In our previous study, we analyzed a library of 845 P. pastoris clones transformed with an integrative GFP expression cassette for their productivity characteristics (Schwarzhans et al., 2016a). Based on the assayed features, interesting clones were selected for genome sequencing. Some strains were selected, because they displayed GFP expression levels that far exceeded the one predicted based on their gene copy number (GCN). One of these was strain JPS664 (EMBL FBUC01000000). With a GCN = 1 a normalized GFP expression level of ca. 1 was to be expected. However, a normalized expression level of 2.4 ± 0.3 was found. Consequently, JPS664 was selected for genome sequencing. After sequencing, it was revealed that JPS664 did not contain an expression cassette in its chromosome. Rather, a ligation of the GFP cassette to a 1.4 kb fragment of mitochondrial DNA was found. As a result of the fusion, a circular 7.3 kb plasmid was formed that was named “pMito” (Figure 1, EMBL LT724168). Due to the plasmid character of pMito, an ARS was suspected to be encoded on the mtDNA fragment. The mitochondrial DNA shows 100% identity to the bases 27,552–28,993 of the mitochondrial genome of P. pastoris CBS7435 containing a fragment of the COX1 (cytochrome c oxidase I) gene. The last 73 bp of the second COX1 exon as well as 1369 bp of the second COX1 intron were detected on the segment. The combination of a novel mtDNA in vivo ligation to an expression cassette in P. pastoris, potential ARS activity, and the apparently high suitability of the plasmid for recombinant protein production prompted us to conduct further experiments on pMito.

First, the plasmid character of pMito had to be validated in order to ensure that it was not an artifact based on genome sequencing and assembly. To this end, a PCR assay with two sets of directly adjacent, diverging primer pairs was designed. This way, a PCR product of the same size as the predicted pMito would validate its circular structure. One primer pair binds in the HIS4 region of pMito and the other in the mtDNA fragment. In Figure 2A the PCR assay visibly indicates the presence of a full length circular plasmid as shown in Figure 1. In addition, the plasmid stability of pMito was investigated (Figure 2B). Under selective conditions pMito is well-maintained at up to 96.2 ± 2.7%, while under non-selective conditions the plasmid stability was lowest at 20.2 ± 1.4%. In combination with the specific growth rate, a growth associated loss and increase of pMito content is apparent, respectively. On the one hand, pMito content rose, while cells grew in selective media and stagnated or fell when they reached the stationary phase, indicating that cells containing pMito lost their growth advantage over plasmid-free cells. On the other hand, cells grown in non-selective media quickly lost pMito during the exponential growth phase and exhibited constant plasmid contents in the stationary phase, indicating a growth advantage of plasmid-free cells. Taken together, the plasmid stability and PCR assay clearly validated the circular plasmid character of pMito.

It was therefore highly likely, that the mtDNA fragment in pMito conferred ARS activity. Using the Find Individual Motif Occurrences (FIMO) tool (Grant et al., 2011), the mtDNA fragment was scanned for known ARS consensus sequence (ACS) motifs. Both AT- and GC-rich ACS motifs from P. pastoris as well as the 11 and the 17 bp ACS motifs of S. cerevisiae were included (Newlon and Theis, 1993; Theis and Newlon, 1997; Liachko et al., 2014). Despite the low GC-content (22%) of the mtDNA fragment, one putative GC-ACS was found in addition to several AT-ACS and S. cerevisiae ACS with a total of 33 matches (Table 1). A more detailed description of the FIMO matches can be found in Supplementary Table S1. By mapping the FIMO matches to the sequence of the mtDNA fragment, multiple clusters of predicted ARS become visible (Figure 3). All putative ARS were found in the COX1 intron on the mtDNA. Especially in the regions at 1256–1300 bp and 701–753 bp matching sequences were found, with up to 6 and 4 cumulative matches, respectively. Here, similar sequences to the P. pastoris AT-ACS, S. cerevisiae 11 bp ACS and S. cerevisiae 17 bp ACS were detected. The accumulation of different types of ACS indicates a high likelihood of actual ARS functionality in these regions. Additionally, between 2 and 3 cumulative matches can be found in the regions of 130–184, 347–363, and 1043–1083 bp. In these regions as well, ARS activity could be present. Although P. pastoris ACS motif matches were found, no region of the mtDNA fragment could be directly aligned via BLASTn to the library of (putative) ARS sequences on the P. pastoris GS115 chromosomes (Liachko et al., 2014). This indicates that the ARS on pMito, while similar to the ARS found on the chromosomes, has its own distinguishable sequence. A preliminary FIMO scan of the mitochondrial genome of P. pastoris CBS7435 (GenBank: FR839632) identified 20 GC-ACS and over 500 AT-ACS motif occurrences (data not shown).

