC118.

Biosystems model 4701A, equipped with an on-line HPLC apparatus model 120A.

DNA sequencing. The DNA sequence of the gene encoding for the ZymomonasEl enzyme was analyzed by the dideoxy method of Sanger et al.20) The entire sequence of the insert contained in pUSHll was analyzed in both directions using an overlapping set of Exonuclease Ill-generated deletions that had been cloned into derivatives of phage Ml3.

Results and Discussion Cloning of the gene encoding a sucrose- Fig. 2. Deletion Plasmids Derived from pZSH1. hydrolyzing enzyme The specific activity of sucrose-hydrolyzing enzyme in cell-free extracts prepared from E. coli clones are indicated. Plasmids containing our BamHIlibrary of The plasmids, pUSHl, pUSH5, pUSH6, and pUSHll, Z. mobilis chromosomal DNAin pZA22were were constructed by inserting a DNAfragment indicated screened for the presence of a sucrose- in the figure into pUC118, and pUSH3 and pUSH4 into hydrolyzing enzyme in E. coli, and seven pUC119. Uppercase letters show only relevant restriction positive clones were obtained. One of these endonuclease sites in the Z. mobilis DNAinserted. Deleted clones, which contained a plasmid designated and remaining portions of the insert in derivatives are pZSHl, was chosen for further investigation. indicated by lines and open boxes, respectively. Arrows This plasmid contained a 7.85-kb BamHI- indicate the transcriptional direction of lac promoter. Abbreviations: B, BamUl; C, Clal; EV, EcoRV; H, Hindlll; fragment inserted into pZA22, as shown in Fig. S, Sad. 1. To identify the coding region of the enzyme on the 7.85-kb fragment, we did subcloning of pUC118to construct pUSHl. A deletion experiments using pUC118 or pUC119 as a mutant, pUSH3,was constructed by inserting vector, as shown in Fig. 2. First, the 7.85-kb the 5.4-kb Hindlll-Bamm fragment of PUSH l fragment on pZSHl was inserted into the into pUC119 with the same transcriptional BamHl site that is located in multi-clonine sites direction of lac promoter. Plasmid pUSH4was 1386 H. Yanase et al constructed by inserting the 3.4-kb EcoRY- Ill to produce pUSHll. The results suggested BamHl fragment of pUSHl into the EcoR- that the gene encoding for a sucrose-hy- I(filled)-^mHI site of pUC119 with the drolyzing enzymeand its ownpromoter were opposite direction of lac promoter. Plasmid located on the 2.25-kb fragment in pUSHll. pUSH5 was constructed by inserting the As the cell-free extracts of E. coli HB101 2.85-kb Sacl-BamUl fragment of pUSHl into carrying pUSHl 1 showed about 25-fold higher pUC118, and the 2.1-kb CM(filled)-BamHI activity than that of E. coli carrying pUSHl, fragment of pUSHl inserted into pUC118at an excess region that inhibited expression of the EcoRI(filled)-BamlII site to generate the gene in E. coli seemed to be deleted by pUSH6. The deletion mutant pUSH5retained Exonuclease III. The restriction map of the sucrose-hydrolyzing activity in E. coli. Next, the 2.85-kb of Sacl~BamHl fragment on pUSH5 was further deleted approximately 500bp from the BamHIsite by Exonuclease

Fig. 4. Disc Gel Electrophoresis of the Cloned El Fig. 3. Chromatograms of the Sucrose-hydrolyzing Enzyme. Enzymes from Z. mobilis Z6C and E. coli HB101 Carrying (A) Polyacrylamide gel electrophoresis. Purified enzyme pUSHll on a DEAE-Toyopearl 650M Column. was put on a 7.5% gel column and run at pH4.3 for A, Z. mobilisZ6C (-O-); B, E. co/i HB101 withpUSHl 1 40min at 5mAper gel. The direction of electrophoresis (-#-) or without (-O-)à" Cell-free extracts prepared was from upper (anode) to lower (cathode). from washed cells of E. coli or Z. mobilis from 100ml of (B) SDS-polyacrylamide gel electrophoresis. Purified culture were put on the column (1.0 x 12cm) equilibrat- enzyme treated with SDSat 100°C for 5min was put on ed with 20mMphosphate buffer, pH 7.0, washed with a 10% gel containing 0.1% SDS, are run at 7mAper gel 10ml of the same buffer, and then eluted with 50ml of for 5 hr. The direction of electrophoresis was from upper a linear gradient NaCl from 0 to 0.5m ( ), collecting (cathode) to lower (anode). Arrows indicate molecular 1 ml per tube. weights ( x 10~3) of marker proteins.

