A toolkit for facilitating markerless integration of expression cassettes in Komagataella phaffii via CRISPR/Cas9

The CRISPR/Cas9 system described by Gassler et al. [22] was used to generate plasmids co-expressing an sgRNA and the human codon optimised Cas9 enzyme, to target double-strand breaks at each of the three integration sites 4576, PFK1, and ROX1. These three integration sites correspond to intragenic regions of the genome selected based on a previous study [21], where they were shown to be neutral integration sites that could be successfully targeted with CRISPR/Cas9 technology. The plasmids were named CRISPi_04576, CRISPi_PFK1, and CRISPi_ROX1, respectively (Table 1 and Fig. 1A). Transient episomal plasmid maintenance in K. phaffii was mediated by the Saccharomyces cerevisiae CEN6/ARS4 locus and a geneticin resistance marker.

Backbone (BB) donor helper plasmids corresponding to the BB3 level in the hierarchical cloning system employed in the GoldenPiCS toolkit [11] (crBB3 plasmids) were generated for cloning of one or two expression units (Table 1). The donor plasmids harboured either BpiI restriction sites and fusion sites Fs1 and Fs4 for cloning of one transcription unit (crBB3_14 plasmids; Fig. 1A), or BsaI restriction sites and fusion sites FsA and FsC for cloning of two transcription units (crBB3_AC plasmids, Fig. 1B). The cloning sites were flanked by 500 bp sequences for homology-directed repair (homology regions) comprising the sequences upstream and downstream of each CRISPR/Cas9 targeting site. To insert three or more transcription units, the existing linkers may be replaced with linkers containing fusion sites FsA-FsD, FsA-FsE, and so on. A hairpin was inserted upstream of the cloning sites, to minimize any transcriptional interference from upstream endogenous genes following insertion into the K. phaffii genome. The donor plasmids may be linearized prior to transformation into K. phaffii using EcoRI, XhoI, BamHI, or NotI restriction digestion.

The crBB3 plasmids can be used for BpiI or BsaI Golden Gate assembly of any sets of inserts between the homology regions, providing that the outermost fragments harbour the Fs1/Fs4 (GGAG/CGCT) or FsA/FsC (GATC/AATT) fusion sites, respectively. All fragments to be assembled should be devoid of the type IIs restriction enzyme recognition sites (BpiI and/or BsaI), as well as at least one of the restriction sites used for linearization of the donor plasmid prior to transformation.

To evaluate the efficiency of integration using the three chosen target sites, three donor DNA cassettes for expression of eGFP were constructed. The selected promoter and terminator modules were both from the K. phaffii THD3 gene. The eGFP expression cassette was inserted into the three crBB3_14 donor helper plasmids containing homology regions for integration near genes 4576, PFK1, and ROX1, generating donor cassette plasmids crBB3_04576_eGFP, crBB3_PFK1_eGFP, and crBB3_ROX1_eGFP, respectively (Table 1).

K. phaffii was co-transformed with donor cassette plasmids linearized by EcoRI digestion (2 µg, corresponding to 1.1 µg donor cassette DNA) and CRISPi plasmids containing the appropriate sgRNA (100 ng). Transformation reactions were selected for clones carrying CRISPi plasmid by plating on selective agar after 1, 2, and 3 h of recovery after electroporation. Integration of eGFP expression cassettes was evaluated by screening of clones for green fluorescence in an endpoint assay after 48 h of growth (Fig. 2). A total of 2234 clones were tested, up to 100 clones for each reaction and time-point.

The transformation efficiencies differed significantly between integration sites (Tukey’s post hoc test, p ≤ 0.015), with the lowest number of transformants obtained at locus 4576 (Fig. 2A). However, reactions targeting the 4576 site obtained the highest integration efficiencies, with an average of 66% green fluorescent clones (Fig. 2B). Integration efficiencies also differed significantly across integration sites (Tukey’s post hoc test, p < 0.001).

Integration of eGFP into the 4576 locus resulted in 150 (SE ± 49) successful integrations per µg donor cassette DNA, while the corresponding results for the PFK1 and ROX1 loci were 289 (SE ± 88) and 812 (SE ± 158), respectively (average values across all time-points). Both transformation and integration efficiencies generally increased slightly from 1 to 3 h of recovery after electroporation (Fig. 2), although the increase in integration efficiency was not statistically significant (ANOVA, p = 0.086). However, there was a significant difference in transformation efficiency between the 1 and 3 h time-points (Tukey’s post hoc test, p < 0.022).

