The expression of cDNA clones of yeast M1 double-stranded RNA in yeast confers both killer and immunity phenotypes

Secretion of a bacterial cellulase by yeast

Gene fusions were constructed between a yeast expression plasmid and a Cellulomonas fimi DNA fragment encoding an endo-1,4-beta-D-glucanase or carboxymethylcellulase. Yeast transformed with the recombinant plasmids secreted carboxymethylcellulase activity. Secretion of active enzyme was greatly increased when the leader of a secreted yeast protein, the Kl toxin, was inserted immediately upstream of and in frame with the bacterial cellulase sequence. This is the first step in constructing a functional cellulase complex in Saccharomyces cerevisiae. It also provides an excellent system for the detailed examination of the determinants of protein secretion because of the ease with which secreted cellulase can be detected.

Immunity to K1 killer toxin: internal TOK1 blockade

K1 killer strains of Saccharomyces cerevisiae harbor RNA viruses that mediate secretion of K1, a protein toxin that kills virus-free cells. Recently, external K1 toxin was shown to directly activate TOK1 channels in the plasma membranes of sensitive yeast cells, leading to excess potassium flux and cell death. Here, a mechanism by which killer cells resist their own toxin is shown: internal toxin inhibits TOK1 channels and suppresses activation by external toxin.

Studies on strong and weak killer phenotypes of wine yeasts: production, activity of toxin in must, and its effect in mixed culture fermentation

Two different killer phenotypes were detected among K+ (killer) yeasts isolated from spontaneous wine fermentations using a plate bioassay. The two phenotypes differed in their degree of killer activity, and were designated as SK+(strong killer) and WK+(weak killer). Strains showing either phenotype were assayed for expression of killer activity under different growth conditions. Growth in must negatively affected expression of the killer activity of both phenotypes. The supernatant fluids from must cultures showed a lower killing effect than those from yeast phosphate dextrose broth (YPDB) cultures. The ability of the two K+ phenotypes to prevail on K-sensitive yeasts was studied in mixed-culture fermentation experiments. Under these conditions, only strains showing SK+ phenotype were able to prevail on the K-sensitive yeasts. These results suggest that the K+ phenotype could play a relevant role in spontaneous fermentations provided that the strain exhibits an SK+ phenotype, and that the latter phenotype should be preferred when selected K + strains are to be used as fermentation starters.

Cloning, sequencing and expression of a full-length cDNA copy of the M1 double-stranded RNA virus from the yeast, Saccharomyces cerevisiae

Strains of the budding yeast, Saccharomyces cerevisiae, may contain one or more cytoplasmic viruses with double-stranded RNA (dsRNA) genomes. The killer phenomenon in yeast, in which one cell secretes a killer toxin that is lethal to another cell, is dependent upon the presence of the L-A and M1 dsRNA viruses. The L-A viral genome encodes proteins for the viral capsid, and for synthesis and encapsidation of single-stranded RNA replication cycle intermediates. The M1 virus depends upon the L-A-encoded proteins for its capsid and for the replication of its killer-toxin-encoding genome. A full-length cDNA clone of an M genome has been made from a single dsRNA molecule and shown to encode functional killer and killer-immunity functions. The sequence of the clone indicates minor differences from previously published sequences of parts of the M1 genome and of the complete genome of S14 (an internal deletion derivative of M1) but no unreported amino acid variants and no changes in putative secondary structures of the single-stranded RNA. A 118-nucleotide contiguous segment of the M1 genome has not previously been reported; 92 of those nucleotides comprise a segment of A nucleotides in the AU-rich bubble that follows the toxin-encoding reading frame.

Expression cassettes for formaldehyde and fluoroacetate resistance, two dominant markers in Saccharomyces cerevisiae

We employed two genes in constructing yeast expression cassettes for dominant, selectable markers. The Saccharomyces cerevisiae gene SFA1, encoding formaldehyde dehydrogenase, was placed under the control of the GPD1 promoter and CYC1 terminator. The Moraxella sp. strain B gene dehH1, encoding fluoroacetate dehalogenase, was placed under the control of both the GPD1 and CYC1 promoters. With these constructs it was possible to select directly for yeast strains resistant to either formaldehyde or fluoroacetate. Both selective agents are completely metabolized and inexpensive, making them very useful in the pursuit of yeast gene functions and for industrial applications. An additional advantage of the formaldehyde dehydrogenase marker is that it is an S. cerevisiae gene, thus allowing 'all yeast' constructs.

