CRI-SPA: a high-throughput method for systematic genetic editing of yeast libraries

Abstract Biological functions are orchestrated by intricate networks of interacting genetic elements. Predicting the interaction landscape remains a challenge for systems biology and new research tools allowing simple and rapid mapping of sequence to function are desirable. Here, we describe CRI-SPA, a method allowing the transfer of chromosomal genetic features from a CRI-SPA Donor strain to arrayed strains in large libraries of Saccharomyces cerevisiae. CRI-SPA is based on mating, CRISPR-Cas9-induced gene conversion, and Selective Ploidy Ablation. CRI-SPA can be massively parallelized with automation and can be executed within a week. We demonstrate the power of CRI-SPA by transferring four genes that enable betaxanthin production into each strain of the yeast knockout collection (≈4800 strains). Using this setup, we show that CRI-SPA is highly efficient and reproducible, and even allows marker-free transfer of genetic features. Moreover, we validate a set of CRI-SPA hits by showing that their phenotypes correlate strongly with the phenotypes of the corresponding mutant strains recreated by reverse genetic engineering. Hence, our results provide a genome-wide overview of the genetic requirements for betaxanthin production. We envision that the simplicity, speed, and reliability offered by CRI-SPA will make it a versatile tool to forward systems-level understanding of biological processes.


INTRODUCTION
Baker's yeast Sacchar om y ces cer evisiae is an important biological model organism and a key production host in industrial biotechnology.S. cerevisiae was the first eukaryote to be fully sequenced ( 1 , 2 ), and its simple lifecycle as a single cell organism with a highly de v eloped genetic toolbox has positioned this yeast as a frontrunner in the field of systems biology.Howe v er, the unpredictability of phenotypic effects resulting from combinations of different genetic traits still challenges our fundamental understanding of yeast and its engineering.The introduction of CRISPR for yeast engineering ( 3 ) promises to relie v e this bottleneck by accelerating the systematic construction and testing of genetic variants.For example, in numerous genome-wide engineering projects, CRISPR-based strategies have been de v eloped to create and screen libraries of strains (4)(5)(6)(7)(8)(9)(10)(11).
e91 Nucleic Acids Research, 2023, Vol.51, No. 17 These screens generally follow a one-pot workflow where the library is pooled, edited and challenged before being resolved by FACS and next generation sequencing ( 12 , 13 ).Howe v er, library pooling is prone to bias as it selects for fast growth, and traits of interest might be lost if they are accompanied by a reduction in growth fitness.Similarl y, w hen the best candidates are recovered from a one-pot screen, the top pool might be sa tura ted by a few strongest variants, which limits the full network of interactions to be uncovered ( 14 ).A more useful output would be achie v ed if the members of the library could be assessed individually to produce the full ov ervie w of the genetic effects impacting the system.
Aside from one-pot approaches, screening can be achie v ed in a systematic manner by introducing a genetic modification in all strains of an existing yeast library ( 15 , 16 ); and to date, Synthetic Genetic Array (SGA) has been the state of the art method for such systematic analyses ( 17 ).SGA queries strains individually in a process based on arr ay ed mating, meiosis, sporulation and marker selection for the desired gene combination.This method has been particularly valuable for the identification of genetic fitness interactions among all double and a selected number of triple gene knock-outs ( 18 , 19 ).Howe v er, the reliance of SGA on meiosis is tedious, and the dependency of markers introduces experimental limitations.For example, sporulation is slow, 4-7 days, and the efficiency is difficult to control.Moreover, the high level of meiotic recombination required to ensure correct chromosome segregation in the first meiotic division ( 20 ) is per se undesirable.Firstly, as meiotic r ecombination mix es donor and r ecipient genomes, the genetic background of the donor and recipient strains need to be identical.Secondly, due to meiotic segregation, all genetic features need to be flanked by a genetic marker allowing selection of the desirable gene combination after meiosis.Such markers may influence expression of neighboring genes; and if many genes are studied, become a limitation for multiple xing.Thir dly, the high le v els of meiotic Homologous Recombination (HR) may also set the stage for unwanted chromosome rearrangements if the strains suffer from repeated use of genetic elements, e.g.promoters.Lastly, flawed meiosis is known to frequently generate aneuploids ( 21 , 22 ), which may complicate further analyses.A faster method, independent of meiosis and marker selection, is ther efor e desirable.
We have developed a new screening platform that allows a genetic trait to be queried in large arrayed yeast libraries.Our method (Figure 1 ), CRI-SPA, combines Clustered Regularly Interspaced Short Palindromic Repeats ( CRI SPR) technology ( 23 ); ( 3 ) with s electi v e p loidy a blation, SPA ( 24 ).In CRI-SPA, a (marker-free) genetic feature of interest is efficiently transferred from a single donor strain to the strains in a library in a process involving mating, Cas9 induced gene conversion, and haploidization by SPA.CRI-SPA can be massi v ely parallelized using a pinning robot and executed in less than a week with as little as 4 hours handling time.Since it relies on image analysis for data extraction, the main cost associated with CRI-SPA comes from media and plastic consumables.Here, we use the method to transfer four genes responsible for the synthesis of the yellow plant metabolite betaxanthin, a biosensor for the morphine precursor L-DOPA ( 25 ) into the 4800 strains of the y east knock-out (YKO) libr ary ( 26 ).We demonstr a te tha t multiple-gene transfer by CRI-SPA can be performed with and without selection in a highly reproducible manner.As a result, we unveil the full betaxanthin pathway-host gene interaction landscape and most prominently demonstrate tha t muta tions impairing mitochondrial functions consistently influence betaxanthin le v els.Finally, we show that CRI-SPA scores obtained in screening conditions correlate well with betaxanthin le v els of re v ersed engineered hits in liquid medium.

Strains and media
All strains constructed in this work are listed in Supplementary Table S1.Strains W8164-2B and W8164-2C of the SPA method ( 27 ) were kindly provided by Rodney Rothstein (Department of Genetics and De v elopment, Columbia Uni v ersity, USA).The YKO library was acquired from Invitrogen.YPD and synthetic complete (SC) media, and SC drop out media wer e pr epar ed as described by Sherman et al. ( 28 ), but with 60 mg / l L -leucine.For URA3 counter selection, SC plates were supplemented with 1 g / l 5-fluor oor otic acid (5-FOA) and 30 mg / l uracil.To enhance the red phenotype of ade2 cells, SC medium with only 4 mg / l adenine was used.To pre v ent ade2 cells to be outcompeted by ADE2 cells, growth media were supplemented with 40 mg / l adenine.Galactose and raffinose solutions were sterilized by filtration and used at 2% final concentration.For growth on solid medium, 20 g / l agar was added.For selection on solid medium, plates were supplemented with 200 mg / l geneticin (G418), 100 mg / l nourseothricin (NTC) and / or 200 mg / l hygromycin (HYG) as indicated.Half the concentration of the respecti v e antibiotics were used in liquid cultivations.All strains were v alidated b y diagnostic colony PCR to ensure correct integra tion a t the intended chromosomal locus.For an e xperimental ov ervie w of CRI-SPA (CD) Donor strains construction, see Supplementary Figure S1.

Plasmid construction and PCR
Generall y, DN A assembl y was done by USER-fusion ( 29 , 30 ) employing uracil-specific excision reagent (USER ™) enzyme from New England Biolabs.PCR fragments were generated using Phusion U Hot Start DN A Pol ymerase (Thermo Fisher).Details for construction of individual plasmids are provided in Supplementary Methods S1.All plasmids and primers used in this study are presented in Supplementary Table S2 and Supplementary Table S3, respecti v ely.All plasmid maps are available for download as GenBank files.