The potential of the presented method for generating recombinant microbial strains for food applications was evaluated through the extracellular production of chicken egg ovalbumin, an ingredient of high relevance for the food industry due to its nutritional and functional properties [32]. GoldenPiCS was designed mainly for the engineering of metabolic pathways [11], rather than for expressing heterologous proteins for secretion. Consequently, the kit does not contain BB1-level modules for secretion signal sequences. For secreted proteins, the secretion signal and the gene encoding the mature protein may be combined in a Golden Gate assembly reaction into the empty backbone plasmid BB1_23, or synthesized as a complete gene cassette containing both a compatible signal sequence and the mature coding sequence. Alternatively, two separate BB1-level plasmid modules can be designed for the secretion signal and the coding sequence, which can then be simultaneously inserted into the BB2 or BB3-level vector backbones. This latter approach, which facilitates testing different signal sequences (and can be expanded to include a signal peptide library [12]), was used in the current study to construct an expression cassette for the chicken ovalbumin gene (OVA) (Supplementary Figure S1). The S. cerevisiae α-mating factor (α-MF) sequence, carrying the L42S mutation mediating increased protein secretion [33], was selected as the secretion signal. The selected promoter and terminator were the same as for the eGFP expression cassette, and the selected target locus was 4576.

Co-transformation into K. phaffii with the CRISPi_04576 plasmid (100 ng) and the linearized donor cassette plasmid (1.65 µg, corresponding to 1 µg donor cassette DNA) resulted in 143 and 157 colony forming units (cfu) per µg donor cassette DNA after 2 and 3 h of recovery, respectively. This transformation efficiency was slightly lower than that obtained for insertion of eGFP in the 4576 integration site, in which on average 214 and 283 cfu µg⁻¹ donor cassette DNA were obtained at the corresponding time-points. PCR screening of randomly selected OVA expression cassette transformants (n = 61) showed that the overall targeting efficiency was 47%, increasing from 40 to 51% after 2 and 3 h of recovery.

Twelve clones were examined for recombinant ovalbumin protein production by growth in 50 mL defined medium for 48 h in shake flasks, to optical density at 600 nm (OD600) of 25.5 (SE ± 0.38). SDS-PAGE and Western blots for supernatant samples showed that two proteins detected by the anti-OVA antibodies, of approximate sizes 45 and 47 kDa, were secreted into the culture medium in all clones (Fig. 3). Based on previous studies [34, 35], the two bands are expected to represent ovalbumin variants with different glycosylation patterns. The mean calculated titer (including both bands) was 35 (SE ± 1.3; range 26–40) mg L⁻¹ and the OVA titer normalized to biomass was 1.38 (SE ± 0.06; range 1.1–1.7) mg L⁻¹ OD600⁻¹. This shows that there was limited variation in protein expression levels between the 12 examined clones, and constitutes a proof of concept that the current approach may be used to produce secreted food proteins.

Next, it was of interest to examine whether protein expression levels varied between the 4576, PFK1, and ROX1 target sites, and/or among different clones for the same integration site. For this analysis, the data from the 48 h endpoint fluorescence screening assay were used. Fluorescence values for clones expressing eGFP, as indicated by green fluorescence detection, were normalized by dividing the relative fluorescence units (RFU) by biomass (OD600). The resulting normalized fluorescence values ranged from approximately 300 to 45,000 RFU OD600⁻¹, with the majority of tested clones (51%) falling within the range 800 to 1400 RFU OD600⁻¹ (Fig. 4A and Supplementary Figure S2). Within this range, the average and median fluorescence for clones with eGFP integrations targeting the ROX1 site was 9–10% lower than those at the 4576 and PFK1 sites. The highest proportion of clones with very high eGFP fluorescence levels (>8000 RFU OD600⁻¹) was obtained for the 4576 integration site (13%), while the lowest proportion (1.4%) was obtained for PFK1.

