## Introduction

Fungal infections have emerged over the past decades as a serious threat for human health1,2,3. Among all human fungal pathogens, Candida albicans is the most frequently encountered species in hospitals. Though a normal resident of the human gastrointestinal or genital tracts, C. albicans can under certain circumstances proliferate and cause superficial benign infections such as oral or vaginal thrush. However, the combination of local factors (broad-spectrum antibiotics), underlying diseases (chemotherapy-induced mucosal erosion, neutropenia) and invasive procedures (abdominal surgery, central venous catheter) favours translocation of C. albicans across the mucosa and dissemination into the bloodstream4. Recent epidemiological surveys have reported that invasive candidiasis accounts for 75% of all invasive fungal infections5,6,7 and Candida species are now recurrently found among the five leading causes of healthcare-associated bloodstream infections8,9,10. The invasive candidiasis mortality rate (>40%) in immunocompromised patients is higher than that of any bacterial sepsis11,12,13. Because of the increase of hospitalised patients with serious underlying diseases, especially those undergoing treatment for haematological malignancies or hospitalised in ICUs, systemic candidiasis has become a life-threatening challenge for human health.

C. albicans is well known for its ability to form biofilms on medical devices, a process that contributes to relapse or persistence of candidiasis despite antifungal treatments14. Indeed, C. albicans biofilms are crucial for pathogenesis by reducing the sensitivity of the fungus to both antifungal agents and the host immune system11. The molecular processes allowing C. albicans biofilms to avoid antifungal toxicity are mechanistically complex. The resistance towards antifungal drugs is largely due to their trapping by the biofilm extracellular matrix (ECM) and, to some extent, the upregulation of efflux pumps15. Upregulation of both ABC and MFS transporters contributes to the cellular export of antifungal components and thus avoids their toxicity. Biofilms are embedded in the ECM, which protects them from environmental insults, creating a steric hindrance that prevents drug penetration inside the biofilm16. In addition, ergosterol production is decreased in mature biofilm as compared to planktonic cells rendering the microbial community less susceptible to azoles and polyenes than planktonic cells17. The acquisition of new knowledge on the establishment of fungal biofilm will allow the development of innovative strategies to counter the threat posed by these microbial communities.

Biofilm formation is a process tightly regulated in C. albicans that occurs in response to several environmental clues such as surface contact, CO2 levels or nutrient starvation18,19,20. This biological process is orchestrated by a « core » of nine transcription factors that finely regulate lipid metabolism, filamentation and above all adhesion21,22. Adhesion of C. albicans to a substrate is mediated by several polypeptides that belong to three families of adhesins: Als, Hwp1 and Iff/Hyr23. Some proteins among these families, namely Als1, Als2, Als3, Als5, Hwp1, Hwp2 and Eap1, were described to impact biofilm formation23. Except for Als2, all adhesins of C. albicans involved in biofilm establishment contain β-aggregation-prone sequences suggesting that they may be assembled as amyloid structures24,25,26. Amyloid fibres correspond to filamentous protein aggregates with an organised structure. The assembly of an amyloid fibre usually consists of two steps: first, amyloids are nucleated by the homomultimerisation of a protein through the amyloid-forming region. Subsequently, the amyloid fibre is elongated by the stacking of hundreds of monomers on top of each other. Structurally, amyloid fibres are characterised by a high β-sheet content as well as by the presence of the typical cross-β arrangement, which is a hallmark of such quaternary structures27. The cross-β structure is characterised by two reflections detected at 4.7 Å and ~10 Å by X-ray diffraction, which respectively corresponds to the inter-strand and inter-sheet spacing27.

Amyloid structures have long been thought to be harmful to living organisms. However, recent reports have demonstrated that some proteins in the amyloid form can acquire new properties that could be useful and beneficial in the cellular context, a phenomenon called functional amyloids28. For instance, the curli system in E. coli consists of amyloid fibres using the proteins CsgA and CsgB as substrates. These fibres extended in the extracellular matrix of the bacteria are involved in biofilm formation and surface colonisation29. Another example is the S. cerevisiae pyruvate kinase, Cdc19, which forms amyloid fibres aggregating into stress granules under different types of stress. This phenomenon protects the protein from degradation and facilitates the resumption of the cell cycle once the stress is over30.

Results based on electron microscopy, as well as heterologous expression of C. albicans proteins in S. cerevisiae, have revealed that Als1, Als3, Als5 and Eap1 genuinely adopt an amyloid-like structure22,23. Over the last decade, knowledge gathered mainly on Als1 and Als5 from C. albicans has tended towards a model in which these adhesins are unfolded by shear forces generated by liquid flow25, unmasking the amyloid-forming regions and allowing lateral and homotypic amyloid interactions to occur in the C. albicans cell wall, thus creating a very dense area of adhesins, called amyloid nanodomain31,32. Finally, these nanometre-sized nanodomains create adhesive patches on the cell surface, enabling cell–cell interactions through adhesins Ig-like domain32. Pga59 is another adhesin found in C. albicans that belongs to the Hwp1 family. Originally, Pga59 was identified by a computational screen that aimed to find all GPI-anchored proteins in several fungi species, including C. albicans33. The proteins found in the in silico analysis, with no functional name, were given the acronym Pga for predicted GPI-Anchored protein associated with a number i.e. Pga5933. Pga59 is a cell wall-associated protein harbouring both N- and O-glycosylation that was originally described to maintain cell wall integrity in C. albicans34. Previous results from our laboratory have shown that PGA59 is highly expressed in C. albicans biofilms34, and that overexpression of PGA59 has a positive impact on biofilm formation by increasing the dry weight of the microbial community35. Furthermore, atomic force microscopy experiments have revealed that PGA59 overexpression is also associated with higher adhesion forces on the cell surface, resulting in increased adhesion to a plastic substrate upon PGA59 upregulation35. However, the molecular mechanisms underlying Pga59 function in biofilm formation remain to be understood.

