Candida albicans, the most prevalent human fungal pathogen, was considered an obligate
diploid that carried recessive lethal mutations throughout the genome. Here, we demonstrate
that C. albicans has a viable haploid state that can be derived from diploid cells
under in vitro and in vivo conditions and appears to arise via a concerted chromosome
loss mechanism. Haploids undergo morphogenetic changes like those of diploids including
the yeast-hyphal transition, chlamydospore formation, and a white-opaque switch that
facilitates mating. Haploid opaque cells of opposite mating type mate efficiently
to regenerate the diploid form, restoring heterozygosity and fitness. Homozygous diploids
arise spontaneously by auto-diploidization and both haploids and auto-diploids display
a similar reduction in fitness, in vitro and in vivo, relative to heterozygous diploids,
suggesting that homozygous cell types are transient in mixed populations. Finally,
we constructed stable haploid strains with multiple auxotrophies that will facilitate
molecular and genetic analyses of this important pathogen.
Background
The opportunistic fungal pathogen C. albicans has been studied extensively since the
1800’s and has been considered a strictly diploid organism with no haploid state.
The diploid nature of the organism has complicated genetic and genomic analyses of
C. albicans biology and virulence. Studies in the 1960’s through the early 1980’s
debated the ploidy of C. albicans, and proposed that a haploid state existed based
on cell size heterogeneity
1
, parasexual genetics
2
and estimates of DNA content by fluorimetry
1,3
. Haploid clinical isolates were reported
2,3
, but were later found to be diploid in genome content or were shown to be non-albicans
Candida species (T. Suzuki, personal comm. and Fig. S1). Research supporting the diploid
nature of the organism included DNA content measurements that suggested a genome size
similar to that of diploid Saccharomyces cerevisiae
4,5
, molecular studies that demonstrated the need to delete two gene copies to produce
a null mutant
6,7
, genetic studies that demonstrated heterozygosity of specific alleles
8,9
, and ultimately the complete genome sequence, which revealed heterozygosity throughout
much of the genome
10
. This, together with the predominantly clonal nature of C. albicans within an individual
host
11,12
, was interpreted as evidence that the organism spent most, if not all, of its life
cycle in a diploid state.
The ‘obligate diploid’ nature of C. albicans was proposed to be due to recessive lethal
mutations dispersed throughout the genome
13,14
. However, studies of C. albicans chromosome monosomy
15,16
, recombination
17
and haplotype mapping
18
demonstrated that homozygosis of certain chromosomes can occur, arguing against this
balanced lethal mutation hypothesis. A diploid-tetraploid parasexual cycle was also
discovered in C. albicans and shown to involve a switch to the ‘opaque’ physiological
state that renders cells mating-competent
19,20
, conjugation between opaque diploids to form tetraploids
21,22
, and subsequent ploidy reduction resulting in diploid progeny that often carry multiple
aneuploid chromosomes
18,23
. Importantly, specific aneuploid chromosomes can provide a selective advantage under
stressful conditions such as exposure to antifungal drugs
24
. Parasexual ploidy reduction in tetraploids occurs via a non-meiotic process termed
‘concerted chromosome loss’. This process can facilitate the rapid generation of diversity
through the production of homozygous and aneuploid progeny, under conditions where
outcrossing in the host is unlikely
25
. If a haploid state for C. albicans were to exist, it could facilitate the elimination
of lethal alleles from the population. Furthermore, mating between different haploids
could promote adaptation to changing conditions within the mammalian host. With the
advent of whole genome approaches that can distinguish ploidy states, it is now possible
to ask if C. albicans exists, even transiently, in the tetraploid or the haploid state.
