- Email: [email protected]
Improving secretion of recombinant proteins from yeast and mammalian cells: Rational or empirical design?
432 forum
Improving secretion of recombinant proteins from yeast and mammalian cells: rational or empirical design? Workshop The efficient secretion of authentic recombinant proteins from eukaryotic cells requires more than simply targeting the proteins to the endoplasmic reticuhim (ELL) and directing their translocation into the lumenal compartment of the ElL. Once within the ElK lumen, a secretory protein must be correctly folded, may need to undergo one or more post-translational modifications and, in some cases, assemble into a functional oligomer, if it is ever to leave the ElL and continue through the secretory pathway (Fig. 1). There is an extensive cellular machinery responsible for ensuring that these processes occur efficiently and accurately with endogenous secretory proteins, but when a cell is genetically engineered to express high levels of a heterologous secretory protein, one or more of these processes may become rate-and/or yield-limiting. Can such problems be overcome by modulating the cellular levels of the soluble and membrane-bound proteins that normally mediate these steps? Several recent publications have provided a first indication that the secretory performance of both the yeast Saccharomyces cerevisiae and mammalian cells in producing recombinant proteins can indeed be improved by such rational manipulations. The major rate-limiting step in constitutive protein secretion in eukaryotic cells occurs at the point of exit from the ElK lumen into the Golgi, and it is at this point that 'quality control' is exerted 1. Thus, proteins that are either incorrectly folded, or modified, or have been assembled into non-native, highmolecular-weight aggregates, are prevented from leaving the ElL and are destroyed by the proteolytic system intrinsic to the E R (Re£ 2). While this form of'garbage disposal' is of obvious benefit to a cell, it can be problematic if one wishes to genetically engineer the cell to secrete high levels o f a heterologous protein. Such heterologous proteins may have a propensity to fold incorTIBTECHNOVEMBER1994 (VOL12)
rectly because the level of a necessary, soluble, folding or modification factor is too tow to cope with the elevated throughput of secretory proteins. Alternatively, the protein may simply be unable to fold correctly because one (or more) of the soluble factors required for the authentic post-translational modification of the protein is missing. This would particularly be the case if one were trying to secrete a mammalian protein from yeast, as the ElL proteins required for folding the mammalian protein may be absent, or at too-low levels, in the yeast ElK (lLefi 3).
Key events: disulphide-bond formation and quality control Efforts to date have focused on improving the processing of recombinant proteins within the yeast and mammalian ElL, with particular attention being paid to two welldefined soluble E R components: protein disulphide isomerase (PDI), an enzyme that catalyses the formation of native disulphide bonds in secretory proteins4; and BiP/GlLP78, an Hsp70 homologue, which appears to act, in part, as a molecular chaperone for secretory proteins passing through the ElL (Ref. 5). Disulphide-bond formation is a covalent modification and is probably the post-translational modification undergone most generally by proteins entering the secretory pathway. Native disulphide-bond formarion is an integral aspect of the protein-folding pathway, and plays a significant part in protein assembly, as many secretory proteins (for example, antibodies, procollagens) are oligorners of two or more polypeptide chains held together by interchain disulphide bonds. In mammalian secretory cells, PDI is relatively abundant whereas, in yeast, it represents <0.05% of the total cellular protein 6, probably reflecting the low endogenous level of disulphide-bonded proteins secreted by this simple eukaryote.
Overexpressing secretory pathway proteins in yeast That S. cerevisiae can efficiently secrete multiply disulphide-bonded mammalian proteins is not in doubt; for example, impressive yields of authentic, correctly folded human serum albumin (HSA) can be secreted from S. cerevisiae7. Two recent papers have, however, demonstrated that overexpression of PDI in S. cerevisiae can improve the levels of secretion of a variety of multiply disulphide-bonded heterologous proteins that are otherwise secreted at low levels. In one report, secretion of the leech protein antistasin, a 119 amino acid polypeptide with the potential to form ten intrachain disulphide bridges, was increased some threefold if human PDI was co-expressed in the same cell8. Co-overexpression of yeast PDI resulted in an almost 25-fold increase (H. Markus et al., unpublished). In an analogous study, Robinson et al.9 showed that the overexpression of yeast PDI in yeast resulted in tenfold higher levels of secretion of human platelet-derived growth factor (PDGF) and a fourfold increase in secretion of the Schizosaccharomyces pombe acid phosphatase. Interestingly, two of the eight disulphide bridges within the secreted PDGF homodimer are interchain disulphide bonds. Robinson et al. 9 also reported that overexpression of PDI did not enhance the secretion of all the heterologous proteins tested; the secretion of human granulocytecolony-stimulating factor (GCSF) was largely unaffected. Nevertheless, these two studies provide evidence that yeast strains overexpressing PDI may be better hosts for secreting heterologous disulphide-bonded proteins.
