|Home | About | Journals | Submit | Contact Us | Français|
This paper describes the serendipitous discovery and first characterization of a new resistant cell type from Dictyostelium, for which the name aspidocyte (from aspis: Greek for shield) is proposed. These cells are induced from amoebae by a range of toxins including heavy metals and antibiotics, and were first detected by their striking resistance to detergent lysis. Aspidocytes are separate, rounded or irregular-shaped cells, which are immotile but remain fully viable; once the toxic stress is removed, they revert to amoeboid cells within an hour. Induction takes a few hours and is completely blocked by the protein synthesis inhibitor cycloheximide. Aspidocytes lack a cell wall and their resistance to detergent lysis is active, requiring continued energy metabolism, and may be assisted by a complete cessation of endocytosis, as measured by uptake of the dye FM1-43. Microarray analysis shows that aspidocytes have a distinct pattern of gene expression, with a number of genes up-regulated that are predicted to be involved in lipid metabolism. Aspidocytes were initially detected in a hypersensitive mutant, in which the AMP deaminase gene is disrupted, suggesting that the inductive pathway involves AMP levels or metabolism. Since aspidocytes can also be induced from wild-type cells and are much more resistant than amoebae to a membrane-disrupting antibiotic, it is possible that they are an adaptation allowing Dictyostelium cells to survive a sudden onslaught of toxins in the wild.
The success of micro-organisms in resisting environmental toxins and antibiotics may be a major determinant of their overall success. For free-living organisms this resistance would be important in competition for ecological niches, and for pathogens in blunting the effect of antibiotic treatment. Strategies employed by bacteria include the expression of resistance genes, the formation of dormant spores (Errington, 2003) or protected biofilms (Hall-Stoodley et al., 2004; Hoffman et al., 2005) and the maintenance of a small population of slow-growing ‘persister’ cells, which can survive a sudden antibiotic onslaught (Balaban et al., 2004).
Dictyostelium amoebae probably face a similar challenge in the wild, since their natural habitat of soil contains a variety of antibiotic-producing bacteria and fungi and may also suffer from contamination with heavy metals and chemical residues. Social amoebae also compete with each other for resources (Horn, 1971; Ketcham et al., 1988), and their recently appreciated capacity for making polyketides (Austin et al., 2006; Eichinger et al., 2005; Serafimidis & Kay, 2005) suggests that some of this competition is mediated by amoebocidal chemicals (Horn, 1971) and defences against these chemicals.
The Dictyostelium discoideum genome reveals a wide array of potential resistance genes (Burlando et al., 2002; Eichinger et al., 2005) and this organism can also form physically resistant macrocysts and spores, encased in cellulosic coats (Kessin, 2001; Raper, 1984). However, macrocysts and spores only differentiate from starving cells, and both require the cooperation of hundreds to thousands of cells and around 24 h to form. Their major ecological function appears to be long-term survival and dispersal (and sexual recombination for macrocysts) rather than a short-term protection against toxins (Huss, 1989; Wallace & Raper, 1979). Microcysts are known from other species of social amoeba, but have not been reported from D. discoideum (Raper, 1984). The naked amoebae of this species are therefore a somewhat vulnerable state, for which some means of protection against chemicals and antibiotics would be advantageous.
The discovery is reported here of a new resistant cell type, the aspidocyte, by whose formation both growing and starving Dictyostelium cells may defend themselves against toxic attack.
D. discoideum Ax2 (Watts & Ashworth, 1970) was used as the wild-type strain in most experiments; the construction of mutant strains is described below. Strains were grown in axenic medium (Watts & Ashworth, 1970) supplemented with vitamins (0.1 mg cycancobalamin l−1, 0.02 mg biotin l−1, 0.2 mg riboflavin l−1) or in association with Klebsiella aerogenes (K. pneumoniae) on SM-agar plates (Kay, 1987). Growth and development were at 22°C. Dictyostelium firmibasis and Dictyostelium mucoroides were a kind gift from Dr Pauline Schaap, University of Dundee, UK.
