Engineering the thermotolerant industrial yeast Kluyveromyces marxianus for anaerobic growth

Current large-scale, anaerobic industrial processes for ethanol production from renewable carbohydrates predominantly rely on the mesophilic yeast Saccharomyces cerevisiae. Use of thermotolerant, facultatively fermentative yeasts such as Kluyveromyces marxianus could confer significant economic benefits. However, in contrast to S. cerevisiae, these yeasts cannot grow in the absence of oxygen. Response of K. marxianus and S. cerevisiae to different oxygen-limitation regimes were analyzed in chemostats. Genome and transcriptome analysis, physiological responses to sterol supplementation and sterol-uptake measurements identified absence of a functional sterol-uptake mechanism as a key factor underlying the oxygen requirement of K. marxianus. Heterologous expression of a squalene-tetrahymanol cyclase enabled oxygen-independent synthesis of the sterol surrogate tetrahymanol in K. marxianus. After a brief adaptation under oxygen-limited conditions, tetrahymanol-expressing K. marxianus strains grew anaerobically on glucose at temperatures of up to 45 °C. These results open up new directions in the development of thermotolerant yeast strains for anaerobic industrial applications.

In terms of product volume (87 Mton y -1 ) 1,2 , anaerobic conversion of carbohydrates into ethanol by the 33 yeast Saccharomyces cerevisiae is the single largest process in industrial biotechnology. For 34 fermentation products such as ethanol, anaerobic process conditions are required to maximize product 35 yields and to minimize both cooling costs and complexity of bioreactors 3 . While S. cerevisiae is applied in 36 many large-scale processes and is readily accessible to modern genome-editing techniques 4,5 , several 37 non-Saccharomyces yeasts have traits that are attractive for industrial application. In particular, the high 38 maximum growth temperature of thermotolerant yeasts, such as Kluyveromyces marxianus (up to 50 °C 39 as opposed to 39 °C for S. cerevisiae), could enable lower cooling costs 6-8 . Moreover, it could reduce the 40 required dosage of fungal polysaccharide hydrolases during simultaneous saccharification and 41 fermentation (SSF) processes 9,10 . However, as yet unidentified oxygen requirements hamper 42 implementation of K. marxianus in large-scale anaerobic processes 11-13 . 43 In S. cerevisiae, fast anaerobic growth on synthetic media requires supplementation with a source of 44 unsaturated fatty acids (UFA), sterols, as well as several vitamins [14][15][16][17] . These nutritional requirements 45 reflect well-characterized, oxygen-dependent biosynthetic reactions. UFA synthesis involves the oxygen-46 dependent acyl-CoA desaturase Ole1, NAD + synthesis depends on the oxygenases Bna2, Bna4, and Bna1, 47 while synthesis of ergosterol, the main yeast sterol, even requires 12 moles of oxygen per mole. 48 Oxygen-dependent reactions in NAD + synthesis can be bypassed by nutritional supplementation of 49 nicotinic acid, which is a standard ingredient of synthetic media for cultivation of S. cerevisiae 17,18 . 50 Ergosterol and the UFA source Tween 80 (polyethoxylated sorbitan oleate) are routinely included in 51 media for anaerobic cultivation as 'anaerobic growth factors' (AGF) 15,17,19 . Under anaerobic conditions, S. 52 cerevisiae imports exogenous sterols via the ABC transporters Aus1 and Pdr11 20 . Mechanisms for uptake 53 and hydrolysis of Tween 80 by S. cerevisiae are unknown but, after its release, oleate is activated by the 54 acyl-CoA synthetases Faa1 and Faa4 21,22 . 55 Distribution of consumed glucose over biomass and products in chemostat cultures of S. cerevisiae (left 132 column) and K. marxianus (right column), normalized to a glucose uptake rate of 1.00 mol·h -1 . Numbers 133 in boxes indicate averages of measured metabolite formation rates (mol·h -1 ) and biomass production 134 rates (g dry weight·h -1 ) for each aeration and AGF supplementation regime. 