Detection of haploid C. albicans cells
Haploid C. albicans was serendipitously discovered during experiments to follow loss
of heterozygosity (LOH) at multiple independent loci. Using a multiply marked derivative
of SC5314, the laboratory reference strain
10
, we selected for an initial LOH event at the GAL1 locus by growth on 2-deoxygalactose
(2-DOG) and subsequently screened for additional LOH events at four other heterozygous
loci. Importantly, growth in 2-DOG does not affect LOH rates (Fig. S2) and as such,
the selection for cells with an LOH event does not artificially induce LOH. Amongst
the ~2500 Gal− isolates, 42 exhibited additional LOH events and intriguingly one strain,
Haploid I, was homozygous for all of the markers tested, as well as for multiple SNPs
on every chromosome. Analysis of DNA content in Haploid I by flow cytometry indicated
that the genome had half the amount of DNA of that in a diploid control (Fig. 1a).
Once we became aware that haploid C. albicans cells could be detected, we used flow
cytometry to screen isolates from many sources, including in vitro stresses,
26
from which one haploid isolate (Haploid IV) was identified amongst small colonies
growing in the presence of fluconazole (in the halo of an E-test strip), a commonly
used antifungal drug. We also screened C. albicans cells isolated from in vivo mouse
models of candidemia
27
and candidiasis
28
and discovered several additional isolates with 1N DNA content. The overall population
of post-in vivo isolates screened for ploidy included ~300 isolates from YPD plates
(Gal+), and ~740 isolates from the 2-DOG selection plates (Gal−) and only the DOG-resistant
colonies contained isolates that were haploid. We found that 3.2% of DOG-resistant
colonies from the OPC model were haploid (10 out of 312), while 1.2% of DOG-resistant
colonies from the systemic model were haploid (5 out of 431). Since LOH frequencies
in vivo are ~10−3, 29
, this indicates that detectable haploids appeared at a frequency of 1–3 per 100,000
cells. Loss of GAL1 is often, but not always associated with the haploid state (e.g.
Haploid IV), especially in diploid progenitors that are heterozygous for this locus.
In total, we characterized eleven strains with haploid (or close to haploid) flow
cytometry profiles (Fig. 1a, S3 and Table S1) that were derived from multiple independent
and varied sources.
Since previous reports of haploid C. albicans
3
included misidentified species (Fig. S1), we determined if these newly identified
isolates were bona fide C. albicans by PCR amplifying and sequencing the mating-type-like
(MTL) locus. All isolates contained sequences identical to either the MTL
a
or MTLα alleles of SC5314 (Fig. S4 and data not shown). Next, we analyzed the haploid
genomes by hybridization to SNP/CGH arrays
30
which measure allelic ratios and also infer relative chromosome copy numbers. Six
isolates were completely euploid (Fig. 1b and S5); in the remaining five haploids,
one or two chromosomes (Chr6 and Chr7) were disomic. These disomic chromosomes were
heterozygous except in one isolate (Haploid IV) where the disomic chromosome (Chr6)
was homozygous.
Based upon the C. albicans haplotype map
30
, each haploid chromosome primarily contained alleles from only one parental homolog
with few, if any, obvious crossovers detectable (Fig. 1b and S5). This implies that
haploids did not arise through conventional meiosis, which usually requires at least
one crossover per chromosome
31
. The few crossovers detected likely arose by mitotic recombination prior to selection
for LOH events. For example, Haploid I was isolated following selection for loss of
GAL1, which is within the crossover region on Chr1. Furthermore, in C. albicans haploids
(Fig. 1b and S5), the majority of disomic chromosomes were heterozygous, which would
appear following segregation of homologs in meiosis I. Such meiosis I segregation
is not seen in Drosophila male meioses, which lack recombination
32
; it is seen in Candida lusitaniae meiotic progeny, which undergo high levels of recombination
33
. The low level of genetic recombination together with the presence of heterozygous
disomic chromosomes remains most consistent with random homolog segregation in C.
albicans. Accordingly, we propose that haploids arise via a concerted chromosome loss
mechanism akin to that described for the diploid-tetraploid C. albicans parasexual
cycle
23
, although a non-conventional meiotic program cannot be completely discounted.
The existence of haploids refutes the argument that recessive lethal alleles are present
on each C. albicans homolog
13,14
. Nonetheless, the absence of some chromosomal homologs from the haploid progeny suggests
that recessive lethal alleles may exist. Indeed, only the ‘b’ homologs for Chr3, Chr4,
Chr6, Chr7 and most of Chr1 were detected in the haploids analyzed (Fig. 1b and S5).
