In a previous work, we proved the capability of the evolved α9H2 leader [42] to improve the secretion by S. cerevisiae of diverse laccases compared to other evolved signal peptides [46]. The α9H2 leader differs from the native α-factor preproleader, αnat leader from now on (Fig. 1b, Fig S1), in seven mutations (Fig. 1c) accumulated through subsequent evolution campaigns. Mutations Aα9D, Fα48S, Sα58G, Gα62R were added during the directed evolution of Pycnoporus cinnabarinus laccase (PcL) for functional expression in S. cerevisiae [32], and Aα87T during the evolution of PM1 laccase (PM1L [33]); all accumulated in the leader sequence of 7D5 chimeric laccase after DNA shuffling of evolved PcL and PM1L [44]. The Aα20T and Qα32H mutations were selected during subsequent evolution of 7D5 laccase to obtain PK2 variant [42]. It is worth noting that αnat and α9H2 leaders contain 2 extra mutations (Lα42S and Dα83E) with respect to the original MFα1 gene [18]. Both mutations come from the α-factor preproleader of Invitrogen (inserted in pPICZα plasmids [35]). In addition, the αnat leader we used here holds a EcoRI restriction site that was introduced to facilitate genetic engineering and encodes for a Glu-Phe extra sequence downstream the spacer and before the foreign protein (Fig. 1b, c).
First, the secretory potentials of the α9H2 leader was compared with the αnat leader for laccase production. Both signal sequences, fused to the CDS of PK2 and ApL laccases were cloned in the pJRoC30 expression vector to transform S. cerevisiae cells. Yeast clones were grown in 96-well plates in SEM laccase expression medium [45] and the secreted activity was determined by the oxidation of ABTS (absorbance peak at 418 nm). While both construction gave detectable laccase activity, α9H2 leader provided significantly higher laccase activity levels than αnat leader, roughly twofold for PK2 and 12-fold for ApL (Fig. S2), confirming previous results obtained with ApL [46]. Due the superiority of α9H2, it was used as upper reference leader in this study. Two engineering strategies were carried out: (i) a bottom-up process over αnat to study the individual effect of mutations accumulated in α9H2 and others, and their epistatic interactions, and (ii) a top-down process over α9H2 to get rid of possible deleterious mutations accumulated during the in vitro evolution pathway that could mask the effect of real beneficial mutations.
Bottom-up design of αnat leader
Site-directed mutagenesis on αnat leader was performed to independently assess the effect of the following mutations: (i) the seven mutations accumulated in α9H2 leader (Aα9D, Aα20T, Qα32H, Fα48S, Sα58G, Gα62R, Aα87T) that were individually added to αnat leader, (ii) the two mutations found in αnat from pPICZα plasmid (Lα42S, Dα83E) that were removed individually, and (iii) four potentially beneficial mutations (Rα2S, Tα24S, Lα44S and Eα86G) selected in previous studies [32, 33] that were added individually to αnat leader. The resulting 13 single-mutated αnat leaders were fused to each laccase CDS (PK2 and ApL), cloned and expressed in S. cerevisiae. Laccase activities secreted by ten replicates of each clone grown in 96-well plates were screened with ABTS as substrate. The average laccase activity of each single mutant was normalized to the parental activity (with αnat leader), and similarities and differences in secretion were statistically supported by the Tukey's range test. Satisfactorily, most of the mutations showed the same behaviour for the production of both laccases (Fig. 2). Clearly beneficial mutations were localised in the pre-region or near it; Rα2S, Aα9D and Aα20T augmented around twofold the secretion of both laccases, whereas Tα24S mutation improved their secretion dissimilarly (1.3-fold for PK2 and threefold for ApL). On the other hand, Qα32H, Lα44S, Fα48S, Sα58G had no effect on the secretion of PK2 and ApL, while Gα62R mutation had a detrimental effect on PK2 secretion (0.57-fold) and neutral effect on ApL. Mutations located in the spacer region had different effects: Eα86G seemed to have no influence on laccase secretion, while Aα87T highly improved the activity levels of ApL (2.3-fold) but not of PK2. Removal of Lα42S mutation decreased laccase secretion to 0.8-fold (ApL) and 0.4-fold (PK2) the activities detected with αnat leader. Reversion of Dα83E had no effect. Nevertheless, both mutations, beneficial Lα42S and neutral Dα83E, were maintained in next assays because they were originally present in αnat leader from Invitrogen and every substitution selected during the evolution to α9H2 leader could have had epistatic interactions with them.