135 Table 1 | Physiology of S. cerevisiae CEN.PK113-7D and K. marxianus CBS6556 in glucose-grown 136 chemostat cultures with different aeration and anaerobic-growth-factor (AGF) supplementation 137 regimes. Cultures were grown at pH 6.0 on synthetic medium with urea as nitrogen source and 7.5 g·L -1 138 glucose (aerobic cultures) or 20 g·L -1 glucose (oxygen-limited cultures) as carbon and energy source. 139 Data are represented as mean ± SE of data from independent chemostat cultures for each condition. 140 The AGFs ergosterol (E) and Tween 80 (T) were added to the media as indicated. Cultures were aerated 141 at 500 mL·min -1 with gas mixtures containing 21·10 4 ppm O 2 (O21·10 4 ), 840 ppm O 2 (O840) or < 0.5 ppm 142 O 2 (O0.5). Tween 80 was omitted from media used for aerobic cultivation to prevent excessive foaming. 143 Ethanol measurements were corrected for evaporation ( Supplementary Fig. 1). Positive and negative 144 biomass-specific conversion rates (q) represent consumption and production rates, respectively. 145

Absence of sterol import in K. marxianus 217
To test the hypothesis that K. marxianus lacks a functional sterol-uptake mechanism, uptake of 218 fluorescent sterol derivative 25-NBD-cholesterol (NBDC) was measured by flow cytometry 43 . Since S. 219 cerevisiae sterol transporters are not expressed in aerobic conditions 20 and to avoid interference of 220 sterol synthesis, NBDC uptake was analysed in anaerobic cell suspensions (Fig. 4a). Four hours after 221 NBDC addition to cell suspensions of the reference strain S. cerevisiae IMX585, median single-cell 222 fluorescence increased by 66-fold (Fig. 4bc). In contrast, the congenic sterol-transporter-deficient strain 223 IMK809 (aus1Δ pdr11Δ) only showed a 6-fold increase of fluorescence, probably reflected detergent-224 resistant binding of NBDC to S. cerevisiae cell-wall proteins 43,44 . K. marxianus strains CBS6556 and 225 NBRC1777 did not show increased fluorescence, neither after 4 h nor after 23 h of incubation with NBDC 226 (< 2-fold, Fig. 4bc, Supplementary Fig. 7). 227 Sterol uptake by S. cerevisiae, which requires cell wall proteins as well as a membrane transporter, has 244 not yet been fully resolved 42,43 . Instead of expressing a heterologous sterol-import system in K. 245 marxianus, we therefore explored production of tetrahymanol, which acts as a sterol surrogate in 246 strictly anaerobic fungi 45 . Expression of a squalene-tetrahymanol cyclase from Tetrahymena 247 thermophila (TtSTC1), which catalyzes the single-step oxygen-independent conversion of squalene into 248 tetrahymanol (Fig. 5a), was recently shown to enable sterol-independent growth of S. cerevisiae 46 . 249 TtSTC1 was expressed in K. marxianus NBRC1777, which is more genetically amenable than strain 250 CBS6556 47 . After 40 h of anaerobic incubation, the resulting strain contained 2.4 ± 0.4 mg·(g biomass) -1 251 tetrahymanol, 0.4 ± 0.1 mg·g -1 ergosterol and no detectable squalene, while strain NBRC1777 contained 252 3.5 ± 0.1 mg·g -1 squalene and 3.4 ± 0.2 mg·g -1 ergosterol (Fig. 5b). In strictly anaerobic cultures on sterol-253 free medium, strain NBRC1777 grew immediately after inoculation but not after transfer to a second 254 anaerobic culture (Fig. 5c), consistent with 'carry-over' of ergosterol from the aerobic preculture 19 . The 255 tetrahymanol-producing strain did not grow under these conditions ( Fig. 5c) but showed sustained 256 growth under severely oxygen-limited conditions that did not support growth of strain NBRC1777 (Fig.  257 5de). Single-cell isolates derived from these oxygen-limited cultures (IMS1111, IMS1131, IMS1132, 258 IMS1133) showed instantaneous as well as sustained growth under strictly anaerobic conditions ( Figure  259 5f and 5g). Tetrahymanol contents in the first, second and third cycle of anaerobic cultivation of isolate 260 IMS1111 were 7.6 ± 0.0 mg·g -1 , 28.0 ± 13.0 mg·g -1 and 11.5 ± 0.1 mg·g -1 , respectively (Fig. 5b), while no 261 ergosterol was detected. 262 To identify whether adaptation of the tetrahymanol-producing strain IMX2323 to anaerobic growth 263 involved genetic changes, its genome and those of the four adapted isolates were sequenced 264 (Supplementary Table 1). No copy number variations were detected in any of the four adapted isolates. 