Initial aim was to identify specific genetic elements (promoter, leading sequence and terminator), as well as culturing conditions (medium composition, temperature, duration of culture) that would lead to optimal protein secretion in genetically distant S. cerevisiae strains. Moreover, the appropriate protocols had to be established that would guarantee robust and sensitive detection of secreted proteins. In total 12 different constructs were designed and tested on 20 different S. cerevisiae strains. Once the most promising construct was selected, a large scale screen with 250 S. cerevisiae strains was performed for β-glucosidase and AH3-mIgA secretion. This screen showcased that ≈70% of natural isolates were able to secrete higher amounts of protein than the most common lab strains, namely S288c and CEN.PK. Furthermore, protein secretion is a cargo-dependent phenotype, as there was no correlation between secretion of β-glucosidase and AH3-mIgA.
Next, three strains which exhibited high protein secretion for both β-glucosidase and AH3-mIgA, as well as two strains each of which was a good secretor of a single protein were selected and crossed in a “Round Robin” scheme to create in total two "Round Robins" and eight unique hybrids. Each of these hybrids was used for Bulk Segregant Analysis (BSA) and altogether 750 F1 haploid segregants per cross were phenotyped for their capacity to secrete proteins. Importantly, crossing of parents capable to secrete high amounts of proteins resulted to “Best parent heterosis”, since ≈10% of the segregants secreted up to Log2fold ≥ 1.5 more protein than their parents.
These "superior" segregantsas well as an equal pool of "inferior" segregants (i.e segregants that secreted less protein than the original parents) were used for a QTL study, where 8 loci which could be linked to the secretion of both cargos were identified. Aiming to identify specific alleles and genes which are responsible for the improved protein secretion phenotype Reciprocal Hemizygosity Analysis (RHA) was performed for 4 of the identified loci. Therefore, genes responsible for the observed phenotype were detected, while allele swap between the parental strains pinpointed the causative alleles. Several SNPs detected in these alleles were engineered in two different strain backgrounds, a commonly used lab strain and a bioethanol strain which in the initial screen exhibited limited secretion capacity. The most promising of these SNPs led to significant increase of protein secretion and as a result the newly engineered strains could perform equally well with the best identified candidates of the original high-throughput screen.
