Moderate expression and activity of flocculins underlie the characteristic flocculation phenotype of Saccharomyces pastorianus Authors: D. G. L. Ignacia, N. X. Bennis, C. Carqueija Cardoso, A. A. J. Potma, E. Knibbe, P. A. de Groot, I. Westerlaken, M. van den Broek, J.-M. G. Daran Published: 2026 ABSTRACT ABSTRACTFlocculation is a key technological trait in lager brewing, governing fermentation performance, yeast recovery, and beer quality. In the allo-aneuploid hybrid yeast Saccharomyces pastorianus, the genetic basis of flocculation remains poorly resolved due to its complex dual sub-genome architecture. Here, we systematically re-annotated and functionally characterized the complete FLO gene repertoire of the Group II strain CBS 1483. Thirteen FLO genes were identified, including allelic variants and a previously uncharacterized adhesin, Flo12, containing a Hyphal_reg_CWP domain instead of the canonical PA14 lectin-binding domain. Structural modeling revealed strong conservation of Ca²+-binding residues in PA14 domains, alongside repeat-region diversification likely contributing to functional variability. Using optogenetic expression in a FLO-null background, we demonstrated that SpcI-FLO9-1 and SpcI-FLO9-2_1 are the strongest drivers of flocculation, exhibiting NewFlo-like sugar sensitivity. Transcriptomic analysis during 17°P wort fermentation showed dynamic induction of these genes coinciding with flocculation onset. Surprisingly, deletion of both loci in CBS 1483 did not abolish but only delayed sedimentation in wort, accompanied by improved maltose utilization and attenuation. These findings reveal functional redundancy and compensatory mechanisms within the FLO network of lager yeast, highlighting the genetic complexity underlying flocculation, and providing a molecular framework to inform yeast selection, strain development, and optimization of the lager fermentation processes.IMPORTANCEFlocculation, the process by which yeast cells aggregate and settle, is essential for producing clear, high-quality lager beer, and for efficient yeast recovery during brewing. However, the genetic basis of this trait in lager yeast has remained poorly understood because these strains possess unusually complex hybrid genomes. In this study, we systematically identified and characterized the complete set of flocculation genes in the industrial lager yeast Saccharomyces pastorianus CBS 1483. We demonstrated that lager yeast flocculation is not controlled by a single dominant gene, but instead emerges from the combined action of several moderately active adhesion proteins that are expressed at low levels during fermentation. Surprisingly, deleting the two strongest candidate genes only delayed, rather than eliminated, sedimentation, revealing a robust compensatory network that preserves brewing performance. These findings refine the current understanding of yeast flocculation and provide a molecular framework for developing brewing strains with improved fermentation efficiency, product consistency, and flavor quality. INTRODUCTION Flocculation, the reversible, non-sexual and Ca2+-dependent aggregation of yeast cells into multicellular clusters (1), is a defining physiological trait of brewing yeasts and a crucial determinant of beer fermentation performance, product quality, and process efficiency (2). Brewing yeasts are traditionally classified as either top-fermenting ale strains, which associate with CO₂ bubbles and rise during fermentation, or bottom-fermenting lager strains, which aggregate and sediment at the end of fermentation (3). While flocculation is important across brewing traditions, it is especially critical in lager brewing, where the timely and consistent aggregation and sedimentation of bottom-fermenting yeasts is crucial to achieving predictable fermentation, efficient yeast recovery, and the characteristic clarity of lager beers (4). The timely onset of flocculation is especially critical for efficient biomass removal. Premature yeast flocculation can lead to incomplete attenuation, particularly through insufficient consumption of maltotriose, a key late-stage sugar in wort, resulting in elevated residual sweetness and lower alcohol content (5, 6). Premature yeast flocculation may lead to elevated acetaldehyde levels and reduce late-formed alcohols and esters essential for a clean lager flavor profile (7). Early flocculation also limits yeast-mediated reduction of diacetyl to acetoin and 2,3-butanediol, leading to buttery off-flavors (8, 9). As lager yeasts often produce elevated sulfur compounds, particularly hydrogen sulfide, adequate yeast–beer contact time is required for their reabsorption or reduction (10–12). These processes depend on yeast metabolic activity, cell density, and fermentation duration (13). Conversely, weak or delayed flocculation hampers clarification and yeast harvesting, prolongs conditioning through cold storage, and increases costs, downstream processing, and tank occupancy (14). Extended suspension also reduces yeast vitality for re-pitching and promotes autolysis-derived off-flavors. In large-scale breweries using cylindroconical fermenters, predictable flocculation yields significant economic benefits, as tank time is one of the major cost drivers in beer production (15). Therefore, moderate and predictable flocculation kinetics are essential for classic lager flavor profiles. To ensure successful flocculation during brewing, it is critical to understand the factors that influence the onset and efficiency of flocculation. Flocculation phenotypes are shaped by a complex interplay of environmental, physiological, genetic, and epigenetic factors (16–34). At the molecular level, flocculation in yeast is primarily mediated by flocculins, cell wall–anchored proteins containing a lectin-like domain. In the presence of Ca²+, flocculins bind to mannose residues of α-mannan polysaccharides on glycosylated cell wall proteins of neighboring cells, thereby driving cell–cell adhesion and aggregate formation (35). These flocculins are encoded by the FLO adhesin gene family (36, 37). In Saccharomyces cerevisiae, FLO genes generally fall into two categories: FLO1, FLO5, FLO9, and FLO10 are responsible for cell–cell adhesion, while FLO11 is critical for pseudo-hyphal formation, invasive growth, substrate adhesion, and biofilm formation. The expression of each FLO gene causes a unique phenotype, showing differences in floc morphology, sedimentation rate, sugar sensitivity, adhesion strength, and cell surface hydrophobicity (33, 38, 39). These can be broadly placed into three categories: The Flo1 phenotype, showing strong flocculation mainly inhibited by mannose; the NewFlo phenotype, which is inhibited by multiple sugars, including mannose, glucose, sucrose, maltose, and maltotriose that competitively bind to the carbohydrate-binding domain (33, 39–41); and the mannose-insensitive phenotype (42). Brewing strains often exhibit the preferable NewFlo phenotype (20, 40, 43), where flocculation can start only once the fermentable sugars have been nearly exhausted. However, in the lager yeast Saccharomyces pastorianus, the situation is more complex. S. pastorianus is an allo-aneuploid interspecific hybrid yeast that originated from a natural hybridization between the domesticated ale yeast S. cerevisiae and the cold-tolerant wild species Saccharomyces eubayanus (44, 45). This hybridization has led to the evolution of two lineages, the Saaz (group I) and Frohberg (group II) strains, differing in the ploidy of each sub-genome complement and exhibiting variations in chromosome copy number across strains (46, 47). This hybrid nature has resulted in a high degree of genetic complexity, with large-scale genomic rearrangement and recombination between the two sub-genomes, leading to the development of novel genes and phenotypes not observed in the parental strains (28, 48, 49). Each sub-genome carries its own collection of FLO genes (46). With each strain showing its own unique genomic composition, understanding the role of FLO genes in S. pastorianus becomes an increasingly complicated endeavor. To elucidate the molecular basis of flocculation in Group II lager yeast, we reanalyzed and compiled the complete FLO gene inventory of S. pastorianus CBS 1483 (46). After verifying gene integrity, the encoded protein sequences were analyzed for structure–function relationships. To support these in silico predictions, each identified flocculin was reintroduced individually into a S. cerevisiae platform strain lacking all native FLO genes (33) and placed under precise optogenetic control, enabling systematic dissection of their roles in floc formation, sugar sensitivity, and cell-surface hydrophobicity. Transcriptomic analysis of S. pastorianus CBS 1483 during wort fermentation then revealed which flocculins are expressed in vivo. Their contributions were validated through targeted gene deletions, and the resulting strains were phenotyped in standard wort