Cryptic Genes for Cellobiose Utilization in Natural Isolates of Escherichia coli

The ECOR collection of natural Escherichia coli isolates was screened to determine the proportion of strains that carried functional, cryptic and nonfunctional genes for utilization of the three beta-glucoside sugars, arbutin, salicin and cellobiose. None of the 71 natural isolates utilized any of the beta-glucosides. Each strain was subjected to selection for utilization of each of the sugars. Only five of the isolates were incapable of yielding spontaneous beta-glucoside-utilizing mutants. Forty-five strains yielded cellobiose+ mutants, 62 yielded arbutin+ mutants, and 58 strains yielded salicin+ mutants. A subset of the mutants was screen by mRNA hybridization to determine whether they were expressing either the cel or the bgl beta-glucoside utilization operons of E. coli K12. Two cellobiose+ and two arbutin+-salicin+ strains failed to express either of these known operons. It is concluded that there are at least four gene clusters specifying beta-glucoside utilization functions in E. coli populations, and that all of these are normally cryptic. It is estimated that in any random isolate the probability of any particular cluster having been irreversibly inactivated by the accumulation of random mutations is about 0.5.     

Biochemical Genetics of the Cryptic Gene System for Cellobiose Utilization in Escherichia coli K12

The cellobiose catabolic system of Escherichia coli K12 is being used to study the role of cryptic genes in microbial evolution. Wild-type E. coli K12 do not utilize the beta-glucoside sugars, arbutin, salicin and cellobiose. A Cel+ (cellobiose utilizing) mutant which grows on cellobiose, arbutin, and salicin was isolated previously from wild-type E. coli K12. Biochemical assays indicate that a cel structural gene (celT) specifies a single transport protein that is a beta-glucoside specific enzyme of the phosphoenolpyruvate-dependent phosphotransferase system. The transport protein phosphorylates beta-glucosides at the expense of phosphoenolpyruvate. A single phosphoglucosidase, specified by celH, hydrolyzes phosphorylated cellobiose, arbutin, and salicin. The genes of the cel system are expressed constitutively in the Cel+ mutant, whereas they are not expressed at a detectable level in the wild-type strain. The transport and hydrolase genes are simultaneously silenced or simultaneously expressed and thus constitute an operon. Cel+ strains which fail to utilize one or more beta-glucosides express the transport system at a lower level than do Cel+ strains which grow on all three beta-glucosides. Other strains inducibly express a gene which specifies transport of arbutin but not the other beta-glucosides. The arbutin transport gene, arbT, maps outside of the cel locus.     

Identification of a new locus in the Escherichia coli cotransduction gap that represents a new genetic component of the L-asparagine utilization system

A temperature-sensitive Escherichia coli mutant defective for the ability to utilize L-asparagine as a sole nitrogen source was isolated after N-methyl-N'-nitro-N-nitrosoguanidine mutagenesis. The mutation (asu) produces two distinct phenotypic effects. Mutant strains grow poorly at high temperature on minimal plates containing asparagine as the sole nitrogen source; this effect is greatly exacerbated by the presence of methionine. Mutant strains utilize L-asparagine as a nitrogen source three to four times more efficiently at permissive temperatures than the wild-type strains. The mutation maps at 32.4 min on the E. coli chromosome, within the E. coli cotransduction gap. Mutant strains produce normal amounts of thermo-stable L-asparaginase I activity. The mutation therefore affects a component of the asparagine utilization system other than the catabolism of asparagine within the cell; it probably affects asparagine uptake.     

Anaerobic Fusiform-shaped Bacteria Isolated from the Caecum of Conventional Mice

Twenty-four strains of anaerobic fusiform-shaped bacteria were isolated from the caecum of conventional mice. Ten of the isolates belonged to the genus Clostridium , nine to Fusobacterium and five to Bacteroides. The clostridia were put into eight biotypes on the basis of their ability to: hydrolyze mucin and esculin; produce indole and hydrogen; utilize pyruvate, and ferment arabinose, cellobiose, fructose, glucose, galactose, maltose, mannose and N-acetylglucosamine and on the acid end-products of such fermentations. The fusobacteria were assigned to six biotypes based on their ability to hydrolyze casein, mucin and starch, and to ferment arabinose, lactose, maltose and N-acetylglucosamine. Each of the bacteroides was deemed to be a separate biotype because of the differences in their ability to hydrolyze casein and starch, grow in bile medium, utilize pyruvate, and to ferment arabinose, maltose, xylan and N-acetylglucosamine. Five of the Fusobacterium isolates resembled F. russii. The remaining 19 isolates did not resemble previously described species of anaerobic bacteria.     