In the standard method, cells were washed free of growth medium or bacteria by low-speed centrifugation from KK2 (16.5 mM KH2PO4, 3.9 mM K2HPO4, 2 mM MgSO4, pH 6.1) and resuspended at 106 ml−1 in spore medium: 10 mM MES, 20 mM KCl, 20 mM NaCl, 1 mM CaCl2, 1 mM MgCl2, pH 6.2, with 100 μg streptomycin ml−1 and distributed into tissue culture dishes (1.5 ml per 3.5 cm diameter dish). After 1 h, 2 or 5 mM cAMP and the compounds indicated in the text were added and incubation continued for 12–18 h, or as indicated in the text. Aspidocytes were scored following lysis of amoebae with 0.3% or 0.2% Cemulsol NP12 (Rhone Poulenc) for at least 4 min, either by phase-contrast microscopy, or by serial dilution onto SM-agar plates with bacteria, and subsequent plaque counting.
Aspidocytes were produced for microarray analysis using HM1158 (amdA) cells (see below) grown in axenic medium with phosphate buffer replaced by 20 mM MES pH 6.2 since phosphate is inimical to aspidocyte formation, and in experiments not shown, aspidocytes were also induced in axenic medium. After harvesting and washing, cells were resuspended at 5×106 ml−1 in spore medium, shaken at 180 r.p.m. and aspidocytes induced for 6 h with either 25 μM CdCl2/10 μM diethylstilboestrol (DES) (five biological replicates, 67% induction) or 250 μM thiabendazole (two biological replicates, 49% induction). Each biological replicate was accompanied by an uninduced control sample.
For electron microscopy, aspidocytes and control cells were produced in the standard conditions and fixed with 1% glutaraldehyde in 20 mM MES, 10 mM NaCl, 1 mM MgCl2, 1 mM CaCl2, pH 6.2 for 3 h on ice, followed by osmication and preparation for electron microscopy by standard methods, all performed by Dr Jeremy Skepper, University of Cambridge.
Confocal images and time-lapse films were taken using a Bio-Rad Radiance system mounted on a Nikon inverted microscope.
REMI mutagenesis (Kuspa & Loomis, 1992) of Ax2 cells used pBRS1 vector and DpnII introduced by electroporation. A pooled library of ~180000 independent transformants, selected with 10 μg Blastocidin ml−1, was subjected to three rounds of selection for mutants insensitive to the endogenous stalk inducer DIF. Cells at 106 ml−1 in spore medium plus 15 mM 8-Br-cAMP, 25 μM CdCl2 and 100 nM DIF-1 [1-(3,5-dichloro-2,6-dihydroxy-4-methoxyphenyl)-1-hexanone] (10 ml per 9 cm diameter dish) were incubated for 20 h, then 0.3% Cemulsol NP12 was added to lyse amoebae. Surviving spores (and aspidocytes) were spun down and grown up in bulk on bacteria for the next round of selection, or as clones, for examination. The site of REMI insertion was identified in strain HM556 as the DpnII restriction site 57 bases downstream of ATG of the coding sequence by inverse PCR (Keim et al., 2004) and DNA sequencing. A knockout vector, pIS11, was made from this strain by plasmid rescue following HindIII digestion of the nuclear DNA, and used to disrupt the gene in Ax2, creating strain HM1158. Strain HM1158 was transformed with the ABD-GFP reporter (Pang et al., 1998) by standard methods.
Total RNA was extracted using RNeasy kits (Qiagen) and 25 μg RNA per sample per microarray labelled by direct incorporation of Cy3- or Cy5-conjugated dCTP (Amersham) using reverse transcriptase (Superscript III, Invitrogen). Samples were paired and hybridized on custom microarrays, printed with ~9300 features in duplicate (http://www.sanger.ac.uk/PostGenomics/PathogenArrays/Dicty/protocols/). Five independent replicates of aspidocytes induced with CdCl2/ DES were compared directly with uninduced controls. Three of these replicates contributed one array each (two replicates in one dye-orientation, one in the other), while two contributed two arrays (of both dye-orientations), making seven arrays in total. The latter two replicates also formed part of a larger design comparing (A) control cells with (B) CdCl2/DES-induced aspidocytes, (C) thiabendazole-induced aspidocytes, and (D) the starting population of vegetative cells. Samples were compared in a loop of A with B, B with C, C with D, and D with A. Two independent replicates were carried out, each in both dye-orientations, giving a total of 16 arrays for this experiment.
