Population dynamics and the evolution of antifungal drug resistance in Candida albicans


Correspondence: Katy C. Kao, Department of Chemical Engineering, Texas A&M University, 3122 TAMU, College Station, TX 77843-3122, USA. Tel.: 979 845 5571; fax: 979 845 6446; e-mail: kao.katy@gmail.com


Candida albicans is an important human fungal pathogen. Resistance to all major antifungal agents has been observed in clinical isolates of Candida spp. and is a major clinical challenge. The rise and expansion of drug-resistant mutants during exposure to antifungal agents occurs through a process of adaptive evolution, with potentially complex population dynamics. Understanding the population dynamics during the emergence of drug resistance is important for determining the fundamental principles of how fungal pathogens evolve for resistance. While few detailed reports that focus on the population dynamics of C. albicans currently exist, several important features on the population structure and adaptive landscape can be elucidated from existing evolutionary studies in in vivo and in vitro systems.


Evolution allows each organism to survive and adapt to changing environments and thus is the driving force behind the biodiversity on earth. The discovery and use of antibiotics is a major advancement in modern medicine. However, the widespread use of antimicrobial agents results in the emergence of drug-resistant strains among previously drug-susceptible populations. These drug-resistant strains arise in the population during the exposure to the antimicrobial agent through a process of adaptive evolution. During adaptive evolution, mutants arise spontaneously, and through a process of natural selection, the adaptive mutant will expand in the population until either it becomes the dominant clone or a fitter clone arises in the population. Depending on the selective pressure, adaptive landscape, frequency of beneficial mutations and population size, the population structure may be complex and consist of multiple-resistant genotypes. Currently, only a few studies have looked at drug resistance from the perspective of population dynamics, and thus, several important questions underlying the evolution of drug resistance are largely unanswered, including (1) the rate at which beneficial mutations arise in the population during the exposure to antifungal agent, (2) the effect of the population size on the evolutionary process of drug resistance, (3) correlation, if any, between the evolutionary trajectory (the order of occurrence of adaptive mutations) and the fitness level achieved by the population, and (4) the adaptive landscape (collection of the fitness effects of each resistant genotype) under the selective pressure. A picture of the population dynamics (the changing genotypic landscape within the microbial population in the presence of antibiotics) will provide valuable insights into the aforementioned questions and contribute to the elucidation of the fundamental principles underlying how microbial pathogens evolve resistance to antimicrobial agents.

Among human fungal pathogens, Candida spp. is recognized as a major challenge in public health, causing potentially life-threatening invasive infections in immunocompromised patients. Candida spp. is the fourth most common cause of blood stream infections with a mortality rate approaching 50% in US hospitals (Zaoutis et al., 2005; Pfaller & Diekema, 2007). The species distribution among clinical Candida isolates varies depending on the geographic regions, with Candida albicans (C. albicans) being the most commonly isolated species in Candidaemia according to a 10.5-year global survey (Pfaller et al., 2010), from the lowest frequency (48.9%) in North America to the highest one (67.9%) in European; however, there is an upward trend in the frequency of isolation of non-albicans species (NAC), likely due to reduced susceptibility to antifungal agents in some NAC (Lai et al., 2008; Pfaller & Diekema, 2010; Pfaller, 2012). In the management of fungal infections, there have been significant recent advances in antifungal therapy, including the introduction of a new generation of antifungal agents, the use of combination therapy, and improved standardization of susceptibility testing; however, drug resistance still poses a challenge in the management and treatment of fungal infections (Kanafani & Perfect, 2008; Chapeland-Leclerc et al., 2010; Pfaller, 2012). In the United States, the treatment associated with Candidemia cost more than US $1 billion annually (Beck-Sague & Jarvis, 1993; Miller et al., 2001). The high mortality rate, the rapid development of drug resistance, and the high cost associated with therapeutic treatment make Candida spp. a medically important group of fungal pathogens.