To confirm that clones differed in eGFP fluorescence levels and to evaluate its impact on cell growth, approximately 20 of the fluorescent clones obtained for each integration site were randomly selected for further analysis in a kinetic fluorescence growth assay. Again, a cluster of strains with relatively low fluorescence levels could be seen, in the range 1400 to 1700 RFU OD600⁻¹ after 40 h of growth, while a subset of tested clones reached higher expression levels (Fig. 4B and Supplementary Figure S3). The large variation in green fluorescence levels among individual clones suggested that the CRISPR/Cas9 approach resulted in variable eGFP expression cassette copy numbers.

Specific growth rates were calculated during the exponential phase for the clones subjected to the kinetic fluorescence growth assay. During the initial phase (from onset of exponential phase to 7.5 h), no significant differences were observed between the specific growth rate of the wild-type control strain, which was 0.13 (SE ± 0.006) h⁻¹, and those of the randomly selected fluorescent clones, which ranged from 0.09 (SE ± 0.04) to 0.18 (SE ± 0.01) h⁻¹. For subsequent periods, specific growth rates did not differ from the control strain for the majority of clones, with a few exceptions. Namely, a significantly lower growth rate was observed compared with the control strain for two of the 4576 clones and three of the ROX1 clones (Supplementary Table S1).

These results confirmed that most of the strains with insertions in 4576, PFK1 and ROX1 grew normally on defined DM1-2 medium, using glucose as a carbon source. Thus, in general, the integrated gene cassettes did not affect cell growth and metabolism of the parental strain after gene insertion (Supplementary Figure S4), except for in a limited number of clones.

The differences in fluorescence levels and growth patterns summarized in the previous sections were further explored by means of whole genome sequencing (WGS) analysis. One clone for each of the three integration sites each with low, medium, and strong eGFP fluorescence levels were selected for analysis (labelled with arrows in Fig. 4B). Data on fluorescence levels and growth kinetics for these nine clones are presented in Table 2 and growth curves are shown in Supplementary Figure S4D. Illumina reads were mapped to the K. phaffii reference genome sequences [36] modified to contain the expected insertions of eGFP expression cassettes, and to the crBB3_04576_eGFP, crBB3_PFK1_eGFP, and crBB3_ROX1_eGFP donor cassette plasmids sequences. Results are shown in Fig. 5A–F and Supplementary Tables S2 to S7, and sequencing and read mapping quality metrics are presented in Table 3.

For the three sequenced clones exhibiting low fluorescence levels (4576_37, PFK1_21, and ROX1_97), BLAST analysis did not indicate the presence of alternative integration sites or multiple insertions. Additionally, PCR analysis covering the entire locus, using primers binding outside of the 5′ and 3′ homologous regions, indicated that each of these low fluorescence clones contained one copy of the eGFP expression cassette. The mean normalized mapping read coverage of eGFP in the low fluorescence clones, relative to the mean coverage of the entire K. phaffii reference genome, ranged from 1.4× to 1.5× (Supplementary Tables S2 to S7). Given that the theoretically expected read coverage of eGFP would be 1× for single insertions, these elevated values are likely attributable to uneven read coverages across the chromosomes, a common phenomenon [37]. Therefore, in the remainder of this text, gene read coverages are reported as ‘coverage relative to eGFP’, meaning that read coverages are normalized to the mean eGFP coverage of the low fluorescence clone for the same target locus (i.e., the coverages of eGFP in the low expression clones are set to 1).

Since the promoter and terminator used in the expression cassette were native to K. phaffii (both linked to the TDH3 gene) these sequence fragments were expected to be present twice in clones with single copies of the integrated expression cassettes. When reads were mapped to the donor cassette plasmid references (which contain only one copy each of the TDH3 promoter and terminator), the mean coverage of the promoter and terminator regions of the expression cassettes was approximately twice as high as the mean eGFP coverage; ranging from 1.6× to 1.8× for the promoter and from 2.0× to 2.4× for the terminator (Supplementary Tables S2, S4, and S6). As shown above, the three clones resulted in similar fluorescence levels, ranging from 1473 to 1588 RFU OD600⁻¹ after 40 h of growth in the kinetic assay (Fig. 4B and Table 2). Together, these data strongly suggested that these three clones contained one copy of the eGFP expression cassette, inserted into the intended target site by CRISPR/Cas9-mediated via homology-directed repair (double crossover cassette exchange).