fermentations. Altogether, this molecular framework will inform yeast selection, strain development, and precision fermentation strategies, ensuring that lager production remains consistent, efficient, and aligned with the sensory expectations of brewers and consumers. MATERIALS AND METHODS Strains and growth conditions All strains used in this study are listed in Table 1. The Saccharomyces cerevisiae strains used in this study originate from the CEN.PK background (50), while the S. pastorianus strains originate from the CBS 1483 background (46). Yeast stocks were stored in 30% vol/vol glycerol at −80°C. Yeast cultures were inoculated from frozen stocks into 500 mL shake flasks containing 100 mL liquid YPD medium containing 10 g·L−1 yeast extract, 20 g·L−1 peptone, and 20 g·L−1 glucose, and incubated at 30°C for S. cerevisiae and 20°C for S. pastorianus in an Innova 44 incubator shaker (Eppendorf, Nijmegen, the Netherlands) set at 200 rpm. For the flocculation tests, strains were inoculated in YPD. For the selection of transformants, YPD medium was supplemented with G418, nourseothricin, hygromycin B, or in combination at concentrations of 200 mg·L−1, 100 mg·L−1, and 100 mg·L−1, respectively. For selection of the pUDR868 (amdSYM marker), (NH4)2SO4 was replaced by 0.6 g·L−1 acetamide as a nitrogen source and 6.6 g L−1 K2 SO4 to compensate for sulfate in SMD, resulting in SMD-AC (51). The pH in all the media was adjusted to 6.0 with KOH. Solid media were prepared by adding 2% bacto agar to the abovementioned media. For the optogenetics experiments, an identical light setup was used for the induction of the S. cerevisiae strains containing an optogenetics circuit as previously described (33). Cultivations in E.B.C. tall tubes were performed in full malt wort (81 g·L−1 maltose, 24 g·L−1 maltotriose, 19 g·L−1 glucose, and 8 g·L−1 fructose) at 17°P (Heineken, Zoeterwoude, the Netherlands) at 12°C. The wort was supplemented with CaCl2 at a final concentration of 59.3 mg L−1. TABLE 1Strains used in this studyStrain nameGenotypeParental strainOriginCEN.PK113-7DMATα MAL2-8C(50)CBS 1483Group II bottom-fermenting yeast, isolated from Heineken in July 1927(44)CDFM21L.1Saccharomyces eubayanus isolated from Tibet(52)IMX2897MATα MAL2-8C X 2::(pTEF1_VP16-EL222A79Q_tCYC1, pC120_QF2_tACT1, p5xQUASf_GFP_tADH1, pCCW12_QS_PSD_tENO1) can1Δ::Spycas9-NatNT2 flo1Δ flo5Δ flo9Δ flo10Δ flo11ΔIMK1060(33)IMX2912MATα MAL2-8C X 2::(pTEF1_VP16-EL222A79Q_tCYC1, pC120_QF2_tACT1, p5xQUASf_ScI-FLO1_tADH1, pCCW12_QS_PSD_tENO1) can1Δ::Spycas9-NatNT2 flo1Δ flo5Δ flo9Δ flo10Δ flo11ΔIMX2897(33)IMX3006MATα MAL2-8C X 2::(pTEF1_VP16-EL222A79Q_tCYC1, pC120_QF2_tACT1, p5xQUASf_SpeXII-FLO1-5_tADH1, pCCW12_QS_PSD_tENO1) can1Δ::Spycas9-NatNT2 flo1Δ flo5Δ flo9Δ flo10Δ flo11ΔIMX2897This studyIMX3007MATα MAL2-8C X 2::(pTEF1_VP16-EL222A79Q_tCYC1, pC120_QF2_tACT1, p5xQUASf_SpeVI-FLO10_tADH1, pCCW12_QS_PSD_tENO1) can1Δ::Spycas9-NatNT2 flo1Δ flo5Δ flo9Δ flo10Δ flo11ΔIMX2897This studyIMX3012MATα MAL2-8C X 2::(pTEF1_VP16-EL222A79Q_tCYC1, pC120_QF2_tACT1, p5xQUASf_SpcI-FLO9-1_tADH1, pCCW12_QS_PSD_tENO1) can1Δ::Spycas9-NatNT2 flo1Δ flo5Δ flo9Δ flo10Δ flo11ΔIMX2897This studyIMX3013MATα MAL2-8C X 2::(pTEF1_VP16-EL222A79Q_tCYC1, pC120_QF2_tACT1, p5xQUASf_SpeVI-FLO12_tADH1, pCCW12_QS_PSD_tENO1) can1Δ::Spycas9-NatNT2 flo1Δ flo5Δ flo9Δ flo10Δ flo11ΔIMX2897This studyIMX3101MATα MAL2-8C X 2::(pTEF1_VP16-EL222A79Q_tCYC1, pC120_QF2_tACT1, p5xQUASf_SpcI-FLO9-2_1_tADH1, pCCW12_QS_PSD_tENO1) can1Δ::Spycas9-NatNT2 flo1Δ flo5Δ flo9Δ flo10Δ flo11ΔIMX2897This studyIMX3102MATα MAL2-8C X 2::(pTEF1_VP16-EL222A79Q_tCYC1, pC120_QF2_tACT1, p5xQUASf_SpcI-FLO9-2_2_tADH1, pCCW12_QS_PSD_tENO1) can1Δ::Spycas9-NatNT2 flo1Δ flo5Δ flo9Δ flo10Δ flo11ΔIMX2897This studyIMX3104MATα MAL2-8C X 2::(pTEF1_VP16-EL222A79Q_tCYC1, pC120_QF2_tACT1, p5xQUASf_SpeI-FLO1-3_1_tADH1, pCCW12_QS_PSD_tENO1) can1Δ::Spycas9-NatNT2 flo1Δ flo5Δ flo9Δ flo10Δ flo11ΔIMX2897This studyIMX3105MATα MAL2-8C X 2::(pTEF1_VP16-EL222A79Q_tCYC1, pC120_QF2_tACT1, p5xQUASf_SpeI-FLO1-3_2_tADH1, pCCW12_QS_PSD_tENO1) can1Δ::Spycas9-NatNT2 flo1Δ flo5Δ flo9Δ flo10Δ flo11ΔIMX2897This studyIMX3109CBS 1483 SpcI-flo9-2Δ::KanMXCBS 1483This studyIMX3185CBS 1483 SpcI-flo9-1Δ::NatNT2 SpcI-flo9-2Δ::KanMXCBS 1483This study Strains used in this study Molecular biological techniques All primers designed in this study are shown in Table S1. The primers were ordered from Sigma-Aldrich (Burlington, MA, USA). RepliQa HiFi ToughMix (QuantaBio, Beverly, MA, USA) was used for obtaining fragments for sequencing and genomic integration according to the supplier’s instructions. For diagnostic purposes, PCR using DreamTaq PCR Master Mix (Thermo Fisher Scientific, Waltham, MA, USA) was utilized according to the supplier’s instructions. Plasmid isolations were performed using the GeneJET Miniprep kit (Thermo Fisher Scientific). All PCR products were purified from 1% TAE agarose using the Zymoclean Gel DNA Recovery Kit (Zymo Research, Irvine, CA, USA). PCR products were sequenced using Oxford Nanopore technology (Oxford Nanopore, Oxford, United Kingdom) by Plasmidsaurus (South San Francisco, CA, USA). All plasmids were propagated and stored using E. coli XL-1 Blue. Yeast genomic DNA for obtaining the FLO genes for integration and diagnostic purposes was isolated using the lithium acetate-SDS method (53). Plasmid construction All plasmids used in this study are found in Table 2. For the targeting of SpcI-FLO9-1 and SpcI-FLO9-2, the gRNA expressing plasmids pUDP359 and pUDP361 were constructed using the pUDP002 plasmid and cloning strategy described by Juergens et al. (54). The plasmids pUD1506 and pUD1584 contain the single-guide RNA (sgRNA) sequences targeting SpcI-FLO9-1 and both alleles of SpcI-FLO9-2, respectively, flanked by a hammerhead ribozyme at the 5′ end, a hepatitis delta virus ribozyme at the 3′ end, and inward-facing BsaI sites on both sides for integration into pUDP002. The insertion of the gRNA sequences into the pUDP002 backbone was done using Golden Gate reactions as described by Juergens et al. (54). Insertion into the pUDP002 backbone was diagnosed using primers 11200/11201. All plasmids were sequenced using Oxford Nanopore Technology (Oxford Nanopore) by Plasmidsaurus. TABLE 2Plasmids used in this studyNameRelevant characteristicsPurposeReferencepUDR8682μm ori bla amdSYM gRNA-GFP gRNA-GFPExpression of gRNA targeting GFP(33)pUDP002ori ampR panARSOPTAgTEF1p-hph-AgTEF1t ScTDH3p-BsaI-BsaI-ScCYC1t AaTEF1p-Spcas9D147Y P411T-ScPHO5tExpression of Cas9 and gRNA(54)pUD1506ori KanRBsaIHH-SpcI-FLO9-1-grna1-HDVBsaIContains gRNA targeting SpcI-FLO9-1GeneArtpUD1584ori KanRBsaIHH-SpcI-FLO9-2-grna1-HDVBsaIContains gRNA targeting SpcI-FLO9-2GeneArtpUDP359ori ampR panARSOPTAgTEF1p-hph-AgTEF1t ScTDH3p-HH-SpcI-FLO9-1-grna1-HDV-ScCYC1t AaTEF1p-Spcas9D147Y P411T-ScPHO5tExpression of Cas9 and gRNA targeting SpcI-FLO9-1This studypUDP361ori ampR panARSOPTAgTEF1p-hph-AgTEF1t ScTDH3p-HH-SpcI-FLO9-2-grna1-HDV-ScCYC1t AaTEF1p-Spcas9D147Y P411T-ScPHO5tExpression of Cas9 and gRNA targeting SpcI-FLO9-2This studypMEL132 μm bla KanMX gRNA-CAN1.YSource of KanMX cassette(55)pMEL152 μm bla NatNT2 gRNA-CAN1.YSource of NatNT cassette(55) Plasmids used in this study Strain construction Construction of Opto-FLO strains The construction of the CEN.K113-7D strains expressing FLO genes under optogenetic control was done according to the previously described workflow (33). Each FLO gene was PCR amplified as a repair fragment from the genomic DNA of CBS 1483 using primers that introduced 60 bp flanks homologous to the 5xQUASfp and ADH1t regions of the OptoQ-AMP5 circuit to allow for integration using Cas9-mediated homologous recombination. IMX2897 was transformed with 500 ng of repair fragment and pUDR868 containing a gRNA targeting the GFP locus in the OptoQ-AMP5 system. This resulted in the strains IMX3006 (SpeXII-FLO1-5), IMX3007 (SpeVI-FLO10), IMX3012 (SpcI-FLO9-1), IMX3013 (SpeVI-FLO12), IMX3101 (SpcI-FLO9-2_1), IMX3102 (SpcI-FLO9-2_2), IMX3104 (SpeI-FLO1-3_1), and IMX3105 (SpeI-FLO1-3_2). A nested PCR was performed to obtain the open reading frames for SpcI-FLO9-2, SpeI-FLO1-2, and SpeI-FLO1-3 using primer combinations 12145/12146, 20961/19618, and 12150/19618, respectively. The primer pairs 20407/20708, 20918/21654, 20411/20412, 20413/20414, 20745/20746, 20747/20748, 21653/21654, and 21652/21679 were used to amplify the repair fragments for SpcVI-FLO1-1, SpeI-FLO1-3, SpeVI-FLO12, SpcI-FLO9-1, SpcVI-FLO10, SpcXII-FLO1-5, SpeXI-FLO1-4, and SpcI-FLO9-2, respectively, from genomic DNA or nested PCR amplicons. Successful integration into the optoQ-AMP5 circuit was confirmed via PCR using the primer pair 14455/19907 and sequencing by Plasmidsaurus. The amplicons for SpcVI-FLO1-1, SpeI-FLO1-2, and SpeXI-FLO1-4 were only used for sequencing. Knockout of SpcI-FLO9-1 and SpcI-FLO9-2 in CBS 1483 To knock out SpcI-FLO9-1 and both alleles of SpcI-FLO9-2 in CBS 1483, antibiotic resistance cassettes were obtained via PCR with 60-bp flanking regions