Table I. Purification of the Sucrose-hydrolyzing Enzyme from E. coli HB101Carrying pUSHl l Total protein Total activity Specific activity , Yield eP (mg) (units) (units/mg) ° (% ) Cell-free extract 10,840 22, 140 2.0 1 100 (NH4)2SO4 ppt 5,487 23,160 4.2 2.1 105 (30-50% saturated) DEAE-Toyopearl 650M 258 17,750 68.4 33.5 80 CM-Sephadex C-50 1 5.8 5,080 322 158 23 1st (NH4)2SO4 ppt 1.58 1,989 1,259 617 9 2nd (NH4)2SO4 ppt 0.16 801 2,288 1,122 4 Z. mobilis Intracellular Invertase Gene 1387

Fig. 5. Sequence and Translation of the Z. mobilis El Gene. The -35 region, - 10 region, and Shine-Dalgarno (S.D.) region deduced are underlined. 1388 H. Yanase et at.

pUSHll is shown in Fig. 1. The resultant These results suggested that the enzymewas a 2.25-kb fragment in pUSHl l was sequenced. /?-D-fructofuranosidase, called invertase. The enzyme showed the maximumactivity at pH Characterization of the gene product 6.5 and was stable between pH 5.0 and 7.0. A preliminary experiment suggested that Z. The highest reaction rate was observed at 37°C, mobilis Z6Cproduced three different sucrose- and the enzymewas thermolabile. The enzyme hydrolyzing enzymes, El, E2, and E3. To activity was inhibited by 1mMof mercuric clarify which enzymewas coded for in the compounds such as HgCl2, HgO, PCMB, 2.25-kb fragment of pUSHll, the cell-free mercuric acetate, and mersalyl acid. extracts prepared from E. coli HB101carrying To find whether El of the E. coli trans- pUSHll was chromatographed on a DEAE- formant was identical with the El enzyme of Toyopearl 650Mcolumn. The chromatogram Z. mobilis Z6C or not, the El enzyme was of the cell-free extract from Z6C cells (Fig. 3A) partially purified from the cell-free extract of gave three distinct peaks of sucrose-hydrolyz- Z6Ccells, and compared with the enzymeof ing activity; the first peak was responsible for the E. coli transformant. The molecular weight intracellular El enzyme, while the second and of the El enzyme was 59,000 by Sephadex third peaks were responsible for extracellular G-200 gel filtration. The enzyme hydrolyzed E2 and E3 enzymes, respectively. As most of sucrose, raffinose, and inulin with a relative E2 and E3 enzymes were present in the washing ratio of100 : 46 : < 1. Themolecularweightand solution, the amount of E2 and E3 enzymes the substrate specificity of the cloned enzyme found in this cell-free extract was equivalent were identical with those of the El enzyme, to about 10% of the total of both enzymes. suggesting that the 2.25-kb fragment coded for From the elution profile shown in Fig. 3B, it the intracellular invertase El ofZ. mobilis Z6C. was clear that the El gene in Z. mobilis Z6C was cloned in the 2.25-kb fragment on Sequence of the intracellular invertase El gene pUSHll. The sequence of the 2.25-kb DNAfragment The El enzyme was purified from cells ofE. on pUSHll containing El enzyme was coli carrying pUSHll about 1,122-fold with a analyzed in both directions, as shown in Fig. yield about 4%, as shown in Table I. The 5. A single open reading frame was found enzymepreparation was shownto be homoge- starting with an ATGcodon at position 371 neous by the criteria of polyacrylamide gel from the.if/wdlll end and ending with a TAA electrophoresis, as shown in Fig. 4. The termination codon at position 1,907, including molecular weight of the enzyme was 58,000 by Sephadex G-200 gel filtration, and 57,500 by SDS-polyacrylamide gel electrophoresis. It was a basic protein with an isoelectric point of pH 8.5. The first 20 N-terminal amino acids of the enzyme obtained by Edmandegradation were Met-Glu-Ser-Pro-X-Tyr-Lys-Asn-Leu-Ile-Lys- Ala-Glu-Asp-Ala-Gin-Lys-Lys-Ala-Gly-. The enzyme catalyzed the hydrolysis of sucrose to liberate glucose and fructose with a equimolar ratio, and also raffinose to fructose and melibiose. In addition, it catalyzed the Fig. 6. Dot Matrix Analysis ofHomology of the Ammo hydrolysis of inulin, but not maltose, meletzi- Acid Sequence between the Zymomonas Sucrose-hydrolyz- tose, or levan (data not shown). The relative ing Enzyme El and the E. coli Sucrose Hydrolase. activities for these substrates were 100 for Criteria for scoring are defined as 3 or more out of 10 sucrose, 54 for raffinose, and < 1 for inulin. amino acid residues. Z. mobilis Intracellular Invertase Gene 1389