Antimicrobial resistance has become increasingly important in antifungal therapy. Resistance to nearly all major antifungal agents has been reported in clinical isolates of Candida spp. (Marr et al., 1998; Sanglard & Odds, 2002; Katiyar et al., 2006), which poses a major public health concern as the arsenal of antifungal agents is limited. Single nucleotide polymorphism, loss-of-heterozygosity (LOH) and gross chromosomal rearrangements have been found to be important processes in the development of drug resistance (Selmecki et al., 2006, 2008, 2009). Research within the past couple of decades has identified numerous drug resistance mechanisms. Mutations in drug targets, such as ERG11 in fluconazole resistance and FKS1 in echinocandin resistance (Loffler et al., 1997; Lamb et al., 2000; White et al., 2002; Park et al., 2005), and mutations that alter the expression level of efflux pumps, such as CDR1, CDR2 and MDR1 in resistance to azoles, are commonly found in drug-resistant isolates (Sanglard et al., 1997; Wirsching et al., 2000, 2001). Aneuploidy has also been associated with the acquisition of drug resistance in many clinical isolates (Selmecki et al., 2006, 2008), such as isochromosome 5L in fluconazole resistance (Selmecki et al., 2008). In addition, mutations leading to the change in the membrane composition, alteration in the ergosterol biosynthesis pathway, and induction in biofilm formation are also correlated to increased resistance to fluconazole (Kelly et al., 1996; Nolte et al., 1997; Loffler et al., 2000; Chandra et al., 2001). Although the resistance mechanisms have been extensively studied, there are still drug-resistant mechanisms yet to be identified; for example, approximately half of the drug-resistant strains have unknown mechanisms of resistance in one collection of clinical isolates (White et al., 2002).

Given the importance of Candida spp. in public health and the paucity of systematic analysis on the emergence of drug resistance in fungal pathogens from the evolutionary perspective, in this review, we focus on the existing literature related to population dynamics of C. albicans, the most well-studied Candida spp., in the presence of antifungal agents in in vivo and in vitro systems. The analysis and discussion based on C. albicans also largely apply to other Candida spp.

Systems used to study population dynamics

Clinical isolates from a single patient throughout the course of treatment offer a unique look at the adaptive evolution of the organism in vivo. However, variables such as the genetic composition and size of the founding fungal pathogen population cannot be controlled in patient studies. In addition, such time-course patient samples are rare and generally only one clone is isolated and analysed at each time point; thus, the amount of population dynamics information that can be gained is limited as it is not possible to determine the population frequency of each allele at each time point, nor is it possible to estimate the time at which each allele arose in the population. Animal studies involving infecting mice with C. albicans offer control over the initial genotype and size of the fungal population, although the effective size of the population inside the host has yet to be accurately determined. Studies using murine models have looked at the ability of a resistant genotype to dominate the population by varying its' initial fraction in the infecting population (Andes et al., 2006). Animal models have also been used to determine the emergence of drug resistance using different dosing regimens (Andes et al., 2006). These in vivo systems provide a more realistic situation, including physically structured distribution of fungal pathogens, to study the emergence of drug resistance within a host; but it is difficult to control factors such as the host physiological states and the population size of the pathogen inside the host in these in vivo systems.

Evolutionary studies in vitro offer more controlled and reproducible environments for studying population dynamics during long-term exposure to antifungal agents. The genotypes of the starting population, the population size, and the selective pressure during the evolution can be readily controlled, allowing reproducibility of the environmental conditions for each experiment. There are two major types of systems used for in vitro evolution studies: batch serial transfer or continuous cultures. In batch serial transfer experiments, the population is grown either on solid or liquid media, and a small fraction is serially transferred to fresh media containing the antifungal agent periodically (Cowen et al., 2000). The population undergoes different growth phases during batch cultivation, as the nutrient content of the environment and the physiological state of the cell both change as a function of time. Continuous culture systems (using chemostats), on the other hand, provide a more constant environment, which helps to keep the population in physiological steady state. The effective population sizes in continuous systems are also generally larger than that of batch systems. Both these systems have been used to study the emergence of drug resistance in C. albicans (Cowen et al., 2000; Huang et al., 2011). Molecular studies of isolates from both in vivo and in vitro systems have shown that starting from a single drug-susceptible genotype, multiple resistance mechanisms are involved in the emergence of drug resistance and that the same mechanisms can be found both in experimental populations and clinical isolates. Thus, while the environmental conditions used in in vitro systems may not exactly mimic those of in vivo systems, the resistance mechanisms of the fungal pathogen have not been found to be different from those of in vivo systems; and they can provide important and useful information by exploring the population dynamics during the emergence of drug resistance.