Construction of yeast strains
All y east tr ansformations were carried out using the LiAc / SS Carrier DNA / PEG method described by Gietz and Schiestl ( 31 ).When antibiotic resistance genes were used as markers, cells w ere allow ed to recover for 2 hours in liquid YPD prior to plating on solid selecti v e media.Step 1 , a CD strain is pinned onto all strains of the library plates and incubated for mating.Step 2 , diploids are exhausted for glucose via growth on raffinose and selected for the marker in the library strains (AB1) and the marker on the CRI-SPA plasmid (AB2).In the diploid stage, the target site in the recipient strain is cleaved by Cas9-sgRNA.Repair of the resulting DNA DSB by gene conversion using the corresponding donor site in the CD strain as template transfers the genetic feature to the recipient locus.Note that the target site of the Cas9-sgRNA in the recipient chromosome is destroyed by insertion of the genetic feature of interest.Step 3 , galactose induced SPA eliminates the donor chromosomes.Step 4, haploid cells containing only recipient chromosomes are selected on 5-FOA.Step 5, the modified library strains are obtained by selecting for AB1, which marks the YKO deletion, and optionally, by selecting for the genetic feature (AB3).
U ni v ersal C RI-SPA strains, UC strains, were made by inserting a gene-targeting substrate containing cas9 and LEU2 into chromosomal integration site X-3 of W8164-2B ( MAT α) and W8164-2C ( MAT a ).The gene-targeting substrate was liberated from pHO8 by digestion with NotI (Thermo Scientific).The UC+:XII-5 ␣ strain (see Supplementary Figure S1 and Supplementary Table S1) was made from the UC strain (UC1 ␣) by inserting an additional copy of Kl URA3 between expression site XII-5 and the telomere.Insertion was achie v ed via co-tr ansfor ation with sgRNA plasmid pHO25 and a linear Kl URA3 repair substr ate liber ated from pCfB390 ( 32 ) using primers HOP181 and HOP182.C RI-SPA d onor (CD) strain were made from UC strains.CD-ade2 was constructed by first transforming the UC strain (UC1 ␣) with sgRNA plasmid pCfB2311 ( 33 ) and a linear repair substrate for disruption of ADE2 via insertion of the hphNT1 cassette, which was generated by PCR using plasmid pMEL12 and primers HOP64 and HOP65.The resulting strain was then further modified to harbor an additional Kl URA3 marker between the disrupted ADE2 gene and the telomere by transformation with sgRNA plasmid pHO22 and a linear Kl URA3 repair substrate amplified from pCfB390 by primers HOP175 and HOP176, yielding CD-ade2 .
CD-Btx was constructed from UC+:XII-5 ␣ by transforming with the integrati v e plasmid pBTX1 harboring the betaxanthin-synthesizing genes CYP76AD1 and DOD, an NTC selectable marker, negative feedback resistant AR O4 K229L , AR O7 G141S and an sgRN A expression cassette for targeting the XII-5 site.The BY-Btx strain was made by transforming the NotI excised insert of pBTX2, coding for CYP76AD1 and DOD and the negati v e feedback resistant AR O4 K229L , AR O7 G141S , into strain BY4741.
Introduction of CRI-SPA-hit knockouts into the BY-Btx strain was performed by replacing their Open Reading Frame (ORF) with a kanMX cassette ( 34 ).The kanMX cassette was amplified from pCfB2312 ( 32 ) and fused to targeting sequences up-and downstream of the relevant ORF.For the first round of KOs (DLD2, KGD2, MIP1, RCR2, YER084W, COX12, QCR10, SHY1, SGM1, VPS34, SSM4, UBC7, RPN4), the 400 bp up-and downstream targeting sequences were amplified using BY4741 genomic DNA as a template and by using primers with 40bp overhangs homologous to the kanMX cassette for amplification.The upstr eam, downstr eam, and kanMX amplicons were transformed into BY-Btx and fused in vivo by HR.For the second round of KOs (PRO2, VMA16, GET4, PRY1, STV1, YCR101C, YLR271W, GLR1, HEM25), the up and down homology sequences of the relevant ORFs were fused to the kanMX cassette by USER cloning.USER primers were used to generate PCR fragments of the up-and downstream sequences of the ORFs using genomic DNA from BY4741 as template.The relevant PCR fragments were USER-fused ( 29 )  coli and inserted into the USER cloning site of pCCM023, a modified version plasmid pCfB2909 lacking XII5-up and XII5-down homology sequences.The Up:: kanMX ::Dw gene-targeting substrates wer e r eleased from purified plasmids by digestion with NotI prior to transformation into BY-Btx.All transformant BY-Btx strains were selected on solid YPD medium containing 200 mg / l G418.All KOs were verified by PCR amplifying the upstream and downstream integration sites of the kanMX cassette.For both amplifications, a primer annealing outside and a primer annealing inside of the cassette were used.

CRI-SPA high-throughput pin replication protocol
Automatic pin replication was carried out using the high thr oughput pinning r obot ROTOR HDA fr om Singer Instruments (United Kingdom), along with replica pinning pads (RePads) and rectangular petri dishes (PlusPlates) from the same company.
For parallel mating of the CD Strain to the strains of the yeast deletion library, a single colony of the UDS was inoculated in 25 ml YPD supplemented with the appropriate antibiotic in order to maintain selective pr essur e for the gRNA plasmid.The UDS w as grown overnight, w ashed by centrifugation, and resuspended with 25mL sterile water to remove the antibiotic.A volume of 150 l of cell suspension was dispensed in wells of a 96-well plate serving as a source plate for the screen.From this source plate, the UDS was pinned in 1536 format on the mating plates.The arr ay ed str ains of the deletion collection were pinned on top of the UDS.Plate shuffling was introduced at this stage to address experimental artifacts (see below).Mating was allowed to proceed from 8 h to overnight.Strains were then pin replicated onto a series of selecti v e plates as outlined in Figure 1 .

Image acquisition and processing
The data from our screen was extracted from images acquired with a specialized imaging system (Phenobooth, Singer) keeping acquisition parameters fixed (lighting power, camera brightness, gain, exposure, hue, sa tura tion, white balances).To extr act y ellow color intensity, we applied a heuristic filter taking the geometric mean of Value and Sa tura tion of image pixels in the HSV color space (Supplementary Figure S2A).This assumed that the colony's Hue was fixed (i.e. a colony can only harbour variants of yellow) which was our observation on YPD media.
We repurposed the mask functions of an image analysis package ( 35 ), to assign pixels to colonies on a plate image.The number of pixels assigned to a colony was used as a measure of colony size.Pixels belonging to a colony were filtered and the average filtered colony pixel value was used as the yellow intensity score.

Randomisation of positional artifacts, outlier r emo v al, data normalization
We observed that colony size and yellowness scor e wer e subject to positional effects known to affect colony size in high density array screens ( 36 ).We adopted a colony position randomization strategy to randomize plate, agar quality and neighbor effects.This randomization was done when pinning the YKO library from its storage 384 format to the screen's 1536 quadruplicate format.Each one of the four 384 arrays from a gi v en library plate was pinned on a differ ent scr eening plate (Supplementary Figur e S2B).As a result, each colony quadruplicate was positioned on a differ ent scr een plate and exposed to randomized agar qualities and neighbor effects.Still, row-column colony positions were not altered by this randomization, and colonies wer e ther efor e still affected by non-random edge effects.These were corrected by setting the median of colonies in outer frames equal to that of center colonies ( 37 ).After edge corr ection, the scr een's data was center ed by subtracting the screen's mean from each data point and normalized by dividing all data points by the standard deviation of the screen.After this normalization, outliers within quadruplicates were detected with Grubbs' test for outliers and deleted.Genes r epr esented by two colonies or less were deleted from the anal ysis.Finall y, the mean of the colony replicate was used as the final score for each gene.

