GoldenPiCS: a Golden Gate-derived modular cloning system for applied synthetic biology in the yeast Pichia pastoris

Background State-of-the-art strain engineering techniques for the host Pichia pastoris (syn. Komagataella spp.) include overexpression of homologous and heterologous genes, and deletion of host genes. For metabolic and cell engineering purposes the simultaneous overexpression of more than one gene would often be required. Very recently, Golden Gate based libraries were adapted to optimize single expression cassettes for recombinant proteins in P. pastoris. However, an efficient toolbox allowing the overexpression of multiple genes at once was not available for P. pastoris. Methods With the GoldenPiCS system, we provide a flexible modular system for advanced strain engineering in P. pastoris based on Golden Gate cloning. For this purpose, we established a wide variety of standardized genetic parts (20 promoters of different strength, 10 transcription terminators, 4 genome integration loci, 4 resistance marker cassettes). Results All genetic parts were characterized based on their expression strength measured by eGFP as reporter in up to four production-relevant conditions. The promoters, which are either constitutive or regulatable, cover a broad range of expression strengths in their active conditions (2–192% of the glyceraldehyde-3-phosphate dehydrogenase promoter P GAP), while all transcription terminators and genome integration loci led to equally high expression strength. These modular genetic parts can be readily combined in versatile order, as exemplified for the simultaneous expression of Cas9 and one or more guide-RNA expression units. Importantly, for constructing multigene constructs (vectors with more than two expression units) it is not only essential to balance the expression of the individual genes, but also to avoid repetitive homologous sequences which were otherwise shown to trigger “loop-out” of vector DNA from the P. pastoris genome. Conclusions GoldenPiCS, a modular Golden Gate-derived P. pastoris cloning system, is very flexible and efficient and can be used for strain engineering of P. pastoris to accomplish pathway expression, protein production or other applications where the integration of various DNA products is required. It allows for the assembly of up to eight expression units on one plasmid with the ability to use different characterized promoters and terminators for each expression unit. GoldenPiCS vectors are available at Addgene. Electronic supplementary material The online version of this article (10.1186/s12918-017-0492-3) contains supplementary material, which is available to authorized users.