Next, we analysed the potential synergism between beneficial mutations Rα2S, Aα9D, Aα20T and Tα24S. Based on a proximity criteria, double (αR2S,A9D and αA20T,T24S) and quadruple (αR2S,A9D,A20T,T24S) mutants of αnat leader were obtained and fused to ApL and PK2 laccases. In addition, we built a double mutant (αE86G,A87T) at the spacer region. Again, ten replicates of each S. cerevisiae clone expressing the aforementioned constructions were grown in 96-well plates; the activities of the supernatants were measured and the corresponding average activities normalized to the laccase activity obtained with the αnat leader (Fig. 3a, b). All data were supported by Tukey´s range test. The αR2S,A9D leader diminished laccase secretion with respect to αnat leader (to 0.2-fold for PK2 and 0.9-fold for ApL). The αA20T,T24S leader favoured ApL secretion as compared to αnat leader (1.4-fold), but the combination of both mutations was detrimental compared to the activity levels obtained with the single-mutated leaders αA20T and αT24S. Conversely, the use of αA20T,T24S leader with PK2 resulted in similar improvement than that obtained with αT24S. The quadruple mutant αR2S,A9D,A20T,T24S led to minimal laccase levels (not detectable with PK2 and 0.3-fold with ApL). Surprisingly, αE86G,A87T notably enhanced production of both laccases, around twofold for PK2 and 12-fold for ApL, suggesting a positive epistatic effect between both mutations.
Afterwards, to allow exploration of other possible advantageous combinations between Rα2S, Aα9D, Aα20T, Tα24S, Eα86G and Aα87T mutations, the 6 individual mutants, 3 double and 1 quadruple mutated α-factor leaders were subjected to in vivo recombination in S. cerevisiae, using PK2 as model laccase (Fig. S3a). After screening 1600 clones of the library, laccase activities were normalized to the activity obtained with the αnat leader (Fig. S3b). Besides, αE86G,A87T leader was included in the comparison as upper reference because it had produced one of the highest total activity improvements with PK2 (and the highest for ApL, Fig. 3a, b). The five fittest clones carried αA9D,A20T,T24S and αA9D,A20T leaders. Mutation Rα2S was discarded for future assays since it seemed to be incompatible with the others (Fig. 3a, b).
Finally, since the combination of the winner set of mutations αA9D,A20T and αA9D,A20T,T24S with one of the best (αE86G,A87T) have not been selected from the in vivo DNA recombination assay, we synthesised two final leaders: αA9D,A20T,T24S,E86G,A87T and αA9D,A20T,E86G,A87T to evaluate their joint effect. It was confirmed that αA9D,A20T,T24S and αA9D,A20T leaders significantly raised laccase secretion with respect to αnat (Fig. 3c, d), being the increment more pronounced with ApL (10–12-fold) than with PK2 (threefold). Conversely, based on Tukey´s range test, the production of none of the two laccases tested was improved by the addition of Eα86G and Aα87T mutations to these leaders. We therefore discarded the latter mutations from the final optimised signal peptide. Finally, it was evidenced the neutral effect of Tα24S mutation, so it was discarded as well. In conclusion, we selected αA9D,A20T as the optimised leader from the bottom-up process, because it remarkably surpassed the secretion potential of αnat leader to values similar (12-fold for ApL) or better (threefold for PK2) than those obtained with α9H2 leader.
Top-down design of α9H2 leader
In the top-down approach we aimed to obtain an optimised and simplified version of the α9H2 leader by removing possible deleterious or neutral mutations that could have been introduced during its in vitro evolution pathway and might mask the effect of beneficial mutations accumulated in the signal peptide. In a first cycle, mutations Qα32H, Fα48S, Sα58G and Gα62R were individually reverted from the α9H2 leader since their neutral effect on laccase secretion were confirmed during the bottom-up approach. The resultant α-factor leaders (Signal Peptides) were named as follows: SP1 (Hα32Q mutation), SP2 (Sα48F mutation), SP3 (Gα58S mutation) and SP4 (Rα62G mutation) (Fig. 4a). SP1, SP3 and SP4 had no significant effect on the secretion of PK2 or ApL, as compared to the laccase activities detected with α9H2 leader. SP2 did not improved ApL production, but raised 1.3-fold the production of PK2, suggesting a possible deleterious effect of Fα48S mutation in α9H2 leader for secretion of this laccase (Fig. 4b).
