Structural and biochemical characterization of the N-terminal domain of flocculin Lg-Flo1p from Saccharomyces pastorianus reveals a unique specificity for phosphorylated mannose



A. Henriksen, Novo Nordisk, Novo Nordisk Park, DK-2760 Måløv, Denmark

Tel: +45 3079 4704



The mechanism of yeast flocculation is generally considered to be mediated through the interaction of cell surface flocculins and mannan carbohydrates. In the present study, the crystal structure of the soluble 25-kDa lectin domain of flocculin 1 from brewer's yeast (Lg-Flo1p) was resolved to 2.5 Å, and its binding specificity towards oligosaccharides was investigated by fluorescence spectroscopy. Lg-Flo1p displays broad specificity towards sugars and has a 14-fold higher affinity for mannose 1-phosphate and glucose 1-phosphate compared to their unphosphorylated counterparts. Based on the results of a structural analysis, we propose that this higher affinity is the result of a charge interaction with a lysine residue in a carbohydrate-binding loop region, NAKAL, unique to NewFlo type flocculins. This raises the possibility of a unique mechanism of flocculation in NewFlo type yeast, which recognizes phosphorylated cell surface mannans.


Structural data have been deposited in the Protein Data Bank under accession number 4GQ7.


carbohydrate-binding loop 1


carbohydrate-binding loop2


epithelial adhesion 1


flocculin 1 from Saccharamyces cerevisiae


flocculin 5 from Saccharamyces cerevisiae


outer loop 1


outer loop 2


outer loop 3


flocculin 1 from brewer's yeast


Protein Data Bank


One of the best-studied examples of cell to cell adhesion is the flocculation of baker's yeast. This interaction is defined as the nonsexual and calcium-dependent aggregation of cells, mediated by the binding of cell surface lectins (flocculins) to surface carbohydrates (mannans) of neighbouring cells [1-7]. The lectin–carbohydrate interaction is often characterized by rather low affinities, which are assumed to be compensated for by an avidity effect and/or a high local concentration of carbohydrates at the cell surface [8]. The genetics of flocculation in the laboratory yeast Saccharomyces cerevisiae are well characterized and involve the expression of cell wall flocculins from at least one of four structural genes: FLO1, FLO5, FLO9, or FLO10 [6].

The flocculation phenotype of yeast is commonly described with respect to whether flocculation can be inhibited by mannose or by other sugars as well. This disruption of flocculation occurs as the sugar competitively displaces the cell wall mannans from the flocculin binding site. The phenotype governing flocculation is usually categorized as being Flo1-type or NewFlo-type and depends on whether the yeast flocculation can be inhibited by mannose, as in the first case, or by a broad range of sugars including mannose and glucose, as in the latter case [9]. Lg-FLO1, a flocculin gene homologous to FLO1 first discovered in Saccharomyces pastorianus but, more recently, also reported in S. cerevisiae [10], has been shown to be responsible for the NewFlo phenotype [11].

Flocculation of yeast is very relevant in biotechnology applications, specifically as a separation tool to remove yeast cells from fermented products, such as beer, wine and ferments used in the distilling industry [12]. NewFlo phenotype yeast is of particular importance in brewing because its sensitivity to wort sugars (glucose, maltose, maltotriose) ensures that flocculation is delayed until the fermentable sugars are depleted [13].

Flocculation is a complex process dependent on many factors, including the regulation of flocculation genes [14-17], cell surface factors (charge and hydrophobicity) [18] and the growth environment, such as pH, temperature, ethanol, oxygen and calcium availability [1, 2, 9, 19]. To fully understand this adhesion process, it is essential to understand all of the aspects involved, including the structural and functional properties of the flocculins. Until recently, little was known about the detailed biochemical properties of the flocculin–carbohydrate interactions because few studies had been conducted outside the complex cellular environment. Recent biochemical and structural studies involving the Flo1-type flocculins flocculin 1 from S. cerevisiae (Flo1p) and flocculin 5 from S. cerevisiae (Flo5p) have attempted to characterize the binding specificity of the flocculins towards terminal yeast mannans [20, 21] and to explain the structural basis for calcium dependency and mannose recognition [21].