265 Only strain IMS1111 showed two non-conservative mutations in coding regions: a single-nucleotide 266 insertion in a transposon-borne gene and a stop codon at position 350 (of 496 bp) in KmCLN3, which 267 encodes for a G1 cyclin 48 . The apparent absence of mutations in the three other, independently adapted 268 strains indicated that their ability to grow anaerobically reflected a non-genetic adaptation. 269 represented with C1-C5. Strains NBRC1777 (wild-type, upward red triangles), IMX2323 (TtSTC1, cyan 280 downward triangle), and the single-cell isolates IMS1111 (TtSTC1, orange circles), IMS1131 (TtSTC1, blue 281 circles), IMS1132 (TtSTC1, yellow circles), IMS1133 (TtSTC1, purple circles). S. cerevisiae IMX585 282 (reference, purple circle) and IMX1438 (TtSTC1, orange circles). c, Extended data with double inoculum 283 size is available in Supplementary Fig. 10. d, Extended data is available in Supplementary Fig. 9a. 284

Test of anaerobic thermotolerance and selection for fast growing anaerobes 285
One of the attractive phenotypes of K. marxianus for industrial application is its high thermotolerance 286 with reported maximum growth temperatures of 46-52 °C 49,50 . To test if anaerobically growing 287 tetrahymanol-producing strains retained thermotolerance, strain IMS1111 was grown in anaerobic 288 sequential-batch-reactor (SBR) cultures (  Table 1). These data show that TtSTC1-expressing K. marxianus can grow 295 anaerobically at temperatures up to at least 45 °C. 296 supplemented with 20 g·L -1 glucose and 420 mg·L -1 Tween 80 at pH 5.0. a, Experimental design of 299 sequential batch fermentation with cycles at step-wise increasing temperatures to select for faster 300 growing mutants, each cycle consisted of three phases; (i) (re)filling of the bioreactor with fresh media 301 up to 100 mL and adjustment of temperature to a new set-point, (ii) anaerobic batch fermentation at a 302 fixed culture temperature with continuous N 2 sparging for monitoring of CO 2 in the culture off-gas, and 303 (iii) fast broth withdrawal leaving 7 mL (14.3 fold dilution) to inoculate the next batch. b, Maximum 304 specific estimated growth rate (circles) of each batch cycle for the three independent bioreactor 305 cultivations (M3R blue, M5R orange, M6L grey) with the estimated number of generations. The growth 306 rate was calculated from the CO 2 production as measured in the off-gas and should be interpreted as an 307 estimate and in some cases could not be calculated. The culture temperature profile (dotted line) for 308 each independent bioreactor cultivation (blue, grey, orange) consisted of a step-wise increment of the 309 temperature at the onset of the fermentation phase in each batch cycle. c, Representative section of 310 CO 2 off-gas profiles of the individual bioreactor (M5R) cultivation over time with CO 2 fraction (orange 311 line) and culture temperature (grey dotted line), data of the entire experiment is available in 312 Supplementary Fig. 11 (Data availability). 313

Discussion 314
Industrial production of ethanol from carbohydrates relies on S. cerevisiae, due to its capacity for 315 efficient, fast alcoholic fermentation and growth under strictly anaerobic process conditions. Many 316 facultatively fermentative yeast species outside the Saccharomycotina WGD-clade also rapidly ferment 317 sugars to ethanol under oxygen-limited conditions 26 , but cannot grow and ferment in the complete 318 absence of oxygen 11,13,25 . Identifying and eliminating oxygen requirements of these yeasts is essential to 319 unlock their industrially relevant traits for application. Here, this challenge was addressed for the 320 thermotolerant yeast K. marxianus, using a systematic approach based on chemostat-based quantitative 321 physiology, genome and transcriptome analysis, sterol-uptake assays and genetic modification. S. 322 cerevisiae, which was used as a reference in this study, shows strongly different genome-wide 323 expression profiles under aerobic and anaerobic or oxygen-limited conditions 51 . Although only a small 324 fraction of these differences were conserved in K. marxianus (Fig. 2), we were able to identify absence 325 of a functional sterol import system as the critical cause for its inability to grow anaerobically. Enabling 326 synthesis of the sterol surrogate tetrahymanol yielded strains that grew anaerobically at temperatures 327 above the permissive temperature range of S. cerevisiae. 