Theories governing the population structure during adaptive evolution

Although C. albicans is a diploid organism with mating type-like loci (Hull & Johnson, 1999) and a parasexual cycle (Bennett & Johnson, 2003), it is still considered to be asexual because of the lack of observed haploid state, spore formation, and many other processes related to sexual reproduction (Olaiya & Sogin, 1979; Riggsby et al., 1982). Therefore, during evolution, C. albicans can be assumed to be evolving asexually. And because there is no evidence that C. albicans can transfer resistance genes horizontally, it is assumed that resistance mechanisms are acquired via mutations.

There are currently three major theories describing the population structure during asexual evolution (Fig. 1). The first is the successional-fixation regime (Desai et al., 2007) (Fig. 1a), also called clonal replacement, which generally applies to small populations when the evolution is mutation-limited and the time between establishments of successive mutations is much greater than the time it takes for the fixation of a beneficial mutation. Thus, in the case that the population structure can be described by the clonal replacement model, most mutations are lost and only one mutation can become established leading to one selective sweep at a time; therefore, the population is assumed to be homogeneous except during the periods when the beneficial mutant is sweeping through the population. The second theory is called clonal interference (Fig. 1b) or sometimes called one-by-one clonal interference because it is assumed that only one mutation can become fixed at a time. This occurs when mutations are established faster than the rate of fixations, multiple beneficial mutants can coexist and compete against each other until the one with the greatest fitness advantage outcompetes all the other genotypes and become the next founding genotype for subsequent evolution. The population is thus heterogeneous except immediately after the complete sweep by the fittest mutant. This theory focuses on the competition between mutations with different fitness effects (Gerrish & Lenski, 1998; Orr, 2000; Gerrish, 2001; Kim & Stephan, 2003; Campos & de Oliveira, 2004; Wilke, 2004) and assumes that mutations cannot be stacked in the same genetic background before the fixation of the most-fit mutation. However, the size of a typical laboratory microbial population is large enough to support multiple beneficial mutations occurring in same lineage before the first mutation in that lineage can fix (Desai & Fisher, 2007), which is the basis of the third theory: the ‘multiple-mutation’ model (Desai et al., 2007) (Fig. 1c). Multiple theoretical and experimental studies in other organisms have indirectly suggested the importance of this multiple-mutation effect (Yedid & Bell, 2001; Shaver et al., 2002; Bachtrog & Gordo, 2004). A study using Saccharomyces cerevisiae evolving under carbon source limitation showed experimental support for this theory (Desai et al., 2007). Therefore, depending on the size of the population, the rate of mutation, time required for the establishment of a beneficial mutation, the fitness distribution of the mutations, and other important factors, evolution dynamics in C. albicans during long-term exposure to antifungal agents may be described by one, or combinations, of the theories mentioned above. Because without exact genotype information, it is difficult to differentiate between the one-by-one clonal interference model and the multiple-mutation model, we will use the general term clonal interference to describe a heterogeneous evolving population structure.

Figure 1.

Three models governing the population structure during adaptive evolution of asexual populations: Clonal replacement (a), one-by-one clonal interference (b), and multiple-mutation model (c). Parental alleles are small cased (a, b, c) and beneficial alleles are capitalized (A, B, C).

Clonal interference observed in C. albicans during the emergence of antifungal drug resistance

In the seminal paper on C. albicans adaptive evolution during antifungal drug exposure, Cowen et al. (2000) evolved 12 parallel populations, six in the absence and six in the presence of fluconazole for 330 generations, and isolated clones throughout the course of the evolution. Based on known mechanisms of fluconazole resistance, the isolated clones were screened for mutations in potential target genes (Cowen et al., 2000); in one population, a neutral mutation identified in an earlier adaptive mutant was not found in a later isolated adaptive mutant, clearly suggesting the presence of clonal interference in the fluconazole-exposed population (Cowen et al., 2000). The argument for the presence of clonal interference in C. albicans populations evolving in the presence of antifungal drug was unambiguously determined by our recent work using an adaptive evolution method called visualizing evolution in real-time (VERT) to help track the population dynamics in an evolving population (Huang et al., 2011). VERT involves the use of a set of different fluorescently marked isogenic strains as the initial population in adaptive evolution. The occurrence of an adaptive event (the occurrence and expansion of an adaptive mutant) in the population can be visually observed by the expansion of the fluorescently marked subpopulation containing the adaptive mutant. Thus, if the population dynamics follows the clonal replacement model, the first expanding subpopulation will take over the entire population. However, if clonal interference is present in the evolving populations, then the subpopulations will expand and contract as different adaptive clones compete for expansion. Figure 2 shows an example of the VERT data for C. albicans evolving in the presence of stepwise increases of fluconazole in a chemostat system (Huang et al., 2011). The use of VERT also allowed us to estimate the frequency at which adaptive mutants arise in the population. We found that the frequency of adaptive events increased in the presence of the drug. Interestingly, the frequency of adaptive events appears to be independent of drug concentration, at least within the drug concentration used in our study (Fig. 2b and c); approximately 9 and 10 adaptive events were observed in the populations exposed to lower and higher (two times higher) concentrations of fluconazole, respectively.