homologous to the promoter and terminator of the FLO genes. For SpcI-FLO9-1, the primer pairs 21638/21639 and 22083/22084 were used to amplify the KanMX and NatNT2 markers from pMEL13 and pMEL15 (55), respectively, while for SpcI-FLO9-2, the primer pairs 21646/21647 were used for the amplification of the KanMX marker from pMEL13. CBS 1483 was transformed with 500 ng of pUDP361 and 1 µg of the KanMX resistance cassette to create the single knockout strain IMX3109 (SpcI-flo9-1Δ::KanMX) via electroporation as previously described (56). Then, the strain IMX3109 was transformed with 500 ng of pUDP359 and 1 µg of the NatNT2 cassette to construct the strain IMX3185 (SpcI-flo9-1Δ::KanMX, SpcI-flo9-2Δ::NatNT2). Successful integration of the KanMX cassettes into the SpcI-FLO9-1 locus was confirmed using the primer pairs 21615/3811, while the correct integration of the NatNT2 cassette into the SpcI-FLO9-1 locus was confirmed using the primer pairs 21615/21616. Fermentation in tall E.B.C. tubes Fermentations in tall European Brewing Convention tubes (E.B.C.) (57) were performed using S. pastorianus strains CBS 1483 and IMX3185 as previously described by Bennis et al. (58). The fermentations were run in duplicates and were stopped after 16 days. Cell count was measured using a Nucleocounter YC-100 (Chemometec A/S, Allerød, Denmark), and wort density was measured using a DMA 35 basic handheld density meter (Anton Paar, Graz, Austria). RNA extraction and transcriptomics The tall E.B.C. tube fermentation of CBS 1483 was sampled at days 1, 2, 5, 6, 9, 12, and 15 for RNA extraction. The total RNA was extracted, and the purity, integrity, and concentrations of the samples were determined as previously described (33, 59). cDNA libraries were prepared using the TruSeq RNA v2 kit (Illumina, San Diego, CA, USA). Paired-end sequencing of twofold 150 bp was performed on a 300 bp PCR-free insert library using a NovaSeq system (Illumina) at Macrogen (Amsterdam, the Netherlands). Duplicate biological samples were prepared and sequenced, generating an average of 20.5 million reads per sample. The sequencing reads were processed as previously described (60). The CBS 1483 assembly (46) was used as a reference. To differentiate between the allelic variants of SpcI-FLO9-2, allele-specific SNP variation positioned at 21 bp upstream of the 3′ end of the gene was quantified by viewing the bam files in IGV, and the ratio of each allele was calculated as a percentage over the total amount of reads at that location for all samples. This percentage was then used to calculate the fraction of reads belonging to each allele over the total amount of mapped reads at the target locus. Normalization was performed using the trimmed mean of M (TMM) method (61) for between-sample comparisons and reads per kilobase of transcript per million reads mapped (rpkm) for in-sample comparisons (62, 63). Differential expression was calculated from TMM values using the EdgeR package (63). The SpcXIV-ALG9 gene, which was shown to have stable expression across a wide range of conditions, was used as a reference (64). Transcriptome RNA seq data of CBS 1483 grown in 17oP wort are available at NCBI under BioProject PRJNA1441968. Flo sequence analysis The DNA and protein sequences of the FLO genes of Saccharomyces cerevisiae CEN.PK113-7D (BioProject [http://www.ncbi.nlm.nih.gov/bioproject/] accession PRJNA393501 [65]), Saccharomyces eubayanus CDFM21L.1 (accession PRJNA528469 [60]), and Saccharomyces pastorianus CBS 1483 (accession PRJNA266750 [46]) were obtained from their respective annotated assemblies. The protein sequences were analyzed according to the previously described workflow (33). The following nomenclature was used to discern between the multiple FLO genes across the three Saccharomyces species and sub-genomes: S(p) c I ‒ FLO9 ‒ 2_1. The first letter, S, indicates the Saccharomyces genus. Saccharomyces pastorianus is indicated with a p. The species or subgenome is indicated with either a c for cerevisiae or an e for eubayanus, followed by the chromosome on which the locus is located (“I” in this example). Finally, the gene name is indicated (FLO9 in this example), followed by the copy number found in the genome (2 in this example) and allele (1 in this example), if applicable. Protein structural modeling Full-length Flo proteins with the signal peptide trimmed were modeled using AlphaFold 3 (66). Structures were simulated in the presence of Ca²+ and mannose-α1,2-mannose-α1,2-mannose. Structural models were visualized using the PyMOL Molecular Graphics System, Version 3.0.3 (Schrödinger, LLC, New York, NY, USA). Microscopy Image acquisition was conducted using a Zeiss AXIO Imager.Z1 microscope (Carl Zeiss AG, Oberkochen, Germany). The floc structure of the induced strains IMX3012 (SpcI-FLO9-1), IMX3013 (SpeVI-FLO12), IMX3101 (SpcI-FLO9-2_1), and IMX3102 (SpcI-FLO9-2_2) was examined by taking samples from the sedimented cells following the flocculation assay. Microscopic analysis of fractions was performed using Zeiss ZENBlue software (Carl Zeiss AG). Cells were imaged at 200× and 400× magnification for floc size analysis. The area of the flocs was measured using ImageJ as a measure of particle size. All particles smaller than 90 μm2 were filtered out to remove aggregates smaller than two aggregated budding cells. Statistical significance was calculated by comparing the geometric means using a lognormal one-way ANOVA test. The number of cell aggregates analyzed per strain ranged from 59 to 147. The data for CEN.PK113-7D, IMX2897 (floΔ), and IMX2912 (ScI-FLO1) were obtained from our previous study (33). Flocculation assays and hydrophobicity assessment For the flocculation assays and hydrophobicity assessments, the same methods as previously described (33) were utilized. The flocculation coefficient was determined by calculating the difference in ratio between the optical density (OD660) of deflocculated cells (A) and samples that were supplemented with CaCl2 to a concentration of 1% (B). For the sugar inhibition assays, the flocculation buffer was supplemented with 1 M of sugars, with the exception of maltotriose at 500 mM. In the case of the cell surface hydrophobicity assessment, the difference in ratio between the OD660 of deflocculated cells incubated with (C) and without (A) the addition of dodecane was used. The percentage of hydrophobicity was calculated as the relative change in absorbance of the aqueous layer before and after vortexing. The OD660 was determined using a Jenway 7200 scanning spectrophotometer (Cole-Parmer Inc, Chicago, IL, USA). Whole-genome sequencing The genomic DNA of the strain IMX3185 was isolated using the QIAGEN Genomic DNA isolation buffer set in combination with the Genomic-tip 100/G columns (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. The concentration, integrity, and purity of the samples were determined using the Qubit 2.0 (Thermo Fisher Scientific), TapeStation 2200 (Agilent Technologies), and Nanodrop (Thermo Fisher Scientific), respectively. All genomic DNA samples were sequenced using Nanopore technology (Oxford Nanopore, Oxford, United Kingdom). Sequencing data are available at the NCBI repository (https://www.ncbi.nlm.nih.gov/) under BioProject PRJNA1441332 (Table S2). The raw Illumina reads were mapped to the CBS 1483 (46) genome using the Burrows-Wheeler (BWA) tool (23) and further processed using SAMtools (67). For mapping of IMX3185 reads, an extra contig was added to match the expected genomic integration of the antibiotic resistance cassettes for each strain. Chromosomal copy number was estimated by the Magnolya algorithm (68). The sequences were analyzed by visualizing the bam files in the Integrative Genomics Viewer (IGV) software (version 2.8.9) (69). Analytical measurements Fermentable sugars (glucose, fructose, sucrose, maltose, and maltotriose) and ethanol concentrations were measured by HPLC (Agilent Technologies, 1260 HPLC system) equipped with a Bio-Rad HPX-42A column (300 × 7.8 mm, 25 μ) (Bio-Rad Laboratories, Hercules, CA, USA) and a 1260 Refractive Index Detector. The column was operated at 75 °C, and the elution was performed using milliQ at 0.5 mL·min−1 for 30 min. Compounds were measured using a RID at 35°C. Samples were 0.2 μm filter-sterilized before analysis (58, 70). RESULTS Composition of the FLO gene complement in S. pastorianus CBS 1483 Genomic re-analysis of Saccharomyces pastorianus CBS 1483 revealed the presence of 11 putative FLO genes (SpcI-FLO9-1, SpcI-FLO9-2, SpcVI-FLO1-1, SpcIX-FLO11-1, SpeI-FLO1-2, SpeI-FLO1-3, SpeVI-FLO10, SpeVI-FLO12, SpeIX-FLO11-2, SpeXI-FLO1-4, and SpeXII-FLO1-5, see Materials and Methods for gene nomenclature) (46) (Fig. 1). These genes were named based on the sequence similarity of their encoded proteins to S. cerevisiae Flo proteins and the presence of characteristic flocculin motifs (Table 3). In the CBS 1483 assembly (ASM1102231v1) (46), several mutations were identified in FLO genes that introduced