TAQing2.0 for genome reorganization of asexual industrial yeasts by direct protein transfection

Introduction of exogenous proteins into living cells by the CPP method

Methods for introducing proteins, enzymes, and peptides into live cells using CPP-based protein transfection reagents have been developed primarily for animal cells, which lack cell walls. Meanwhile, a protein has also been reported to be successfully transfected into plants²⁷. Our goal is to establish a method to induce genome rearrangements by transfecting endonuclease proteins into non-conventional yeasts such as Cu. However, to our knowledge, similar attempts in yeast and fungi have not been reported. We thus attempted first to introduce a foreign protein by the CPP method after spheroplasting by removing the cell wall of Sc and then further to find optimal conditions for Cu. An experiment was conducted to introduce β-galactosidase (β-Gal) into cells using spheroplasts prepared from the Sc strain S288c and the Cu strain NBRC0988. Intracellular transfection was quantitatively analyzed by the activity of β-Gal in cell extracts. A method (Xfect method) using Xfect, a transfection reagent, was employed³⁸. Briefly, β-Gal and Xfect were pre-mixed to form a complex, which was then incubated with yeast cells.

In both yeasts, the β-Gal activity of cell extracts after removal of cell debris was increased when spheroplasts were incubated with Xfect (+) and β-Gal. More importantly, the live cells with intact cell walls treated with Xfect (+) and β-Gal were also found to show higher activity than those without Xfect (−) (Fig. 1a). In addition, the Xfect-treated cells of both yeasts exhibited similar levels of β-Gal activity irrespective of the spheroplasting. We did not observe any severe toxicity of the Xfect reagent for both yeast species (Supplementary Fig. 1). Notably, the β-Gal activity in the cell extract prepared from intact cells was generally higher in Cu than in Sc, indicating that protein transfection by the CPP method is more efficient in Cu.

These experiments alone cannot exclude the contribution of proteins adsorbed to the cell membrane. Following the introduction of β-Gal into cells by the CPP method, the cell surface was treated with trypsin to digest and remove proteins adsorbed to the cell wall. The β-Gal activity was then measured to assess the amount of intracellular protein uptake^32,40. The results showed that a sufficient amount of β-Gal activity was detected, even in cells treated with trypsin, indicating that β-Gal was certainly transfected into cells (Fig. 1b). Green fluorescent protein (GFP), which is fused to a membrane-penetrating peptide, has been reportedly transfected in live cells of C. albicans via endocytosis³². Endocytosis is considered to be the primary mode of action for intracellular uptake of foreign proteins or peptides using Xfect as a carrier³⁸. These results led us to propose a minimally invasive method of protein transfection into yeast cells with intact cell walls by Xfect method.

The intracellular transfection of β-Gal was then detected using live-cell imaging to support the above colorimetric results. The β-Gal/Xfect complexes were mixed with live cells of Sc or Cu, excess complexes were removed by washing, and then SPiDER-βGal, a fluorescent reagent, was added. SPiDER-βGal has cell membrane permeability, is converted to a quinone methide intermediate by an enzyme–substrate reaction with β-Gal, and exhibits fluorescence and intracellular retention by covalent binding with SH groups of nearby proteins⁴¹. Fluorescence signals were consequently noted inside cells and around the periphery of cells after addition of Xfect (+) for both Sc and Cu (Fig. 1c). In addition, almost no fluorescence signals were detected when no Xfect (−) or SPiDER-βGal (−) was added (Supplementary Fig. 2). These results demonstrated that the fluorescence image detected was not a false-positive, such as autofluorescence derived from cell stress associated with Xfect treatment.