The structure of yeast cell wall mannan consists of a long linear α-(1,6) linked mannose backbone to which α-(1,2) and α-(1,3) linked mannose side chains are attached. From structural studies of Flo5p, the flocculin recognizes the terminal α-(1,2) linked mannose side chains [21]. However, in some yeasts, such as S. cerevisiae and the pathogenic yeast Candida albicans, the outer chains of the yeast mannans are further modified by mannosylphosphate, resulting in a net negative charge on the yeast cell surface [22]. In S. cerevisae, these mannosylphosphate residues are added on to α-(1,2) side chains of cell wall mannans by a Golgi mannosyltransferase encoded by the MNN6 gene [23]. The role of mannosylphosphorylation is not well understood. However, it has been proposed that it may play a role as a virulence determinant in C. albicans [23]. In the present study, we examined the role of mannose phosphorylation on flocculation.

In the present study, we focused on the biochemical characterization of a NewFlo-type flocculin from the brewer's yeast Saccharomyces carlsbergensis (syn. S. pastorianus), which was previously cloned, expressed and purified. We present the first crystal structure of the lectin domain of flocculin 1 from brewer's yeast (Lg-Flo1p) resolved to 2.5 Å and compared it with the sugar-complexed structure of Flo5p and related epithelial adhesion 1 (Epa1p) domain from Candida glabrata. We also investigated the sugar-binding specificity of Lg-Flo1p towards a number of mono-, di- and polysaccharides derivatives, including phosphorylated sugars.

Results and Discussion

Sugar specificity of Lg-Flo1p

The interaction between the purified lectin domain of Lg-Flo1p and various mono-, di- and polysaccharides was investigated by fluorescence spectroscopy (Table 1 and Supporting information, Doc. S1). One characteristic of the NewFlo phenotype yeast is that flocculation can be inhibited by both mannose and glucose. As expected, Lg-Flo1p was shown to bind to both monosaccharides but displayed a 7.5-fold higher affinity for mannose (0.77 mm) compared to glucose (5.8 mm). Stereoisomers of mannose, differing in configuration at carbons 3 and 4 (i.e. altrose, talose and gulose) resulted in the abolishment of binding. These observations confirm the importance of the C3 and C4 hydroxyl configuration in saccharide binding, as proposed in previous studies [9, 21, 24]. One particular important observation, which will be discussed further, was the 10-fold decrease in KD, when mannose/glucose was phosphorylated at the C1 position.

Table 1. Dissociation constants for the interaction between Lg-Flo1p and different sugars at pH 5.0
SugarKD (mm)
  1. aCould not be solubilized to more than 10 mm. b Could not be solubilized to more than 100 mm. c Could not be measured above 100 mm as a result of absorption of the sugar at 275 nm.

Mannose 1-phosphate0.057 ± 0.0015
Mannose0.77 ± 0.028
Methyl-α-mannoside3.2 ± 0.067
Methyl-β-mannoside10 ± 0.56
N-acetyl mannosamine51 ± 2.2
Mannose 6-phosphate> 10a,
Glucose 1-phosphate0.41 ± 0.025
Glucose5.8 ± 0.80
Methyl-α-glucoside9.9 ± 0.21
N-acetyl glucosamine13 ± 1.2
Methyl-α-glucoside36 ± 3.4
Glucose 6-phosphate> 100c
Fructose14 ± 0.91
Altrose, talose, gulose> 100b
α-(1,6)-mannobiose3.0 ± 0.33
α-(1,3)-mannobiose3.9 ± 0.98
α-(1,2)-mannobiose4.5 ± 0.38
β-(1,4)-mannobiose47 ± 5.5
β-(1,4)-glucobiose(cellobiose)26 ± 1.3
α-(1,6)-glucobiose (isomaltose)28 ± 0.63
α-(1,1)-glucobiose (trehalose)31 ± 3.3
α-(1,4)-glucobiose (maltose)80 ± 7.0
Sucrose11 ± 0.54
α-(1,6)-(α-1,3)-mannopentaose3.2 ± 0.19
α-(1,6)-(α-1,3)-mannotriose3.5 ± 0.27
α-(1,4)-(α-1,4)-maltotriose59 ± 1.9