frameshifts and premature stop codons. However, systematic amplicon resequencing demonstrated that many of these apparent mutations originated from sequencing errors in the published assembly. Resequencing led to the correction of five genes: SpcI-FLO9-1, SpcI-FLO9-2, SpeI-FLO1-2, SpeI-FLO1-3, and SpeVI-FLO12, which, with the exception of SpeI-FLO1-2, were found to encode full-length proteins. In addition, two allelic variants of SpcI-FLO9-2 and SpeI-FLO1-3 were identified and designated SpcI-FLO9-2_1 and _2, and SpeI-FLO1-3_1 and _2. The SpcI-FLO9-2 alleles differed in length by 522 bp, accounted for by a large difference in repeat type 1 and repeat type 2 counts, whereas the SpeI-FLO1-3 alleles were identical in length but exhibited 49 single-nucleotide variations, resulting in 17 amino acid substitutions (Table 3; Fig. S1). In total, 13 unique FLO genes were predicted, of which 4 are assumed to be non-functional due to premature stop codons: SpcVI-FLO1-1, SpcIX-FLO11-1, SpeI-FLO1-2, and SpeXI-FLO1-4 (Table 3). For comparison, the FLO gene complement of S. eubayanus CDFM21L.1 (52, 60) was also examined. Seven FLO genes were identified (Table 3): SeI-FLO1-1, SeI-FLO1-2, SeX-FLO11, SeVI-FLO12, SeVI-FLO10, SeXI-FLO1-3, and SeXII-FLO1-4. Among these, only SeXI-FLO1-3 contained a premature stop codon, resulting in a truncated protein. (A) Simplified depiction of chromosomal locations of all FLO genes identified in the CBS 1483 genome. With the exception of the FLO11 genes, all FLO genes are located in the telomeric region. FLO genes are indicated with a black line on the chromosomes. (B) Multiple sequence alignment of full-length Flo proteins encoded in S. cerevisiae CEN.PK113-7D, S. pastorianus CBS 1483, and S. eubayanus CDFM21L.1 using Clustal Omega (71). TABLE 3Overview of relevant protein features of Flo proteins identified in S. pastorianus CBS 1483, S. cerevisiae CEN.PK113-7D (33), and S. eubayanus CDFM21L.1aSpeciesStrainProteinFull ORFLength (AA)Signal sequenceLigand-binding domainTypeCentral domainGPI anchor site (homology)GPI anchor site (NetGPI)Repeat type 1Repeat type 2Repeat type 3Repeat type 4Repeat type 5Repeat type 6Repeat type 7 S. pastorianus CBS 1483SpcI-Flo9-1Yes9881–24113–248PA14274–498G964N967526216-- S. pastorianus CBS 1483SpcI-Flo9-2_1Yes1,0431–24112–248PA14274–543G1019N1022645214-- S. pastorianus CBS 1483SpcI-Flo9-2_2Yes8691–24112–248PA14274–318G845N848176212-- S. pastorianus CBS 1483SpcVI-Flo1-1No3451–23119–255PA14---1------ S. pastorianus CBS 1483SpcIX-Flo11-1No3501–2143–193Flo11---4*------ S. pastorianus CBS 1483SpeI-Flo1-2No3381–22112–248PA14---1------ S. pastorianus CBS 1483SpeI-Flo1-3_1Yes7531–22111–247PA14274–421G730S7272-3116-- S. pastorianus CBS 1483SpeI-Flo1-3_2Yes7531–22111–247PA14274–421G730S7272-3116-- S. pastorianus CBS 1483SpeVI-Flo10Yes1,1251–20108–249PA14381–624G1102N110518*-3---- S. pastorianus CBS 1483SpeVI-Flo12Yes1,2531–179–298HRCWD417–752G1231S12271-1--21*4* S. pastorianus CBS 1483SpeIX-Flo11-2Yes1,3021–2157–207Flo11293–946G1281G128145*5*----- S. pastorianus CBS 1483SpeXI-Flo1-4No3521–22129–262PA14---1------ S. pastorianus CBS 1483SpeXII-Flo1-5Yes9631–23103–251PA14285–329G940G9401-21225- S. cerevisiae CEN.PK113-7DScI-Flo1Yes1,6751–24112–248PA14274–1173G1652G16522023513-- S. cerevisiae CEN.PK113-7DScI-Flo9Yes9591–24112–248PA14274–453G936G936443213-- S. cerevisiae CEN.PK113-7DScVIII-Flo5Yes1,0751–24112–247PA14274–633G1052G105282326-- S. cerevisiae CEN.PK113-7DScIX-Flo11Yes1,3671–2431–207Flo11313–879G1346G134641*7*2*---- S. cerevisiae CEN.PK113-7DScXI-Flo10Yes1,1691–24123–270PA14302–606G1146N115410*2*32*--- S. eubayanus CDFM21L.1SeI-Flo1-1Yes7531–24111–247PA14274–422G730S7272-3-15- S. eubayanus CDFM21L.1SeI-Flo1-2Yes7531–22111–247PA14274–422G730S7272-3-15-- S. eubayanus CDFM21L.1SeIX-Flo11Yes1,2031–2143–193Flo11293–847G1182G118235*5*----- S. eubayanus CDFM21L.1SeVI-Flo12Yes1,0451–179–298HRCWD417–544G1022S10191-1--8*4* S. eubayanus CDFM21L.1SeVI-Flo10Yes9971–2099–262PA14381–547G974N97711*-2---- S. eubayanus CDFM21L.1SeXI-Flo1-3No361–2299–263PA14---------- S. eubayanus CDFM21L.1SeXII-Flo1-4Yes1,0211–2398–259PA14285–329G998G9981-21355- a Relevant domains were identified using the InterPro database (72). Repeating motifs were identified using RADAR, Clustal Omega 1.2.4, and manual curation as previously described (33). The central domain consists of the repeat type 1 repeating motifs. The GPI anchor site was determined based on sequence alignment and homology with the known sites in S. cerevisiae flocculins. GPI-anchor site residues were also predicted using NetGPI (73), showing a discrepancy between tool- and homology-based predictions. Repeating motifs that deviate from the Flo1 type repeats are marked with an asterisk (*). Notably, Flo12 contains an alternative ligand-binding domain, the Hyphally Regulated Cell Wall Domain (HRCWD), two unique repeating motifs, and shares the Flo1 repeat type 1 and type 3 motifs. SpeXII-Flo1-5 and SeXII-Flo1-4 contain a unique repeat type 2. The S. eubayanus Flo1-like flocculins generally lack the repeat types 2 and 4. Overview of relevant protein features of Flo proteins identified in S. pastorianus CBS 1483, S. cerevisiae CEN.PK113-7D (33), and S. eubayanus CDFM21L.1a Relevant domains were identified using the InterPro database (72). Repeating motifs were identified using RADAR, Clustal Omega 1.2.4, and manual curation as previously described (33). The central domain consists of the repeat type 1 repeating motifs. The GPI anchor site was determined based on sequence alignment and homology with the known sites in S. cerevisiae flocculins. GPI-anchor site residues were also predicted using NetGPI (73), showing a discrepancy between tool- and homology-based predictions. Repeating motifs that deviate from the Flo1 type repeats are marked with an asterisk (*). Notably, Flo12 contains an alternative ligand-binding domain, the Hyphally Regulated Cell Wall Domain (HRCWD), two unique repeating motifs, and shares the Flo1 repeat type 1 and type 3 motifs. SpeXII-Flo1-5 and SeXII-Flo1-4 contain a unique repeat type 2. The S. eubayanus Flo1-like flocculins generally lack the repeat types 2 and 4. Protein sequence alignment showed that the predicted flocculins could be broadly classified into four groups: Flo1-like (including Flo5 and Flo9, boxes 1 and 2), Flo10-like (boxes 3, 4, and 5), a novel Flo12-like group (box 6), and the Flo11-like (box 7) (Fig. 1B). Box 1 (Fig. 1; Table S3) contains S. cerevisiae Flo1-like flocculins, while box 2 contains the S. eubayanus Flo1-like proteins. Boxes 3 and 4 comprise proteins that share partial similarity with the Flo1-like flocculins but are more divergent; this group includes S. cerevisiae Flo10 and Flo proteins encoded by genes located on chromosomes VI and XII of S. eubayanus, along with their corresponding homologs in CBS 1483. Notably, box 6 includes a novel, distinct hypothetical flocculin designated Flo12, which exhibits structural similarity but limited identity compared to other flocculins. Finally, box 7 includes the Flo11-like proteins. Structure of S. pastorianus flocculins Flo proteins are well known to contain three functional domains: an N-terminal carbohydrate-binding PA14 domain, a central serine/threonine-rich tandem repeat region that stabilizes the PA14 domain and stems it toward the extracellular space, and a C-terminal GPI anchor required for anchoring the protein to the cell wall (74–78). The domain structure of the 13 identified flocculins was analyzed (Table 3; Fig. S2). PA14 site alignment The N-terminal carbohydrate-binding domain of flocculins belongs to the PA14 domain family (Pfam: PF07691) (79) and contains conserved Ca2+ coordinating sub-domains required for mannoside binding (75, 77, 80–82). Flo11 proteins share a similar overall architecture but contain a distinct “Flo11” domain (Pfam: PF10182) (38, 83) and were therefore excluded from our analyses. Multiple sequence alignment of flocculins from S. pastorianus CBS 1483, S. eubayanus CDFM21L.1, and S. cerevisiae CEN.PK113-7D showed strong conservation of residues essential for Ca2+ and carbohydrate recognition (Fig. 2). The Flo1 subdomain was conserved across Flo1-like proteins, except in ScIX-Flo10 and the flocculins encoded on CHRVI and CHRXII in CBS 1483 and CDFM21L.1. Key residues within the Flo1 subdomain (Q98) and the L3 loop (K194–I204) were maintained (84). Ca2+-binding loop 1 (CBL1; residues 157–161), including the DcisD motif (D160–D161), which is critical for Ca2+ docking (74), was conserved in all proteins except SeXII-Flo1-4, which contains a conservative D to E substitution. Multiple sequence alignment of the N-terminal carbohydrate PA14-binding site of flocculins from S. cerevisiae CEN.PK113-7D, S. eubayanus CDFM21L.1, and S. pastorianus CBS 1483. The PA14 domain from Epa1 from Nakaseomyces glabratus and the original Protective Antigen from Bacillus anthracis were included as references. The PA14 domain is highly conserved across all