Background
The yeast Pichia pastoris (syn. Komagataella spp.) is frequently applied for the production of heterologous proteins, most of which are efficiently secreted [1]. It is also favored for the production of membrane proteins [2], pharmaceuticals and chemical compounds [3] and as a model organism for biomedical research [4]. New genetic tools for P. pastoris such as promoters, signal peptides, selection markers, Flp-frt/Cre-lox recombination and CRISPR/Cas9 have been reviewed recently [3]. Compared to other yeast species, P. pastoris is distinguished by its methylotrophy, its Crabtree-negative metabolism, its growth to very high cell densities, the low number and concentration of secreted host cell proteins [5] and the availability of many genetic tools and industrially relevant strains (humanized N-glycosylation, protease deficiency) [6]. Genomic integration into specific loci is usually applied by using 5′ and 3′ homologous regions and is crucially depending on the avoidance of repetitive homologous regions and the use of well-purified vector DNA [7].
For recombinant protein production in P. pastoris, different highly efficient promoter systems were established (reviewed by Weinhandl et al. [8]) and applied to produce up to several grams per liter of secreted heterologous products, covering proteins intended for biopharmaceutical purposes as well as industrial enzymes [9]. The methanol utilization (MUT) pathway of P. pastoris is very efficient and the corresponding genes are highly induced on methanol [10,11]. MUT promoters (e.g. P AOX1 , P DAS1/2 and P FLD1 ) as well as strong constitutive promoters from highly expressed genes (such as P GAP , derived from the glyceraldehyde-3phosphate dehydrogenase gene TDH3 and the promoter of translation elongation factor P TEF1 ) are frequently applied [12]. Nevertheless, there is still room to further improve productivity and/or protein quality. Besides increasing transcriptional strength by using strong promoters and higher gene copy numbers of the expression cassettes, also cell engineering to boost the host's folding and secretion capacity or to provide precursors and energy for these processes proved to be beneficial for enhancing product titers (reviewed by Puxbaum et al. [1]). Sometimes such cell engineering approaches require the simultaneous overexpression of more than one gene to reach their full potential (e.g. Nocon et al. [13], Delic et al. [14]). On the other hand, gene knock-outs might be necessary to avoid detrimental processes such as transport to the vacuole or proteolysis (e.g. Idiris et al. [15]). Both of these genetic manipulations require extensive cloning and transformation efforts, which makes them rather time-consuming and tedious.
Today, advanced synthetic biology tools are applied in all fields of microbiology. New cloning methods such as Gateway®, Gibson Assembly and Golden Gate cloning, together with genome editing techniques like CRISPR/ Cas9 and TALEN (transcription activator-like effector nucleases), enable efficient and highly specific cell engineering and thereby revolutionized the whole field [16]. Golden Gate cloning is based on type IIs restriction enzymes (which are cutting outside of their recognition sequence) and offers important benefits: it does not require long flanking DNA, it uses efficient one-pot reactions, allows scar-less cloning and is cost-saving compared to many other advanced techniques [17]. In Golden Gate Assembly (GGA), the two different type IIs restriction endonucleases BsaI and BpiI are used which yield four base pair overhangs outside of their recognition sequence. These overhangs can be freely designed and are termed fusion sites (Fs). These fusion sites enable base pair precise assembly of genetic parts such as promoters, coding sequences (CDS) and transcription terminators. By using simultaneous restriction and ligation in efficient one-pot cloning reactions, rapid assembly of multiple DNA fragments is achieved [17].
Recently, Obst et al. [18] and Schreiber et al. [19] reported the use of Golden Gate cloning for the generation of libraries of expression cassettes in P. pastoris, which were tested for the production of reporter proteins by the assembly of standardized parts such as promoters, ribosome binding sites, secretion signals and terminators in a fast and efficient way. These studies aimed to optimize a single transcription unit for the production of one heterologous protein of interest (either a fluorescent reporter or an antimicrobial peptide). Vogl et al. [20] used Gibson assembly with a set of MUTrelated promoters and novel transcription terminators for the overexpression of multiple genes in P. pastoris and could show a strong effect of the inserted promoters when overexpressing the carotenoid pathway (crtE, crtB, crtI, and crtY). Our study extends the versatile Golden Gate technique for all applications in P. pastoris where the simultaneous integration of multiple DNA products is required (e.g. cell engineering, pathway expression, protein production, co-expression of cofactors) and aims beyond the mere assembly of single expression cassettes for the heterologous protein of interest.
For this purpose, we adapted the Golden Gate based modular cloning (MoClo) introduced by Weber et al. [21], to create the GoldenPiCS (Golden Gate derived P. pastoris cloning system) vector toolkit. GoldenPiCS is part of a universal system termed GoldenMOCS, standing for Golden Gate-derived Multiple Organism Cloning System [22]. The GoldenMOCS platform enables versatile integration of host specific parts such as promoters, terminators, and resistance cassettes, origins of replication or genome integration loci to adapt the plasmid to the needs of the experiment and the host cell to be engineered. Here, we present the GoldenMOCS-subsystem GoldenPiCS designed for use in P. pastoris and the characterization of its individual genetic parts using eGFP (enhanced green fluorescence protein) as a reporter. Vectors of these systems were deposited at Addgene as GoldenPiCS kit (#1000000133).