Arrays were scanned using an Axon Instruments Genepix 4000B scanner and the resulting images quantified (Genepix 3.0, Axon Instruments) and analysed using the Bioconductor package Limma (Gentleman et al., 2004; Smyth, 2004, 2005). Background fluorescence was subtracted (Kooperberg et al., 2002), and the resulting log-ratios normalized by the print-tip loess function of Limma. In the larger design above, expression levels for each treatment were related back to uninduced cells in axenic medium by linear model fitting (using generalized least squares) and comparisons of interest then inferred. The significance of apparent differential expression was assessed by a Bayesian approach (Smyth, 2004), controlling the false discovery rate (Benjamini & Hochberg, 1995) at 5%. Gene identifiers given in the text refer to the annotated genome accessible at GeneDB (http://www.genedb.org/genedb/dicty/index.jsp) and Dictybase (http://dictybase.org/).
The dendrogram was based on the Spearman correlations of relative expression values for genes significantly different from starting vegetative cells in any other of the conditions. This filtering step was used to exclude noisy data from the analysis. The function spearman.dist in the Bioconductor package bioDist (B. Ding, R. Gentleman & V. Carey – bioDist: Different distance measures. R package version 1.2.0) was used to convert the correlation scores into a distance metric, which was used to generate a tree using phylip (Felsenstein, 2005).
The heatmap was generated using the heatmap.2 function of the R package gplots [Gregory R. Warnes. Includes R source code and/or documentation contributed by Ben Bolker and Thomas Lumley (2005). gplots: Various R programming tools for plotting data. R package version 2.1.1] using the euclidean distances between expression values and the ‘complete’ agglomeration method to order the genes.
Quantitative real-time RT-PCR was carried out using a Stratagene Mx3000P instrument and Invitrogen Superscript III Platinum SYBR Green One-Step qRT-PCR kit, according to the manufacturer's instructions, with the exception that reverse transcription was prolonged to 20 min. PCR cycling conditions were 95°C for 10 min followed by 40 cycles of 94°C for 30 s, 52 or 58°C (depending on primers) for 45 s, 68°C for 45 s. After this, products were briefly melted, annealed and remelted to obtain a dissociation curve. Only primers that resulted in a single peak were accepted. The threshold cycle (Ct) for each amplification was determined by the amplification-based threshold method, and baseline corrected by the adaptive baseline method of the Stratagene Mx3000P software. Primers (forward, reverse): DDB0167376 GCTATGACCCCCAAGGATTCACAAAATACAGC and GCTCTACAGGTCCGATATCATTCTCTTCGACT; DDB0167446 AGATTTTAGGTTGTCTTGATTCAG and AGAGATTGTAGAAATGACTTTTGC; DDB0216948 GTAGCTGGTAAACAATCAGAAACT and AGATAGGATTTGAAAAATCAGTTG; DDB0190107 GCTATGACCCCCGAATTTGGTTCCTGTAAAAA and GCTCTACAGGTCATTTACCTGGGAATACACAA; DDB0217770 GCTATGACCCCCTCAAGTGTTATCAAATGCAC and GCTCTACAGGTCTCCTCTGAAAGATCCAATAG; DDB0217855 GCTATGACCCCCCAAATCATTCTCTCCCTCTA and GCTCTACAGGTCTTACTACGGAAAAGATCACC; DDB0206063 GCTATGACCCCCGTCTTAAAGGAAGGTGTTGA and GCTCTACAGGTCTACTGGACGATAGTTAAGGG.
For each gene of interest a standard curve of Cts for four quantities of total RNA was produced in duplicate and a linear model fitted using the lm() function in R. Cts for two different quantities of both control RNA (cells starved in spore medium) and aspidocyte RNA (cells treated with Cd2+ and DES) in duplicate were read off the standard curve, giving four measurements of starting mRNA concentration for each gene in each RNA. The geometric mean ratio between aspidocyte and control RNA samples for each gene of interest was normalized by subtracting the geometric mean of ratios obtained similarly for three reference genes (for which there was little evidence of differential expression in the array data – Dictybase IDs: DDB0167376, DDB0217770, DDB0206063).