Directed evolution of cellobiose utilization in Escherichia coli K12

The cellobiose catabolic system of Escherichia coli K12 is being used to study the role of cryptic genes in evolution of new functions. Escherichia coli does not use beta-glucoside sugars; however, mutations in several loci can activate the cryptic bgl operon and permit growth on the beta-glucoside sugars arbutin and salicin. Such Bgl+ mutants do not use cellobiose, which is the most common beta-glucoside in nature. We have isolated a Cel+ (cellobiose-utilizing) mutant from a Bgl+ mutant of E. coli K12. The Cel+ mutant grows well on cellobiose, arbutin, and salicin. Genes for utilization of these beta-glucosides are located at 37.8 min on the E. coli map. The genes of the bgl operon are not involved in cellobiose utilization. Introduction of a deletion covering bgl does not affect the ability to utilize cellobiose, arbutin, or salicin, indicating that the new Cel+ genes provide all three functions. Spontaneous cellobiose negative mutants also become arbutin and salicin negative. Analysis of beta-glucoside positive revertants of these mutants indicates that there are separate loci for utilization of each of the beta-glucoside sugars. The genes are closely linked and may be activated from a single locus. A fourth gene at an unknown location increases the growth rate on cellobiose. The cel genes constitute a second cryptic system for beta-glucoside utilization in E. coli K12.      

Possible mechanisms underlying the slow lactose fermentation phenotype in Shigella spp.

A Southern hybridization analysis revealed that the region homologous to Escherichia coli lacZ was present on the chromosomal DNAs of beta-galactosidase-positive Shigella strains, such as Shigella dysenteriae serovar 1 and Shigella sonnei strains, whereas this region was absent from chromosomal DNAs of beta-galactosidase-negative strains of Shigella flexneri and Shigella boydii. We found that the lacY-A region was deficient in S. dysenteriae serovar 1 and believe that this is the reason for the slow fermentation of lactose by this strain. S. sonnei strains possessed the region which hybridized with E. coli lacY-A despite their slow hydrolysis of lactose. The whole lactose-fermenting region was cloned from S. sonnei and compared with the cloned lac operon of E. coli K-12. Both clones directed the synthesis of beta-galactosidase in an E. coli K-12 strain lacking indigenous beta-galactosidase activity (strain JM109-1), and we observed no difference in the expression of beta-galactosidase activity in S. sonnei and E. coli. However, E. coli JM109-1 harboring the lactose-fermenting genes of S. sonnei exhibited the slow lactose fermentation phenotype like the parental strain. S. sonnei strains had no detectable lactose permease activities. E. coli JM109-1 harboring the lactose-fermenting genes of S. sonnei had a detectable permease activity, possibly because of the multicopy nature of the cloned genes, but this permease activity was much lower than that of strain JM109-1 harboring the lac operon of E. coli K-12. From these results we concluded that slow lactose fermentation by S. sonnei is due to weak lactose permease activity.     

Characterization and Nucleotide Sequence of the Cryptic Cel Operon of Escherichia Coli K12

Wild-type Escherichia coli are not able to utilize beta-glucoside sugars because the genes for utilization of these sugars are cryptic. Spontaneous mutations in the cel operon allow its expression and enable the organism to ferment cellobiose, arbutin and salicin. In this report we describe the structure and nucleotide sequence of the cel operon. The cel operon consists of five genes: celA, whose function is unknown; celB and celC which encode phosphoenolpyruvate-dependent phosphotransferase system enzyme IIcel and enzyme IIIcel, respectively, for the transport and phosphorylation of beta-glucoside sugars; celD, which encodes a negative regulatory protein; and celF, which encodes a phospho-beta-glucosidase that acts on phosphorylated cellobiose, arbutin and salicin. The mutationally activated cel operon is induced in the presence of its substrates, and is repressed in their absence. A comparison of proteins encoded by the cel operon with functionally equivalent proteins of the bgl operon, another cryptic E. coli gene system responsible for the catabolism of beta-glucoside sugars, revealed no significant homology between these two systems despite common functional characteristics. The celD and celF encoded repressor and phospho-beta-glucosidase proteins are homologous to the melibiose regulatory protein and to the melA encoded alpha-galactosidase of E. coli, respectively. Furthermore, the celC encoded PEP-dependent phosphotransferase system enzyme IIIcel is strikingly homologous to an enzyme IIIlac of the Gram-positive organism Staphylococcus aureus. We conclude that the genes for these two enzyme IIIs diverged much more recently than did their hosts, indicating that E. coli and S. aureus have undergone relatively recent exchange of chromosomal genes.     

