High Foam Phenotypic Diversity and Variability in Flocculant Gene Observed for Various Yeast Cell Surfaces Present as Industrial Contaminants

: Many contaminant yeast strains that survive inside fuel ethanol industrial vats show detrimental cell surface phenotypes. These harmful effects may include ﬁlamentation, invasive growth, ﬂocculation, bioﬁlm formation, and excessive foam production. Previous studies have linked some of these phenotypes to the expression of FLO genes, and the presence of gene length polymorphisms causing the expansion of FLO gene size appears to result in stronger ﬂocculation and bioﬁlm formation phenotypes. We performed here a molecular analysis of FLO1 and FLO11 gene polymorphisms present in contaminant strains of Saccharomyces cerevisiae from Brazilian fuel ethanol distilleries showing vigorous foaming phenotypes during fermentation. The size variability of these genes was correlated with cellular hydrophobicity, ﬂocculation, and highly foaming phenotypes in these yeast strains. Our results also showed that deleting the primary activator of FLO genes (the FLO8 gene) from the genome of a contaminant and highly foaming industrial strain avoids complex foam formation, ﬂocculation, invasive growth, and bioﬁlm production by the engineered ( ﬂo 8 ∆ :: Ble R / ﬂo8 ∆ :: kanMX ) yeast strain. Thus, the characterization of highly foaming yeasts and the inﬂuence of FLO8 in this phenotype open new perspectives for yeast strain engineering and optimization in the sugarcane fuel-ethanol industry.


Introduction
The renewable Brazilian sugarcane first-generation fuel-ethanol industry is highly competitive, when compared with ethanol production processes from other crops (e.g., corn and sugarbeet), as it shows the highest percentage of greenhouse gas emissions' reduction, highest energy balance and yields per hectare, and the lowest production costs [1,2]. The productivity of the Brazilian fuel ethanol industry has increased steadily since 1975, an increase possibly due to several improvements, including the selection of new sugarcane varieties with an increased amount of sugarcane biomass per hectare and also the amount of ethanol produced from each ton of sugarcane, as the industrial sector responsible for this, including the fermentation process, has also reached high industrial efficiency [3][4][5].
In Brazil, over 70% of the milling and distilling companies perform the Melle-Boinot process, a fed-batch fermentation with high cell density that allows higher production stability when compared to the continuous mode [6,7]. At the end of each fermentation, the yeast cells are centrifuged, treated with diluted sulphuric acid, and over 90% of the yeast cells are reused from one fermentation to the next to ensure the high cell density middle domain that contains many tandem repeats extremely rich in serine and threonine (and also proline) residues that is highly glycosylated, and a C-terminal domain were the glycosylphosphatidylinositol (GPI) anchor is added for localizing the adhesins to the cell surface, although the adhesins may also be covalently linked to the non-reducing end of β-1,6-glucans of the cell wall via a GPI remnant [38,41,42].
The N-terminal domain differs significantly between FLO11 and the other flocculin genes. The structure of the N-terminal domain of the flocculin encoded by FLO11 (Pfam entry PF10182, https://pfam.xfam.org, accessed on 20 May 2020) is composed of three subdomains, including a hydrophobic apical region, a β-sandwich of the fibronectin type III domain, and the neck subdomain [43]. Two aromatic bands (containing tryptophan and tyrosine residues) that are present at the ends of the adhesion domain mediate homotypic and hydrophobic adhesion to a range of solid surfaces and the air-liquid interface of flor (or velum) in sherry-like wine fermentations [42][43][44][45][46]. The N-terminal domain of the other flocculins (e.g., FLO1 or FLO5) have a lectin-like structure (PA14 domain, Pfam entry PF07691) that binds (in a Ca 2+ -dependent manner) to mannose oligosaccharides present in the yeast cell surface [42,47,48]. Consequently, FLO11 and the other FLO genes play distinct roles in adhesion, flocculation, invasive growth, biofilm formation, and other cell surface phenotypes of yeast cells [40,49]. A further level of complexity is determined by the serine/threonine-rich middle domain, which contains intragenic tandem repeats of DNA sequences that trigger recombination, causing variations in gene size, and the expansion of this middle domain contributes to stronger flocculation, biofilm formation, and other cell surface phenotypes of yeast cells [39, 50,51].
The first yeast gene described as specifically involved in foam formation was the AWA1 gene, encoding for a cell-wall GPI-linked and highly glycosylated serine/threoninerich protein from sake-producing yeasts [52]. As with FLO genes, the serine/threonine-rich domain in AWA1 can also vary in size, contributing to different foaming phenotypes of sake yeasts [53,54]. More recently, primers based in the AWA1 gene allowed the identification of two other highly homologous genes, the FPG1 gene responsible for foam formation in a wine yeast strain [55] and the CFG1 gene involved in foam production by S. pastorianus lager brewing yeasts [56].
In the present report, we analyzed several cell-surface properties (flocculation and cellwall hydrophobicity, foam and biofilm production, invasive growth) of a group of 15 wild and indigenous yeast strains isolated from sugarcane industrial fuel-ethanol production plants, including strains with good or bad (e.g., premature flocculation, excessive foam production) industrial characteristics. We also determined gene length polymorphisms in some adhesin genes (FLO1 and FLO11) and analyzed possible correlations with the phenotypic diversity found in this group of contaminant industrial yeast strains. Finally, we showed that a highly foaming industrial fuel-ethanol yeast strain loses its capacity to produce foam when this FLO8 gene is deleted, indicating that the expression of flocculins is involved in problematic (abundant and persistent) foam formation during industrial fermentation by this contaminant S. cerevisiae yeast.