Because it is proposed that flocculins recognize the terminal mannose residues of yeast surface mannans, the specificity of Lg-Flo1p towards α-(1,2)-, α-(1,3)- and α-(1,6)-dimannoses was examined. It was notable that Lg-Flo1p displayed a four- to seven-fold weaker affinity for the disaccharides than for mannose, suggesting that the extra sugar ring causes some steric hindrance, which is not compensated by additional stabilizing interactions. This binding trend is also mirrored in Flo1p studies where mannose binds 2.5- to 3.5-fold tighter (0.058 mm) compared to the dimannoses [20]. These results, however, are in contrast to Flo5p, where preferential binding is towards α-(1,2)-mannobiose (3.5 mm) compared to mannose (29 mm), and α-(1,3) and α-(1,6) dimannoses display no binding. Because deglycosylation has been shown to affect the substrate affinity of Flo1p [20], one possible explanation for the different binding trends among the three flocculins could be explained by the glycosylation state: Lg-Flo1p and Flo1p flocculins were expressed recombinantly in yeast and are thus glycosylated, whereas Flo5p was expressed in Escherichia coli and unglycosylated. It is, however, also possible that the Flo5-subdomain plays a role in disaccharide binding.

The branched mannoses had an affinity equal to α-(1,6)-mannobiose, suggesting that flocculin does not have an extended binding site capable of making stabilizing interactions wit several sugar units. The dissociation constant of mannan could not be determined as a result of absorption of the solution at a broad part of the spectrum, including the excitation wavelength, causing a pronounced decrease in the fluorescence intensity. Furthermore, no mannan–flocculin interaction was observed when investigated by surface plasmon resonance or when running the protein through a mannan column (data not shown). These binding observations support the idea that flocculin binds the outermost sugars of branched mannans [25].

Another consideration is that, depending on the strain and growth state, yeast mannans contain phosphodiester linkages between mannose units in the outer branches, including the α-(1,6) linkage [26]. The strong affinity of Lg-Flo1p towards mannose 1-phosphate may suggest that binding to the terminal phosphodiester linkages of yeast mannan plays a role in flocculation. The proposed structural basis for phosphodiester recognition is described further below.

Effect of Ca2+, pH and ionic strength on mannose binding

Additionally, we investigated how Lg-Flo1p-mannose interaction is affected by calcium, pH and ionic strength, which are important environmental factors affecting yeast flocculation [1-7, 9].

Calcium is known to be essential for flocculation. Binding of calcium was monitored indirectly by comparing the fluorescence of the fluorescent calcium chelator 2-([2-[bis(carboxymethyl)-amino]-5-methylphenoxy]methyl)-6-methoxy-8-bis(carboxymethyl)aminoquinoline (Quin2) [6, 27] in the absence and presence of flocculin. Titration of calcium-free flocculin with mannose showed no effect in the fluorescence of flocculin, indicating that mannose could not bind in the absence of calcium. The dissociation constant of Ca2+ to Lg-Flo1p was determined to be 14 ± 3 nm (Supporting information, Fig. S1). At this dissociation concentration and under normal physiological conditions, it is most likely that the flocculins are saturated with calcium.

The influence of pH on the interaction with mannose was investigated by determining the dissociation constant values at various pH in the range 3.25–9.0 (Supporting information, Fig. S2). A pH optimum for binding was found at pH 5.0, with binding decreasing rapidly at lower pH. The pH profile for binding of mannose in vitro closely correlated with the pH profile of brewer's yeast strain CG2164 flocculation in vivo (Supporting information, Fig. S2). This weaker binding at low pH likely results from decreased calcium binding as a result of the protonation of carboxylate groups and unfolding of the protein, as confirmed by the pH effect on the fluorescence spectrum of flocculin itself and CD measurements (data not shown).

Finally, fluorescence studies showed that the ionic strength did not have any pronounced effect on the binding between Lg-Flo1p and mannose, although it did have an effect on mannose 1-phosphate binding. When the mannose is phosphorylated, the nature of the binding becomes more electrostatic and therefore more dependent on the ionic strength (Table S1).