Possible mechanisms underlying the slow lactose fermentation phenotype in Shigella spp.

A Southern hybridization analysis revealed that the region homologous to Escherichia coli lacZ was present on the chromosomal DNAs of beta-galactosidase-positive Shigella strains, such as Shigella dysenteriae serovar 1 and Shigella sonnei strains, whereas this region was absent from chromosomal DNAs of beta-galactosidase-negative strains of Shigella flexneri and Shigella boydii. We found that the lacY-A region was deficient in S. dysenteriae serovar 1 and believe that this is the reason for the slow fermentation of lactose by this strain. S. sonnei strains possessed the region which hybridized with E. coli lacY-A despite their slow hydrolysis of lactose. The whole lactose-fermenting region was cloned from S. sonnei and compared with the cloned lac operon of E. coli K-12. Both clones directed the synthesis of beta-galactosidase in an E. coli K-12 strain lacking indigenous beta-galactosidase activity (strain JM109-1), and we observed no difference in the expression of beta-galactosidase activity in S. sonnei and E. coli. However, E. coli JM109-1 harboring the lactose-fermenting genes of S. sonnei exhibited the slow lactose fermentation phenotype like the parental strain. S. sonnei strains had no detectable lactose permease activities. E. coli JM109-1 harboring the lactose-fermenting genes of S. sonnei had a detectable permease activity, possibly because of the multicopy nature of the cloned genes, but this permease activity was much lower than that of strain JM109-1 harboring the lac operon of E. coli K-12. From these results we concluded that slow lactose fermentation by S. sonnei is due to weak lactose permease activity.     

A Fourth Escherichia Coli Gene System with the Potential to Evolve β-Glucoside Utilization

Escherichia coli K12 is being used to study the potential for adaptive evolution that is present in the genome of a single organism. Wild-type E. coli K12 do not utilize any of the beta-glucoside sugars arbutin, salicin or cellobiose. It has been shown that mutations at three cryptic loci allow utilization of these sugars. Mutations in the bgl operon allow inducible growth on arbutin and salicin while cel mutations allow constitutive utilization of cellobiose as well as arbutin and salicin. Mutations in a third cryptic locus, arbT, allow the transport of arbutin. A salicin+ arbutin+ cellobiose+ mutant has been isolated from a strain which is deleted for the both the bgl and cel operons. Because the mutant utilized salicin and cellobiose as well as arbutin, it is unlikely it is the result of a mutation in arbT. A second step mutant exhibited enhanced growth on salicin and a third step mutant showed better growth on cellobiose. A fourfold level of induction in response to arbutin and a twofold level of induction in response to salicin was observed when these mutants were assayed on the artificial substrate p-nitrophenyl-beta-D-glucoside. Although growth on cellobiose minimal medium can be detected after prolonged periods of time, these strains are severely inhibited by cellobiose in liquid medium. This system has been cloned and does not hybridize to either bgl or cel specific probes. We have designated this gene system the sac locus. The sac locus is a fourth set of genes with the potential for evolving to provide beta-glucoside utilization.     

Lactobacilli isolated from the stomach of conventional mice.

Twenty strains of lactobacilli isolated from the stomach of conventional mice were tested for their ability to ferment or hydrolyze substrates that may be present in the stomach habitat. The lactobacilli could be placed in four groups (A to D) depending on their ability to ferment N-acetylglucosamine, dextrin, cellobiose, gum arabic, and xylan. The majority of the isolates belonged to groups A and D. Group A strains did not resemble previously described Lactobacillus species, but group D strains were identified as L. leichmannii. A representative group A isolate colonized the surface of the nonsecretory epithelium of the stomach of gnotobiotic mice; a group D isolate did not.     

