Bet-hedging antimicrobial strategies in macrophage phagosome acidification drive the dynamics of Cryptococcus neoformans intracellular escape mechanisms

The fungus Cryptococcus neoformans is a major human pathogen with a remarkable intracellular survival strategy that includes exiting macrophages through non-lytic exocytosis (Vomocytosis) and transferring between macrophages (Dragotcytosis) by a mechanism that involves sequential events of non-lytic exocytosis and phagocytosis. Vomocytosis and Dragotcytosis are fungal driven processes, but their triggers are not understood. We hypothesized that the dynamics of Dragotcytosis could inherit the stochasticity of phagolysosome acidification and that Dragotcytosis was triggered by fungal cell stress. Consistent with this view, fungal cells involved in Dragotcytosis reside in phagolysosomes characterized by low pH and/or high oxidative stress. Using fluorescent microscopy, qPCR, live cell video microscopy, and fungal growth assays we found that the that mitigating pH or oxidative stress reduced Dragotcytosis frequency, whereas ROS susceptible mutants of C. neoformans underwent Dragotcytosis more frequently. Dragotcytosis initiation was linked to phagolysosomal pH, oxidative stresses, and macrophage polarization state. Dragotcytosis manifested stochastic dynamics thus paralleling the dynamics of phagosomal acidification, which correlated with the inhospitality of phagolysosomes in differently polarized macrophages. Hence, randomness in phagosomal acidification randomly created a population of inhospitable phagosomes where fungal cell stress triggered stochastic C. neoformans non-lytic exocytosis dynamics to escape a non-permissive intracellular macrophage environment.


Introduction
Cryptococcus neoformans is a pathogenic yeast that can reside in the phagolysosome of host macrophages [1]. Macrophages are critical cells in the pathogenesis of cryptococcosis, being involved in the containment of, and extrapulmonary dissemination of infection [2]. Macrophages are also involved in the pathogenesis of latent infection where the organism can survive for a long time in granulomas. Hence, C. neoformans survival in macrophages is critical for both the persistence and dissemination of infection. The yeasts, inhaled from the environment, manage intracellular survival by modifying the phagolysosomal environment in their own favor. The capsule of C. neoformans is comprised of several subunits including glucuronic acid and glucuronoxylomannan (GXM) which act as weak acids capable of buffering the environment to pH~5 [3,4]. C. neoformans also produces urease which disrupts phagolysosomal acidification by breaking down urea into carbon dioxide and ammonia, a weak base [5]. Additionally, C. neoformans are capable of exiting host macrophages through lytic egress, nonlytic exocytosis (Vomocytosis) [6][7][8], or lateral cell-to-cell transfer (Dragotcytosis) [9].
We recently observed that macrophages use a bet hedging strategy in phagosomal acidification to maximize the antimicrobial properties of acidic pH for controlling the growth of ingested microbes [10]. By increasing the diversity of possible phagolysosomal pH, a population of macrophages will optimize its chances to inhibit pathogen growth based on pH alone. We also noticed that even small perturbations in this system could disrupt the bet-hedging strategy and that different pathogens employ various strategies to tip the odds in their own favor. C. neoformans interferes with the stochastic modulation of phagosomal pH by capsule buffering and urease activity [3,5]. Phagosomal pH varies with macrophage polarization such that M2 polarized macrophages are less hostile [10], in terms of pH, to ingested C. neoformans yeasts.
A fascinating aspect of the interaction with C. neoformans with macrophages is the phenomenon of fungal cell transfer between two macrophages, a process we have recently termed 'Dragotcytosis' [9]. Dragotcytosis results from coordinated sequential exocytosis and phagocytosis events between adjacent macrophages [9]. We also noted that Dragotcytosis is favorable to C. neoformans as macrophages with Dragotcytosis blockaded yield fewer colony forming units after 24 h than those who were not blockaded. However, Dragotcytosis is also potentially beneficial to the host cell if the continued fungal intracellular residence would result in more cellular damage. Non-lytic exocytosis events avoid lytic escape, resulting in fewer lysed host cells, a type of relationship observed in other amoeba ingested microbes like Mycobaterium marinum [11,12]. Finally, we note that Dragotcytosis is an active process, only occurring with live C. neoformans [6]. When taken together, these findings suggest Dragotcytosis is triggered by C. neoformans for a purpose beneficial to the yeast. However, it is unknown what specifically triggers Dragotcytosis or how transferring between macrophages confers a benefit Notably, previous observations have established that phagolysosomal pH can modulate non-lytic exocytosis frequency, supporting the idea that pH has an important and complex role in the regulation of this system [13,14].
In addition to phagolysosomal pH there are other stressors in the phagosome that could conceivably affect the rate of Dragotcytosis. The oxidative burst and the release of reactive oxygen species into the phagolysosome are also potentially important. Both ROS generation and nitric oxide synthase (NOS) activity are upregulated in M1 macrophages in response to phagocytosis [15][16][17]. NOS metabolizes arginine into nitric oxide and citrulline. Conversely, Arginase (Arg), upregulated in M2 macrophages) hydrolyzes arginine to ornithine and urea. Increased activity of either Arg or NOS will consume available arginine, resulting in lower efficacy of the other. Thus, M2 polarized macrophages have lower NOS activity and a less significant oxidative burst, which is thought to be one of the main reasons M2 macrophages are less effective at killing pathogens.
In this study, we analyzed the consequences of macrophage polarization and phagolysosomal stress on C. neoformans pathogenesis and outcome. We hypothesized that macrophages containing C. neoformans phagolysosomes that caused significant stress to the yeast by acidity and/or oxidative stress would be more likely to trigger Dragotcytosis events than other C. neoformans containing phagolysosomes. In this regard, we tested whether the more fungicidal environment in M1 macrophages was more likely to trigger Dragotcytosis than the more permissive intracellular environment of M2 macrophages. Our goal was evaluating the hypothesis that Dragotcytosis is a fungal driven mechanism for escape of a hostile environment. We found that both hostile pH and reactive oxygen species (ROS) in the phagolysosome drive Dragotcytosis and use modelling to show the potential increasing benefit of Dragotcytosis events as a function of resident phagolysosome hostility.

