Physarum polycephalum is a lower eukaryote belonging to the amoebozoa group of organisms that forms macroscopic, multinucleate plasmodial cells during its developmental cycle. Plasmodia can exit proliferative growth and differentiate by forming fruiting bodies containing mononucleate, haploid spores. This process, called sporulation, is controlled by starvation and visible light. To genetically dissect the regulatory control of the commitment to sporulation, we have isolated plasmodial mutants that are altered in the photocontrol of sporulation in a phenotypic screen of N-ethyl-N-nitrosourea (ENU) mutagenized cells. Several non-sporulating mutants were analyzed by measuring the light-induced change in the expression pattern of a set of 35 genes using GeXP multiplex reverse transcription–polymerase chain reaction with RNA isolated from individual plasmodial cells. Mutants showed altered patterns of differentially regulated genes in response to light stimulation. Some genes clearly displayed pairwise correlation in terms of their expression level as measured in individual plasmodial cells. The pattern of pairwise correlation differed in various mutants, suggesting that different upstream regulators were disabled in the different mutants. We propose that patterns of pairwise correlation in gene expression might be useful to infer the underlying gene regulatory network.
Physarum polycephalum is an acellular slime mold that phylogenetically belongs to the amoebozoa group of organisms (Baldauf & Doolittle 1997; Baldauf et al. 2000; Barrantes et al. 2010). Physarum can differentiate into various structurally, functionally, and biochemically different cell types that occur in temporal order in the course of the branched life cycle (Burland et al. 1993). At one stage of the life cycle, haploid, mononucleate amoebae, also called myxamoebae, hatch from mature spores. Myxamoebae feed on bacteria and propagate by mitosis. Two amoebae of different mating types may mate with each other to form a diploid zygote. The zygote develops into a diploid multinucleate plasmodium, which grows as long as sufficient nutrients are available. During growth, the protoplasmic mass continually increases and the nuclei suspended in the vigorously streaming protoplasm synchronously divide by undergoing a closed mitosis, while the plasmodial cell does not divide. A plasmodium covering a 9 cm diameter Petri dish, for example, then contains between 107 and 108 nuclei, which are still naturally synchronous in cell cycle progression and during differentiation (Rusch et al. 1966; Sachsenmaier et al. 1972; Werenskiold et al. 1988).
In response to starvation, the plasmodium has two options of development to exit from growth and proliferation by forming dormant, drought-resistant cell types. The developmental decisions are under sensory control and involve the differential regulation of developmental genes (Putzer et al. 1983, 1984; Martel et al. 1988). The plasmodium may either form macro-cysts also called sclerotia or it may sporulate. During sporulation the entire protoplasmic mass synchronously develops into fruiting bodies. In the fruiting bodies the protoplasm is cleaved, the nuclei undergo meiosis and give rise to mononucleate, haploid spores.
Experimentally, sporulation can be triggered with a pulse of far-red light, which is sensed by a phytochrome-like photoreceptor (Starostzik & Marwan 1995a; Lamparter & Marwan 2001). Approximately 4–6 h after the inductive light pulse, the plasmodium has crossed the point of no return to become irreversibly committed to sporulation. Before commitment, the plasmodium can return to the proliferative state when it is supplemented with glucose (Starostzik & Marwan 1994, 1995b). The transcriptomic changes that occur in committed cells have been analyzed by massive parallel sequencing (Barrantes et al. 2010).
The Mendelian genetics of Physarum is well established (for review see [Dee 1987]). By treatment of mononucleate, haploid amoebal cells with UV or chemical mutagens, mutants can be obtained, crossed with wild type amoebae and, after meiotic recombination in a sporulating plasmodium, segregants can be screened for phenotypes and genotypes (Schedl & Dove 1982; Dee 1987). Massive parallel sequencing combined with polymerase chain reaction (PCR)-based techniques would allow modern genetic analyses to be performed at the molecular level.