Results and discussion
GoldenPiCS, our Golden Gate-derived P. pastoris cloning system, consists of three hierarchical backbone (BB) levels for flexible generation of overexpression plasmids containing multiple transcription units (up to eight per plasmid), four different selection markers and five loci either for targeted genome integration or episomal plasmid maintenance (Fig. 1). We mainly designed the system to enable advanced cell engineering or the expression of whole metabolic pathways. In the lowest cloning level, individual parts such as promoters, coding sequences (e.g. reporters or GOIs) and transcription Fig. 1 Assembly strategy and hierarchical backbone levels of the cloning systems GoldenMOCS and GoldenPiCS. In the microorganism-independent general platform GoldenMOCS, DNA products (synthetic DNA, PCR products or oligonucleotides) are integrated into BB1 by a BsaI Golden Gate Assembly and fusion sites Fs1, Fs2, Fs3 and Fs4. Fusion sites are indicated as colored boxes with corresponding fusion site number or letter. Basic genetic elements contained in backbone 1 (BB1) can be assembled in recipient BB2 by performing a BpiI GGA reaction. The transcription units in BB2 are further used for BsaI assembly into multigene BB3 constructs. Single transcription units can be obtained by direct BpiI assembly into recipient BB3 with fusion sites Fs1-Fs4. Fusion sites determine module and transcription unit positions in assembled constructs. Thereby, fusion sites Fs1 to Fs4 are used to construct single expression cassettes in BB2 and are required between promoter (Fs1-Fs2), CDS (Fs2-Fs3) and terminator (Fs3-Fs4). Fusion sites FsA to FsI are designed to construct BB3 plasmids and separate the different expression cassettes from each other. The FSs are almost randomly chosen sequences and only FS2 has a special function, because it includes the start codon ATG. GoldenPiCS additionally includes module-containing BB1s specific for P. pastoris: 20 promoters, 1 reporter gene (eGFP) and 10 transcription terminators, and recipient BB3 vectors containing different integration loci for stable genome integration in P. pastoris and suitable resistance cassettes (Additional file 2) terminators are incorporated into backbone 1 (BB1) plasmids, which are subsequently assembled to one transcription unit in BB2. This is followed by the assembly of multiple expression units into one BB3, which is designed for subsequent genome integration in P. pastoris (four different selection markers and four loci for targeted genome integration are available). Contrary to other cloning techniques, Golden Gate Assembly avoids the need for excessive sequencing, because BB2 and BB3 constructs are assembled by ligation instead of overlapextension PCR (only BB1 inserts require sequencing after assembly). Correct ligation is assured by defined fusion sites (see Fig. 1): Fusion sites Fs1 to Fs4 are linked to individual parts by PCR and required for their assembly into a transcription unit (BB2) e.g. promoters with coding sequences ('CATG' , fusion site Fs2). Fusion sites FsA to FsI are used for the assembly of multiple expression units in BB3 e.g. for fusing the first transcription unit to the second ('CCGG' , fusion site FsB). The different BB3 vectors with fusion sites FsA-FsC, FsA-FsD, to FsA-FsI are designed for the assembly of two, three, … up to eight transcription units in one single plasmid. Internal BsaI and BpiI restriction sites must be removed in all modules by introducing point mutations, with consideration of the codon usage of P. pastoris.
Previously, strain engineering approaches with P. pastoris often relied on pGAPz, pPIC6 (Invitrogen) or related expression vectors, which harbor only one transcription unit (Fig. 2). Cloning of concatemers containing more than two transcription units proved to be highly time consuming and also led to unpredictable integration events when using repetitive promoter and terminator sequences [23,24]. Therefore, conventional overexpression of n genes requires n cycles of preparing competent cells and transforming them, and the use of n different selection markers. For overexpression of three factors plus screenings using consecutive transformations, the procedure would take at least 31 days (four days for transformation and re-streak, five days for screening and two days to prepare competent cells; Fig. 2, upper panel). Furthermore, selection markers can be removed and recycled, with the cost of an additional cycle of competent-making and transformation (Fig. 2, middle panel). In addition to that, the use of different integration loci must be included to prevent 'loop-out' incidents of integrated DNA. Alternatively, cotransformation of multiple vectors can be considered, but this requires the use of several independent selection markers, otherwise in our experience transformation efficiency is low and it is very unpredictable if all vectors get integrated into the genome [25]. Our backbone BB3 Golden Gate plasmids can carry multiple transcription units and hence significantly simplify and shorten the procedure to one single transformation step. Integration Fig. 2 Comparison of conventional and Golden Gate based strain engineering strategies for P. pastoris. Overexpression of multiple genes (GOIs) in P. pastoris using conventional cloning plasmids requires several rounds of competent making (2 days), transformation (4 days including second streak-out) and clone screening (5 days) and takes at least 31 days for three genes with three selection markers. Alternatively, multiple vectors can be co-transformed at once, but resulting transformation efficiencies are usually very low and clonal variation increases. Appropriate flanking sites for the selection marker (loxP or FRT sites) enable marker recycling by recombinases (Cre or Flp, respectively), but require one more round of competent making and transformation which takes at least eight additional days. Golden Gate plasmids carry several transcription units at once and thereby enable transformation and screening in only nine days of multiple transcription units plus screening lasts only nine days (Fig. 2, lower panel). The thereby generated strains benefit from decreased generation numbers and milder selection procedures.
Recombination events by repetitive sequences disturb full vector integration in P. pastoris Initially, we started with Golden Gate vectors carrying up to four transcription units with the same promoter and transcription terminator (P GAP and ScCYC1tt). High clonal variation prompted us to analyze gene copy numbers (GCN) of integrated genes and we found that individual transcription units were lost. Similar results have been obtained with P GAP or P AOX1 based multicopy vectors [24], thus showing that incomplete vector integration is not due to the Golden Gate backbone. We positively confirmed post-transformational integration stability for three of the transformants in three consecutive shake flask-batch cultivations without selection pressure (gene copy numbers were stable for more than 15 generations; Additional file 1: Table S1). Therefore, we conclude that repetitive homologous sequences within the expression vector (P GAP and ScCYC1tt sequences) resulted in recombination events (internal 'loop out') during transformation.
The high occurrence of incomplete vector integration prompted us to establish a collection of 20 different promoters and 10 transcription terminators (Table 1), in order to avoid repetitive sequences when creating constructs carrying multiple expression units. Promoters and terminators were selected based on the transcriptional regulation and expression strength of their natively controlled genes in published microarray experiments [26]. All of them were screened in several production-relevant conditions (Additional file 1: Table S2). In addition to established promoters [3,12], we selected novel yet uncharacterized promoter sequences based on their expression behavior in transcriptomics data from P. pastoris cells cultivated on different carbon sources [26]. By applying this collection we aimed to gain the ability to fine-tune the expression of integrated genes. Ideally, promoters for cell engineering purposes should cover a wide range of expression strengths and allow constitutive as well as tunable expression. Transcription terminators were selected from constitutively highly expressed genes [26], many of them being derived from ribosomal protein genes. Also ribosomal genes were reported to be regulated at the level of mRNA stability [27], which is one of the main functions of the 3'UTR contained in the transcription terminator fragments.
Validation of Kozak sequence mutations in the P GAP sequence of P. pastoris In the GoldenMOCS setup, fusion site Fs2 that links the promoter to the GOI contains the start codon ATG and part of the Kozak sequence, which is important for translational initiation in eukaryotes. As this fusion site is a fixed variable in the Golden Gate system, we evaluated the effect of the '-1' position of the P GAP promoter (position in front of the start codon, Fig. 3d) on reporter gene expression in P. pastoris. Due to the 'CATG' fusion site, the native ' A' in position '-1' of the P GAP promoter ('-8' to '-1': AAAACACA) is changed to 'C' (AAAA-CACC). While P GAP variants with ' A' , 'T' and 'C' at position '-1' performed similarly, the variant with 'G' resulted in a lower eGFP level (P GAP _GATG; about 40% lower compared to the other variants). The Kozak consensus sequence of P. pastoris was analyzed and found to be similar to that of S. cerevisiae, which is rich in ' A' and poor in 'G' bases (Fig. 4). Based on these results, eight bases of the A-rich Kozak consensus sequence were tested in an additional P GAP variant (P GAP _A 8 ATG; position '-4' and '-2' replaced by ' A': AAAAAAAA) and a slightly increased expression of eGFP was found (Fig. 3d). Nevertheless, we chose the 'CATG' fusion site for our GoldenPiCS system and kept the 'GCTT' fusion site for the GOI-terminator assembly.