Combining the data of multiple screens
To combine the data of different screens, the edge-corrected data for each individual screen was normalized by subtracting the screen mean and by dividing colony scores by the screen standard deviation.At this point, the individual datasets were pooled and outliers removed with Grubbs' test.After outlier removal, mean yellow intensity and size of gene-deletion strains with three or more colonies was used as the final yield and fitness, respecti v ely, (pipeline and data can be accessed on the repository https://github.com/pc2912/CRI-SPA repo ).

Betaxanthin fluorescence quantification
Fluorescence in SC medium was measured by a Synergy Mx plate reader (Agilent BioTek) using excitation of 465 / 20 nm and emission of 525 / 20 nm as described previously by Savitskaya et al. ( 14 ).

Gene enrichment analysis and GO term graphs
Gene enrichment analysis was performed using GOA-TOOLS ( 38 , 39 ) to generate the gene groups indicated in Figure 5 .The default Benjamini-Hochberg correction was used to account for the number of tests, and a P-value of 0.05 was used as a threshold to accept GO terms as enriched.For visualizing results at a systems-le v el, a graph was constructed where nodes representing GO terms were connected if they shared associated genes.The size of the nodes was drawn as a function of the significance of the term enrichment and the edge opacity was drawn in relation to the number of genes shared between terms.Nodes wer e color ed according to their description or the description of their ancestors (i.e.descriptions contained key words such as 'mitochondria', 'translation', ect.).Finally, the resulting networks were built in networkX, transferred to Cytoscape ( 40 ) with py4cytoscape and minor manual adjustments were made for visibility (scripts available at https://github.com/pc2912/CRI-SPArepo ).

Statistical tests
Statisticall y significant overla ps of hits identified in screen A and screen B were determined by using a test based on hypergeometric distribution: p h( x ≥ k | M , n , N ).M is the number of genes quantified in both screen A and screen B, and n = 192 is the number of genes defined as hits in Screen A. N = 192 is the number of genes defined as hits in screenB.Finally, k is the number of overlapping gene hits between the two screens.Statistically significant increases (or decreases) in yellowness and fluorescence between BY-Btx and BY-Ref were determined by a one-sided t -test.Significant differences in yellowness score on solid YPD medium and fluorescence in SC medium between BY-Btx and re v ersed engineered hit mutants were identified with a one-sided Welshtests.

Basic components of the CRI-SPA system
CRI-SPA is based on four components: (i) U ni v ersal C RI-SPA strains (UC strains), see Supplementary Table S1, which are SPA strains ( 24 ) equipped with a cas9::LEU2 cassette, see Materials and Methods, in integration site X-3 ( 41 ).A SPA strain contains a Kl URA3 -GAL1 p cassette next to all of its sixteen centromeres; and all chromosomes of a SPA strain can ther efor e be destabilized on media containing galactose.Hence, in diploids where a SPA strain serves as one of the parent strains, SPA chromosomes can be selecti v ely eliminated on galactose medium followed by counterselection on 5-FOA ( 24 ).As a result, a haploid strain composed by chromosomes solely originating from the non-SPA parent strain is achie v ed.We hav e constructed UC strains of both mating types.(ii) CRI-SPA vectors, see Supplementary Table S2, which are selectable 2 based plasmids encoding a desirable gRNA.CRI-SPA v ectors enab le formation of Cas9-gRNA comple xes that serve two purposes.Firstly, they media te integra tion of a genetic feature of interest into a specific site in a UC strain; and secondly, they enable HR mediated transfer of the genetic feature during CRI-SPA (see below).In addition, the marker gene on the CRI-SPA vector facilitates selection for diploid cells during the CRI-SPA procedure, see below.(iii) C RI-SPA D onor strains (CD), see Supplementary,Table S1, which are UC strains that contain the genetic feature of interest in a defined chromosomal locus, a CRI-SPA vector encoding a gRNA matching the wild-type sequence of the modified locus, and a Kl URA3 gene between the genetic featur e of inter est and the telomer e of the modified chromosome.This additional Kl URA3 gene is used to counterselect undesir ed r ecombination e v ents that ma y accompan y Cas9 induced gene conversion, see below and Supplementary Figure S3.(iv) A CRI-SPA compatible recipient strain libr ary.The str ains in the libr ary need to contain a genetic marker allowing selection for diploid cells when they are mated to the CD strain.Moreover, in the current version of our CRI-SPA system, the library needs to be composed of strains that are ura3 as SPA depends on counterselection of Kl URA3 .In the present study, we have used the genome-wide YKO collection as the recipient library ( 26 ).The strains in this library are all ura3 Δ0 and the individual gene deletions are marked by kanMX ( 42 ), which can be selected by addition of G418 to the medium.

Experimental steps of the CRI-SPA method
The CRI-SPA procedure is performed in fiv e steps during six days (Figure 1 ).In step 1, The CD strain and the recipient libr ary str ains mate overnight on YPD medium to form diploid cells.In step 2, diploid cells are selected via a complementary marker setup for 48 h on solid raffinose.In the present study, this was achie v ed by replica-pinning the colonies from YPD to solid raffinose medium containing G418 and either nourseothricin (NTC) or hygromycin (HYG) to select for kanMX marked deletions delivered by the parental library strains and the natMX -or hphNT1based CRI-SPA plasmid deli v er ed by par ental CD strains, respecti v ely.In diploid cells, the Cas9-sgRNA CRISPR nuclease produces a DNA DSB at the recipient locus in the libr ary str ain chromosome, and this DNA DSB is r epair ed through HR using the corresponding modified locus of the CD strain as a repair template.Most commonly, repair is expected to proceed via synthesis-dependent strandannealing ( 43 ) resulting in desirable transfer of the genetic feature from the donor to the recipient chromosome by gene conv ersion.Howe v er, we note that undesirable chimeric donor / recipient chromosomes may be produced as the result of repair by gene conversion accompanied by a crossover or if repair is mediated by break induced replication ( 44 ).In CRI-SPA, these undesirable repair outcomes are counter-selected in step 4 (also see Supplementary Figure S3).The suboptimal and non-repressing carbon source raffinose ( 45 ) was used to exhaust glucose and to provide a slo w gro wth step thereby allo wing more time for CRISPR mediated gene transfer compared to growth on glucose.In step 3, the absence of glucose r epr ession allows for the sudden induction of the GAL1 promoter as the colonies are replica-pinned from raffinose to solid galactose medium ( 45 ).Acti v e GAL1 promoters disrupt the centromeres of all CD strain chromosomes , hence , inducing CD strain chromosome loss as the cells are dividing ( 24 ).In step 4, cells that have lost all donor chromosomes are selected by transferring the strains to solid SC medium containing 5-FOA.Note, this medium also counterselects undesired recombinant strains resulting from DNA DSB repair involving crossing-over or break-induced replication at the target locus (Supplementary Figure S3).As a result, step 4 generates haploid strains that solely contain the chromosomes originating from the recipient strain as well as the genetic feature of interest.In the final step 5, recipient cells are selected by the marker of the library, in this case by the kanMX marker of the YKO library to eliminate CD cells (if any) that did not mate, and which survi v ed kanMX and 5-FOA selection.Optionally, the genetic feature of interest may also be selected for if it is accompanied by a marker.This may eliminate unmodified recipient cells (if any) that have escaped the selection step for diploids.Conveniently, we note that CRI-SPA can be executed by hand or scaled with a pinning robot to e91 Nucleic Acids Research, 2023, Vol.51, No. 17 PAGE 6 OF 14 accommodate a range of budgets and throughputs.A more detailed scheme describing the fiv e CRI-SPA steps is provided in Supplementary Methods S2.