Foam Quantification
Yeast cells were cultivated in rich YPD medium until 1 ± 0.5 g (dry cell weight, DCW)/L, harvested by centrifugation (7.000× g, 5 min) and washed twice with cold, sterile water. A cellular suspension of 10 g DCW/L was prepared in 5 g/L Na 2 SO 4 and added to molasses-based medium (final sugar and other components' concentrations as indicated above), incubated at room temperature under constant agitation (100 rpm) during 8 h, and foam formation was volumetrically monitored during the fermentation. Foam formation was expressed as the percentage of foam volume in relation to the fermentation media volume.

Flocculation Tests
To test cell-cell adhesion, flocculation assays were performed based on the protocol described by Wang et al. [59]. Cells were grown in molasses medium until the late exponential growth phase (3 ± 0.5 g DCW/L), harvested by centrifugation, and washed twice with 50 mM sodium citrate, 50 mM EDTA, and pH 3.0 buffer, followed by two washings with cold, distilled water. After centrifugation, cells were suspended (5 g DCW/L) in 100 mM sodium acetate, pH 4.0 buffer, and their ability to flocculate in the presence or absence of 10 mM CaCl 2 was determined. Briefly, cells were vortexed during 10 s and allowed to stand for 10 min. Aliquots of 30 µL taken from below the meniscus (at 0 and 10 min) were used to determine (after proper dilution) the cell density at 570 nm. Flocculation was expressed as the percentage of sedimented cells after 10 min, relative to the initial value.

Cellular Hydrophobicity Assay
Cells were grown in molasses-based medium as described above, washed twice in cold, distilled water, and suspended in water to an absorbance of 0.8 at 570 nm. Hexane was added to the cellular suspensions at a proportion of 1:3 (v:v) and vigorously mixed. The relative difference in cell density of the aqueous phase was determined by measuring the absorbance at 570 nm before and after vortexing and considered as the hydrophobicity percentage of the cell surface.

Biofilm and Invasive Growth Determination
Biofilm formation was analyzed as described by Reynolds and Fink [60]. The yeast strains were inoculated onto YPD plus 3 g/L agar plates with a sterile toothpick. The plates were maintained resting at room temperature, and the yeast colony was measured and photographed after 5 days. For invasive growth assays, the cells were streaked onto standard YPD plus 2% agar plates, grown at 28 • C for 5 days, and then photographed. Tap water was used to gently wash non-invasive cells from the surface of the agar, and then the plate was photographed again [40].