BLAST analysis of the PFK1 medium fluorescence clone (PFK1_66) identified one contig covering the junction between two tandem donor cassettes fused in a head-to-tail direction. The 406 bp long contig covered the last part of the downstream homology region, the downstream EcoRI site used to linearize the donor helper plasmid, and the first part of the upstream homology region. The mapping read coverages indicated that two copies of eGFP, three copies of the TDH3 promoter and terminator regions, and two copies of each homology region were present (Fig. 5C, D and Supplementary Tables S4 and S5). Sequences matching the remaining parts of the crBB3 donor cassette E. coli plasmid backbone, including the cloramphenicol resistance gene (cat) and the E. coli replication origin (ori), were absent from the genome. These data suggested that two tandem eGFP expression cassettes, interspaced by copies of the two homology regions, had been inserted into the PFK1 target site in the medium fluorescence clone.

The WGS analysis of the PFK1 strong fluorescence clone (PFK1_72) suggested that this genome contained three copies of the eGFP expression cassette (Fig. 5C, D and Supplementary Tables S4 and S5). The genome also appeared to contain two or more copies of the plasmid backbone containing cat and ori. Contigs were identified that spanned both EcoRI sites found on the original crBB3 plasmid. Furthermore, contigs were identified that covered fusions of the donor cassette and the plasmid backbone fragments in the opposite direction than in the original plasmids, i.e., in head-to-head and tail-to-tail configurations. This suggests that multiple fragments had randomly integrated into the target site through ligation of EcoRI-generated cohesive ends.

The PFK1 medium and strong fluorescence clones showed roughly double (2277 RFU OD600⁻¹) and triple (3745 RFU OD600⁻¹) fluorescence levels compared to the PFK1 low fluorescence clone (1473 RFU OD600⁻¹) after 40 h of growth in the kinetic assay (Fig. 4B and Table 2). This supports the conclusion based on the WGS analysis indicating that these two clones contained two and three copies of the eGFP expression cassette.

Copies of the plasmid backbone containing cat and ori, as well as unexpected EcoRI fusions of eGFP donor cassette and the plasmid backbone fragments, were also found in the 4576 and ROX1 medium and strong fluorescence clones (4576_38, 4576_35, ROX1_87, and ROX1_96). For example, three clones contained tandem copies of the plasmid backbone fragment fused together in head-to-tail direction at the EcoRI site. The number of copies of the donor cassette did not strictly correlate with the number of copies of the plasmid backbone fragments (see e.g. Figure 5E and Supplementary Tables S2, S4, and S6). This further supported the hypothesis that multiple fragments harbouring EcoRI cohesive ends had randomly inserted into the target site as part of homology-directed repair after CRISPR/Cas9 cleavage.

The mapping read coverage suggested that the 4576 and ROX1 medium fluorescence clones (4576_38 and ROX1_87) contained 5 and 2 copies of the eGFP expression cassette, respectively. The corresponding strong fluorescence clones (4576_35 and ROX1_96) were estimated to have 14 and 8 copies, respectively, based on mapping read coverage, while the fluorescence levels measured in the kinetic assay were approximately 10 and 6 times higher than for the low fluorescence clones (14 955 and 8944 RFU OD600⁻¹, respectively; Table 2). It is important to note, however, that these estimates become progressively less precise with increasing copy numbers.

The ROX1 target site was located on the q arm of chromosome 2, 36.6 kbp from the distal end of the 2.4 Gbp long chromosome. In the ROX1 medium fluorescence clone the mapping coverage for the region downstream of the integration site (upstream of the upstream homology region) was zero (Fig. 5F), showing that this entire fragment was lost. This corresponds to a loss of 1.5% of the chromosome or 0.4% of the entire genome. In addition to the telomeric region, the missing part of the chromosome contained two rRNA genes and 15 annotated protein coding genes, including ROX1 itself, which encodes a heme-dependent repressor of hypoxic genes [36, 38].

The loss of the distal chromosomal region was associated with a significantly reduced specific growth rate compared with the control strain. Specifically, during 9–11 h of cultivation, the average growth rate was 0.12 h⁻¹, compared to 0.17 h⁻¹ for the control strain. Similarly, from 11 to 13 h of cultivation, the average growth rate was 0.11 h⁻¹, compared to 0.16 h⁻¹ for the control strain (see Supplementary Table S1). At least one other clone out of the 19 ROX1 clones tested in the kinetic growth assay (ROX1_78) showed the same growth pattern as the sequenced clone, as well as a significantly reduced specific growth rate (Supplementary Fig. S4C and Supplementary Table S1).