In this study, we have shown that Pga59 is assembled into amyloid structures upon adhesion, and we have demonstrated that Pga59 amyloid formation has a positive impact on cell adhesiveness as well as on biofilm establishment in C. albicans. In addition to the work on Als proteins, our study provides further strong evidence that amyloid structures play a role in fungal cell adhesion to a substrate. Furthermore, we describe for the first time that Pga59-based amyloid structures have a direct impact on biofilm formation in C. albicans, and thus constitute a promising avenue for the development of new antifungal strategies.

## Results

### The D1 domain of Pga59 triggers β-amyloid assembly in vitro

Using the in silico AMYLPRED2 tool36,37, we identified three potential amyloid-forming domains in the Pga59 mature sequence, D1 from position 28 to 35 (28IATTVVTI35), D2 from 48 to 56 (48VTTGVTTVT56) and D3 from 61 to 66 (61TYTTYC66) (Fig. 1a). Further analysis with TANGO, one of the programmes run by AMYLPRED2, showed that D1 was the only domain presenting a high β-aggregation potential (Fig. 1a). Within D1, the highest β-aggregation potential (~74%) was predicted for two valines at positions 32 and 33 (Fig. 1a). A synthetic eight amino acid-long peptide overlapping this sequence (residues 28 to 35) was obtained, as well as peptides incorporating mutations of either V32 or V33 or both into asparagine residues (V32N, V33N and V32,33N, respectively). To confirm the amyloidogenic properties of the D1 sequence, transmission electron microscopy (TEM) observation of the negatively stained synthetic peptides was performed after incubation at 37 °C for 16 h. The WT peptide of the D1 domain formed a large micrometre-sized aggregate with organised regions exhibiting a fibrillar morphology (Fig. 1b). The single mutant peptides V32N and V33N were still able to adopt a fibrillar morphology reminiscent of an amyloid fibre. However, the peptide in which both valines had been substituted for asparagines was unable to adopt an elongated fibrillar morphology (Fig. 1b). To determine if the fibrillar aggregates observed in TEM corresponded to amyloid structures, we performed in vitro Thioflavin T (ThT) assays on the same set of synthetic peptides. The WT peptide displayed a very high fluorescence intensity (~30k arbitrary units (AU)) generated by amyloid structures stained with ThT (Fig. 1c). The V32N mutation had a limited impact, reducing the fluorescence intensity by twofold to ~15k AU while the V33N mutation enhanced it by 1.4-fold to ~44k AU (Fig. 1c). However, the fluorescence intensity of the V32,33N peptide upon ThT staining was significantly reduced to a value of 987 AU, i.e. more than 30 times less than to the WT (Fig. 1c). As the WT D1 peptide of Pga59 self-assembles in fibrillar quaternary structures that are positive to ThT, we sought to determine if these aggregates are genuine amyloids exhibiting a typical cross-β signature. Hence, we used X-ray fibre diffraction to investigate the amyloid nature of the D1 WT peptide. The X-ray diffraction pattern showed two main reflections at 4.7 and 10 Å (Fig. 1d). Such a pattern is characteristic of the so-called cross-β arrangement, one of the structural hallmarks of amyloid fibrils27. This clearly indicates that the Pga59 D1 domain is assembled into amyloid fibres made of β-strands in its aggregated state. The cross-β pattern reveals a repetition of β-strands that are parallel to the main fibril axis and separated by 4.7 Å. The β-sheets are perpendicular to the fibril axis with a spacing of 10 Å. Taken together, these results revealed that the D1 domain of Pga59 is sufficient to craft a cross-β amyloid fibre and that the presence of valine at positions 32 and 33 is critical for this process.

### Recombinant full-length Pga59 is assembled as an amyloid structure

We next examined if the ability of the D1 domain to form an amyloid structure is conserved with the full-length Pga59 protein. To do so, we produced Pga59 and Pga59 V32,33N (in which both valine residues were replaced by asparagines) fused to an N-terminal His-tag (His6-Pga59 and His6-Pga59V32,33N, respectively) in Escherichia coli SHuffle T7 and purified the recombinant Pga59 proteins (rPga59) by affinity chromatography. We chose to work with the double mutant since the amyloid assembly was blocked only in the D1V32,33N peptide (Fig. 1b, c). Both versions of the rPga59 were purified from bacterial inclusion bodies (Fig. 2a) and their purity was assessed by MALDI-TOF mass spectrometry. For His6-Pga59, the mass analysis revealed the presence of a single intense peak at 9, 839 Da, which corresponds to the expected mass of the WT recombinant protein. Regarding His6-Pga59V32,33N, a single intense peak was observed at the expected mass of 9, 869 Da (Supplementary Fig. 1). The mass analysis demonstrates that both WT and mutant rPga59 have been purified to a high degree. This result allows us to propose that the gel migration of His6-Pga59V32,33N, noticeably slower than that of His6-Pga59, is most likely due to charge variations affecting the migration of the mutant protein compared to the WT. Subsequently, a single signal was detected in the elution fraction by an α-His antibody, confirming that polypeptides eluted from the affinity chromatography matrix were genuinely either His6-Pga59 or His6-Pga59V32,33N (Fig. 2b). To evaluate the ability of rPga59 to adopt an amyloid shape, we conducted ThT assays and TEM. Following refolding, both the WT and the mutant version of rPga59 were incubated at 37 °C and subsequently either treated with ThT or negatively stained with uranyl acetate and observed by electron microscopy. His6-Pga59 displayed a strong ThT signal (~25k AU, Fig. 2c) and adopted elongated fibrillary structures (purple arrow, Fig. 2d). On the other hand, His6-Pga59V32,33N produced a 2.5-fold weaker ThT signal (~10k AU) and no fibrillary aggregates were observed (Fig. 2c, d). These results suggest that the amyloid assembly capabilities of His6-Pga59V32,33N are greatly reduced compared to His6-Pga59. X-ray diffraction was also used to confirm the amyloid nature of His6-Pga59, and typical signatures of β-amyloid fibre were observed for His6-Pga59, with an X-ray diffraction pattern showing reflections at 4.7 and 10 Å (Supplementary Fig. 2). In conclusion, ThT staining, TEM and X-ray diffraction measurements demonstrate that E. coli-produced Pga59 is assembled as a typical cross-β amyloid structure in vitro.

