Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Yeast cells possess a remarkable capacity to adhere to abiotic surfaces, cells and tissues. These adhesion properties are of medical and industrial relevance. Pathogenic yeasts such as Candida albicans and Candida glabrata adhere to medical devices and form drug-resistant biofilms. In contrast, cell–cell adhesion (flocculation) is a desirable property of industrial Saccharomyces cerevisiae strains that allows the easy separation of cells from the fermentation product. Adhesion is conferred by a class of special cell wall proteins, called adhesins. Cells carry several different adhesins, each allowing adhesion to specific substrates. Several signalling cascades including the Ras/cAMP/PKA and MAP kinase (MAPK)-dependent filamentous growth pathways tightly control synthesis of the different adhesins. Together, these pathways trigger adhesion in response to stress, nutrient limitation or small molecules produced by the host, such as auxin in plants or NAD in mammals. In addition, adhesins are subject to subtelomeric epigenetic switching, resulting in stochastic expression patterns. Internal tandem repeats within adhesin genes trigger recombination events and the formation of novel adhesins, thereby offering fungi an endless reservoir of adhesion properties. These aspects of fungal adhesion exemplify the impressive phenotypic plasticity of yeasts, allowing them to adapt quickly to stressful environments and exploit new opportunities.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Medical and industrial relevance of fungal adhesion
Apart from stabilizing and shielding cells from mechanical forces, the cell wall also serves as a tool for microbes to interact with their environment – and vice versa. One of the most critical functions of the cell surface is its ability to adhere to other cells and surfaces. Adhesion prevents cells from being washed away when they find themselves in a nourishing environment, and allows them to form biofilms that offer protection from hazardous conditions. Pathogenic yeasts exploit their capacity to adhere to abiotic surfaces such as plastic prostheses to gain access to the bloodstream and internal organs of patients (Kojic and Darouiche, 2004). Prostheses and catheters can also serve as carriers for fungal biofilms and thus provide an internal reservoir of highly drug-resistant infective cells (Kojic and Darouiche, 2004). In contrast to the mostly harmless superficial infections, disseminated fungal infections have a high mortality rate (> 40%) (Wisplinghoff et al., 2004), placing fungal adhesion centre stage in today's medicine. Fungal adhesion is also of considerable economic importance for food-processing companies, because adherent fungi can form highly resistant biofilms in industrial installations. On the positive side, cell–cell adhesion of industrial brewing and wine yeast is often exploited as a convenient and cost-effective way to separate biomass from various fermentation products (Verstrepen et al., 2003). At the end of the fermentation process, when all available sugars have been converted into ethanol and carbon dioxide, the yeast cells start adhering to each other to form macroscopic ‘flocs’ consisting of many thousands of cells. Cell–cell adhesion between yeast cells is therefore often called ‘flocculation’. Depending on the yeast strains used, the yeast flocs rapidly sediment to the bottom (‘lager’ strains) or float to the surface (‘ale’ strains), thereby facilitating their removal from the medium.
One of the most remarkable features of yeast adhesion is the phenotypic variability and plasticity. Yeast cells are able to quickly adapt their adhesion properties to new environments. Moreover, the adhesion behaviour of closely related strains can vary dramatically. In this paper, we first review the mechanisms of yeast adhesion. We then discuss the signalling pathways that regulate adhesion and the various genetic and epigenetic mechanisms that generate variability and plasticity of this crucial phenotype.
Fungal adhesion is conferred by specialized cell-surface proteins
Adhesion is conferred by specialized cell-surface proteins called ‘adhesins’ or ‘flocculins’ that bind specific amino acid or sugar residues on the surface of other cells or promote binding to abiotic surfaces (Fig. 1). All fungal adhesins share a common three-domain structure. The C-terminal part of adhesins contains a glycosylphosphatidylinositol (GPI)-anchor addition site and links the adhesin to the cell wall (Bony et al., 1997; Kapteyn et al., 1999). (Fig. 1). The N-terminal part of adhesins protrudes from the cell surface and often contains a carbohydrate or peptide binding domain (Kobayashi et al., 1998; Groes et al., 2002; Rigden et al., 2004). The large middle domain of adhesins is characterized by the presence of multiple serine- and threonine-rich repeats encoded by conserved DNA sequences that serve as a source of variability by triggering frequent recombination events through which new adhesin alleles are generated (see further). In many cases, calcium ions enable the adhesins to achieve their active confirmation (Stratford, 1992).