flocculins. SeVI-Flo10, SpeVI-Flo10, SeXII-Flo1-4, SpeXII-Flo1-5, and ScXIFlo10 show distinct variations in key amino acid residues relevant for Ca2+ binding and carbohydrate recognition compared to the remaining Flo1-like flocculins. Sequence alignment was done using Clustal Omega (71) and visualized using Jalview (85). The L3 loop was also conserved, except in flocculins encoded on CHRVI and CHRXII, which lacked several hallmark residues. The D202 residue, which influences L3 loop flexibility and thereby carbohydrate specificity (Fig. 2), was only conserved in flocculins belonging to box 1 (Fig. 1B). Variations in CBL2 (e.g., C248N, W250S) may likewise further influence Ca²+ and carbohydrate recognition (74, 84, 86). Overall, the PA14 domain is highly conserved across Flo proteins from both parental species and their hybrid, with ScI-Flo10 and the CHRVI/CHRXII S. eubayanus-derived proteins representing notable outliers that may use alternative ligand-recognition mechanisms. Repeat analysis The central domain of Flo proteins consists of tandem repeats that modulate binding strength and ligand specificity (74, 76, 77, 80, 82). We previously reported five distinct motifs in Flo1-like proteins from S. cerevisiae CEN.PK113-7D, whereas Flo10 and Flo11 contained four and three types, respectively, with Flo10 sharing one motif with the Flo1-like proteins (33). Repeat type composition was determined for all Flo proteins from S. pastorianus CBS 1483 and S. eubayanus CDFM21L.1 (Table 3). All Spc-Flo1-like proteins (box 1; Fig. 1) contained the same repeat types as their S. cerevisiae counterparts. Flo proteins from S. eubayanus contained Flo1 repeat types 1, 3, and 5 but lacked types 2 and 4, except SpeXII-Flo1-5 and SeXII-Flo1-4, which possessed Flo1 repeat types 1, 3, 4, and 5, plus an additional motif resembling an extended type 5 repeat. SpcI-Flo9-1, SpcI-Flo9-2, SpeI-Flo1-3, and SpeXII-Flo1-5 all contain a high number of repeat type 5, differing starkly from the S. cerevisiae proteins. These findings indicate that, while the overall repeat organization of Flo proteins is conserved between Saccharomyces cerevisiae and eubayanus, specific repeat variants have evolved. Flo12 is a hypothetical protein with similar structural organization as yeast adhesins One protein, Flo12, was particularly distinct, sharing very low sequence identity with other flocculins (Fig. 1). A BLASTP analysis identified only uncharacterized proteins in other closely related Saccharomyces species, showing little similarity to any other known protein. Although Flo12 contained multiple internal repeats and a predicted C-terminal GPI anchor site, it notably lacks the characteristic N-terminal PA14 domain. Instead, it harbored a hyphally regulated cell wall protein N-terminal domain (Hyphal_reg_CWP, Pfam: PF11765) (Table 3), previously identified in Candida albicans adhesins involved in cell wall organization and host–fungus interactions (87, 88), and the cell wall protein Haze Protection Factor 1 (HPF1) and its homolog HZY1 in S. cerevisiae (89–91). A Phyre2.2 (92) search revealed 85% structural similarity with 100% confidence to Adhesin-like wall protein 1 (Awp1) α-domain from Candida glabrata, while a protein sequence alignment of the two domains only showed 24.9% sequence similarity. Notably, the main tandem repeating motif in the central domain is distinct from the other flocculins, consisting of a serine/threonine-rich 16 amino acid sequence (Table S4). Downstream of the central domain region lies another novel repeating motif, followed by single Flo1 type 1 and type 3 repeats. The function of Flo12 is unknown, but it may contribute to the complex flocculation phenotype of S. pastorianus. Structural models highlight similarities and variations among flocculation proteins Sequence alignment revealed substantial conservation within the PA14-binding domain, with limited yet potentially functionally relevant differences in key regions (Fig. 2). To explore the structural consequences of these variations, biomolecular structure predictions were simulated using Alphafold 3 (66) (Fig. 3). The predicted models displayed the canonical β-sandwich fold characteristic of the PA14 domains (74, 77, 79, 93) and Ca2+ coordination by conserved residues in CBL1 and CBL2. Structural alignment using US-align (94) showed high structural conservation across all PA14 domains (Fig. S2). Deviations in structure stem largely from the sequence variations observed in the Flo1 subdomain and L3 loop (Fig. 2), which may impact carbohydrate binding specificity (36, 74, 86). All PA14 domain-containing Flo proteins were predicted to bind mannose-α1,2-mannose-α1,2-mannose, a proxy of the D1 arm of N- and O-linked mannan chains of glycosylated proteins on the yeast surface. Alignment of predicted structures of the PA14 domains of flocculins (grey) with ScI-Flo1 (light blue) and structure of the Hyphal_reg_CWP domain of Flo12. The Flo1 subdomain is indicated in blue and green for ScI-Flo1 and the aligned flocculin, respectively. The relevant loops are indicated in purple and orange for ScI-Flo1 and the aligned flocculin, respectively. Ca2+ is indicated in cyan. (A) ScI-Flo9. (B) ScXI-Flo10. (C) SpcI-Flo9-1. (D) SpcI-Flo9-2_1. (E) SpcI-Flo9-2_2. (F) SpeI-Flo1-3. (G) SpeVI-Flo10. (H) SpeXII-Flo1-5. (I) The predicted structure of the Hyphal_reg_CWP domain in Flo12 strongly resembles Candida adhesins, displaying the characteristic β-helix and α-crystallin structures. The Hyphal_reg_CWP domain of Flo12 was predicted to comprise a right-handed β-helix consisting of nine rings of three β-strands forming three stacked β-sheets in parallel, followed by an α-crystalline-like fold (Fig. 3I). A disulfide bond was predicted to bind the C-terminal section to the β-helix. This structure is characteristic of fungal adhesins such as the virulence factors, adhesin-like wall proteins (Awp) and Hyr/Iff-like (Hil) proteins, involved in adherence to host endo- and epithelial cells, and mediating biofilm formation in Candida spp. (87, 88). Based on these characteristics, Flo12 is likely involved in an adhesion phenotype. Functional assessment of S. pastorianus CBS 1483 flocculation genes using optogenetic control in a FLO-null strain background To assess the functionality of the identified FLO genes, each of the eight full-length FLO genes, excluding FLO11 (Table 3), was inserted into the optogenetic OptoQ-AMP5 circuit and characterized following a previously described workflow (33). As a benchmarking reference, S. pastorianus CBS 1483 cells harvested in the stationary phase exhibited a flocculation coefficient of 91.4% ± 1.2%, while CEN.PK113-7D and IMX2897 (floΔ) showed a flocculation coefficient of 7.2% ± 2.6% and 1.9% ± 4.2%, respectively (Fig. 4A). The OptoQ-AMP5 strains were grown in the dark and induced with light for 2 h, after which flocculation coefficients were determined. In the dark conditions, the flocculation coefficients ranged from −3.1% ± 5.7% to 26.1% ± 6.6%. After 2 h of induction, IMX3012 (SpcI-FLO9-1) and IMX3101 (SpcI-FLO9-2_1) were the only two strains showing a significant increase in flocculation coefficient, reaching 73.6% ± 8.8% and 65.7% ± 12.6%, respectively. However, these values remained lower than the coefficient of flocculation exhibited by IMX2912, which expresses the S. cerevisiae ScI-FLO1 (33), and reached 96.6% ± 2.3%. Since all strains have the same transcriptional regulation, differences in flocculation strength are likely attributable to intrinsic differences in flocculin activity/affinity. Although flocculin expression was not quantified, the sequence of each expression cassette was verified and revealed no mutations or structural variations. Therefore, these differences are unlikely to result from defects in the expression constructs and may instead be mitigated by increasing flocculin surface expression. To test this hypothesis, we modified the assay by extending the light induction period to 24 h. Under these conditions, IMX3012 (SpcI-FLO9-1) and IMX3101 (SpcI-FLO9-2_1) reached flocculation of 98.1% ± 0.1% and 94.6% ± 0.2%, respectively, values comparable to IMX2912 (ScI-FLO1) after only 2 h of induction. Under the extended induction regime, IMX3013 (SpeVI-FLO12) and IMX3102 (SpcI-FLO9-2_2) also showed substantial increases, reaching 57.9% ± 9.8% and 80.1% ± 3.5%, respectively. IMX3104 also showed a significant increase in flocculation coefficient after 24 h, reaching 28.5% ± 0.8%, but did not show any measurable increase in floc formation. None of the remaining strains exhibited a significant increase in flocculation coefficient, and microscopy confirmed the absence of visible flocs. Among the flocculating strains, floc sizes were generally smaller than those of IMX2912 (ScI-FLO1), which exhibited an interquartile range (IQR) of 693–9,363 μm2 (Fig. 4B; Fig. S4). After 24 h of induction, IMX3102 (SpcI-FLO9-2_2) formed the largest flocs with an IQR of 144–1,070 μm2, consistent with its flocculation coefficient. IMX3101 (SpeVI-FLO9-2_1) showed smaller floc sizes with an IQR of 150–464 μm2. Notably, although IMX3013 (SpeVI-FLO12) displayed an increase in flocculation coefficient, the geometric mean of the floc size was not significantly different from IMX2897 (floΔ). However, IMX3013 showed an IQR of 107.6–202.5 μm2 after 24 h of light induction, showing a slight increase in larger particles appearing more frequently (Fig. 4B; Fig. S4). This minor increase may be enough to trigger sedimentation. Visual inspection of the flocculation assay tube revealed very small, sand-like particles. Together, these results highlight the diversity of flocculation phenotype among S. pastorianus FLO genes and suggest substantial variation in aggregation potential of the encoded flocculins. (A) Flocculation coefficients of CEN.PK113-7D, CBS 1483, IMX2897 (floΔ), and optogenetics strains expressing ScI-FLO1 and CBS 1483 FLO genes sampled after 2 and 24 h of growth under light and dark conditions. CBS 1483 was sampled at the stationary phase. For each condition, strains were grown in duplicate cultures, and triplicate technical replicates were performed to determine the flocculation coefficient for each culture. IMX3012 (SpcI-FLO9-1) and IMX3101 (SpcI-FLO9-2_1) showed a significant increase in flocculation coefficient after 2 h of induction, while IMX3013 (SpeVI-FLO12) and IMX3102 (SpcI-FLO9-2_2) only showed a significant increase after 24 h of induction (*P < 0.05, **P < 0.01, ****P < 0.0001). IMX3104 (SeI-FLO1-3_1) showed a minor significant increase after 24 h of induction, but no significant increase in floc formation was observed. All other strains show no significant increase relative to dark conditions. (B) Floc size distribution based on the area of cell aggregates imaged at 200× magnification using microscopy. The floc sizes of IMX2912 (ScI-FLO1) show a larger range and geometric mean relative to the flocs induced by CBS 1483 FLO genes. All strains form significantly larger floc sizes relative to IMX2897 (floΔ) (****P < 0.0001), with the exception of IMX3013 (SpeVI-FLO12), despite showing a moderate flocculation coefficient. All particles smaller than 90 μm2 were filtered out to remove aggregates smaller than two aggregated budding cells. Statistical significance was calculated using a lognormal one-way ANOVA test. Each box represents the 25th to 75th percentiles, with the whiskers indicating the minimum and maximum values. The black line represents the median, while the red line represents the geometric mean of the samples. The number of cell aggregates analyzed per strain ranged from 59 to 147. The data for CEN.PK113-7D, IMX2897 (floΔ), and IMX2912 (ScI-FLO1) in panels A and B were obtained from our previous study (33). Sugar inhibition phenotypes of CBS 1483 flocculins S. pastorianus strains are known to exhibit a NewFlo flocculation phenotype, showing sensitivity to various sugars, e.g., mannose, maltose, glucose, and sucrose, low pH conditions, and proteinases (40, 77, 86). To identify which of the flocculation genes are involved in flocculation in CBS 1483, the sensitivity of the functional flocculins SpcI-Flo9-1, SpcI-Flo9-2_1, and SpcI-Flo9-2_2 to various fermentable sugars found in wort was determined as previously described (33) (Fig. 5A). All the sugars tested completely abolished flocculation in CBS 1483. Notably, we demonstrated the sensitivity of S. pastorianus to fructose and maltotriose. IMX3012 (SpcI-FLO9-1) showed sensitivity mainly to mannose, sucrose, and, to a lesser degree, fructose, maltose, and maltotriose. Additionally, IMX3101 (SpeVI-FLO9-2_1) and IMX3102 (SpcI-FLO9-2_2) showed different inhibition characteristics. IMX3101 (SpeVI-FLO9-2_1) was sensitive to mannose, sucrose, maltose, and, to a lesser degree, maltotriose, while IMX3102 (SpcI-FLO9-2_2) showed an additional sensitivity to fructose and glucose. Remarkably, none of the tested strains exhibited the stark inhibition by all the sugars as was observed for CBS 1483. Sugar inhibition profiles of (A) CBS 1483, IMX3012 (SpcI-FLO9-1), IMX3101 (SpcI-FLO9-2_1), IMX3102 (SpcI-FLO9-2_2), and (B) mannose and Ca2+ sensitivity of IMX3013 (SpeVI-FLO12). The optogenetic strains were induced until measurable flocculation occurred: 2 h for IMX3012 (SpcI-FLO9-1) and IMX3101 (SpcI-FLO9-2_1), and 24 h for IMX3102 (SpcI-FLO9-2_2) and IMX3013 (SpeVI-FLO12). (A) CBS 1483 showed a strong sensitivity to all the tested sugars. IMX3012 (SpcI-FLO9-1), IMX3101 (SpcI-FLO9-2_1), and IMX3102 (SpcI-FLO9-2_2) all exhibited similar sugar sensitivity profiles, showing strong sensitivity to mannose, with weak to moderate sensitivity to the remaining sugars. (B) IMX3013 (SpeVI-FLO12) showed Ca2+-independent flocculation that was moderately sensitive to mannose. As IMX3013 (SpeVI-FLO12) exhibited flocculation behavior, we tested it for Ca2+ requirement and mannose sensitivity (Fig. 5B). There was no significant reduction in flocculation coefficient in the samples without Ca2+ supplementation (58% ± 6.4%) relative to the control (60% ± 13.2%). Interestingly, 1 M mannose caused a reduction of 42% in flocculation coefficient relative to the control, suggesting sugars may interact with the Hyphal_reg_CWP domain. CBS 1483 flocculins exhibit varying impact on cell surface hydrophobicity Flocculins are known to contribute to cell wall hydrophobicity to varying degrees (33, 38, 95, 96). Using the previously described dodecane partitioning assay, the effect of the CBS 1483 flocculins on the cell surface hydrophobicity was determined (Fig. 6). Flocculating CBS 1483 cells grown to the stationary phase displayed a hydrophobicity percentage of 46.5% ± 7.3%, while the non-flocculating CEN.PK113-7D showed a hydrophobicity percentage of 18.3% + 11.0%. CBS 1483 was comparable to the strongly flocculating IMX2912 (ScI-FLO1), which exhibited a hydrophobicity percentage of 51.3% ± 3.5%. Notably, IMX3006 (SpeXII-FLO1-5) and IMX3007 (SpeVI-FLO10) showed a strong increase in cell surface hydrophobicity, reaching values of 76.5% ± 2.3% and 55.0% ± 3.3%, respectively, despite their encoded Flo proteins failing to initiate flocculation (Fig. 6A). In contrast, the flocculating strains IMX3012 (SpcI-FLO9-1), IMX3101 (SpcI-FLO9-2_1), and IMX3102 (SpcI-FLO9-2_2) did not show a significant increase in cell surface hydrophobicity relative to dark conditions. The strain IMX3013 (SpeVI-FLO12), which expresses a novel Flo protein lacking a PA14 lectin-binding domain, exhibited a strong increase in hydrophobicity, reaching 75.8% ± 1.0%. IMX3104 (SpeI-FLO1-3_1) and IMX3105 (SpeI-FLO1-3_2) showed no significant increase in cell surface hydrophobicity. Together, these data illustrate that increases in cell surface hydrophobicity do not necessarily coincide with cellular aggregation, highlighting distinct mechanistic contributions of different flocculins. Determination of cell surface hydrophobicity of CEN.PK113-7D, CBS 1483, and optogenetic strains expressing ScI-FLO1 (33) and the identified CBS 1483 FLO genes grown under light and dark conditions. The hydrophobicity percentage is calculated by the ratio of cells partitioning between an aqueous phase and a dodecane organic phase. All bars represent average values of experiments using biological duplicates and measured in technical duplicates. The CEN.PK113-7D bar represents the average of all control experiments (33). IMX3006 (SpeXII-FLO1-5), IMX3007 (SpeVI-FLO10), and IMX3013 (SpeVI-FLO12) showed a significant increase in hydrophobicity in light conditions relative to dark conditions (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). CBS 1483 and IMX2912 (ScI-FLO1 light) showed a similar increase in hydrophobicity. All strains grown in darkness showed no significant difference in hydrophobicity relative to CEN.PK113-7D. Error bars represent the standard deviation. Statistical significance was calculated using multiple unpaired two-sided t-tests. P-values were corrected for multiple comparisons using the Holm-Šidák method. The data for CEN.PK113-7D and IMX2912 (ScI-FLO1) were obtained from our previous study (33). FLO gene expression analysis in 17° Plato wort fermentation of CBS 1483 in E.B.C. tall tubes To identify which of the flocculin genes contribute to the strong flocculation phenotype observed in group II strain CBS 1483, a transcriptome analysis was performed by sampling cells at various time points during a 17° Plato wort fermentation conducted in E.B.C. tall tubes for 15 days (Fig. 7). The S. pastorianus CBS 1483 strain started to flocculate around day 5 (Fig. 7A). Samples were collected on days 1, 2, 5, 6, 9, 12, and 15 for transcriptome analysis. After 5 days of fermentation, the cell counts peaked and subsequently declined, showing a 72% decrease by day 6 and complete sedimentation of the biomass by day 8 (Fig. 7A). The onset of sedimentation coincided with the depletion of most fermentable sugars after 6 days (Fig. 7A and B). At the end of fermentation, residual concentrations of maltose and maltotriose remained at 2.4 