The isolation and characterization of mutants is greatly facilitated by a temperature-sensitive allele, which is closely linked to the mating type locus matA, and which causes apogamic development (Dee 1966, 1987; Wheals 1970; Mohberg 1977). By apogamic development, a haploid amoebal cell can directly transform into a haploid plasmodium without the need of mating. This provides the practical advantage that plasmodia arising from amoebal mutant clones can be directly screened for phenotypes, while crossing of the mutants with other strains and subsequent Mendelian analysis is still possible at the temperature that blocks apogamic development.
Here, we isolate mutants in the photocontrol of sporulation and show that these mutants display altered gene expression patterns in response to light and altered patterns in the pairwise correlation of differentially expressed genes.
Materials and methods
Screen for mutants altered in the photocontrol of sporulation
Amoebae of the apogamic wild-type strain WT31 were mutagenized with N-ethyl-N-nitrosourea (ENU) as described (Sujatha et al. 2005). After mutagenesis, cells were allowed to undergo several cell divisions and after forming a confluent amoebal layer, develop into cysts. The cysts were used to prepare glycerol stocks of the mutant library. All manipulations were performed under sterile conditions (Sujatha et al. 2005).
Aliquots from a glycerol stock of a mutant library were thawed and diluted 100- to 1000-fold in order to obtain approximately 20–50 amoebal colonies per plate. The cyst suspension was examined under the microscope without a cover slip to verify that cysts were suspended individually and did not form any clumps. Before use, microscope slides were thoroughly cleaned with ethanol to remove any detergent from the surface that might dissociate potential clumps of cysts during sample preparation. Avoiding clump formation is essential in order to make sure that each amoebal colony emerges as a clone of a single cell. 100 μL of the diluted amoebal cyst suspension were mixed on a plate with 100 μL of a standardized formalin-killed E. coli suspension that served as a food source for the amoebae (FKB; [Dee 1986]) and the droplet was spread evenly over a DSPB agar plate (diluted semidefined phosphate buffered medium; [Anderson et al. 1997]) by streaking with the conical tip of a 10-mL glass pipette. The plates were incubated at 24°C for 48 h to facilitate hatching of amoebae from cysts and subsequently at 30°C to allow amoebal growth while suppressing the unwanted formation of haploid plasmodia by apogamic development (Bailey et al. 1987, 1992; Bailey 1995). After several days at 30°C, amoebal colonies became visible. Well separated individual colonies were picked by suspending them in a droplet of 10 μL of FKB by pipetting up and down several times. The droplet with the picked cells was split into two parts each mixed with a droplet of 100 μL FKB that had been placed on a DSPB as well as on a small (diameter = 4.5 cm) DSDM (diluted semidefined medium; [Dee et al. 1997]) agar plate. The suspensions were spread over the plates to obtain even amoebal growth.
The DSPB plate was incubated for 2 days at 24°C and subsequently for 10 days at 30°C to allow amoebal growth and encystment. Each plate with encysted amoebae was flooded with 10 mL of a 15% (v/v) aqueous glycerol solution. The cysts were suspended with the help of a 10-mL glass pipet and a 1-mL aliquot was frozen at –80°C to preserve the amoebal clone.
The DSDM plate was incubated at 24°C for several days to allow amoebal growth and the apogamic development of small plasmodia, which were inoculated on a SDM plate (semidefined medium; [Dee et al. 1997]), to obtain a fully grown macro-plasmodium from which further cultures were inoculated.
Each plasmodial clone was then inoculated onto one SDM plate covered with a sheet of filter paper (diameter = 70 mm, 85 g/m2; Macherey Nagel 640 m, supplied by Carl Roth, Germany). Each plate was incubated for several days at 24°C until it was completely covered by a fully grown plasmodium. The filter paper with the plasmodium was transferred to a sporulation agar plate for starvation at 22°C in complete darkness. The incubation temperature is critical to avoid light-independent sporulation in the wild type. After 6 days of starvation in the dark, plasmodia were examined for spontaneous (i.e. light-independent) sporulation under dim green safe light to identify potential mutants that sporulate spontaneously. Plasmodia that had not sporulated after starvation (the vast majority) were irradiated for 30 min with far-red light (≥700 nm; 13 W/m2) as described (Starostzik & Marwan 1998). Those plasmodia that had not sporulated on the next day were immediately exposed to a second 30-min pulse of far-red light, and eventually to a third one if the second pulse did not cause sporulation within one day. The screening procedure (with potentially three subsequent light exposures, one exposure each day) was repeated twice for all clones. This screening was found to be selective enough to identify potential mutants in the photocontrol of sporulation that cause a tight phenotype.