328 A short adaptation phase of tetrahymanol-producing K. marxianus strains under oxygen-limited 329 conditions reproducibly enabled strictly anaerobic growth. Although this ability was retained after 330 aerobic isolation of single-cell lines, we were unable to attribute this adaptation to mutations. In 331 contrast to wild-type K. marxianus, a non-adapted tetrahymanol-producing strain did not show 'carry-332 over growth' after transfer from aerobic to strictly anaerobic conditions and adapted cultures showed 333 reduced squalene contents (Fig. 5). These observations suggest that interactions between tetrahymanol, 334 ergosterol and/or squalene influence the onset of anaerobic growth and that oxygen-limited growth 335 results in a stable balance between these lipids that is permissive for anaerobic growth. 336 Comparative genomic studies in Saccharomycotina yeasts have previously led to the hypothesis that 337 sterol transporters are absent from pre-WGD yeast species 11,52 . While our observations on K. marxianus 338 reinforce this hypothesis, which was hitherto not experimentally tested, they do not exclude 339 involvement of additional oxygen-requiring reactions in other non-Saccharomyces yeasts. For example, 340 pyrimidine biosynthesis is often cited as a key oxygen-requiring process in non-Saccharomyces yeasts, 341 due to involvement of a respiratory-chain-linked dihydroorotate dehydrogenase (DHOD) 53,54 . K. 342 marxianus, is among a small number of yeast species that, in addition to this respiration dependent 343 enzyme (KmUra9), also harbors a fumarate-dependent DHOD (KmUra1) 55 . In K. marxianus the activation 344 of this oxygen-independent KmUra1 is a crucial adaptation for anaerobic pyrimidine biosynthesis. The 345 experimental approach followed in the present study should be applicable to resolve the role of 346 pyrimidine biosynthesis and other oxygen-requiring reactions in additional yeast species. 347 Enabling K. marxianus to grow anaerobically represents an important step towards application of this 348 thermotolerant yeast in large-scale anaerobic bioprocesses. However, specific growth rates and biomass 349 yields of tetrahymanol-expressing K. marxianus in anaerobic cultures were lower than those of wild-type 350 S. cerevisiae strains. A similar phenotype of tetrahymanol-producing S. cerevisiae was proposed to 351 reflect an increased membrane permeability 46 . Additional membrane engineering or expression of a 352 functional sterol transport system is therefore required for further development of robust, anaerobically 353 growing industrial strains of K. marxianus 56 . 354 cultivation, synthetic medium was supplemented with ergosterol (10 mg·L -1 ) and Tween 80 (420 mg·L -1 ) 367 as described previously 14,17,19 . 368

Expression cassette and plasmid construction 369
Plasmids used in this study are described in (Table 4). To construct plasmids pUDE659 (gRNA AUS1 ) and 370 pUDE663 (gRNA PDR11 ), the pROS11 plasmid-backbone was PCR amplified using Phusion HF polymerase 371     Boroslicate glass, Thermo Fisher Scientific) at 70 °C. As internal standard 5α-cholestane (Sigma-Aldrich) 499 was added to the saponified biomass suspension. Subsequently tert-butyl-methyl-ether (tBME, Sigma-500 Aldrich) was added for organic phase extraction. Samples were extracted twice using tBME and dried 501 with sodium-sulfate (Merck, Darmstadt, Germany) to remove remaining traces of water. The organic 502 phase was either concentrated by evaporation with N 2 gas aeration or transferred directly to an 503 injection vial (VWR International, Amsterdam, the Netherlands). The contents were measured by GC-FID 504 using Agilent 7890A Gas Chromatograph (Agilent Technologies, Santa Clara, CA) equipped with an 505 Agilent CP9013 column (Agilent). The oven was programmed to start at 80 °C for 1 min, ramp first to 280 506 °C with 60 °C·min -1 and secondly to 320 °C with a rate of 10 °C·min -1 with a final temperature hold of 15 507 min. Spectra were compared to separate calibration lines of squalene, ergosterol, α-cholestane, 508 cholesterol and tetrahymanol as described previously 46 . 