Cloning and expression of a cDNA copy of the viral K28 killer toxin gene in yeast

The killer toxin K28, secreted by certain killer strains of the yeast Saccharomyces cerevisiae is genetically encoded by a 1.9 kb double-stranded RNA, M-dsRNA (M28), that is present within the cell as a cytoplasmically inherited virus-like particle (VLP). For stable maintenance and replication, M28-VLPs depend on a second dsRNA virus (LA), which has been shown to encode the major capsid protein (cap) and a capsid-polymerase fusion protein (cap-pol) that provides the toxin-coding M-satellites with their transcription and replicase functions. K28 toxin-coding M28-VLPs were isolated, purified and used in vitro for the synthesis of the single-stranded M28 transcript, which was shown to be of plus strand polarity and to bind to oligo(dT)-cellulose, indicating that M28(+)ssRNA contains an internal A-rich tract. Strand separation of the 1.9 kb M28-dsRNA and direct RNA sequencing of its 3' ends was performed in order to obtain specific DNA oligonucleotides that could be used as primers for cDNA synthesis. The nucleotide sequence of the toxin-coding M28-cDNA identified a single open reading frame (ORF) coding for a polypeptide of 345 amino acids, which contained two potential Kex2p/Kex1p processing sites and three potential sites for protein N-glycosylation. The toxin-coding cDNA was cloned and expressed in sensitive non-killer strains under the control of the yeast PGK promoter. Upon transformation, this construct conferred the complete K28 phenotype, demonstrating that both toxin and immunity determinants are contained within the cloned cDNA. In vitro translational analysis of the M28(+)ssRNA in vitro transcript identified the primary gene product of M28 as a K28 preprotoxin of 38 kDa (M-p38).

Industrial production of heterologous proteins by fed-batch cultures of the yeast Saccharomyces cerevisiae

This review concerns the issues involved in the industrial development of fed-batch culture processes with Saccharomyces cerevisiae strains producing heterologous proteins. Most of process development considerations with fed-batch recombinant cultures are linked to the reliability and reproducibility of the process for manufacturing environments where quality assurance and quality control aspects are paramount. In this respect, the quality, safety and efficacy of complex biologically active molecules produced by recombinant techniques are strongly influenced by the genetic background of the host strain, genetic stability of the transformed strain and production process factors. An overview of the recent literature of these culture-related factors is coupled with our experience in yeast fed-batch process development for producing various therapeutic grade proteins. The discussion is based around three principal topics: genetics, microbial physiology and fed-batch process design. It includes the fundamental aspects of yeast strain physiology, the nature of the recombinant product, quality control aspects of the biological product, features of yeast expression vectors, expression and localization of recombinant products in transformed cells and fed-batch process considerations for the industrial production of Saccharomyces cerevisiae recombinant proteins. It is our purpose that this review will provide a comprehensive understanding of the fed-batch recombinant production processes and challenges commonly encountered during process development.

New developments in fungal virology

Although viruses are widely distributed in fungi, their biological significance to their hosts is still poorly understood. A large number of fungal viruses are associated with latent infections of their hosts. With the exception of the killer-immune character in the yeasts, smuts, and hypovirulence in the chestnut blight fungus, fungal properties that can specifically be related to virus infection are not well defined. Mycoviruses are not known to have natural vectors; they are transmitted in nature intracellularly by hyphal anastomosis and heterokaryosis, and are disseminated via spores. Because fungi have a potential for plasmogamy and cytoplasmic exchange during extended periods of their life cycles and because they produce many types of propagules (sexual and asexual spores), often in great profusion, mycoviruses have them accessible to highly efficient means for transmission and spread. It is no surprise, therefore, that fungal viruses are not known to have an extracellular phase to their life cycles. Although extracellular transmission of a few fungal viruses have been demonstrated, using fungal protoplasts, the lack of conventional methods for experimental transmission of these viruses have been, and remains, an obstacle to understanding their biology. The recent application of molecular biological approaches to the study of mycoviral dsRNAs and the improvements in DNA-mediated fungal transformation systems, have allowed a clearer understanding of the molecular biology of mycoviruses to emerge. Considerable progress has been made in elucidating the genome organization and expression strategies of the yeast L-A virus and the unencapsidated RNA virus associated with hypovirulence in the chestnut blight fungus. These recent advances in the biochemical and molecular characterization of the genomes of fungal viruses and associated satellite dsRNAs, as they relate to the biological properties of these viruses and to their interactions with their hosts are the focus of this chapter.