CRI-SPA donor strain construction and test for cas9 induced DNA DSB formation efficiency
CD strains are constructed in three simple steps, see Supplementary Figure S1.Firstly, an additional Kl URA3 marker is inserted into a locus between the integration site of the genetic feature of interest and the telomere.This is achie v ed in a CRISPR mediated process based on a gRNA plasmid expressing the a ppropriate gRN A. Secondl y, the resulting strain is cured of the gRN A plasmid.Finall y, the CD strain is obtained by inserting the genetic feature into the desired locus in a second CRISPR mediated reaction assisted by the CRI-SPA vector.
Successful CRI-SPA mediated allele transfer depends on highly efficient DNA DSB induction at the target locus by a CRISPR nuclease directed by a specific gRNA encoded from the CRI-SPA vector.Hence, prior to a CRI-SPA experiment, it is important to demonstrate the efficiency of a desired CRISPR nuclease / gRNA complex.This can be judged by performing a TAPE (Technique to Assess Protospacer Efficiency) experiment ( 46 ), which examines lethality of a UC strain after transformation with a CRI-SPA plasmid in the absence or presence of an efficient repair template.Compared to a reference plasmid, a CRI-SPA vector suitable for a CRI-SPA experiment needs to produce transformants at numbers that are se v eral fold higher in the presence than in the absence of a repair template, see Figure S4.Note that the TAPE test can conveniently be combined with CD strain construction by using a repair templa te tha t is designed to introduce the genetic feature of interest into the desired locus of the UC strain.

T r ansfer of ade2 ␦::hphNT1 into a set of arra y ed genedeletion mutants
To examine the potential of CRI-SPA, we first investigated whether it is possible to transfer a HYG selectable ade2 ::hphNT1 cassette from a CD-ade2 strain to a subset of mutants from the YKO library containing G418 selectable gene-deletion cassettes.Ad vantageousl y, ade2 Δ strains produce an easy-to-score red phenotype ( 47 , 48 ), which may be epistatically blocked by mutations upstream in the purine pathway, e.g.ade3 ( 49 ).Successful interchromosomal transfer of ade2 ::hphNT1 into YKO library mutants should ther efor e produce mostly r ed colonies, but should also be able to identify epistatic genetic interactions, like ade3 , which should appear as white colonies.Hence, we set out to investigate whether CRI-SPA would be able to quickly identify mutants that are epistatic to ade2 by screening for double mutants that form white colonies.
The CD-ade2 strain was constructed as part of a TAPE test demonstra ting tha t the NTC selectable ADE2 CRI-SPA vector (pHO24) is able to produce a gRNA that allows Cas9 to efficiently cleave the ADE2 locus (Supplementary Figure S4).Specifically, we obtained > 6-fold more transformants in the presence (915 transformants) than in the absence (145 transformants) of the ade2 ::hphNT1 cassette r epair template.Next, the r esulting CD-ade2 strain was used in a CRI-SPA experiment to screen the strains on plate 9 of the YKO library ( MAT a ).This plate contains 376 gene deletions, including the negati v e interaction control ade3 : : kanMX .The ade2 ::hphNT1 transfer experiment was performed as a 2 × 2 quadruplicate using 1536 format.In all growth steps, except the final one, media was supplemented with additional adenine to reduce the negati v e fitness effects due to the ade2 deletion ( 50 ).On the final plate, the medium contained reduced adenine le v els to speed-up de v elopment of red color, see MATERIALS AND METH-ODS.
After the CRI-SPA experiment, virtually all quadruplicates were red indica ting tha t the ade2 ::hphNT1 cassette was efficiently transferred to the vast majority of all mutant strains of the library, see Figure 2 .Only a few mutant strains did not produce any growth after CRI-SPA and such mutants may r epr esent strains that failed to go through CRI-SPA, e.g. if they failed to mate.For a few quadruplicates, some of the colonies appeared white / pinkish.Further r estr eaking on solid selecti v e media (G418 and HYG) demonstra ted tha t they contained a mix of cells that were able to form white and red cells.Moreover, cells forming white colonies were able to grow in the absence of adenine and diagnostic PCR demonstrated that their genotype was ADE2 and ade2 ::hphNT1 .Further, full genome sequence analysis of selected purified white strains showed that they were aneuploid or diploid strains containing two copies of the target chromosome of which only one has recei v ed a copy of the ade2 ::hphNT1 cassette, see Supplementary Figure S5.Formation of such cells may reflect that many strains in the gene deletion library suffer from aneuploidy ( 51 ).Some mutants may be particularly prone to this phenomenon and the most prominent case of a 'false whitish' phenotype was observed for scp160 , which is known to display ploidy instability ( 47 , 52-53 ).Alternati v ely, they may be deri v ed from diploid cells resulting from endoreduplication of the recipient genome, which may happen during SPA ( 27 ), see Supplementary Figure S6.Although formation of ADE2 / ade2 ::hphNT1 cells potentially could complicate the CRI-SPA analysis as they formed white cells, in particular as they display a fitness advantage over ade2 ::hphNT1 cells, this was generally not the case.Indeed, the ade2 ::hphNT1 ade3 : : kanMX quadruplet was easily identified as the only example of a quadruplet where all four colonies were purel y w hite, see Figure 2 .Hence, we belie v e that formation of this cell type is quite rare and they may likel y onl y impact experiments if they display a significant fitness ad vantage.Alto gether, we conclude tha t tha t the CRI-SPA method was able to ef ficiently transfer the ade2 ::hphNT1 cassette from a CD to a recipient library and that it was able to robustly identify the negati v e ade2 ::hphNT1 ade3 : : kanMX interaction.