Fig. 7. Comparison of the AminoAcid Sequence of the ZymomonasSucrose-hydrolyzing EnzymeEl, Sucrose Hydrolase (rafD) from E. coli,6) Sucrase (sacA) from B. subtilis^ Levanase (sacC) from B. subtilis,8) and Invertase (suc2) from S. cerevisiae.9) Identical residues in the first 278 amino acids are boxed. The arrow indicates a cystein residue in the putative active site.

1,536bp and coding for a protein with the consensus sequence of promoter region was molecular weight calculated to be 58,728. This also predicted as A/G*****C/AT*G***---10 was almost identical with the molecular weight to 16 bases---TA*T/AG*A/T*T. In the El of the El protein. The first 20 N-terminal gene the sequence ofAAAAGCCTTGCGA-- amino acid residues predicted from the 8 bases---TATAGAAAAstarting at 120 bases nucleotide sequence was also identical with upstream from the start codon seemed to be a those of the El enzyme obtained by Edman promoter region, as shownin Fig. 5. The results degradation. of deletion analysis also implied that a promoter Although we have not yet confirmed the that could be recognized by E. coli RNA initiation site of the El gene, we tried to find polymerase was upstream of the El gene. A a candidate for the Shine-Dalgarno (S.D.) sequence of TTGTCT---15 bases---TATAAT sequence and the promoter region, - 35 region starting at 55 bases upstream from the start and - 10 region, of the gene. Recently, pyruvate codon that seemed to adhere to the -35 and decarboxylase (pdc), alcohol dehydrogenase - 10 consensus sequence of E. coli promoter (adhB), glyceraldehyde-3-phosphate dehydro- was found. genase (gap), phosphoglycerate kinase (pgk), and phosphatase (phoC) of Z. mobilis have Comparison of the El gene with other /?-D- been sequenced, and its S.D. sequence and fructofuranosidase genes promoter region were implied by pond et al.21) As mentioned above, the El enzyme was an The consensus S.D. sequence was seemedto intracellular invertase, jS-D-fructofuranosidase. be *GGAG*10 to 16 bases upstream from the The predicted amino acid sequence of the El translational start codon. Then by searching the gene was compared to that of other four analogous sequence upstream of the El gene, /?-D-fructofuranosidases, the intracellular su- the sequence, AGGCAG, starting 1 1 bases up- crose hydrolase encoded by rafD on the stream from the translational start codon was plasmid-borne raf operon from E. coli,6) the found, and deduced as the S.D. sequence. The intracellular sucrase encoded by sacA,n) extra- 1390 H. Yanase et al. cellular levanase encoded by sacC from Bacillus References subtilis,8) and the extracellular invertase H. Yanase, M. Yasui, T. Miyazaki and K. Tonomura, encoded by suc2 from Saccharomyces cer- Agric. Biol. Chem., 49, 133 (1985). evisiae.9) The composition of a dot matrix L. Viikari, Appl. Microbiol. BiotechnoL, 20, 118 analysis among these enzymes