Despite their common features, there are many different yeast adhesins. First, each cell carries a family of (slightly) different specialized adhesins that provide the cells with an array of adhesion properties (Guo et al., 2000; Sheppard et al., 2004; Verstrepen et al., 2004a). Second, different yeast species carry different families of adhesins that reflect the species’ lifestyle. The benign brewer's yeast Saccharomyces cerevisiae, for example, carries five FLO (flocculation) genes: FLO1, FLO5, FLO9, FLO10 and FLO11 (Teunissen and Steensma, 1995). FLO1, 5, 9 and 10 confer cell–cell adhesion (flocculation), while FLO11 is responsible for adhesion to substrates (Guo et al., 2000). In contrast, the adhesin families in pathogens like Candida albicans (ALS and EAP genes) and Candida glabrata (EPA genes) confer adhesion to mammalian host tissues (Hoyer, 2001; De Las Penas et al., 2003; Li and Palecek, 2003; Kumamoto and Vinces, 2005). Figure S1 shows a phylogenetic tree of the various adhesin genes as well as some related genes in higher eukaryotes.
Recent studies have identified some of the subtle differences in specificity between the various adhesins. Guo et al. (2000) selectively overexpressed the S. cerevisiae FLO1, FLO10 and FLO11 genes and found that expression of FLO1 results in strong flocculation, expression of FLO10 confers weak flocculation, and FLO11 expression does not produce flocculation at all (Fig. 2A). Instead, expression of FLO11 results in adherence to agar and plastic (Fig. 2B). Similarly, subtle differences in the binding specificity of the C. albicans Als1 and Als5 adhesins to small heptapeptides have been described (Klotz et al., 2004). Furthermore, Als proteins heterologously produced in S. cerevisiae confer distinct adhesion profiles towards human proteins and cells (Sheppard et al., 2004). Another indication of the different specificities among the Candida adhesins is that the various adhesin genes are differentially expressed under different host and model conditions. More specifically, the C. albicans ALS2, 3, 6, 7 and 9 are highly expressed and ALS4 and ALS5 are repressed in models of vaginal candidiasis (Cheng et al., 2005). In contrast, oral specimens show strong expression of ALS1, 2, 3, 4, 5 and 9, while ALS6 and ALS7 are repressed (Green et al., 2004). This differential expression probably reflects the ability of cells to express only the appropriate, most effective adhesins in any given situation.
Different mechanisms of adhesion
While all fungal adhesins make cells bind to other cells or surfaces, their modes of action differ. Adhesion can be divided into two main groups: lectin-like adhesion (sugar sensitive) and sugar-insensitive adhesion. As the name suggests, lectin-like adhesion depends on the lectin-like binding of the adhesin to sugar residues on the surface of other cells. Adhesins of this group have a lectin-like carbohydrate binding domain (pfam no. 06660) in their N-terminus. Addition of certain sugars competitively inhibits adhesion and provides an easy way to determine the sugar specificity of the adhesins. Examples of lectin-like adhesins include the S. cerevisiae FLO gene products (except Flo11) (Stratford, 1992; Guo et al., 2000) as well as the EPA gene products in the human pathogen C. glabrata (Cormack et al., 1999). Pathogenic yeasts bind to glycosides on mammalian cells such as N-acetyl lactosamine (Cormack et al., 1999), while S. cerevisiae cells recognize mannose oligomers on their own surface, resulting in aggregation (Kobayashi et al., 1998). In S. cerevisiae, lectin-like adhesion is further divided into two sub-categories, Flo1 and NewFlo. The Flo1 group only binds mannose sugars, whereas the NewFlo type binds various sugars, including mannose, glucose and glucose oligomers such as maltose (Sato et al., 2002). Most industrial brewing strains are of the NewFlo type; competitive inhibition conferred by sugars in the medium prevents them from flocculating before all fermentable sugars are converted into ethanol, which is exactly what brewers and winemakers want.