Construction of Flo8∆ Knockout Strains
Yeast transformation was performed by the lithium acetate method [44]. The FLO8 gene was deleted according to the polymerase chain reaction (PCR)-based gene replacement procedure [62]. Briefly, the kanMX cassette from plasmid pUG6 (Table 1) was amplified with primers M1FLOF1 and M1FLO8R1 (Table 1), and the resulting PCR product of 1779-bp was used to transform competent yeast cells [63] from the laboratory strain S288C. After 2-h cultivation on YP-2% glucose, the transformed cells were plated on the same medium containing G-418 and incubated at 28 • C. G-418-resistant isolates were tested for proper genomic integration of the kanMX cassette at the FLO8 locus by analytical colony PCR using primers (V1FLO8F and INFLO8R; Table 1) that yielded a 2857-bp fragment from a normal FLO8 locus or with primers (V3FLO8F and VKan R R, Table 1) that amplified a 1394-bp fragment if the kanMX cassette was correctly integrated at this locus and replaced the FLO8 gene, yielding strain S288C-flo8∆::kanMX. Using the same strategy above, the FLO8 gene was deleted from the genome of the laboratory strain CEN.PK2-1C using primers M2FLO8F1 and M2FLO8R1 (Table 1) to amplify the Ble R gene from plasmid pUG66 [62]. The resulting PCR product of 1344-bp was used to transform competent yeast cells and zeocin-resistant isolates were tested for proper genomic integration of the Ble R gene at the FLO8 locus by Southern analysis (see below) or analytical colony PCR using primers V1FLO8F and INFLO8R, as described above, or with primers LIM2FLO8R and VBLE R F ( Table 1) that amplified a 956-bp fragment if the FLO8 gene was deleted and replaced by Ble R , yielding strain CEN.PK2-1C-flo8∆::Ble R . To delete both copies of the FLO8 gene present in the industrial diploid strain FT859, we used a long-flanking-homology approach [62] by using primers LIM1FLO8F and LIMFLO8R (Table 1) to amplify a 2118-bp fragment form the DNA of strain S288C-flo8∆::kanMX or primers LIM2FLOF and LIM2FLO8R (Table 1) to amplify a 1646-bp fragment from the DNA of strain CEN.PK2-1C-flo8∆::Ble R . After selection of transformants in G-418 and zeocin plates, the proper genomic integration of the kanMX and Ble R genes at the FLO8 locus was confirmed by Southern analysis (see below) or analytical colony PCR, as described above, yielding the industrial-modified strain CFY02B-B (FT859 flo8∆::kanMX/flo8∆::Ble R ).

PFGE, Chromosome Blotting, and Hybridization
Yeast chromosomes were prepared as previously described [64] from 1 mL of yeast cells pre-grown in YPD medium and collected at the stationary phase of growth. Cells were washed with 10 mM Tris-HCl, pH 7.5, containing 50 mM EDTA, and resuspended in the 0.4 mL of 4 mM Tris-HCl, pH 7.5, containing 95 mM EDTA, 130 µg/mL of Zymolyase 20 T (Sigma-Aldrich Brasil Ltda., São Paulo, SP, Brazil), and 0.7% of molten (42 • C) lowmelting-point agarose. After solidification in a mold (Amersham Pharmacia Biotech do Brasil Ltda., São Paulo, SP, Brazil), the agarose blocks were immersed in 10 mM Tris-HCl, pH 7.5, containing 0.5 M EDTA, and incubated at 37 • C for 8 h. Following a subsequent incubation in 10 mM Tris-HCl, pH 9.5, containing 0.5 M EDTA, 1% N-lauroylsarcosine, and 2 mg/mL proteinase K at 50 • C overnight, the blocks were washed in 10 mM Tris-HCl, pH 7.5, containing 50 mM EDTA, and stored at 4 • C in the same buffer. Each low-melting-point agarose block was transferred to a 1% agarose gel in 50 mM Tris-HCl, pH 8.3, containing 50 mM boric acid and 1 mM EDTA. Pulsed field gel electrophoresis (PFGE) was performed at 8 • C using a Gene Navigator pulsed field system (Amersham Pharmacia Biotech Pharmacia Biotech do Brasil Ltda., São Paulo, SP, Brazil) for a total of 27 h at 200 V. The pulse time was stepped from 70 s after 15 h to 120 s for 12 h. Following electrophoresis, the gel was stained with ethidium bromide and photographed (Gel Doc™ XR, Bio-Rad Laboratories, Hercules, CA, USA).
The chromosomes separated by PFGE were transferred to a nylon filter (Hybond-N + , GE Healthcare, Barueri, SP, Brazil) by capillary blotting [61], and labeling of DNA probes (see below), including the pre-hybridization, hybridization, stringency washes, and chemiluminescent signal generation and detection was performed by using a AlkPhos kit (GE Healthcare, Barueri, SP, Brazil) as recommended by the manufacturer. After hybridization, an autoradiography film (Hiperfilm™ ECL, Kodak, GE Healthcare, Barueri, SP, Brazil) was exposed to the membrane for 1 to 2 h before it was developed. Images were obtained with Image Lab Software (Gel Doc™ XR, Bio-Rad Laboratories, Hercules, CA, USA) and annotated with Microsoft PowerPoint. Probes were generated by PCR using DNA from strain S288C or plasmids pUG6 and pUG66 as templates and primers (Table 1) INFLO8F and INFLO8F for the FLO8 gene, HbFLOsF and HbFLOsR for FLO1, HbFLO11F and HbFLO11F for the FLO11 gene, SDKanF and SDKanR for kanMX, and SDBleF and SDBleR for the Ble R gene, respectively.