### Heterologous expression of Pga59 triggers amyloid structure assembly in the budding yeast cell wall

We next sought to determine whether amyloid fibre assembly of Pga59 is a biological process relevant in yeast. We first expressed HA3-Pga59 and HA3-Pga59V32,33N in S. cerevisiae to avoid any background due to amyloid production by other proteins in C. albicans such as Als1 or Als5, among others23. Using immunofluorescence staining, we showed that both the WT HA3-Pga59 and the HA3-Pga59V32,33N mutant proteins were properly localised at the cell surface of S. cerevisiae, while no signal was observed in yeast cells expressing the empty vector (Fig. 3a). We also confirmed that both proteins were produced at similar levels (Fig. 3b). The molecular weight of Pga59 is higher when expressed in S. cerevisiae as protein glycosylation occurs in yeast and not in E. coli. We triggered cell adhesion by incubating S. cerevisiae cells expressing either HA3-Pga59 or HA3-Pga59V32,33N with BSA-coated magnetic beads, and assessed the formation of amyloid structures using ThT staining in vivo. S. cerevisiae cells expressing HA3-Pga59 displayed a ThT-fluorescent signal at the cell surface, as indicated by colocalization with a cell wall marker, namely Concanavalin A, Alexa FluorTM 594 conjugate (Fig. 3c). In contrast, S. cerevisiae cells expressing either the empty vector or HA3-Pga59V32,33N did not display a cell surface-localised ThT signal. Here, the ThT fluorescence was restricted to weak cytoplasmic foci that did not colocalize with the Concanavalin A signal (Fig. 3c). Altogether, these data suggest that Pga59 is involved in the assembly of amyloid structures in the fungal cell wall. Interestingly, the replacement of valines by asparagines at positions 32 and 33 in Pga59, shown above to impair the formation of amyloid structures in vitro, also prevents Pga59-induced assembly of parietal amyloid structures in S. cerevisiae.

### The amyloid property of Pga59 positively impacts adhesion and biofilm formation in C. albicans

The involvement of Pga59 in biofilm formation has already been shown. Indeed, in a continuous flow microfermentor model, overexpression of Pga59 was shown to lead to thicker biofilms than the wild type35. In the same experimental setting, the pga59 knock-out strain forms a biofilm twice lighter than that formed by cells expressing Pga59 (Supplementary Fig. 4), thus confirming the implication of Pga59 in this process. To investigate the impact of Pga59 amyloid properties on biofilm establishment, we assessed the biofilm formation capacity in a continuous flow fermentor system of strains overexpressing either PGA59 or pga59V32,33N. Inducing the overexpression of PGA59 led to a significantly higher dry biofilm biomass (170 ± 27 mg) as compared to the same strain in uninduced conditions (100 ± 9 mg) (Fig. 6c). In contrast, when pga59V32,33N was overexpressed, the dry biofilm biomass dropped to 67 ± 6 mg compared to the identical strain without overexpression of pga59V32,33N (105 ± 15 mg) (Fig. 6c). The overexpression of PGA59 is associated to an increase of 70% in fungal biofilm production while the overexpression of pga59V32,33N is linked to a decrease of 37% of the biofilm biomass both compared to strains without overexpression of PGA59 and pga59V32,33N respectively.

Overall, these results show that amyloid properties of Pga59 are important for adhesion to both abiotic surfaces and human cells, as well as biofilm establishment in C. albicans.

## Discussion

In the present work, we have investigated the molecular mechanisms underlying Pga59 function regarding biofilm formation in C. albicans. Pga59 belongs to the Hwp1 family of adhesins comprised of 12 proteins characterised by a conserved 42 amino acid-long region, the molecular function of which has not been uncovered yet23,34. Results from several members of this family, namely Hwp1, Hwp2, Eap1 and Rbt1 have revealed the presence of amyloid-forming regions within the 42 amino acid domains, suggesting a role of this specific domain in amyloid formation20,40.

Here, an in silico analysis revealed that the Pga59 42 amino acid domain contains three potential amyloid-forming regions, only one of which exhibits β-aggregation potential. Using an in vitro approach, we demonstrated that Pga59 can be assembled into β-amyloid fibres. Furthermore, we showed that this phenomenon is relevant in cellulo where Pga59-dependent amyloid structures are crafted upon adhesion. Importantly, we also demonstrated that amyloid structures assembled in a Pga59-dependent manner is important for fungal cell adhesion and effective biofilm formation.

Several adhesins have already been described in the literature as having an impact on biofilm formation in C. albicans23. Therefore, the relevance of Pga59 in relation to other adhesins in biofilm formation must be discussed. Biofilm establishment in C. albicans is a biological process that includes four steps: (i) the adhesion of yeast cells to a surface, (ii) the initiation step where yeast cells start to filament, (iii) the maturation phase to constitute an intricate network of yeast cells encased within an ECM and finally (iv) the dispersion to release yeast cells from the biofilm and colonise new area46. Als1 is under the control of the Efg1 transcription factor and hence its expression occurs during filamentation at the maturation step47. Moreover, immunofluorescence microscopy clearly shows that Als1 is mainly located at the hyphae cell wall47. Hwp1, Hwp2 and Als3 are hyphae-specific proteins expressed once the filamentation is engaged during the maturation step20,48,49. Thus, in a fungal biofilm context, these adhesins were suggested to maintain microbial community cohesion by making cell–cell connections. However, in the present report, we showed that adhesion properties provided by Pga59 through cell surface amyloid formation start to occur prior to filamentation (Fig. 4). It is thus suggested that amyloid structures crafted by using Pga59 contribute to the initial adhesion of fungal cells to the substrate which chronologically would happen before the intervention of Als1, Hwp1, Hwp2 and Als3 adhesins. Further studies will be required to validate this statement.