Analysis of GoldenPiCS promoter and terminator strength and regulation using eGFP
All promoters and terminators were characterized for their capacity for expression of the intracellular reporter eGFP in appropriate conditions (Additional file 1: Table S2): glycerol or glucose excess ("G" and "D", respectively, maximum specific growth rate μ MAX~0 .22 h −1 ) as present in batch cultivation, limiting glucose ("X", 12 mm glucose feed beads, specific growth rate μ~0.04 h −1 ) and methanol feed ("M", μ MAX up to 0.1 h −1 ), the latter two representing conditions as encountered during fed batch cultivation. P GAP was used as reference to evaluate the expression strength of the promoters. P TEF2 , P GPM1 , P RPP1B , P PDC1 , P POR1 , P ADH2 , P FBA1-1 , P RPL2A , P LAT1 , P PFK300 and P MDH3 were confirmed to have a constitutive regulation with a range of eGFP expression of 2-192% of P GAP in all tested conditions ( Fig. 3a and b). Promoters responsive to thiamine (P THI11 ), glycerol (P GUT1 ) and methanol (P AOX1 , P DAS1 , P DAS2 , P FDH1 , P SHB17 and P ALD4 ) were well repressed (0% of P GAP ) and induced (to 18-160% of P GAP ) in the repressed and induced conditions, respectively (Table 1 and Fig. 3b). The selected transcription terminator sequences did not have large effects on eGFP expression levels (tested with P GAP , normalized to termination with ScCYC1tt, Table 1 and Fig. 3c). However, eGFP levels were about 20% lower compared to the other terminators when using ScCYC1tt (also reported recently in [7]), chr4_0883tt and RPL2Att. Recently, a set of transcriptional terminators derived from MUT-and other metabolic genes of P. pastoris was tested in combination with expression under control of P AOX1 and a broader Mean expression from microarray (μA) data were extracted from [26]. The relative eGFP expression is related to expression under control of P GAP in the respective condition G 2% glycerol, X glucose feed bead (limited glucose), D 2% glucose, M methanol range of expression levels from 60 to 100% relative to the AOX1 terminator was observed [20]. Compared to that, transcription terminators of GoldenPiCS, which were mainly selected from ribosomal genes (reported to be regulated at the level of mRNA stability [27]), appear to result in more uniform expression levels.