Interestingly, a similar homolog bias was seen in diploid parasexual progeny derived
from SC5314
18
. In contrast, homologs of Chr5, which carries the MTL, appeared in equal numbers
(six MTL
a and five MTLα) and were entirely of one or the other parental haplotype. Similarly,
both homologs of ChrR and Chr2 were observed. As such, homologs from more than half
the chromosomes (Chr3, Chr4, Chr6, Chr7 and most of Chr1) potentially carry at least
one recessive lethal mutation and may limit the frequency with which viable haploids
appear. Accordingly, we propose that haploidization may provide an effective mechanism
for eliminating recessive lethal mutations from the predominantly diploid populations
of C. albicans.
Auto-diploidization of haploids
In the course of these studies, prolonged propagation of haploid isolates yielded
cultures with a mixed population of haploid and diploid cells (Fig. 1c). Subsequent
colony purification yielded distinct haploid and/or diploid populations (Fig. 1d).
Surprisingly, some haploid isolates also diploidized during the process of strain
shipping, which involved storage, transit, and revival from partial dehydration on
a solid surface. For example, a potential haploid isolate identified in Taiwan was
sent to Minnesota on sterile filter paper. Once revived and grown in liquid culture,
the genome was diploid (Fig. 1e) and homozygous for all SNPs (Fig. 1b, ‘Auto-dip.’),
suggesting auto-diploidization from a haploid phase. While homozygous diploids could
arise through mitotic defects or by self-mating
34
, mating between cells of the same mating-type was not detectable, as discussed below.
Thus, we suggest that auto-diploids arise through mitotic defects that may be analogous
to the auto-diploidization events in vertebrate haploid stem cell cultures
35–37
.
Haploid Morphogenesis and Mating
Consistent with reduced ploidy content in other yeast species
38
, haploid C. albicans cells were smaller, on average, for both cell and nuclear size
compared to diploid cells (Fig. 2a and S6). The bud site selection patterns of haploids,
like diploid C. albicans were predominantly axial, with a small fraction of cells
displaying a bipolar budding pattern (Fig. 2b)
39
. Significantly, under conditions that induce morphogenesis of diploid C. albicans
to form true hyphae, pseudohyphae, and chlamydospores
40,41
, similar morphogenetic events occurred in haploids (Fig. 2c).
C. albicans diploid cells that are homozygous at the MTL can undergo a transcriptionally
induced switch to the opaque state that renders them competent for mating
19,42
. Similarly, haploids readily switched from the white (Wh) to the opaque (Op) state,
as detected by colony sectoring (Fig. 2d). Haploid opaque cells had elongated opaque-like
cell morphology
43
, yet were considerably smaller than diploid opaque cells (Fig. 2d). Importantly,
haploid opaque cells of opposite mating type were mating competent. Mating between
cells with complementary genetic markers resulted in the formation of heterozygous
diploids (Fig. 2e, Op x Op). In contrast, neither opaque cells of the same mating
type, nor white cells of opposite mating type, mated efficiently (Fig. 2e). This is
consistent with opposite mating types and the opaque state being prerequisites for
conventional mating
19,42
. Thus, we propose that haploids participate in a non-meiotic haploid-diploid parasexual
cycle.
Reduced growth and virulence of haploids
The products from haploid a and α mating were heterozygous for MTL (Fig. S7), near-diploid
in genome content (Fig. S7), and likely inherited the extra aneuploid copies of Chr6
and/or Chr7 from their MTLα parent. All the haploid isolates grew significantly slower
than SC5314 (Fig. 3a and S8, p < 0.001 for all haploids). However, the mating products
grew significantly faster than either of their haploid parents (Fig. 3a), as expected
if heterozygosity restores fitness through the complementation of recessive alleles
44
. Consistent with this, auto-diploids exhibited growth rates indistinguishable from
their haploid progenitors (Fig. 3a and S8), with the single exception of Haploid VIII
(p = 0.02). We propose that the low fitness of haploids, as well as their corresponding
auto-diploids, is a consequence of unmasking recessive alleles that reduce growth
potential. Furthermore, the variability in fitness between haploids is likely due
to the specific combination of alleles inherited by each haploid isolate.
