Inclusion body formation

When overexpressing recombinant proteins, inclusion bodies can be observed in different host systems, for example, prokaryotes, yeast or higher eukaryotes. It is interesting to note that even endogenous proteins, when overexpressed, can result in inclusion body formation (Gribskov et al., 1983). This is an indication that high expression rates are the primary reason for the aggregation of proteins, regardless of the expression system or the host used. On the other hand, proteins with a high hydrophobic

content are more predisposed to aggregation due to increased intermolecular interactions of regions of the folding polypeptide chain (Hartl et al., 2002).

The majority of the P. torridus proteins whose expression was attempted in E.

coli with T7 promoter based vectors were accumulated as inclusion bodies (Table 10).

Two general approaches were applied in attempting to overcome this problem – (i) purification of the obtained inclusion bodies and subsequent refolding or (ii) escaping their formation. The second approach included the use of different growth conditions in order to decrease the rate of protein synthesis, employment of different expression system (weak promoter) or the construction of fusion proteins with a highly hydrophilic and soluble partner protein. Additionally, coexpression of one or more of the E. coli heat shock proteins – GroEL, GroES, DnaK,DnaJ and GrpE was tested. These proteins, although constitutively expressed in E. coli, are synthesised at increased levels under stress conditions and are considered responsible for the proper folding of nascent polypeptide chains (Gething et al., 1992).

As an alternative to E. coli as the host for expression experiments, expression in S. solfataricus using the viral vector SSV1 and in the yeast S. cerevisiae for one P.

torridus protein was tested (sections C.2.4 and C.2.5).

The outcome of using all of the refolding techniques described in chapter C.2.1.

was a soluble but inactive enzyme. These soluble forms of the proteins were unstable and easily denatured at moderate temperatures (data not shown), indicating non-native protein structures. It has to be noted, however, that this approach is highly empirical and laborious. Therefore, it is probable that untested conditions exist which would lead to obtaining a particular P. torridus enzyme in its functional, native state after refolding.

The use of the commercially available kits also led to inactive proteins, at least for the two enzymes used for screening – the two ORFs annotated as β-galactosidase (ORFs 810 and 615).

When the ORF 810 coding sequence was fused in-frame with the E. coli genes nus or mal the resulting polypeptides were expressed in a soluble form (section C.2.3).

The same result, i.e. soluble protein, was obtained when the gluconate dehydratase ORF 1383 was expressed as a fusion with the E. coli mal gene (Fig. 15). It is obvious that the presence of the N- terminal fusion peptide sequences are responsible for the different processing of these proteins in E. coli, since the same ORF expressed without a tag resulted exclusively in inclusion body formation. This effect was dependent neither on the promoter used nor on the genetic background of the host. This can be deduced from

the facts that the Nus Tag system is actually based on a modified pET vector and that the same E. coli Rosetta strain was used for the expression from both p24-810 and pN-810. On the other hand, when the two fusion constructs of ORF 810 are compared, a different level of solubility of the encoded polypeptides could be observed. The use of the PTAC promoter in the expression vector pM-810 (maltose-binding protein fusion) resulted in a greater fraction of soluble protein compared to the fraction obtained by the T7-based pN-810 (NusA fusion, Figures 16 and 17). This difference can be attributed either to the promoter or to the fusion partner. As NusA has been reported to be one of the most soluble proteins in E. coli (Harrison et al., 2000) it is more probable that the use of the weaker PTAC promoter was of greater impact on soluble fusion protein production than the fusion partner. The fusion proteins, however, were also inactive, both before and after the cleavage of the fusion partner. Similar to the results from the refolding experiments, soluble forms were produced which were unstable and easily precipitated after heat treatment at 60°C (Fig. 14B).

The coexpression of the E. coli chaperones encoded by the vector pG-KJE8 (Nishihara et al., 1998) also did not result in an active form of the tested protein β-galactosidase encoded by ORF 810 (data not shown).

Because β-galactosidase activity was repeatedly detected in P. torridus cell extracts (not shown) it is interesting to compare the amenability to overexpression between the P. torridus ORF 810 product and its ortholog in S. solfataricus (LacS). The two polypeptides share 52% amino acid sequence similarity and produce reciprocal best BLAST hits. However, LacS from S. solfataricus has been expressed in an active form in E. coli 5α, using a pUC19-based vector (Haseltine et al., 1999). Also, when using the SSV1 based expression system, the control vector (containing lacS) did complement the Lac^- phenotype of S. solfataricus PH1-16, while the vector containing the P. torridus 810 gene did not, as inferred by measuring the β-galactosidase activity in the primary transformation mixture (data not shown). The inability of the ORF 810 construct to complement the Lac^- phenotype in S. solfataricus is most probably due to lack of expression. Of course the possibility of a wrong annotation of ORF 810 also exists. The other candidate gene that could account for the measured native β-galactosidase activity is ORF 615 whose translation product shares 22 % amino acid sequence similarity with both LacS and the ORF 810-encoded polypeptide. The results from the expression experiments with this ORF, however, were similar to the ones obtained with ORF 810.

Despite the overall amino acid sequence similarity between S. solfataricus LacS and P. torridus ORF 810 polypeptide, an important difference can be observed when the hydrophobicity plots of the two encoded proteins are compared. As it can be seen from Figure 36, they have a similar distribution pattern along the polypeptide chain, except for a highly hydrophobic N-terminal region in the P. torridus protein.

lacS

810

-4.0

+4.0 -4.0

+4.0

Fig. 36. Hydrophobicity analysis of the N-terminal 250 amino acids of S. solfataricus LacS and the predicted P. torridus ORF 810 β-galactosidase. The abscissa represents Kyte-Doolottle values, hydrophilic residues have a negative score. Each value is the average of the values of 5 adjacent residues and is plotted at the middle residue.

This hydrophobic N-terminus may be the decisive factor that contributes to the different processing of the proteins compared above both in E. coli and in S.

solfataricus. This is supported also by the successful expression of soluble N-terminal fusion constructs of ORF 810 polypeptide.

Having in mind the low intracellular pH of P. torridus, one reason for the failure to obtain properly folded polypeptides in these experiments could be the neutral cytoplasmic pH of the different expression hosts. Further possible reasons include the requirement of specific chaperones (at least when E. coli was used), a high hydrophobic content of the tested proteins, or a combination of the above.