The expression cassette portion of pMito, excluding the mtDNA fragment, was also scanned for ACS motif occurrences, as described above. In spite of being approximately four times as long as the mtDNA fragment, only five matches were found. They are detailed in Supplementary Table S2 and the mapping result is shown in Supplementary Figure S1. Interestingly, three of these matches accumulate within a 43 bp region of the AOX1 terminator and two occur in the 3′ UTR AOX1 element. Therefore, we conducted transformation experiments with different variants of the plasmid.

Using the original pMito from JPS664 as template, four different variants were created via PCR (Figure 4). They differed in two aspects. Firstly, the constructs pMito-I and pMito-II encompass the complete plasmid, while pMito-ΔUTR (untranslated region) omits the 3′ UTR AOX1 region and pMito-ΔGOI (gene of interest) does not contain the AOX1 promoter, gfp gene and AOX1 terminator. Secondly, different loci were used for linearization. pMito-ΔUTR and pMito-ΔGOI were linearized by removing the aforementioned segments. pMito-I was linearized inside the mtDNA fragment and pMito-II inside the HIS4 gene. In accordance with our previous recommendations, P. pastoris was transformed with linear, PCR amplified DNA (Schwarzhans et al., 2016b). This method also offered the advantage to assay P. pastoris capability for in vivo circularization of the transformed DNA. In the case of pMito-II, only a successfully circularized plasmid could bestow histidine prototrophy, since the HIS4 gene was split onto the distal ends of the transformed, linear DNA.

All four constructs produced transformants (Figure 5A). Using the PCR assay described in the previous chapter, the presence of full-length circular plasmids could be confirmed in strains of all four constructs (Figure 5B). The successful transformation of pMito-II confirmed that P. pastoris was capable of in vivo ligating the linear DNA into circular plasmids. Depending on the aim of an experiment, this ability can be used, e.g., for DNA assembly purposes. Transformation of pMito-I, pMito-II, and pMito-ΔUTR resulted in comparable efficiencies, with an average of about 190 colony forming units (cfu)/μg. In contrast, an efficiency more than 10-fold higher at ca. 3000 cfu/μg was encountered using pMito-ΔGOI.

Since the original strain JPS664 in which pMito was found exhibited favorable productivity characteristics, it was of interest to investigate if these properties could be replicated. To this end, 16 clones transformed with pMito-I were randomly selected and submitted to further analysis. The productivity of the selected clones was assayed in relation to the original JPS664 and the reference strain JPS066, which contained a single expression cassette integrated into the AOX1 locus on the genome (Schwarzhans et al., 2016a). All selected pMito-I clones produced high amounts of GFP, markedly exceeding the reference strain at least two- and up to fourfold (Figure 6A). Applying the Student’s t-test (two-sided), it was determined that all pMito clones, and the original strain JPS664, produced GFP at significantly higher levels than JPS066, with the vast majority of pMito-I strains scoring a p-value < 0.01. A highly uniform expression level of pMito-I strains is visible, with all but one clone (pMito-I 2) exhibiting normalized expression levels in the range of 2.5–3.9. It seemed, that from the normalized expression levels an upper limit of GFP productivity can be deduced. To investigate whether the gene dosage could explain this behavior all clones were subjected to GCN analysis. As can be seen in Figure 6B, expression level and GCN did not correlate. While GCN values between 0.6 and 5.5 were encountered, the gene dosage of a strain had apparently no impact on the expression level.

Not only the expression level but also the localization of the product is of interest. For correct post-translational modification and secretion of the recombinant protein, it must first be present in the cytoplasm, so it can translocate to the endoplasmic reticulum and Golgi apparatus. A post-translational translocation of GFP to the mitochondria was unlikely due to the absence of a corresponding signal sequence. However, it was possible that copies of pMito were present in the mitochondria, especially for JPS664. Here, the exact origin of pMito was uncertain and could result in GFP accumulation in the mitochondria, complicating product capture. Living cells of the P. pastoris CBS7435 wild type (WT), JPS664, JPS066 and strain pMito-I 10 were analyzed via super-resolution fluorescence microscopy. In all GFP expressing strains the recombinant protein was present in the entire cytoplasm (Figure 7). Two major compartments absent of GFP fluorescence are distinguishable. They are likely the peroxisome and the mitochondria. This and more findings are analyzed in more detail in Section “Discussion”.

As previously detailed, ACS motifs from S. cerevisiae could be matched to the mtDNA segment of pMito. Therefore, we were interested to see if pMito could also be used as an episomal vector in S. cerevisiae. A variant of the plasmid pYES2 was constructed, replacing the 2 μm origin with the mtDNA fragment of pMito, resulting in pYES2-Mito. For the purpose of better comparison with the Pichia experiments, S. cerevisiae was transformed with linear, PCR amplified constructs. Both constructs facilitated creation of uracil prototroph S. cerevisiae clones (Figure 8A) and the presence of circular plasmids was confirmed via PCR (Figure 8B). Surprisingly, the transformation efficiency of pYES2-Mito was almost sixfold higher as for pYES2.