To date, Als1 and Als5 are the only proteins described in C. albicans to form amyloid nanodomains and trigger adhesion in response to shear forces generated by a liquid flow31. Yet, the fact that Pga59 colocalized with amyloid structures upon adhesion leaves open the possibility for the existence of amyloid nanodomains constituted of Pga59 adhesins (Fig. 4). The β-aggregation-prone amyloid core sequence IVIVATT from Als1 is the minimal sequence to trigger amyloid links between homologous adhesins45. Here we showed that the IATTVVTI octapeptide from Pga59 is necessary and sufficient to form fibrillar structures24,25. Although the amyloid core sequences of Als1 and Pga59 are different, they are mainly comprised of isoleucines, threonines and valines, which are the residues most often encountered in amyloid-forming cores24,25. Therefore, the similar biochemical properties of both sequences suggest that Pga59 could behave as Als1 and generate amyloid nanodomains upon adhesion.

Amyloid structures extracted from adherent C. albicans are crafted in a Pga59-dependent manner and contain Pga59. In vitro, amyloid fibres are made up of several hundred monomers of the same protein50. However, in bacteria, some functional amyloids incorporate several proteins. For instance, the curli fibre of E. coli is mainly composed of CsgA but also includes the minor curli subunit CsgB51,52. Furthermore, a fibrous structure produced within the ECM of Streptococcus mutans biofilms is positive to immunogold labelling using antibodies targeting WapA and SMU_63c, suggesting that both proteins are assembled in an amyloid aggregate by the bacteria53. To date, the protein composition of Pga59-dependent amyloid structures found in the C. albicans cell wall upon adhesion is still unclear. The behaviour of the pga59 knock-out strain, as well as the strain producing Pga59V32,33N, let us suggest that Pga59 is not the only protein found in cell surface amyloids of adherent cells. Indeed, although the loss of either Pga59 itself or its ability to form amyloid structures greatly reduced the ThT staining on adherent cells, a weak ThT signal persisted (Fig. 4). Amyloid nanodomains are thought to be exclusively constituted by adhesins homotypic interactions. However, most of the work performed on amyloid nanodomains relied on heterologous expression of C. albicans Als adhesins in S. cerevisiae. Thus, the composition of amyloid nanodomains in the C. albicans cell wall is still not fully elucidated and it is thus possible that Pga59 is assembled within amyloid nanodomains with other adhesins at the cell surface of C. albicans upon adhesion.

The assembly of amyloid-like materials in the ECM is a common mechanism used by bacterial species (i.e., E. coli, P. aeruginosa) to form a biofilm. Here, we have demonstrated that the formation of amyloid materials positively impacts biofilm establishment in C. albicans. Cells expressing Pga59V32,33N, a mutant version of Pga59 unable to form Pga59-dependent amyloid structures in vivo, produce significantly thinner biofilms compared to a WT strain. Using atomic force microscopy, we had previously shown that overexpression of Pga59 is associated with higher adhesion forces35. Altogether, these results let us speculate on a model where Pga59 assembled as an amyloid structure creates adhesion forces which are beneficial for biofilm establishment.

Amyloid structures are directly involved in the pathophysiology of several human diseases54. Anti-amyloid compounds inhibiting the assembly of amyloid fibres are thought to be effective molecules to fight diseases where amyloid structures contribute to the pathology55. Regarding fungal biofilms, an amyloid inhibitory peptide targeting the amyloid-forming region of Als5 was previously shown to greatly reduce C. albicans biofilm formation in vitro38. Cells expressing pga59V32,33N showed a tremendous reduction in their abilities to both produce amyloid structures and to adhere on surfaces compared to strains expressing the wild-type version of PGA59 (Figs. 46). Adherent cells treated with the V32,33N mutant peptide showed reduced amyloid formation as well as smaller cellular aggregates compared to C. albicans cells treated with the Pga59 WT peptide (Fig. 7a). Additionally, biofilm biomass was lowered when cells were incubated with the Pga59 mutant peptide compared to untreated cells or cells cultivated with the Pga59 WT peptide (Fig. 7b). Adding the mutant peptide weakens adhesion forces between cells most probably by reducing amyloid assembly upon adhesion, and this negatively impacts biofilm formation. Thus Pga59-based peptides harbouring the V32,33N mutation could constitute a promising avenue to develop new antifungal strategies by lowering amyloid assembly and thus impairing the adhesion process and biofilm formation. Further, as Als5-derived peptides have been shown to inhibit cell–cell adhesion38, blocking amyloid assembly with a combination of peptides targeting several amyloid proteins important for adhesion could lead to a stronger strategy to block biofilm establishment of the pathogenic yeast C. albicans.