In a second evolution cycle, given the aforementioned detrimental effect of Fα48S mutation in α9H2 leader for PK2 laccase (Fig. 4b), and of Gα62R observed in the bottom-up approach also for PK2 (Fig. 2), both mutations were simultaneously reverted in SP5. It increased secretion of PK2 similarly to SP2 (with only reversion of Fα48S mutation), whereas no improvement in ApL levels were observed respecting α9H2. Two more leaders were designed in parallel to assess the effect of Qα32H and Sα58G mutations on SP5 environment; each had three reverting mutations: SP6 (Hα32Q, Sα48F, Rα62G) and SP7 (Sα48F, Gα58S, Rα62G). Alike SP5, SP6 and SP7 provided similar levels of PK2 than SP2, confirming the detrimental effect of Fα48S mutation on α9H2 leader for laccase secretion.
Lastly, the combined absence of the four mutations Qα32H, Fα48S, Sα58G and Gα62R was assayed in SP8 leader. SP8 showed no effect on ApL laccase secretion, whereas the simultaneous reversion of the four mutations rendered significant higher (1.8-fold) PK2 laccase levels as compared with α9H2 leader. Despite the fact that Fα48S seemed to be the only deleterious amino acid change in α9H2 leader for PK2 laccase secretion (SP2, Fig. 4b), the four aforementioned mutations have a larger deleterious effect together than separately. Thus, mutations Aα9D, Aα20T and Aα87T seem to be responsible for the greater secretory potential of α9H2 with respect to αnat leader.
Selection of a final optimised α-factor leader
The two final leaders selected from the bottom-up (αA9D,A20T) and top-down (αA9D,A20T,A87T) pathways were compared for secretion of ApL and PK2 laccase by S. cerevisiae cultured in flasks. The αnat and α9H2 leaders were included in the assay as lower and upper references. In this case, the minimum laccase expression medium (SEM) utilised in the micro-fermentations was replaced by a richer medium (EB) because the reduced growth of the yeast in SEM could limit laccase production in flasks [45]. We aimed as well to check the reproducibility of the results obtained in other culture medium and conditions. Optical densities (Fig. S4) and laccase activities (Fig. 5) of the cultures were monitored for 4 days. All S. cerevisiae clones grew similarly, but they produced dissimilar laccase activities. After 4 days of incubation, αA9D,A20T,A87T leader fused to PK2 laccase provided up to 4800 U/L with ABTS, while αA9D,A20T yielded around 3700 U/L, which respectively represent 18-fold and 14-fold higher activities than that detected with αnat leader, and 1.8-fold and 1.4-fold improvements respecting laccase levels detected with α9H2 leader (Fig. 5a). On the other hand, αA9D,A20T,A87T, αA9D,A20T and α9H2 leaders fused to ApL gave rise to similar laccase levels (around 260 ABTS U/L), which represent a 26-fold improvement respecting the laccase activity detected with αnat (Fig. 5b). The superior secretory potential of the three engineered leader sequences with respect to αnat was therefore confirmed. Furthermore, the similar (ApL) or markedly improved (PK2) laccase levels obtained with the optimised leaders with respect to the α9H2 leader, pointed out the essential role that Aα9D and Aα20T mutations play in the superior secretory capability of α9H2 leader. By contrast, the dissimilar results obtained with αA9D,A20T or αA9D,A20T,A87T in the production of the two laccases suggested that the variable effect exerted by Aα87T mutation may be influenced by the sequence of the fused protein.
Taking all this into account, αA9D,A20T was selected over αA9D,A20T,A87T leader. Since αA9D,A20T leader carried also mutations Lα42S and Dα83E (Invitrogen), we double checked their contribution in αA9D,A20T leader context by individually discarding them from the selected leader fused to PK2 and ApL. As previously shown, the absence of Lα42S had a strong negative effect on laccase production (0.5-fold reduction), whereas we confirmed the neutral effect of Dα83E (Fig. S5). Mutations Aα9D, Aα20T, Lα42S, Dα83E were included in the optimised all-purpose leader for further assays, named αOPT from now on.
Expression of other enzymes
The secretory potential of αOPT leader was evaluated for the production by S. cerevisiae of other fungal oxidoreductases like two more laccases from Pleurotus eryngii (PeL) [46] and Pycnoporus cinnabarinus (PcL) [32], an aryl-alcohol oxidase from P. eryngii (AAO) [47] and a versatile peroxidase (VP) from P. eyringii [48]. Besides, we assayed it with fungal hydrolases such as two β-glycosidases (BGL2 and BGL3) from Talaromyces amestolkiae [49, 50] and a sterol esterase (OPE) from Ophiostoma piceae [51]. To this aim, the native signal peptides were removed and replaced by αOPT, αnat and α9H2 leaders for enzyme expression in S. cerevisiae (the two latter used as lower and upper references). Yeast cells (ten replicates of each clone) were grown in 96-well plates in SEM, and the secreted enzyme activities were measured and normalized to the activities obtained with αnat leader (Fig. 6).