Sugar-independent adhesion, on the other hand, is mediated by adhesins that bind peptides instead of sugars, or increase the cell-surface hydrophobicity, thereby promoting hydrophobic interactions between the cells and certain abiotic surfaces (Kang and Choi, 2005). Examples include the Flo11 adhesin in S. cerevisiae, which confers hydrophobicity-based adhesion to abiotic surfaces (Guo et al., 2000) and the Als proteins in the pathogen C. albicans, which recognize certain peptides in host cells (Klotz et al., 2004).
Adhesion is induced by various environmental triggers
As mentioned above, the adhesion-encoding genes are not constitutively expressed. Instead, adhesion is under tight transcriptional control by several interacting regulatory pathways. The adhesion genes are activated by diverse environmental triggers like carbon and/or nitrogen starvation, or changes in pH or ethanol levels (Verstrepen et al., 2003; Sampermans et al., 2005). The switch from non-adherence to adherence probably allows yeasts to adapt to stress. Activation of FLO11 upon nitrogen starvation, or for example, allows fungi to adhere to and penetrate substrates in an attempt to forage for new nutrients (Kron, 1997; Gagiano et al., 2002). Flocculation, on the other hand, might protect the cells in the middle of the flocs from the environment. In addition, the flocs sediment in the medium or float to the surface, and may therefore provide a means of passive transport in the medium, away from the stress.
Apart from being a stress-defence mechanism, adhesion is also crucial for fungal pathogenesis: fungi need to adhere to the appropriate host tissues in order to establish an infection site. Prusty et al. (2004) show that FLO11 is also activated by the elevated concentrations of indole acetic acid (auxin) that occur at plant wound sites, suggesting that FLO11 activation by indole acetic acid might be crucial for feral yeast cells to infect wound sites in plants. Similarly, activation of adhesins in animal pathogens occurs when the cells perceive an opportunity for infection, enabling cells to adhere to the appropriate tissue and establish a colony or biofilm of infectious cells (Verstrepen et al., 2004a; Domergue et al., 2005).
Adhesion is controlled by several signalling pathways
Among the best-studied aspects of fungal adhesions are the various signalling cascades that translate environmental conditions into a proper transcriptional response of the adhesin genes. The pathways regulating adhesion were originally dissected for the S. cerevisiae FLO11 gene, the only FLO gene that is active in the laboratory Σ1278b strain. It is generally expected that the other S. cerevisiae FLO genes as well as adhesins in pathogenic fungi are controlled by similar (but not identical) pathways to those that control FLO11. Indeed, recent studies indicate that at least in Candida spp., many ALS genes are regulated by orthologues of the pathways known to regulate adhesion in S. cerevisiae (Lengeler et al., 2000; Liu, 2001; Kaur et al., 2005; Maidan et al., 2005).
FLO11-mediated cell-surface adhesion is triggered by certain stress factors and/or nutrient limitation. At least three well-known signalling cascades regulate FLO11 expression in response to environmental changes: the Ras-cAMP pathway, the MAP kinase (MAPK)-dependent filamentous growth pathway and the main glucose repression pathway (Madhani and Fink, 1997; Rupp et al., 1999; Gagiano et al., 2002; Vyas et al., 2003; Schwartz and Madhani, 2004). Figure 3 depicts these signal transduction pathways. A fourth pathway, the so-called ‘Target of Rapamycin’ (TOR) pathway, has recently also been implicated in FLO11 regulation (not shown). This pathway is believed to respond to nitrogen starvation, but the exact details are not yet known (Cutler et al., 2001). A fifth (partial) pathway involves the transcription factors Sok2, Phd1 and Ash1, which seem to function in an epistatic pathway. However, the details are not yet understood (Gimeno and Fink, 1994; Ward et al., 1995; Pan and Heitman, 2000) (not shown). In addition, a plethora of other genes have been shown to affect yeast adhesion in large-scale genetic screens. Some of these genes may not be directly related to adhesion, but merely influence adhesion in an indirect way.