All plasmids were constructed and propagated in E. coli Top 10 [Invitrogen, USA; genotype: F- mcrA Δ(mrr-hsdRMS-mcrBC) Φ80lacZΔM15 ΔlacX74 recA1 araD139 Δ(araleu)7697 galU galK rpsL (StrR) endA1 nupG]. E. coli cultivations were carried out in LB (Lysogeny Broth) medium supplemented with 100 μg/mL ampicillin. Experiments involving P. pastoris employed the WT strain CBS7435 (identical to NRRL Y-11430 and ATCC76273), obtained from the Spanish Type Culture Collection (Valencia, Spain) under the designator CECT11047, as well as the histidine-auxotrophic strain P. pastoris CBS7435 (ΔHIS4) from the Austrian Center of Industrial Biotechnology (Graz, Austria). S. cerevisiae INVSc1 (genotype: MATa HIS3D1 LEU2 TRP1-289 URA3-52 MAT HIS3D1 LEU2 TRP1-289 URA3-52) from Invitrogen (Waltham, MA, USA) was used for Saccharomyces experiments. Shake flask cultivations of both P. pastoris and S. cerevisiae were performed using YPD (Yeast Peptone Dextrose) medium, as well as BMD (Buffered Minimal Dextrose) medium (Invitrogen, 2010) for P. pastoris and SC minimal medium (Invitrogen, 2008) for S. cerevisiae. If necessary, BMD medium was supplemented with 4 mg/L L-histidine, and SC medium was supplemented with 0.05 g/L L-histidine, 0.1 g/L L-leucine, and 0.1 g/L L-tryptophan. P. pastoris expression levels were assayed in 96-deep-well plates with 2.4 mL total volume (Eppendorf, Hamburg, Germany) according to Weis et al. (2004) and Hartner et al. (2008), using BMD, BMM2 (Buffered Minimal Methanol) and BMM10 medium. By applying this method, the cells are first grown in BMD and expression is induced with BMM2 and BMM10 by maintaining a 0.5% (v/v) methanol concentration in the culture medium. All yeast cultivations were carried out at 28°C, with shake flasks (baffled) being agitated at 120 min^-1 and deep-well plates at 340 min^-1.

Genomic DNA (gDNA) was isolated from yeast cultures using the MasterPure^TM Yeast DNA Purification kit (Epicentre, Madison, WI, USA). The method used for genome sequencing of relevant P. pastoris strains was recently described (Schwarzhans et al., 2016a). In short, gDNA quality was assayed via gel-electrophoresis and gDNA of sufficient quality was quantified using the Quant-iT PicoGreen dsDNA kit by Invitrogen (Waltham, MA, USA). From samples of high quality and quantity, paired-end libraries were prepared by applying the TruSeq sample preparation kit (Illumina, San Diego, CA, USA). The libraries were sequenced on an Illumina MiSeq system. Raw data was de novo assembled using the GS De Novo Assembler (Version 2.8, Roche, Basel, Switzerland) with default settings. The assembled draft genome of P. pastoris JPS664 can be found under FBUC01000000, and the finalized sequence of pMito under LT724168 in the EBI database.

For bioinformatic analysis the BLASTn algorithm (Altschul et al., 1997) and a local database including the pAHBgl-GFP vector sequence were used for database comparison. Only hits with a sequence identity of 100% and an e-value > 1 × 10^-20 were further analyzed in more detail. If necessary, gaps in the vector were closed via an in silico approach with CONSED (Gordon et al., 1998; Küberl et al., 2011; Wibberg et al., 2011). This approach allowed to determine the exact locus for the expression cassette in the sequenced P. pastoris genome.

Primers were designed in SnapGene (GSL Biotech, Chicago, IL, USA) and their sequences can be found in Supplementary Table S3. PCR procedures employed the Phusion^® High-Fidelity DNA Polymerase (New England Biolabs, Ipswich, MA, USA). For agarose gel electrophoresis analysis, the GeneRuler DNA Ladder Mix by Thermo Scientific (Waltham, MA, USA) was used as size marker. The P. pastoris strain JPS664 was created as described in Schwarzhans et al. (2016a) with a BglII digested pAHBgl-GFP plasmid. pMito and its variants were amplified via PCR using the original pMito as template. Potentially due to secondary structures, the variants pMito-ΔUTR and pMito-ΔGOI consistently resisted PCR amplification. As a compromise, primers were designed that bind slightly inside the region targeted for omission. In consequence, pMito-ΔUTR and pMito-ΔGOI still contain 63 and 65 bp of the targeted region, respectively. pYES2 was obtained from Invitrogen (Waltham, MA, USA). It can be used for the transformation of S. cerevisiae with an episomal vector containing a 2 μm sequence for replication and an URA3 selection marker (Invitrogen, 2008). For the construction of pYES2-Mito the mtDNA fragment on pMito was PCR amplified and combined via Gibson Assembly (Gibson et al., 2009) with the linearized pYES2 without 2 μm sequence.