Structural comparison of Lg-Flo1p with Flo5p

Overall structure

The crystal structure of Lg-Flo1p, determined to 2.5 Å, was solved by molecular replacement using the related Flo5p flocculin carbohydrate-binding domain as a search model. Overall, the structure is very similar to the crystal structure of Flo5p [Protein Data Bank (PDB) code: 2XJP] and can be superposed with a Cα-rmsd of 0.30 Å over 216 residues. Similar to Flo5p, Lg-Flo1p is comprised of a PA14-like domain (Pfam: PF07691) consisting of a β-sandwich fold made up of two antiparallel β-sheets and an L-shaped region composed of the N- and C-terminal regions. A Flo5 subdomain insertion after β-strand 5, which is present in Flo1p and Flo5p proteins (Fig. 1), is replaced by a short highly flexible loop 2 (L2) in Lg-Flo1p (Figs 1 and 2).

Figure 1.

Multiple sequence alignment of selected flocculin and Epa carbohydrate-binding domains. Alignment was generated using clustalw and prepared using espript [44]. Secondary structural elements of Lg-Flo1p were assigned using dssp [45]. Important structural features are highlighted, including inner CBL 1 and CBL2 (green), the Flo5 subdomain and the outer loops L1, L2 and L3 flanking the carbohydrate-binding site.

Figure 2.

Overall structure of Lg-Flo1p and a comparison with related structures. (A) Cartoon representation of Lg-Flo1p, coloured according to temperature factor (B-factor). The colour spectrum is denoted using blue as the minimum and red as the maximum B-factor values. The CBL1 and CBL2 loops of the carbohydrate-binding site are indicated by arrows and the DcisD motif coordinating the calcium ion (red sphere) is represented by sticks. Three outer loops flanking the carbohydrate-binding site are denoted as L1, L2 and L3. Two N-acetyl glucosamine residues refined as fully occupied sites are shown as grey sticks. (B) Superposition of the Lg-Flo1p (blue), Flo5p apo (grey) and Epa1p (tan) structures. A disulfide bond from the L1 to L2 loop in the Epa1p structure is represented by orange sticks.

Carbohydrate-binding site

The carbohydrate-binding site of flocculins is characterized by two inner carbohydrate-binding loops (CBL1 and CBL2) and outer loops (L1, L2 and L3) (Figs 1 and 2). CBL1 and CBL2 are both involved in Ca2+ binding and, additionally, in ligand binding in the case of CBL1 (Fig. 3). CBL1 contains a unique DcisD binding motif featuring a cis-peptide bond between two highly conserved Asp residues that coordinate with Ca2+ and the 3′ and 4′ OH groups of hexose ligands. The CBL2 loop is variable in sequence (Fig. 1) and contains a conserved Asn residue involved in Ca2+ binding, and less conserved residues that are involved in Ca2+ coordination via their main chain carbonyl groups (Fig. 3A).

Figure 3.

Comparison of carbohydrate-binding sites of Lg-Flo1p, Flo5p and Epa1p. (A) Superposition of Lg-Flo1p (green) and apo-Flo5 structures (grey). CBL1 and CBL2 loops of Lg-Flo1p are coloured light green and dark green, respectively. Coordination with the calcium (Lg-Flo1p)/sodium (apo-Flo5) ion (red sphere) is indicated by dashed lines. (B) Superposition of Lg-Flo1p and Flo5-mannose structures. Mannose is represented by an orange stick figure and important interactions with mannose are indicated by dashed lines. Colouring of protein residues is identical to (A). (C) Superposition of Lg-Flo1p and Epa1p (tan) structures. A galactose moiety of lactose is coloured purple.