Since a higher fluorescence intensity was observed in the cell periphery in Fig. 1c, the possibility that the β-Gal/Xfect complexes adhere to the cell membrane and react with SPiDER-βGal on the membrane to become fluorescent could not be ruled out. The cells were thus reacted with the β-Gal/Xfect complexes and suspended in citrate buffer (pH 4.0) after washing to inactivate β-Gal on the cell membrane. The citrate buffer was immediately replaced with normal phosphate-buffered saline (PBS) (pH 7.4), followed by addition of SPiDER-βGal for cell observation. Strong fluorescence signals at the cell periphery were attenuated, but intracellular fluorescence signals were still detected at a high level (Fig. 1c). These results demonstrated that fluorescence signals detected in the yeast cells were not derived from β-Gal adsorbed on the cell membrane but were attributable to intracellularly transfected β-Gal. The Xfect was thus found to function as a protein carrier and β-Gal to be intracellularly delivered while maintaining its activity.

Next, the differences between Sc and Cu were assessed in terms of the intracellular delivery effects of Xfect. The concentration of SPiDER-βGal was 0.2 μM for Cu, compared with 1.0 μM for Sc, as shown in Fig. 1c. This is because the fluorescence signals are saturated following addition of 1.0 μM of SPiDER-βGal to Cu under the same excitation laser intensity and exposure time conditions as for Sc. The intracellular fluorescence level was significantly higher in Cu than in Sc (Fig. 1d) when the level was measured quantitatively in the experiment, Xfect/citrate-treated cell images, shown in Fig. 1c. The β-Gal/Xfect complexes are thus seemingly more efficiently incorporated in Cu cells than in Sc cells. This speculation is consistent with the results of the colorimetric method shown in Fig. 1a, b.

We have thus established a method in which proteins/enzymes can be easily transfected into yeast cells while maintaining their activity even in Cu, a non-laboratory type yeast with a strong chitin cell wall.

Preferable conditions for protein transfection into Cu by the Xfect method

The transfection efficiency of a foreign protein is an important factor in the delivery of an active protein to the cellular target organelle at an appropriate timing and the continuous expression of its physiological function. We therefore searched preferable conditions for protein transfection by altering buffer reagents. The use of HEPES, a Good’s buffer, has been reported to increase the efficiency of protein transfection into mammalian cells⁴². Thus, the optimal conditions for transfection into Cu were investigated on the basis of the intracellular activity of β-Gal delivered. According to a common protocol for animal cells, Xfect should be used at 60 to 80% cell confluence. For budding yeast in the above section, the Xfect method was applied to cells in the exponential (log) phase of growth in yeast-peptone-dextrose (YPD) liquid medium. We then compared three types of media, YPD, synthetic defined (SD), and yeast extract with supplements (YES). The highest β-Gal activity was observed in YPD medium (Supplementary Fig. 3a). Next, the efficiency of the Xfect method was assessed at each stage of proliferation in YPD medium in the early exponential phase (OD₆₀₀ value up to 1), middle exponential phase (value 1–10), late exponential phase (value 10–20), and stationary phase (value 20–23), and the highest β-Gal activity was detected when cells in the early exponential phase (value up to 1) were used (Supplementary Fig. 3b).

We furthermore examined the optimal conditions for the transfection of β-Gal/Xfect into Cu cells. In the standard procedure, cells are suspended in PBS and incubated at 37 °C for 60 min. When various exposure times and treatment temperatures were tested, we found 30 to 120 min and 20 to 37 °C were appropriate, respectively (Supplementary Fig. 3c, d). We also examined the composition of the buffer used to suspend cells (Supplementary Fig. 3e–g); the highest β-Gal activity was observed with salt-free 2-morpholinoethanesulphonic acid (MES) (pH 6.0) (Supplementary Fig. 3e, f). We note that this result is likely to reflect an optimal condition for protein transfection, rather than for β-Gal activity, since β-Gal had similar activity between pH 5.0 and pH 10.0, regardless of the buffer type (Supplementary Fig. 3g). Thus, the most important factor in the optimization was the extracellular environment at the time of protein transfection (Supplementary Fig. 3f).