## Methods

### Yeast strains and culture conditions

C. albicans and S. cerevisiae strains used in this study are listed in Table 1. Strains were routinely cultivated in YPD medium (1% yeast extract, 2% peptone and 2% glucose) at 30 °C or SD minimal medium (0.67% yeast nitrogen base without amino acids, 2% glucose) with supplements (arginine 20 μg mL−1, uridine 40 μg mL−1, histidine 20 μg mL−1). C. albicans strains were transformed using the lithium acetate-PEG protocol56. Briefly, an overnight culture was diluted to OD600 = 0.2 in 50 mL YPD and grown at 30 °C until OD600 = 0.6–0.8. Cells were rinsed first in ice-cold 10X TE (100 mM Tris pH 7.5/10 mM EDTA), then in ice-cold 1X TE/100 mM lithium acetate, resuspended in 200 μL 1x TE/100 mM lithium acetate and incubated for 1 h on ice; 50 μL of competent cells were mixed with 1 μg DNA, 50 μg of salmon-sperm DNA and 300 μL of a 40% PEG 3000/1X TE/100 mM lithium acetate solution. The transformation mix was then incubated O/N at 30 °C; cells were then heat-shocked for 15 min at 44 °C and washed with 500 μL of SD medium; cells were pelleted by centrifugation and resuspended either in 300 μL of SD medium and plated on selection plates for prototrophic selection, or in 1 mL YPD, the cells incubated at 30 °C for 3–4 h before plating on YPD medium supplemented with 200 μg mL−1 nourseothricin. S. cerevisiae strains were routinely transformed with the lithium method57. In short, cells were grown as described above for C. albicans, but rinsed in water and resuspended in 1 mL of water. One hundred μL of the cells were used per transformation as follows: cells were pelleted by centrifugation and resuspended in 360 μL transformation mix (30% PEG 3000; 100 mM LiAc, 50 µg salmon-sperm DNA; up to 1 μg DNA). Cells were heat-shocked at 42 °C for 40 min, pelleted, resuspended in 300 μL of water and plated on SC-Ura medium (0.67% yeast nitrogen base without amino acids, Dropout mix without uracil and 2% glucose). For biofilm formation in a continuous flow fermentor system, C. albicans cells were cultivated in GHAUM medium (SD minimal medium supplemented with arginine 20 μg mL−1, uridine 40 μg mL−1, histidine 20 μg mL−1 and methionine 40 μg mL−1) with or without 50 μg mL−1 of doxycycline.

### DNA constructs

Primers used to generate the DNA constructs are listed in Table 2. The mutant ORF PGA59V32,32N was created by fusion PCR using CIpSAT1::PGA59 as a template34. Primers A and B and primers C and D were used to amplify two overlapping fragments with the Q5 High-Fidelity DNA polymerase (NEB) (98 °C-30s; (98 °C-10 s; 59 °C-45 s; 72 °C-60 s) x36; 72 °C-2 min), and the fusion PCR (98 °C-30 s; (98 °C-10 s; 59 °C-45 s; 72 °C-90 s) x36; 72 °C-2 min) was performed by using 300 ng of each fragment as templates and primers A and D. Similarly, a PCR fusion was performed to engineer the GFP-tagged PGA59V32,32N mutant using pGEM-T::(PGA59-GFP, ARG4) as a template34; two overlapping fragments were obtained with primers E and B and primers F and C, respectively, using the Q5 High-Fidelity DNA polymerase as above. The fusion PCR was performed as before, using primers E and H. Then, PGA59V32,32N and GFP-PGA59V32,32N were cloned in the KpnI and SalI sites of CIpSAT2 to generate CIpSAT2::PGA59V32,33N and CIpSAT2::GFP-PGA59V32,33N, respectively.

For expression of HA3-Pga59 and HA3-Pga59V32,33N proteins in the cell wall of S. cerevisiae, PGA59 (primers G and H) and PGA59V32,32N (primers G and H) were PCR-amplified from CIpSAT1::PGA59 and CIpSAT1::PGA59V32,33N, respectively, with the Q5 High-Fidelity DNA polymerase (98 °C-30 s; (98 °C-10 s; 59 °C-45 s; 72 °C-45 s) x36; 72 °C-2 min). We then used the Gateway recombinase-based cloning system (Thermo Fisher Scientific) to introduce the fragments in pBC54258, according to the manufacturer’s instructions. The resulting vectors have been named pBC542(URA3 + PGA59) and pBC542(URA3 + PGA59V32,33N).

### Prediction of amyloidogenic domains in Pga59 and other related proteins

The amyloidogenic domains in Pga59 full-length primary sequence were predicted using the AMYLPRED2 (http://aias.biol.uoa.gr/AMYLPRED2/) and TANGO (http://tango.crg.es) tools36,37. The primary sequence of Pga59 was obtained from Uniprot (https://www.uniprot.org/) with the following accession number: Q5AF39.

### In vitro assembly of fungal amyloid fibres

In order to investigate amyloid fibres formation in vitro, synthetic octapeptides corresponding to the wild type (NH2-IATTVVTI-COOH) and mutant versions V32N (NH2-IATTNVTI-COOH), V33N (NH2-IATTVNTI-COOH) and V32,33N (NH2-IATTNNTI-COOH) of the Pga59 D1 domain were purchased from Thermo Fisher Scientific. Peptides were resuspended at 100 μM in DMSO. Next, each peptide was diluted at 5 μM in a buffer composed of 20 mM Tris-HCl pH 7.4, 150 mM NaCl, 40 μM thioflavin T and incubated at 37 °C for 16 h.

For rHIS6-Pga59 and rHIS6-Pga59V32,33N purified from Escherichia coli, 5 μM of recombinant protein was incubated at 37 °C for 16 h in the same solution used to trigger the aggregation of D1 peptides.