Aspidocytes were discovered by chance in a screen for mutants that cannot respond to DIF-1 (Morris et al., 1987), a signal molecule that induces stalk cell differentiation during Dictyostelium development. The screen was based on the ability of DIF-1 to suppress the alternative fate of spore formation in monolayer cultures (Kay, 1989), and hence unresponsive mutants were sought as those still able to form spores in the presence of DIF-1. Spores were selected at the end of the induction by adding a non-ionic detergent to lyse any remaining amoebae, stalk cells being already dead. We varied from previous work (Thompson et al., 2004) by adding a low concentration of Cd2+ ions to the culture to increase the overall efficiency of cell differentiation (Serafimidis & Kay, 2005), but by happenstance this led instead to the isolation of a mutant that forms aspidocytes (see below).
After three rounds of induction with DIF-1 followed by detergent selection, an unusual mutant was obtained from a library of insertional mutants. Cells of this mutant readily survived a detergent selection (0.3% Cemulsol NP12), but they did not form spores as expected, nor did they resemble vacuolated stalk cells: instead rounded, or irregular-shaped cells remained (Fig. 1A, BB).). Because of their resistance to detergent lysis, we named these cells aspidocytes (from aspis: Greek for shield).
As the original aspidocyte-forming strain was created by REMI mutagenesis (Kuspa & Loomis, 1992), the affected locus should be marked by the insertion of the REMI plasmid. We therefore identified this locus by rescue and sequencing of the flanking DNA (Keim et al., 2004) and found that that the disrupted gene (amdA) encoded the enzyme AMP deaminase. To confirm that the mutant phenotype is due to disruption of amdA, we redisrupted this gene by homologous recombination in the wild-type Ax2 background, giving strain HM1158, and also obtained a disruption in the Ax4 background from another laboratory (Chae et al., 2002). Both strains formed aspidocytes as readily as the original mutant (not shown) and as these strains were so similar, only HM1158 was used in subsequent experiments. Development of these strains is not normal: they produce small fruiting bodies and far fewer spores than the wild-type, as has already been described (Chae et al., 2002).
To understand the conditions necessary to induce aspidocytes, we first varied single components from the original medium in which they formed. This medium contained 8-Br-cAMP, which could be replaced by cAMP without affecting aspidocyte yield, thus giving the complete medium shown in Table 1. In this medium about 70% of HM1158 amoebae were converted to aspidocytes. For each condition aspidocytes remaining after detergent treatment were scored both by phase-contrast microscopy and by plating them clonally with bacteria for viability. Omission of components in turn from the medium showed that the key aspidocyte inducer is cadmium, and that DIF-1 acts synergistically with it (Table 1). cAMP has only a minor impact but was usually included in assays for convenience as its presence inhibits aggregation of the starving cells. The concentrations of Mg2+, Ca2+, K+ and Na+ were then independently varied and aspidocytes found to form quite efficiently in a wide range of ionic conditions, including the complete omission of divalent cations (not shown). The conditions of Table 1 are already close to optimal and were therefore retained.
A variety of potential aspidocyte inducers was then tested over a range of concentrations, both alone and in the presence of a low concentration of either DIF-1 or Cd2+ ions, to detect possible synergistic interactions (Table 2). Aspidocytes were scored as detergent-resistant cells and given as a percentage of input cells for convenience.
Inducers include: the heavy metal ions Cd2+ and Hg2+ (Co2+, Mn2+, Ni2+ and Zn2+ were much less effective or inactive at 0.2 mM; not shown); thiabendazole, a mitotic inhibitor in Dictyostelium (Welker & Williams, 1980); a group of azole anti-fungal agents that block sterol synthesis, represented by clotrimazole and econazole (Lupetti et al., 2002); sublethal doses of cycloheximide (protein synthesis) and actinomycin D (RNA synthesis) and G418, the aminoglycoside antibiotic widely used as a selective agent in Dictyostelium transformation (Nellen et al., 1984). In addition, the non-physiological stalk cell inducers DES, miconazole and zearalenone are strongly active and may work through inhibition of the plasma membrane proton pump (Gross et al., 1983, 1988; Kay et al., 1986; Pogge-von Strandmann et al., 1984).
Synergy is clearly evident between some combinations of inducers, such as 25 μM CdCl2 and 100 nM DIF-1, where summing their individual potencies would predict that the two compounds together would give less than 3% aspidocytes, whereas in fact they give 66% aspidocytes. DIF-1 is also strongly synergistic with low concentrations of cycloheximide, actinomycin D and G418, while 25 μM CdCl2 is only strongly synergistic with thiabendazole.