In general, αOPT leader provided enzyme secretion levels significantly higher than those obtained with αnat leader. In some cases, the increments in enzyme production obtained with αOPT leader were remarkable: 10–20-fold higher levels for PeL, PcL, ApL and OPE than those obtained with αnat leader (Fig. 6). As regards α9H2 leader, it significantly enhanced the production of all tested laccases respecting the use of αnat leader, but to a lower or at most similar extend than αOPT leader. Moreover, α9H2 performance with the rest of enzymes was not as good, being in general similar or worse than αnat leader (e.g. 0.35-fold for VP and 0.07-fold for AAO). Taking all this into account, αOPT leader emerges as a general signal peptide suited for efficient expression of fungal enzymes in S. cerevisiae.
Combinatorial saturation mutagenesis on the spacer region
Mutations Eα86G and Aα87T were ruled out from the aforementioned “universal” αOPT leader due to their dissimilar effect on secretion of ApL or PK2 laccases which might be related to the different fused protein sequences. We hypothesised that positions 86th and 87th of the spacer region would play a crucial role in the secretory potential of the signal peptide, and, therefore, they may well be hotspots for engineering the α leader towards the production of a particular recombinant enzyme. To test this hypothesis, positions 86th and 87th of αOPT leader fused either to PK2 or ApL were subjected to combinatorial saturation mutagenesis (CSM), covering all possible amino acid combinations, and the activities of the mutant libraries expressed in S. cerevisiae were screened with ABTS. Population of clones with parental-like activity (inside parent’s confidence interval) were minor in both CSM 86/87 libraries, whereas most clones (53% for PK2 and 69% for ApL) exhibited lower activity than parental αOPT leader, and clones with higher laccase activities represent a 32% in PK2 library and 5% in ApL library (Fig. 7a).
On the other hand, we randomised positions 58/59 and 68/69 of two N-glycosylation sites (Asn in positions 57th and 67th) of the pro-region of the α leader [19, 52], in such a way that the consensus pattern Asn-X-Ser/Thr was conserved. While Asn was maintained, positions 58 and 68 were mutated by whatever amino acid except for Pro and positions 59 and 69 were restricted to Ser or Thr. We used the resulting CSM N-Gly58/59 and N-Gly68/69 libraries (built on αOPT-PK2 and αOPT-ApL) as reference of presumably neutral libraries, and compared the results from their screening with those obtained from the CSM86/87 libraries under the criteria “the larger population of clones with parental-like activity, the less impact the mutated sites have on enzyme secretion”. By contrast to CSM 86/87 libraries, most of the clones (50–60%) exhibited parental-like activities (Fig. 7a), confirming that the 2nd position of N-glycosylation sites was not so relevant for α leader engineering as 86th and 87th positions were. In addition, although clones with improved activities were also found in CSM N-Gly58/59 and 68/69 libraries, the improvements detected were significantly lower. Moreover, the plain shape of their activity landscapes remarks the “neutral” nature of these libraries by contrast with the hill trend of CSM86/87 landscapes (Fig. 7b).
The best amino acid substitutions selected from each CSM 86/87 library were different for PK2 laccase (Eα86T/Aα87N; Eα86D/Aα87N and Eα86D/Aα87G) and ApL (Eα86A/Aα87P; Eα86T/Aα87K and Eα86S/Aα87R). The clones providing the highest secreted activity improvements (αOPT Eα86T/Aα87N for PK2 and αOPT Eα86A/Aα87P for ApL) were cultivated in flask to test laccase production. Production of PK2 laccase was raised roughly twofold and 30-fold as compared with the activity levels provided by αOPT leader and αnat leader, respectively; while for ApL production, the improvements were around 1.3-fold and 34-fold, respectively (Fig. 7c).
Finally, we purified PK2 laccase produced with αOPT and αOPT Eα86T/Aα87N as leaders in S. cerevisiae flask cultures (Fig. S6). In both cases, after deglycosylation with Endo H, the enzyme showed a molecular weight around 53 KDa, coincident with its theoretical MW (Fig. S7). The enzymes purified from both cultures showed also identical specific activities with ABTS regardless of the signal peptide used: 405 ± 23 U/mg and 423 ± 34 U/mg for the enzyme secreted with αOPT and αOPT Eα86T/Aα87N, respectively. With this data and the laccase activity units detected in the culture broths (2800 U/L with αOPT and 4800 U/L with αOPT Eα86T/Aα87N), we determined that the total mg of PK2 laccase secreted with αOPT Eα86T/Aα87N was roughly twice as high the amount of enzyme secreted with αOPT.