We finally established an optimization procedure by combining each condition that showed the maximum β-Gal transformation efficiency, that is a condition in which Cu is incubated in YPD medium until the early exponential growth phase and brought into direct exposure to the β-Gal/Xfect complexes in the salt-free MES (pH 6.0). The efficiency of β-Gal transfection by the original procedure was 231 mUnits while 607 mUnits in the optimized one, indicating that the latter had 2.6-fold higher efficiency (Fig. 2a). Such improved effects by the optimized protocol were again confirmed by the SPiDER-based measurements of the intracellular β-Gal activity (Fig. 2b, c).

We further assessed whether the optimized protocol were also effective for proteins other than β-Gal. According to the quantitative comparison of Alexa Flour 488-conjugated goat IgG (Alexa488-IgG) with a mean molecular weight of 160 kDa, the transfection efficiency was higher in MES than in PBS (Fig. 2d, e). GFP was assessed in the same manner, and the transfection efficiency was found to be higher when MES was used as a cell suspension buffer in comparison with PBS (Fig. 2f, g). Furthermore, foreign proteins were transfected into Cu cells only when those proteins were incubated with Xfect (Fig. 2d, f). In conclusion, the optimization procedure of the Xfect method for Cu was applicable for various proteins.

Genome breeding of Cu (torula yeast) using the protein-based TAQing system (TAQing2.0)

We have developed the TAQing system, which simultaneously induces multiple genomic DSBs via transient heat-activating/inducing of TaqI in Sc and A. thaliana cells¹⁶. The DNA cleavage activity of TaqI is maintained at a low level at room temperature but markedly activated at a higher temperature¹⁶. Since one TCGA sequence theoretically appears per 256 base pairs in the genome, breaks can be introduced at many genomic sites, and this allows for the induction of multisite chromosomal rearrangements by facilitating rejoining of the cleavage sites. In addition, we analyzed the genomic DNA sequence of Sc mutants obtained by the TAQing system, and found that SNVs were limited, but insertions-deletions (InDels), homologous/nonhomologous chromosome translocations (TLs), and CNVs occurred frequently. The recognition sequence (TCGA) was often detected at the break points in interchromosomal TLs. The TAQing system can efficiently generate mutant strains for quantitative phenotypes known to involve a large number of genes¹⁶. In fact, it has been shown that a variety of mutants with altered phenotypes can be obtained in a very short period, such as yeast with both xylose-assimilation ability and high-temperature fermentation ability and plants with various morphological changes¹⁶. The TAQing system thus enables rapid genome improvement even by skipping sexual reproduction processes including meiosis and sporulation.

We thus considered that application of the TAQing system to Cu, which has been widely used as edible/fodder yeast and whose safety has been approved by the United States Food and Drug Administration⁴³, would overcome the challenges of industrial use that only limited mutations are obtained in classical breeding of this yeast, since Cu is a high-order polyploid non-conventional yeast that has lost their sporulation ability. We also considered that combination of the above CPP method and the TAQing system may solve the problem of transfecting foreign DNA into yeasts for edible/fodder use.

We thus designed another version of TAQing system using the CPP-mediated protein transfection to Cu. Since Cu reproduces asexually, chromosome-level recombination is limited to accidental events during mitosis and its frequency is considered extremely lower than homologous recombination during meiosis. Large-scale genomic rearrangements induced by the TAQing system are therefore expected to produce valuable mutants than ever obtained. We additionally expected to take advantage of the TAQing system, which induces genomic rearrangements more effectively in polyploid species¹⁶, since the Cu genome is estimated to be polyploid^43,44. From these, we proposed alternative derivative of the TAQing system based on the direct protein delivery into the cell nucleus (the TAQing2.0 system) (Fig. 3a).