Figure 2.

Evolutionary dynamics of experimental populations evolved in no fluconazole (a) (Huang et al., 2011), lower concentration of fluconazole (b), and higher concentration of fluconazole (c) (Huang et al., 2011). The coloured bars represent the relative fraction of each coloured subpopulation: RFP (red), GFP (green), YFP (yellow) as determined using Fluorescent Activated Cell Sorter (FACS). FLU: The concentration of fluconazole in the feed. The observed adaptive events (consecutive increase in the relative proportion of a coloured subpopulation in two or more data points) are numbered for each population at the end of each expansion (the generation at which the expanding subpopulation is at its highest frequency).

Is clonal interference also present during the emergence of drug resistance in C. albicans in vivo? Transcriptional analysis of several target genes in a series of 17 isolates from an AIDS patient showed sequential stacking of resistance mechanisms in isolates obtained throughout the course of treatment (White, 1997), suggesting the population structure in vivo during the course of treatment may be governed by the clonal replacement model. However, this may not always be the case. A series of nine clinical isolates of C. albicans isolated from a bone marrow transplant patient, who underwent a series of antifungal drug treatment (Marr et al., 1997), were analysed for LOH at predicted alleles and gross chromosomal rearrangements (Coste et al., 2006; Selmecki et al., 2008). Results from these analyses clearly showed a heterogeneous population where multiple resistance alleles coexist, demonstrating that clonal interference also occurs in vivo.

Even though the evolving population is heterogeneous, a question of interest is whether there is convergence in their resistance mechanisms within and between parallel populations. Prior study in S. cerevisiae evolving under glucose-limited condition showed that in one evolving population, adaptive mutants from different lineages evolved similar mechanisms of adaptation based on both transcriptional and genotypic analyses (Kao & Sherlock, 2008). Unfortunately, there exist few studies of time-course samples in C. albicans currently. In C. albicans, studies of in vitro isolates evolved in the presence of fluconazole found different replicate populations reached different fluconazole MIC levels, suggesting a divergence in resistance mechanisms between different populations (Cowen et al., 2000). Further transcriptome studies of the same series of in vitro evolved isolates demonstrated similarities and divergences in potential resistance mechanisms between different lineages (Cowen et al., 2002); and while evidence seems to suggest that similar resistance mechanisms are present in isolates from the same population, because of the small number of time-course samples analysed, it is not clear whether there is convergence in resistance mechanisms among isolates within the same population.

During the emergence of drug resistance, each mutation that arises represents a step along the fitness landscape. An important question is whether, starting from the same point on the fitness landscape (same genotype), parallel populations will converge in their evolutionary trajectories (whether they will traverse similar paths along the fitness landscape). Although no detailed studies exist currently to answer this question definitively, some prior experimental evidence suggests that early steps in the evolutionary trajectory may ‘influence’ the population down certain evolutionary paths. We will discuss some of the evidence here.

First, similarities in gene expression profiles between several parallel populations were observed in transcriptome studies of the in vitro evolved populations by Cowen et al. (2002). Specifically, in two parallel populations they analysed, the transient changes in transcriptional expression profiles from time point isolates were very similar (Cowen et al., 2002), suggesting that convergence in evolutionary trajectories may occur. A study with parallel populations of S. cerevisiae subjected to either stepwise increases in or a single high concentration of fluconazole found similar mechanisms arising in independent populations under the same selection scheme (Anderson et al., 2003), suggesting that selection regimen may determine resistance mechanisms involved and that these resistance mechanisms possibly converge in parallel populations in S. cerevisiae. The other evidence comes from more detailed genotypic analysis of the same series of C. albicans isolates by Selmecki et al. (2008), where they found populations that have obtained an isochromosome 5 mutation early in the evolution achieved higher MIC at the end compared with the populations that did not, suggesting that the early evolutionary steps may influence the final evolutionary outcome.