Significantly, the diploid mating products from haploid crosses grew more slowly than
the highly heterozygous diploid SC5314. The generation of robust and diverse diploid
progeny from haploid mating would require outcrossing, which is predicted to occur
infrequently
11,12
, as many humans are colonized with a single strain of C. albicans. As it appears
that some lethal mutations have accumulated on one of the two homologs for more than
half the chromosomes (Fig. 1 and S5), mating between related haploids would not reestablish
complete genome heterozygosity. Thus, even though the formation of haploids potentially
eliminates lethal mutations, continued inbreeding between related haploids would perpetuate
homozygosity of most chromosomes and consequently would reduce overall fitness.
Consistent with reduced growth rates in vitro, Haploid II was avirulent in a mouse
model of systemic candidiasis (Fig. 3b) and was cleared from the mouse after ten days.
However, after 48 hours of infection with a haploid strain, cells could be recovered
(Fig. S9a) and the majority of these isolates remained haploid. Subsequent in vivo
experiments compared the colony forming units (CFUs) of haploids, auto-diploids and
SC5314 isolated from mouse kidneys 48 hours post-inoculation in individual and direct
competition experiments (Fig. 3c and S9b). The number of CFUs recovered from mice
inoculated with haploids and auto-diploids was several orders of magnitude lower than
the number of CFUs recovered from mice inoculated with SC5314 (Fig. 3c). Furthermore,
competition experiments indicated that haploids and auto-diploids exhibited similar
fitness in vivo, whereas both of these forms were outcompeted by heterozygous diploids
(Fig. S9b). Similar to growth rates in vitro, the low fitness of haploids and their
corresponding auto-diploids in vivo, indicates that it is not the diploid state, per
se, that is beneficial to growth but rather extensive allelic heterozygosity, as evidenced
with SC5314. From these data, we hypothesize that haploid formation may not be rare,
but that haploids likely represent a very small fraction of the overall population
due to their low competitive fitness relative to heterozygous diploids.
Haploid strains as genetic tools
The haploid state greatly facilitates experimental approaches such as classical genetic
screens for recessive alleles and a single-round of gene-knockout phenotyping. To
illustrate this potential, a set of auxotrophic C. albicans strains was derived from
a relatively stable haploid, GZY792 (Fig. 4a and S10) and genes important for morphogenesis
in diploids (HGC1, RVS167, SLA1, SEC3 and ACE2) were deleted in a single step. The
resulting mutants exhibited morphogenesis defects closely resembling those of the
corresponding diploid null mutants (Fig. 4c)
45–48
. Furthermore, because the deleted genes map to all eight chromosomes (Fig. 4b), the
ability to delete each gene in a single step confirms that all chromosomes were monosomic
in the parental haploid.
Concluding remarks
In summary, C. albicans can no longer be considered to be an obligate diploid. Rather,
it has the ability to form haploids that subsequently mate to form diploids or undergo
auto-diploidization. The lower fitness of haploids and auto-diploids suggests they
will not persist in the population and thus should be detected only rarely. Nonetheless,
efficient mating between viable haploids can produce heterozygous diploids that have
increased fitness, presumably due to the complementation of detrimental recessive
alleles. Furthermore, the reduction to a haploid state can serve as a vehicle to eliminate
recessive lethal alleles from a heterozygous diploid population. We propose that C.
albicans rapidly generates genetic diversity by producing a broad range of different
ploidy states, including haploid, diploid, tetraploid, and aneuploid
25
. Each of these ploidy states are mating competent and genetic outcrossing can introduce,
albeit infrequently, further genetic diversity into the population. While the ploidy
reduction mechanism(s) used by C. albicans remain an enigma, the discovery of a haploid
form and the potential for a haploid-diploid parasexual cycle significantly expands
our ability to manipulate, and thereby better understand, this opportunistic pathogen.
