{"paper_id":"b268a0db-fd97-487f-b408-9c6db4bcbd6e","body_text":"1 \nCombinatorial Design Testing in Genomes with POLAR-seq \n \nKlaudia Ciurkot1,2, Xinyu Lu1,2, Anastasiya Malyshava1,2, Livia Soro1,2, Aidan Lees1,2, Thomas E. \nGorochowski3, Tom Ellis1,2,4 \n \n1 Imperial College Centre for Synthetic Biology, Imperial College London, London, UK  \n2 Department of Bioengineering, Imperial College London, London, UK \n3 School of Biological Sciences, University of Bristol, 24 Tyndall Avenue, Bristol, UK \n4 Associate Faculty Program, Wellcome Sanger Institute, UK \n \nCorrespondence to Tom Ellis: t.ellis@imperial.ac.uk \n \n \nAbstract \nSynthetic biology projects increasingly use modular DNA assembly or synthetic in vivo recombination \nto generate diverse combinatorial libraries of genetic constructs for testing. But as these designs \nexpand to multigene systems it becomes challenging to sequence these in a cost-effective way that \nreveals the genotype to p henotype relationships in the libraries. Here, we introduce a new quick, \nlow-cost method designed for assessing combinational designs of genome -integrated multigene \nconstructs that we call Pool of Long Amplified Reads (POLAR) sequencing. POLAR -seq takes \ngenomic DNA isolated from library pools and uses long range PCR to amplify target genomic regions \nup to 35 kb long containing combinatorial designs. The pool of long amplicons is then directly read \nby nanopore sequencing with full length reads then used to identify the gene content and structural \nvariation of individual genotypes in the library and read count indicating how abundant a genotype \nis within the pool. Using yeast cells with loxP -containing synthetic gene clusters th at rearrange in \nvivo in the presence of Cre recombinase, we demonstrate how POLAR-seq can be used to identify \nglobal patterns from combinatorial experiments, find the most abundant genotypes in a pool and also \nbe adapted to sequence-verify gene clusters from isolated strains. \n \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 2 \nIntroduction \nSynthetic biology and metabolic engineering aim to create engineered host organisms that perform \nnew tasks or produce a high-value compounds. Typically, this genetic engineering centres around \ninserting sets of genes and regulatory elements into the genome , usually building -up a cluster of \nsynthetic genes at a single locus 1,2. For metabolic engineering, c onstructing multi-step enzymatic \npathways requires the balance  between achieving sufficient production  of the target molecule, \nachieving cofactor equilibrium, while also minimizing metabolic burden and the accumulation of toxic \ncompounds in the cell3. The classic approach to this relies on design-build-test-learn cycles where a \ncell factory is built and optimized in iterative cycles , an altogether  time-consuming task 4,5. In \nresponse to this, many tools have emerged from synthetic biology that utilise modular DNA assembly \nto rapidly build combinatorial librarie s of genetic de signs6,7. When combined with functional \nscreening assays, these tools enable researchers  to test the optimal composition of genes for \npathways and genetic circuits, allowing scientists to evaluate a large number of parts  in a single \nexperiment3,4,8,9. As well as determining the DNA part choices that optimise the transcriptional3,10–12 \nand translational13,14 levels of genes (e.g. via promoter and 5’UTR choices), combinatorial libraries \nalso allow researchers to test the impact of the position and orientation of each gene within a cluster \nor set15.  \n \nCombinatorial diversity of a set of genes encoding an engineered function is typically achieved during \nthe DNA assembly steps, using pools of parts to generate pools of gene-encoding plasmids3,8,16. The \nfinal pool of multigene plasmids is then transformed into the host cell, with the DNA ideally being \ndesigned so that the cluster of genes ends up stably integrated into the host genome at a known \nlocus for reliable expression. However, an alternative approach for combinatorial library construction \nis to first place all the genes and regulatory elements for a function into a locus in the host genome \nand then use targeted recombination at this locus to generate combinatorial diversity in vivo, so that \nthe initial synthetic gene set quickly becomes thousands of different designs in the growing cell \npopulation.  \n \nThe widest-known approach for in vivo recombination of synthetic genes is SCRaMbLE (Synthetic \nChromosome Rearrangement and  Modification by LoxPsym - mediated Evolution)  where \nheterologous expression of Cre recombinase rearranges DNA within synthetic gene regions that \ncontain the LoxP site s that it recognises 17. Originally, SCRaMbLE was developed for the  Sc2.0 \nproject, a collaborati ve project  to construct a  Saccharomyces cerevisiae strain with synthetic \nchromosomes and ultimately  a fully synthetic genome . Following th e Sc2.0  design, synthetic \nchromosomes contain loxPsym sites within the 3’ UTR of all non-essential genes18, and these enable \ninducible Cre-mediated gene deletion , duplication  and rearrangements  within synthetic \nchromosomes in vivo.  \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 3 \nBut while the original use of SCRaMbLE was to trigger genome-wide deletions in yeast to better \nunderstand gene function, it has also demonstrated use for those seeking to rapidly optimise genetic \ndesign of specific functions encoded by synthetic genes and gene clusters.  For example, \nSCRaMbLE in yeast has been  used to generate a  combinatorial library of strains encoding -\ncarotene biosynthesis pathway designs, within which strains with a 5 -fold increase in -carotene \ntitres were identified 19. By having LoxPsym sites flanking the four enzyme-encoding genes, \nSCRaMbLE shuffles the gene position, orientation, and copy number, leading to diverse designs \nwith altered gene expression19,20. More recently, a SCRaMbLE-inspired in vivo optimisation system \ndeveloped by Cautereels et al. has been described that enables Gene Expression Modification by \nLoxPsym-Cre Recombination (GEMbLeR) . This system places genes with a panel of regulatory \nelements into a locus in the yeast genome and uses Cre recombinase to shuffle which promoter and \nterminators DNA parts flank a gene, diversifying its expression in a population of cells20.  \n \nWith combinatorial library approaches increasingly used in synthetic biology to optimise for \nengineered novel functions and high yields of biosynthetic products , a critical bottleneck that has \nemerged is the speed in which the genotypes can be determined. In combinatorial approaches, a \nphenotypic or fluorescence-associated screen is used to assay thousands of strain designs from a \nlibrary to isolate the best performing cells21–23. The underlying DNA sequence of these cells is then \ndetermined to identify the genotype -to-phenotype relationships that explain the best designs. This \napproach works well when the DNA region undergoing combinatorial design is short and so can be \nresolved by Sanger sequencing methods (up to 1 kb) or by short-read Illumina amplicon sequencing \n(up to 500 bp)24.  \n \nHowever, when combinatorial DNA constructs introduced into cells are multi -gene length these \nmethods fall short, and this is especially true when the library DNA is integrated into the host genome \nor is a cluster of genes diversified by SCRaMbLE and related methods. Typically, the researcher is \nleft with no choice but to perform whole -genome sequencing of strains just to resolve the best -\nperforming designs, despite the DNA encoding the synthetic regions being a tiny fraction of the \ngenome. Due to the high cost of this, only a few strains are usually sequenced, meaning that all the \nsequence-to-phenotype information associated with all other pos sible designs, including the worst \nperformers, is unknown: representing a major loss for using any downstream learning approaches, \nincluding machine learning. Clearly there is a major gap between our abilities to affordably sequence \na few engineered genome s in the million bp scale and millions of combinatorial DNA designs that \nare under 1 kb in length. A targeted method that sequences hundreds or thousands of strains, but \nonly in the region of the genome where combinatorial diversity is introduced is needed. \n \nTo help bridge this gap, O’Connell and co-workers recently described a two-step sequencing method \nwhere multi-gene length combinatorial library DNA is designed to include a DNA barcode and the \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 4 \nplasmid pool post-library construction is subjected to long read sequencing to determine the DNA \nparts combination associated with each barcode. Following phenotypic screening of the plasmid \nlibrary in cells, the short barcode region is then amplified for short-read sequencing and the barcode \nassociated with each cell is then mapped to pre-sequenced longer design25. While this approach is \nespecially powerful, it requires investment in multiple rounds of sequencing. A more elegant and \npotentially cheaper approach would be to simply just use long -read sequencing to  reveal the full \nsequence of the DNA in the screened cells without any pre-requirements. This could take advantage \nof the ability to use barcoding primers with nanopore sequencing so that DNA c onstructs derived \nfrom different cells could be multiplexed in single sequencing run. As an alternative to commercial \nbarcoding kits, Currin et al. proposed amplification of a construct up to 10 kb with tailed primers \nencoding unique barcodes26. Sequencing solutions also exist for combinatorial libraries of plasmids, \nisolated from individual cell cultures or pools and prepped using transposase-based chemistry27,28. \nThe crucial consideration in screening construct libraries is sequencing solely t he target region , \nhowever methods remain limited for genomically integrated constructs exceeding a standard PCR \nrange.  \n \nHere, we introduce a new sequencing method specifically designed to resolve combinational designs \nof genome -integrated multi -gene DNA c onstructs that we call Pool of Long Amplified Reads \n(POLAR) sequencing. POLAR -seq, demonstrated here using engineered yeast libraries,  relies \nsimply on just three experimental steps;  (1) isolation of high molecular weight genomic DNA from \ncells screened from combinatorial libraries, (2) ultra-long PCR amplification o f the engineered \ngenomic region, and (3) long -read nanopore sequencing (Figure 1A,B). Reads covering the full \nlength of the engineered region are then selected based on presence of primer binding regions using \nPorechop29 (Figure 1C) and reads shorter than the size of the smallest amplicon detected on an \nagarose gel are removed. Genotypes are revealed by annotating each read with Liftoff 30, allowing \nthe arrangement and content of the DNA parts  in the synthetic region of the genome in the  cell \npopulation to be analysed with custom scripts.  \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 5 \n \n \nFigure 1: Experimental steps and data analysis of POLAR -seq. A. Overview of POLAR -seq \nworkflow for combinatorial libraries: (1) Parts for composing  of a  gene expression cassette are \nassembled in a one-pot reaction to create a diverse library of constructs  inserted into the genome. \n(2) Colonies of desired phenotype are selected for screening. (3) Cells are grown and subjected to \ngenomic DNA isolation which is used as a template in long-range PCR to amplify synthetic cluster. \n(4) PCR amplicons are prepped for sequencing using the ligation kit and sequenced on nanopore \nplatform. B. Overview of POLAR -seq workflow for in vivo recombined libraries: (1) Parental strai n \nwith a synthetic cluster that contains a set of genes (colour) flanked by loxP sites (black triangle) is \nsubjected to SCRaMbLE. Cells also contain a SCRaMbLE  reporter plasmid encoding Cre \nrecombinase (in grey) and flipped GFP reporter (green) between two  pairs of loxP sites (black and \ngreen triangles). Cre -mediated recombination leads to rearrangements within the cluster and \nreversal of GFP orientation (green) which turns on  cell fluorescence. (2) GFP positive cells that \nunderwent SCRaMbLE re-arrangements can be retrieved by FACS. Steps (3) and (4) are performed \nas described in panel A. C. Workflow for data analysis steps post-sequencing and the resulting files, \nas described in the methods. Figure created with BioRender.com. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 6 \nResults \nEstablishing long range PCR from engineering yeast genome samples \nTo develop a method to selectively sequence DNA from a defined region of a host genome, we \nfocused on optimisation of conditions for long range PCR from genomic targets, specifically testing \nthis with a S. cerevisiae yeast strain containing a 28 kb  synthetic gene cluster integrated into the \nURA3 locus at chromosome V. Long range PCR amplification is possible up to 30 kb and beyond, \nbut is known to be challenging for several reasons, especially if the DNA used as a template has \npoor integrity. Therefore, to prevent the use of damaged DNA and to remove any PCR inhibitors \nfrom the template DNA substrate, our first priority was to employ a genomic DNA isolation method \nthat maintains high-molecular DNA fragments (>50 kb), as described in the methods section. The \nmethod of Denis et al. allows to extract yeast DNA with N50 of 50 kb while maintaining purity suitable \nfor long reads sequencing. 