Foam Production by Industrial Yeast Strains
We analyzed 15 industrial yeast strains isolated during fermentative fuel-ethanol production processes in Brazil. This group included two strains (PE-2 and CAT-1) with good industrial fermentation characteristics, such as non-flocculent cells with high fermentative capacity, dominance of the industrial fermenters, and production of very small amounts of foam during industrial fuel-ethanol production, and, consequently, these two strains are used by more than half of the industrial fuel-ethanol companies in Brazil [4,19,23]. The other 13 industrial strains (BAT-1 and FT278 through FT859) showed some undesired industrial characteristics, such as premature flocculation and/or sedimentation or excessive foam production during industrial fermentation processes, generally leading to problematic/stuck fermentations with residual sugars not completely fermented in the vats. Figure 1 shows the kinetics of foam production during fermentation of molasses-based medium by three representative strains (CAT-1, BAT-1, and FT281). Strain CAT-1 produced less than 10% (v/v) of foam during fermentation, while strains BAT-1 and FT281 produced significant amounts of foam (more than 60% for strain BAT-1 and up to 90% for strain FT281), with significant differences in the kinetics of foam production. Strain FT281 started to produce foam early during fermentation, and the maximum production peaked after 2 h, but, soon after that, the foam started to decrease and disappeared after 4 to 5 h of fermentation ( Figure 1). Strain BAT-1 started to produce foam earlier during fermentation, but, once the foam accumulated, it persisted all along the 8 h of fermentation ( Figure 1). We observed the presence of high amounts of yeast cells attached to this highly stable, thick, dense, and persistent foam (data not shown). Figure 2 shows the maximum amount of foam produced by all the 15 industrial yeast strains analyzed. Some of them (strains CAT-1, PE-2, FT279, FT540, and FT718) were considered as poor foam producers (less than 20% of the total volume), while the other 10 strains produced 35% or more of foam during fermentation, the majority with a stable and persistent foam all along the fermentation (the exceptions were strains FT281, FT705, and FT714, see Figures 1 and 2).

Flocculation and Cellular Hydrophobicity of Industrial Yeast Strains
Yeast strains were grown in molasses-based medium and flocculation assays were performed in the presence or absence of 20 mM Ca 2+ , and at least three different patterns of yeast flocculation were observed (Figure 3). Six industrial strains showed a high cellular sedimentation (>60% of flocculation), which was mostly calcium independent and only one of them (strain FT718) was not a strong foam producer (see Figure 2). A second group included four industrial yeast strains that presented calcium-dependent flocculation, as the sedimentation rate in absence of calcium was lower than 30%, but increasing to approximately 50% after Ca 2+ addition ( Figure 3). Finally, the remaining five strains (including strains CAT-1 and PE-2) were either non-flocculent or showed an insignificant flocculation rate (less than 30%) in the presence or absence of calcium.