The protein size that can pass through nuclear pores by diffusion is reportedly 9 nm in diameter or less than 60 kDa in molecular weight⁴⁵. The estimated molecular weight of TaqI is ~31 kDa, which can easily clear the hurdle, though we fused TaqI with a nuclear localization signal (NLS: PKKKRKV) of the SV40 large T antigen at its N-terminus. This NLS-tag is also expected to minimize intracellular degradation and rapidly deliver/localize in the target cell nucleus.

NLS-TaqI with a 6×His-tag for purification was expressed with a pET system in Escherichia coli, and then NLS-TaqI was purified under high concentrations (3.3–4.7 mg/mL) in a cobalt resin affinity column (Supplementary Fig. 4a). The purified NLS-TaqI showed the same cleavage pattern as commercially available TaqαI (New England Biolabs Inc.), demonstrating that no changes occurred in the substrate specificity (Supplementary Fig. 4b).

Cell viability is a parameter for assessing the induction of DSBs by transfected NLS-TaqI. The viability of TaqI-transfected Sc cells after heat treatment has been reported to decrease to 20% or less in haploids and 40% in diploids¹⁶. The NLS-TaqI/Xfect complex was introduced into Cu cells using the optimized procedure described above. The cells were washed and incubated in YPD medium at 30 °C for 30 min for recovery. Subsequently, the Cu cells were incubated at 38 °C for 90 min to temporarily and partially activate TaqI. As controls, we employed samples incubated at 30 °C for 90 min or those transfected with nontoxic bovine serum albumin (BSA). The diluted cells were then plated on YPD agar medium, and the viable colonies were counted (Fig. 3b, c).

When the viability of cells treated with the Xfect reagent alone was designated as 100%, the viability score of cells incubated with the NLS-TaqI/Xfect complex decreased to 24.1% at 30 °C and markedly to 8.5% at 38 °C (Fig. 3c). Increasing the incubation temperature resulted in marked reduction of cell viability. Thus, the activity of transfected NLS-TaqI was indeed temperature dependent. Meanwhile, the viability of the BSA-treated control unchanged at much higher levels at 30 and 38 °C. Possibly, either the thermal resistance of cells was increased or the cytotoxicity of Xfect was reduced owing to incubation with BSA. These results suggested that NLS-TaqI was transfected into Cu cells and DSBs were induced in the genome.

Application of the TAQing system to Sc induces cellular morphological alterations¹⁶. Changes in chromosome size in these cells were often noted by pulsed-field gel electrophoresis (PFGE)¹⁶. For analysis of Cu after applying the TAQing2.0 system (TAQed Cu), we picked up clones with smaller colony size followed by selecting strains with cellular morphological changes. Of 1352 colonies of TAQed Cu formed on YPD agar medium, 38 small colonies were observed under a bright-field microscope. Six of these strains had changes in their flocculation phenotype. They were consecutively subcloned 5 times in YPD medium. After the passages, 2 strains still maintained their aggregability (Fig. 3d). We further analyzed these 2 strains (AG4 and AG9) exhibiting stable strong aggregability after the passages. RNA-seq experiments for AG4 and AG9 strains supported their hyper-flocculation phenotypes, since we observed marked increase in the expression of flocculation genes such as CuFLO1 and CuFLO5 in these strains (Supplementary Fig. 5 and Supplementary Table 1a).

Changes in chromosome size were analyzed by PFGE for Cu wild type, AG4 and AG9 strains: Cu wild type (WT), AG4, 5-round passaged AG4 (AG4×5), AG9, and 5-round passaged AG9 (AG9×5). Both AG4 and AG9 had altered chromosome sizes (Fig. 3e and Supplementary Fig. 6). No changes were detected in the band sizes before and after the 5-round passages, indicating that the genomic structure, once rearranged, was stably maintained at least over several rounds of passages. Figure 3e showed that two bands, indicated by leftwards red arrow, had been clearly shifted up in both AG9 and AG9×5, and this would be perhaps due to large-scale structural alterations around rDNA in ChrI, partial duplication, or aneuploidy in ChrIII (described in details below).