31  \n \nAnother known challenge for long range PCR is the presence of secondary structure and high GC \ncontent in the DNA to be amplified, as both can prevent primers from binding correctly and can cause \npremature termination of strand elongation. GC content was not expected to be of concern for the \nwork described here, as the clusters we amplified had a normal 40% GC content. A further challenge \nis the design of primer pairs specific for the desired target, where it is important that designs do not \ngenerate non-specific amplicons. To avoid this, we therefore initially examined three sets of primers \nwith different binding regions.  \n \nFinally, s election of a DNA polymerase is crucial for successful generation of ultra -long PCR \namplicons (Figure 2A )32. For this purpose, w e tested a range of commercially available DNA \npolymerases expected to be able to do long amplicons. Two of the tested polymerases, LA Taq HS \nand Herculase II generated PCR products from 18 kb up to 35 kb (Figure 2B). LA Taq HS is a mix \nof Taq polymerase and a DNA polymerase exhibiting 3’→5’ exonuclease activity  with hot -start \nmediated by a monoclonal antibody against Taq Polymerase. In the further experiments we \ncontinued all work with LA Taq HS which achieved higher PCR product yields compared to \nHerculase II (Figure 2B). Importantly, all PCRs were also set to a total volume of 25 µL to avoid any \nthermal gradients.  PCRs were then further optimized to find conditions that prevent formation of \nunspecific amplicons, and this led to us a protocol with primer concentrations reduced to a final \nconcentration of 0.1 µM and the amount of template DNA provided to each reaction reduced to 20 \nng (Figure 2C, Supplementary Method 2).  \n \nLong range PCR from libraries of yeast cells with rearranged genomic DNA \nHaving established a protocol for long -range PCR from isolated yeast genomic DNA , we next set \nout to show that this can amplify long-range DNA from synthetic gene clusters in the yeast genome \nwhen taking diverse pools of yeast cells, rather than just taking from single colonies of yeast cultured \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 7 \nto high density. To demonstrate this we grew two yeast strains, known as strain C and strain H, \nwhich have each been constructed for other projects to have a synthetic multi-gene cluster integrated \ninto the URA3 locus of their genomes (see SI Note 1). Strain C’s cluster is 28 kb long and contains \n9 genes with roles in the control of the cell cycle in yeast, an essential function  but with gene \nredundancy. Strain H contains an 18 kb cluster of 7 genes encoding a biosynthesis pathway that is \nessential to be fully intact in cells grown in minimal media. In both cases the synthetic clusters have \nbeen designed with loxP sites between each gene so that genes in the cluster can be rearr anged, \ndeleted and duplicated in vivo through expression of Cre recombinase in the nucleus. The cluster in \nstrain H uses symmetrical loxP sites (loxPsyms) to allow all types of rearrangements, but the cluster \nin strain C uses non-symmetrical loxP sites and so heavily favours deletion events.  \n \nCells of strain C and strain H were transformed with a plasmid construct that expresses Cre \nrecombinase and shuttles this into the nucleus to recombine genomic loxP sites when induced by -\nestradiol18. This plasmid, when used in a SCRaMbLE with Sc2.0 yeast strains, allows for rapid \ninduction of Cre and leads to deletions, duplications, and/or translocations of genes flanked by loxP \nsites. Strain C and H cells were grown in flasks in minimal media and given -estradiol to induce \nSCRaMbLE with the intention of rapidly diversifying the content and arrangement of their cluster \ngenes in the large population of cells in the flask. \n \nUsing a SCRaMbLE reporter construct that reverses GFP -encoding ORF DNA into an expressing \norientation in response to Cre -Lox recombination, we next sorted the yeast by fluorescence to \ncapture the cells in the population most likely to have had SCRaMbLE events in their genome . \nSorting of the GFP+ cells was performed by FACS to isolate population with DNA rearrangements \nand high-quality genomic DNA from the pools of sorted yeast cells was then obtained. Within these \ngenomic DNA samples, should be the DNA sections encoding the clusters after their diverse \nrearrangements. \n \nLong range PCR was then attempted as before, but now using pool genomic DNA samples as \ntemplate. For both the C and H strain pools, the PCR protocols proved successful in amplifying the \ncluster DNA regions. The PCR of the cluster DNA from the sorted C strain pool led to amplification \nof DNA fragments of 7 distinguishable size group from 12 kb up to 28 kb in length, the original length \nof the C cluster. (Figure 2D). Each size group represents clusters with a different number of deleted \ngenes. Meanwhile, PCR of cluster DNA from the sorted H strain pool gave amplicon products mostly \naround 20 kb in length, but with two groups of shorter amplicons (approximately 12 kb and 18 kb) \nalso visible by agarose gel electrophoresis (Figure 2D).  \n \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 8 \n \n \nFigure 2: Long-range PCR amplification of genomic regions. A. A multigene synthetic construct \nintegrated into a genomic locus with a selectable marker gene (red) is flanked by primer binding sites \n(yellow) that are used for PCR amplification of clusters up to 35 kb in length. B. Comparison of LA \nTaq Hot Start and Herculase II polymerases for amplification of 23 to 35 kb regions using primers \nKC001/2, KC003/4, KC005/6. The annealing temperature of all PCR reaction with Herculase II was \nset to 52.5 ℃. Extension time was set to 15 min for all PCR products generated with LA Taq, while \nthis parameter was adjusted for reactions with Herculase II (30.1 – 35.2 kb: 18 min, 23.1 – 28.2 kb: \n14 min). Amplicons were diluted x10 and 10 µL were loaded on the gel. PCR amplification with KOD \nXtreme produced unspecific products and repliQa HiFi Tough failed to generate any products (not \nshown). C. Optimization of PCR amplification with LA Taq HS. Reduction of primers concentration \nfrom 0.5 µM to 0.1 µM prevents formation of unspecific prod ucts. D. PCR amplification of C and H \nclusters from pools of cells subjected SCRaMbLE and FACS-sort.  \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 9 \nLong-read sequencing reveals genotypes of SCRaMbLEd clusters \nWe next assessed whether PCR libraries generated from DNA extracted from pools of SCRaMbLEd \ncells were suitable for nanopore sequencing. Amplicons from long-range PCR from the C strain pool \nwere prepared for sequencing using the ligation kit SQK-LSK109 and sequencing was performed on \nR9.4.1 Flongle flow cells which allows a rapid, low-cost experiment while providing up to 2 Gb of  \ndata. To determine whether sequencing on Flongle underrepresents certain genotype s, we also \nconducted sequencing of the amplified C clusters on R10.4.1 MinION flow cell.  \n \nSequencing of the pool of C clusters on both types of flow cells was successful and after a series of \ndata analysis steps (see Methods) we ended up obtaining over 2000 high-quality annotated reads \nthat cover the full length of the synthetic cluster via the Flongle sequencing, and over 100 times more \nreads (>245,000) via the MinION sequencing ( Table 1). In both cases the reads were suitable for  \ndetermining the C cluster genotypes of the cells from the sorted post-SCRaMbLE pool. Importantly, \nthe sequencing of the long amplicons allowed for rapid identification of Cre/loxP mediated gene \ndeletion combinations within the C cluster. It led to the discovery of 432 unique C cluster genotypes \nusing the MinION flow cell data and 81 unique genotypes using the Flongle data. Importantly, the 81 \ngenotypes recognized from data obtained with Flongle device overlapped with those found via  \nMinION flow cell sequencing, where the least frequent genotypes represented only 0.02% of the \ntotal reads and therefore only a tiny fraction of the yeast cells in the sorted pool.  \n \n \nTable 1: Overview of the nanopore sequencing data generated and processed for sorted \nstrain pools of SCRaMbLEd C and H clusters. \n \nLibrary C cluster C cluster H cluster \n Flow cell Flongle MinION Flongle \n Chemistry v9 v14 v9 \n Bases generated [Mb] 232.06 26,850 658.75 \n Reads passed 11,765 1,820,119 36,986 \n Reads demultiplexed 3,341 413,702 6,635 \n Reads above threshold 2,839 332,090 6,401 \n Reads annotated 2,084 245,895 4,050 \n Detected genotypes 81 432 127 \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 10 \nNext, the set of 10  most abundant genotypes in the data were visualized and the count of reads \ncorresponding to each genotype was compared between runs (Figure 3A). These 10 genotypes \ntogether account for around 75% of all reads from the sequencing runs and all contain at least one \ngene deletion from the C cluster, with some showing up to 4 deleted genes. For all 10 genotypes the \npercentages of reads from the MinION and Flongle experiments were in close agreement, with \nvariability between sequencing runs only expected to be caused by pipetting errors, for example \nwhile setting up the PCRs and loading the sequencing flow cells. This shows that either flow cell is \nsuitable for these experiments. Genotypes i–iii were by far the most abundant, together accounting \nfor 45.8% of MinION reads and 47.2% Flongle reads , presumably due to the c ells with these \ngenotypes being most abundant in the sorted cell pool. Overall, t his initial analysis shows that the \nfrequency of SCRaMbLE rearrangements in a synthetic yeast cluster can be estimated from pooled \nsamples by long-range PCR and long read sequencing.  \n \nRearrangements within the cluster can be tracked globally  \nOver 99 .5% of the reads  from the C cluster pool  contained a gene deletion , with only  4.4% \nadditionally containing a gene duplication. To investigate which genes are deleted from the cluster \nmost frequently, we counted the occurrence of each gene in all annotated reads (Figure 3B). The \nfrequency of genes in the cluster obtained from the MinION and Flongle sequencing closely \nmatched. As shown in the Figure 3B, the most frequently deleted gene was C7 (found in only 20% \nof the detected genotypes), followed by C 8 (40%) and C9 (50%). Whereas genes C4, C5 and C6 \nwere those most frequently kept in the cluster; with C4 and C6 gene detected in 91% and 98% of \nthe reads, respectively , and gene C5 being found in 100% of the reads . This kind of analysis is \nespecially useful for quantifying gene essentiality, for example in a growth condition. Gene C5 in this \ncase proving to be essential (found in 100% of reads), and gene C7 proving to be non-essential (only \nfound in 20% of reads). \n \nImportantly, long-read sequencing also reveals the combinations of genes that can be deleted within \nthe same cluster, something difficult to achieve from a pooled sample by short -read sequencing or \nby RNAseq. Analysis of common gene deletion combinations (Figure 3C) showed that genes C7 \nand C1 were commonly deleted together, as were C7 and C8, and also genes C2 and C3 . \nVisualization of deletion combinations  also reveals those  rarely or never  deleted together, for \nexample genes C4 and C8. \n \n \n \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 11 \n \n \nFigure 3: POLAR-seq of the C cluster subjected to combinatorial deletions. A. Genotype of the \nparental strain and ten most abundant rearranged genotypes discovered by sequencing on MinION \nand Flongle flow cells. The percentage represents number of reads discovered for a given genotype \nand normalized by all the annotated reads. B. Frequency of genes maintained in the C cluster after \ncombinatorial deletions expressed as percentage (count of gene annotation normalized by the total \nnumber of annotated reads). Data points represent frequency from the two independent sequencing \nruns (MinION and Flongle). C. Combinations of gene knockouts recorded in rearranged C clusters. \nHeat map was prepared for the 20 most abundant genotypes . The colour intensity indicates the \nnumber of reads detected for the given genotype.  \n \n \n \nPOLAR-seq allows to study gene inversions and translocations \nHaving shown with the C strain that POLAR-seq identifies deletion genotypes in a post-SCRaMbLE \nlibrary, we test ed whether it serves to identify other structural variations . For this we used the H \nstrain, where the cluster genes  are flanked by symmetrical loxPsym sites  that permit gene \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 12 \nduplication, translocations and inversions as well as deletions. Amplified cluster DNA from a post-\nSCRaMbLE library of these cells was sequenced on Flongle and data analysis from this gave over \n4000 cluster-length annotated reads that identified 127 different genotypes (Table 1). These were \nthen classed based on the type of rearrangements seen in the reads ( Figure 4A) with the most \ncommon being genotypes with deletions, duplications and inversions all occurring ( 48% of all \ngenotypes) and the second most common being genotypes with just duplications (39%). Deletions \nwere almost never seen because all 7 genes in the cluster are essential for cell viability in the \nconditions they were cultured in. \n  \nIn total 92% of all read s showed a duplication of gene H5, and this was reflected in the 5 most \nabundant genotypes found in read data (Figure 4B). Genotype i showed just a single duplication of \nH5 while the other four top genotypes ( ii-v) showed duplication in association with inversion and/or \ntranslocations of the other genes  Further analysis based on all of the determined genotypes was \nused to reveal the frequency of each gene being found to have moved position in the cluster, inverted \nits orientation or being duplicated (Figure 4C). This kind of global analysis of gene rearrangements \nhas not been demonstrated before in any published SCRaMbLE experiments and is not possible to \ndo by standard PCRTag analysis33 or the recently described LoxPTag34 method. POLAR-seq in \nparticular shows its advantage in detecting duplications, which are not easily identifiable by non -\nsequencing approaches.  Importantly, the high frequency of detecting duplications in the dataset \nhere also relieves concerns that the PCR -based target enrich ment used by POLAR -seq might \nstruggle with clusters with repetitive DNA. \n \nFinally, to  confirm whether genotype abundance is consistent with the single cells , w e next \nsequenced 8 randomly picked yeast colonies plated out from the same post-SCRaMbLE library. \nColonies were cultured in 2 mL of YPD medium, genomic DNA isolated and cluster sequence PCR \namplified using LA Taq polymerase and primers KC 007 and KC 015. These primers match the  \nprimers KC005 and KC006 used in POLAR-seq but have an additional 25 base tail (SI Table 1) that \nserves as a barcode for demultiplexing sequencing reads from combined samples, according to the \nmethod published by Currin et al 35. Three of the tested colonies exhibited genotype with a H5 \nduplication (i), which was found in 37.5% of reads generated from sequencing pool (Figure 4B). The \nnext most abundant two genotypes (ii, iii), were detected in 2 colonies each and genotype v was \nfound in 1 colony. Genotype iv was not found among 8 sequenced colonies. Overall, the frequency \nat which cluster genotypes were present in this small random selection of colonies broadly matched \nthe abundance of reads for the genotypes from our  POLAR-seq experiment, giving us confidence \nthe read numbers in our analysis are representative of the number of cells with that genotype in the \nsampled pool. \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 13 \n \n \nFigure 4: POLAR-seq of the H cluster subjected to SCRaMbLE. A.  Types of detected re -\narrangements in the H cluster. Each genotype was classified depending on re-arrangement type and \nthe number of reads representing each type was normalized by the total number of annotated reads. \nB. Genotype of the parental cluster and top five post-SCRaMbLE genotypes detected from a pool of \nFACS-sorted cells , either  normalized by the total number of annotated reads  (pool) or from  8 \nrandomly picked colonies from the pool (colonies). C. Relationship between gene copy  number, \ngene orientation, and gene position in clusters sequenced from the post-SCRaMbLE pool of cells. \nEach heatmap shows the frequency of finding each gene (H1 -H7) in positions 1 through 10 of a \nsequenced cluster, where the parental strain has genes H1-H7 in positions 1 to 7 respectively, all in \nforward orientation and single gene copy. Left heatmaps show the frequency of finding the first/only \ncopy of a gene  in each position in a  forward (top) or reverse ( bottom) orientation. Right heatmaps \nshow the frequency of finding genes in each position that are duplicated in forward (top) or reverse \n(bottom) orientations. \n \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 14 \nDiscussion \nHere, we present a method for long -range amplification and long -read sequencing of genome -\nintegrated synthetic DNA regions that is suitable for a pool of FACS-sorted cells. By sequencing the \nDNA of an entire pool of cells, a larger population can be screened to identify rare genotypes, which \nmay not  be detected by analysis of individually isolated strains. When applied to SCRaMbLEd \nsynthetic gene clusters in yeast, the POLAR-seq method allows rapid and efficient identification of \nstructural rearrangements within the cluster including duplications, deletions, inversions, and \ntranslocations. As the method simply requires just genomic extraction, PCR amplificat ion and then \nlow coverage nanopore sequencing, it is quick and affordable. Considering the cost of reagents, kits \nand flow cells, the method reveals yeast cluster genotypes from a sorted library at an approximate \ncost of $ 0.09 per genotype when using a MinION flow cell, and $ 1.5 per genotype when using a \nFlongle, assuming that 25 annotated full -length reads is sufficient to identify a cluster . A Flongle \nPOLAR-seq experiment costs only $150 in total, and can also be adapted as i n Figure 4 b to \nsequence barcoded amplified DNA from up to 24 isolated colonies, representing a cost of less than \n$10 per strain to determine individual cluster genotypes. \n \nIn developing this method, we demonstrated the use  of POLAR -seq to study Cre -mediated \nrearrangements and deletions within loxP site-containing synthetic clusters in S. cerevisiae cells \nshowing that the method was especially suited for revealing duplications, a rearrangement difficult \nto assess with classic short read sequencing approaches. POLAR-seq is however, not limited to just \nstructural rearrangement applications . POLAR-seq could also be used to reveal the optimal part \ncomposition of any DNA encoded constructs so long as the desired function of this can be screened \nand retrieved , e.g., by FACS -based cell sorting. For example, instead