Flocculation and Cellular Hydrophobicity of Industrial Yeast Strains
Yeast strains were grown in molasses-based medium and flocculation assays were performed in the presence or absence of 20 mM Ca 2+ , and at least three different patterns of yeast flocculation were observed (Figure 3). Six industrial strains showed a high cellular sedimentation (>60% of flocculation), which was mostly calcium independent and only one of them (strain FT718) was not a strong foam producer (see Figure 2). A second group included four industrial yeast strains that presented calcium-dependent flocculation, as the sedimentation rate in absence of calcium was lower than 30%, but increasing  to approximately 50% after Ca 2+ addition ( Figure 3). Finally, the remaining five strains (including strains CAT-1 and PE-2) were either non-flocculent or showed an insignificant flocculation rate (less than 30%) in the presence or absence of calcium. In order to quantify the cell wall hydrophobicity of the industrial strains, we analyzed the cellular partitioning into an organic solvent such as hexane. Our results showed a significant variability in cellular hydrophobicity among the industrial yeast strains ( Figure 4), but, again, strains CAT-1 and PE-2 were the ones with the lowest cellular hydrophobicity. Among those strains with high cellular hydrophobicity (60% or more of cells partition into the organic solvent), it is worth to note that all of them had strong flocculation phenotypes (see Figure 3) and only one (strain FT718) was a poor foam producer (see Figure 2).   In order to quantify the cell wall hydrophobicity of the industrial strains, we analyzed the cellular partitioning into an organic solvent such as hexane. Our results showed a significant variability in cellular hydrophobicity among the industrial yeast strains (Figure 4), but, again, strains CAT-1 and PE-2 were the ones with the lowest cellular hydrophobicity. Among those strains with high cellular hydrophobicity (60% or more of cells partition into the organic solvent), it is worth to note that all of them had strong flocculation phenotypes (see Figure 3) and only one (strain FT718) was a poor foam producer (see Figure 2).   Figure 5 shows some patterns of invasive growth and biofilm formation by dustrial strains and in Table 2 are scored the size of mats and the degree of invas into YPD agar plates for all strains. The industrial strains CAT-1 and PE-2 show nificant attachment to the agar surface and, while the other strains were either si even less invasive than these two strains, only one (strain FT278) attached firmly plate surface. Despite the highly variable size of the biofilms formed by the stra ameters ranging from 1.2 to 4.5 cm after 5 days), the mats they produced were and did not form highly structured "fluffy" biofilms, as described earlier for sever laboratory and wild yeast strains [60,65,66]. Nevertheless, biofilm formation seem directly correlated to flocculation, since all strong flocculent industrial strains w to form biofilms larger than 3.5 cm and all non-flocculant strains formed biofilms than 1.5 cm.   Figure 5 shows some patterns of invasive growth and biofilm formation by the industrial strains and in Table 2 are scored the size of mats and the degree of invasiveness into YPD agar plates for all strains. The industrial strains CAT-1 and PE-2 showed significant attachment to the agar surface and, while the other strains were either similar or even less invasive than these two strains, only one (strain FT278) attached firmly into the plate surface. Despite the highly variable size of the biofilms formed by the strains (diameters ranging from 1.2 to 4.5 cm after 5 days), the mats they produced were smooth and did not form highly structured "fluffy" biofilms, as described earlier for several other laboratory and wild yeast strains [60,65,66]. Nevertheless, biofilm formation seems to be directly correlated to flocculation, since all strong flocculent industrial strains were able to form biofilms larger than 3.5 cm and all non-flocculant strains formed biofilms smaller than 1.5 cm. Invasive growth, scored as adhesion to agar surfaces, was determined by patching the yeast cells on YPD plates and incubation for 5 days. The plate was documented before and after gentle washing with water. Strains BAT-1 and FT645 were considered non-adhesive ("─" in Table 2), while the other strains showed different strengths of invasive growth (scored "+" to "+++" in Table 2).