Furthermore, this study reveals how whole-genome analyses can lead to a re-evaluation
of common assumptions about genomic structure in microbial organisms.
Supplemental Methods
Haploid screening from the mouse models of infection
Bloodstream infections were performed by injecting 106 cells of parent strain AF7
(gal1Δ::URA3/GAL1), into the tail vein of 13 outbred ICR male mice (22–25 g, Harlan,
Indianapolis)
27
. When moribund (at 5–7 days), mice were anesthetized using isofluorane, euthanized,
and both kidneys were removed. Kidneys were combined, homogenized with 1 mL of water.
1:1000 dilutions of kidney homogenates were plated for total cell count onto YPD,
and 1:10 dilutions of the same homogenate were plated onto 2-DOG medium to obtain
Gal− colony counts at 3 days.
For the oropharyngeal model of infection, Balb/C mice were immunosuppressed with cortisone
on days −1, 1 and 3 of infection. Calcium alginate swabs were saturated with a suspension
of 106 cells/ml of YJB9318 (gal1Δ::URA3/GAL1) under their tongues for 75 min. Mice
were sacrificed at 1, 2, 3, and 5 days post-infection. 1:1000 dilutions of tongue
tissue homogenates were plated for total cell count onto YPD, and 1:10 dilutions of
the same homogenate were plated onto 2-DOG medium to obtain Gal− colony counts at
3 days. Importantly, to confirm that 2-DOG resistant cells only arose during in vivo
passage and were not selected for by plating on 2-DOG medium, a gal1/gal1 strain and
a gal1/GAL1 heterozygote strain were plated onto 2-DOG medium and observed for growth.
2-DOG resistant colonies grew up from the gal1/gal1 mutant but not from the GAL1/gal1
heterozygote by day 3 after plating
27
. Therefore, we used a cut-off of day 3 for picking 2-DOG resistant colonies after
plating.
Approximately 300 colonies were transferred from YPD plates and ~740 colonies from
2-DOG plates (312 from OPC and 431 BSI isolates) into 96-well plates containing 50%
glycerol and stored at −80°. All C. albicans cells isolated following in vivo passaging
were analyzed by flow cytometry to determine cell ploidy.
Flow cytometry preparation and analysis
Mid-log phase cells were harvested, washed and resuspended in 50:50 TE (50mM Tris
pH8 : 50mM EDTA) and fixed with 95% ethanol. Cells were washed with 50:50 TE and treated
with 1 mg/ml RNAse A and then 5 mg/ml Proteinase K. Cells were washed with 50:50 TE
and resuspended in SybrGreenI (1:85 dilution in 50:50 TE) incubated overnight at 4°.
Stained cells were collected and resuspended in 50:50 TE and analyzed using a FACScaliber.
Whole genome ploidy was estimated by fitting DNA content data with a multi-Gaussian
cell cycle model that assumes the G2 peak has twice the fluorescence of the G1 peak
and that minimizes S-phase cell contribution to the error function. Ploidy values
were calculated by comparing the ratio of peak locations in experimental samples to
those of diploid and tetraploid controls.
Mating Assays
Opaque or white cells were mixed together in equal cell numbers and incubated on Spider
media
34
for 18 hours prior to replica-plating onto SDC −Ade + Nat (to select for mating products),
as well as YPAD, YPAD + Nat, SDC −Ade and SDC −Ade + Nat to detect parental auxotrophies
and then photographed 24 hours later.
In vitro growth assays
Strains were grown in YPD media supplemented with adenine, uridine and histidine in
a 96-well microtiter plate and optical density was measured every 15 minutes with
a plate reader (Tecan Sunrise) for 24 hours. Doubling times were calculated as previously
described
30
.
In vivo assays
For virulence assays, eight mice per C. albicans strain were inoculated with ~6.0
× 105 CFUs by tail vein injection
49
. For survival assays, three mice per strain were inoculated by tail vein injection,
both kidneys were harvested at 48h and CFUs were determined by plating on YPAD. Ploidy
of randomly selected isolates was determined by flow cytometry.
Supplementary Material
1
2