While at least some of the signalling pathways regulating adhesion are relatively well characterized, relatively little is known about the upstream sensors of the various signalling cascades. It is clear that these sensors must be triggered by conditions that activate adhesion, such as nutrient limitation, stress or the proximity of a suitable infection site or surface. Recently, the mucin-like transmembrane protein Msb2 has been shown to function as an upstream sensor for the MAPK-dependent filamentous growth pathway (Cullen et al., 2004). Msb2 might function as a sensor for stress, but the exact triggers of Msb2 remain unknown. The ammonium permease Mep2 has also been suggested as an upstream receptor for the MAPK pathway (Gagiano et al., 1999). Mep2 is believed to trigger the pathway in response to nitrogen starvation (Fig. 3A). Other reports indicate that activation of FLO11 by nitrogen depletion relies on a MAPK-independent pathway, in which two of the known regulators of nitrogen metabolism, Gcn2 and Gcn4, play a key role (Braus et al., 2003). This pathway requires Flo8, indicating that the pathway merges or at least interacts with the cAMP/PKA pathway (Fig. 3B). In C. albicans, physical contact with a surface activates the mitogen-activated protein kinase Mkc1, which in turn actives adhesion and biofilm formation (Kumamoto, 2005). Hence, Mkc1 allows cells to determine their location within a host and sense whether they are in physical contact with a suitable carrier to adhere to and form a biofilm. This mechanism bears interesting similarities to the so-called ‘contact inhibition’ that regulates cell proliferation and behaviour in multicellular organisms.
Even when the sensors and signalling cascades for certain environmental factors are known, the picture is not straightforward. For example, adhesion is triggered by carbohydrate depletion (Sampermans et al., 2005). This is explained by the role of the main glucose repression pathway, which represses FLO11 as long as glucose is available in the medium (Fig. 3C). However, the Ras/cAMP/PKA pathway requires glucose or sucrose for activation (Fig. 3B). Hence, in a naïve implementation of the models, glucose would repress adhesion through the main glucose repression pathway, while on the other hand, glucose would be required for induction of the FLO genes because it is needed in order to activate the Ras/cAMP/PKA pathway. It becomes increasingly clear that our models of signalling are somewhat simplistic and that pathways should not be regarded as single, independent entities, but rather as integrated systems working together to control adhesion (Gagiano et al., 2002; Pan and Heitman, 2002; Schwartz and Madhani, 2004). The small G-protein Ras, for example, activates both the Ras/cAMP/PKA and the MAPK pathways. Another common component of various signalling cascades is Mss11. This protein seems to act as a central regulator, integrating the input of all pathways (van Dyk et al., 2005) (Fig. 3). Taken together, a more holistic study of the different pathways is needed to reveal the precise relationship between the environment, the signalling cascades and FLO gene expression.
Fungal adhesion is also controlled epigenetically
Apart from the different signalling cascades that regulate adhesin genes, the genes are often also under epigenetic control (Frieman and Cormack, 2004; Halme et al., 2004). For example, a homogeneous S. cerevisiae population will contain some cells in which FLO11 is transcribed and some in which it is silent. The expression state of FLO11 is metastable, and usually inherited from mother to daughter cells for several generations. However, the expression state is fully reversible, and cells regularly switch between the states. It is unknown if this only happens in a newly formed daughter cell, or if mature cells can also switch between expression states.
Replacing the FLO11 promoter with another promoter or relocating FLO11 with its endogenous promoter to another chromosomal location abolishes the epigenetic regulation. Hence, the epigenetic regulation of FLO11 seems to be both promoter-specific and dependent on the subtelomeric location of FLO11. The promoter-dependent effect relies, at least partially, on Ras/cAMP target proteins, such as the transcriptional repressor Sfl1, and the histone deacetylase Hda1 (see Fig. 3). The latter is also believed to play a role in the position-dependent silencing of FLO11.
FLO11 is not the only adhesin gene that is under epigenetic control. A similar regulation has also been shown for the S. cerevisiae FLO10 and the C. glabrata EPA genes (De Las Penas et al., 2003; Halme et al., 2004). In contrast to FLO11, silencing of FLO10 does not depend on Hda1. Instead, other histone deacetylases (namely Hst1 and 2) and the telomeric silencing regulators Sir3 play a key role. This latter mechanism is similar to the transcriptional silencing of the C. glabrata EPA genes. Silencing of the subtelomeric EPA gene clusters is established by the Yak1 kinase and its effector, the Rap1/Sir3/Sir4 chromatin-remodelling complex (De Las Penas et al., 2003; Iraqui et al., 2005).