Ethics statement
Murine experiments were carried out in accordance with approved IACUC guidelines (protocol MO21H124). Mice were sacrificed using CO 2 asphyxiation.
20% L-929 cell conditioned supernatant, 0.1% beta-mercaptoethanol for 6 days at 37˚C and 9.5% CO 2 . BMDMs were used for experiments within 5 days after differentiation. BMDMs were activated with 0.5 ug/mL LPS and 10 ng/mL IFN-γ for M1 polarization or 20 ng/mL IL-4 for M2 polarization for 16 h prior to experiments.

Phagolysosomal pH measurement
Phagolysosomal pH was measured using ratiometric fluorescence imaging involving the use of pH-sensitive probe Oregon green 488 as described in prior studies [5]. Briefly, Oregon green 488 was first conjugated to mAb 18B7 using Oregon Green 488 Protein Labeling Kit (Thermo-Fisher). BMDMs were plated at a density of 1.25 × 10 5 cells/well on 24-well plate with 12 mm circular coverslip. Cells were activated with 0.5 μg/ml LPS and 100 U/ml IFN-γ or 20 ng/mL IL-4 as previously described at 37˚C in a 9.5% CO 2 atmosphere overnight. Prior to infection, 2 d old live H99, heat killed H99, or anti-mouse IgG coated polystyrene bead (3.75 × 10 6 cells or beads/ml) were incubated with 10 μg/ml Oregon green conjugated mAb 18B7 for 15 min. Macrophages were then incubated with Oregon green conjugated mAb 18B7-opsonized particles in 3.75 × 10 5 cryptococcal cells or beads per well. Cells were either centrifuged immediately at 350 x g for 1 min and incubated at 37˚C for 10 min or incubated at 4˚C for 30 min to synchronize ingestion and allow phagocytosis. Extracellular cryptococcal cells or beads were removed by washing three times with fresh medium, a step that prevents the occurrence of new phagocytic events. As an additional safeguard against new phagocytic events fresh media was supplemented with AlexaFluor 568 conjugated mAb 18B7 for 1 h to label extracellular particles. Samples were collected at their respective time points after phagocytosis by washing the coverslip twice with pre-warmed HBSS and placing it upside down on a MatTek petri dish (35 mm with 10 mm embedded coverslip well) with HBSS in the microwell. Images were taken by using Olympus AX70 microscopy with objective 40x at dual excitation 440 nm and 488 nm, and emission 520 nm. Images were analyzed using MetaFluor Fluorescence Ratio Imaging Software. Fluorescence intensities were used to determine the ratios of Ex488 nm/Ex440 nm that were converted to absolute pH values using a standard curve where the images are taken as above but intracellular pH of macrophages was equilibrated by adding 10 μM nigericin in pH buffer (140 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2, 5 mM glucose, and appropriate buffer � pH 5.0: acetate-acetic acid; pH 5.5-6.5: MES; � pH 7.0: HEPES. Desired pH values were adjusted using either 1 M KOH or 1 M HCl). The pH of buffers was adjusted at 3-7 using 0.5-pH unit increments.

Cryptococcal capsule measurements
Capsule measurements were acquired by measuring exclusion zones on India Ink slides and phase contrast microscopy. To determine differences between polarized macrophage incubations, C. neoformans were harvested after being ingested by macrophages for 24 h. Extracellular yeasts were first removed by washing the cells 3 times with 1 mL HBSS. Macrophages were lifted from their plates, centrifuged at 350 g for 10 min, and resuspended in 1 mL distilled H 2 O. Cells were then passed through a 27 ¾ gauge needle 10 times and left incubating for 20 total min to ensure lysis. After lysis C. neoformans were pelleted via centrifugation at 2300 g for 5 min and resuspended in 50 μL of PBS. Slides were prepared using 8 μL of cell mixture and 1.5 μL India ink, then imaged on an Olympus AX70 at 20x objective. Capsules and cell bodies were measured using a previously published measuring program [20].

NOS and arginase activity measurements
BMDMs were seeded at 10 6 cell/well in 6-well treated tissue culture plates and activated overnight for M0, M1, or M2 polarization as previously described. Prior to infection, 2 d old live H99 (10 6 cells/well) were incubated with 10 μg/ml mAb 18B7 for 10 min. Macrophages were then incubated with opsonized particles at MOI of 1. Cells were either centrifuged immediately at 350 g for 1 min or incubated at 4˚C for 30 min to synchronize ingestion and cultures were incubated at 37˚C for 10 min to allow phagocytosis. Extracellular cryptococcal cells were removed by washing three times with fresh medium, a step that prevents the occurrence of new phagocytic events. After 24 hours, cell supernatant was collected and the BMDMs were lysed with distilled water and 10 passages through a syringe with 23G needle. The supernatant was tested for NO levels using Greiss reagent kit (Millipore-Sigma G4410). Cell lysates were tested for arginase activity with arginase activity assay kit (Millipore-Sigma MAK112).