Overexpressing secretory pathway proteins in mammalian cells Can overexpression of resident ElL proteins such as PDI and BiP improve the levels of proteins secreted from mammalian cells? Using a structured kinetic model for monoclonal antibody (mAb) synthesis and secretion by hybridoma cells, Bibila and Flickinger 1°,11 predicted that the assembly of antibody molecules in the E R is the most likely rate-limiting step in antibody secretion in fast-growing cells. As antibody assembly requires the formation of disulphide bonds between heavy and light chains, they further predicted that engineering higher levels of PDI synthesis in hybridoma © 1994, Elsevier Science Ltd
433
f o t'lA lq'l cells would increase both the specific antibody-secretion rate and the final antibody yield u . Although no experimental studies on the effectiveness of PDI overexpression in mammalian cells have yet been reported, there have been studies on the consequences of altering the levels of B i P / G R P 7 8 on secretion from mammalian cells. Surprisingly, reducing the levels of B i P / G R P 7 8 in Chinese hamster ovary (CHO) cells improved the secretion of an underglycosylated form of tissue plasminogen activator (tPA) (Refi 12). By contrast, overexpression of B i P / G R P 7 8 in these same cells resulted in a block in the secretion of certain other heterologous proteins 13. These data suggest that B i P / G R P 7 8 may play an important quality-control function, with free B i P / G R P 7 8 binding to misfolded or aggregated polypeptides in the ER, thereby preventing their further passage through the secretory pathway. This idea has been supported by recent studies on the 'unfolded protein response' pathway in yeast TM and mammalian cells is. In this response, the accumulation of unfolded proteins in the E R triggers the expression of genes encoding the major soluble E R proteins including B i P / G R P 7 8 and PDI. Hence, reducing the levels of free BiP/GP,,P78 would result in a higher proportion of secretion-incompetent polypeptides escaping the E R and being secreted. T h e problem then faced is that, although the levels of secreted protein are increased, they may consist of a large proportion of incorrectly folded or modified polypeptides, which is not what the biotechnology industry wants to produce. Although the consequences of overexpressing the yeast BiP/GR.P78 gene (KAR2) on heterologous protein secretion have not been reported, overexpression of the KAR2 gene in yeast cells accumulating unfolded proteins in their E R results in a reduction in the levels of unfolded protein 16. Little attempt has been made to optimize other post-translational modification processes in yeast or mammalian cells by this rational genetic manipulation. This is due, in part, to a lack of fundamental knowledge of the identity and functional significance of the factors that mediate these processes. A particular future target must be glycosylation, as many studies have shown that abnormal or inefficient glycosylation
Endoplasmic reticulum
Golgi
Secretory vesicles
Figure 1 Secretory proteins are synthesized on ribosomes that are tightly bound to the rough endoplasmic reticulum (ER) and then translocated into the lumen of the rough ER, where the signal sequence is cleaved. Disulphidebond formation, and the addition of high-mannose oligosaccharides to specific Asp residues, also occurs within the ER. 'Quality control' of incorrectly folded or aggregated proteins is carried out by proteolytic degradation within the ER: this, therefore, is the point at which major loss of heterologous protein can occur. Proteins that pass the quality check move from the rough ER to the Golgi complex via membrane vesicles. Further proteolytic cleavage or post-translational modifications, some of which play a role in targeting the protein to its final destination, occur in the Golgi. Some secretory proteins are directed to the secretory vesicles that fuse constitutively with the plasma membrane, thus exocytosing their contents.