Effect of 2,6-diaminopurine resistance mutations on adenine and adenosine assimilation by cells of an adenine-dependent strain of Escherichia coli K-12]

In purine-requiring strain of Escherichia coli K-12 defective in purine nucleoside phosphorylase (pur, pup) mutants (designated apt) have been obtained that are resistant to 2,6-diaminopurine on guanine-containing medium and incapable to utilize adenine for their growth at 42degreesC, but they are still sensitive to the analogue and can utilize adenine at 28degreesC. It has been shown that the introduction of the corresponding apt mutations in the genome of adenine-requiring strains impaired the ability of these strains to grow on both adenine and adenosine at 42degreesC.     

Maintenance of the cellobiose utilization genes of Escherichia coli in a cryptic state

The genes for cellobiose utilization are normally cryptic in Escherichia coli. The cellobiose system was used as a model to understand the process by which silent genes are maintained in microbial populations. Previously reported was (1) the isolation of a mutant strain that expresses the cellobiose-utilization (Cel) genes and (2) that expression of those genes allows utilization of three beta- glucoside sugars: cellobiose, arbutin, and salicin. The Cel gene cluster has now been cloned from that mutant strain. In the course of locating the Cel genes within the cloned DNA segment, it was discovered that inactivation of the Cel-encoded hydrolase rendered the host strain sensitive to all three beta-glucosides as potent inhibitors. This sensitivity arises from the accumulation of the phosphorylated beta- glucosides. Because even the fully active genes conferred some degree of beta-glucoside sensitivity, the effects of cellobiose on a series of five Cel+ mutants of independent origin were investigated. Although each of those strains utilizes cellobiose as a sole carbon and energy source, cellobiose also acts as a potent inhibitor that reduces the growth rate on glycerol 2.5-16.5-fold. On the other hand, wild-type strains that cannot utilize cellobiose are not inhibited. The observation that the same compound can serve either as a nutrient or as an inhibitor suggests that, under most conditions in which cellobiose will be present together with other resources, there is a strong selective advantage to having the cryptic (Cel0) allele. In those environments in which cellobiose is the sole, or the best, resource, mutants that express the genes (Cel+) will have a strong selective advantage. It is suggested that temporal alternation between these two conditions is a major factor in the maintenance of these genes in E. coli populations. This alternation of environments and fitnesses was predicted by the model for cryptic-gene maintenance that was previously published.      

Widespread distribution of deletions of the bgl operon in natural isolates of Escherichia coli

A deletion that includes the bgl (beta-glucoside utilization) operon of Escherichia coli was originally detected in several rarely occurring natural isolates that utilize cellobiose. Here I show that bgl deletions are present in 95% of the Cel+ isolates obtained from diverse sources. They are also present in 29% of the Cel- strains in two different collections of natural isolates of E. coli. At least three versions of bgl deletions are present in E. coli populations. In the most common version approximately 8 kb of DNA around the bgl region of E. coli K12 is replaced by a specific 6.5-kb DNA fragment. In another version a deletion of similar length is not replaced by the same sequence. A third version involves deletion of approximately 14 kb without the replacement fragment being present. The distribution of these deletions suggests that the version 1 deletion occurred very early in the history of E coli. It also appears likely that there is selection for bgl deletions in Cel+ strains of E. coli. The presence of the version 1 deletion within distantly related phylogenetic groups of E. coli provides evidence for recombination within natural populations of E coli.      

Genes for ribitol and D-arabitol catabolism in Escherichia coli: their loci in C strains and absence in K-12 and B strains.

Escherichia coli C strains can grow at the expense of the two natural pentitols ribitol and D-arabitol, sugar alcohols previously thought not to be utilized by E. coli. E. coli strains K-12 and B cannot utilize either compound. The genetic loci responsible for pentitol catabolism in E. coli C, designated rtl and atl, are separate and closely linked. Each lies between metG and his and is highly co-transducible with metG and with a P2 prophage attachment site. rtl and atl readily can be transduced into E. coli K-12 or B strains, in which they integrate at, or very near, their E. coli C location. Transduction also can be used to insert rtl and atl into certain E. coli K-12 F' plasmids. No recombination between E. coli C strains and either K-12 or B strains occurs within the rtl-atl genetic region after interstrain conjugations or transductions. No cryptic rtl or atl genes in K-12 or B strains can be detected by complementation, recombination, or mutagenesis. These results are consistent with the view that the rtl-atl portion of the E. coli C chromosome has no counterpart in E. coli K-12 or B and may have been obtained from an extrageneric source. Detailed biochemical and genetic comparisons of penitol utilization in E. coli and Klebsiella aerogenes are in progress. The ability to catabolize xylitol is conferred upon E. coli C strains by a mutation at or adjacent to the rtl locus, whereas in E. coli K-12 or B strains harboring rtl an additional mutation at a separate locus is required for xylitol utilization.     