Cryptococcus neoformans viability and growth inhibition
BMDMs were seeded (10 6 cells/well) in 6-well tissue culture plates. The cells were activated overnight (16 h) with IFNγ (0.02 μg/mL) and (0.5 μg/mL) LPS for M1, IL-4 (20 ng/mL) for M2, or unstimulated for M0. The cells were then infected with a 2 d culture of C. neoformans opsonized with 18B7 (10 μg/mL) at an MOI of 1 for 2 or 24 h. At the respective timepoint, the supernatant was collected and total yeast cells as well as GFP positive yeast were counted by hemocytometer and fluorescence microscopy using an Olympus AX70. C. neoformans ingested by adherent macrophages were imaged on a Zeiss Axiovert 200M inverted scope in a live cell incubator chamber. Viability was calculated as the proportion of yeast cells expressing GFP-actin. A second measure of viability was acquired by staining collected yeast on a hemocytometer in 1:1 0.4% trypan blue solution (Sigma Aldrich) and enumerating total blue cells compared to total cells.
The collected, conditioned supernatant was then pelleted at 2300 x g for 5 min to remove debris and extracellular crypto. Supernantants were seeded with 5 x 10 4 cell / mL of C. neoformans and absorbance (600 nm) as well as fluorescence (GFP 488/510) was read at 0 and 24 h. Samples were prepared on glass slides with India Ink negative stain and imaged with an Olympus AX70. Growth was characterized by increase in OD600 and viability was measured as the proportion of GFP positive cells.

Modelling
To simulate the effect of Dragotcytosis on a population of C. neoformans in phagolysosomes we generated 10,000 hypothetical phagolysosomal pH values based on the distribution of observed phagolysosomal pH in M1 polarized bead containing macrophage phagolysosomes. Each value < 4 was replaced one time by randomly determining a new phagolysosomal pH from the same distribution to simulate a Dragotcytosis event from the initial macrophage to a random new one.

Data analysis for stochastic signatures
Discrimination of deterministic vs. stochastic dynamics was achieved using the previously characterized permutation spectrum test [21] and the methods outlined in our previous work [10]. In short, processed datasets are aligned in a vector and separated into subgroups by a scanning window of 4 units. Each of the overlapping four-unit segments are then assigned a value from 0 to 3 based on their relative size (the largest value being 3 and the least being 0) in what is referred to as an ordinal pattern. The frequencies for each pattern are calculated for the entire dataset. Determinism is characterized by observations of ordinal patterns that do not exist (forbidden patterns), whereas stochastic dynamics are characterized by every ordinal pattern existing at a non-zero frequency. In this work, the original data is the time interval between the initiation of each instance of host cell escape within an experiment.

qPCR
BMDMs were seeded at 10 6 cell/well in 6-well treated tissue culture plates and activated overnight for M0, M1, or M2 polarization as previously described. Prior to infection, 2 d old live H99 or inert beads (10 6 particles/well) were incubated with 10 μg/ml mAb 18B7 for 10 min. Macrophages were then incubated with opsonized particles at MOI 1. Cells were either centrifuged immediately at 350 x g for 1 min or incubated at 4˚C for 30 min to synchronize ingestion and cultures were incubated at 37˚C for 10 min to allow phagocytosis. Extracellular cryptococcal cells were removed by washing three times with fresh medium, a step that prevents the occurrence of new phagocytic events. After 24 hours, BMDMs were resuspended in TRIzol reagent and frozen at -80˚C. Total RNA was isolated from frozen cell cultures using the TRIzol reagent per the manufacturer's suggestions (InVitrogen, Carlsbad, CA). The RNA was precipitated in isopropanol and then subjected to additional purification using the RNeasy RNA isolation kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. RNA was reverse transcribed to generate cDNA using a First-Strand Synthesis kit (Amersham) and random hexamers as primers. The cDNA was then used as the template for quantitative PCR using an iCycler Thermal Cycler Real-Time PCR machine (Bio-Rad, Hercules, CA). The products of PCR amplification were detected with Syber green fluorescent dye and the relative expression of each gene of interest expressed with reference to that of glyceraldehyde phosphate dehydrogenase (GAPDH). The PCR products were analyzed on Tris-Acetate-EDTA agarose gels to confirm its correct size. Analysis was performed with commercially available and optimized Qiagen primers (Table 1).

Macrophage preference assay
To ascertain whether C. neoformans preferred residence in macrophages as a function of host cell polarization we carried out cell transfer experiments in macrophage populations consisting of mixed M0, M1, and M2 polarized cells. BMDMs were seeded at 3 x 10 4 cells per well in a 24 well tissue culture plate loaded with circular coverslips then activated overnight for M1

PLOS PATHOGENS
Macrophage phagosome acidification bet-hedging drives dynamics of cryptococcal escape mechanisms polarization as previously described. Three separate populations of BMDMs were seeded in dishes large enough to account for an additional 3 x 10 4 cells per well per condition and activated overnight for M0, M1, or M2 polarization as previously described. The initial population of M1 BMDMs on coverslips was infected with 6 x 10 4 18B7 opsonized C. neoformans yeasts per well for 1 h. The remaining three populations of uninfected BMDMs were stained with CellTracker CMFDA (Green). After the 1 h infection the three labeled populations of BMDMs were lifted with cellstripper and added to the unlabeled, infected M1 coverslips so that each coverslip had two populations of macrophages: 1. Unlabeled and infected M1 and 2. Labeled but uninfected M0, M1, or M2. The coverslips were then incubated for 24 h at 37˚C and 9.5% CO 2 for 24, 48, or 72 h. The cells were then fixed with 4% paraformaldehyde for 10 min at room temperature and coverslips were mounted on slides with ProLong Gold Antifade Mountant (Thermofisher) and imaged on an Olympus AX70. For each field of view the total number of each macrophage population was counted along with how many infected macrophages in each population. To correct for floating cells and ensure a proper count of total cells, a hemocytometer was used to enumerate the total number of floating cells in each sample. We assumed any floating cells were previously evenly distributed across the original sample and thus used the percent of floating cells to estimate the number of missing cells in any given sample using the formula below:

Modeling dragotcytosis dynamics
A population of 40,000 cells was generated consisting of 10,000 cells each of infected M1, uninfected M1, uninfected M2, and uninfected M0 macrophages. During each iteration of Dragotcytosis each infected cell had a random chance of experiencing Dragotcytosis or Vomocytosis based on observed frequencies in vitro. A Vomocytosed yeast was considered removed from the modeling experiment as we did not have an accurate way to model reuptake, and a Dragotcytosed yeast was randomly transferred to one of the four types of macrophage cells with no bias introduced. 10 rounds of sequential Dragotcytosis were modeled and the entire computational experiment was replicated 100 times.