As seen in Table 2, the fluorescence levels in clones with multiple copies of the eGFP expression cassette, relative to that in low fluorescence single insertion clones, were generally lower than predicted based on the copy numbers estimated from the WGS analysis (listed in Table 3). For the ROX1 medium fluorescence level clone, however, which had two eGFP expression cassettes, the RFU OD600⁻¹ value was approximately 2.9 times higher than in the low fluorescence clone. The reduced growth rate did not sufficiently explain this discrepancy, as the absolute relative fluorescence was still 2.7 times higher in this clone compared to in the ROX1 low fluorescence clone.

There was no evidence in the sequencing data, for any of the nine sequenced clones, suggesting that the eGFP expression cassette had inserted into unintended locations on the chromosome. The only fragments containing unexpected fusions were those described above, spanning the EcoRI sites of the donor plasmid. Unfortunately, the short read sequencing approach cannot confirm nor disprove the presence of single-crossover insertions of donor plasmids into the TDH3 promoter, since the 300 bp long reads were shorter than the 493 bp promoter fragment cloned on the donor plasmid. However, no sequence data was identified containing parts of both the TDH3 gene and any of the downstream homology regions, nor parts of both eGFP and the region downstream of the TDH3 terminator in the wild type genome. This indicates that insertions into the 202 bp long TDH3 terminator had not occurred.

Escherichia coli NEB 10-beta Competent E. coli (New England Biolabs) were used for plasmid construction. E. coli was cultured in tryptone soya broth (TSB; Oxoid), plated on 1.5% TSB agar (Oxoid), and grown at 37 °C. For selection in E. coli, 50 µg mL⁻¹ kanamycin sulfate, 100 µg mL⁻¹ ampicillin, or 25 µg mL⁻¹ chloramphenicol (all from Sigma-Aldrich) were used.

The K. phaffii strain used was ordered from ATCC through LGC Standards as Komagataella phaffii Kurtzman (ATCC 76273, NRRL Y-11430, CBS 7435). K. phaffii was routinely cultured in YPD broth (BD Difco from Fisher Scientific) or on 1.5% YPD agar plates (Oxoid). Agar plates were incubated at 30 °C. Submerged cultures were grown in 500 mL baffled Erlenmeyer flasks (Bellco Biotechnology). Geneticin (G418; Sigma-Aldrich) at 500 μg mL⁻¹ was used for selection of CRISPi plasmids.

The DM1-2 defined medium, used for kinetic growth assays, is a modification of the DM1 medium [74] containing 1.7 g L⁻¹ Yeast Nitrogen Base, 7 g L⁻¹ urea, 10 g L⁻¹ KH₂PO₄ pH 6.5, 22 g L⁻¹ D(+)-Glucose·H₂O (all from Sigma-Aldrich), 1× Gibco MEM amino acids solution, 1× Gibco Cholesterol Lipid Concentrate (both from ThermoFisher Scientific) and 100 µL L⁻¹ PTM1s trace mineral solution. PTM1s was prepared by dissolving 0.2 g L⁻¹ biotin, 0.02 g L⁻¹ boric acid, 0.92 g L⁻¹ CoCl₂·6H₂O, 6 g L⁻¹ CuSO₄·5H₂O, 65 g L⁻¹ FeSO₄·7H₂O, 3 g L⁻¹ MnSO₄·H₂O, 0.08 g L⁻¹ NaI, 0.2 g L⁻¹ Na₂MoO₄·2H₂O, 20 g L⁻¹ ZnCl₂, 5 mL L⁻¹ H₂SO₄ (all from Sigma-Aldrich) in milliQ-H₂O and subsequent sterile filtration.

Primers, genes, and gene strands were ordered from Eurofins Genomics. Restriction enzymes and ATP were from New England Biolabs. Plasmids were isolated using the QIAprep Spin Miniprep kit (Qiagen). Templates for PCR of E. coli clones were generated by transferring small samples from single colonies to PCR tubes followed by heating for 1 min at 800 W in a microwave oven. Templates for PCR of K. phaffii clones were generated by resuspending small samples of single colonies in 20 µL 0.1 M NaOH, incubating for 25 min at 99 °C in a ThermoCycler, and using 1 µL in each PCR reaction.