g·L−1 and 13.6 g·L−1, respectively (Fig. 7B), and a final gravity of 4.1 was reached (Fig. 7A). Out of the nine full-length FLO genes (including allelic variants, and excluding FLO11 and the truncated FLO genes) detected on CBS 1483 genome (Table 3), eight could form a detectable transcript (Fig. 7C), namely SpcI-FLO9-1, SpcI-FLO9-2_1, SpcI-FLO9-2_2, SpeI-FLO1-3_1, SpeVI-FLO12, SpeVI-FLO10, and SpeXII-FLO1-5 (Table 3). Transcriptomic analysis of FLO gene expression in CBS 1483 during a duplicate 17°P wort fermentation at 12°C in E.B.C. tall tube fermenters. (A) Cell count (cells·mL−1) of CBS 1483 and wort density (°P) during the 15-day-long fermentation. Red arrows indicate sampling timepoints for transcriptomic analysis. The cell count drops dramatically after 5 days, indicating flocculation, and coincides with attenuation of the wort. (B) Fermentable sugars, glycerol, and ethanol concentrations during the CBS 1483 fermentation. (C) Transcription profile of detected FLO genes during the CBS 1483 fermentation expressed as TMM values. (D) Fold changes of FLO genes between and day 6 relative to day 2. Overall, FLO genes’ expression levels were relatively low, with a subset showing a dynamic expression pattern coinciding with the onset of flocculation and sedimentation at day 5 (Fig. 7C; Table S5). Expressed in reads per kilobase of transcript per million reads mapped (rpkm), FLO gene transcripts never exceeded 38.5 ± 0.92 rpkm. Although slightly higher than the median transcript level across all samples (23.36 rpkm), the FLO transcripts were at least an order of magnitude lower than maltose transporter genes SpcII-MAL31, which reached 743.9 ± 68.42 rpkm at day 6, or glycolytic genes SpcIV-TPI1 and SpeIII-cIII-PGK1, which reached expression levels of 8013.3 ± 184.38 and 5828.0 ± 33.74 rpkm, respectively, on the first day. After SpeVI-FLO10, SpcI-FLO9-2_1 showed the highest rpkm value at 19.3 ± 9.31 on day 15, followed by 19.0 ± 0.90 and 18.7 ± 2.90 on days 5 and 6, respectively. We monitored more specifically the changes between day 2 and day 6. SpcI-FLO9-1 and SpcI-FLO9-2_1 both exhibited a 4.5-fold increase in transcript levels on day 6 relative to day 2 (Fig. 7D). Notably, SpcI-FLO9-2_2 showed very low expression relative to SpcI-FLO9-2_1, indicating preferential allelic expression. SpeI-FLO1-3 and SpeXII-FLO1-5 showed a 3.2- and 3.5-fold increase on day 6 relative to day 2. Notably, SpeVI-FLO10 shows the highest expression with 38.5 ± 0.92 rpkm on day 1, while the cells are still in free suspension. After 6 days, transcript levels of all FLO genes declined. An increase was also observed on day 15, reaching expression levels comparable to those reached at day 5, which corresponded to the planktonic cellular fraction that still had to flocculate. These results suggest that SpcI-FLO9-1, SpcI-FLO9-2, followed by SpeI-FLO1-3 and potentially SpeVI-FLO10, are likely the main contributors to the pronounced flocculation phenotype observed in CBS 1483. The novel SpeVI-FLO12 gene and SpeXII-FLO1-5 exhibited very low levels of expression and are likely not involved in the observed flocculation phenotypes. Knockout of SpcI-FLO9-1 and SpcI-FLO9-2 did not abolish flocculation in CBS 1483 but improved overall brewing performance Data integration revealed that among the flocculin genes SpcI-FLO9-1, SpcI-FLO9-2_1, SpcI-FLO9-2_2, and SpeVI-FLO12, all previously associated with initiation of a flocculation phenotype, only SpcI-FLO9-1 and SpcI-FLO9-2_1 exhibited detectable transcriptional activity (Fig. 7). To determine whether these genes were the sole genetic determinants of flocculation in CBS 1483, a double knockout strain lacking both SpcI-FLO9-1 and SpcI-FLO9-2 was constructed, yielding strain IMX3185 (SpcI-flo9-1Δ::NatNT2 SpcI-flo9-2Δ::KanMX). IMX3185 and CBS 1483 were then cultivated in 17° Plato wort in E.B.C. tall tubes to test the effects of the knockouts under brewing-relevant conditions (Fig. 8). As expected, the parental strain exhibited flocculation onset at day 5, followed by a progressive decline in planktonic biomass until day 10. Surprisingly, IMX3185 also exhibited flocculation, with a profile similar to that of CBS 1483, albeit delayed by approximately 24–30 h. By day 10, planktonic cells were largely sedimented in both strains. Consistent with this delayed flocculation, IMX3185 showed slightly improved attenuation. The mutant strain reached a final gravity of 4.25°P, compared to 4.6°P for CBS 1483, which correlated with a lower residual maltose concentration at the end of fermentation (4.7 ± 0.4 g L⁻¹ versus 6.81 ± 0.5 g L⁻¹, respectively). Overall, these results suggest that compensatory mechanisms are likely activated to counteract the absence of SpcI-FLO9-1 and SpcI-FLO9-2, thereby partially restoring the flocculation phenotype under wort fermentation conditions. Measured cell count (cells·mL−1) (A), wort density (°P) (B), and maltose (C) and maltotriose concentrations (D) of CBS 1483 and the double knock-out strain IMX3185 (SpcI-FLO9-1Δ::NatNT2 SpcI-FLO9-2Δ::KanMX) during a 17°P wort fermentation at 12°C in E.B.C. tall tubes. DISCUSSION Flocculation in Saccharomyces species has traditionally been framed as a phenotype driven by highly potent, dominant flocculins such as Flo1, whose overexpression is sufficient to induce rapid and extensive cell aggregation (1, 30, 97). However, our systematic analysis of the FLO repertoire of Saccharomyces pastorianus CBS 1483 challenges this classical single-driver model. Instead, our data support a distributed architecture in which moderately active adhesins, expressed at relatively low levels, collectively shape the characteristic sedimentation behavior of lager yeast. Optogenetic characterization in a FLO-null background (33) demonstrated that SpcI-Flo9-1 and SpcI-Flo9-2_1 possess clear aggregation capacity. However, their intrinsic performance remains below that of canonical ScI-Flo1 under identical transcriptional control. The most likely factor underlying this difference is protein architecture and, more specifically, the number of tandem repeats. Previous studies have established that flocculin activity is directly related to repeat number, with a lower number of repeats generally resulting in reduced flocculation capacity (98). Consistent with this observation, the ScI-FloI from CEN.PK113-7D is 1,675 amino acids long and contains 20 type 1 repeats, whereas the SpcI-Flo9-1 and SpcI-Flo9-2_1 proteins from S. pastorianus CBS 1483 are only 1,043 and 869 amino acids long, and contain six and one type I repeat, respectively. These structural differences likely account for the observed variation in aggregation activity. More intriguingly, they suggest that despite the importance of flocculation in the brewing process, the domestication of lager yeast did not necessarily favor evolution or retention of the most active flocculins. This observation may indicate that an optimal rather than a maximal level of flocculation was selected during the adaptation of S. pastorianus to industrial brewing environments. Therefore, only prolonged induction allowed these proteins to reach comparable flocculation coefficients, indicating lower intrinsic binding strength. SpcI-Flo1-2_2 and the structurally distinct SpeVI-Flo12 showed limited or delayed aggregation potential. Furthermore, several flocculins substantially increased cell surface hydrophobicity without triggering macroscopic flocculation, highlighting that adhesive surface properties and intercellular aggregation are not equivalent phenomena (74, 84). It is important to note that CBS 1483 and the FLO-null platform strain belong to different species, which may influence protein folding, glycosylation, and flocculin function at the cell surface. Differences in cell wall composition and surface architecture may therefore contribute to the observed flocculation phenotypes. Comparative studies have revealed variation in mannoprotein content, cell wall elasticity, polysaccharide structure, and carbohydrate composition among yeast strains (99–101). In addition, differences in the spatial organization of flocculins may influence flocculation behavior (84). However, the contribution of these factors remains poorly understood. Comparative analyses across different strain backgrounds may help identify additional mechanisms underlying FLO-mediated flocculation. Transcriptomic analysis during 17°P wort fermentation revealed that FLO genes are expressed at modest levels relative to highly expressed metabolic genes. Even at peak induction around day 5, coinciding with the onset of sedimentation, transcript levels remained low. No single FLO gene exhibited dominant transcription that would indicate a master flocculin controlling the phenotype. Instead, flocculation onset correlated with moderate induction of multiple adhesins, most notably SpcI-FLO9-1 and SpcI-FLO9-2_1, alongside contributions from SpeI-FLO1-3 and potentially SpeVI-FLO10. This coordinated yet restrained expression pattern