To verify the phenotypes and the stability of potential mutants, plasmodia were regrown from glycerol stocks of amoebal clones. From each candidate strain, 10 amoebal subclones were picked and allowed to apogamically develop into plasmodia. Five plasmodia of each subclone were used to verify the phenotype as described for the screening procedure. After successful verification, a plasmodial strain developed from one subclone of each identified mutant strain was taken into liquid shaken culture for the generation of spherules and for further phenotypic characterization. Cultures on plates were incubated in Memmert ICP 800 incubators (Memmert, Schwabach, Germany). As the incubation temperatures are critical, they were verified using a calibrated mercury thermometer with 0.1°C resolution. The incubators were supplied with 0.5 volumes of air per hour, which was dehumidified by passing the airflow through a reflux condenser with steel-made cooling coil thermostated at 2°C. Supply with dry air contributes to the reproducible culture conditions and helps to avoid that plasmodia creep out of the culture dishes at high humidity.
Growth and preparation of sporulation-competent plasmodia for gene expression studies
A 5-L fermenter (Minifors, Infors HT, Bottmingen, Switzerland) with 1.5 L of growth medium (Daniel & Baldwin 1964) was inoculated with 2 vol % of a 3.5-day-old shaken culture. Plasmodia were grown for 4 days at 24°C, supplied with 1 L of air per min, and stirred at 250 rpm with a marine propeller. Microplasmodia were harvested, washed twice with salt medium, and applied to starvation agar plates (9 cm diameter) with niacin and niacinamide (Daniel & Rusch 1962) as described (Starostzik 1995; Starostzik & Marwan 1998). A ring of 1 g of cell paste (fresh weight) was applied to the centre of each plate with the help of a motor-driven 50-mL syringe coupled to an automatic device for rotating the agar plate around its axis. Plates were incubated for 6 days at 22°C in complete darkness. The incubation temperature is critical in order to avoid unwanted spontaneous sporulation in response to heat.
Induction of sporulation, preparation of total RNA, and multiplex gene expression profiling
Sporulation of competent plasmodia was induced by a far-red light pulse (30 min, λ≥700 nm, 13 W/m2) as described (Starostzik & Marwan 1998). At 6 h after the onset of the light pulse, three quarters of each plasmodium were harvested with a small glass spoon (Roth, Karlsruhe, FRG), shock-frozen in liquid nitrogen and stored at –80°C for RNA isolation. One quarter was maintained in the dark for another 18 h to reveal the developmental decision (i.e. whether or not sporulation occurred). Control plasmodia were treated identically except that the light pulse was omitted (dark control). All manipulations were done under sterile conditions and under green safe light as described (Kroneder et al. 1999). For extraction of total RNA, approximately 100 mg fresh weight material of a single plasmodium was used. Samples were homogenized quickly at room temperature for 10 s in 3-mL peqGOLD TriFast (PeqLab; Nr 30-2020) using a glass potter and incubated for further homogenization for 1 min at 50°C. RNA was extracted with phenol/CHCl3 using 15 mL Phase Trap tubes according to manufacturer's instructions (PeqLab; Nr 30-0150A-01). Afterwards the RNA was precipitated with ethanol, and digested with rDNase to remove DNA contamination as described (Hoffmann et al. 2012).
Multiplex RT–PCR reactions were performed to simultaneously amplify 35 transcripts and the amplification products were quantified through separation of the fragments on a Beckman Coulter 8-capillary sequencer (CEQ 8800) as described (Hoffmann et al. 2012). Each series of samples to be directly compared to each other was analyzed with the same gene expression profiling (GeXP) kit to ensure consistent results. For each GeXP kit used, a new calibration curve was established as described in order to correct for variations in the quality of the kit components.