The structure of Lg-Flo1p was compared with that of the apo and mannose bound structures of Flo5p (PDB codes: 2XJQ and 2XJP) (Fig. 3A,B). In Lg-Flo1p, the residues involved in the coordination of Ca2+ are CBL1 residues D134 and D133 and CBL2 residues K199 and L201. In the Lg-Flo1p structure, the side-chain carboxy-groups of D133 and D134 are aligned horizontally and separated by a distance of 2.6 Å (Fig. 3A). This is in contrast to the apo-Flo5p structure, where D160 of the DcisD motif is not involved in the coordination of Na+ (Fig. 3A) but only flips and coordinates the cation when Ca2+ is present in the binding site (Fig. 3B) [9, 21]. Additionally, in the mannose bound state, the D161 side chain is flipped by approximately 70° to form an additional hydrogen-bond interaction with C3-OH of the hexose ligand (Fig. 3B). Although no ligand was detected in the Lg-Flo1p binding site, the DcisD configuration is most similar to that of the mannose bound form of Flo5p, and only requires the rotation of the D134 carboxy-group to allow interaction with the sugar during binding. One possible explanation why D133 maintains coordination with Ca2+ is that the R203 residues (Fig. 3A), comprising a Thr in Flo5p, restrict the flexibility of D133 via an NE–carboxy-group interaction.

Structural comparison of Lg-Flo1p with Epa1p

Epithelial adhesins are another member of the PA14 superfamily, and are responsible for mediating cell–host interactions via their lectin domain. The crystal structure of Epa1p from the pathogenic yeast C. glabrata was first solved by Ielasi et al. [28] (PDB code: 2A3X) and reveals structural conservation with the overall structure of Lg-Flo1p (rmsd Cα = 1.7 Å over 220 residues). The carbohydrate-binding architectures of Epa1p and Lg-Flo1p are similar in that both lack a Flo5 subdomain but vary considerably in some of the outer loop structures. In Lg-Flo1p, the Flo5 subdomain is replaced by a highly mobile loop L2. However, in Epa1p, L2 is stabilized via a disulfide linkage connection to loop 1 (L1) (Fig. 2). In the Flo5p and Lg-Flo1p structures, L1 is shorter by seven residues and no disulfide linkage is made to the Flo5p subdomain or L2 loop, respectively (Fig. 2). Apart from the DcisD motif of CBL1, there is low sequence and structural conservation around the carbohydrate-binding site (Figs 1 and 3c). Compared to Lg-Flo1p, the binding cleft of Epa1p is narrowed through contributions of bulky side chains of W179 from loop L1, R226 and Y228 from CBL2, and W198 from loop L3, providing an explanation for the higher specificity of Epa1p towards glycan molecules, as suggested by Ielasi et al. [28].

FLO versus NewFlo substrate specificity

Previous studies have shown that certain mutations can convert Flo phenotypes to NewFlo [11, 13]. Mutational and structural studies with Flo5p have shown Q98 in the Flo5p-subdomain to be directly involved in the discrimination between mannose and glucose [14-17, 21]. From the structure of Flo5p, it was observed that residue Q98 from Flo5 subdomain interacts with the 2′ OH of mannose (Fig. 3B) and, furthermore, is displaced in the presence of glucose. The absence of a Flo5p domain, and thus discriminating Gln residue in Lg-Flo1p, is most likely the reason for its ability to recognize both mannose and glucose sugars.

D202T, N224A and W228L mutation in Flo5p were also shown to lack discrimination between mannose and glucose. These residues are T175, N197 and L201, respectively, in Lg-Flo1p. N197 and L201 are located in the CBL2 loop region comprising residues NAKAL in Lg-Flo1p (Fig. 1). In Lg-Flo1p the N197 side-chain OD2 atom is directly involved in coordinating the Ca2+ ion, whereas K199 and L201 interact via their backbone carbonyl oxygens (Fig. 3A). T175 is further removed from the carbohydrate-binding site and is located on a variable loop (L3) in the structure (Figs 1 and 4). From a structural comparison of the Lg-Flo1p, Flo5p-mannose and D202T-Flo5p-glucose structures, all three loops adopt different conformations. L3 displays two different conformations in the Flo5p-man structure but, in the D202T mutant, the loop is locked in a conformation further away from the binding site. In the Lg-Flo1p structure, L3 adopts a conformation close to the carbohydrate-binding site (Fig. 4). However, it should be noted that the L3 loop displays high B-factors compared to the rest of the protein (Fig. 1A), and thus may be quite flexible in a liquid environment. Molecular dynamic studies on Epa1p have also shown this L3 region to reorganize and become more ordered upon ligand binding [18, 29]; thus, modifications in this area could potentially affect this process.

Figure 4.