509 used for all samples. The gating strategy is given in Supplementary Fig. 8. Fluorescence of a strain was 530 determined by a sample of cells from independent shake-flask cultures and compared to cells from 531 identical unstained cultures of cells with the exact same chronological age. The staining experiment of 532 the strains IMX585, CBS6556 and NBRC1777 samples was repeated twice for reproducibility, the mean 533 and pooled variance was subsequently calculated from the biological duplicates of the two experiments. 534 The NBDC intensity and cell counts obtained from the NBDC experiments are available for re-analysis in 535 Supplementary Data set 1, and raw flow cytometry plots are depicted in Supplementary Data set 2. 536 Long read sequencing, assembly, and annotation 537 Cells were grown overnight in 500-mL shake flasks containing 100 mL liquid YPD medium at 30 °C in an 538 orbital shaker at 200 rpm. After reaching stationary phase the cells were harvested for a total OD 660 of 539 600 by centrifugation for 5 min at 4000 g. Genomic DNA of CBS6556 and NBRC1777 was isolated using 540 the Qiagen genomic DNA 100/G kit (Qiagen, Hilden, Germany) according to the manufacturer's 541 instructions. MinION genomic libraries were prepared using the 1D Genomic DNA by ligation (SQK-542 LSK108) for CBS6556, and the 1D native barcoding Genomic DNA (EXP-NBD103 & LSK108) for NBRC1777 543 according to the manufacturer's instructions with the exception of using 80% EtOH during the 'End 544 Repair/dA-tailing module' step. Flow cell quality was tested by running the MinKNOW platform QC 545 (Oxford Nanopore Technology, Oxford, UK). Flow cells were prepared by removing 20 μL buffer and 546 subsequently primed with priming buffer. The DNA library was loaded dropwise into the flow cell for 547 sequencing. The SQK-LSK108 library was sequenced on a R9 chemistry flow cell (FLO-MIN106) for 48 h. 548 Base-calling was performed using Albacore (v2.3.1, Oxford Nanopore Technologies) for CBS6556, and for 549 NBRC1777 with Guppy (v2.1.3, Oxford Nanopore Technologies) using dna_r9.4.1_450bps_flipflop.cfg. 550 CBS6556 reads were assembled using Canu (v1.8) 69 , and NBRC1777 reads were assembled using Flye 551 which was subsequently used as evidence for both CBS6556 and NBRC1777 genome annotations. 581 RNAseq libraries were mapped into the CBS6556 genome assembly described above, using bowtie 582 (v1.2.1.1) 76 with parameters (-v 0 -k 10 --best -M 1) to allow no mismatches, select the best out of 10 583 possible alignments per read, and for reads having more than one possible alignment randomly report 584 only one. Alignments were filtered and sorted using samtools (v1.3.1) 77 . Read counts were obtained 585 with featureCounts (v1.6.0) 78 using parameters (-B -C) to only count reads for which both pairs are 586 aligned into the same chromosome. 587 Differential gene expression (DGE) analysis was performed using edgeR (v3.28.1) 79 . Genes with 0 read 588 counts in all conditions were filtered out from the analysis, same as genes with less than 10 counts per 589 million. Counts were normalized using the trimmed mean of M values (TMM) method 80 , and dispersion 590 was estimated using generalized linear models. Differentially expressed genes were then calculated 591 using a log ratio test adjusted with the Benjamini-Hochberg method. Absolute log2 fold-change values > 592 2, false discovery rate < 0.5, and P value < 0.05 were used as significance cutoffs. The anaerobic growth ability of the yeast