Routine verification of plasmids and strains was performed by PCR of the targeted loci using Platinum HotStart PCR Master Mix (2X) (Invitrogen) and 0.2 µM each of forward and reverse primers. PCR primers with annealing sites located outside of the integration sites on plasmids are listed in Supplementary Table S8. Sequencing of plasmid constructs was performed using Sanger sequencing with the BigDye Terminator v3.1 Cycle Sequencing Kit and the Genetic Analyzer 3500 instrument (both Applied Biosystems). PCR screening of OVA transformants and eGFP expression clones subjected to WGS was performed with primers covering the 5′ site of the targeted sequence to detect transformants and primers covering the entire locus (binding outside of the 5′ and 3′ homologous regions) to identify wild-type clones. The PCR primers are listed in Supplementary Table S8.

Plasmids utilized in the present work were constructed by means of Golden Gate assembly. The Type IIS restriction endonuclease BbsI, which has identical recognition and cleavage specificities as BpiI, was used in the current study to generate ssDNA sticky ends. BbsI Golden Gate assembly reactions were performed using 0.5 µL BbsI-HF (10 U), 0.1 µL T4 ligase (40 U), 2 µL rCutSmart Buffer (10×), 2 µL ATP (10 mM) in 20 µL volumes. BsaI Golden Gate assembly reactions were similarly performed using 1 µL NEB Golden Gate Assembly Mix (BsaI-HFv2) and 2 µL T4 DNA Ligase Buffer (10×). Reactions used ~50 fmol vector and ~100 fmol of each insert and were incubated for 30 cycles of 5 min each at 37 °C and 16 °C, followed by heat inactivation for 6 min at 60 °C, except for construction of the ovalbumin expression cassette, where ~10 fmol vector and ~50 fmol of each insert was used, and the reaction was incubated for 100 cycles of 5 min at 37 °C and 3 min at 16 °C. Further details on the specific Golden Gate assembly reactions used to construct the plasmids employed in the present work can be found in subsection “Plasmid construction” below.

Plasmids containing donor cassettes were linearized prior to transformation into K. phaffii by digestion of 5 µg plasmid with 5 µL EcoRI-HF in 250 µL reaction volumes. Complete digestion was confirmed by agarose gel electrophoresis, the DNA was concentrated by ethanol precipitation with 0.1× volumes of 3 M Na-acetate (pH 4.8–5.2), dissolved in 10 µL dH₂O, and the DNA concentration determined using NanoDrop.

Electrocompetent K. phaffii cells were prepared essentially as described [39]. Cultures were inoculated at 0.1% (v/v) from glycerol stocks (15% glycerol) prepared from exponential phase cultures and grown at 28 °C with shaking at 120 rpm to optical density at 600 nm (OD600) between 1.2 and 1.5 (~10⁷ cells mL⁻¹). Cells were centrifuged at 500 g for 5 min at room temperature, gently resuspended in 25 mL freshly made room-temperature sterile-filtered resuspension buffer (100 mM lithium acetate, 10 mM dithiothreitol, 0.6 M sorbitol, and 10 mM Tris–HCl, pH 7.5), and incubated at 20 °C with gentle mixing at 100 rpm for 20–30 min. Cells were centrifuged as before, gently resuspended in 750 µL ice-cold 1 M sorbitol, and transferred to 2-mL microtubes placed on ice. They were then centrifuged at 4 °C and washed twice with 750 µL ice-cold 1 M sorbitol, before final addition of 50 µL ice-cold 1 M sorbitol and gentle resuspension. Aliquots of 80 µL, kept on ice, were used the same day.

Electrocompetent K. phaffii cells were co-transformed with 100 ng CRISPi plasmid and linearized crBB3 donor plasmid, corresponding to 1.0 or 1.1 µg donor cassette DNA. Cells were electroporated in 0.2-cm gap cuvettes (Bio-Rad) at 1.5 kV, 25 µF, and 200 Ω in a MicroPulser (Bio-Rad). 1 mL ice-cold YPD medium diluted 1:1 in 1 M sorbitol was added to the cells immediately after electroporation. Cells were allowed to regenerate for 1–3 h at 28 °C with gentle shaking at 100 rpm before selecting on YPD agar containing G418 to select for transformants harbouring CRISPi plasmids. At each time-point, 300 µL was plated on a total of six agar plates. Transformants were counted after 4 to 7 days of incubation. The CRISPi plasmid (containing ARS replication region) was cured by growth on non-selective media. For transformation with the eGFP donor cassettes, three independent experiments (biological replicates) were performed.