Evaluation of genome integration efficiency of GoldenPiCS multigene constructs
Vectors containing up to five transcription units without repetitive homologous sequences, including P GAP _eGFP_ScCYC1tt in different positions of the vector as readout, resulted in complete vector integration for more than 97% of all P. pastoris transformants, Fig. 4 Kozak consensus sequences of P. pastoris and S. cerevisiae. The sequences were retrieved by the regulatory sequence analysis tool (RSAT) and illustrated using weblogo although we observed a slight efficiency decrease with increasing distance from the selection marker (Fig. 5a). Relative eGFP levels were very similar for all tested constructs. In contrast, just 56% eGFP positive clones were obtained when a control vector containing twice the identical transcription unit with P GAP _eGFP_ScCYC1tt in between was used (only 9 out of 16 clones contained an integrated copy). To further increase our repertoire for strain engineering purposes, Golden Gate constructs containing a single eGFP transcription unit (P GAP _eGFP_ScCYC1tt) targeted to different integration loci (AOX1tt, RGI2, ENO1 or NTS) were analyzed. The eGFP levels were similar with all tested constructs expressing from different genomic loci in P. pastoris (Fig. 5b).

Example of multi-gene construct assembly with GoldenPiCS
Efficient genome editing by the CRISPR/Cas9 system was shown in many organisms including P. pastoris [28]. However, efficiencies and applicability were not uniformly high when using different targets or approaches. We applied GoldenPiCS to assemble different alternatives of the two transcription units of humanized Cas9 (hcas9) and single guide RNA (sgRNA) on one single episomal plasmid and test them for their efficiency to perform InDel mutations in P. pastoris (Fig. 6). The assembled BB3 plasmids were episomally maintained in P. pastoris by using the S. cerevisiae CEN/ARS locus instead of a genome integration locus [29]. Initially, we tested sgRNA expression with the SNR52 promoter (RNAPIII promoter capable to express non-coding RNA) and the SUP4 terminator from S. cerevisiae [30], and hcas9 controlled by P ScTEF and ScCYC1tt, but we could not obtain InDel mutations in P. pastoris. Next, we tried the strong RNAPII promoter P GAP and flanking self-splicing hammerhead (HH, 5′) and hepatitis delta virus (HDV, 3′) ribozyme sequences for correct processing of the sgRNA [31] and observed an efficiency of up to 90% when targeting eGFP, similar as described by Weninger et al. [28]. To reduce potential loop-out problems of the Cas9 transcription unit encountered during expression in P. pastoris, we exchanged the ScCYC1tt terminator of the sgRNA transcription unit for the P. pastoris-derived transcription terminator RPS25Att to avoid repetitive sequences. Regarding the expression of hcas9, we obtained similarly high efficiencies with different promoters, however, growth was weaker with P ScTEF1 while it was almost unaffected when using P LAT1 or P PFK300 . Targeting efficiency was mostly dependent on the applied sgRNA sequence, as we found large differences for several examples: At least two different sgRNAs designed by CHOP CHOP [32] were tested for each target. In all cases, they resulted in different efficiencies for InDel formation, e.g. two different sgRNAs each targeting AOX1 and DAS2, which are non-essential on glucose, resulted in largely different efficiencies of 38% vs. 100% and 0% vs. 100%, respectively. Therefore, we recommend to test at least two different sgRNAs for each target sequence.
With this example we demonstrate the suitability of GoldenPiCS to assemble several expression cassettes on one vector and to rapidly create new variants by exchanging parts like promoters, terminators or expression sequences. The GoldenPiCS based vectors for the described CRISP/Cas9 approach are available as separate kit at Addgene (Gassler et al. 2018.) This allowed rapid optimization of a CRISPR/Cas9-BB3 for efficient InDel mutations in P. pastoris within a short time. clones (quadruple combinations), respectively. TU1-4 were cloned with the promoters and transcription terminators P POR1 /RPS3tt, P PDC1 /IDP1tt, P ADH2 /RPL2Att and P MDH3 /TDH3tt, respectively. Coding sequences of TU1-4 were derived from various non-essential intracellular-protein coding genes of P. pastoris with a length of 500-2000 bp. The influence of the genomic integration locus was analyzed using vectors containing only one transcription unit (P GAP _eGFP_ScCYC1tt) (b). In both cases clones were screened under glucose surplus