Of proteins leads to their misfolding and concomitant retention in the E R (Refs 1, 17). It is possible that glycosylation per se may be part of the chaperoning process that is essential for correct protein folding in the E R (Ref.. 18). Selection versus rational design o f hosts
While this approach to improving the efficiency of secretion from yeast and mammalian cells by rational genetic design of the host cell is beginning to bear fruit, there is still a place for the empirical approach. Foremost here is the isolation of mutant strains or cell lines that show elevated levels of secretion of a particular recombinant protein. This approach has already been successfully applied to the genetically tractable & cerevisiae3, by the isolation, after random chemical mutagenesis, of'super-secretors'. Surprisingly, the limited analysis of such mutants carried out to date has not uncovered mutations in genes encoding components of the secretory pathway. For example, one such mutant carried a mutation in the PMRI gene encoding a Ca>-transporting ATPase located in the Golgi 19.
Such an empirical approach might be fruitful when looking for proteins which, when over- or underexpressed, increase the rate and yield of secretory proteins. This is borne out by the recent remarkable finding that overexpression of the polyubiquitin gene UBI4 in yeast leads to a sevenfold increase in the levels of the human leucocyte elastase inhibitor, elafin, secreted from the strain employed 2°. As ubiquitination of proteins is a key initiating step during intracellular proteolysis, one might have predicted that underexpression, rather than overexpression, 0fubiquitin would improve secretion levels. This finding may indicate a secondary chaperoning function for ubiquitin in the yeast secretory pathway 2°. That the achievable levels of secretion for any one particular recombinant protein can be improved by a combination of rational and empirical genetic manipulation of the host cell's secretory pathway is clear. Yet, there remains much to be learnt about the major ratelimiting and yield-hmiting steps in the secretory pathways of the eukaryotic hosts favoured by the biotechnology industry for the TIBTECHNOVEMBER1994 (VOL12)
434
f o rbl m
secretion of recombinant proteins. Only then will generic 'super-secreting' strains become available for expressing a wide range o f recombinant secretory proteins.
References 1 Hurtley, S. M. and Helenius, A. (1989) Annu. Rev. Cell. Biol. 5, 277-307 2 Klausner, P,. D. and Sitia, IZ. (1990) Cell 62, 611-614 3 l
Letter
846--851 7 Quirk, A. V. et al. (1989) Biotechnol. Appl. Biochem. 11,273-287 8 Schultz, L. D. et al. (1994) Ann. NYAcad. Sci. 721,148-157 9 R.obinson, A. S., Hines, V. and Wittrup, K. D. (1994) Bio/Technology 12, 381-385 10 Bibila, T. A. and Flickinger, M. C. (1992) Biotechnol. Bioeng. 39, 251-261 11 Bibila, T. A. and Flicldnger, M. C. (1992) Biotechnol. Bioeng. 39, 262-272 12 Dorner, A., Krane, M. and Kaufi'nan, R. (1988) Mol. Cell. Biol. 8, 4063-4070 13 Dorner, A., Wastey, L. and Kaufman, P,.. (1992) EMBOJ. 11, 1563-1571 14 Shamu, C. E., Cox, J. S. and Walter, P. (1994) Trends Cell Biol. 4, 56-60 15 Dorner, A. J., Wasley, L C., Raney, P., Haugejorden, S., Green, M. and Kaufman,
The dogs that did not bark Sometimes it is what an article does not say that may mislead readers as to the status of a scientific field. This is perhaps inevitable as authors of minireviews frequently have to restrict their selection o f papers discussed to remain within the journal's specified limits. In a recent issue o f Trends in Biotechnology, Montague and Morris 1 discuss a variety of applications o f artificial neural networks (ANNs) in biotechnology, including a brief study from their own institution of the use of A N N s in the analysis of pyrolysis mass spectral (PyMS) data for microbial identification. W e think that TIBTECH's