For the endpoint fluorescence screening assay, up to 100 single colonies from each eGFP cassette transformation reaction and plating time-point (picked at random) were used to inoculate 200 µL of YPD cultures in black clear bottom 96-well plates (Greiner Bio-One) with lids. For each reaction and time-point, all clones were picked if fewer than 100 colonies were present; otherwise, 100 clones were selected. Plates were incubated with shaking at 28 °C and 180 rpm for 48 h. Measurements of OD600 (to confirm growth) and fluorescence at 484 nm excitation and 510 nm emission, read from the plate bottom, were obtained in a Synergy H1M instrument (BioTek Instruments / Agilent Technologies).

Targeting efficiency was calculated as the percentage of fluorescent clones above a threshold set by the readings obtained for negative controls (relative fluorescence units [RFU] above 600, obtained with gain set to 50). For evaluation of normalized fluorescence (RFU OD600⁻¹), only clones with OD600 > 1 were included.

For the kinetic fluorescence growth assay, seed cultures were prepared by inoculating single colonies from YPD agar plates containing G418 into 5 mL YPD and incubating overnight at 28 °C and 200 rpm. Overnight seed cultures were diluted to 0.5% (v/v) in 200 µL DM1-2 defined medium in 96-well black clear bottom plates (Greiner Bio-One) with lids. The side of plates with lids was covered with parafilm to minimize volume losses due to evaporation and incubated in the Synergy H1M instrument at 28 °C, with slow-speed 1 mm-amplitude orbital mode mixing (10 Hz) at 559 cpm. OD600 and RFU were measured every 5 min as described above. Two to four biological replicates were used for each transformant. Growth and fluorescence data were represented using the python libraries Matplotlib v3.9.3 [77], NumPy v2.2.0 [78] and Pandas v2.2.3 [79].

Statistical differences in transformation and integration efficiencies were evaluated using Minitab v.22 software and three-way analysis of variance (ANOVA) with integration site, sampling time-point, and biological replicate as factors. The tested null hypotheses were no differences between integration sites and no differences between time-points. After rejection of a null hypothesis (p < 0.05), Tukey’s post hoc test for pairwise comparisons was performed to compare differences between groups. The full results for this statistical analysis is shown in Supplementary Tables S13–S16.

Statistical analyses for the kinetic fluorescence growth assay were performed in GraphPad Prism 10.3.1. A significance level of p = 0.05 was employed to evaluate differences in group means. When the data had normal distribution and homoscedastic variance, ordinary one-way ANOVA was applied, followed by Dunnett’s test. Kruskal–Wallis test (one-way ANOVA on ranks) followed by Dunnett’s T3 test was used when the data was not normally distributed. The full results for the ANOVA analysis is shown in Supplementary Tables S17–S19.

For production of recombinant ovalbumin in K. phaffii, overnight 50 mL YPD seed cultures from single colonies were used to inoculate 100 mL of DM1-2 defined medium in 500 mL baffled Erlenmeyer flasks covered with cotton plugs, to obtain a starting OD600 of 0.005. Cultures were grown at 25 °C and 200 rpm for 48 h before supernatants were harvested by centrifugation at 17000 g for 4 min at 4 °C and snap-frozen on liquid N₂.

Supernatant samples or standard protein samples (albumin from chicken egg white; Sigma-Aldrich) were mixed with Invitrogen Bolt LDS Sample buffer and 100 mM DTT, and heated at 70 °C for 10 min before loading onto SDS-PAGE gels; 12-well Invitrogen Bolt Bis–Tris Plus mini protein gels, 4–12%, 1.0 mm, WedgeWell format. Invitrogen BenchMark pre-stained protein ladder (10 µL; ThermoFisher Scientific) and Amersham ECL Plex fluorescent Rainbow marker (5 µL; Cytiva) were used as molecular weight markers for total protein staining and Western blot, respectively. Gels were run in Invitrogen Bolt MOPS SDS running buffer for 40 min at 200 V. For total protein staining, gels were washed for 3 × 5 min with milliQ-H₂O, incubated for 1 h with Invitrogen SimplyBlue SafeStain (all from ThermoFisher Scientific) and washed with milliQ-H₂O overnight before imaging in a Perfection 4990 Photo scanner (Epson).