Structure and heterologous expression of the Ustilago maydis viral toxin KP4

Killer toxins are polypeptides secreted by some fungal species that kill sensitive cells of the same or related species. In the best-characterized cases, they function by creating new pores in the cell membrane and disrupting ion fluxes. Immunity or resistance to the toxins is conferred by the preprotoxins (or products thereof) or by nuclear resistance genes. In several cases, the toxins are encoded by one or more genomic segments of resident double-stranded RNA viruses. The known toxins are composed of one to three polypeptides, usually present as multimers. We have further characterized the KP4 killer toxin from the maize smut fungus Ustilago maydis. This toxin is also encoded by a single viral double-stranded RNA but differs from other known killer toxins in several respects: it has no N-linked glycosylation either in the precursor or in the mature polypeptide, it is the first killer toxin demonstrated to be a single polypeptide, and it is not processed by any of the known secretory proteinases (other than the signal peptidase). It is efficiently expressed in a heterologous fungal system.

Precise mapping and molecular characterization of the MFT1 gene involved in import of a fusion protein into mitochondria in Saccharomyces cerevisiae

Garrett et al. [Mol. Gen. Genet. 225 (1991) 483-491] recently reported that an Atp2-lacZ fusion protein was transported into mitochondria in yeast, thus identifying the MFT1 (mitochondrial fusion targeting) gene as a genomic fragment which complements a mutation (mft1) that failed in targeting a fusion protein into mitochondria. They mapped this gene to the ORF, which we have independently identified as a gene homologous to the cyc07 gene, which is expressed specifically in the S phase during the plant cell cycle. We have mapped the MFT1 gene precisely and found that this gene should correspond to the neighboring ORF, rather than the ORF they identified.

Role of the gamma component of preprotoxin in expression of the yeast K1 killer phenotype

K1 killer strains of Saccharomyces cerevisiae secrete a polypeptide toxin to which they are themselves immune. The alpha and beta components of toxin comprise residues 45-147 and 234-316 of the 316-residue K1 preprotoxin. The intervening 86-residue segment is called gamma. A 26-residue signal peptide is removed on entry into the endoplasmic reticulum. The Kex2 protease excises the toxin components from the 290-residue glycosylated protoxin in a late Golgi compartment. Expression of a cDNA copy of the preprotoxin gene confers the complete K1 killer phenotype on sensitive cells. We now show that expression of immunity requires the alpha component and the N-terminal 31 residues of gamma. An additional C-terminal extension, either eight residues of gamma or three of four unrelated peptides, is also required. Expression of preprotoxin terminating at the alpha C-terminus, or lacking the gamma N-terminal half of gamma causes profound but reversible growth inhibition. Inhibition is suppressed in cis by the same 31 residues of gamma required for immunity to exocellular toxin in trans, but not by the presence of beta. Both immunity and growth inhibition are alleviated by insertions in alpha that inactivate toxin. Inhibition is not suppressed by kex2, chc1 or kre1 mutations, by growth at higher pH or temperature, or by normal K1 immunity. Inhibition, therefore, probably does not involve processing of the alpha toxin component at its N-terminus or release from the cell and binding to glucan receptors. Some insertion and substitution mutations in gamma severely reduce toxin secretion without affecting immunity. They are presumed to affect protoxin folding in the endoplasmic reticulum and translocation to the Golgi.