readers might also be interested to know that we have demonstrated that much greater power in the use o f ANNs in analysing PyMS data may be obtained from exploiting them in effecting the quantitative (bio)chemical analysis of microbial (and indeed other) systems. Thus, we and our collaborators have shown that the combination o f A N N s and PyMS may be used to quantify indole product~pn in bacteria 2, biopolymers in binary 3,4 and tertiary mixtures s, the production of recombinant proteins in whole cells o f Escherichia coli 6, in mixed microbial cultures s, and in the rapid and quantitative screening of cultures and fermenter broths for the overproduction of metabolites 7. Inter alia, we have also been the first to apply PyMS and A N N s to the successful identification of the adulteration o f extra-virgin olive oils 8,9, and in demonstrating that canine isolates
TIBTECHNOVEMBER1994 (VOL12)
of Propionibacterium aches strains are the same as human wild-type strains 1°,11. Recent reviews of the general approach are available 12,13. In another article in the same issue o f Trends in Biotechnology, Konstantinov and c o - w o r k e r s TM survey a number of approaches that have been taken for monitoring the biomass concentration of animal-cell cultures in real time, including the direct measurement of the electrical capacitance o f animal-cell suspensions. They state, on the basis of an earlier study by the first author using a two-terminal device made available to him by the Kobe Steel Company (Kobe, Japan), that: 'practical applications of the method have not been very successful, chiefly due to the strong influence of the "resistance" component, which masks the capacitance', and that: 'as a result of the relatively low cell concentrations and high medium conductivity, it is unlikely that the capacitance method will find serious application in monitoting mammalian cell cultures'. The influence o f conductivity on dielectric measurements o f this type is well known (and referred to as 'electrode polarization'), and is minimized by using a four-terminal configuration 15,16, as is in fact exploited in the other device that Konstantinov et al. cite in this context, the Abet Instruments (Aberystwyth, UK) Biomass Monitor. This (patented) approach and instrumend v,18 has been widely and successfully applied in media and in systems of substan-
R. J. (1990) J. Biol. Chem. 265, 22029-22034 16 Kohno, K., Normington, K., Sambrook,J., Gething, M-J. and Moil, K. (1993) Mol. Cell. Biol. 13, 877-890 17 Gallagher, P. J., Henneberry, J. M., Sambrook, J. F. and Gething, M-J. H. (1992)J. Virol. 66, 7136-7145 18 Helenius, A. (1994) MoL Biol. Cell 5, 253-265 19 Rudolph, H. K. et al. (1989) Cell 58, 133-145 20 Chen, Y., Pioti, D. and Piper, P. W. (1994) Bio/Technology 12, 819-823
Mick F. Tuite Robert B. Freedman Research School of Bioscienees, University of Kent, Canterbury, Kent, UK CT2 7NJ.
tial conductivity at laboratory and industrial scales, including bacterial and yeast cultures 1v-22, bacterial biofilms 23, immobilized ceils 24 and filamentous organisms in liquid and solid-substrate fermentations 2s-28. In particular, as existing instruments can accept conductivities up to 2 4 m S c m -1, it has been very successfully exploited in both human blood 29, and in animal-cell cultures 30,31 in isotonic media. The Biomass Monitor has a particularly high sensitivity for animal ceils, as the dielectric increment (capacitance) per unit biomass scales linearly with the cell radius 32-34and, interestingly, in the present context, we have also shown that multifrequency dielectric data can be analysed effectively using A N N s (Re£ 35). These and other facts would suggest to us not only that it is very likely indeed 'that the capacitance method will find serious application in monitoring mammalian cell cultures', but that it has, in fact, already done so.