showed, in (1984). M. Zachariou and R. K. Scopes, /. BacterioL, 167, general, stronger homology in the N-terminal 863 (1986). portion than in the C-terminal portion (data E. A. Dawes and D. W. Ribbons, Biochem. J., 98, not shown). In particular, the El enzymegave 804 (1966). an extensive homologyto the E. coli sucrose E. W. Lyness and H. W. Doelle, BiotechnoL Lett., 5, hydrolase, as shown in Fig. 6. Aligned 345 (1983). C. Aslanidis, K. Schmid and R. Schmitt, J. BacterioL, sequences of the N-terminal portion of these 111, 6753 (1989). enzymes are shown in Fig. 7. A comparison of A. Fouet, A. Klier and G. Rapoport, Gene, 45, 221 the El enzyme with the E. coli sucrose (1986). hydrolase indicated that the score of identity I. Martin, M. Debarbouille, E. Ferrari, A. Klier and was 50.7% in the first 280 amino acid residues. G. Rapoport, Mol. Gen. Genet., 208, 177 (1987). R. Taussig and M. Carlson, Nucleic Acids Res., ll, The comparison with other enzymesshowed 1943 (1983). the score was about 40% in all cases. Several N. Misawa, T. Okamoto, K. Nakamura, K. homologous sequences are shown in this figure Kitamura, H. Yanase and K. Tonomura, Agric. Biol. with a boxed frame. Chem., 50, 3201 (1986). Theaminoacid sequence around a cysteine H. Yanase, J. Kurii and K. Tonomura /. Ferment. residue at aa position 230, which was suggested TechnoL, 66, 409 (1988). T. Maniatis, E. F. Fritsch and J. Sambrook, to be an active site of levanase by Martin et "Molecular Cloning," A Laboratory Manual, Cold al. 8) Trp(Tyr)-Glu-Cys-Pro-Asp(Gly)-Leu, was Spring Harbor Laboratory, New York, 1982. found in the El enzyme with a minor H. Yanase, T. Kotani and K. Tonomura, Agric. Biol. modification. Martin et al. have indicated a Chem., 50, 3139 (1986). tempting speculation that sucrase, levanase, H. Saito and K.-J. Miura, Biochim. Biophys. Ada, 72, 619 (1963). and invertase have evolved by fusion of two H. Yanase, M. Masuda, T. Tamaki and K. DNAsegments corresponding to the N- and Tonomura, J. Ferment. Bioengineer., 70, 1 (1990). C-terminal parts of the protein and that the B. R. Bochner and M. A. Savageau, Appl. Environ. DNAsequence encoding the N-terminal part Microbiol., 33, 434 (1977). Of these proteins derived from a Common17) °- H- Lowry,N. J. Rosebrough,A. L. Farrand R. J. Randall, /. Biol. Chem., 193, 265 (1951). ancestor.8) There was no conflict between our B. J. Davis, Ann. N.Y. Acad. ScL, 121, 404 (1964). data obtained from the El enzyme of 19) U. K. Laemmli, Nature, 227, 680 (1970). Zymomonasand their speculation. 20) F. Sanger, S. Nicklen and A. R. Coulson, Proc. Natl. Acad. Sci. U.S.A., 74, 5463 (1977). Acknowledgment. This work was supported by a 21) J.L.Pond,C.K.Eddy,K.F.Mackenzie,T.Conway, Grant-in-Aid for Scientific Research from the Ministry of D- J- Borecky and L. O. Ingram, /. Bacteriol., ill, Education, Science and Culture of Japan. 767 (1989). 18)