Yeast strains were provided by the ARS culture collection (Peoria, IL), the Phaff Yeast Culture Collection (Davis, CA), the American Type Culture Collection (Manassas, VA), and the CBS-KNAW Culture Collection (Utrecht, NL) and are listed in Supplementary Data Sheet 1 . Yeast strains were maintained on YPD (20 g/L glucose, 20 g/L peptone, 10 g/L yeast extract), YM (10 g/L glucose, 5 g/L peptone, 3 g/L yeast extract, 3 g/L malt extract), or PD (20 g/L glucose, 4 g/L potato extract) medium supplemented with 20 g/L agar as appropriate at 30°C. All yeast strains were stored in 15% glycerol at -80C. The ammonium-secreting Azotobacter vinelandii nifL mutant (Mus et al., 2022) (kindly provided by Florence Mus, Washington State University, WA, USA) was grown aerobically at 30°C and 200 rpm in a 250 mL baffled Erlenmeyer flask in a rotary shaker. Medium used for selection and co-culture experiments include Yeast Nitrogen Base without amino acids and ammonium sulfate (YNB; BD, USA) supplemented with 20 g/L glucose and either 3 g/L (NH₄)₂SO₄ or NaNO₃ without pH adjustment, modified Burk’s medium (mBG; 20 g/L glucose; 0.2 g/L KH₂PO₄; 0.8 g/L K₂HPO₄; 0.2 g/L MgSO₄•7H₂O; 0.13 g/L CaCl₂•2H₂O; 0.000253 g/L Na₂MoO₄•2H₂O; 0.00145 g/L FeSO₄•7H₂O, pH 6.0), modified Burk’s medium with nitrogen (mBGN; 20 g/L glucose; 3 g/L (NH₄)₂SO₄; 0.2 g/L KH₂PO₄; 0.8 g/L K₂HPO₄; 0.2 g/L MgSO₄•7H₂O; 0.13 g/L CaCl₂•2H₂O; 0.000253 g/L Na₂MoO₄•2H₂O; 0.00145 g/L FeSO₄•7H₂O, pH 6.0) and a derivative with mixed carbon sources representative of lignocellulosic hydrolysate (mBGX; 20 g/L glucose; 20 g/L xylose; 0.2 g/L KH₂PO₄; 0.8 g/L K₂HPO₄; 0.2 g/L MgSO₄•7H₂O; 0.13 g/L CaCl₂•2H₂O; 0.000253 g/L Na₂MoO₄•2H₂O; 0.00145 g/L FeSO₄•7H₂O, pH 6.0) (Mus et al., 2022). For co-culture experiments yeast and A. vinelandii cultures were washed twice with mBGX to remove spent medium and resuspended in fresh mGBX with A. vinelandii and the yeast, each to OD₆₀₀ = 0.001. Co-cultures were grown aerobically at 30°C and 200 rpm in 250 mL baffled Erlenmeyer flask in a rotary shake for 7 to 10 days.

Industrial yeasts were cultivated individually in YPD, YM, and PD medium in 96-well plates covered with breathable membranes at room temperature on an orbital plate shaker for 4 days at 800 rpm after which most cultures had reached stationary phase. Individual cultures of different strains from each medium type were then mixed in equal volume amounts and passaged at 1:100 volume into fresh YPD, YM, and PD medium. Mixed cultures were grown overnight at 30°C and 200 rpm in 250 mL baffled Erlenmeyer flasks in a rotary shaker. Cultures were then aliquoted as 15% glycerol stocks in cryo-tubes, flash-frozen in liquid N₂ and stored at -80°C as the industrial yeast library starting material. For all library experiments, a fresh cryo-stock was thawed on ice prior to 1:100 inoculation in fresh medium. For selection experiments the industrial yeast library and A. vinelandii cultures were washed twice and resuspended in fresh medium with each inoculum to OD₆₀₀ = 0.001. Selection experimental cultures were grown aerobically at 30°C and 400 rpm in 24-well plates with breathable membranes on an orbital plate shaker for 4 days prior to passaging 1:100 volume to fresh medium. Passaging was performed three times prior to collection of cell pellets by centrifugation at 13,000xg. The supernatant was removed for HPLC analysis and cell pellets were stored at -80C for further analysis. All selection experiments were performed in at least triplicate.

Genomic DNA was isolated from cell pellets using a yeast genomic DNA purification kit (AMRESCO, Solon, OH). The fungal internal transcribed spacer (ITS) region was amplified from genomic DNA samples by PCR with primers ITS1_F (5’-CTTGGTCATTTAGAGGAAGTAA-3’) and ITS4_R (5’-TCCTCCGCTTATTGATATGC-3’) (Toju et al., 2012). Oligonucleotides were purchased from IDT (Coralville, Iowa). PCR reactions were assessed for quality by gel electrophoresis and purified using a QIAquick PCR purification kit (Qiagen, Hilden, Germany). ITS amplicons were sequenced using the Illumina-based amplicon-EZ service (Genewiz Inc, South Plainfield, NJ). To assess mapping quality, FASTQ files of ITS amplicons were combined from all samples and mapped to ITS sequences present in the UNITE and UNITE+S databases (Abarenkov et al., 2024), and the NCBI ITS RefSeq Database (Schoch et al., 2014) using Bowtie2 (Langmead and Salzberg, 2012) in end-to-end or local alignment mode and and Samtools (Danecek et al., 2021) to assess mapping statistics. Read counts were converted to log₂ scale and median normalized within datasets prior to hierarchical clustering using average linkage with the correlation distance metric and visualization as heat-maps using InfernoRDN (Polpitiya et al., 2008).