In addition to these silencing/de-silencing mechanisms, Halme et al. (2004) reported another way by which silent adhesins can be activated. The IRA1 and IRA2 genes show an unusually high rate of nonsense mutation (10−3). Inactivation of these genes causes overactivation of the Ras/cAMP/PKA pathway, which in turn leads to de-silencing of the FLO10 gene. The exceptionally high rates of IRA inactivation suggest that these mutations represent an additional mechanism that yeast cells employ to generate variability in the expression pattern of the adhesin family.
Epigenetic silencing and de-silencing is not always a stochastic switching mechanism. In a remarkable paper, Domergue et al. (2005) demonstrate that chromatin remodelling is used to induce expression of the C. glabrata EPA6 in the host urinary tract. This de-silencing probably results from a reduction of NAD+ levels in the environment, which in turn affects the activity of the NAD+-dependent histone deacetylase Sir2. Due to the higher levels of nicotinic acid (the NAD precursor), activation of EPA6 does not occur in cells during bloodstream infection. Hence, cells seem to use the low nicotinic acid/NAD+ levels as a cue for establishing infection sites, a process requiring adhesion and thus EPA expression.
Epigenetic silencing of adhesins may serve multiple goals. First, silencing of adhesion in a subset of the population allows a balance between adhering, colonizing cells and non-adhering cells that can disseminate and find new sites for colonization. Epigenetic switching also helps cells to proactively anticipate new conditions in fluctuating environments (Kussell and Leibler, 2005). Furthermore, by only expressing a subset of adhesins, yeast cells can switch between appropriate adhesion phenotypes, allowing them to adhere to specific surfaces only. Lastly, for pathogenic fungi, stochastically variegated expression of different subsets of exposed cell-surface proteins may allow evasion of the host immune system.
Recombination of intragenic repeats generates novel adhesins
Each yeast strain only carries a limited number of slightly different adhesin genes. However, the variability in adhesins isolated from closely related strains is often much larger. Analysis of clinical isolates of C. albicans, for example, identified 60 different alleles of ALS7 (Zhang et al., 2003). Various industrial S. cerevisiae brewer's strains carry sets of FLO genes that significantly vary in length (Verstrepen et al., 2005). This variability is also reflected in evolutionary studies: the adhesin genes were identified as some of the fastest expanding group of paralogues in the genomes of S. cerevisiae and the related Kluyveromyces waltii (Kellis et al., 2004). This genetic variability results in a remarkable phenotypic variation in adhesion phenotypes between different (but closely related) strains and species (Verstrepen et al., 2003).
Recent research shows that the highly repetitive DNA sequences in the middle part of the adhesin genes are the driving force behind the creation of novel adhesins (Verstrepen et al., 2004a; 2005). Figure 4 shows a box plot of the FLO1 gene, revealing the exceptional symmetry of the tandem repeats within the FLO1 sequence. Because of their repetitive nature and their high sequence similarity, the repeats trigger frequent slippage and/or recombination events during DNA replication. This leads to removal or addition of repeat units, and thus contraction or expansion of the respective adhesin gene. Longer adhesins generally confer greater adherence, while smaller adhesins usually result in decreased adhesion, possibly because the N-terminal domain remains buried in the cell wall (Frieman et al., 2002; Verstrepen et al., 2005). Similar to epigenetic silencing, this mechanism may also allow fungi to generate variability within a population and attune their adhesion characteristics. In addition, recombination events between repeats of different adhesion genes could generate chimeric forms, which are differently regulated and might have other adhesion properties. In this way, the limited set of adhesins carried by one cell represents sort of a toolbox with which many other adhesins can be constructed to allow adherence to novel substrates. For pathogens such as C. albicans and C. glabrata, frequent recombination of cell-surface genes has the additional advantage of creating cell-surface variability (Verstrepen et al., 2004a; 2005). In combination with the variable expression of adhesins, the frequent recombination events provide fungi with an ever-changing outer coat, thereby keeping one step ahead of the host's adaptive immune system. The mechanism is different from that of trypanosomes, which continuously express different subsets of the hundreds of VSG (variant surface glycoproteins) genes through recombination and metastable gene silencing (Barry and McCulloch, 2001; Scherf et al., 2001).