Discussion
The analysis of hundreds of indigenous S. cerevisiae strains isolated from industrial sugarcane fuel-ethanol producing plants in Brazil revealed that more than 67% of the strains produced excessive amounts of foam. Only 20% of the strains did not present undesirable features, including foam, premature flocculation or inability to metabolize all Invasive growth, scored as adhesion to agar surfaces, was determined by patching the yeast cells on YPD plates and incubation for 5 days. The plate was documented before and after gentle washing with water. Strains BAT-1 and FT645 were considered non-adhesive ("-" in Table 2), while the other strains showed different strengths of invasive growth (scored "+" to "+++" in Table 2). 1 Diameter of the mat/biofilm. 2 "-" indicates absence of a characteristic (or gene), and "+" to "+++" the presence and strength of the characteristic (see Figure 5). 3 The expected amplified fragment, based on the genome of strain S288C, is 4.3 Kb for FLO1 and 3.1 Kb for FLO11.

Presence and Flocculin Gene Size Polymorphisms
We designed primers to amplify the highly variable and repetitive serine/threoninerich middle domain present in several GPI-linked cell-wall proteins (Table 1). Considering the highly foaming phenotypes present by some members of this group of industrial yeast strains, we initially tried to amplify and verify fragment length polymorphisms in the already described AWA1 foaming gene [52][53][54]. However, we could not amplify any fragment for this gene from the DNA of any of the industrial yeast strains analyzed. Thus, we focused or analysis on two other important flocculin genes, FLO1 and FLO11 ( Table 2). The FLO1 gene was found in 66% of the industrial yeast strains analyzed and presented significant length polymorphism in the serine/threonine-rich middle domain among the strains (sizes varied from 0.8 to 5.0 Kb), and only two strains (FT281 and FT697) showed PCR fragments longer that the expected 4.3 Kb fragment that is predicted from the known genome of strain S288C. Regarding FLO11, only strain FT697 showed no amplified fragment, and the other analyzed industrial strains showed DNA fragments of varying length polymorphisms (from 0.5 to 2.9 Kb), all of them with lower sizes than the expected size of 3.1 Kb obtained for this gene from the genome of strain S288C.
Nevertheless, we analyzed possible significant correlations (using STATISTIX 8.0 ® Analytical Software, Tallahassee, FL, USA) between the size polymorphisms of these two genes and the phenotypic variability shown by the corresponding industrial strains (Figures 1-5, Table 2). There was a significant correlation of FLO11 length polymorphisms and foam production, cellular hydrophobicity, and flocculation (r 2 ≥ 0.50 and p ≤ 0.05, Spearman nonparametric test). For FLO1 length polymorphisms, the more significant correlation observed was with cellular hydrophobicity (r 2 ≥ 0.62 and p ≤ 0.05) and, to a lesser extent, with foam production (r 2 ≥ 0.42 and p ≤ 0.05). Consequently, there was also a significant correlation between foam production during fermentation and the cellular hydrophobicity (r 2 ≥0.50 and p ≤ 0.05), while this last parameter was significantly correlated with flocculation (in the presence or absence of calcium, r 2 ≥ 0.70 and p ≤ 0.05) and biofilm formation (r 2 ≥ 0.42 and p ≤ 0.05, Pearson parametric test).