of rational ly selecting of \npromoter parts for each gene in a multigene metabolic pathway, all DNA parts for this pathway could \nbe assembled in a one-pot reaction, integrated into yeast and the optimal designs for function can \nthen be determined by combining a functional screen with POLAR-seq. A further possibility afforded \nby POLAR-seq is in quantifying the relative fitness of each genotype within a library. Genotypes with \nthe highest percentage of reads correspond to the strains with the highest abundance in the sampled \npool (Figure 4B), and so performing POLAR -seq at sequential time points when a library pool is \ngrown in a condition of interest could be used to reveal the strain genotypes that outgrow others due \nto better fitness. Being able to measure function and fitness from thou sands of variants in a single \nexperiment is likely to open up new opportunities in machine learning for synthetic DNA design. \n \nA limitation of PCR amplification, however, is frequency of introduced errors, especially for low fidelity \npolymerases such as Taq. Therefore, we would recommend against using POLAR-seq to determine \nvariation in libraries at the single base level, rather than at the parts and structure level as used here. \nBase mutations do not prevent our analysis from correct recognition of cluster’s genes  as this has \nbeen set to accept up to 12.6% of erroneous bases within a sequence. In applications sensitive to \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 15 \nSNPs, the PCR step in POLAR-seq can be improved using the high fidelity Herculase II polymerase \n(1 error in 777 kb). Another method to improve quality of sequencing data is the chemistry of the \nselected library prep kit and the flow cell. During the preparation of this manuscript, only v9 Flongle \nflow cells were available , however these can be now substituted with v14 flow cells.  With this \nadvance, the raw read accuracy in simplex now is reported to exceed Q28 (99.8%)36,37. \n \nFinally, the POLAR-seq method is limited by the size of the region of interest, as this must be suitable \nfor PCR amplification. As it stands POLAR-seq is not suitable to study content and structural diversity \nin genomic regions  with lengths beyond 35 k b. Other methods available for region -specific \nenrichment of genome sequencing include Oxford Nanopore Technology’s adaptive sampling \nmethod, and CRISPR/Cas9-based ‘catching’ of specific genomic regions38–40. However, these work \nby exclusion of unwanted sequence rather than amplification of target DNA, meaning that orders of \nmagnitude more genomic DNA needs to be harvested from samples to achieve the same coverage. \nPCR amplification is therefore much better suited to this task, and hopefully with continued \nimprovements in long range DNA polymerases, or by adapting rolling -circle amplification methods, \nthe lengths of genomic regions suitable for POLAR-seq analysis will increase in the future. \n \n \n \nAcknowledgments  \nThis research was supported by a  Wellcome Trust Discretionary Award (221267/Z/20/Z) providing \nfunding for K.C. and T.E., a Chinese Scholarship Council (CSC) PhD scholarship to X.L., and a \nDarwin Trust of Edinburgh PhD scholarship to A.M. T.E.G. was supported by a Royal Society \nUniversity Research Fellowship grant URF\\R\\221008 and a Turing Fellowship from The Alan Turing \nInstitute under EPSRC grant EP/N510129/1. \n \nDeclarations of Interests \nK.C. is now an employee of Oxford Nanopore Technologies but was solely employed by Imperial \nCollege London during the time generating the data included in this paper. All other authors declare \nno conflicts of interest. \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 16 \nMethods \nStrains and Media \nAll yeast strains used in this study are derivatives of BY4741 yeast (MATa his3Δ1 leu2Δ0 met15Δ0 \nura3Δ0). Yeast strain C is a haploid strain generated by genetic engineering for an independent \nresearch project and has a synthetic cluster of 9 cell cycle related genes assembled and integrated \ninto the URA3 locus on chromosome V. Yeast strain H is also a haploid strain with a synthetic cluster \nof 7 metabolism related genes assembled and integrated into the URA3 locus. \n \nYeast extract Peptone Dextrose (YPD) media (10 g L-1 yeast extract (VWR), 20 g L-1 peptone (VWR), \n20 g L -1 glucose (VWR)) was used for culturing of yeast strains without the SCRaMbLE reporter , \nunless otherwise stated. Synthetic Complete media (SC; 6.7 g L-1 Yeast Nitrogen Base without amino \nacids (Sigma Aldrich), 1.4 g L-1 Yeast Synthetic Drop-out Medium Supplements without L-uracil, L-\ntryptophan, L -histidine, L -leucine, 20 g L -1 glucose (Sigma Aldrich )) was use d for auxotrophic \nselection. Amino acids such as 20 mg L -1 L-tryptophan, 20 mg L -1 L-histidine, 20 mg L -1 uracil and \n120 mg L -1 L-leucine were supplemented into SC media depending on the required auxotrophic \nselection. For growth on plates, media were supplemented with 20 g L-1 bacto-agar (VWR). \n \nSCRaMbLE \nC and H yeast strains were transformed with plasmids for β-estradiol-induced Cre expression and \nfor GFP expression in response to nuclear Cre activity. Strains were  then grown overnight in 2 mL \nSC complete media with appropriate selection (30°C, 250 rpm). Cultures were diluted to an OD600 of \n~0.2 in 5 mL SC complete media with appropriate selection and grown for 4 hours. SCRaMbLE was \ninduced by addition of β-estradiol (dissolved in DMSO) to a final concentration of 1 μM. Cultures \nwere grown for an additional 4 hours before being washed twice in water and resuspend in 1 mL \nPBS for Fluorescence-Activated Cell Sorting (FACS). \n \nFluorescence-Activated Cell Sorting (FACS) \nFACS sorting was performed on the BD FACSAria III Cell Sorter (BD Biosciences) to select for yeast \ncells with GFP fluorescence. Yeast cells, washed and resuspended in PBS buffer post-SCRaMbLE, \nwere transferred to a 5 mL FACS tube (Invitrogen) and diluted to appropriate density with PBS buffer \nfor FACS sorting. The 70 μm nozzle was selected for the sorting. Around 1 million GFP+ cells sorted \nfrom the FACS instrument were collected in a 15 mL Falcon centrifuge tube. A subset of cells was \nimmediately inoculated i nto appropriate media and grown for 2 -3 days to reach saturation (30°C, \n250 rpm). The remaining cells were spun down at 4000 rpm for 20 min. The pellet was resuspended \nin 0.25 mL PBS and stocked in 25% glycerol (final concentration) at -80°C. For the FACS analysis, \nyeast cells were firstly selected based on morphology (FSC -A vs SSC -A). Single cells were then \nselected based on a double doublet-discrimination (FSC-A vs FSC-H and SSC-W vs SSC-H). Single \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 17 \nGFP+ cells were then selected based on GFP expression (FS C-A vs GFP). GFP expression was \nselected with the 530/30 bandpass filter. \n \nGenomic DNA isolation from yeast \nGenomic DNA was isolated according to the method published by Denis et al.  31 with following \nmodifications: yeast culture was grown until cells reached OD=5 -10, zymolyase was replaced with \nlyticase (Sigma Aldrich, 600 U per 1 mL of OD=1) and all centrifugation steps were performed at \n4000 g ( Supplementary Method 1 ). The adapted pro tocol was also tested in minimized scale, \nmaking use of 2 mL of yeast culture and proportionally reduced reagents.  \n \nPCR amplification \nInitial testing used four DNA polymerases: LA Taq Hot Start (TaKaRa, Shiga Japan), Herculase II \n(Agilent, Santa Clara, United States), KOD Xtreme (Sigma Aldrich, Saint Louis, United States) and \nrepliQa HiFi ToughMix (QuantaBio, Beverly, United States). PCR reactions were set up in 25 µL \nvolume according to manufacturers’ protocol using 20 ng of genomic DNA as template and reduced \nconcentration of primers (final concentration 0.1 µM). Amplification was verified by gel \nelectrophoresis on 0.5% agarose gel run at 50 V for 5 h. PCR products were purified using AMPure \nbeads (1.8x volume of the PCR reaction) and quantified using Qub it dsDNA Broad Range kit and \nQubit 2.0 Fluorometer. Prior to sequencing, quality of DNA was evaluated by measuring absorbance \nusing Nanodrop spectroscopy. \n \nNanopore Sequencing \nPools of amplicons were prepared for sequencing using the NEBNext Companion Module and the \nligation kit SQK-LSK109 or SQK-LSK114. The DNA library was sequenced on the Flongle (FLG001) \nor MinION flow cell (FLO-MIN114) using the MinION Mk1B device. In each case data was collected \nduring sequencing with the latest version of MinKnow (22.05.5 – 23.04.6). \n \nData Analysis \nBasecalling sequencing data and Demultiplexing. Raw data was basecalled with Guppy (6.1.5 - \n6.5.7) using high-accuracy model and only reads with Q-score above 9 were kept for further analysis. \nTo select reads that span full amplicons we used the demultiplexing function of Porechop 29. Reads \nwere demultiplexed to firstly find binding regions of primer KC005 and subsequently of primer KC006 \nusing default settings ( barcode_threshold 75 and barcode_diff 5 ). This step ensure s that partial PCR \nproducts and reads corresponding to shredded DNA fragments are not further considered. \n \nFiltering sequencing reads. Demultiplexed reads in fastq format were next transformed into fasta \nsequence and filtered based on size. The threshold i s set based on the smallest detectable PCR \nproduct, specifically 11 kb for the C cluster and 10 kb for the H cluster. The remaining reads were \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 18 \nsorted by length and renamed numerically with funannotate41. Reads were split into individual fasta \nfiles using splitfasta.  \n \nAnnotating sequencing reads . Reads were annotated by Liftoff 30 from the provided reference \nsequence that contains all genes present in the parental cluster design. Liftoff generated a gff file for \neach read and these were uploaded as Pytho n dataframe with gffpandas. This allowed retrieval of \nthe position of each gene and could be also used to determine the orientation of a read based on \nthe fact that both C and H clusters contain an auxotrophic marker at the 3’ end. Thus, the read \norientation can be defined by presence of this marker on the forward or the reverse stand. If the \nmarker is on the reverse strand and appears as the last gene, the reverse complement sequence is \ngenerated for the given fasta file with BioPython SeqIO library. Subsequently, the generated \nsequences were re-annotated. \n \nFiltering GFF files. The GFF files were then combined for the reads encoding the forward strand \nand filtered based on the following criteria: 1) the first annotation must be within the defined distance \nA from the start of the read; 2) the gap between two annotations must be lower than distance B; 3) \nthe sequence iden tity and coverage of each gene annotated in a read must be above the 87.4% \nthreshold defined based on Q-score. 4) The reads must encode DNA sequence from the auxotrophic \nmarker as the last annotated gene. For criteria 1 to 3, values were determined based on the expected \nerror of 12.6% when Q -score=9 is set as a threshold for reads binned as passed in high -accuracy \nbase-calling. This corresponds to a threshold of 87.4% of correct bases in each read. Distance A \n(criterium 1) was calculated based on the distance from the start of the amplicon to the first annotated \ngene plus 12.6% buffer on each (corresponding to the allowed sequencing error when Q -score=9). \nDistance B was calculated as the distance between genes in the cluster plus 12.6% buffer on each \nend. \n \nData visualization. After filtering, GFF files that fulfilled the abovementioned criteria were loaded as \na Python dataframe with gffpandas. Only gene names were maintained and were next merged into \na single line, allowing to count the same genotypes. Select ed genotypes were visualized with \nDnaFeaturesViewer42. Plots were then created with Python packages: matplotlib, and plotly. \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 19 \nReferences  \n1. Lian, J., Mishra, S. & Zhao, H. Recent advances in metabolic engineering of \nSaccharomyces cerevisiae: New tools and their applications. Metabolic Engineering vol. 50 \n85–108 Preprint at https://doi.org/10.1016/j.ymben.2018.04.011 (2018). \n2. Da Silva, N. A. & Srikrishnan, S. Introduction and expression of genes for metabolic \nengineering applications in Saccharomyces cerevisiae. FEMS Yeast Research vol. 12 197–\n214 Preprint at https://doi.org/10.1111/j.1567-1364.2011.00769.x (2012). \n3. Du, J., Yuan, Y., Si, T., Lian, J. & Zhao, H. Customized optimization of metabolic pathways \nby combinatorial transcriptional engineering. Nucleic Acids Res 40, (2012). \n4. Naseri, G. & Koffas, M. A. G. Application of combinatorial optimization strategies in synthetic \nbiology. Nature Communications vol. 11 Preprint at https://doi.org/10.1038/s41467-020-\n16175-y (2020). \n5. Gurdo, N., Volke, D. C., McCloskey, D. & Nikel, P. I. Automating the design-build-test-learn \ncycle towards next-generation bacterial cell factories. N Biotechnol 