References 1 Montague, G. and Morris, A. J. (1994) Trends Biotechnol. 12, 312-324 2 Goodacre, R. and Kell, D. B. (1993) Anal. Chim. Acta 279, 17-26 3 Goodacre, R., Edmonds, A. N. and Ke/l, D. B. (1993)J. Anal. Appl. Pyrol. 26, 93-114 4 Neal, M.J., Goodacre, R. and Kell, D. B. (1994) Proc. World Congress Neural Networks, San Diego, pp. 1-318-1-323 5 Goodacre, R., Neal, MI J. and Kell, D. B. (1994) Anal. Chem. 66, 1070-1085 6 Goodacre, Ik., Karim, A., Kaderbhai, M. and Kell, D. B. (1994) J. Bioteehnol. 34, 185-193 7 Goodacre, R., Trew, S., Wrigley-Jones, C., Neal, M. J., Porter, N. and Kell, D. B. Biotechnol. Bioeng. (in press)
Improving secretion of recombinant proteins from yeast and mammalian cells: rational or empirical design? Workshop The efficient secretion of authentic recombinant proteins from eukaryotic cells requires more than simply targeting the proteins to the endoplasmic reticuhim (ELL) and directing their translocation into the lumenal compartment of the ElL. Once within the ElK lumen, a secretory protein must be correctly folded, may need to undergo one or more post-translational modifications and, in some cases, assemble into a functional oligomer, if it is ever to leave the ElL and continue through the secretory pathway (Fig. 1). There is an extensive cellular machinery responsible for ensuring that these processes occur efficiently and accurately with endogenous secretory proteins, but when a cell is genetically engineered to express high levels of a heterologous secretory protein, one or more of these processes may become rate-and/or yield-limiting. Can such problems be overcome by modulating the cellular levels of the soluble and membrane-bound proteins that normally mediate these steps? Several recent publications have provided a first indication that the secretory performance of both the yeast Saccharomyces cerevisiae and mammalian cells in producing recombinant proteins can indeed be improved by such rational manipulations. The major rate-limiting step in constitutive protein secretion in eukaryotic cells occurs at the point of exit from the ElK lumen into the Golgi, and it is at this point that 'quality control' is exerted 1. Thus, proteins that are either incorrectly folded, or modified, or have been assembled into non-native, highmolecular-weight aggregates, are prevented from leaving the ElL and are destroyed by the proteolytic system intrinsic to the E R (Re£ 2). While this form of'garbage disposal' is of obvious benefit to a cell, it can be problematic if one wishes to genetically engineer the cell to secrete high levels o f a heterologous protein. Such heterologous proteins may have a propensity to fold incorTIBTECHNOVEMBER1994 (VOL12)
rectly because the level of a necessary, soluble, folding or modification factor is too tow to cope with the elevated throughput of secretory proteins. Alternatively, the protein may simply be unable to fold correctly because one (or more) of the soluble factors required for the authentic post-translational modification of the protein is missing. This would particularly be the case if one were trying to secrete a mammalian protein from yeast, as the ElL proteins required for folding the mammalian protein may be absent, or at too-low levels, in the yeast ElK (lLefi 3).
Key events: disulphide-bond formation and quality control Efforts to date have focused on improving the processing of recombinant proteins within the yeast and mammalian ElL, with particular attention being paid to two welldefined soluble E R components: protein disulphide isomerase (PDI), an enzyme that catalyses the formation of native disulphide bonds in secretory proteins4; and BiP/GlLP78, an Hsp70 homologue, which appears to act, in part, as a molecular chaperone for secretory proteins passing through the ElL (Ref. 5). Disulphide-bond formation is a covalent modification and is probably the post-translational modification undergone most generally by proteins entering the secretory pathway. Native disulphide-bond formarion is an integral aspect of the protein-folding pathway, and plays a significant part in protein assembly, as many secretory proteins (for example, antibodies, procollagens) are oligorners of two or more polypeptide chains held together by interchain disulphide bonds. In mammalian secretory cells, PDI is relatively abundant whereas, in yeast, it represents <0.05% of the total cellular protein 6, probably reflecting the low endogenous level of disulphide-bonded proteins secreted by this simple eukaryote.
Overexpressing secretory pathway proteins in yeast That S. cerevisiae can efficiently secrete multiply disulphide-bonded mammalian proteins is not in doubt; for example, impressive yields of authentic, correctly folded human serum albumin (HSA) can be secreted from S. cerevisiae7. Two recent papers have, however, demonstrated that overexpression of PDI in S. cerevisiae can improve the levels of secretion of a variety of multiply disulphide-bonded heterologous proteins that are otherwise secreted at low levels. In one report, secretion of the leech protein antistasin, a 119 amino acid polypeptide with the potential to form ten intrachain disulphide bridges, was increased some threefold if human PDI was co-expressed in the same cell8. Co-overexpression of yeast PDI resulted in an almost 25-fold increase (H. Markus et al., unpublished). In an analogous study, Robinson et al.9 showed that the overexpression of yeast PDI in yeast resulted in tenfold higher levels of secretion of human platelet-derived growth factor (PDGF) and a fourfold increase in secretion of the Schizosaccharomyces pombe acid phosphatase. Interestingly, two of the eight disulphide bridges within the secreted PDGF homodimer are interchain disulphide bonds. Robinson et al. 9 also reported that overexpression of PDI did not enhance the secretion of all the heterologous proteins tested; the secretion of human granulocytecolony-stimulating factor (GCSF) was largely unaffected. Nevertheless, these two studies provide evidence that yeast strains overexpressing PDI may be better hosts for secreting heterologous disulphide-bonded proteins.