PLOS PATHOGENS
Macrophage phagosome acidification bet-hedging drives dynamics of cryptococcal escape mechanisms

Statistical analysis
Specific statistical tests for each experiment are denoted in the figure legends with tests for multiple hypotheses. Graphical error bars indicate 95% confidence intervals of the respective measurement. When comparing frequency of host cell exit strategies, we compared mutant strains to the wild-type strain with a test of equal proportions and Bonferroni multiple hypothesis correction.

Macrophage polarization state affects frequency of dragotcytosis uncoupled from vomocytosis
Given our prior findings showing different acidification dynamics in differently polarized macrophages [10], we investigated whether macrophage polarization effected the frequency of Dragotcytosis. First, to ensure that our results were not confounded by differing population densities of macrophages resulting in higher or lower frequencies of events, we compared Dragotcytosis frequency to macrophage density and found that while the total number of Dragotcytosis events increases with total number of macrophages, as expected, there was no correlation between the frequency of events overall (S1 Fig). Vomocytosis and Dragotcytosis frequencies in macrophages infected with C. neoformans strain H99 correlated (Pearson correlation 0.41 with P = 0.055 and Spearman correlation 0.48 with P = 0.025), consistent with the notion that both processes are coupled (S1D Fig). Analysis of the pH distribution in cryptococcal phagosomes, M2 macrophages are more hospitable that M1 macrophages C. neoformans, having a higher proportion of macrophages at optimal growth pH (~5.5) and few phagolysosomes at inhibitory ranges (� 4) (Fig 1A). Using M0 macrophages as a baseline measure of Dragotcytosis frequency, we find that M1 polarization increases Dragotcytosis frequency while M2 polarization completely abrogates Dragotcytosis. (Figs 2 and S2). This result correlates the frequency of Dragotcytosis with the relative hospitality of each type of macrophage phagolysosome. When taken in context with the previously reported insight that Dragotcytosis is an active process which is beneficial for C. neoformans survival [9], these data suggest that Dragotcytosis is more likely to be initiated in hostile phagolysosomes.

Cryptococcus neoformans Capsule and cell body size are unaffected by polarization
The C. neoformans capsule is one of the most important virulence factors and determinants of infection outcome and a powerful modulator of phagolysosomal pH [3]. Capsule growth could modulate phagolysosomal pH by mechanically stressing the phagolysosomal membrane as it enlarges, causing the phagolysosome to become leaky [22]. Several aspects of cryptococcal pathology and phagolysosomal outcome are correlated to capsule size including buffering potential [3,22]. To investigate whether the changes in pH in differently polarized macrophages were due to capsule differences, we measured capsule sizes after infection of differently polarized macrophages. We found that the average and median capsule sizes did not appreciably differ between populations of C. neoformans ingested by differently polarized macrophages, nor did the average and median cell body sizes (S3 Fig). Furthermore, if phagolysosomal pH was modulated by capsule size, we would expect to observe a correlation between cell radius and phagolysosomal pH. We found no indication of a correlation between particle size and phagolysosomal pH, even if we began measuring clusters of multiple particles within a single phagolysosome (S4 Fig).

PLOS PATHOGENS
Macrophage phagosome acidification bet-hedging drives dynamics of cryptococcal escape mechanisms The dashed line at pH 4 represents the point at which pH inhibits C. neoformans replication while the gray area between pH 5 and 6 represents the optimal growth pH for C. neoformans. Hospitality of each population is estimated by the number of phagolysosomes within each of these regions. Data was gathered from 3 independent replicates with n of 1490, 499, and 1419 for M0, M1, and M2 respectively. B. Arg-1 activity of differently polarized BMDMs infected (gray) or uninfected (black) with C. neoformans. M2 have the highest overall activity, and activity is promoted with infection in M1 populations. Data was gathered from 9 independent experiments with an n of 9 each. C. NOS activity of differently polarized BMDMs infected (gray) or uninfected (black) with C. neoformans. M1 have the highest overall activity, and activity is decreased with infection in both M0 and M2 populations. Significance determined by 2-way ANOVA with Tukey's HSD comparisons. Data was gathered from 9 independent experiments with an n of 9 each. D.

PLOS PATHOGENS
Macrophage phagosome acidification bet-hedging drives dynamics of cryptococcal escape mechanisms