DIF-1 could be efficiently substituted in synergy experiments by the closely related DIF-2 and DIF-3, which also induce stalk cell differentiation (Masento et al., 1988), but not so well by its biosynthetic precursors THPH, Cl-THPH and dM-DIF-1 (Kay, 1998) which have greatly reduced stalk-cell-inducing activity (Table 3).
Aspidocytes can also be induced in conditions suitable for normal development: starving HM1158 cells plated on agar containing either 10 μM miconazole or 300 μM thiabendazole remain as an even lawn and after 24 h have formed more than 50% aspidocytes (not shown). Thus it appears that aspidocyte formation can override normal development.
Initially aspidocytes could only be efficiently induced in the amdA mutant strains, and the parental strain Ax2 gave less than 10% induction (Table 1). However, a combination of 25 μM Cd2+ and 10 μM DES worked well for starving Ax2 cells, giving 69±4.9% aspidocytes in the conditions of Table 1, and also allowed 10% aspidocyte induction in growth medium with the amdA mutant.
Aspidocytes, again defined as cells resistant to 0.3% Cemulsol NP12, can also be induced by this combination from starving cells of the wild-type isolates NC4 and V12M2 of D. discoideum, and from the related species D. firmibasis and D. mucoroides. All these strains produced approximately 5% aspidocytes, by viability after detergent treatment, using 50 μM CdCl2 and 10 μM DES in the conditions of Table 1 (not shown).
Aspidocyte formation takes one to several hours depending on the inducer (Fig. 2A, CC).). Time-lapse filming shows that the amoebae move actively for a while, but then gradually cease moving and forming pseudopodia, though cytoplasmic movements can still be seen. These cells remain fully viable, and if the inducers are washed away, they once again become detergent-sensitive (Fig. 2B) and many of them resume amoeboid movement within 1 h (see movies 1 and 2, available as supplementary data with the online version of this paper).
The slow evolution of detergent resistance as aspidocytes form contrasts with the rapid response of Dictyostelium cells to certain other stresses such as osmotic shock (Zischka et al., 1999) and suggests that changes in gene expression may be involved in aspidocyte formation. As a first test of this idea, we asked whether inhibitors of transcription and translation inhibit aspidocyte formation. Cycloheximide strongly inhibits protein synthesis in early development at the concentration used in Fig. 2C, but is not immediately lethal, as cells can recover even after 24 h of treatment; actinomycin D inhibits transcription, but mRNA production is only incompletely inhibited (Firtel et al., 1973; Mizukami & Iwabuchi, 1970). Fig. 2C shows that aspidocyte production is completely inhibited by cycloheximide and partially inhibited by actinomycin D, consistent with an essential requirement for protein synthesis and an important one for RNA synthesis.
We next used micro-arrays carrying 9300 probes to compare expression profiles of aspidocytes with their uninduced controls and with the starting vegetative cells. Induction was performed in suspension for 6 h without cyclic-AMP using either thiabendazole or CdCl2/DES.
Fig. 3A shows a Venn diagram of all the genes significantly (P<0.05) up- or down-regulated in control and thiabendazole- or CdCl2/DES-treated cells, compared to vegetative amoebae. Overall, the expression of 1076 genes changes significantly in one or more condition, compared to the starting cells. The majority of these genes are common to control and induced cells and presumably reflect their common response to starvation. However, a group of 157 genes changes significantly in both aspidocyte samples compared to their direct control, thus perhaps defining a core group of aspidocyte-regulated genes. This is illustrated by the dendrogram based on genewise Spearman correlations amongst the control and aspidocyte samples (Fig. 3B).
The genes differentially expressed in aspidocytes (P<0.05) compared to control cells are shown in the heat map (Fig. 3C) and listed in Supplementary Tables S1 and S2, available with the online version of this paper. These results were validated by RT-PCR in which two up-regulated and two down-regulated genes were compared using a group of three unregulated genes as the internal standard (Fig. 4). The ranking of the PCR results completely supports the trends apparent in the microarray data, which if anything appear to underestimate the magnitude of the changes; the overall (Pearson) correlation between the two sets of log2 ratios is 0.993.
The top 20 core genes significantly up-regulated in aspidocytes by CdCl2/DES and by thiabendazole are listed in Table 4 (for a more extensive list see Supplementary Table S1). They include a number whose predicted function is in lipid metabolism, including fatty acid synthase, phosphatidic acid phosphatase and an MBOAT acyltransferase, all suggesting possible membrane modification in aspidocytes. There is more overlap in the genes down-regulated in the two induction conditions (Supplementary Table S2), but their varied nature is less informative as to how aspidocyte metabolism differs from control cells.