Immunity and resistance to the KP6 toxin of Ustilago maydis

The KP6 toxin of Ustilago maydis, encoded by segmented double-stranded (ds) RNA viruses, is lethal to sensitive strains of the same species and related species. The toxin consists of two polypeptides, alpha and beta, synthesized as a single preprotoxin, which are not covalently linked. Neither polypeptide alone is toxic, but killer activity can be restored by in vitro and in vivo complementation. Killer-secreting strains are resistant to the toxin they produce. Resistance is conferred by a single recessive nuclear gene. This study describes a search for cytoplasmic factors that may confer resistance, also referred to as immunity. The approaches used to detect cytoplasmic immunity included transmission of dsRNA and transmission of virus particles to sensitive cells by cytoduction, cytoplasmic mixing in diploids and infection with viruses. An alternative approach was also used to express cloned cDNAs of the KP6 toxin-encoding dsRNA and of the alpha and beta polypeptides. The results indicated that no immunity to KP6 can be detected. While KP6, alpha and beta polypeptides were expressed by resistant cells, neither KP6 nor beta were expressed in sensitive strains. The alpha polypeptide was expressed in sensitive cells, but it did not confer immunity. These results suggest that neither the preprotoxin nor the alpha or beta polypeptides confer immunity and thus beta may be the toxic component of the binary toxin.

Genetic analysis of maintenance and expression of L and M double-stranded RNAs from yeast killer virus K28

The killer phenotype expressed by Saccharomyces cerevisiae strain 28 differs from that of the more extensively studied K1 and K2 killers with respect to immunity, mode of toxin action and cell wall primary toxin receptor. We previously demonstrated that the M28 and L28 dsRNAs found in strain 28 are present in virus-like particles (VLPs) and that transfection with these VLPs is sufficient to confer the complete K28 phenotype on a dsRNA-free recipient cell. We also demonstrated that L28, like the L-A-H species in K1 killers, has [HOK] activity required for maintenance of M1-dsRNA, and predicted that M28 would share with M1 dependence on L-A for replication. We now confirm this prediction by genetic and biochemical analysis of the effects of representative mak, ski and mkt mutations on M28 maintenance, demonstrating that M28 replication resembles M1 in all respects. We also show that L28 is an L-A-H species lacking [B] activity, and that M28 excludes both M1 and M2 from the same cytoplasm. Stable coexpression of K28 phenotype from M28 and of K1 phenotype from an M1-cDNA clone was demonstrated. Exclusion, therefore, acts at the level of dsRNA replication, presumably reflecting competition for the L-A-H encoded capsid and cap-pol fusion protein, rather than reflecting incompatibility of toxin or immunity expression. Finally, we show that expression of active K28 toxin, but not of K28 immunity, requires the Kex2 endoprotease.

Yeast killer virus transcription initiation in vitro

Killer virions isolated from infected Saccharomyces cerevisiae cells contain an RNA polymerase activity which catalyzes the transcription in vitro of positive polarity RNAs from the L-A and M double-stranded RNA genomic segments of the virus. The RNA polymerase can initiate transcription in vitro with gamma-thio-GTP, whose thiophosphate group is found on the 5' terminus of transcripts. Transcripts produced in vitro by the virion-associated RNA polymerase in the presence of 7mGpppG are significantly more active as translational templates than are transcripts produced in its absence. However, unlike Escherichia coli RNA polymerase transcripts from viral cDNA made in the presence of 7mGpppG, transcripts produced by viral RNA polymerase in the presence of 7mGpppG fail to bind to antibody against 7mG.

Kex2-dependent processing of yeast K1 killer preprotoxin includes cleavage at ProArg-44

The K1 killer toxin of Saccharomyces cerevisiae consists of 103- and 83-residue alpha and beta components whose derivation, from a 316-residue precursor preprotoxin, requires processing at the alpha N-terminus (after ProArg-44), the alpha C-terminus (after ArgArg-149) and at the beta N-terminus (after LysArg-233). These processing events occur after translocation to the Golgi and have been investigated using beta-lactamase fusions. Signal peptidase cleavage of the precursor, predicted to occur after Ala-26, was confirmed by N-terminal sequence analysis of Ala-34 and Ile-52 fusions. Cleavage at all of the other predicted processing sites, including ProArg-44, is dependent on activity of the Kex2 protease. A fourth Kex2-dependent cleavage occurs at LysArg-188. Implications for the specificity of Kex2 cleavage and preprotoxin processing are discussed.

The K2-type killer toxin- and immunity-encoding region from Saccharomyces cerevisiae: structure and expression in yeast

The cDNA copies of M2-1, the larger heat-cleavage product of M2 double-stranded (ds) RNA, have been synthesized, cloned, sequenced and expressed in yeast. This sequence, in combination with the known terminal sequence of M2-1 dsRNA, identifies a translation reading frame for a 362-amino-acid protein of 38.7 kDa, similar in size to the one of several protein species produced from M2-1 dsRNA in vitro translation. The expression of this cDNA clone in yeast confers both killer and immunity phenotypes.