Overexpressing secretory pathway proteins in mammalian cells Can overexpression of resident ElL proteins such as PDI and BiP improve the levels of proteins secreted from mammalian cells? Using a structured kinetic model for monoclonal antibody (mAb) synthesis and secretion by hybridoma cells, Bibila and Flickinger 1°,11 predicted that the assembly of antibody molecules in the E R is the most likely rate-limiting step in antibody secretion in fast-growing cells. As antibody assembly requires the formation of disulphide bonds between heavy and light chains, they further predicted that engineering higher levels of PDI synthesis in hybridoma © 1994, Elsevier Science Ltd
433
f o t'lA lq'l cells would increase both the specific antibody-secretion rate and the final antibody yield u . Although no experimental studies on the effectiveness of PDI overexpression in mammalian cells have yet been reported, there have been studies on the consequences of altering the levels of B i P / G R P 7 8 on secretion from mammalian cells. Surprisingly, reducing the levels of B i P / G R P 7 8 in Chinese hamster ovary (CHO) cells improved the secretion of an underglycosylated form of tissue plasminogen activator (tPA) (Refi 12). By contrast, overexpression of B i P / G R P 7 8 in these same cells resulted in a block in the secretion of certain other heterologous proteins 13. These data suggest that B i P / G R P 7 8 may play an important quality-control function, with free B i P / G R P 7 8 binding to misfolded or aggregated polypeptides in the ER, thereby preventing their further passage through the secretory pathway. This idea has been supported by recent studies on the 'unfolded protein response' pathway in yeast TM and mammalian cells is. In this response, the accumulation of unfolded proteins in the E R triggers the expression of genes encoding the major soluble E R proteins including B i P / G R P 7 8 and PDI. Hence, reducing the levels of free BiP/GP,,P78 would result in a higher proportion of secretion-incompetent polypeptides escaping the E R and being secreted. T h e problem then faced is that, although the levels of secreted protein are increased, they may consist of a large proportion of incorrectly folded or modified polypeptides, which is not what the biotechnology industry wants to produce. Although the consequences of overexpressing the yeast BiP/GR.P78 gene (KAR2) on heterologous protein secretion have not been reported, overexpression of the KAR2 gene in yeast cells accumulating unfolded proteins in their E R results in a reduction in the levels of unfolded protein 16. Little attempt has been made to optimize other post-translational modification processes in yeast or mammalian cells by this rational genetic manipulation. This is due, in part, to a lack of fundamental knowledge of the identity and functional significance of the factors that mediate these processes. A particular future target must be glycosylation, as many studies have shown that abnormal or inefficient glycosylation
Endoplasmic reticulum
Golgi
Secretory vesicles
Figure 1 Secretory proteins are synthesized on ribosomes that are tightly bound to the rough endoplasmic reticulum (ER) and then translocated into the lumen of the rough ER, where the signal sequence is cleaved. Disulphidebond formation, and the addition of high-mannose oligosaccharides to specific Asp residues, also occurs within the ER. 'Quality control' of incorrectly folded or aggregated proteins is carried out by proteolytic degradation within the ER: this, therefore, is the point at which major loss of heterologous protein can occur. Proteins that pass the quality check move from the rough ER to the Golgi complex via membrane vesicles. Further proteolytic cleavage or post-translational modifications, some of which play a role in targeting the protein to its final destination, occur in the Golgi. Some secretory proteins are directed to the secretory vesicles that fuse constitutively with the plasma membrane, thus exocytosing their contents.