### Expression and purification of rPga59

PGA59 codon-optimised synthetic gene lacking the SPER (signal peptide) and SPGPI (propeptide used to post-translationally modify a protein with a GPI anchor) coding regions were produced by GeneArt Gene Synthesis (Thermo Fisher Scientific). NdeI and XhoI restriction sites added at the N- and C-terminal ends, respectively, were used to excise the gene from the initial pMAT vector; the fragment was inserted into the pET28a(+) expression vector previously linearised with the same sites and the ligation was transformed into One Shot TOP10 Chemically Competent E. coli cells (Thermo Fisher Scientific) according to the manufacturer’s instructions. The pET28a(+)-PGA59 construct was verified by sequencing. The pET28a(+)-PGA59V32,33N vector was produced by GeneArt Gene Synthesis (Thermo Fisher Scientific). Both plasmids were transformed in E. coli SHuffle® T7 Express lysY Competent cells according to the manufacturer’s instructions, and selected on LB containing 30 mg L−1 kanamycin. For HIS6-Pga59 and HIS6-Pga59V32,33N production, 50 mL of overnight bacterial cultures were inoculated into 2.5 L baffled flasks containing 1 L of LB supplemented with 30 mg L−1 of kanamycin. Cultures were grown by shaking at 220 rpm at 30 °C until OD600 reached 0.6. PGA59 expression was induced with 1 mM IPTG at 37 °C for 4 h. Bacterial cells were sedimented by centrifugation at 3315×g for 20 min, weighted and stored at −80 °C. Cell pellets were thawed on ice and subsequently resuspended in a 50 mM Tris-HCl pH 8400 mM NaCl, 0.1% Triton X-100, 1 mM DTT lysis buffer supplemented with cOmpleteTM Mini EDTA-free protease inhibitor (Roche) at a ratio of 3 mL/g of pellet. One mg mL−1 of lysozyme was added to the mixture to allow cell lysis at 4 °C for 45 min. Then, 50 μg mL−1 of DNase I was added and lysates were incubated for 45 min at 4 °C, followed by centrifugation at 27,000×g for 15 min. The resulting pellets were washed two times for 1 h at 4 °C with 25 mL of lysis buffer containing 1% of Triton X-100 and dissolved overnight at 4 °C with 25 mL of lysis buffer containing 6 M of guanidine hydrochloride. HIS6-Pga59 and HIS6-Pga59V32,33N polypeptides were then purified on a 1 mL Protino® Ni-NTA column (Macherey-Nagel) using the $${{{\ddot{\mathrm A}}}}$$KTA Express chromatography system at 4 °C, and a linear gradient of imidazole to a final concentration of 0.5 M. The presence of both peptides was assessed at different steps of the purification by loading 10 μL of rPga59 onto a 4–20% gradient Criterion™ gel (Bio–Rad), using 10 μL of Multicolour Low Range protein ladder (Euromedex) as a molecular weight marker, and either staining 30 min with InstantBlue® Coomassie protein (Abcam) followed by a water rinse, or transferring onto a PVDF membrane with an iBlot2 (Thermo Fisher Scientific) using programme 0 (step 1: 20 V for 1 min, step 2: 23 V for 4 min, step 3: 25 V for 2 min); the membrane was blocked with 5% skimmed milk in 1x PBS for 1 h at room temperature. After rinsing in 1X PBS, the membrane was incubated for 1 h with the A7058 anti-His-tag antibody (Sigma-Aldrich) diluted 1/1000 in 1x PBS. The membrane was then washed, treated with Clarity® Western ECL Substrate (Bio-Rad) and observed using the ChemiDoc MP imaging system (Bio-Rad). Uncropped and unprocessed scans of the gel and the blot are given in the Source Data file. The fractions that contained rPga59 peptides were dialysed against a 20 mM Tris-HCl pH8, 150 mM NaCl, 4 M guanidine hydrochloride solution supplemented with a glutathione redox couple (10 mM reduced, 1 mM oxidised) using a 3 mL Slide-A-Lyzer G2 Dialysis Cassette, 3.5 K MWCO (Thermo Fischer Scientific). Dialysis was performed at 4 °C overnight under gentle agitation. The dialysis process was repeated 3 times by decreasing the quantity of guanidine hydrochloride from 4 to 0 M. The protein concentrations in the refolded fractions were estimated at 280 nm with a spectrophotometer.

### Thioflavin T assays

Amyloid fibres made either of Pga59 peptides or rHIS6-Pga59, or directly extracted from C. albicans, were stained with Thioflavin T (ThT), a dye specific for amyloid structures, as follows. Amyloid fibres were diluted (5 μM for peptides and rHIS6-Pga59; 100 μg of amyloid mixtures enriched from C. albicans) in 300 μL Tris buffer (20 mM Tris-HCl pH8 and 150 mM NaCl) containing 40 μM of ThT (stock solution: 3.3 mM in ethanol). Subsequently, 100 μL of each mixture was deposited in a black flat-bottom 96-wells plate (Greiner) and incubated for 16 h at 37 °C. Then the fluorescence intensity was monitored with a TECAN Infinite 200 Pro spectrophotometer using an excitation wavelength of 440 nm and emission wavelength of 496 nm at 37 °C to evaluate amyloid fibril formation.

For in vivo ThT staining, C. albicans strains were grown overnight in a YPD medium at 30 °C. Then, cultures were diluted to an OD600 of 0.3 in 4 mL of fresh YPD medium and allowed to grow at 30 °C for an additional 4 h. At the mid-logarithmic phase, C. albicans strains were washed twice with sterile PBS and resuspended in 1 mL of SD medium. Subsequently, 108 yeast cells were incubated with 106 magnetic beads treated overnight with 1 mg/mL−1 heat-denatured BSA on a benchtop rotator for 1 h to trigger adhesion or left untreated (negative control)24,59. Then, all fungal cultures were treated with 25 μM of ThT and incubated at room temperature on a benchtop rotator for 1 h. For testing the effect of WT and mutant peptides on cell adhesion, 5 μM of either NH2-IATTVVTI-COOH or NH2-IATTVVTI-COOH were added during the incubation with ThT.

S. cerevisiae BY4742 strains expressing either the WT allele of Pga59 (CEY679), the Pga59V32,33N mutant (CEY680) or the empty vector (VIF205) were grown in SC-Ura medium at 30 °C. Overnight cultures were diluted in 4 mL of fresh SC-Ura medium to an OD600 of 0.5 and incubated with agitation at 30 °C for 4 h. S. cerevisiae cultures were centrifuged at 200 × g for 3 min and resuspended in 1 mL of SC-Ura medium. Then, cells were simultaneously treated with BSA-coated magnetic beads, 25 μM ThT and 50 μg mL−1 Concanavalin A, Alexa Fluor 594 Conjugate. After 1 h of incubation on a benchtop rotator at room temperature, cells were washed two times with SC-Ura medium and observed by epifluorescence microscopy.

Fungal cells (C. albicans and S. cerevisiae) were observed with an ApoTome microscope (Carl Zeiss) equipped with both OlympusTM 40X X-Apo (0.95 NA) and OlympusTM 100X X-Apo (1.45 NA) objectives. Pictures were recorded with a Nikon DXM1200F digital camera. Thioflavin T signals were recorded with the blue filter, while the Concanavalin A, Alexa Fluor 594 Conjugate fluorescence was recorded with the red filter.