Yeast dsRNA viruses: replication and killer phenotypes

The cytoplasmic L-A dsRNA virus of Saccharomyces cerevisiae consists of a 4.5 kb dsRNA and the two gene products it encodes; the capsid (cap) and at least one copy of the capsid-polymerase (cap-pol) fusion protein. Virion cap-pol catalyses transcription of the plus (sense)-strand; this is extruded from the virus and serves as messenger for synthesis of cap and cap-pol. Nascent cap-pol binds to a specific domain in the plus strand to initiate encapsidation and then catalyses minus-strand synthesis to complete the replication cycle. Products of at least three host genes are required for replication, and virus copy number is kept at tolerable levels by the SKI antivirus system. S. cerevisiae killer viruses are satellite dsRNAs that use a similar encapsidation domain to parasitize the L-A replication machinery. They encode precursors of secreted polypeptide toxins and immunity (specific resistance) determinants and are self-selecting. Three unique killer types, K1, K2 and K28, are currently recognized. They are distinguished by an absence of cross-immunity and by toxin properties and lethal mechanisms; while K1 and K2 toxins bind to cell-wall glucan and disrupt membrane functions, K28 toxin binds to mannoprotein and causes inhibition of DNA synthesis.

Expression in yeast of a cDNA copy of the K2 killer toxin gene

A cDNA copy of the M2 dsRNA encoding the K2 killer toxin of Saccharomyces cerevisiae was expressed in yeast using the yeast ADH1 promoter. This construct produced K2-specific killing and immunity functions. Efficient K2-specific killing was dependent on the action of the KEX2 endopeptidase and the KEX1 carboxypeptidase, while K2-specific immunity was independent of these proteases. Comparison of the K2 toxin sequence with that of the K1 toxin sequence shows that although they share a common processing pathway and are both encoded by cytoplasmic dsRNAs of similar basic structure, the two toxins are very different at the primary sequence level. Site-specific mutagenesis of the cDNA gene establishes that one of the two potential KEX2 cleavage sites is critical for toxin action but not for immunity. Immunity was reduced by an insertion of two amino acids in the hydrophobic amino-terminal region which left toxin activity intact, indicating an independence of toxin action and immunity.

Protein transport and compartmentation in yeast

Many newly synthesized proteins must be translocated across one or more membranes to reach their destination in the individual organelles or membrane systems. Translocation, mostly requiring an energy source, a signal on the protein itself, loose conformation of the protein and the presence of cytosolic and/or membrane receptor-like proteins, is often accompanied by covalent modifications of transported proteins. In this review I discuss these aspects of protein transport via the classical secretory pathway and/or special translocation mechanisms in the unicellular eukaryotic organism Saccharomyces cerevisiae.

Genetic and molecular approaches to synthesis and action of the yeast killer toxin

The K1 killer toxin of Saccharomyces cerevisiae is a secreted, virally-coded protein lethal to sensitive yeasts. Killer yeasts are immune to the toxin they produce. This killer system has been extensively examined from genetic and molecular perspectives. Here we review the biology of killer yeasts, and examine the synthesis and action of the protein toxin and the immunity component. We summarise the structure of the toxin precursor gene and its protein products, outline the proteolytic processing of the toxin subunits from the precursor, and their passage through the yeast secretory pathway. We then discuss the mode of action of the toxin, its lectin-like interaction with a cell wall glucan, and its probable role in forming channels in the yeast plasma membrane. In addition we describe models of how a toxin precursor species functions as the immunity component, probably by interfering with channel formation. We conclude with a review of the functional domains of the toxin structural gene as determined by site-directed mutagenesis. This work has identified regions associated with glucan binding, toxin activity, and immunity.

Yeast KEX1 gene encodes a putative protease with a carboxypeptidase B-like function involved in killer toxin and alpha-factor precursor processing

The yeast KEX1 gene product has homology to yeast carboxypeptidase Y. A mutant replacing serine at the putative active site of the KEX1 protein abolished activity in vivo. A probable site of processing by the KEX1 product is the C-terminus of the alpha-subunit of killer toxin, where toxin is followed in the precursor by 2 basic residues. Processing involves endoproteolysis following these basic residues and trimming of their C-terminal by a carboxypeptidase. Consistent with the KEX1 product being this carboxypeptidase is its role in alpha-factor pheromone production. In wild-type yeast, KEX1 is not essential for alpha-factor production, as the final pheromone repeat needs no C-terminal processing. However, in a mutant in which alpha-factor production requires a carboxypeptidase, pheromone production is KEX1-dependent.