strains was tested on SMG-urea with 50 g·L -1 glucose at pH 6.0 619 with Tween 80 prepared as described earlier. The growth experiments were started from aerobic pre-620 cultures on SMG-urea media and the anaerobic shake flasks were inoculated at an OD 660 of 0.2 621 (corresponding to an OD 600 of 0.14). In order to minimize opening the anaerobic chamber, culture 622 growth was monitored by optical density measurements inside the chamber using an Ultrospec 10 cell 623 density meter (Biochrom, Cambridge, UK) at a 600 nm wavelength. When the optical density of culture 624 no longer increased or decreased new shake-flask cultures were inoculated by serial transfer at an initial 625 OD 600 of 0.2. 626

Laboratory evolution in low oxygen atmosphere 627
Adaptive laboratory evolution for strict anaerobic growth was performed in a Bactron anaerobic 628 workstation (BACTRON BAC-X-2E, Sheldon Manufacturing) at 30 °C. 50-mL Shake flasks were filled with 629 40 mL SMG-urea with 50 g·L -1 glucose and including 420 mg·L -1 Tween 80. Subsequently the shake-flask 630 media were inoculated with IMX2323 from glycerol cryo-stock at OD 660 < 0.01 and thereafter placed 631 inside the anaerobic chamber. Due to frequent opening of the pass-box and lack of catalyst inside the 632 pass-box oxygen entry was more permissive. After the optical density of the cultures no longer 633 increased, cultures were transferred to new media by 40-50x serial dilution. For IMS1111, IMS1112, 634 IMS1113 three and for IMS1131, IMS1132, IMS1133 four serial transfers in shake-flask media were 635 performed after which single colony isolates were made by plating on YPD agar media with hygromycin 636 antibiotic at 30 °C aerobically. Single colony isolates were subsequently restreaked sequentially for 637 three times on the same media before the isolates were propagated in SM glucose media and glycerol 638 cryo stocked. 639 To determine if an oxygen-limited pre-culture was required for the strict anaerobic growth of IMX2323 640 strain a cross-validation experiment was performed. In parallel, yeast strains were cultivated in 50-mL the CO 2 signal dropped to 70% of the maximum reached in each batch. The dilution factor of each 659 empty-refill cycle was 14.3-fold (100 mL working volume, 7 mL residual volume). The first batch 660 fermentation was performed at 30 °C after which in the second batch the temperature was increased to 661 42 °C and maintained at for 18 consecutive sequential batches. After the 18 batch cycle at 42 °C the 662 culture temperature was again increased to 45 °C and maintained subsequently. Growth rate was 663 calculated based on the CO 2 production as measured by the CO 2 fraction in the culture off-gas in 664 essence as described previously 83 . In short, the CO 2 fraction in the off-gas was converted to a CO 2 665 evolution rate of mmol per hour and subsequently summed over time for each cycle. The corresponding 666 cumulative CO 2 profile was transformed to natural log after which the stepwise slope of the log 667 transformed data was calculated. Subsequently an iterative exclusion of datapoints of the stepwise 668 slope of the log transformed cumulative CO 2 profile was performed with exclusion criteria of more than 669 one standard deviation below the mean. 670

Statistics 682
Statistical test performed are given as two sided with unequal variance t-test unless specifically stated 683 otherwise. We denote technical replicates as measurements derived from a single cell culture. Biological The raw RNA-sequencing data that supports the findings of this study are available from the Genome 694 Expression Omnibus (GEO) website (https://www.ncbi.nlm.nih.gov/geo/) with number GSE164344. 695 Whole-genome sequencing data of the CBS6556, NBRC1777 and evolved strains were deposited at NCBI 696 (https://www.ncbi.nlm.nih.gov/) under BioProject accession number PRJNA679749. 697

Code availability 698
The code that were used to generate the results obtained in this study are archived in a Gitlab 699 repository (https://gitlab.tudelft.nl/rortizmerino/kmar_anaerobic).