Establishment of an efficient CRI-SPA based betaxanthin scr eening platf orm
We next sought to demonstrate the usefulness of the CRI-SPA technology in a metabolic engineering context.We attempted to map out the genetic r equir ements, at single gene resolution, for the production of betaxanthin, a plant metabolite deri v ed from the shikimate pathway.Betaxanthin is yellow and fluorescent; and has previously been used as a biosensor and a proxy for production of the morphine precursor L-DOPA in yeast ( 25 ).To test whether betaxanthin production could be meaningfully evaluated in a CRI-SPA screen using the YKO library, we firstly introduced a fiv e-gene betaxanthin-cassette (Btx-cassette) into BY4741, producing strain BY-Btx.The Btx-cassette harbors CYP76AD1 and DOD ( 25 ), which are required for production of betaxanthin, the dominant mutant genes AR O4 K229L , AR O7 G141S , w hich alleviate the negati v e feedback regulation of the shikimate pathway ( 25 ); ( 54 ), as well as the selectable natMX marker.As expected for a strain producing betaxanthin, BY-Btx, unlike the r efer ence strain BY-Ref (BY4741), produced yellow colonies (Figure 3 A); and liquid cultures were yellow and fluorescent (Figure 3 B).Importantly, betaxanthin production did not appear to influence fitness, as BY-Btx displayed essentially the same growth rate as BY-Ref Supplementary Figure S7.
We next tested whether significant differences in betaxanthin production in the YKO library background could be scored in a simple colon y-based assa y suiting the format and throughput of a CRI-SPA experiment.In our hands, the colony color phenotype was the easiest to score with a color filter, extracting color intensity from RGB pixel values, see 'Image acquisition and processing' in Materials and Methods and Supplementary Figure . S2A).Using this method, we demonstrated that the yellowness-score of colonies obtained with the betaxanthin producing strain was around 5-fold higher ( P ≤ 10e-10) than the score obtained with colonies of the r efer ence strain, see Figure 3 A. We also examined production of betaxanthin in liquid SC-medium by measuring betaxanthin fluorescence in a plate-reader assay using a method previously described by Savitskaya et al. ( 14 ).In this setup, the BY-Btx strain was 6-fold ( P ≤ 0.007) mor e fluor escent than the r efer ence strain, see Figur e 3 B, demonstra ting tha t the method could be used as a complementary assa y f or char acterizing specific Btx-hit str ains identified by yellow color de v elopment on solid medium.In summary, we have established a readily scalable image analysis method able to quantify betaxanthin production based on yellow color intensity; and we note that yellow color and fluorescence are widely accepted betaxanthin detection methods ( 14 , 25 , 55 ).
To set the stage for a genome-scale betaxanthin CRI-SPA experiment, we firstly constructed a betaxanthin CD (CD-Btx) that harbors the NTC selectab le fiv e-gene Btxcassette in the XII-5a expression site, a hphNT1 selectable CRI-SPA plasmid (pHO29), which encodes a gRNA that efficiently targets empty XII-5a sites ( 56 ), and a Kl URA3 marker positioned between XII-5a and the telomere.Secondly, gi v en the considerab le size, 9.3kb, of the Btx-cassette we performed a pilot experiment to assess how efficient it is transferred from the CD-Btx strain to the mutant strains of plate 9 of the YKO library, which we also used in the ade2::hphNT1 transfer experiment described above.Lastly, as the Btx-cassette, unlike ade2 ::hphNT1 , does not confer a significant fitness defect in BY-Btx the YKO reference strain, we decided to explore the possibility of transferring the Btx-cassette without selecting for its natMX marker.Hence, we performed the pilot CRI-SPA Btxcassette transfer experiment in quadruplicate in the absence of nourseothricin.Out of the 376 strains on the library plate, 373 strains ( > 99%) produced colonies after CRI-SPA, and 367 ( > 97%) produced all four colonies, see Supplementary Figure S8.Encouragingly, the vast majority of colonies were yellow indicating highly efficient transfer of the Btx-cassette despite that we did not a ppl y nourseothricin for its selection during any of the steps in the CRI-SPA procedure.
CRI-SPA colonies are not clones, and str ain-to-str ain betaxanthin production comparisons can ther efor e be compromised if some colonies are contaminated with significant amounts of other cell types producing le v els of betaxanthin that are different as compared to the le v els of the modified recipient mutant cells.Since we did not select for the Btx-cassette in the final CRI-SPA step, we rea-e91 Nucleic Acids Research, 2023, Vol.51, No. 17 At the top, a scheme illustrating the Btx-cassette and how it is transferred from the CD strain to a library recipient strain.To the left, diagrams with color codes indicating selection applied at the different CRI-SPA steps.G418 selects for the kanMX marker in the gene-deletion cassettes of the YKO library, HYG selects for the CRI-SPA plasmid, NTC selects for the Btx-cassette, and 5-FOA counter-select donor strain chromosomes, which all contain Kl URA3 markers.( D ) In the middle, yellow intensity ranking of all YKO mutant strains after CRI-SPA mediated Btx-cassette transfer.To the left, mosaic images of the four individual replicates obtained with selected YKO mutants displaying higher , similar , or lower yellowness as compared to BY-Btx.Mutants in red frames were purified and restreaked on solid YPD medium, see images to the right.soned that the largest source of error would be contributed by white unmodified recipient cells, which were not eliminated during CRI-SPA steps 2-3, and which would reduce the yellowness score of the mutant colony.To explore the extent of this potential problem, we picked cells from each of the four colonies obtained for each strain on the final CRI-SPA screen plate (YPD-G418) and replica spotted a pproximatel y 1000 (two of the colonies) and 10 000 cells (the other two colonies) on YPD + G418 (control) and YPD + G418 + NTC plates, see Supplementary Figure S8.Contaminating unmodified YKO cells should form white colonies on YPD + G418, but not on YPD + G418 + NTC medium.For 360 of the strains (96%), sufficient viable cells wer e transferr ed ( > 100 colonies produced in three of the four trials) to allow for further analysis.Next, we qualitati v ely compared the number of colonies formed on YPD + G418 and YPD + G418 + NTC plates for all trials.This analysis identified eight strains ( ∼2% of the 360 strains), which in one or more trials contained > 1% unmodified YKO library cells in the colonies of the final CRI-SPA screening plate.Most of these colonies contained little if any modified YKO cells and these trials ther efor e r epr esent rare flawed CRI-SPA transfer trials.We note that for six of these eight strains ≥ two of the four trials were not contaminated with unmodified YKO cells.Accordingly, we conclude that transfer of unmodified YKO cells through the CRI-SPA procedure is a minor problem.
We also reasoned that yellow CD-Btx and diploid cells, which survi v ed CRI-SPA steps 4-5, would compromise strain-to-strfain betaxanthin production comparisons if they were present in large amounts in the CRI-SPA colonies.We ther efor e examined the final CRI-SPA plates for colonies that contained CD-Btx and were diploid.In parallel with the experiment presented above, cells were ther efor e spotted in the same way, but on solid SC-Ura medium, which allows CD-Btx and diploid cells to grow, and on solid SC-Ura + G418 medium, which allows diploid, but not CD-Btx, cells to grow.These analyses demonstrated that basically no CD-Btx cells survi v e the CRI-SPA screen.Some diploid cells survi v e, but they constitute less than 0.1% of the cells in the colonies.For a few strains, one of the four trials contained ∼1% diploid cells.Hence, we conclude that carryover of CD-Btx and diploid cells to the final CRI-SPA plate constitutes a minor problem in this CRI-SPA experiment.Lastly, we note that the scp160 strain, which produced a false phenotype in the ade2 ::hphNT1 cassette transfer experiment failed to produce more than a fe w survi ving cells after CRI-SPA in this e xperiment.

Introduction of the betaxanthin pathway into a genome-wide library of gene-deletion strains
With the CRI-SPA based betaxanthin-pathway transfer system and a betaxanthin quantification method in place, we then used CD-Btx to transfer the Btx-cassette to the 4787 unique deletion strains of the YKO library using CRI-SPA in quadruplicate amounting to a total volume of 19148 transfer e v ents.To address positional effects known to affect colony size in high density array screens ( 36 , 37 ), we r andomized the YKO libr ary arr ays (in 384 format) when scaling up to 1536 quadruplicate format.Specifically, each of the four 384 arrays from a gi v en library plate were pinned on a different screening plate (see Supplementary Figure S2B and Materials and Methods for full description of the image analysis, colony randomization and data processing pipeline).
In the first set of experiments, we employed NTC selection in the final step of CRI-SPA as a safety measure to ensure that all cells contain the Btx-cassette.Encouragingly, the vast majority of the colonies appeared yellow after CRI-SPA indica ting tha t the CRI-SPA media ted transfer of the betaxanthin pathway into the library was successful (Figure 3 C).We extracted the yellowness score of all colonies and after removing remaining positional artifacts and outliers (see MATERIALS AND METHODS including accompanying python scripts available on our CRI-SPA repository), 4224 strains ( ≈ 88% of the library) produced ≥ three viable CRI-SPA colonies and were henceforth included in our analysis.
We finally ranked the mean yellowness for all genes (Figure 3 D) and found that the screen produced a strong yellowness signal to noise ratio (the Z -scores of the 16 top and bottom hits was 40.83, -6.93; respecti v el y.).This anal ysis demonstra ted tha t our screen can produce strains with a large range of betaxanthin synthesis abilities.
Since CRI-SPA colonies are not monoclonal, we isolated thr ee candidates r epr esenting thr ee differ ent color le v els and r estr eaked them on solid YPD medium with appropriate selection.As expected, the individual streaks of cells formed colonies that were uniformly yellow indicating little, if any, presence of contaminating cell types.Importantly, the color differences between the sets were easily distinguished on pla te images, valida ting the position of the mutant strains in the color ranking scheme (Figure 3 D).