A family of yeast expression vectors containing the phage f1 intergenic region

The construction and characterization of a family of yeast expression vectors is described. They have the following features: plasmid replication and selection (ApR) in Escherichia coli, packaging of single-stranded (ss) DNA upon infection of E. coli with a filamentous helper phage, replication in Saccharomyces cerevisiae based on the 2 mu plasmid origin of replication (ori), selection in yeast by complementation of LEU2 (pVT-L series, size 6.3 kb) or URA3 gene (pVT-U series, size 6.9 kb) and seven unique restriction sites for cloning within an 'expression cassette' which includes the promoter and 3' sequence of the ADH1 gene. The multiple cloning site as well as the ori and intergenic region of the phage f1 have been cloned in two orientations for convenient gene cloning and ssDNA strand selection. As a result any of these eight vectors can be chosen for cloning, expressing genes in yeast, sequencing and mutagenesis without the need for recloning into specialized vectors.

The internal polyadenylate tract of yeast killer virus M1 double-stranded RNA is variable in length

The 1.8-kbp M1 double-stranded (ds) RNA from type 1 killer strains of Saccharomyces cerevisiae contains an internal 200-bp adenine- and uracil-rich region. We have previously demonstrated that this region consists primarily of adenine residues on the plus strand of M1 dsRNA and on the full-length, in vitro synthesized (+) transcript (denoted m) of M1 dsRNA, neither of which contains 3'-terminal polyadenylate. We now show that there is variability in the length of the polyadenylate tracts of m transcripts synthesized in vitro by virions purified from either of the K1 diploid killer strains A364A X S7 or A364A X 1384. This variability reflects size differences seen in the corresponding M1 dsRNA genomes which, along with other data presented, localizes the variability in the length of M1 dsRNA to the adenine- and uracil-rich region.

Yeast killer toxin: site-directed mutations implicate the precursor protein as the immunity component

Yeast killer toxin and a component giving immunity to it are both encoded by a gene specifying a single 35 kd precursor polypeptide. This precursor is composed of a leader peptide, the alpha and beta subunits of the secreted toxin, and a glycosylated gamma peptide separating the latter. The toxin subunits are proteolytically processed from the precursor during toxin secretion. Using site-directed mutagenesis, we have identified a region of the precursor gene necessary for expression of the immunity phenotype. This immunity-coding region extends through the C-terminal half of the alpha subunit into the N-terminal part of the gamma glycopeptide. Mutations in other parts of the gene allow full immunity but produce precursors that fail to be processed. The precursor can therefore confer immunity, and we propose that it does so in the wild type by competing with mature toxin for binding to a membrane receptor.

Yeast arginine permease: nucleotide sequence of the CAN1 gene

The yeast CAN1 gene, thought to encode arginine permease, has found use in genetics as a selectable locus. We have sequenced the cloned CAN1 gene, which contains an open reading frame of 1770 nucleotides, encoding a polypeptide of calculated molecular weight 65,766. Disruption of this open reading frame largely abolishes CAN1 gene expression, while subcloned fragments of the open reading frame hybridize strand-specifically to a 2.3 kb yeast RNA message. The encoded protein has no leader signal sequence, and is highly hydrophobic, with a possible twelve membrane-spanning domains, several of which have the high hydrophobic moments seen in channel-forming or permease proteins. This protein structure is consistent with the CAN1 product being the plasma membrane arginine permease.

Analysis and utilization of the preprotoxin gene encoded in the M1 double-stranded RNA of yeast

The yeast type 1 toxin is a secreted protein which, along with immunity to the toxin, is encoded by an encapsidated double-stranded RNA (dsRNA) as a precursor protein, the preprotoxin. This chapter reviews recent work that has led to sequencing of complementary DNA (cDNA) copies of the preprotoxin gene, expression of toxin and toxin-immunity from cDNA, and use of the cDNA to effect secretion from yeast of a bacterial cellulase.