Computational methods, statistics, and software used
GEDI maps were computed with the Gene Expression Dynamics Inspector v2.1, a platform-independent software package (http://www.childrenshospital.org/research/ingber/GEDI/gedihome.htm) as described in the following. The expression values of the set of 35 genes as measured in all samples of a given strain (wild type or mutant) were normalized to the average expression level of the reference genes dspA, ribB, and damA. Then, the gene expression data measured for all plasmodial samples (wild type and mutants) were combined into one comprehensive dataset on which the algorithm was executed. Accordingly, the relative changes in the colour patterns of the GEDI maps are fully comparable between all plasmodia of the entire dataset. A grid size of 4 × 4 was chosen so that the algorithm assigned at least one of the 35 genes to each of the 16 tiles. The resulting miniclusters defined genes that are coregulated in the wild type and the mutants.
To define genes that were up- or downregulated in response to far-red light, the data were analyzed using the parameter-free Mann–Whitney U-Test (Mann & Whitney 1947). The U-value Ua characterizing the difference between the plasmodia of group a (e.g. dark control) with group b (e.g. far-red irradiated) was calculated according to
where na and nb are the sample sizes of group a and group b, respectively, and Ra is the sum of ranks in group a. A threshold for the U-value of Ua ≧0.05 or ≦ 0.05 was chosen to indicate whether the light and the dark treated plasmodia form significantly different groups with respect to the relative concentration of any given transcript.
Pearson Correlation Test
The Pearson correlation (Rovine 1997) was used to reveal any pairwise correlation in the expression levels of the 35 genes. For 11 genes selected from the GEDI map, the correlation coefficient r was calculated according to
where x and y are the expression levels of two different genes i and sx and sy are the standard deviations of the transcript abundance of gene x and gene y, respectively. Thresholds of r ≤ −0.8 or r ≥0.8 were chosen for negative or positive correlation to be significant.
Isolation of mutants in the photocontrol of sporulation
Libraries of haploid amoebae mutagenized with ENU were screened for mutants in the photocontrol of sporulation (Fig. 1). Amoebal clones obtained as single colonies on agar plates were picked twice. One sample was used to propagate the amoebal clone for preservation as glycerol stock. In parallel, amoebae were allowed to develop apogamically into a plasmodium, which was grown to a macroplasmodium and then starved for 6 days in complete darkness. During the period of starvation, wild type plasmodia become competent for the induction of sporulation by far-red light, blue light, or heat shock (Starostzik & Marwan 1995b). After 6 days, plasmodia were screened for spontaneous, that is, light-independent sporulation. Plasmodia were inspected visually for altered plasmodial morphology and subsequently exposed to a far-red light pulse, which is saturating in inducing sporulation in the wild type. Plasmodia that spontaneously sporulated in the dark and those that did not sporulate in response to the far-red pulse were subjected to rounds of verification in order to identify mutants in the photocontrol of sporulation that display tight phenotypes. We screened libraries containing amoebae that were mutagenized once or twice, respectively. From 820 clones of the once mutagenized library two spontaneously sporulating mutants were obtained. Out of 500 clones from the twice mutagenized library, seven nonsporulating mutants were obtained (Table 1). Amoebal subclones retrieved from the glycerol stocks confirmed plasmodial phenotype and stability of the identified mutant strains. Mutants obtained in the present screen and those isolated in previous screens (Starostzik & Marwan 1998; Sujatha et al. 2005) were found to be phenotypically stable.
Table 1. Mutant strains and relevant phenotypes of the strains used
Strongly reduced sporulation in response to far-red
Strongly reduced sporulation in response to far-red
Strongly reduced sporulation in response to far-red
Reduced sporulation in response to far-red
Strongly reduced sporulation in response to far-red
Gene expression analysis uncovers different molecular phenotypes of nonsporulating mutants
Transcriptomic studies (Barrantes et al. 2010) as well as quantitative multiplex RT–PCR (GeXP) analyses with RNA samples prepared from individual plasmodial cells (Hoffmann et al. 2012) have shown that at 6 h after induction by a far-red light pulse, that is, shortly downstream of the point of no return, the expression level of many genes has already changed drastically. Some genes are differentially regulated even at 2 h (Hoffmann et al. 2012) which is far upstream the point of no return where sporulation of the induced plasmodium can still be prevented by feeding glucose (Starostzik & Marwan 1994, 1995b). This suggests that sensory signal transduction, signal processing at the protein level, and gene regulation at least to some extent are chronologically interwoven. Moreover, not all plasmodial cells that show a light-induced transcriptional response necessarily sporulate (Hoffmann et al. 2012). Therefore, it seemed likely that at least some of the nonsporulating mutants might have altered gene expression patterns and/or respond at the transcriptome level to the light stimulus even if sporulation subsequently cannot occur.