The immediately striking feature of aspidocytes is their resistance to lysis by non-ionic detergents at concentrations that rapidly lyse amoebae. This resistance depends on the presence of 1–2 mM Mg2+ in the medium (not shown) and is not absolute, as aspidocytes, unlike spores, progressively lyse over a few hours in 0.3% Cemulsol NP12 (Fig. 5A). Aspidocytes are also more resistant than control amoebae to other non-ionic detergents such as Triton X-100 or Tween 20, but again they lyse over time.
One possible explanation for this detergent resistance would be if aspidocytes were able to form a cellulosic cell wall, similar to spores. However, staining for cellulose with Calcofluor showed that aspidocytes lack a cell wall (not shown) and this was confirmed by transmission electron microscopy. Under the electron microscope the plasma membrane of aspidocytes appears identical to that of control cells and the only ultrastructural difference noted was in the mitochondria, which tended to have more distinct cristae (not shown).
The cortical layer of F-actin adds mechanical strength to the plasma membrane, and we wondered whether modification of the actin cytoskeleton might be a feature of aspidocytes, possibly contributing to their detergent resistance. The F-actin cytoskeleton was visualized using a GFP fusion protein that specifically binds F-actin (Pang et al., 1998). Most aspidocytes have a continuous cortex of F-actin superficially similar to control cells. However, about 10% have a unique double layer of F-actin under the plasma membrane, which we have never observed in control cells (Fig. 1C).
Having eliminated the possibility of a cell wall, we considered the possibility that aspidocyte detergent resistance might be an active process. To test this idea we used azide or cooling to 0°C to block ATP production. In both cases, detergent rapidly lyses preformed aspidocytes (Fig. 5A), showing that their resistance is an active process. Control experiments show that azide alone does not lyse amoebae (Patterson & Spudich, 1995) or aspidocytes (not shown) in short incubations.
Resistance may also be linked to a complete cessation of endocytosis, which can be demonstrated using the membrane-partitioning dye FM1-43 (Aguado-Velasco & Bretscher, 1999). This rapidly stains the plasma membrane of aspidocytes and control cells, and is subsequently internalized in control cells by endocytosis to stain internal membranes. In aspidocytes there is no detectable internal staining even after 20 min incubation, showing that endocytosis is completely blocked (Fig. 1D). Similar results were obtained with the related dye FM4-64 (not shown).
Our working hypothesis is that the function of aspidocytes is to resist levels of environmental toxins that would normally overwhelm amoebae. If this idea is correct, then pre-formed aspidocytes should be able to survive an antibiotic treatment that would kill amoebae. We tested this idea using two toxins, which in preliminary experiments could kill Dictyostelium amoebae within 1 h (most of the toxins tested, such as cycloheximide, thiabendazole and daunomycin had little effect on cell viability in this time). Fig. 5B shows that aspidocytes are strongly resistant to killing by amphotericin B, a membrane-disrupting fungicide produced by streptomycetes (Hartsel & Bolard, 1996), whereas control cells are sensitive. In contrast, their resistance to bleocin, another streptomycete toxin, but one that damages DNA, is only modestly greater than control cells (not shown). Preformed aspidocytes therefore appear to be strongly resistant to at least one class of antibiotics.
Here we report the discovery of what appears to be a new, resistant cell type in Dictyostelium. Aspidocytes clearly differ from spores, stalk cells, macrocysts and microcysts in morphology and in their lack of a cellulosic cell wall. They differ from amoebae in their striking detergent resistance and their lack of movement and endocytosis, and they differ from cells undergoing non-vacuolar cell death in their prolonged viability (Huang et al., 2006; Kosta et al., 2004). Aspidocyte formation from amoebae requires protein synthesis and results in a distinct pattern of gene expression, all suggestive of a distinct cell type or state. Once induced, aspidocytes persist for hours without apparent change, but it should also be emphasized that, like the prestalk and prespore cells formed in development, aspidocyte formation is fully reversible: if the inducers are removed, the cells rapidly resume movement and are fully viable.