Of proteins leads to their misfolding and concomitant retention in the E R (Refs 1, 17). It is possible that glycosylation per se may be part of the chaperoning process that is essential for correct protein folding in the E R (Ref.. 18). Selection versus rational design o f hosts
While this approach to improving the efficiency of secretion from yeast and mammalian cells by rational genetic design of the host cell is beginning to bear fruit, there is still a place for the empirical approach. Foremost here is the isolation of mutant strains or cell lines that show elevated levels of secretion of a particular recombinant protein. This approach has already been successfully applied to the genetically tractable & cerevisiae3, by the isolation, after random chemical mutagenesis, of'super-secretors'. Surprisingly, the limited analysis of such mutants carried out to date has not uncovered mutations in genes encoding components of the secretory pathway. For example, one such mutant carried a mutation in the PMRI gene encoding a Ca>-transporting ATPase located in the Golgi 19.
Such an empirical approach might be fruitful when looking for proteins which, when over- or underexpressed, increase the rate and yield of secretory proteins. This is borne out by the recent remarkable finding that overexpression of the polyubiquitin gene UBI4 in yeast leads to a sevenfold increase in the levels of the human leucocyte elastase inhibitor, elafin, secreted from the strain employed 2°. As ubiquitination of proteins is a key initiating step during intracellular proteolysis, one might have predicted that underexpression, rather than overexpression, 0fubiquitin would improve secretion levels. This finding may indicate a secondary chaperoning function for ubiquitin in the yeast secretory pathway 2°. That the achievable levels of secretion for any one particular recombinant protein can be improved by a combination of rational and empirical genetic manipulation of the host cell's secretory pathway is clear. Yet, there remains much to be learnt about the major ratelimiting and yield-hmiting steps in the secretory pathways of the eukaryotic hosts favoured by the biotechnology industry for the TIBTECHNOVEMBER1994 (VOL12)
434
f o rbl m
secretion of recombinant proteins. Only then will generic 'super-secreting' strains become available for expressing a wide range o f recombinant secretory proteins.
References 1 Hurtley, S. M. and Helenius, A. (1989) Annu. Rev. Cell. Biol. 5, 277-307 2 Klausner, P,. D. and Sitia, IZ. (1990) Cell 62, 611-614 3 l
Letter
846--851 7 Quirk, A. V. et al. (1989) Biotechnol. Appl. Biochem. 11,273-287 8 Schultz, L. D. et al. (1994) Ann. NYAcad. Sci. 721,148-157 9 R.obinson, A. S., Hines, V. and Wittrup, K. D. (1994) Bio/Technology 12, 381-385 10 Bibila, T. A. and Flickinger, M. C. (1992) Biotechnol. Bioeng. 39, 251-261 11 Bibila, T. A. and Flicldnger, M. C. (1992) Biotechnol. Bioeng. 39, 262-272 12 Dorner, A., Krane, M. and Kaufi'nan, R. (1988) Mol. Cell. Biol. 8, 4063-4070 13 Dorner, A., Wastey, L. and Kaufman, P,.. (1992) EMBOJ. 11, 1563-1571 14 Shamu, C. E., Cox, J. S. and Walter, P. (1994) Trends Cell Biol. 4, 56-60 15 Dorner, A. J., Wasley, L C., Raney, P., Haugejorden, S., Green, M. and Kaufman,
The dogs that did not bark Sometimes it is what an article does not say that may mislead readers as to the status of a scientific field. This is perhaps inevitable as authors of minireviews frequently have to restrict their selection o f papers discussed to remain within the journal's specified limits. In a recent issue o f Trends in Biotechnology, Montague and Morris 1 discuss a variety of applications o f artificial neural networks (ANNs) in biotechnology, including a brief study from their own institution of the use of A N N s in the analysis of pyrolysis mass spectral (PyMS) data for microbial identification. W e think that TIBTECH's readers might also be interested to know that we have demonstrated that much greater power in the use o f ANNs in analysing PyMS data may be obtained from exploiting them in effecting the quantitative (bio)chemical analysis of microbial (and indeed other) systems. Thus, we and our collaborators have shown that the combination o f A N N s and PyMS may be used to quantify indole product~pn in bacteria 2, biopolymers in binary 3,4 and tertiary mixtures s, the production of recombinant proteins in whole cells o f Escherichia coli 6, in mixed microbial cultures s, and in the rapid and quantitative screening of cultures and fermenter broths for the overproduction of metabolites 7. Inter alia, we have also been the first to apply PyMS and A N N s to the successful identification of the adulteration o f extra-virgin olive oils 8,9, and in demonstrating that canine isolates