Extracellular metabolites including glucose and xylose were quantified by HPLC. Ten microliters of the samples filtered with 0.2 µm syringe filters were analyzed for 45 min using an Aminex HPX-87H ion exclusion column with a 4.5 mM H₂SO₄ flow of 0.55 ml/min. The temperature of the column was 50°C. The refractive index was measured with a Waters 2414 refractive index detector at 30°C as well as the UV absorption at 210 nm. For microscopy, 10ul of coculture was taken and observed under the stereomicroscope-Leica MZ16 (Leica Microsystem Ltd, Bannockburn, IL, USA) at day 7. All images were captured at the same magnification (x100).

We developed a diverse library of industrially relevant yeast species and strains to enable rapid screening for bioconversion hosts with desirable phenotypes, including substrate utilization, tolerance to product toxicity as well as adaptability to various process conditions that impact process economics and greenhouse gas emissions such as temperature, pH, and osmolyte concentration (Navarrete and Martínez, 2020; Sun and Alper, 2020; Rodriguez-Ocasio et al., 2022). The library encompasses representatives from both the Saccharomycotina subphylum, where most industrial yeasts are found (Riley et al., 2016), and the Basidiomycete yeasts, particularly from Agaricomycotina and Pucciniomycotina, to capture a broad range of metabolic diversity (Johnson, 2013; Riley et al., 2016; Andreu et al., 2022; Suchova et al., 2022). Our library comprises 249 strains from 191 species across 82 genera ( Supplementary Data Sheet 1 ), with an emphasis on genera known for their potential in converting industrial and agricultural wastes into valuable bioproducts. These include Yarrowia, Lipomyces, Rhodotorula, Debaryomyces, and Trichosporon ( Figure 2A ). Where possible, we prioritized strains with publicly available genome sequences and annotations to facilitate future genomic analyses.

To construct the library each yeast was cultivated on three rich general purpose medium formulations (YPD, YM, and PD) for four days to ensure growth to or near stationary phase prior to mixing the yeast cultures in equal volumes. The mixed library was then diluted into fresh YPD, YM, and PD and allowed to grow overnight prior to collecting and freezing the library as glycerol stock aliquots. Genomic DNA was prepared from four samples of the industrial yeast library prior to storage to assess representation and quality. Universal primers for the fungal internal transcribed spacer (ITS) were used to amplify the ITS region from the pooled library, sequenced as amplicons, and aligned to sequences from high quality fungal ITS databases. Three databases were evaluated as sources of ITS sequences; the NCBI ITS RefSeq Database (Schoch et al., 2014) the UNITE database for molecular identification of fungi (Abarenkov et al., 2024), and UNITE+S which contains additional singletons. Amplicon sequences were aligned to ITS sequences in each database using end-to-end or local alignment modes to evaluate trade-offs between accuracy and quantity of mappings. As expected, local alignments mapped a greater number of the amplicon reads though this came at the expense of some unique mappings. In end-to-end mode more amplicons aligned to the NCBI database than the UNITE databases ( Figure 3A ). We further assessed the quality of end-to-end and local alignments by comparing the ratio of reads mapped to high count targets in the NCBI database and found that for expected yeast genera present in the industrial yeast library (Yarrowia, Pichia, Candida, ect.) the ratio was approximately 1:1 whereas filamentous fungi not present in the library (Mucor, Fusarium, Penicillium, Loekoeslaszloa, ect.) were identified only in local alignment mode ( Figure 3B ). Finally, we compared the ratio of reads mapping to high-count targets in the three databases in end-to-end mode. The overall quality of identifications was similar with some nuances in the naming of genera (e.g. Pichia in NCBI vs. Issatchenkia in the UNITE databases) and occasional erroneous alignment to filamentous fungi not present in the industrial yeast library even in end-to-end mode (e.g. Alternaria in UNITE+S and Penicillium in all 3 databases) ( Figure 3C ). From these results we concluded that alignment to the NCBI ITS RefSeq Database in end-to-end mode produced the highest quality output and was used for all further analyses ( Supplementary Data Sheet 2 ).

ITS amplicon sequences from Industrial yeast library samples collected during library construction were used to evaluate the overall composition of the stored library. ITS sequences from 73% of the genera included in the library design were identified in the physical library prior to storage suggesting overall maintenance of diversity through the use of diverse media for library preparation and broad detection with the possibility that PCR bias may contribute to loss of detection for some genera. Notable expansion was observed for some rapidly growing Ascomycete genera (Pichia, Candida, and Yarrowia) while overall representation of the Basidiomycetes declined ( Figure 2B ). Interestingly, 89% of the genera included in the library design were identified in follow-up experiments examining library composition in conditions with stronger selective pressure (data not shown). This suggests that while some strains may be present at levels lower than the threshold of detection for amplicon sequencing in the library stock, they can enrich and be quantified in environmental conditions where they are more competitive.