Macrophage polarization modulates oxidative response and affects C. neoformans viability
To determine whether the macrophage oxidative response varied between polarization states in a manner that would support our hypothesis on the triggers of Dragotcytosis frequency, we measured NOS and Arg1 activity between macrophage populations. M2 macrophages had the lowest NOS activity and the highest Arg1 activity, while M1 macrophages had the highest NOS activity with the lowest Arg1 activity. (Fig 1B and 1C). This data supports the notion that M2 macrophage phagolysosomes are more hospitable for C. neoformans. Additionally, by comparing enzymatic activity between macrophages infected or uninfected with C. neoformans we found that infection lowers NOS activity in M0 and M2 populations while increasing Arg1 activity in M1 populations (Fig 1B and 1C). Each of these changes would result in more hospitable phagolysosomes for C. neoformans. We used GFP expression behind an actin promoter as a surrogate for viability for internalized C. neoformans in macrophages. Whether or not GFP negative cells are truly "dead" is a philosophical question beyond the scope of this paper, but a cell that does not express actin would certainly be incapable of division. Thus, we explored viability in terms of fungistatic rather than fungicidal qualities. We found that, at two hours post infection, C. neoformans ingested by M1 or M0 macrophages were less viable than those in M2 macrophages (Fig 1D). Additionally, we investigated the viability of extracellular C. neoformans and found that at 24 h post infection extracellular yeasts cultured with M1 macrophages are still inhibited (Fig 1E  and 1F). This is consistent with prior reports that macrophages can inhibit extracellular C. neoformans cells [23,24]. We hypothesize this inhibition is due to the buildup of NO in the extracellular media. We found no extracellular crypto at 2 hours post infection consistent with our observations that exocytosis occurs hours after ingestion.
Finally, to investigate whether macrophage polarization was altered populationally during C. neoformans infection, we performed qPCR on RNA harvested from differently polarized macrophages 24 h after infection. We found that C. neoformans infection alone, in M0 macrophages, correlates with a populational skewing toward M2 polarization (Fig 1G). This effect is augmented in M2 macrophages activated beforehand with IL-4 suggesting that C. neoformans may actively skew macrophages toward the most hospitable state. We did not see such synergism when macrophages were stimulated with IFNγ and LPS, or when macrophages were not stimulated. It is unclear at this time how the polarization skewing is achieved. C. neoformans may actively encourage M2 macrophage polarization, possibly through secretion of intracellular proteins that affect macrophage function [25]. Alternatively, increased lysis frequency of M1 polarized macrophages could skew the population by simply removing M1 polarized macrophages (S2 Fig).

Phagolysosome pH is associated with dragotcytosis frequency
To probe whether phagosomal pH modulated Dragotcytosis frequency rather than other downstream effects of polarization, we altered phagosomal pH in M1 macrophages with various drugs. Chloroquine is a weak base that localizes to phagolysosomes and can be used to alkalize macrophage phagolysosomes in vitro, and at 6 μM buffers phagolysosomes to a pH optimal C. neoformans growth. Bafilomycin A1 is a V-ATPase inhibitor which prevents protons from being pumped into the organelle. Alkalization of M1 cryptococcal phagosomes with either drug abrogated Dragotcytosis (Fig 2), consistent with the notion that phagolysosomal acidification triggers lateral cell transfer. Interestingly, treatment with sub-inhibitory concentrations of Fluconazole (20 μg / μL) and Amphotericin B (0.5 μg / μL) also reduced the frequency of Dragotcytosis. Unlike chloroquine, bafilomycin, and 1400W, which do not affect

PLOS PATHOGENS
Macrophage phagosome acidification bet-hedging drives dynamics of cryptococcal escape mechanisms the viability of fungal cells at the concentration used, Amphotericin B and Fluconazole both inhibit C. neoformans. The antifungal drug effects thus support our previous finding that only live C. neoformans are capable of initiating Dragotcytosis (S5 Fig).

Disrupting macrophage anti-fungal function lowers dragotcytosis frequency
If Dragotcytosis is triggered by fungal cell stress, then further disrupting the macrophages' ability to inhibit C. neoformans would be expected to lower its overall frequency. We investigated this using two methodologies. First, we infected macrophages with C. neoformans induced to grow large capsules. The sheer size of cell as well as the anti-phagocytic properties of the capsule are known to inhibit macrophage uptake and promote fungal survival. Furthermore, the C. neoformans capsule was previously shown to have buffering potential and raise the phagolysosomal pH, and consistent with that observation we observed that cryptococcal cells with induced larger capsules had higher phagolysosomal pH than those with smaller capsules (S6 Fig) and acapsular C. neoformans resulting in lower phagolysosomal pH [10]. We found that capsule induced C. neoformans Dragotcytosed less frequently and Vomocytosed dramatically more often compared to wild-type, resulting in no overall change when comparing both types of events together (Fig 2A). Next, we infected macrophages with wild-type C. neoformans at 30 0 C, a temperature optimal for the fungus and suboptimal for the macrophage. We again found that Dragotcytosis was abrogated (Fig 2A). Taken together these results support the hypothesis that Dragotcytosis is induced when C. neoformans are stressed or overwhelmed by host macrophage defenses. Finally, we investigated whether melanized C. neoformans Dragotcytosed at different frequencies as melanin is known to protect from phagolysosomal stressors [26][27][28]. We found that melanized yeast underwent Dragotcytosis at a reduced rate (Fig 2A). We confirmed this finding with two alternative methods: using NOX2 knockout mice that cannot make NO and by inhibiting oxidative burst with 1400W. In both cases, Dragotcytosis frequencies were reduced (Fig 2A). Taken together, these data suggest that Dragotcytosis could be triggered by general cellular stress within the phagolysosome rather than pH specifically.

Oxidative stress modulates dragotcytosis frequency
To further probe whether oxidative stress had a role in triggering host cell escape, we investigated several mutant C. neoformans strains deficient in various oxidative stress mitigation pathways including: Superoxide Dismutases 1 and 2 (SOD1, SOD2), Catalases 1 and 3 (Cat1, Cat3), Thiorexodin 1 (TXN), and Glutathione Peroxidases 1 and 2 (GPx and GPx2, 19]. We first observed that the mutant library parental strain KN99α undergoes Dragotcytosis at a significantly reduced rate compared to strain H99 (P < 0.0001 via test of equal proportions, Fig  2). ΔSOD1, ΔCat1, ΔCat3, ΔTXN, ΔGPx, and ΔGPx2 strains of C. neoformans underwent Dragotcytosis at an increased frequency compared to the parental strain KN99α, while Dragotcytosis frequency in ΔSOD2 and ΔTXN2 remained unchanged (Figs 2B and S7). Furthermore, both supplementing an M1 skewed macrophage population with 1400W prior to infection with ΔSOD1 and infecting NOX2 knockout mice completely abrogated the increased Dragotcytosis phenotype (S8 Fig). Overall, these data support the notion that Dragotcytosis is triggered by a stressed yeast, and that increasing stress on the yeast results in increased Dragotcytosis.
To investigate whether the frequency of Dragotcytosis is increased significantly enough in these knockout mutant strains (ΔSDO1, ΔSOD2, ΔCat1, ΔCat3, ΔTXN, ΔTXN2, ΔGPx, and ΔGPx2) to completely offset their increased susceptibility to ROS, we measured total cell

PLOS PATHOGENS
Macrophage phagosome acidification bet-hedging drives dynamics of cryptococcal escape mechanisms counts and cell viability of intracellular and extracellular yeast during BMDM infection. We found that the total cell population was unchanged, but the mutant strains suffered reduced viability compared to wild type (S9 Fig).