74, 1–15 (2023). \n6. Mitchell, L. A. et al. Versatile genetic assembly system (VEGAS) to assemble pathways for \nexpression in S. cerevisiae. Nucleic Acids Res 43, 6620–6630 (2015). \n7. Lee, M. E., DeLoache, W. C., Cervantes, B. & Dueber, J. E. A Highly Characterized Yeast \nToolkit for Modular, Multipart Assembly. ACS Synth Biol 4, 975–986 (2015). \n8. Smanski, M. J. et al. Functional optimization of gene clusters by combinatorial design and \nassembly. Nat Biotechnol 32, 1241–1249 (2014). \n9. Schaerli, Y. & Isalan, M. Building synthetic gene circuits from combinatorial libraries: \nScreening and selection strategies. Molecular BioSystems vol. 9 1559–1567 Preprint at \nhttps://doi.org/10.1039/c2mb25483b (2013). \n10. Yuan, J. & Ching, C. B. Combinatorial assembly of large biochemical pathways into yeast \nchromosomes for improved production of value-added compounds. ACS Synth Biol 4, 23–\n31 (2015). \n11. Blount, B. A., Weenink, T., Vasylechko, S. & Ellis, T. Rational diversification of a promoter \nproviding fine-tuned expression and orthogonal regulation for synthetic biology. PLoS One \n7, (2012). \n12. de Boer, C. G. et al. Deciphering eukaryotic gene-regulatory logic with 100 million random \npromoters. Nat Biotechnol 38, 56–65 (2020). \n13. Wang, H. H. et al. Programming cells by multiplex genome engineering and accelerated \nevolution. Nature 460, 894–898 (2009). \n14. Redden, H., Morse, N. & Alper, H. S. The synthetic biology toolbox for tuning gene \nexpression in yeast. FEMS Yeast Research vol. 15 Preprint at https://doi.org/10.1111/1567-\n1364.12188 (2015). \n15. Georgakopoulos-Soares, I. et al. Transcription factor binding site orientation and order are \nmajor drivers of gene regulatory activity. Nat Commun 14, (2023). \n16. Iverson, S. V., Haddock, T. L., Beal, J. & Densmore, D. M. CIDAR MoClo: Improved MoClo \nAssembly Standard and New E. coli Part Library Enable Rapid Combinatorial Design for \nSynthetic and Traditional Biology. ACS Synth Biol 5, 99–103 (2016). \n17. Dymond, J. & Boeke, J. The saccharomyces cerevisiae SCRaMbLE system and genome \nminimization. Bioeng Bugs 3, 168–171 (2012). \n18. Dymond, J. S. et al. Synthetic chromosome arms function in yeast and generate phenotypic \ndiversity by design. Nature 477, 471–476 (2011). \n19. Wu, Y. et al. In vitro DNA SCRaMbLE. Nat Commun 9, (2018). \n20. Cautereels, C. et al. Combinatorial optimization of gene expression through recombinase-\nmediated promoter and terminator shuffling in yeast. Nat Commun 15, 1112 (2024). \n21. Hill, B. D., Prabhu, P., Rizvi, S. M. & Wen, F. Yeast Intracellular Staining (yICS): Enabling \nHigh-Throughput, Quantitative Detection of Intracellular Proteins via Flow Cytometry for \nPathway Engineering. ACS Synth Biol 9, 2119–2131 (2020). \n22. Huttanus, H. M. et al. Targeted mutagenesis and high-throughput screening of diversified \ngene and promoter libraries for isolating gain-of-function mutations. Front Bioeng Biotechnol \n11, (2023). \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint \n\n 20 \n23. Liu, M., Zhang, J., Liu, X., Hou, J. & Qi, Q. Rapid Gene Target Tracking for Enhancing b-\nCarotene Production Using Flow Cytometry-Based High-Throughput Screening in Yarrowia \nlipolytica. Appl Environ Microbiol 88, (2022). \n24. Hernandez Hernandez, D. et al. Improved Combinatorial Assembly and Barcode \nSequencing for Gene-Sized DNA Constructs. ACS Synth Biol 12, 2778–2782 (2023). \n25. O’connell, R. W. et al. Ultra-high throughput mapping of genetic design space. \ndoi:10.1101/2023.03.16.532704. \n26. Currin, A. et al. Highly multiplexed, fast and accurate nanopore sequencing for verification of \nsynthetic DNA constructs and sequence libraries. Synth Biol 4, (2019). \n27. Lood, C., Gerstmans, H., Briers, Y., van Noort, V. & Lavigne, R. Quality control and \nstatistical evaluation of combinatorial DNA libraries using nanopore sequencing. \nBiotechniques 69, 379–383 (2020). \n28. Emiliani, F. E., Hsu, I. & McKenna, A. Multiplexed Assembly and Annotation of Synthetic \nBiology Constructs Using Long-Read Nanopore Sequencing. ACS Synth Biol 11, 2238–\n2246 (2022). \n29. Wick, R. R., Judd, L. M., Gorrie, C. L. & Holt, K. E. Completing bacterial genome assemblies \nwith multiplex MinION sequencing. Microb Genom 3, (2017). \n30. Shumate, A. & Salzberg, S. L. Liftoff: Accurate mapping of gene annotations. Bioinformatics \n37, 1639–1643 (2021). \n31. DENIS, E. et al. Extracting high molecular weight genomic DNA from Saccharomyces \ncerevisiae. Protoc Exch (2018) doi:10.1038/protex.2018.076. \n32. Jia, H., Guo, Y., Zhao, W. & Wang, K. Long-range PCR in next-generation sequencing: \nComparison of six enzymes and evaluation on the MiSeq sequencer. Sci Rep 4, (2014). \n33. Richardson, S. M. et al. Design of a Synthetic Yeast Genome. Science vol. 355 \nhttps://www.science.org (2017). \n34. Lindeboom, T. A. et al. An Optimized Genotyping Workflow for Identifying Highly \nSCRaMbLEd Synthetic Yeasts. ACS Synth Biol 13, 1116–1127 (2024). \n35. Currin, A. et al. Highly multiplexed, fast and accurate nanopore sequencing for verification of \nsynthetic DNA constructs and sequence libraries. Synth Biol 4, (2019). \n36. Sereika, M. et al. Oxford Nanopore R10.4 long-read sequencing enables the generation of \nnear-finished bacterial genomes from pure cultures and metagenomes without short-read or \nreference polishing. Nat Methods 19, 823–826 (2022). \n37. Ni, Y., Liu, X., Simeneh, Z. M., Yang, M. & Li, R. Benchmarking of Nanopore R10.4 and \nR9.4.1 flow cells in single-cell whole-genome amplification and whole-genome shotgun \nsequencing. Comput Struct Biotechnol J 21, 2352–2364 (2023). \n38. Payne, A. et al. Readfish enables targeted nanopore sequencing of gigabase-sized \ngenomes. Nat Biotechnol 39, 442–450 (2021). \n39. Kovaka, S., Fan, Y., Ni, B., Timp, W. & Schatz, M. C. Targeted nanopore sequencing by \nreal-time mapping of raw electrical signal with UNCALLED. Nat Biotechnol 39, 431–441 \n(2021). \n40. Gilpatrick, T. et al. Targeted nanopore sequencing with Cas9-guided adapter ligation. Nat \nBiotechnol 38, 433–438 (2020). \n41. Almutairi, H. et al. Chromosome-scale genome sequencing, assembly and annotation of six \ngenomes from subfamily Leishmaniinae. Sci Data 8, (2021). \n42. Zulkower, V. & Rosser, S. DNA features viewer: A sequence annotation formatting and \nplotting library for Python. Bioinformatics 36, 4350–4352 (2020). \n  \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted June 6, 2024. ; https://doi.org/10.1101/2024.06.06.597521doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}