Fixed nitrogen, typically in the form of nitrate or ammonium, is essential for microbial growth in biological processes. Conventionally, ammonia produced by the energy and CO₂ intensive Haber-Bosch process serves as the nitrogen source for commercial bioprocesses (Smith et al., 2020; Matt, 2023). In this study, we explored in situ biological fixation of nitrogen by incorporating the nifL mutant of A. vinelandii, which is capable of fixing gaseous N₂ and secreting ammonia under aerobic conditions (Mus et al., 2022). The inclusion of A. vinelandii as a nitrogen-fixing biocatalyst offers an alternative to supplying offsite ammonia produced through the Haber-Bosch process, potentially reducing the carbon footprint of bioprocesses. We screened the industrial yeast library using a selective enrichment approach to identify yeasts capable of partnering with A. vinelandii, as well as those able to grow on various nitrogen sources. This was followed by amplicon sequencing of the ITS region to quantify competitive yeasts present in the selected libraries ( Figure 4A ). The industrial yeast library was passaged on YNB with either ammonium or nitrate as the nitrogen source, mBGN (with ammonium), and mBG in co-culture with A. vinelandii (using N₂ gas as the only nitrogen source). After three passages, we isolated genomic DNA from the cell pellet, amplified the ITS region, and sequenced it to quantify yeast enrichment in different selective environments.

Highly competitive Ascomycete yeasts were enriched in the control cultures with NH₄ as the nitrogen source (e.g. Scheffersomyces, Debaryomyces, and related yeasts) while more metabolically diverse yeasts were enriched on NO₃ as the nitrogen source such as the xylose consuming Myxozyma (Pomraning et al., 2019), Rhodotorula (Tiukova et al., 2019), and Spathaspora (Hou, 2012). A variety of yeast were enriched after passaging in co-culture with A. vinelandii with N₂ as the nitrogen source. These included Tremella, Apiotrichum, Metschnikowia, Kondoa, Zygosaccharomyces, Cutaneotrichosporon, Trichosporon, Effuseotrichosporon, Fellomyces, Yarrowia, and Rhodotorula species ( Figure 4B ). Many of the strains enriched in co-culture with A. vinelandii are considered to be oleaginous yeasts that accumulate greater than 20% of their biomass as lipids when grown in nitrogen limited conditions including Trichosporon, Yarrowia, and Rhodotorula strains (Zhu et al., 2012; Gao et al., 2014; Pomraning et al., 2015). The oleaginous phenotype is controlled by both nitrogen quantity and quality (Pomraning et al., 2016; Pomraning et al., 2017), and a high C:N ratio with ammonium as a nitrogen source is typically used to induce lipid accumulation (Kerkhoven et al., 2016). In our co-culture system with A. vinelandii, yeast growth is limited by the rate of N₂ fixation and secretion of usable ammonium. This natural nitrogen limitation may provide an inherent mechanism for controlling lipid production processes in oleaginous yeasts while simultaneously supplying sufficient nitrogen to support essential housekeeping metabolism for optimal productivity. This balance offers a potential advantage for bioprocess design, potentially eliminating the need for precise external nitrogen control and allowing for more robust and efficient lipid production systems.

We co-cultured 10 strains of oleaginous Y. lipolytica with A. vinelandii to confirm the ability of heterotrophic yeasts to grow in N₂ fixing conditions. Prior to co-culture, A. vinelandii was precultured in aerobic N₂ fixing conditions while the yeast strains were passaged in N-starvation conditions to induce a state of scavenging and eliminate carry-over of NH₄. Both partners were washed to remove any extracellular nitrogen and co-cultivation initiated in 50 mL shake-flask cultures at low density in aerobic N₂ fixing conditions in mBGX medium with glucose and xylose as carbon sources to mimic sugars present in enzymatically hydrolyzed lignocellulosic biomass, and to differentiate between sugars utilized by xylose-consuming strains of Y. lipolytica (Quarterman et al., 2017; Walker et al., 2018) and A. vinelandii which does not utilize xylose (Millan et al., 2020). Cultures without A. vinelandii were included for each yeast to control for any additional growth observed without the N₂ fixing partner that we speculate is due to carryover of intracellular nitrogen reserves. After 7 days, all Y. lipolytica-containing cultures showed additional glucose consumption compared to A. vinelandii control monocultures, but no consumption of xylose. Four co-cultures (strains NRRL Y-11853, NRRL Y-7208, NRRL Y-7317, and NRRL YB-618) consumed significantly more glucose than A. vinelandii alone ( Figure 4C ). Light microscopy confirmed the presence of both yeast and A. vinelandii cells in co-cultures at harvest time ( Figure 4D ) indicating growth of both species in the diazotroph/heterotroph partnership.