Time to host exit initiation is stochastic for exocytosis events
If Vomocytosis and/or Dragotcytosis are triggered by hostile pH and final pH is determined stochastically from a normal distribution [10], then these host cell exocytosis events should also display stochastic dynamics resembling those of the upstream pH trigger. To explore whether the stochasticity of phagolysosome acidification is preserved in this system, we measured the time at which each process occurs throughout a series of videos of macrophages infected with C. neoformans. We found that while the time of exocytosis events was not normally distributed, these events were triggered stochastically with no forbidden ordinal patterns observed (Fig 3A and 3B). The absence of forbidden ordinals implies that the process is not deterministic, ruling out chaotic dynamics.

Simulation data suggests dragotcytosis at low ph benefits cryptococcal cells
We hypothesized that triggering Dragotcytosis, or non-lytic exocytosis, in phagolysosomes of low pH would benefit C. neoformans cells by lowering the total number of yeasts within inhibitory phagolysosomes. The argument for non-lytic exocytosis is intuitive, in that the yeast leaves the phagolysosome altogether and could only result in fewer inhibitory phagolysosomes. Dragotcytosis however is more nuanced as there is no obvious way for C. neoformans to know whether the new phagolysosome will be more or less hostile to the yeast. To explore this hypothesis, we modeled phagolysosomal pH distributions based on observed pH measurements of bead containing BMDM phagolysosomes and the established bet-hedging strategy of host macrophages in silico. We found that if a C. neoformans yeast were to trigger Dragotcytosis even just a single time if their original phagolysosome dropped to pH < 4, it would be enough to drop the proportion of inhibitory phagosomes from~15% to~1% (Fig 3C and 3D). The actual effect on C. neoformans survival would be even more pronounced since Dragotcytosis events are not limited to one event per yeast in vitro/vivo. We then investigated how beneficial Dragotcytosis would be at a range of pH cutoffs, assuming that an ingested C. neoformans would be equally likely to migrate into an M0, M1, or M2 macrophage. In the context of pH, the benefit of a Dragotcytosis event increases steadily the further away from an optimal pH the original phagolysosome is (Fig 3E).

Cryptococcus neoformans does not display an inherent bias for destination macrophages
While there is no obvious and apparent method for C. neoformans to specifically target M2 macrophages as the Dragotcytosis destination, such a system could still be possible. The donor and acceptor macrophages maintain physical contact during the Dragotcytosis process, and an immune synapse could transduce signals detectable by the ingested yeast. To investigate whether C. neoformans shows a preference during Dragotcytosis we compared the rates of transfer from infected M1 macrophages to uninfected M1 or M2 macrophages. If there is a preference, then we would expect to observe an increased amount of C. neoformans residing in M2 macrophages 24 HPI; if there were no preference, we would expect to see equal transfer into M1 and M2 macrophages. Instead, we see roughly an equivalent amount of transfer into M1 and M2 macrophages with slightly less transfer into M0 macrophages (Fig 4A and 4B). Based on a model of Dragotcytosis dynamics, we expect it would take approximately 4 to 5

PLOS PATHOGENS
Macrophage phagosome acidification bet-hedging drives dynamics of cryptococcal escape mechanisms sequential transfer events before we would be able to detect differences in population based on random chance alone (Fig 4C).

Proposed model for triggering dragotcytosis
Dragotcytosis requires live C. neoformans and its frequency is modulated by variables that effect stress on the fungal cell (Table 2). We synthesize these observations into a model whereby stress in the phagolysosome, resulting in fungal cell damage, triggers a program for

PLOS PATHOGENS
Macrophage phagosome acidification bet-hedging drives dynamics of cryptococcal escape mechanisms exiting the host cell. We know that exocytosis is associated with actin flashes [7,29] and posit that stressed cells activate a program that manipulates the host cell to promote a choreography of events resulting in transit from the phagolysosome to outside of the host cell (Fig 5). If these events lead to exocytosis at the cell membrane region in contact with an adjoining macrophage, then non-lytic exocytosis becomes Dragotcytosis. Given that non-lytic exocytosis has been observed in mammalian, fish, insect, and protozoal cells the exit program is likely to target conserved functions in eukaryotic phagocytic cells [6,8,[30][31][32]. As to whether apposition of

PLOS PATHOGENS
Macrophage phagosome acidification bet-hedging drives dynamics of cryptococcal escape mechanisms two macrophages induces changes in the infected cell that are sensed by the fungal cell is unknown and a subject for future investigation.