Pichia pastoris CBS7435 (ΔHIS4) was transformed according to Wu and Letchworth (2004) and S. cerevisiae INVSc1 according to Thompson et al. (1998) using PCR amplified, linear DNA. Per transformation approximately 500 ng of purified (Wizard^® Plus SV Minipreps DNA Purification System, Promega, Madison, WI, USA) DNA was used. After transformation the Pichia cells were immediately spread onto MD (Minimal Dextrose) plates (Invitrogen, 2010), and Saccharomyces cells onto SC plates without uracil (Invitrogen, 2008) in 200 μL aliquots and incubated for 3 days at 28°C. Following the incubation, the total number of transformants was counted and randomly selected clones picked for dilution plating. Single colonies from dilution plating were used for following experiments, including PCR assays for the correct construct, expression screenings and GCN determination. Experiments for assaying the transformation efficiency were carried out in biological triplicates.

The mtDNA fragment on pMito was scanned for ACS motifs via the FIMO tool (Grant et al., 2011) of the MEME suite (Bailey et al., 2009). Both the 11 and 17 bp ACS motifs of S. cerevisiae (Newlon and Theis, 1993; Theis and Newlon, 1997), as well as the GC-ACS and AT-ACS motifs of P. pastoris (Liachko et al., 2014) were used as references.

Precultures of P. pastoris JPS664 were grown overnight under selective conditions in MD medium without L-histidine. On the following day the main cultures were inoculated to OD 0.2. For selective conditions MD medium without L-histidine and for non-selective conditions YPD medium were used, respectively. Samples for OD measurement and the plasmid stability assay were taken every 2 h until 16 h of cultivation and again after 26, 36, and 60 h. OD values were used to calculate the specific growth rate at each time point. Before plating onto YPD plates, the samples for the plasmid stability assay were diluted based on the OD so that about 100–1000 colonies were to be expected per plate (Invitrogen, 2010). After 2 days of incubation, 52 colonies were picked per sampling point and individually washed twice with 200 μL of 9 g/L NaCl. Washed cells were resuspended in 20 μL of 9 g/L NaCl and pipetted onto MD plates without L-histidine. Following 2–3 days of incubation the colonies were counted in order to determine the plasmid stability. The plasmid stability assay was carried out in biological duplicates with technical triplicates each.

The procedures employed for assaying the GFP expression level and GCN were described previously in more detail (Schwarzhans et al., 2016a). In brief, a reference strain containing a single copy of the GFP expression cassette in the AOX1 locus was used for normalization of the GFP/OD expression level. Strains were grown in 96 deep-well plates as described above and values 60 h after the start of the methanol induction are presented in this study. All strains were cultivated in biological triplicates with technical triplicates each. The GCN was determined based on the protocol by Abad et al. (2010) via the 2^-ΔΔCt method (Livak and Schmittgen, 2001), with ARG4 as the calibrator gene. Biological triplicates with technical duplicates each were used in these procedures.

Mitochondria were stained with MitoTracker^® Red CM-H₂XRos (Thermo Scientific, Waltham, MA, USA). The staining procedure was performed as described by Farre et al. (2007), using P. pastoris cells that had been induced with 0.5% (v/v) methanol for 3 days. Per experiment 3 μL of cells were mounted between a microscope slide and a high precision #1.5 coverglass (Marienfeld-Superior, Germany). Images were acquired on a DeltaVision OMX V4 system from GE Healthcare (United Kingdom) with a 60x 1.42 NA oil immersion PlanApoN objective (Olympus, Japan) and sCMOS camera. This setup applies the principal of three dimensional super-resolved structured illumination microscopy (3D SR-SIM), gaining a twofold resolution increase compared to conventional fluorescence microscopy. GFPuv was excited at 488 nm and the emission recorded at 504–552 nm. For MitoTracker^® Red CM-H₂-XRos an excitation wavelength of 568 nm and emission wavelength band of 590–627 nm were employed. Multiple z planes encompassing the entire cell from top to bottom were recorded at a distance of 125 nm. For reference, differential interference contrast (DIC) images were recorded. Super-resolved fluorescent images were reconstructed with the corresponding recorded optical transfer function (OTF) in the softWoRx 6.5.2 software (GE Healthcare, United Kingdom) at a Wiener filter setting of 0.006. The GFP background was adjusted by subtracting the value of the WT, non-expressing strain. Maximum intensities were individually adjusted for optimal representation. The raw data, OTFs and calibration settings can be provided upon request.