Wild type and mutant plasmodia starved for 6 days were irradiated with a 30 min pulse of far-red light. At 6 h after the light pulse, one quarter of each plasmodium was cut off together with its supporting agar slice, returned to the dark and checked on the next day, in order to see whether or not sporulation had occurred (Fig. 2). The other part was immediately shock-frozen in liquid nitrogen for subsequent isolation of RNA. In the dark controls irradiation was omitted but the samples were treated identically otherwise.
A set of 35 genes has been selected based on their differential regulation (up or down) during commitment of the wild type to sporulation and/or for encoding different types of proteins like kinases, lipases, checkpoint proteins, RNA-binding proteins etc. to cover different cellular processes (Table 2) (Hoffmann et al. 2012). Transcripts of the set were amplified by multiplex RT–PCR in the same cap and quantified through separation of the amplified fragments on a capillary sequencer (GeXP analysis) (Liu & Saint 2002; Rai et al. 2008; Hoffmann et al. 2012). The expression levels of each of the 35 genes relative to each other were evaluated for each sample (i.e. for each individual plasmodial cell).
Table 2. Genes analyzed for transcript abundance and their degree of similarity to orthologues in the Swissprot data base
Percentage query coverage
Actin, plasmodial isoform
Probable basic-leucine zipper transcription factor G
Cell division control protein 31
Putative transcriptional regulator cudA
DNA damage-binding protein 1a
Dual specificity protein phosphatase 12
EH domain-containing protein 1
Probable GTPase-activating protein 8
Histone chaperone ASF1A
Probable histone H2B 4
Checkpoint protein hus1 homolog 1 (LigA)
Meiosis protein mei2
Non-histone chromosomal protein 6
Serine/threonine-protein kinase pakC
Proliferating cell nuclear antigen
Phosphatidylinositol 3-kinase 2
Phosphatidylinositol 4-kinase beta
Serine/threonine-protein kinase phg2
Phosphatidylinositol-glycan-specific phospholipase D
Phosphatidylinositol-glycan-specific phospholipase D
Protein phosphatase 2C POL
Physarum specific gene
Pumilio homologue 2
Piwi-like protein 1
Circularly permutated Ras protein 1
Ras-related protein RABD2a
Regulator of G-protein signaling 2
60S ribosomal protein L38
60S ribosomal protein L4-2
Tumor suppressor p53-binding protein 1
Programmed cell death protein 2
Wild type and sporulation-deficient mutants were phenotypically characterized by analyzing, visualizing, and comparing the expression patterns of 35 genes. Data were analyzed and visualized with the Gene Expression Dynamics Inspector (GEDI) (Eichler et al. 2003; Guo et al. 2006), a software to cluster high dimensional gene expression data into a two-dimensional colour map. The method uses so-called self organizing maps (SOM) to generate patterns that are easily accessible to the human eye. The underlying algorithm clusters genes with similar expression patterns according to their Euclidian distance and arranges the resulting miniclusters with similar behavior adjacent to each other. Dynamic changes in gene expression patterns become therefore easily visible.
To generate the GEDI maps shown in Figure 3, we combined the expression values of all 35 genes measured in all plasmodial cells, wild type and mutants, into one dataset and executed the algorithm. As a result, the assignment of genes to the miniclusters and the geometric arrangement of the miniclusters relative to each other is the same in all maps shown. Each map corresponds to the expression pattern found in a different individual plasmodial cell. Therefore, not only the colour coded expression levels of the 35 genes, but also their expression patterns can be directly compared between individual plasmodial cells for the given experimental condition (dark control vs. far-red irradiated) but also between wild type and mutant strains. For the wild type the gene expression pattern differs drastically between the dark controls and the samples taken 6 h after far-red irradiation. Although some of the plasmodia within the far-red irradiated group as well as within the dark control group have quite similar expression patterns, there is also considerable individual variation between different plasmodial cells.