Comparison of the variable L3 loop region of Lg-Flo1p and Flo5p. Structural superposition of Lg-Flo1p (B-factor colouring as in Fig. 2), D202T-Flo5p and Flo5p-man structures. For clarity only, the variable loops (L3) of D202T-Flo5p (pink) and Flo5p-man (grey) with their respective T202 and D202 residues (represented as sticks) are shown. In Lg-Flo1p, the corresponding residue is T175 (green stick). The Flo5p-man loop is found in two different conformations, whereas the D202T Flo5 mutant is locked in one conformation.

It is not evident from the structural analysis how the L3 and CBL2 loop regions are involved in determining Flo from NewFlo behaviour because they are not directly involved in the binding of hexoses, although they might be involved in defining the water structure in the carbohydrate-binding site [1, 2, 9, 19, 21]. Recent studies in Epa1p have shown that subtle structural changes resulting from the modifications of residues in the CBL2 region can result in a relaxation of substrate specificity in the outer subsite through steric accommodation and modification of the electrostatic environment [20, 21, 29]. Because the structure of a Lg-Flo1p-sugar complex has not yet been solved, it remains to be determined how or whether the structure of the CBL2 and L3 loops, and additionally the highly flexible L2 loop (Fig. 1A), will change upon ligand binding.

Recognition of phosphorylated sugars

The flocculation properties of the NewFlo flocculent brewer's yeast strain CG2164 were investigated in vivo and inhibition experiments were performed with mannose 1-phosphate, glucose 1-phosphate, mannose and glucose (Fig. 5). The results obtained show that all four sugars could inhibit flocculation, although at different efficiencies. Glucose was effective in the 100 mm range; glucose 1-phosphate and mannose were slightly better; and mannose 1-phosphate was the most efficient at inhibiting flocculation in the 10 mm range. The flocculation inhibition studies revealed that these four sugar exhibit the same inhibition trends as that exhibited in the in vitro binding studies where mannose 1-phosphate and glucose 1-phosphate both display an approximately 14-fold higher affinity than their nonphosphorylated counterparts (Table 1). This suggests that the inhibition of flocculation is directly linked to the disruption of carbohydrate–lectin interaction by the specified sugar and not by other mechanisms.

Figure 5.

In vivo flocculation inhibition assay. Inhibition of flocculation of the brewers yeast strain CG2164 by various sugars at pH 4.5. Analyzed sugars: mannose 1-phosphate (□), glucose 1-phosphate (○), mannose (■) and glucose (●).

The electrostatic potential surface maps calculated for the Lg-Flo1p structure indicate that the carbohydrate-binding site is positively charged as a result of residue K199 belonging to CBL2 (Fig. 6). When mannose 1-phosphate is modelled in the binding site, the location of the phosphate group is in close proximity to the lysine. Thus, the higher affinity for the 1-phosphorylated glucose/mannose sugars could be a result of a charge interaction of the negative phosphate group with residue K199 in the carbohydrate-binding site of Lg-Flo1p (Fig. 6). It is interesting to note that glucose 6-phosphate and mannose 6-phosphate show no binding to Lg-Flo1p (Table 1), perhaps as a result of steric hindrance of the phosphate with the active site.

Figure 6.

Electrostatic surface representation of (A) Lg-Flo1p and (B) Flo5p carbohydrate-binding sites. The locations of the binding sites are mapped by a mannose molecule (green) in the Flo5p structure and a modelled mannose 1-phosphate molecule (yellow) in the Lg-Flo1p structure. The Lg-Flo1p-mannose 1-phosphate structure was modelled based on the structure of Flo5p–mannose complex. The surface charge distribution is coloured accordingly: red: −1 kT/e and blue: +1 kT/e.

Notably, the electrostatic potential of the Flo1p and Flo5p carbohydrate-binding site is predicted to differ from that of Lg-Flo1p because the CBL2 Lys is replaced by a conserved Val in the Flo1-type flocculins (Figs 1 and 3). To our knowledge, there are no biochemical studies investigating the binding of phosphorylated mannose to Flo5p or Flo1p. However, in vivo studies have shown that the flocculation of laboratory S. cerevisiae strain M24 displaying the Flo phenotype could only be inhibited by mannose and not mannose 1-phosphate, glucose 1-phosphate, maltose, glucose, trehalose or glucose (data not shown). This suggests that the Flo1 phenotype is not sensitive to phosphorylation, in contrast to the NewFlo phenotype, and may be explained structurally by a lack of appropriate groups such as Lys and Arg in the carbohydrate-binding site.