Screen to screen reproducibility and marker-free CRI-SPA
Confident that the screen readout was accurate for a handful of strains, we repeated the entire CRI-SPA experiment to validate its reproducibility.After CRI-SPA, the yellowness of the individual mutants obtained in this experiment, trial 2, were then plotted against the yellowness obtained for the mutants in the first experiment, trial 1.The clear linear correlation between the two trials (PCC = 0.765) demonstrates the reproducibility of CRI-SPA (Figure 4 A).Moreover, we found that, from all successfully labeled genes in the two screens, 86% ( P = 1.09e-220) and 52% ( P = 9.97e-83) of the 192-top and 192-bottom hits were common in the two trials, respecti v el y.Hence, and especiall y for the top-hits, reproducibility was high.
In a third experiment, trial 3, we explored the possibility of transferring the fiv e-gene Btx-cassette from CD-Btx to the full YKO library without selecting for the natMX marker, which is present in the Btx-cassette.In this screen, the CRI-SPA plasmid's restriction of the target site maintained the sole selection for genetic transfer.With this setup, 95.1% of mutants produced at least three colonies after CRI-SPA mediated transfer of the Btx-cassette.Importantly, the yellowness score in this trial had a clear linear correlation with trial 1 (PCC = 0.768, Figure 4 B).Moreover, when the 192 most yellow and 192 least yellow mutants obtained in the two trials wer e compar ed, the top and bottom groups had 75% ( P ≤ 10e-180) and 44% ( P ≤ 10e-69) overlap across scr eens, r espectively (Figur e 4 B).Hence, the above analyses demonstrate that CRI-SPA is reproducible and that the combined effect of a mutation and a genetic feature can be reproduced e v en without selecting for this feature.

Systems perspective of betaxanthin production
In contrast to one-pot screens, CRI-SPA produces a readout for each individual strain within a library.We investigated whether this systematic gene labeling could re v eal ne w system le v el patterns in betaxanthin production.For this analysis, we combined the data of four screens (see supplementary part), obtaining a CRI-SPA score for 4761 genes out of the 4788 present in our library with a median of 15 replicates (16 is maximum) per gene.We first plotted yellowness of all strains against colony size using the latter feature as a proxy for fitness, see Figure 5 .Plotted in this way, the vast majority of cells map in the close vicinity of the WT, indica ting tha t the muta tions in these strains do not impact betaxanthin production (Figure 5 ).We next grouped the remaining mutant strains according to their yellowness score: the 'yellow' group was defined as mutants with the highest le v el of betaxanthin production (1.2 standar d de viation greater than the screen mean, yellow points in Figure 5 ).The 'white' group was defined as mutants with the lowest e91 Nucleic Acids Research, 2023, Vol.51, No. 17 The results of Gene Ontology Enrichment Analyses on se v eral subgroups of genes are shown as graphs for the bottom hits (white, 1.2 std below the screen mean) and top hits (yellow, 1.2 std above the screen mean).As indicated by the bottom right scale, node size indicates the significance of GO terms enrichment, edge transparency indicates the number of shared genes between GO terms.The cyan group did not show any term enrichment.The position of the BY-Btx strain is marked with a black asterisk.le v el of betaxanthin production (1.2 standard deviation below the screen mean, white points in Figure 5 ).We observed that most members of the yellow group displayed a strong fitness penalty.We ther efor e defined a third group of mutants, the 'cyan group', see Figure 5 .This group contains mostl y highl y y ellow str ains with little or no fitness penalty.Ne xt, we e xamined whether specific cellular functions could be linked to betaxanthin le v els by running gene ontology enrichment analyses ( 38 ) within the three groups.The cyan group did not produce any enrichment.This may not be surprising as the hits in this group were the least reproducible (Supplementary Figure S9).The white group, despite its broad fitness range, was enriched for terms describing the degr adation machinery, tr anslation, and vacuolar mechanisms.Most strikingly, the yellow group displayed a dramatic enrichment for genes associated with mitochondrial functions, particularly the respiratory chain, which might explain the fitness penalty observed in this group.Hence, this analysis demonstrates that CRI-SPA is able to identify a no vel netw ork of genes, which impact functions of a specific organelle in the cell, and which, when deleted, lead to reduced fitness and high betaxanthin le v els.

Validation of CRI-SPA hits in liquid culture
Curr ently, CRI-SPA scr eens for phenotypic changes on solid medium and it is possible that a phenotypic hit may potentially vanish in different growth conditions.To explore that possibility, we decided to test how selected Btxhit strains fare in liquid SC-medium.Moreover, to further valida te tha t strains picked up in a CRI-SPA screen truly reflect the impact of the specific mutation provided by the YKO library strain, we re-introduced the mutations in the BY-Btx r efer ence strain by conventional one-step genetargeting ( 57 ) using kanMX to select for the deletion.Hence, we reintroduced 16 top and 6 bottom KO hits in the reference strain BY-Btx, covering a variety of cellular functions and strain fitnesses (Figure 6 A, Supplementary Table S4).To avoid penalizing slow growing strains, and strains with a prolonged lag-phase, we grew the reference and the 22 mutant BY-Btx strains in liquid SC medium with 2% glucose for 7 days.This setup is possible as betaxanthin does not appear to be catabolized during fermentation as its le v els remained stable throughout the stationary phase of a BY-Btx pilot experiment (Supplementary Figure S7).
We then examined all strains for betaxanthin production using a fluorescence readout at the fermentation end-point.The results showed that betaxanthin le v els strongly correlated with the yellowness score obtained for the same mutants on the final CRI-SPA selection plate (PCC = 0.680, see Figure 6 B).For the positi v e hits, of the 13 scoring significantly higher than the WT for yellowness in the CRI-SPA scr een, nine wer e significantly mor e fluor escent than the r efer ence BY-Btx strain, indicating that they also produced significantly increased le v els of betaxanthin in liquid SC medium.Indeed, the two most yellow mutant strains identified in the CRI-SPA screen, ubc7 and ssm4 / doa10 , were 2.6-fold and 2.2-fold more fluorescent than BY-Btx ( ssm4 / doa10 : P ≤ 6 × 10e-4, ubc7 : P ≤ 2 × 10e-3), respecti v ely.Another set of positi v e CRI-SPA Betaxanthin hits were composed by mip1 , shy 1 , YER084W , and cox12 .The corresponding genes all encode mitochondrial proteins showing that dysfunctional mitochondria may also stimulate betaxanthin production in liquid SC medium.For the negati v e hits, the hem25 hit was the only strain displaying significantly r educed fluor escence compar ed to BY-Btx (0.59-fold, P ≤ 0.05).Taken together, we show that CRI-SPA is able to identify gene hits, which perform in growth conditions routinely used for metabolic engineering strain de v elopment.