Conclusions
Advanced synthetic biology tools revolutionized genetic engineering and are applied for many different organisms today. Strain engineering of P. pastoris has been shown to improve bottlenecks of protein synthesis, folding and secretion [1,6]. Recently, we reported that overexpression of different pentose phosphate pathway genes had synergistic effects on production of human superoxide dismutase in P. pastoris [13], as were combinations of individual enzymes involved in redox homeostasis and oxidative protein folding [14]. For such complex cell engineering approaches more efficient strategies are needed. Recently, the carotenoid pathway was introduced into P. pastoris and fine-tuned by using a set of MUT-related promoters and terminators constructed by Gibson assembly [20,33]. At about the same time, we established GoldenPiCS, a Golden Gate based modular cloning system for genetic engineering of P. pastoris. Both systems facilitate the assembly of multiple transcription units with the possibility to fine-tune the expression of each target individually. In our opinion, Golden Gate cloning has crucial advantages such as its low price, broad flexibility and high efficiency. These advantages have inspired several research groups to apply Golden Gate cloning for their purposes. So far, Golden Gate based screening in P. pastoris was dedicated to design and optimize the expression of a single recombinant gene for its production in P. pastoris by high throughput testing of different promoters and secretion signals [18,19]. In contrast, our GoldenPiCS system is aimed to facilitate cell engineering by allowing the overexpression of multiple genes e.g. redox partners or metabolic enzymes that act in a common pathway. Therefore, our system allows for the assembly of up to eight expression units on one plasmid with the ability to use different characterized promoters and terminators for each expression unit. The latter was proven to be essential to obtain stable transformants. The toolbox described by Obst et al. [18] is based on the yeast toolkit (YTK) and is primarily designed to test gene expression of a heterologous protein of interest with different regulatory elements (with a strong focus on the comparison Fig. 6 Overview of CRISPR/Cas9-BB3 plasmids assembled using GoldenPiCS. Transcription units for sgRNA (fusion of crRNA and tracrRNA [42]) and hcas9 (with SV40 nuclear localization sequence at its C-terminus) with various promoters and transcription terminators were assembled in BB3cK_AC (CEN/ARS locus, KanMX selection marker, fusion sites FsA-FsC). The plasmids were successfully used to generate InDel mutations in different genomic loci and eGFP in P. pastoris. The effects of the tested constructs are summarized in the table below the vector scheme. Modules which worked most efficiently are indicated in bold in the vector scheme of different published signal peptides), but is rather limited to a subset of 4-6 strong promoters and just employs two transcription terminators, one thereof taken from the original YTK. It also overlaps with the high throughput screening platform described by Schreiber et al. [19], where two promoters and three different secretion signals were tested for the production of antimicrobial plasmids. The GoldenPiCS system is not primarily aimed for the expression screening of the heterologous protein itself (although it can be used for it), but dedicated to combinatorial cell engineering strategies or the expression of whole metabolic pathways. We have thus added an example on the assembly of CRISPR/Cas9 vectors with 2 expression cassettes, where the issue of stability with repeated sequences (and its solution with the promoter library) is illustrated, as well as the advantage of fast assembly of elements. Aside from the development of the GoldenPiCS toolkit, we have invested effort to assay promoter and terminator strength in different conditions, to validate the effect of the '-1' position in front of the start codon and to present data for the 'loop-out' effect in P. pastoris transformants. Importantly, we found that repetitive sequences on the expression vector lead to unwanted recombination events and therefore must be avoided.
Overall, we present the hierarchical multi-organism modular cloning system (GoldenMOCS) and provide several modules and plasmids for P. pastoris: GoldenPiCS consists of 20 P. pastoris promoters, 10 terminators (all P. pastoris-derived, except for the terminator ScCYC1tt), 4 integration loci (RGI2, ENO1, NTS and AOX1tt) and one locus for episomal plasmid maintenance, as well as 4 resistance marker cassettes (hphMX, natMX, kanMX and ZeoR; the latter with loxP sites). With the currently available set of fusion sites, assembly of up to eight expression units per plasmid is possible. All of these are available through Addgene (please note that the ARS/CEN locus for episomal plasmid maintenance is only part of the CRISPR/Cas9 kit; Gassler et al. 2018) and allow high throughput assembly of multigene constructs for cell and metabolic engineering purposes.