Biological considerations

The phenomenon of flocculation of yeast cells is influenced by various environmental and genetic factors [21, 30] and results from interactions between lectins and mannose chains present on the cell surface [1-5, 9, 21, 24].

We have demonstrated that certain phosphorylated sugars, in particular mannose 1-phosphate, bind tightly to Lg-Flo1p and inhibit flocculation of NewFlo type yeast and not Flo type yeast. We have also shown that the CBL2 lysine residue present only in Lg-Flo1p and not Flo1p or Flo5 may be responsible for the charge interaction with the phosphate group.

When considering the actual sugar abundance during brewing, it is likely that glucose is the main sugar involved in binding to Lg-Flo1p and prevention of the flocculin–yeast mannan interaction. Once fermentation is nearing completion and the glucose from the wort has been depleted, the flocculin is free to interact with mannans from neighbouring cells. As a result of our observations with phosphorylated compounds, we propose that, at least for the NewFlo phenotype mediated by Lg-Flo1p, flocculation can occur through interaction of the flocculin with the cell surface phospho-mannans and that this flocculation mechanism will be more efficient than flocculation via cell surface mannans. Phospho-mannans occur in yeast cell walls and are located at the terminal ends of N-linked oligosaccharides [26]. We propose that Lg-Flo1p could recognize the terminal mannosylphosphate units of certain yeast mannan structures, such as those observed in S. cerevisiae mannosyl transferase mutants MNN1 [25], MNN9 and MNN10 [31]. In these structures, the terminal mannosylphosphates have the C3-OH and C4-OH groups exposed for interaction with Lg-Flo1p residues and the phosphate group attached to the C1 position, analogous to the mannose 1-phosphate structure.

The phenotypic definitions of the Flo1 and NewFlo flocculation types are obviously a very simplified way of describing the flocculation characteristics of a yeast strain. This depends solely on whether flocculation can be inhibited only by mannose or by other sugars as well. The discovery that the phosphorylation of ligands affects the binding affinity adds more diversification to the distinction of flocculation phenotypes. However, we have introduced yet another important factor that should be considered when aiming to understand yeast flocculation.

Materials and methods

Crystal structure determination

The expression, purification and crystallization of Lg-Flo1p, comprising residues 26–245 of Lg-Flo1 (UniProt ID: B3IUA8), have been described previously [20, 32]. The crystal structure of Lg-Flo1p was solved by molecular replacement [20, 33] using a Flo5p structure (PDB code: 2XJV) with a truncated Flo5p subdomain as a model. The Lg-Flo1p structure was refined with several cycles of restrained refinement using refmac 5 [34], alternating with manual model adjustment in coot [35], followed by the addition of ligands [N-acetyl glucosamine (Fig. 2A) and Ca2+] and water [36]. C-terminal residues 243–245 could not be built as a result of poor electron density. A summary of the X-ray data and refinement statistics is provided in Table 2.

Table 2. X-ray data and structure refinement statistics
  1. a Values in parentheses refer to the highest resolution bin. b Based on molprobity validation [43].

Data collection
Space groupP 21 21 21
Unit cell parameters
a, b, c (Å)36.1, 59.9, 82.3
α, β, γ (°)90, 90, 90
Resolution range9.99–2.53
Total number of reflections31822
Number of unique reflections5666
Completeness (%)97.9 (89.1)a
R merge 0.049 (0.212)
II5.9 (3.3)
Number of protein atoms1742
Number of water molecules39
Number of N-acetyl glucosamine ligands2
rmsd bond lengths (Å)0.006
rmsd angles (°)1.020
Mean B-factor (Å2)19.6
Ramachandran plotb (%)
PDB code 4GQ7

The pdb2pqr server [37] was used to prepare the structure for electrostatic calculations, and electrosurface potentials of Lg-Flo1p were calculated by apbs [38] using an ionic strength of 0.15 m. Figures were generated using pymol [39].