It is not surprising that aspidocytes have escaped detection in more than 50 years of laboratory study of Dictyostelium, given the unusual circumstances required for their initial discovery: a hypersensitive mutant, induction with Cd2+ ions and detection by detergent resistance. However, in appropriate conditions aspidocytes can be readily induced in wild-type D. discoideum and related social amoebae.
The detergent resistance of aspidocytes takes at least 1 h to appear, but is not due to the synthesis of a cell wall detectable by light or electron microscopy. It is likely that the lipid or protein composition of the plasma membrane is altered, and also that the cessation of endocytosis in aspidocytes increases their resistance to detergent by denying it access to the internal membranes of the endocytic pathway. However, a key feature of the detergent resistance is that it is not merely passive, but requires continued energy metabolism. The nature of this active process can only be speculated upon: one possibility is that aspidocytes are able to expel detergent molecules from their plasma membrane; another is that they somehow repair the lesions in the membrane caused by detergent.
Aspidocyte formation is most efficient from starving cells and is stimulated by a wide variety of agents, including heavy metals, such as cadmium, inhibitors of sterol, protein or RNA synthesis and of mitosis, and also the endogenous DIF signals and their partial mimics, DES, zearalenone (a phyto-oestrogen) and miconazole (Gross et al., 1983; Wang et al., 1990). In most cases, higher concentrations of the inducer are directly toxic to the cell, or at least block aspidocyte formation – thus for instance a high concentration of the protein synthesis inhibitor cycloheximide completely prevents aspidocyte formation.
DIF-1 at roughly physiological concentrations is a poor inducer of aspidocytes, but can synergize effectively with other inducers. This is surprising at first sight, since DIF-1 was discovered as a signal inducing stalk cell differentiation in later development (Morris et al., 1987). However, there is also evidence that DIF-1 can affect cells that have been starved for only a few hours, as in the aspidocyte inductions (Wurster & Kay, 1990), most notably by causing the rapid tyrosine phosphorylation and nuclear localization of STATc (Fukuzawa et al., 2001). This in turn may link DIF-1 into a stress response pathway, since STATc also mediates osmotic and other stress responses in these cells (Araki et al., 2003). Alternatively, DIF-1 may itself be causing mild cellular stress, as it is known to be a mitochondrial uncoupler at unphysiologically high concentrations (Shaulsky & Loomis, 1995).
This synergistic effect of DIF-1 and the wide variety of aspidocyte inducers suggests that several distinct stress pathways must have their inputs integrated to stimulate aspidocyte formation. An important clue as to the identity of one of these pathways comes from the gene disrupted in the hypersensitive mutant. This encodes the metabolic enzyme AMP deaminase, whose activity is one way of removing AMP from the cytoplasm. Activity of this enzyme increases strongly in early development (Jahngen & Rossomando, 1986) and amdA mutant cells contain no detectable AMP deaminase activity, resulting in higher AMP levels during starvation than in wild-type cells (Chae et al., 2002). This would in turn be expected to activate AMP-activated protein kinase, which is a known transducer of metabolic stress (Hardie & Hawley, 2001), and therefore a candidate for mediating aspidocyte induction
What is known about aspidocytes suggests that they are an adaptation to resist environmental toxins and antibiotics. They are induced by a wide range of such agents, are formed from separate cells either growing or starving, and are rendered resistant to attack on the plasma membrane by antibiotics such as amphotericin by some unknown mechanism. Once the toxic stress has passed, aspidocytes rapidly lose their resistance and resume amoeboid movement. Since aspidocyte formation overrides development into a fruiting body and is much quicker, aspidocytes seem best adapted to resist a sudden flux of toxins. Even if only a small proportion of the population were induced to become aspidocytes – as is the case with wild-type amoebae in our best current conditions – this would still confer a survival advantage compared to a strain that could produce none. Although Dictyostelium is strictly non-pathogenic, other free-living amoebae are opportunistic pathogens (Schuster & Visvesvara, 2004), and it will be interesting to discover whether any of them use a similar strategy to resist drug treatment.
Thanks for discussions with the Dictyostelium group; supported by core funding from the MRC and a Wellcome Trust grant for the Dictyostelium microarray facility.
Microarray data deposited at ArrayExpress (http://www.ebi.ac.uk/arrayexpress/); submission E-SGRP-5.
Two movies and two tables are available as supplementary data with the online version of this paper.