Fungal adhesion is an unusually complex and variable phenotype. First, each cell possesses several slightly different adhesins. Differential expression of these various adhesin genes enables fungi to quickly adapt their adhesive properties to a particular environment. While the exact conditions that induce adhesion are not fully understood, many different genetic and epigenetic signalling cascades are employed to ensure proper regulation of this vital phenotype. Some results indicate that these signalling cascades show significant differences between related species, adding to the complexity of the phenotype (Hoyer et al., 2001). Moreover, the genes encoding fungal adhesins are prone to recombination. The internal tandem repeats in these genes trigger frequent recombination events, leading to the formation of new adhesins with novel phenotypes. Together, these different mechanisms make fungal adhesion one of the most plastic and variable phenotypes, underscoring the need for fungi to adapt their adhesion behaviour to the highly variable environment and to constantly explore new opportunities for infection. It is important to note that the subtelomeric positioning of many of the adhesin genes is no coincidence. The relatively isolated position in the genome allows epigenetic switching and frequent recombination events without the risk of affecting other genes. Moreover, the numerous repetitive sequences in telomeric and subtelomeric regions help to recruit and position chromatin remodelling enzymes such as the Sir proteins, and may stimulate recombination events (Castano et al., 2005; Iraqui et al., 2005; Verstrepen et al., 2005). It would be interesting to see if other subtelomeric gene families, such as some of the S. cerevisiae carbohydrate uptake and metabolism genes, are also subjected to similar variability-inducing mechanisms as the adhesin genes.
Whereas constant switching and adaptation of adhesion may be crucial for survival in natural habitats, the instability of the flocculation profile is a true nightmare for industrial applications that rely on a constant and predictable yeast performance. Brewers and winemakers constantly struggle to get a grip on the flocculation behaviour of their little helpers, often without much success. In principle, recombinant DNA techniques make it possible to create more stable strains with an appropriate flocculation profile (Pretorius and Bauer, 2002). Yeasts carrying a specific FLO gene under transcriptional control of a late-fermentation promoter have proven useful in laboratory-scale fermentations (Ishida-Fujii et al., 1998; Verstrepen et al., 2001) but are currently not acceptable to consumers.
The remarkable plasticity of fungal adhesion also causes concern in medicine. Despite awareness and preventive measures to avoid the formation of highly resistant fungal biofilms, yeast cells are capable of adhering to many of the different materials used in today's hospitals, and as a consequence, yeast infections are becoming more prevalent. Given the variability of the fungal adhesins, directly targeting these cell-surface proteins or using them in vaccines seems a poor strategy. Better targets for new drugs are the less-variable features of adhesion, such as the biosynthesis of the GPI-anchor of the adhesins or the transcriptional and post-transcriptional regulatory mechanisms of adhesion. In a shorter term, the use of anti-fungal coatings on medical devices seems like the most realistic way of fighting disseminated yeast infections.
The authors thank Chia-Yung Wu for the picture used in Fig. 2B and An Jansen and Sharad Ramanathan for the useful comments. Kevin J. Verstrepen is a Fellow of the Harvard University Bauer Center for Genomics Research and a research assistant of the Fund for Scientific Research – Flanders (Belgium) (FWO-Vlaanderen). Frans M. Klis acknowledges the financial support from the EU (Framework VI, GALAR FUNGAIL II and FUNGWALL).
The following supplementary material is available for this article online:
Fig. S1. Phylogenetic tree of cellular adhesins and mucins. The figure shows the phylogenetic similarity of the best-known adhesin genes of S. cerevisiae, S. pastorianus, C. albicans and C. glabrata. Other genes that are not strictly adhesin family members, but do show homology to adhesins, are also included in the analysis. These include the S. cerevisiae a-agglutinin subunits encoded by AGA1 and 2, and the human mucin (MUC) genes. A surprising result of this analysis is the relative isolated position of FLO11, which shows more similarity to the yeast agglutinin subunit Aga1 and the human mucins than to the yeast flocculins encoded by FLO1, 5 and 9. Another remarkable finding is the high similarity between the C. albicans ALS4 gene and the S. cerevisiae uncharacterized open reading frame YFL051C. The tree was constructed using the clustalx software; the robustness of the tree was confirmed with standard bootstrapping techniques.