Analysis of the Erwinia chrysanthemi arb genes, which mediate metabolism of aromatic beta-glucosides.

Erwinia chrysanthemi is one of the few members of the family Enterobacteriaceae that is capable of metabolizing most of the naturally occurring beta-glucosides. We previously isolated the clb genes, which allow the use of the disaccharide cellobiose as well as the aromatic beta-glucosides arbutin and salicin. We report here the isolation of the arb genes, which permit fermentation of the aromatic beta-glucosides only. Establishment of a functional Arb system in Escherichia coli depended on the presence of the phosphotransferase system and on the activation by the cyclic AMP-cyclic AMP receptor protein complex. Strains carrying mini-Mu-induced LacZ fusions to the arb genes were used to analyze arb genes organization and function. Three arb genes (arbG, arbF, and arbB) were identified and organized in this order. Genetic and structural evidence allowed us to assign a phospho-beta-glucosidase and a permease activity to the ArbB and ArbF proteins, respectively. Several Lac+ arb-lacZ insertions were introduced into the E. chrysanthemi chromosome. Both ArbG- and ArbF- strains were unable to ferment the aromatic beta-glucosides, whereas ArbB- strains were impaired only in salicin fermentation. None of the mutations in the arb genes affected cellobiose metabolism. The expression of the arb genes was substrate inducible and required the ArbF permease and, possibly, the ArbG protein. Collectively, our results underline the resemblance between the naturally expressed E. chrysanthemi arbGFB and the cryptic E. coli bglGFB operons, yet the arbG gene product seemed unable to activate E. coli bgl operon expression.     

Functional genes for cellobiose utilization in natural isolates of Escherichia coli.

The genes for utilization of cellobiose are normally cryptic in both laboratory strains and natural isolates of Escherichia coli. A survey of natural isolates of E. coli reveals that functional genes for cellobiose utilization, while rare, are present. The fraction of E. coli that utilized cellobiose ranged from less than 0.01% in human fecal samples to 7% in fecal samples obtained from horses. Samples obtained from sheep, cows, dogs, and pigs contained 0.1 to 0.5% cellobiose-positive E. coli. Neither the previously identified cel genes nor the bgl genes from E. coli K-12 were expressed during growth on cellobiose by any of the 14 naturally occurring Cel+ isolates that were tested. All of the naturally occurring Cel+ isolates possessed a cel operon, but all were deleted for the major portion of the bgl operon. The functional cel+ genes from these natural isolates differed from the mutationally activated cel+ genes obtained in earlier studies in that (i) the mutationally activated cel+ genes were temperature sensitive, while the functional genes were not, and (ii) transport of cellobiose was inducible in the strains carrying functional cel+ genes, while it was expressed constitutively in strains carrying mutationally activated genes.     

DNA sequences and expression in Streptomyces lividans of an exoglucanase gene and an endoglucanase gene from Thermomonospora fusca.

Two genes encoding cellulases E1 and E4 from Thermomonospora fusca have been cloned in Escherichia coli, and their DNA sequences have been determined. Both genes were introduced into Streptomyces lividans, and the enzymes were purified from the culture supernatants of transformants. E1 and E4 were expressed 18- and 4-fold higher, respectively, in S. lividans than in E. coli. Thin-layer chromatography of digestion products showed that E1 digests cellotriose, cellotetraose, and cellopentaose to cellobiose and a trace of glucose. E4 is poor at degrading cellotriose and cleaves cellopentaose to cellotetraose and glucose or cellotriose and cellobiose. It readily cleaves cellotetraose to cellobiose. E1 shows 59% identity to Cellulomonas fumi CenC in a 689-amino-acid overlap, and E4 shows 80% identity to the N terminus of C. fimi CenB in a 441-amino-acid overlap; all of these proteins are members of cellulase family E. Alignment of the amino acid sequences of Clostridium thermocellum celD, E1, E4, and four other members of family E demonstrates a clear relationship between their catalytic domains, although there is as little as 25% identity between some of them. Residues in celD that have been identified by site-directed mutagenesis and chemical modification to be important for catalytic activity are conserved in all seven proteins. The catalytic domains of E1 and E4 are not similar to those of T. fusca E2 or E5, but all four enzymes share similar cellulose-binding domains and have the same 14-bp inverted repeat upstream of their initiation codons. This sequence has been identified previously as the binding site for a protein that regulates induction.     