DISCUSSION
We have developed a highly efficient, low cost and rapid method, CRI-SPA, for parallelized transfer of a genetic feature from a haploid CD strain to any haploid yeast library strain, which is ura3 .In this report, we have focused on gene transfer into the YKO library, but we note that other libraries like the GFP library ( 57 ) and SWAT libraries ( 15 , 16 ) could also be used in CRI-SPA experiments.At the individual le v el, CRI-SPA media ted gene transfer is highly ef ficient and can, unlike SGA, be performed without selecting for the genetic feature of interest.At the global le v el, the efficiency of transfer is also very high and only 62 genes (1.3%) of the library consistently failed to produce any colonies through three screen repeats.As expected, these genes are associated with mating, cell division, and DNA r epair, which r epr esent functions that ar e essential for CRI-SPA (Supplementary Table S7).To this end, we stress that CRI-SPA mediated gene transfer is based on interchromosomal gene conversion.Hence, the donor sequence persists in the diploid state where gene transfer takes place allowing for a long tunable window of opportunity for transfer.This contrasts the scenario in classical gene targeting where the genetic feature is introduced via a linear DNA donor frag-ment or a self-cleaving plasmid system deli v ered by a SPA based method ( 16 , 58-59 ).In those cases, the time window where gene transfer can take place is limited to the lifespan of the donor fragment.
We note that CRI-SPA colonies are not clones and that the colonies may contain other cell types than modified recipient cells.Hence CRI-SPA colonies may potentially contain escapers, unmodified recipient-, CD-and diploid cells on the final selection plate.In the betaxanthin experiment w hen a ppl ying selection for the Btx-cassette, these cell types wer e pr esent in amounts lower than 1% and did not influence the readout significantly.In the corresponding experiment where we did not a ppl y selection for the Btx-cassette, we observed a few cases where large amounts of unmodified recipient strains survi v ed step 2 where diploids are selected.These cases were typically flawed CRI-SPA transfer trials, and we envision that they may form if unusually high amounts of recipient cells are pinned from the starting library plate to the plate where mating takes place.If at the same time, mating is poor, diploid cells may not outcompete unmodified recipient strain on the plate selecting for diploids despite their advantage of selection.We stress that all four trials are rarely tainted by contaminating cells, and if two rounds of CRI-SPA are performed, each with quadruplicate trials, we belie v e that a recipient library can be reliab ly cov ered for ≥99% of the strains.The exceptions will be if the genetic feature of interest introduces a strong fitness disadvantage as this will likely increase the number of undesired other cell types in the CRI-SPA colonies.In these cases, the selection scheme may need further optimization.In the pilot ade2 : :hphNT1 transfer experiments, where introduction of this mutation comes with a fitness penalty, we added additional adenine to the medium to reduce this effect.Ne v ertheless, a fe w colonies were still contaminated by white cells to artificially impact the colony color phenotype.Some of these cells turned out to be diploid recipient YKO cells where only one of the homologous chromosomes had recei v ed the ade2 :: hphNT1 cassette.Although it may be difficult to avoid occasional formation of diploid recipient cells due to endoreduplication, the effect of these cells on the colony phenotype will be minimized if both chromosomes recei v e the genetic featur e of inter est.This e91 Nucleic Acids Research, 2023, Vol.51, No. 17 PAGE 12 OF 14 can be achie v ed by optimizing the gRNA to ensure that all target chromosomes are cut in all cells.Consequently, we recommend verifying Cas9-sgRNA cutting efficiency with TAPE, and we hav e pre viously observ ed fold-differences in transformant numbers in the presence and absence of a repair template that exceed 400 ( 46 ).Another way to ensure that both sister chromatids of the recipient strain will recei v e the modification is to prolong the window of opportunity for Cas9 cleavage.This can easily be achie v ed by extending the duration of the raffinose step where little cell growth takes place.Another source of contaminating cells could in principle be aneuploid cells containing CD chromosomes where the Kl URA3 markers were mutated.The forward mutation rate of endogenous URA3 selecting for 5-FOA resistance is around 5 × 10 −8 ( 60 ).Assuming that the forward muta tion ra te of Kl URA3 markers are similar to the rate of URA3, we find the presence of such cells highly unlikel y.Lastl y, we note that the betaxanthin phenotype of the hit strains may be caused by other accidental genetic alterations in the recipient strain, w hich potentiall y could have been introduced during construction of the gene deletion or during the CRI-SPA procedure.In our Btx-Cassette transfer experiment, we successfully demonstrated that Btxhit strains maintained their phenotype if they were recreated de novo by conventional gene tar geting, sho wing that the phenotype of the hit strain was caused by the specific mutation of the YKO library strain (Figure 6 ).
In principle, CRI-SPA can be used as a tool to forward the understanding of any biological process in yeast as it allows a genetic feature to be combined with large libraries of other mutations.In this report, we demonstrate in a pilot experiment that CRI-SPA can be used to identify epistatic interactions in a genetic network.Specifically, we identified the epistatic interaction between ade2 and ade3 in a set of other mutant strains (Figur e 2 ).Mor e thoroughly, we demonstrated how CRI-SPA can be used to de v elop a metabolic-engineering strategy for production of the plant metabolite betaxanthin by identifying target genes that can be manipulated to increase betaxanthin production.This was achie v ed by using CRI-SPA to transfer a fiv e-genes cassette for betaxanthin production into the YKO library.Our successful transfer of the Btx-cassette to close to all of the strains of the YKO library strains allowed us to provide a systems ov ervie w of functions that influence betaxanthin production in yeast.Importantly, our CRI-SPA screen identified genes that are strongly enriched for GO-terms forming functional networks supporting a mechanistic consensus across hits.For example, our analyses showed that deletion of genes, UBC7 and SSM4 / DOA10 , involved in ubiquitin mediated endoplasmic-reticulum-associated protein degradation, ERAD ( 61 ), benefit betaxanthin production.In this context it is important to note that the betaxanthin pathway includes a p450 enzyme, CYP76AD1, which is N-terminally linked to the ER membrane.P450s are notoriously difficult to produce in fungal production hosts and it is tempting to speculate that folding of CYP76AD1 in yeast is inefficient and ther efor e prone to ERAD.In this model, increased betaxanthin le v els are the result of the higher CYP76AD1 activity that will materialize in the absence of functional ERAD.From a gene-network perspecti v e, we note that Ubc7 is a ubiquitin conjugating enzyme, w hich is C-terminall y tail-anchored to the ER membrane ( 62 ).This fact may explain why get4 is also scored as a positi v e hit as Get4 is part of the GET complex that integrates C-terminal-tail-anchored proteins into the ER membrane ( 63 ).In another part of the gene network connected by CYP76AD1 we find genes encoding mitochondrial proteins.For example, deletion of HEM25 negatively impacts betaxanthin production.Hem25 is a mitochondrial transporter facilitating the import of glycine, a precursor for the biosynthesis of heme.Since heme is an essential cofactor of CYP76AD1 ( 14 , 64 ); it is easy to envision that reduced heme availability might impair the activity of CYP76AD1, and in turn betaxanthin pr oduction.Fr om another point of view, we also stress that our CRI-SPA screen, in addition to wellcharacterized genes, was also able to pick up Btx-hit genes with no assigned function.Such gene hits would be difficult, if not impossible, to identify via traditional metabolicmodel based hit-finding strategies.
Since our betaxanthin CRI-SPA screen was designed to facilitate future betaxanthin cell factory de v elopment, it was important to validate that our CRI-SPA betaxanthin hits identified on solid medium would maintain their phenotype if they were propagated in liquid medium, which is most often used in production.In favor of using CRI-SPA as a valuable starting point in the development of a metabolic engineering strategy, most of the r ecr eated betaxanthin hits influenced betaxanthin production in a manner similar to the original effects observed for the corresponding CRI-SP A colonies.Altogether, CRI-SP A was able to unrav el nov el physiology impacting betaxanthin production and to identify gene targets that can be used to de v elop a metabolic engineering strategy for production of betaxanthin, and potentially of L-DOPA deri v ed compounds.For the present, and for other metabolic engineering projects, we envision that by examining gene-interaction networks identified by a CRI-SPA screen, it will be possible to identify mutations that can be ad vantageousl y combined.In particular, muta tions tha t are not part of the same functional network may be combined to provide additi v e, or e v en synergistic, effects on production of the desired product.We note that not only the positi v e hits can be used in a future stra tegy, nega ti v e hits may point to limiting functions, which may be improved by ov ere xpressing the hit gene.
Our successful marker-free gene-transfer experiment opens promising avenues for CRI-SPA.Bypassing the need for marker selection sets the stage for the transfer of point mutations as well as the multiplex transfer of several genetic fea tures loca ted a t dif ferent positions in the genome at once.In another application, CRI-SPA could be used to transfer a genetic trait to libraries with different genetic backgrounds or e v en closely related species, as long as they can mate with the CD.This is possible as CRI-SPA is independent of meiotic recombination and sporulation.Currently, a limitation of our method is that a recipient library needs to be ura3 to be compatible with the CRI-SPA selection procedure.This could be bypassed by replacing the Kl URA3 markers in the CD strain by another counter-selectable mark er lik e amdS ( 65 ) to produce an e v en more fle xib le v ersion of CRI-SPA.
To conclude, we have demonstrated that CRI-SPA is a powerful high-throughput data generation tool opening up PAGE 13 OF 14 Nucleic Acids Research, 2023, Vol.51, No. 17 e91 a range of new exciting applications both in fundamental genetics and in genome engineering.