Molecular biology
Primer design and in silico cloning was performed using the CLC Main Workbench Version 7.7.3. GoldenPiCS module sequences and backbones are listed in Additional file 2. Custom DNA oligonucleotides and gBlocks (from IDT, BE), restriction enzymes, T4 Ligase, Q5 polymerase (all from New England Biolabs, DE, or Fermentas, DE) and DNA cleanup kits (from Qiagen, DE, and Promega, DE) were used for routine cloning work.

Basic principle and background
Golden Gate cloning [17,34,35], a modular cloning system, was set up for simultaneous overexpression of multiple genes independent of the microorganism and further developed for application in P. pastoris (see Fig. 1 for a schematic overview). We termed the basic system GoldenMOCS, (Golden Gate-derived Multiple Organism Cloning System) and the subsystem specialized for application in P. pastoris was named GoldenPiCS (includes GoldenMOCS plus further developments). The system is generally comprised of three backbone (BB) levels. BB1 constructs harbor the three basic modules (promoters, coding sequences and terminators), BB2 constructs are used to assemble transcription units (promoter + CDS + terminator) and BB3 are used to further combine multiple transcription units (Fig. 1).
Golden Gate cloning employs type IIs restriction enzymes (BsaI and BpiI) which cut outside of their recognition site and enables scarless cloning, assembly of multiple DNA fragments and efficient one-pot cloning reactions (simultaneous restriction and ligation; termed Golden Gate assembly reaction). Therefore, internal BsaI and BpiI restriction sites must be removed from all modules by introducing point mutations, respecting the codon usage of the host organism (P. pastoris codon usage reported by De Schutter et al. [36]).
Fusion sites, modules and plasmids (summarized in Fig. 1 and Additional file 2) Golden Gate cloning applies the two type IIs restriction endonucleases BsaI and BpiI, which yield four base pair overhangs outside of their recognition sequence. These overhangs -termed fusion sites (Fs) -can be freely designed and are used to systematically assemble modules in the GoldenMOCS.
Modules of the GoldenPiCS include basic modules (promoters, CDSs, terminators), resistance cassettes, integration sites, and linkers (containing restriction sites for DNA integration and excision; BB2 linkers additionally contain a 5′ located strong artificial transcriptional terminator (BBa_B1007, modified E. coli thr terminator) to prevent transcriptional read-through from the resistance gene). The following nomenclature is used for basic modules: P XXXn or pXXXn for promoters, YYYn for coding sequences and ZZZntt for terminators. Modules from other organisms are indicated by the initials of the species name, e.g. 'Sc' for S. cerevisiae.
Recipient BB1 and BB2 plasmids were adapted from pIDT-SMART (IDT, BE) and pSTBlue-1 (VWR, DE), respectively (with kanamycin/ampicillin resistance). All recipient backbones are comprised of an origin of replication for E. coli, a resistance cassette (for E. coli or P. pastoris, a linker with BpiI and/or BsaI cloning sites, while BB3 additionally contains a genomic locus for integration or episomal plasmid maintenance for P. pastoris.