In the non-sporulating mutants the response to light was clearly different as compared to the wild type. In PHO1, PHO3, PHO57, and PHO76 there was only a slight still in some of the plasmodia a more pronounced response to the light stimulus. In PHO64 and PHO68 plasmodia the response to far-red at 6 h was very clear but still different as in the wild type.
In strains PHO48 and PHO72 some of the plasmodia sporulated in response to the light stimulus. Nevertheless, we classified this phenotype as tight as well, because the lack of sporulation in some of the plasmodia regularly occurred in response to a stimulus that is highly saturating in the wild type. Note that also in these mutants sporulation was always an all-or-none response for each individual plasmodial cell. The gene expression response of PHO48 and PHO72 to light, which was clearly different to the wild type, supports the use of the gene expression pattern as a mutant phenotype. In these strains (PHO48 and PHO72) a clear gene expression response to the light stimulus was not obvious for all plasmodia and the mutants behaved different as compared to the wild type in this respect as well.
Over the years it very rarely occurred that in strains PHO1 and PHO3 some of the plasmodia sporulated in response to light, suggesting that the mutation in these strains also influences the probability for sporulation to occur rather than abolishes the ability for sporulation per se.
Mutants show different patterns in the correlated expression of genes
The GEDI maps clearly show that there is a relatively high variation in the expression patterns between individual plasmodia of a given experimental group. High variation was even obvious in those cases where only few plasmodia have been analyzed. In some clusters of genes identified with GEDI analysis, up- or downregulated genes cluster together. It appeared that the expression levels of some of these genes are more or less strictly correlated in individual cells in the dark or in far-red irradiated samples although the relative abundance of the transcripts as compared to other messages can vary drastically and also change between dark controls and far-red irradiated plasmodial cells (Fig. 4). To obtain more comprehensive results, we arbitrarily picked from the strongly differentially regulated genes a subset of 11 genes from the GEDI miniclusters to perform a Pearson correlation analysis and to quantify the correlated expression of each gene with each other. In Figure 5 the correlations in gene expression are graphically represented setting a threshold for the correlation coefficient of r = 0.8. In the wild type there is a clear correlation in gene expression levels in and in between the groups of upregulated and downregulated genes. The expression levels of cdcA and psgA in wild type, for example, are inversely correlated and the correlation is only seen in plasmodia of the dark controls (Fig. 4), while cdcA and pwiA are positively correlated in the dark and in far-red irradiated plasmodia. As the overall correlation value of cdcA/psgA pair was below threshold, it is not seen in Figure 5.
In different mutants correlation patterns clearly differ from each other. In PHO72 plasmodia, for example, the number of correlations between genes has drastically increased. This is evident for the sporulated PHO72 plasmodia and even more pronounced in the non-sporulated plasmodia. A similar high degree of correlation is seen in PHO1 while the correlation patterns of PHO1 and PHO3 differ drastically, although the two strains appear to behave somewhat similarly according to the GEDI maps. In the sporulated plasmodia of PHO48 there is correlation among the up- and the downregulated genes, similar to wild type, but in the non-sporulated PHO48 plasmodia, most correlations are lost. Most interestingly, new correlations are found and some genes (in PHO57 and PHO64) are even inversely regulated by light as compared to the wild type. Hence, the non-sporulating mutants differ drastically from the wild type and from each other with respect to the expression correlation patterns of genes that are differentially regulated before and during commitment of the wild type to sporulation.