Discussion
Dragotcytosis is a cellular process by which C. neoformans yeasts within a host macrophage phagolysosome transfer to another, proximal macrophage without lysis of the initial host [6,8,9,13]. During recent investigations into the receptors used in Dragotcytosis and into the phagolysosomal acidification dynamics of macrophages, we found a correlation between phagolysosomal hospitality and the frequency of Dragotcytosis, suggesting that Dragotcytosis is beneficial to, and triggered by, ingested C. neoformans [9,10]. Arguably, Dragotcytosis also benefits the macrophage by relieving it of a fungal cell that causes cellular damage, although the net benefit versus costs of this transfer event for the microbe and host cell remains uncertain. The fact that Dragotcytosis was initiated by the fungal cell led us to hypothesize that fungal discomfort in the phagosome was a trigger for cell-to-cell transfer and in this study, we demonstrate that cellular stress on C. neoformans leads to increased Dragotcytosis frequency.
Previous studies showed that M2 macrophages and Th2 skewed immune responses are more permissive to C. neoformans infection [33][34][35][36]. Our initial observation that M2 macrophages have, on average, higher pH phagolysosomes compared to M1 macrophages when containing inert particles offers a partial explanation of why these macrophages are more hospitable. Specifically, M2 macrophage populations had the fewest phagolysosomes at an inhibitory pH (pH < 4) and the most phagolysosomes in an optimal pH range (5 < pH < 6) for fungal growth [22]. Additionally, we noticed that C. neoformans underwent fewer

PLOS PATHOGENS
Macrophage phagosome acidification bet-hedging drives dynamics of cryptococcal escape mechanisms Dragotcytosis events when residing in M2 macrophage phagolysosomes compared to yeast cells resident in M1 macrophages. From this correlation, we hypothesized that if the phagolysosomal pH dropped too low for an ingested C. neoformans to counter acidity with its polysaccharide capsule or urease activity [3,5,37,38], then it triggered either non-lytic exocytosis or Dragotcytosis as an escape mechanism that would bring the yeast cell to the less acidic extracellular environment or into a phagolysosome of more likely to have a hospitable pH, respectively. Since phagolysosomal pH is stochastically determined with a distribution in which most phagosomes have pH > 4, it is likely that C. neoformans residing in a phagosome with pH < 4 would find a more hospitable home in another macrophage after a Dragotcytosis event [10]. Consequently, we probed how Dragotcytosis could modulate the effectiveness of the macrophage acidification bet hedging strategy via in silico modeling and found that even a single Dragotcytosis event triggered for C. neoformans residing in low pH phagolysosomes would significantly increase the likelihood of finding a less acidic phagosome. These findings could also help explain how C. neoformans replicates faster inside macrophages in vivo and in vitro than in the extracellular space, which has been attributed to faster replication in the acidic pH of the average phagosome [39][40][41]. Despite not all phagosomes being conducive to faster fungal growth, sequential Dragotcytosis events could lead C. neoformans to find phagolysosomes more permissive to rapid growth during an infection. These results fit well with the known data of C. neoformans infection in human macrophages. Human macrophage phagolysosomes acidify with different dynamics than those of mice. For example, human M2 macrophage phagolysosomes acidify to a lower pH than mouse macrophage phagolysosomes [42] and, perhaps as a result, humans are markedly resistant to C. neoformans infection which poses little threat to most immune competent hosts.