The size of C. albicans cellular aggregates in each genetic background upon adhesion was evaluated in Nomarski at a magnification of 20X with the Opera Phenix microscope (Perkin Elmer).

### Negative staining

Ten microlitres of rHIS6-Pga59 samples or D1 synthetic peptides were spotted on glow discharged formvar/carbon grids (FCF200-Cu, DELTA microscopies), negatively stained with 2% uranyl acetate, pH 4.2, analyzed with an FEI Tecnai T12 120 kV transmission electron microscope (FEI Company) equipped with a Gatan Ultrascan US4000 CCD detector.

### Immunogold labelling

C. albicans cells expressing either GFP-PGA59 (CEC56), GFP-pga59V32,33N (CEC6218) or the empty vector (CEC832) were grown in YPD medium. Overnight cultures were diluted to an OD600 of 0.3 in 1 L of YPD medium and subsequently incubated under agitation at 30 °C. Yeasts were centrifuged and resuspended in 50 mL of SD medium. Fungal cells were then treated with BSA-coated magnetic beads for 1 h at room temperature to trigger adhesion or left untreated (negative controls). After incubation, cells were centrifuged and stored at −80 °C. Amyloid structures were extracted from C. albicans as described by ref. 39 Briefly, the pellets were resuspended in 4 mL lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 5 mM MgCl2, 0.1% Nonidet P-40, 1 mM PMSF, and a cOmplete protease inhibitor tablet (Roche)); then 600 μL were distributed in ten 1.5 mL screwcap tubes and 200 μL of 0.5 mm glass beads were added. Subsequently, fungal cells were mechanically disrupted six times for 1 min in a Bullet Blender bead beater (Next Advance) at full speed. The whole-cell extract was first incubated for 10 min with 0.1 mg mL−1 of RNAse A at room temperature, and then with 0.5% of Triton X-100 at 4 °C for 10 min. Samples were centrifuged at 2000 × g for 10 min at 4 °C and supernatants were loaded on a 40% sucrose pad. The protein mixtures were then ultracentrifuged at 200,000 × g, 4 °C for 2 h. Pellets were resuspended in the amyloid-prion resuspension buffer (50 mM Tris-HCl pH 7.4, 100 mM NaCl, 2% SDS, 5 mM DTT and 5% glycerol) and loaded onto an acrylamide gel. Following the migration, amyloids were eluted from the gel with Amyloid-prion buffer R (Tris-HCl pH 7.4, DTT 5 mM and SDS 0.4%). Ten microlitres of enriched amyloid structures were spotted on glow discharged formvar/carbon grids and processed throughout an immunogold labelling procedure as described in Mourer et al.60. Briefly, electron microscopy grids were incubated for 5 min in a droplet of blocking solution (1X PBS, 1% BSA), and then in a solution of 1X PBS, 1% BSA containing the α-GFP antibody (Santa Cruz, Clone B-2 diluted 200 times) for 1 h. After five washes in 1X PBS, 0.1% BSA and five washes in 1X PBS, grids were incubated for 20 min in a solution composed of 1X PBS, 1% BSA and a 1/50 dilution of 10 nm gold particles conjugated to protein A. Grids were washed several times in 1X PBS and incubated in 1X PBS, 1% glutaraldehyde. Finally, grids were washed with distilled water, incubated with 4% uranyl acetate for 30 min, rinsed five times in a droplet of water, and incubated in 2% lead citrate for 2 min. Grids were then rinsed five times in water and dried for at least 24 h at room temperature. Samples were observed with an FEI Tecnai T12 120 kV transmission electron microscope (FEI Company) equipped with a Gatan Ultrascan US4000 CCD detector.

CEC369 (WT), CEC373 (pga59Δ/pga59Δ), CEC832 (pga59Δ/pga59Δ CIpSAT1-PGA59) and CEC6217 (pga59Δ/pga59Δ CIpSAT1-PGA59V32,33N) were used to test the involvement of Pga59 on cellular adhesion as described in ref. 61 as follows. Caco-2 (colon carcinoma) epithelial cells were kindly provided by the Dynamics of Host-Pathogen Interactions Unit, Institut Pasteur. A monolayer of Caco-2 cells was cultivated on coverslips deposited in each well of a TPP® 24-well plate (Merck) in DMEM medium (Gibco) supplemented with 10% FBS for 5 days at 37 °C with 5% CO2. Caco-2 cells monolayers were washed with sterile PBS and then infected with 2 × 105 fungal cells per mL diluted in RPMI medium. After 1 h, non-adherent C. albicans cells were washed off three times with sterile PBS. The remaining adherent cells were cross-linked to Caco-2 cells by overnight incubation at 4 °C with Histofix®. Following the chemical fixation, coverslips were washed with PBS and mounted on a microscope slide with a droplet of ProLongTM Gold antifade (Thermo Fisher Scientific) mounting medium. Fifty random pictures per sample showing an area of 200 μm × 200 μm were recorded in Nomarski on an ApoTome microscope. The number of cells was counted and averaged in each genetic background. The results presented here are representative of three independent experiments. The statistical analysis was performed on GraphPad Prism 9 using a Student’s t-test.