Whole-genome resequencing of Cu mutants obtained by TAQing2.0

We then analyzed the whole-genome sequences of AG4 and AG9. We first determined the reference genome of the original Cu strain to accurately identify TAQing-dependent rearrangements. The draft genome of the Cu WT was previously reported using a 454/Roche sequencer⁴⁶. Rupp and coworkers reported that the ploidies of Cu and Cyberlindnera jadini (C. jadinii) were triploid and diploid, respectively, based on the single nucleotide variation (SNV) analysis of Cu NBRC0988 and its possible ancestor C. jadinii NRRL y-1542⁴⁴. Meanwhile, the frequency of local SNPs in C. jadinii has been reported to be triploid, tetraploid, or haploid (large deletion), instead of diploid in the specific region (several hundred kilobases) of each scaffold.

Since the genome composition of torula yeast remains controversial as described above, we employed a combination of long-/short-read sequencing to obtain a more accurate reference genome sequence (Supplementary Table 2). Thirty contigs (total length: 14,097,946 bases) were generated through both PacBio Sequel and Illumina MiSeq data with the assembler MaSuRCA⁴⁷. Meanwhile, when assembly was performed with the assembler FALCON, 24 primary contigs and 234 alternative haplotigs were obtained.

By the manual connection of the contigs derived from MaSuRCA and FALCON using BLAST⁴⁸ for local alignment search and YASS⁴⁹ for genomic similarity search, we assumed that torula yeast is triploid with six chromosomes (Fig. 4a and Supplementary Table 3). In consistent with this assumption, we detected 5–20 or more repetitive units of the telomeric sequence GGGTGTCT, which is similar to the telomeric repeat of Sc⁵⁰, at each end of these six chromosomes. Notably, we identified an independent short extra-chromosome with telomeric repeats at both ends, consisting of copy of ~700 kb segment of the right arm of chromosome-II (hereafter referred to as ChrII, Fig. 4b).

Genomic sequencing of AG4 and AG9 using Illumina NovaSeq generated data of 4–12 Gb (at least 280-fold coverage, the Cu genome size is 14 Mb). The sequence obtained was mapped to the reference genome to assess large-scale genomic rearrangements based on changes in the frequency and coverage of SNVs. We found that AG4 had two break-induced repairs (BIRs) in ChrI and ChrIV, nine gene conversion (GCV) in ChrI, ChrII and ChrIV, and three large deletion in ChrI and ChrII (Fig. 4b, c and Table 1). AG9 had one BIR in ChrIII, five GCV in ChrI and ChrIII, one large deletion (Del) over 67 kb in ChrIII and one smaller Del about 6 kb in ChrI (Fig. 4b and Table 1). Figure 4d illustrates a large Del in ChrIII and the large Del was found in TaqI-recognition sequence (TCGA), and also confirmed by PCR amplification (Fig. 4d). SNVs and InDels were detected at three sites in AG4 and four sites in AG9 (Table 1). These results showed that AG4 and AG9 had multiple large genomic rearrangements.

Next, to detect nonhomologous chromosomal recombination such as TLs, we focused on reads mapped between different contigs and extracted regions where the boundary sequence was TCGA, the recognition sequence of TaqI. In addition, the coverage of AG4 and AG9 was normalized to the WT one to quantitatively estimate chromosome copy numbers in the extracted regions. Consequently, we found AG4 has one nonhomologous TL (the actual coverage data are shown in Fig. 4a, magenta is AG4). A TL event was detected between ChrI and ChrIV, and their physical linkage was confirmed by PCR amplification (Fig. 4b, e). AG9 had aneuploidy, i.e., ChrIII was tetrasomy, and the right arm of ChrI had an additional copy (partial duplication).

These results showed that the TAQing2.0 system, which is based on direct protein delivery into the cell nucleus, induced TaqI-mediated large-scale genomic rearrangements in Cu and enabled the generation of mutants with an altered flocculation phenotype.