Advances in genomic tools such as tiling arrays, comparative genome hybridization microarrays (array CGH), and ultra-high-throughput sequencing are now allowing researchers to have a better understanding of the genotypic changes associated with adaptation [for review see (Dettman et al., 2012)], such as drug resistance (Selmecki et al., 2010). The applications of these tools to time-course isolates obtained in vitro and in vivo will yield the necessary correlations between genotypic and phenotypic changes in resistant strains and help researchers to gain a firmer grasp on the evolutionary trajectories of fungal pathogens during exposure to antifungal agents.

Factors that affects population dynamics

In addition to the aforementioned factors (e.g. population size, relative fitness coefficients, rate of beneficial mutations, etc.) that contribute to the population dynamics during adaptive evolution, additional factors such as dosing regimens and the mode of action of the antifungal agent may also contribute to the population dynamics during the emergence of drug resistance in C. albicans. A series of in vivo studies in murine model shed some light on the importance of some of these factors on antifungal drug resistance in C. albicans (Andes et al., 2006). Andes et al. (2006) investigated the impact of different fluconazole (a fungistatic agent) dosing regimens, using different dose levels and dosing intervals, on the outgrowth of resistant strain with different initial ratios of drug-resistant and susceptible strains in a murine model; they found a lower but more frequent dosage of fluconazole led to less frequent outgrowth of the resistant strain compared with higher but more infrequent dosage. Another study by the same group revealed a similar effect of dosing regimen on drug resistance emergence when they evolved an initially drug-susceptible strain of C. albicans in a murine model (Andes et al., 2006). Results from these studies suggest different selection strategies may have different impacts on the expansion of drug-resistant genotypes within the population, leading to different population dynamics and ultimately to different evolutionary outcomes. In addition, they found that if the initial population contained at least 10% of the drug-resistant clone, the evolving population behaved phenotypically as entirely drug resistant, suggesting that the population structure prior to drug exposure is an important factor in determining the evolutionary outcome of the population (Andes et al., 2006).

The mode of action of the antifungal agent may also be a contributing factor on the emergence of drug resistance. In vitro studies with aminoglycoside antimicrobials against bacterial pathogens demonstrated the best way to prevent the emergence of the resistance to this drug in cell populations is a high level but infrequent dosing that maximizes the peak concentration of the drug (Blaser et al., 1987), which was different from what was observed in C. albicans with fluconazole (Andes et al., 2006). A possible explanation for the difference in the best dosing strategy in the different systems was proposed by Andes et al. (2006) to be the differences in modes of action on the target organisms. Aminoglycoside antimicrobials have cidal activities against the bacteria tested while fluconazole is a fungistatic agent for C. albicans. The cidal activity of the aminoglycoside antimicrobials can effectively reduce the population size of the pathogens and thus reduce the supply of beneficial mutations. Under this type of selection, genetic drift may play a more important role because of the smaller population sizes, leading to the higher frequency of loss of rare beneficial mutations; thus exposure to a cidal agent may result in a more homogeneous population structure containing few drug-resistant mutants. However, a fungistatic agent may not effectively reduce the size of the population significantly to prevent the emergence of rare beneficial mutations, possibly leading to a more heterogeneous population containing multiple beneficial mutants. Thus, depending on the mode of action of the antimicrobial agent, different population dynamics may emerge. Additional studies with C. albicans using fungicidal agents will help to shed additional insight on the effects of the mode of action of the drug on the population dynamics during drug exposure.