DA T A A V AILABILITY
The top hits, bottom hits and failing genes lists are available as Supplementary Tables 5, 6 and 7, respecti v ely.A sample of 13 screen images, the raw datasets obtained from such images for each screen, the individual and the corrected + filtered datasets are all available on our CRI-SPA github r epository ( https://github.com/pc2912/CRI-SPAr epo ).All scripts used in the analysis and for the drawing of the figures of this work are also available on the CRI-SPA github repository (permanent doi: https://doi.org/10.5281/zenodo.8160760 ).

Figure 1 .
Figure 1.The CRI-SPA gene-transfer system.( A ) Left: The CRI-SPA Donor (CD) strain and the recipient library strains (here YKO).Right: Graphics of the individual genetic components are shown in the legend box.The YKO strains contain a specific deletion marked by the antibiotic marker 1 (AB1) and a target site for the Cas9-sgRNA CRISPR nuclease.The CD strains contain: Kl URA3-GAL1p cassettes at the centr omeres of all chr omosomes , a cas9 gene , a selectable CRI-SPA plasmid maintained by antibiotic marker (AB2) and encoding a sgRNA targeting the insertion site of the genetic feature (blocked in the CD), a genetic feature of interest coupled to an antibiotic marker, AB3 (optional), and a Kl URA3 gene between the genetic feature of interest and the telomere.( B ) The individual steps of the CRI-SPA procedure (see main text for details).Step 1 , a CD strain is pinned onto all strains of the library plates and incubated for mating.Step 2 , diploids are exhausted for glucose via growth on raffinose and selected for the marker in the library strains (AB1) and the marker on the CRI-SPA plasmid (AB2).In the diploid stage, the target site in the recipient strain is cleaved by Cas9-sgRNA.Repair of the resulting DNA DSB by gene conversion using the corresponding donor site in the CD strain as template transfers the genetic feature to the recipient locus.Note that the target site of the Cas9-sgRNA in the recipient chromosome is destroyed by insertion of the genetic feature of interest.Step 3 , galactose induced SPA eliminates the donor chromosomes.Step 4, haploid cells containing only recipient chromosomes are selected on 5-FOA.Step 5, the modified library strains are obtained by selecting for AB1, which marks the YKO deletion, and optionally, by selecting for the genetic feature (AB3).

Figure 2 .
Figure 2. Transfer of the ade2 ::hphNT1 cassette from a CD to plate 9 of the YKO library.CRI-SPA mediated transfer of the ade2 ::hphNT1 cassette to the mutant strains of plate 9 of the YKO library was performed in quadruplicate.The final CRI-SPA plate was photo gra phed and shown in the middle.Selection applied at the different CRI-SPA steps are indicated to the left by the color code.G418 selects for the kanMX marker in the gene deletion cassettes of the YKO library, NTC selects for the CRI-SPA plasmid, HYG selects for the ADE2 disruption cassette ade2 ::hphNT1 , and 5-FOA counter-select donor strain chromosomes, which all contain Kl URA3 markers.The four ade2 ::hphNT1 scp160 : : kanMX and the four ade2 ::hphNT1 ade3 : : kanMX CRI-SPA colonies are framed by blue squares.Magnifications of the ade2 ::hphNT1 ade3 : : kanMX quadruplets are shown to the right as indicated.

PAGE 8 OF 14 Figure 3 .
Figure 3. Betaxanthin production in YKO strains.( A ) Left: images of BY-Ref and BY-Btx colonies in visible light (top) and corresponding filtered images (bottom).Right: quantification of yellowness of BY-Ref and BY-Btx colonies by image analysis, see main text and Supplementary Figure S2 for details. ( B ) Left: image of BY-Ref and the BY-Btx strains in liquid SC medium.Right: quantification of fluorescence produced by BY-Ref and the BY-Btx strains in liquid SC medium.( C ) Final plate of a CRI-SPA Btx-cassette transfer experiment.At the top, a scheme illustrating the Btx-cassette and how it is transferred from the CD strain to a library recipient strain.To the left, diagrams with color codes indicating selection applied at the different CRI-SPA steps.G418 selects for the kanMX marker in the gene-deletion cassettes of the YKO library, HYG selects for the CRI-SPA plasmid, NTC selects for the Btx-cassette, and 5-FOA counter-select donor strain chromosomes, which all contain Kl URA3 markers.( D ) In the middle, yellow intensity ranking of all YKO mutant strains after CRI-SPA mediated Btx-cassette transfer.To the left, mosaic images of the four individual replicates obtained with selected YKO mutants displaying higher , similar , or lower yellowness as compared to BY-Btx.Mutants in red frames were purified and restreaked on solid YPD medium, see images to the right.

Figure 4 .
Figure 4. Reproducibility of betaxanthin CRI-SPA experiments in the presence and absence of pathway transfer selection.( A ) The mean yellow color intensities obtained for each mutant in repeat 1 and repeat 2, which both were performed in the presence of natMX to select for the Btx-cassette genes.( B ) The mean yellow color intensities obtained for each mutant in trial 1 and trial 3. Trial 3 was performed in the absence of natMX .The top and bottom 192 genes and their overlap across screens are colored and displayed in flanking Venn diagrams.Insets show the correlation for colony fitness across screens.

PAGE 10 OF 14 Figure 5
Figure5 Betaxanthin yield and fitness r elationship r e v eals mechanistic patterns in betaxanthin production.Colony size (i.e.fitness) vs Yellow Color intensity (i.e.Betaxanthin Yield).The results of Gene Ontology Enrichment Analyses on se v eral subgroups of genes are shown as graphs for the bottom hits (white, 1.2 std below the screen mean) and top hits (yellow, 1.2 std above the screen mean).As indicated by the bottom right scale, node size indicates the significance of GO terms enrichment, edge transparency indicates the number of shared genes between GO terms.The cyan group did not show any term enrichment.The position of the BY-Btx strain is marked with a black asterisk.

PAGE 11 OF 14 NucleicFigure 6 .
Figure 6.CRI-SPA hits identified on solid medium produce enhanced amounts of betaxanthin in liquid medium.( A ) selected Btx-hit str ains y ellow and fitness scores within the screen.( B ) Correlation between CRI-SPA yellowness score and fluorescence of re v ersed engineered hits after 7 days cultivation in liquid SC medium.Error bars show standard deviations for 16 and 3 replicates for CRI-SPA and liquid Fluor escence, r especti v ely.Mar kers coloring scheme, below inset, is defined by the agreement between CRI-SPA scor e and fluor escence in liquid medium, and significance differ ence from WT r eported by a one-sided Welch's test.
to the kanMX cassette in Esc heric hia