The data obtained show that efficient uPA secretion demands an optimal level of its expression: both low and increased expression did not result in the maximal levels of uPA secretion. Indeed, we observed that the best uPA secreting strains were those containing the uPA expression cassette integrated in a single copy into the MOX or LEU2 genes. In contrast, integrants carrying a single copy of the uPA cassette integrated into a telomere region produced less uPA and were less efficient in uPA secretion. On the other hand, strains with multiple copies of the integrated cassette also secreted less uPA and accumulated it in an inactive form intracellularly. Comparison of strains with uPA expression exceeding an optimal level showed an inverted correlation between the efficiency of its expression and secretion (Table ). This can explain why no mutants with increased uPA expression levels were obtained so far among those producing more of its extracellular form.
We also found that upon overexpression uPA is accumulated in a form of high molecular weight aggregates. The absence of the Golgi modifications in accumulated uPA suggested that its aggregation occurred within the ER. Deletion analysis has shown that the N-terminal domains of uPA were responsible for the elevated levels of its aggregation. Alteration of the N-glycosylation site of uPA also resulted in its intracellular accumulation and aggregation. Thus, it is possible to suggest that both the presence of N-terminal domains and the absence of N-glycosylation caused intracellular accumulation of uPA, which interfered with its proper folding and resulted in subsequent aggregation in the ER. In agreement with this, the presence of the ER retention signal significantly increased aggregation degree of the best secreted uPA variant lacking the N-terminal domains.
Some of the ERAD substrates, e. g. mutant carboxy peptidase Y, can escape degradation and accumulate within the ER if they lack the N-linked oligosaccharide chains [15
]. Thus, it is possible that the misfolded uPA variants lacking the N-glycosylation site also escape degradation and accumulate in the ER, interfering with the folding of newly synthesised molecules due to the "crowding" effect. This may explain why these uPA variants demonstrated worse secretion and higher propensity to aggregate.
The aggregation of uPA on its own is unlikely to be the cause of its poor secretion by yeast cells, since transformants bearing a single copy of the uPA expression cassette integrated into the MOX or LEU2 loci showed both higher secretion rate of uPA and higher degree of its aggregation, than transformant with a single copy of this cassette integrated into a telomere region. Moreover, it is possible that to some extent, the aggregation of uPA even improves its secretion: accumulation in aggregates may decrease uPA crowding in the ER, thus facilitating its proper folding and subsequent secretion.
The increased aggregation of uPA-KDEL in the H. polymorpha opu24 mutant provides an additional evidence for the interference of uPA accumulation within the ER with its folding. We suggest that the degradation rate of uPA-KDEL in this mutant was reduced due to the ERAD lesion. In contrast, the opu24 mutation did not stimulate aggregation of uPA lacking the ER retention signal. This indicates that a significant portion of uPA escaped from the ER, probably due to improved folding, which may be the reason of the "supersecretion" phenotype of this mutant.