is consistent with a network-based model rather than a single-gene determinant (38). The double deletion of SpcI-FLO9-1 and SpcI-FLO9-2 provides further support for this distributed control mechanism. The deletions did not eliminate sedimentation in the E.B.C. tube wort fermentations but merely delayed it by approximately 24–30 h. Ultimately, the mutant sedimented efficiently and even showed improved attenuation due to prolonged sugar consumption. These observations strongly suggest activation of compensatory mechanisms within the FLO network. Such compensation is not unprecedented in yeast biology. Saccharomyces cerevisiae possesses numerous paralogous gene families that exhibit transcriptional compensation upon deletion of a dominant member. A classic example is the pyruvate decarboxylase system: deletion of PDC1, which encodes the major isoenzyme, triggers strong upregulation of PDC5, restoring fermentative flux through autoregulatory mechanisms (102–104). Similarly, deletion of HXK2, the dominant hexokinase under glucose-repressed conditions, induces HXK1 expression (105). Enolase paralogs exhibit comparable behavior, with ENO1 upregulated upon ENO2 deletion (106). Compensation also occurs in ribosomal protein genes (e.g., RPL1A/RPL1B) (107), glucose transporters (HXT family via Snf3/Rgt2 signaling) (108), glycerol-3-phosphate dehydrogenases (GPD1/GPD2) (109), and cytosolic malate dehydrogenases (MDH2/MDH3) (110). These examples illustrate a recurring principle in yeast genetics: functionally overlapping paralogs can buffer perturbations through transcriptional or physiological compensation, thereby maintaining system-level robustness. The FLO gene family, particularly in the hybrid genome of S. pastorianus, is ideally structured for such buffering. CBS 1483 contains multiple subgenome-derived FLO genes, allelic variants with distinct repeat compositions, and structurally diverse adhesins such as Flo12. This redundancy creates a landscape in which loss of dominant contributors may be partially offset by increased relevance, whether transcriptional, post-transcriptional, or functional, of secondary flocculins. One can argue that the selective pressure acting on flocculins differs fundamentally from that acting on metabolic isozymes. Loss of a metabolic isozyme can create an immediate fitness defect by impairing growth, whereas loss of a flocculin does not directly prevent cell proliferation. However, in the context of industrial lager brewing, flocculation is subject to indirect but persistent selection. The brewing process relies on serial re-pitching, whereby flocculated biomass is recovered at the end of fermentation and used to inoculate subsequent batches. Consequently, strains with impaired flocculation may be disadvantaged during repeated propagation cycles, providing a selective pressure to maintain sufficient aggregation capacity. In addition, FLO genes are characterized by extensive tandem-repeat regions that are prone to expansion and contraction. Changes in repeat number can substantially alter flocculin activity and therefore provide a mechanism through which aggregation capacity could be modulated. While we do not claim to have demonstrated a compensatory mechanism, we propose that the selective pressure imposed by industrial re-pitching, together with the structural plasticity of FLO genes, could facilitate the maintenance or restoration of flocculation capacity following the loss or reduced activity of individual flocculins, as observed in this study. Although we did not experimentally validate whether alternative FLO genes were upregulated upon deletion of SpcI-FLO9-1 and SpcI-FLO9-2, compensation in yeast is not always mediated by large transcriptional shifts. In several paralog systems, modest changes in expression, altered metabolic flux, or changes in cellular context are sufficient to restore phenotype (111–113). In wort, environmental parameters such as ethanol accumulation, nutrient depletion (14), pH shifts (17), and Ca²+ availability may alter the effective binding affinity or cell-surface exposure of remaining adhesins (74). Thus, SpeI-Flo1-3 or SpeVI-Flo10, insufficient alone under laboratory conditions, may become functionally significant in the physiological context of fermentation. Importantly, the compensatory phenotype observed here does not restore rapid or excessive flocculation, but rather produces a delayed yet orderly sedimentation profile. From an industrial standpoint, this is advantageous. Excessively strong flocculation can cause premature biomass removal, incomplete maltose and maltotriose consumption, and formation of off-flavors. Conversely, weak flocculation prolongs clarification and increases tank occupancy. The lager yeast phenotype appears finely tuned to initiate sedimentation only after substantial sugar depletion. The NewFlo-like sugar sensitivity of SpcI-Flo9 proteins ensures inhibition during early fermentation stages, while moderate expression and distributed activity prevent abrupt aggregation. Lg-FLO1 is often proposed as the primary determinant of NewFlo-type flocculation in lager yeast strains (40, 41, 43, 86, 97). However, its absence in CBS 1483, consistent with previous reports (46), demonstrates that Lg-FLO1 is not universally required for this phenotype. In CBS 1483, the corresponding locus contains a recombination-derived FLO/HZY1 (YIL169C) hybrid gene, with an additional recombination event on the SpeX-cX chromosome. Such rearrangements have been associated with loss of flocculation in S. pastorianus, indicating that Lg-FLO1 is unlikely to contribute to flocculation in this strain. Despite lacking a functional Lg-FLO1, CBS 1483 clearly displays a NewFlo phenotype. None of the identified FLO genes contain the KALAR/KVLAR motif previously linked to broad sugar recognition, suggesting that this motif is not essential for NewFlo-type flocculation. This is consistent with reports that S. cerevisiae FLO9 and FLO10 can confer NewFlo-like behavior despite lacking canonical sequence features (22, 33, 78), implying that alternative molecular mechanisms underlie sugar sensitivity in flocculins. We therefore propose that the lager flocculation phenotype represents an emergent property of a moderately expressed adhesin network rather than dominance of the strongest flocculins known in S. cerevisiae. The hybrid genome of S. pastorianus may have facilitated this evolutionary outcome by providing a diverse repertoire of adhesins with graded binding capacities. Over centuries of domestication under cold, high-gravity conditions, strains exhibiting balanced sedimentation kinetics, neither too rapid nor too slow, would have been preferentially retained. In this context, compensatory capacity becomes a feature rather than a flaw. The persistence of flocculation upon deletion of major FLO genes demonstrates robustness embedded within the adhesin network. Similar to central carbon metabolism or ribosome assembly systems, redundancy ensures stability under genetic or environmental perturbation. Such robustness likely contributes to the reliability and reproducibility demanded in industrial lager brewing. The assembly of FLO genes is particularly challenging when relying on Illumina short-read sequencing, which remains the most widely used approach for sequencing Saccharomyces pastorianus strains. Members of the FLO gene family share extensive sequence homology while also containing long internal arrays of tandem repeats. Because these highly similar and repetitive regions often exceed the length of individual Illumina reads, assembly algorithms frequently fail to unambiguously assign reads to specific loci or accurately resolve repeat copy numbers. Consequently, FLO genes are prone to fragmentation, repeat collapse, and misassembly in short-read-based genome assemblies, limiting the extension of flocculin analyses to additional S. pastorianus strains. The adoption of long-read sequencing technologies, which generate reads capable of spanning entire repeat arrays and extending into unique flanking regions, should greatly improve the resolution of FLO loci. This would enable more accurate reconstruction of gene structure, repeat number, and copy number variation, thereby facilitating robust comparisons of flocculin repertoires across S. pastorianus strains. In conclusion, our findings support a revised model of flocculation control in S. pastorianus: rather than being driven by a single high-performance flocculin, sedimentation behavior emerges from the collective action of multiple moderately active adhesins whose expression and activity are finely modulated during fermentation. Compensatory interactions within this network buffer genetic perturbations and ensure that biomass depletion proceeds with kinetics optimally aligned to lager fermentation. This systems-level perspective provides a framework for rational strain engineering strategies aimed at tuning, rather than maximizing, flocculation performance.