Deletion of FLO8 Reduces Foam Production by an Industrial Yeast Strain
Considering the correlation between FLO11 (and also FLO1) gene length polymorphisms and foam production during fermentation by the industrial yeast strains, we decided to see the phenotypic consequences of deleting the FLO8 gene, encoding one of the key transcriptional activators of the FLO1 and FLO11 genes [40,67], in one of the industrial yeast strains (FT859), a strain that produces large amounts of persistent foam (Figure 2), while in the other characteristics we analyzed it had intermediate (e.g., flocculation and invasive growth) to low (e.g., cellular hydrophobicity and biofilm formation) values (Figures 3-5, Table 2). We initially deleted the FLO8 gene from two laboratory yeast strains (using different markers, see Materials and Methods) and afterward we used the long-flanking-homology approach [62] to delete this gene from the diploid genome of the industrial strain FT859. Figure 6 shows a PFGE and blotting analysis of the strains, confirming that the FLO8 gene was specifically removed from the genome of the industrialmodified strain CFY02B-B (FT859 flo8∆::kanMX/flo8∆::Ble R ) and also showing the presence of both the FLO1 (chromosome I) and FLO11 (chromosome IX) genes in its genome.
Strain CFY02B-B (FT859 flo8∆/flo8∆) had no difference in cellular hydrophobicity when compared to strain FT859, but showed a pronounced drop in the flocculation capacity (15 and 20% of sedimented cells in the absence and presence of Ca 2+ , compared to 32 and 45% obtained with strain FT859, see Figure 3), biofilm formation (mats of 0.9 cm of diameter, compared to 1.6 cm for strain FT859, see Table 2), invasive growth (strain CFY02B-B lost its capacity to adhere to the agar surface, while strain FT859 does, see Table 2) and, as shown in Figure 7, also lost its capacity to produce high amounts of foam during fermentation, indicating that induction of FLO genes by FLO8 is required for foam production by this contaminant industrial yeast strain. Strain CFY02B-B (FT859 flo8Δ/flo8Δ) had no difference in cellular hydrophobicity when compared to strain FT859, but showed a pronounced drop in the flocculation capacity (15 and 20% of sedimented cells in the absence and presence of Ca 2+ , compared to 32 and 45% obtained with strain FT859, see Figure 3), biofilm formation (mats of 0.9 cm of diameter, compared to 1.6 cm for strain FT859, see Table 2), invasive growth (strain CFY02B-B lost its capacity to adhere to the agar surface, while strain FT859 does, see Table 2) and, as shown in Figure 7, also lost its capacity to produce high amounts of foam during fermentation, indicating that induction of FLO genes by FLO8 is required for foam production by this contaminant industrial yeast strain.   Table 2), while the other strains showed different strengths of invasive growth (scored "+" to "+++" in Table 2).

Discussion
The analysis of hundreds of indigenous S. cerevisiae strains isolated from industrial sugarcane fuel-ethanol producing plants in Brazil revealed that more than 67% of the strains produced excessive amounts of foam. Only 20% of the strains did not present undesirable features, including foam, premature flocculation or inability to metabolize all