Effect of Increased beta-Glucosidase Activity on Virulence of Erwinia amylovora

Plant tissues often contain beta-glucosides that can be enzymatically hydrolyzed to produce toxic aglycones. It has been suggested that the low beta-glucosidase activity found in Erwinia amylovora contributes to bacterial virulence by allowing the bacteria to infect plants that contain beta-glucosides without inducing the formation of toxic aglycones. To test this suggestion, we created strains of E. amylovora which had high beta-glucosidase activities and studied the ability of these strains to cause fire blight disease in pears (Pyrus communis). We isolated spontaneous mutants that were able to utilize beta-glucosides as the sole carbon source and showed that one class had about 10 times as much beta-glucosidase activity as the wild-type strain. In addition, we constructed several plasmids that carry the Escherichia coli bgl operon under the control of a transposon Tn5 promoter that is expressed in E. amylovora. These plasmids were introduced in E. amylovora by transformation. Pathogenesis studies in immature Bartlett pear fruits, etiolated sprouts, and young shoots showed that a 100-fold increase in beta-glucosidase activity does not interfere with normal development of fire blight disease in these model systems.     

Isolation and molecular characterization of high-performance cellobiose-fermenting spontaneous mutants of ethanologenic Escherichia coli KO11 containing the Klebsiella oxytoca casAB operon.

Escherichia coli KO11 was previously constructed to produce ethanol from acid hydrolysates of hemicellulose (pentoses and hexoses) by the chromosomal integration of Zymomonas mobilis genes encoding pyruvate decarboxylase (pdc) and alcohol dehydrogenase (adhB). Klebsiella oxytoca P2 was constructed in an analogous fashion for the simultaneous saccharification and fermentation of cellulose and contains PTS enzymes for cellobiose. In this study, KO11 was further engineered for the fermentation of cellulose by adding the K. oxytoca casAB genes encoding Enzyme IIcellobiose and phospho-beta-glucosidase. Although the two K. oxytoca genes were well expressed in cloning hosts such as DH5 alpha, both were expressed poorly in E. coli KO11, a derivative of E. coli B. Spontaneous mutants which exhibited more than 15-fold-higher specific activities for cellobiose metabolism were isolated. The mutations of these mutants resided in the plasmid rather than the host. Three mutants were characterized by sequence analysis. All contained similar internal deletions which eliminated the casAB promoter and operator regions and placed the lacZ Shine-Dalgarno region immediately upstream from the casA Shine-Dalgarno region. KO11 harboring mutant plasmids (pLOI1908, pLOI1909, or pLOI1910) rapidly fermented cellobiose to ethanol, and the yield was more than 90% of the theoretical yield. Two of these strains were used with commercial cellulase to ferment mixed-waste office paper to ethanol.     

Temperature Sensitivity of Maltose Utilization and Lambda Resistance in Escherichia coli B

Escherichia coli B strains that have acquired the malB region from E. coli K-12 are able to utilize maltose and to adsorb phage lambda when grown at 30 C, but when grown at 40 C they do not absorb phage lambda and are devoid of amylomaltase activity. These Mal(ts) Lam(ts) cells can be mutated or transduced to become able to grow on maltose at 40 C, but they still have no detectable amylomaltase activity nor functional lambda receptors at that temperature. This Mal(40) phenotype is governed by a gene located near or at malA. It is suggested that the temperature sensitivity of both characters results from a defect in malT. However, transduction of malA from E. coli B to E. coli K-12 results in a wild-type phenotype, whereas E. coli B cells that have acquired malA from E. coli K-12 donors are still temperature sensitive for both amylomaltase and lambda-receptor production.