For western blotting, SDS-PAGE gels were blotted onto iBlot 3 nitrocellulose membranes using the Invitrogen iBlot 3 transfer system (both from ThermoFisher Scientific). After transfer, membranes were blocked in TBS-Tween with 2% ECL Prime blocking agent (Cytiva) for 1 h, rinsed twice with TBS-Tween and then incubated overnight at 4 °C with primary antibody (rabbit anti-OVAL antibody; Sigma-Aldrich) diluted 1:2000 in TBS-Tween with 0.2% ECL Prime blocking agent. Membranes were subsequently washed three times for 5 min with TBS-Tween, incubated for 1 h at room temperature with secondary antibody (Amersham ECL Plex Cy5 conjugated goat-anti-mouse IgG; Cytiva) diluted 1:2000 in TBS-Tween with 0.2% ECL Prime blocking agent, and finally washed three times for 5 min with TBS-Tween. Membranes were air-dried before protein bands were visualized in a G-Box gel-doc system (Syngene). Densitometric analysis of the images was performed with the ImageQuantTL software (Cytiva).

For preparation of genomic DNA for Illumina sequencing, cells were lysed using Lysing Matrix B and a FastPrep instrument (both MP Biomedicals) and DNA isolated using the DNeasy Blood and Tissue Kit (Qiagen). Libraries for genome sequencing were prepared using the Nextera XT DNA Sample Preparation Kit (Illumina) and sequenced using paired-end (PE) 2× 300 bp reads on a MiSeq instrument (Illumina).

The reference genomes used for mapping were the four chromosomes contained in the K. phaffii reference genome sequence [36] (Komagataella phaffii CBS 7435 NCBI GenBank assembly GCA_900235035.2), manually modified to contain the expected insertions of each of the three eGFP expression cassettes, in addition to the sequences of the K. phaffii CBS 7435 mitochondrion (GenBank accession NC_015384.1↗) and the two killer plasmids (GenBank accessions MG491503.1↗ [13.1 kb plasmid] and MG491504.1↗ [9.5 kb plasmid]).

Raw reads were filtered on q15 and trimmed of adaptors before they were mapped to the modified versions of the K. phaffii reference genome (described above) and to the sequences of the crBB3_04576_eGFP, crBB3_PFK1_eGFP, and crBB3_ROX1_eGFP donor cassette plasmids (Table 1) using Burrows–Wheeler Aligner (BWA-MEM) v0.7.15 [80]. The alignment was converted to BAM format, sorted, indexed, and duplicates were marked using SAMtools v1.10 [81] and Picard v2.22.3 [82].

The percentage of mapped reads was determined using SAMtools flagstat. The mean read depth coverage of the entire K. phaffii reference genome and the number of positions on each reference with zero coverage was calculated using Picard CollectWgsMetrics. The read depth coverage at each position in each reference was obtained by SAMtools depth (with option -a). Normalized depth of coverage at each position in each reference was obtained by dividing the read depth coverage at each position by the mean read depth coverage for the entire K. phaffii reference genome. ‘Mean normalized coverage’ was calculated as the mean normalized depth of coverage for a specific gene or genetic region. ‘Coverage relative to eGFP’ was calculated by dividing the mean normalized coverage of a specific gene or DNA fragment by the mean normalized coverage of the eGFP gene in the low fluorescence single gene insertion clone, for the corresponding target site (Supplementary Tables S2 to S7).

Genome assembly was performed with SPAdes v3.13.0 [83] and six k-mer sizes (21, 33, 55, 77, 99, and 127). Sequences from the expected integration regions, the region surrounding the TDH3 gene, and the crBB3 donor cassette plasmids were used as queries in BLAST (blastn v2.12.0+) to search for contigs with alternative integration sites or multiple insertions.

The authors declare no competing interests.

Maps and GenBank sequences of key plasmids generated in this study are provided in the Supplementary Information files. The plasmids obtained during the study are available from the corresponding author on reasonable request. The raw Illumina sequencing reads have been deposited to the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) under BioProject number PRJNA1188001.