Discussion
The analysis of hundreds of indigenous S. cerevisiae strains isolated from industrial sugarcane fuel-ethanol producing plants in Brazil revealed that more than 67% of the strains produced excessive amounts of foam. Only 20% of the strains did not present undesirable features, including foam, premature flocculation or inability to metabolize all sugar from the medium [19]. Yeasts growing in nature usually form organized multicellular structures (colonies, biofilms) with specific properties that do not normally arise under laboratory conditions. The cooperation of yeast cells and their growth within a community is beneficial in harmful or stressful conditions and might contribute to successful colonization of new environments. Consequently, these morphological transitions from a single-cell budding yeast that forms into organized structures are regulated by several sensing/signaling pathways that allow the formation of pseudohyphae and biofilms and promote their adhesion to other cells or surfaces [49,68,69].
In the case of the Brazilian sugarcane fuel-ethanol production plants there is not only several stressful conditions during the industrial process, but sucrose, the main sugar found in sugarcane juice that the cells will have to ferment [1], is one of the strongest inducers of pseudophyphal growth and filamentation in yeasts [70,71]. For example, after serial transfers of yeast cells in low sucrose-containing medium, of 12 evolved clones, 11 formed multicellular clumps through incomplete cell separation [72], due mainly to nonsense or frameshift mutations in the ACE2 gene, encoding for the transcription factor required for septum destruction after cytokinesis [73][74][75]. Indeed, even a good industrial fuelethanol strain like PE-2 can eventually show recurrent appearance of mutants displaying a mother-daughter cell separation defect resulting in rough colonies in agar media and fast sedimentation in liquid culture [76]. This diploid strain was shown to be heterozygous for a frameshift mutation in the ACE2 gene (ACE2/ace2-A7) and, thus, loss-of-heterozygosity events at the region of chromosome XII where ACE2 is located can produce ace2-A7/ace2-A7 cells with the undesirable phenotype of cellular aggregation [76]. Nevertheless, other reports indicate that mutations that increase the expression of the flocculin genes FLO1 and FLO11 are the favored routes to flocculation in yeasts [75,77].
Our results showed a great diversity of morphological and cell-surface-associated phenotypes in the group of 15 wild and indigenous yeast strains isolated from industrial fuel-ethanol production processes in Brazil that we analyzed. We not only found Ca 2+dependent and -independent flocculation phenotypes, but also some strains showing significant invasive growth, a characteristic of haploid yeast cells not described in diploid strains (as are the industrial strains analyzed). While most laboratory strains do not show these dimorphic transitions or flocculate due to a mutation that inactivates the transcriptional activator encoded by the FLO8 gene [39, 40,78], the known genome of two Brazilian industrial strains included in our study (strains CAT-1 and PE-2, see [57,58]) revealed that these strains have a functional FLO8 gene, but are non-flocculent and do not produce foam due to absence or significant polymorphisms in the coding regions of their flocculin FLO genes (including FLO1 and FLO11, see Table 2).
Although we did not perform a comprehensive analysis of the FLO genes present in the industrial fuel-ethanol yeast strains, an important correlation was found between FLO11 length polymorphisms in the genome of the strains and foam production, cellular hydrophobicity, and flocculation. FLO11 has been implicated in the interaction of yeasts with air-liquid interfaces [44,51,79,80], which is in agreement with our results that indicate the presence of yeast cells in the highly persistent foam produced by some of the industrial strains (Figures 1 and 2). Although none of the industrial strains analyzed had a FLO11 gene with the predicted size (based on strain S288C, see Table 2), but instead had shorter version of the gene, it is also worth noting that all these strains produced biofilms (of variable sizes) that were smooth, resembling the mats and colonies produced by flo11∆ strains [60,65]. Not only increased central tandem repeat domains of the FLO11 coding region were shown to be responsible for the formation of buoyant biofilms at the medium surface, but also alterations in the FLO11 promoter region have been implicated in floating phenotypes of wild flor yeast strains [44,79,80]. Indeed, the FLO11 promoter is one of the biggest promoters described in yeasts (~2.8 Kb), with several cis-regulatory elements (four activating and at least nine repressing sequences) and numerous trans-acting factors implicated in its regulation, including posttranscriptional and epigenetic control by chromatin remodeling complexes, as well as prions [69,[81][82][83][84]. While expression of the FLO11 gene has been shown to be induced under limiting nutritional conditions [85,86], it will be interesting to analyze the expression of this gene during the fermenting conditions that produce high foaming phenotypes in the different genetic backgrounds of the industrial fuel-ethanol yeast strains analyzed. It is worth noting that yeast strains with flocculating phenotypes can easily appear from non-flocculating yeasts cultures [72,74,77,79,87]. Nevertheless, our results indicate that deleting the primary activator of FLO genes (the FLO8 gene) from the genome of an industrial strain avoids problematic foam formation, flocculation, invasive growth, and biofilm production. From an industrial perspective, it is difficult to imagine deleting all functional flocculin genes in a selected strain to warranty singlecell budding yeasts and, thus, the strategy described in this work (FLO8 deletion) opens new perspectives for yeast strain engineering and optimization in the growing sugarcane fuel-ethanol industry.

Conclusions
Our molecular analysis of FLO1 and FLO11 gene polymorphisms present in contaminant strains of S. cerevisae from fuel ethanol distilleries showing strong foam production during fermentation revealed significant correlations with cellular hydrophobicity, flocculation, and highly foaming phenotypes in these yeast strains. Our results also show that deleting the primary activator of FLO genes (the FLO8 gene) from the genome of a contaminant and highly foaming industrial strain avoids problematic foam formation, flocculation, invasive growth, and biofilm production by the engineered strain.

Data Availability Statement:
The data presented in this study are available in the article.