Mannose 1-phosphate was modelled into the active site of Lg-Flo1p using the mannose coordinates from the Flo5p-mannose structure (PDB code: 2XJP) as a starting scaffold model. A phosphate group was added to the mannose in the O1 position and energy-minimized using the pef95sac force field [40] with moe software (Chemical Computing Group, Montreal, Canada).

KD determination using fluorescence spectroscopy

The interaction between Lg-Flo1p and various mono-, di- and polysaccharides was investigated by fluorescence spectroscopy (Table 1). Fluorescence measurements were performed on an LS50B luminescence spectrometer (Perkin Elmer, Boston, MA, USA) at 21 °C in a 3.5-mL quartz cuvette (Sigma-Aldrich, St Louis, MO, USA). An excitation wavelength of 275 nm, emission wavelength pf 300–400 nm, scan speed of 50 nm·min−1, excitation slit width of 10 nm and emission slit width of 3–7 nm were used depending on the sample. Each binding experiment was a series of measurement points gradually expanded to optimize coverage of the KD (Supporting information, Fig. S3).

The initial sample volume was 2 mL containing 5 μm Lg-Flo1p and buffer (0.08% Brij 35 (Sigma-Aldrich), 50 mm acetic acid, pH 5, and 1 mm CaCl2). This was titrated with 2–80 μL of solution containing 5 μm flocculin, buffer and sugar. At each titration step, the sample was mixed three times with a mixer (Sigma-Aldrich) and measured immediately. The duration of each step was approximately 3 min.

By titration of the sugar, the KD was determined by nonlinear regression of changes in the fluorescence emission of the protein. All titration curves were fitted using grafit 3.09b, for one site ligand binding. The fits were based on a function that accounts for the fact that the sugars contained impurities, which, to varying extents, absorb at the excited wavelength:

display math

where KD = [Flo] × [ligand]/[Flo:ligand], A is the absorption coefficient of impurities, and [Flo:ligand] is the concentration of flocculin bound to ligand.

All sugars were purchased from Sigma-Aldrich, except α-(1,6)-α-(1,3)-mannotriose, α-(1,6)-α-(1,3)-mannopentaose, α-(1,3)-mannobiose, α-(1,6)-mannobiose and α-(1,2)-mannobiose, which were purchased from Dextra Laboratories (Reading, UK). Furthermore, β-(1,4)-mannobi, -tri, -tetra and -pentaose were obtained from MegaZyme (Wicklow, Ireland). d-trehalose and d-maltose were purified by re-crystallization to remove impurities absorbing at 275 nm.

In vivo flocculation measurements

Flocculation of the S. pastorianus brewer's yeast strain CG2164 (Carlsberg, Copenhagen, Denmark) and the laboratory strain M24 (MATa his4-delta24 LEU ADE SUC1 FLO1) of S. cerevisiae [41] was investigated by a slightly modified Helms test [42]. The cells were grown to stationary phase in yeast extract peptone dextrose at 20 °C for 3 days to develop flocculation. Cells from 1 mL of culture were harvested and washed twice in wash buffer (50 mm acetic acid, 50 mm EDTA, pH 4.5) and then resuspended in water. In a 10-mL tube, an appropriate amount of cells was mixed with 10× flocculation buffer (1 m acetic acid, 100 mm CaCl2, pH 4.5) and water to give a final density of cells of D600 = 1. After gentle inversion of the tubes for 2 h at room temperature, they were placed in the upright position and the cells were left to sediment for 5 min. The top most 0.5 mL of sample was collected and mixed with 0.5 mL of 500 mm EDTA as sample A. The remaining cells were brought into suspension by vigorous agitation and a 0.5-mL sample was collected and mixed with 0.5 mL of 500 mm EDTA as sample B. D600 was measured from samples A and B and flocculation was then calculated as 100 × (B − A)/B. The influence on various sugars on the flocculation characteristics was investigated by adding these sugars at various concentrations to the flocculation reaction.


The authors thank the MAX II Beamline I711 for providing research support. This research was funded through contributions from the Carlsberg foundation and DANSCATT.