PLOS PATHOGENS
Macrophage phagosome acidification bet-hedging drives dynamics of cryptococcal escape mechanisms To probe deeper into whether there was a causal association between average phagolysosomal pH of M1 macrophages and the frequency of Dragotcytosis, we treated macrophages with either chloroquine or bafilomycin A1, which increased phagolysosomal pH near the optimal growth range of C. neoformans and comparable to the observed average phagolysosomal pH of M2 macrophage populations. We found that both chloroquine and bafilomycin A1 drastically reduced Dragotcytosis events, supporting the hypothesis that cellular stress is the stimulus for Dragotcytosis. We are confident that the drug treatments themselves did not have a direct effect on C. neoformans growth, virulence, or survivability at the concentrations used but rather modulated the host macrophage response. Chloroquine does not significantly reduce C. neoformans growth or viability unless ingested by macrophages and at concentrations higher than used in these experiments [43]. While high concentrations Bafilomycin A1 has antifungal [44] and even inhibits C. neoformans melanization [45], the concentrations used here are much lower than those associated with antifungal effects [46]. We also treated cells with subinhibitory concentrations of fluconazole and amphotericin B, to explore whether these drugs would also stress the yeast in a way that promotes Dragotcytosis [47]. Interestingly, both antifungal drugs reduced the overall frequency of Dragotcytosis possibly because the concentrations were still too disruptive to fungal metabolism, as C. neoformans growth is inhibited by both drugs, and the yeast was too compromised to transfer. In this regard, both drugs affect fungal membrane ergosterol integrity and fungal lipids may be required for the Dragotcytosis process. Alternatively, ergosterol disruption may not be sensed by the fungal machinery that triggers Dragotcytosis.
To further investigate the oxidative trigger hypothesis, we characterized Dragotcytosis frequencies of several C. neoformans strain KN99α mutants from a mutant knockout library during infection. We first observed that KN99α undergoes Dragotcytosis at a reduced rate compared to H99. We do not know the specific genetic or phenotypic changes responsible for the baseline reduced Dragotcytosis rate but it is not unexpected as KN99α is considered a hypervirulent strain due to multiple factors [48][49][50]. We selected a variety of potential candidates based on known pathways relevant to our cellular stress hypothesis. We found that SOD1, Cat1, Cat3, TXN, GPx, and GPx2 deficient mutants underwent Dragotcytosis much more frequently than the wild-type parental strain, supporting the notion that Dragotcytosis is triggered by the accumulation of oxidative stress. Interestingly, the ΔSOD2 and ΔTXN2 mutants did not exhibit increased frequency of Dragotcytosis which may reflect the nature of the phagolysosomal oxidative stress, as SOD1 and SOD2 are known to have different efficacy against different oxidative stresses [51].
If oxidative stress triggers Dragotcytosis, then reducing stress within the phagolysosome should reduce the frequency of Dragotcytosis. Melanin is known to protect C. neoformans from oxidative stress [52]. Melanized C. neoformans triggered Dragotcytosis at a lower frequency compared to non-melanized C. neoformans. Similarly, fewer Dragotcytosis events were observed in infections of NOX2 knockout mice incapable of oxidative burst and in wild-type macrophages treated with the NOS inhibitor 1400W. These conditions also restored Dragotcytosis frequency in the ΔSOD1 strain to the same frequency as wild-type KN99α. Finally, the cryptococcal capsule is known to protect against both pH and ROS in macrophages. Macrophages infected with C. neoformans with large capsules manifested Dragotcytosis at a reduced rate, though Vomocytosis occurred at an increased rate. It is less clear whether Vomocytosis is triggered by the yeast or host as heat killed C. neoformans have been observed to undergo Vomocytosis, albeit at a drastically reduced rate [7], raising the question of what could trigger one pathway opposed to the other. When macrophage infection was done at a temperature optimal for C. neoformans and suboptimal for macrophages: 30 0 C we again found that Dragotcytosis frequency was reduced.
Following our observations that increasingly stressful conditions increased Dragotcytosis, and that increased Dragotcytosis is associated with C. neoformans survival during these in vitro infections, we investigated whether this egress from hostile phagolysosomes would offset the increased ROS damage sustained in the ROS knockout mutant library. Thus, we measured intracellular and extracellular total cell counts and cell viability of all ROS knockout mutants used in the study. We found that the overall number of yeasts did not significantly change after 24 h infection, but the viability of each ROS mutant was significantly decreased compared to the parental strain after ingestion. Extracellular yeasts were more variable in their viability with ΔSOD2 and ΔGPx suffering no loss in viability.
Next, we considered whether C. neoformans induced a bias toward M2 macrophages as acceptor macrophages in the Dragotcytosis process. Logically it would benefit the yeast as M2 macrophages were found to be more permissive, but a mechanism by which the yeast could sense interaction with an M2 macrophage and initiate the Dragotcytosis process all from within the phagolysosome is difficult to imagine. Thus, we designed an experiment in which infected M1 macrophages would be seeded with labelled and uninfected M1 or M2 macrophages and we would compare the rate of transfer. If there was a preference, we should see an unequal rate of transfer between the two populations. Instead, we observed equal rates of transfer into both macrophage types. We expect that, given enough time, the distribution would shift to favor M2 macrophages by random chance. However, this was not observed due to the infrequency of sequential macrophage-to-macrophage transfers by a single cryptococcal cell during the first 24 hours of experimental observation. Thus, we find it more likely that C. neoformans initiates a transfer without prior knowledge of the polarization state of the receiver macrophage and instead continues to transfer until a hospitable phagolysosome is discovered by chance.
Dragotcytosis was stochastically initiated, reflecting the dynamics of phagolysosome acidification [10]. However, unlike phagosomal acidification, which manifested a normal distribution, the distribution of Dragotcytosis as a function of time was skewed away from normality. Therefore, while phagolysosomal pH contributed to triggering Dragotcytosis, it is likely not the only important trigger, a hypothesis supported by previous observations that inhibiting pH buffering via urease mutation did not increase Dragotcytosis events as would be expected if the process was triggered by pH alone [5]. Instead, Dragotcytosis may be triggered by an accumulation of diverse sources of cellular stress encountered in the phagolysosome. Whereas the endpoint of phagolysosomal acidification represents a bet hedging strategy by macrophages to control ingested microbes [10], C. neoformans also appears to be employing a bet hedging strategy in exocytosis that is triggered by the bet-hedging strategy in macrophage phagolysosome acidification. In fact, C. neoformans may engage in its own bet hedging strategy wherein Dragotcytosis is triggered past a certain threshold of stress in which it becomes more likely that the ingested yeast will migrate to a more hospitable phagolysosome. This results in the interesting situation whereby a defensive strategy by the host cells is co-opted for survival by the fungal cell. Hence, chance outcomes in the phagolysosome trigger events that produce chance outcomes in macrophage exocytosis, implying connectivity between these two cellular processes with regards to their system dynamics. Stochastic processes are random and thus inherently unpredictable, suggesting a fundamental unpredictability to cellular processes that could extend to making host-microbe interactions not predictable.
We previously proposed that Dragotcytosis was the result of sequential Vomocytosis and phagocytosis events on cell adjacent macrophages [9]. If this is the case, then the frequency of Vomocytosis events should correlate with that of Dragotcytosis events as the frequency of any non-lytic escape is determined by overall cellular stress but the decision to undergo Dragotcytosis as opposed to Vomocytosis would rely on additional, more complex, parameters.
Consistent with this thesis, we observed a correlation between Dragotcytosis and Vomocytosis frequencies.
In summary, our results consistently show that conditions that increase or decrease C. neoformans stress are associated with enhanced and reduced Dragotcytosis, respectively. ROS stress appears to be a particularly potent regulator of Dragotcytosis, but it may be impossible to narrow down Dragotcytosis triggers to one specific stressor. Drugs that inhibit phagosome acidification can also inhibit autophagy, and vice versa, making it difficult to parse out the effects of only one of these systems at a time. Even the generation of reactive oxygen species is innately tied to these processes, and phagolysosomal pH itself, with the concentration of ROS in the phagolysosome increasing alongside pH with previously measured concentrations of 50 μM O 2 •at pH 7.4 and 2 μM at pH 4.5 [15]. We find it more likely that a combination of cellular stresses triggers Dragotcytosis. We note how the normally distributed phagolysosomal pH following phagocytosis randomly includes some phagolysosomes that are inhospitable to fungal cells which, in turn, triggers exocytosis with its own stochastic dynamics with a different distribution from normally distributed phagolysosome pHs [10], as other factors contribute to the exit phenomenon and skew the distribution away from normality. Hence, stochastic dynamics in phagosomal acidification beget stochastic dynamics in non-lytic exocytosis, implying a fundamentally unpredictable host-microbe interaction at the level of cellular organelles. Such unpredictability at the cellular level is a likely contributor to our inability to predict the outcome of host-microbe interactions at the organismal level [53].
Supporting information S1 Fig. Frequency