TIBTECHNOVEMBER1994 (VOL12)
of Propionibacterium aches strains are the same as human wild-type strains 1°,11. Recent reviews of the general approach are available 12,13. In another article in the same issue o f Trends in Biotechnology, Konstantinov and c o - w o r k e r s TM survey a number of approaches that have been taken for monitoring the biomass concentration of animal-cell cultures in real time, including the direct measurement of the electrical capacitance o f animal-cell suspensions. They state, on the basis of an earlier study by the first author using a two-terminal device made available to him by the Kobe Steel Company (Kobe, Japan), that: 'practical applications of the method have not been very successful, chiefly due to the strong influence of the "resistance" component, which masks the capacitance', and that: 'as a result of the relatively low cell concentrations and high medium conductivity, it is unlikely that the capacitance method will find serious application in monitoting mammalian cell cultures'. The influence o f conductivity on dielectric measurements o f this type is well known (and referred to as 'electrode polarization'), and is minimized by using a four-terminal configuration 15,16, as is in fact exploited in the other device that Konstantinov et al. cite in this context, the Abet Instruments (Aberystwyth, UK) Biomass Monitor. This (patented) approach and instrumend v,18 has been widely and successfully applied in media and in systems of substan-
R. J. (1990) J. Biol. Chem. 265, 22029-22034 16 Kohno, K., Normington, K., Sambrook,J., Gething, M-J. and Moil, K. (1993) Mol. Cell. Biol. 13, 877-890 17 Gallagher, P. J., Henneberry, J. M., Sambrook, J. F. and Gething, M-J. H. (1992)J. Virol. 66, 7136-7145 18 Helenius, A. (1994) MoL Biol. Cell 5, 253-265 19 Rudolph, H. K. et al. (1989) Cell 58, 133-145 20 Chen, Y., Pioti, D. and Piper, P. W. (1994) Bio/Technology 12, 819-823
Mick F. Tuite Robert B. Freedman Research School of Bioscienees, University of Kent, Canterbury, Kent, UK CT2 7NJ.
tial conductivity at laboratory and industrial scales, including bacterial and yeast cultures 1v-22, bacterial biofilms 23, immobilized ceils 24 and filamentous organisms in liquid and solid-substrate fermentations 2s-28. In particular, as existing instruments can accept conductivities up to 2 4 m S c m -1, it has been very successfully exploited in both human blood 29, and in animal-cell cultures 30,31 in isotonic media. The Biomass Monitor has a particularly high sensitivity for animal ceils, as the dielectric increment (capacitance) per unit biomass scales linearly with the cell radius 32-34and, interestingly, in the present context, we have also shown that multifrequency dielectric data can be analysed effectively using A N N s (Re£ 35). These and other facts would suggest to us not only that it is very likely indeed 'that the capacitance method will find serious application in monitoring mammalian cell cultures', but that it has, in fact, already done so.
References 1 Montague, G. and Morris, A. J. (1994) Trends Biotechnol. 12, 312-324 2 Goodacre, R. and Kell, D. B. (1993) Anal. Chim. Acta 279, 17-26 3 Goodacre, R., Edmonds, A. N. and Ke/l, D. B. (1993)J. Anal. Appl. Pyrol. 26, 93-114 4 Neal, M.J., Goodacre, R. and Kell, D. B. (1994) Proc. World Congress Neural Networks, San Diego, pp. 1-318-1-323 5 Goodacre, R., Neal, MI J. and Kell, D. B. (1994) Anal. Chem. 66, 1070-1085 6 Goodacre, Ik., Karim, A., Kaderbhai, M. and Kell, D. B. (1994) J. Bioteehnol. 34, 185-193 7 Goodacre, R., Trew, S., Wrigley-Jones, C., Neal, M. J., Porter, N. and Kell, D. B. Biotechnol. Bioeng. (in press)