We have isolated Physarum polycephalum mutants that do not sporulate in response to far-red light and performed quantitative analyses of the expression levels of a set of 35 genes with the help of GeXP multiplex RT–PCR. The gene expression patterns were evaluated with GEDI (Eichler et al. 2003; Guo et al. 2006), a software package that clusters genes according to the similarity of expression patterns and visualizes the results with the help of self organizing map algorithms. GEDI maps of individual plasmodial cells indicated that the plasmodia of different non-sporulating mutants responded differently to the light stimulus as well as differently as compared to the wild type (Fig. 3). The GEDI maps also indicated that the gene expression pattern of unstimulated plasmodia (the dark controls) is altered in most of the mutants as compared to the wild type. Even without GEDI mapping it was obvious that the expression levels of some of the genes that are up- or downregulated in the wild type in response to light stimulation were relatively strictly pairwise correlated (Fig. 4). For some gene pairs correlation was not seen in far-red irradiated plasmodia, while it clearly occurred in the dark controls. Correlation was positive or negative, depending on the pair of genes considered. Despite being strongly correlated, the expression levels of some of these genes varied considerably (fivefold or more) between individual plasmodial cells of a given strain when compared to other, not regulated genes in the samples. The individual differences in gene expression patterns of individual plasmodia from the same strain are not due to inhomogeneities within a multinucleate plasmodium, as different samples taken from different parts of the same plasmodium show indistinguishable gene expression patterns (Rätzel and Marwan, unpubl. data, 2012). For a systematic evaluation of the pairwise correlation of transcript levels, we have chosen an arbitrary subset of 11 out of those genes that are strongly differentially regulated in the wild type and calculated the Pearson correlation for each gene as compared to each other out of the set. This analysis revealed that the different mutants displayed different patterns of pairwise correlation (Fig. 5). In some of the mutants, the number of pairwise correlations increased. In other mutants the number of correlations decreased and new patterns of pairwise correlation formed. These results suggest that patterns of correlation in the expression levels of pairs of genes can be used to functionally characterize non-sporulating mutants. In summary, there are three molecular criteria to phenotypically discriminate non-sporulating mutants: (i) the gene expression pattern in unstimulated cells; (ii) the differential regulation of genes in response to stimulation with far-red light; and (iii) the pattern of correlated expression of pairs of genes. Certainly, whether or not these phenotypic traits are specific to the mutation that blocks light-induced sporulation has to be determined for each mutant individually.
How can mutations change the correlated expression of pairs of genes? A possible explanation is given in Figure 6. Let us assume that in the wild type, the transcript level of four genes is under the control of four regulators that are differently connected to the four genes (Fig. 6a). Each of the regulators may receive input from cellular processes while there is no direct or strict coupling between the four regulators. In the example shown, expression of Gene 4 is not correlated with the other three genes as it is controlled by its own regulator (Reg 4), which is neither linked to the other three genes nor to the other three regulators. The expression levels of Gene 1, Gene 2, and Gene 3 are more or less strongly correlated due to the partly overlapping regulatory control by Regulator 1, Regulator 2, and Regulator 3. If the highly connected Regulator 2 is deleted (Fig. 6b), the only correlation that is left is between Gene 1 and Gene 2 as mediated by Regulator 1. The correlation strength between the two genes may be somewhat weaker as in the wild type because the co-regulatory influence of Regulator 2 on the pair of genes is missing. When on the other hand, Regulator 1 (Fig. 6c) or Regulator 3 (Fig. 6d) are deleted, the strengths of pairwise correlations change due to the altered combinatorial influence of the respective regulators that have been disabled by the mutation. At a given threshold for the numerical value of the correlation of expression to be considered as significant, as it was set to prepare Figure 5, the number of correlating connections between the three genes changes accordingly. Based on this rationale it might be possible to reverse engineer (i.e., infer) the layers of regulatory control uncovered by the mutants from altered patterns in the correlated expression of pairs of genes as measured in individual plasmodial cells. In addition, GEDI analysis may assist reverse engineering approaches. To conclude, the experimental data suggest that the increase in the number of gene pairs with correlated expression level is due to the deletion of regulators or of their upstream components.
This work was financially supported by a BMBF FORSYS grant and by the Excellence Initiative of Saxony Anhalt. We thank Dr Joseph Zhou for help with the GEDI analysis and for critical reading of the manuscript. We also thank Mrs Bärbel Lorenz and Mrs Regina Pflug for excellent technical assistance.