Fitness costs associated with antifungal drug resistance

The fitness effect associated with a resistance mutation plays a key role in determining whether the resistant genotype can survive drift and whether it will become dominant in the population (Andersson, 2003; Andersson & Hughes, 2010). It is expected that if drug-resistant mutations carry a fitness cost in the absence of drug, the proportions of the drug-resistant phenotypes will decrease and may even be eliminated from the population when the drug is removed and further compensatory evolution is absent. This type of trade-off in the relative fitness between different environments is commonly observed (Johanson et al., 1996; Schrag & Perrot, 1996; Schrag et al., 1997; Bjorkman et al., 1998, 1999; Sandegren et al., 2008). Several scenarios have been used to describe such differences in fitness effects in different environmental conditions (Elena & Lenski, 2003). The first scenario is antagonistic pleiotropy (AP), which describes mutations that are beneficial in one condition but are deleterious in another environment. The second is mutation accumulation (MA), in which neutral mutations that accumulated in one environment are deleterious in another condition. The third scenario is independent adaptation (IA), which describes mutations with beneficial effects in one environment but neutral in another. And a fourth scenario is cross-adaptation (CA) in which beneficial mutation in one environment is also beneficial in another. Although these fitness trade-off scenarios are commonly observed in natural and experimental systems, few studies have focused on their underlying mechanisms.

Some of these trade-off scenarios are observed in drug-resistant isolates of C. albicans. Evidence of AP in drug-resistant mechanisms was observed in a single isolate from our recent evolutionary study of C. albicans (Huang et al., 2011). In this study, cell populations were evolved under the selective pressures of fluconazole and limiting carbon source (glucose). An adaptive clone isolated from one population (DP-1-M5) showed a significant increase in the relative fitness compared to the parental strain in the presence of drug, but the increased drug resistance had a fitness cost, as the mutant showed a lower relative fitness in the absence of the drug (Table 1), demonstrating a clear case of AP. However, the majority of the isolates from this study fall in the IA or CA categories described above, where mutations that are beneficial in the presence of the drug are either neutral or beneficial in the absence of the drug (see Table 1). This is contrary to results from Cowen et al. (2001); in their study, most isolates with increased fitness in the presence of the drug compared with the parental strain showed neutral or negative fitness in the absence of the drug (AP or IA). Possible explanations for the difference in our observations may be due to the differences in C. albicans strains used for the evolution experiments, the media used for the evolution (yeast nitrogen base vs. RPMI 1640), and the population size and evolution system used (chemostat vs. serial batch transfer). The use of serial batch transfer involves a larger bottleneck effect during each transfer. Thus, it is likely that the majority of the beneficial mutations that arise are lost in the process. In a continuous system, on the other hand, beneficial mutants have a higher probability of being retained in the system for further evolution. However, the exact mechanisms for the fitness trade-offs will require further studies to identify all the underlying adaptive mutations and to characterize their exact fitness effects.

Table 1. Relative fitness coefficients for isolated adaptive mutants compared with parental strains (Huang et al., 2011)Thumbnail image of


Both in vivo and in vitro data have shown C. albicans populations to be heterogeneous and that clonal interference plays an important role in the population structure during exposure to antifungal agents. With the development of VERT, we can now track the population dynamics during adaptive evolution to readily estimate the frequency at which drug-resistant mutants arise in the population and to isolate mutants in a systematic manner. While clinical isolates from patients throughout the course of treatment would be the ideal system to study the emergence of antifungal drug resistance, it is difficult and often not practical to control. In vitro systems using bioreactors offer controlled and more reproducible environments. Combined with VERT, the use of in vitro systems has the potential to determine the evolutionary trajectories of parallel populations and correlation, if any, between the evolutionary trajectory (the order of occurrence of adaptive mutations) and the fitness level achieved by the population. Even though the molecular mechanisms underlying antifungal drug resistance have been extensively studied, there are still a large fraction of azole-resistant clinical isolates that have no known resistance mechanisms (White et al., 2002). With rapid advances in genomics and molecular biology tools, researchers now have the capability to identify the exact mutations in drug-resistant isolates from in vivo and in vitro systems, which will likely lead to identification of additional mechanisms of drug resistance. Indeed, a recent study by Selmecki et al. (2009) identified a segmental trisomy on chromosome 4, which included a gene encoding the NADPH-cytochrome P450 reductase, using array CGH, and may have found a new mechanism for fluconazole resistance. The identification and characterization of these genetic determinants that underlie drug resistance will expand our knowledge on the fitness landscape of drug resistance in C. albicans and other medically important NAC.


The authors would like to acknowledge partial financial support from the National Science Foundation MCB-1054276 and the Texas Engineering Experimental Station.