### Immunofluorescence

S. cerevisiae cells transformed with pBC542_HA3-Pga59 (CEY679) pBC542_HA3-Pga59V32,33N (CEY680) or pBC524 (VIF205) were grown in SC-Ura medium. About 80 × 106 cells were fixed with 4.6% formaldehyde for 40 min at room temperature, washed two times with PBS and then resuspended in 1 mL of PBS. Twenty microlitres of yeast cells were dropped microscope slides previously coated with 0.1% on poly-l-Lysine (Sigma-Aldrich) and were allowed to adhere to the substrate for 30 min at room temperature. After several washes with 1X PBS, cells were covered with 20 μL of 1X PBS, 0.5% BSA for 5 min. The samples were successively incubated with 20 μL of: (i) a monoclonal α-HA antibody (working dilution:1/1000, Santa Cruz clone F-7) in 1X PBS + 0.5% BSA overnight at 4 °C and then (ii) a secondary α-mouse antibody conjugated to Texas Red-X (working dilution: 1/1000, Invitrogen T-862) in 1X PBS + 0.5% BSA for 1 h at room temperature. A drop of ProLongTM Gold antifade mounting medium and a cover slip were successively added to the samples. S. cerevisiae strains were observed with an ApoTome microscope. Cells were visualised with Nomarski and the red channel was used to record the fluorescent signal generated by the secondary antibody.

### RNA extraction and RT-qPCR

About 15 mL of CEC369 (WT), CEC373 (pga59Δ/pga59Δ), CEC832 (pga59Δ/pga59Δ CIpSAT1-PGA59) and CEC6217 (pga59Δ/pga59Δ CIpSAT1-PGA59V32,33N) were cultivated as described in the thioflavin T assays section. Total RNAs of adherent cells were extracted with the RNeasy mini kit (Qiagen) as per the manufacturer’s instructions. Five hundred micrograms of total RNA were used for reverse transcription using the QuantiTect Reverse Transcription kit (Qiagen) according to the manufacturer’s specifications. The qPCR reaction mixtures contained 1 μL of the resulting cDNA, 4 μL of primer mix at 10 pmol μL−1 (primers reverse and forward primers of the target genes, see Table 2), 5 μL of water and 10 μL of 2X SYBR green mix (Eurobio). Q-PCRs were performed in a CFX96 touch real-time PCR detection system (Bio-Rad) apparatus with the following cycle: 50 °C-2 min; 95 °C-10 min; (95 °C-15 s; 59 °C-1 min) x40. The relative expression level of PGA59 in each genetic background was determined using specific primers (pga59F and pga59R, Table 2) and compared to the expression level of ACT1 (primers act1F and act1R, Table 2).

### Biofilm formation in a continuous flow fermentor system

Cells overexpressing either PGA59 (CEC6219) or PGA59V32,33N (CEC6220) were grown in a YPD medium with or without 50 μg mL−1 of doxycycline for 16 h. Each culture was then diluted to an OD600 of 1 in fresh GHAUM medium with or without 50 μg mL−1 of doxycycline and left at room temperature for 30 min, to allow further overexpression. For each condition, plastic slides (Thermanox; Nunc) were immersed in the diluted cultures for 90 min at room temperature to allow adherence of fungal cells to the plastic substrate. The plastic slides were then transferred to glass vessels (1 vessel per condition). Each vessel has two glass tubes inserted to drive the entry of medium and air, while the used medium is evacuated through a third tube62. The flow of the GHAUM medium is controlled by a recirculation pump (Ismatec) set at 0.6 mL min−1 and pushed by pressured air supplied at 105 Pa, conditions minimising planktonic phase growth and promoting biofilm formation. The chambers with the plastic substrate were incubated at 37 °C and biofilms were grown for 40 h. Plastic substrates were subsequently immersed in 1X PBS and vortexed to recover the fungal biofilms. The PBS solution containing the biofilms was vacuum filtered through a 1.2 μm filter (Millipore); the filter was dried out at 65 °C for 24 h and then weighed on a precision scale (Mettler AE200; Mettler Toledo) to obtain the dry mass of the biofilm. Three biological replicates were analysed through a Student’s t-test.

### Biofilm formation in TPP® 6 well culture plates

C. albicans overnight cultures were washed in sterile PBS prior to their dilution at 1 × 106 cells per mL in RPMI medium; then, 6 mL were distributed in TPP® six-well plates and allowed to adhere for 90 min without agitation at 37 °C. Subsequently, non-adherent cells were washed once with sterile PBS and 6 mL of fresh RPMI medium, supplemented with either 5 μM of the WT Pga59 peptide or 5 μM of V32,33N mutant peptide, were added to each well. Fungal cells cultured without peptide were used as control. The plates were covered with BREATHseal™ and biofilms were allowed to form for 48 h at 37 °C under gentle agitation at 110 rpm. Biofilms were then gently washed with sterile PBS and vacuum filtered through a 1.2 μm filter. Filters were dried out at 65 °C for 24 h and then weighed on a precision scale (Mettler AE200; Mettler Toledo) to obtain the dry biofilm biomass.

### X-ray diffraction measurement

Fibre diffraction patterns of fibrils were measured at 4 °C on a Rigaku FRX rotating anode X-ray generator equipped with a Pilatus 200 K hybrid pixel detector (DECTRIS Ltd., Baden-Dättwil, Switzerland) at the copper wavelength. The concentrated hydrated samples were mounted in a MicroLoopsTM from Mitegen (Ithaca, NY, USA) on a goniometer head under the cold nitrogen flow. Each diffraction pattern corresponds to a 360° rotation along the phi axis with an exposure time of 720 s after subtraction of a ‘blank’ image of the same exposure time with only the loop on the goniometer head.

### MALDI-TOF mass spectrometry

His6-Pga59 and His6-Pga59V32,33N integrity and purity were analysed on a Bruker UltrafeXtreme MALDI-TOF/TOF instrument. A volume of 15 μL of protein was passed through a ZipTip C4 and eluted on an MTP 384 ground steel target plate (Bruker-Daltonics, Germany) with 2 μL of 20 mg mL−1 α-Cyano-4-hydroxycinnamic acid in 50% acetonitrile, 0.1% trifluoroacetic acid as matrix solution. Data were acquired using Flexcontrol software (Bruker-Daltonics, Germany) and shots were recorded in positive ion linear mode. Mass spectra were externally calibrated in the m/z range of 5–20 kDa with Protein I (Bruker-Daltonics, Germany) and analysed with the Flexanalysis software (Bruker).

### Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.