Golden Gate assembly -BB1
The three basic modules promoter, CDS and terminator are assembled into BB1 using primers with BsaI sites and two appropriate fusion sites: Fs1-Fs2 to integrate into recipient BB1_12 (for promotor modules), Fs2-Fs3 to integrate into BB1_23 (for CDS modules) and Fs3-Fs4 to integrate into BB1_34 (for terminator modules). Multiple fragments (e.g. to introduce mutations or to create fusion genes) can be assembled in the BB1 assembly reaction by appropriate fusion site design. All inserts which were assembled into BB1 need to be checked by sequencing.

Golden Gate assembly -BB2
Single transcription units (promoter, CDS, terminator) are assembled into a recipient BB2 using the fusion sites Fs1, Fs2, Fs3 and Fs4. Depending on the intended position of the transcription unit in BB3, the appropriate BB2 is used: BB2_AB with FsA-FsB for the first, BB2_BC with FsB-FsC for the second position, etc.

Golden Gate assembly -BB3
Multiple transcription units are assembled into a recipient BB3 using fusion sites appropriate for the number of transcription units (e.g. A-C for two transcription units). For overexpression of a single transcription unit, direct cloning from BB1 into a special BB3, equipped with a BB2 linker with BpiI restriction sites and fusion sites Fs1-Fs4 (e.g. BB3aN_14) can be done.

BB3 creation
De novo assembly of BB3 plasmids can be done using a BpiI Golden Gate Assembly reaction with the following modules: linker, resistance cassette, integration locus and Ori modules -including appropriate flanking fusion sites (Fs1-2, Fs2-3, Fs3-4 and Fs4-1, respectively). In order to create recipient BB3 for direct cloning with BB1, the linker with fusion sites FsA-FsB can be replaced using a BsaI reaction with BB2_ABthereby introducing the BB2 linker with BpiI restriction sites and fusion sites Fs1-Fs4. Amplification and sequencing primers are included in Additional file 2.
BB3 plasmids for CRISPR/Cas9-mediated genome editing were assembled as usual recipient BB3, consisting of resistance cassette, Ori, CEN/ARS locus and a linker containing fusion sites FsA-FsC for integration of two transcription units (Additional file 2 and Fig. 6).

Characterization of genetic parts in P. pastoris
For evaluation of promoter and terminator function (screening), P. pastoris transformants were cultivated at 25°C on a rotary shaker at 280 rpm. Screening conditions were designed to represent bioreactor cultivation phases (Additional file 1: Table S2). Briefly, glycerol and glucose excess conditions ("G", "D") as present in batch cultivation were analyzed at a high growth rate of μ MAX~0 . 22 and an OD 600 of about 3-8. Limiting glucose ("X", 12 mm glucose feed beads, releasing glucose at a non-linear rate of 1.63 • t 0.74 mg per disc, Kuhner, CH) and methanol feed ("M"), representing fed batch conditions, were measured at an OD 600 of about 10 and growth rates around 0.04 h −1 and μ MAX-MeOH (up to 0.1 h −1 ), respectively. Growth rates and biomass increase can roughly be calculated from the substrate yield coefficient, which is Y X/S~0 .5 for μ > 0.05 h −1 on glucose [37] and Y X/S~0 .6 on glycerol [38], while it is lower on sole methanol (Y X/S~0 .4) and methanol culture lag phases are prolonged [39].

Flow cytometry
Analysis of eGFP levels in screenings and corresponding calculations were done as described before [40,41]. Briefly, fluorescence intensity is related to the cell volume for all data points, resulting in specific eGFP fluorescence. Thereof, the population's geometric mean is normalized by subtracting background signal (of non-producing P. pastoris wild type cells) and related to expression under the control of P GAP . Indel mutation screenings with CRISPR/Cas9-BB3's were done in a CBS7435(Mut S ) strain stably expressing eGFP under control of the P GAP promoter (integration in the native P GAP locus). Disruption