Susceptibility of E. dermatitidis to proton, deuteron, and α-particle irradiation
We initiated this study by characterizing the resistance of E. dermatitidis melanized (WT) and non-melanized (Δpks) strains to three ionizing particles with different RBE, LET, and mass (in ascending order: protons, deuterons, and α-particles [40]). To enable comparisons with γ-IR, we considered several factors (see Fig. 1 and Materials and Methods), including: the lower penetration of these particles into materials; the RBE of each particle type in comparison with γ-IR [41, 42]; and the shape of previously constructed survival curves exhibited by E. dermatitidis after irradiation. At doses much greater than 2000 Gray (Gy, J/kg), for example, survival decreases rapidly, and differences between strains become harder to interpret due to the low percentage of recovering colonies produced [31]. Effects of interest, such as would be imparted by melanin protection, are therefore best detected at moderate exposures, so we chose to irradiate each strain with doses equivalent to 500, 1000, and 2000 Gy of γ-IR. After each exposure, cells were collected, their concentration determined, and they were plated after appropriate dilution and grown, to allow for their survival to be measured through counting colony forming units.
Survival was then modeled with the resulting measurements. The best-fit parameter values for the WT strain were: α = 6.38 (95% CI: 0.26, 16.51) kGy− 1, PE (Plating Efficiency) = 0.60 (0.21, 0.99). Corresponding values for the Δpks strain were: α = 2.10 (0.01, 10.04) kGy− 1, PE = 0.32 (0.16, 0.80) (see Methods for explanation of modeling and Supplemental Tables 1 and 2 for plating and modeling data). The fitted dose response curves and data points are shown in Fig. 2. Notably, these differences between strains in terms of parameter values did not reach statistical significance (95% CIs overlapped). However, these results suggest some evidence of heterogeneous radiosensitivity in the irradiated fungal population. The Δpks strain appears to have a smaller α parameter than the WT strain, suggesting that Δpks has lower radiosensitivity. PE, though, was also smaller for Δpks than WT, which may suggest that only some selected “healthy” sub-population of Δpks cells was able to grow even without radiation, and possibly this selection was correlated with improved radioresistance as well. Indeed, Δpks cells have been shown to clump together, possibly due to a decreased ability to completely cleave apart after cell division, which can affect plating efficiency [31, 43]. In all, though, there was no clear protection, or difference at all, in WT cells compared to those without melanin.
The transcriptomic response of E. dermatitidis to particle irradiation
The effect of exposure on cell survival was roughly equivalent between these particles and γ-IR for similar doses, and we considered that differences between these conditions could be accessed by looking at the transcriptomic response for each source. Therefore, next, we performed RNA-seq on cultures recovering from the second highest dose (equivalent to 1000 Gy of γ-IR) for each particle, corresponding to 30–50% cell survival depending on the type of IR. Before collecting RNA we also incubated cells in fresh medium for 1 h to allow for them to mount a transcriptomic response (see Methods).
Each particle irradiation induced a strong and consistent gene expression pattern in both strains (Supplemental Tables 3 and 4). For example, R2 values from comparative analysis of transcripts per million counts ranged from 0.84–0.95 between biological replicate samples as well as WT and Δpks1 cultures under the same condition (IR exposure or control), while the irradiated transcriptome was vastly different from that of controls (Fig. 3A, Supplemental Table 3, and Supplemental Fig. 1). Initial analysis of differential gene expression with an FDR < 0.05 demonstrated that a range of 4088 (2110 up, 1978 down in the Δpks1 deuteron-irradiated culture) to 5305 (2739 up, 2566 down transcripts in the Δpks1 proton-irradiated culture) out of the 9268 transcripts predicted to be encoded by the E. dermatitidis genome [18] were differentially expressed. Overall, 1095 (UP) and 969 (DOWN) genes were seen in every sample at this cutoff level, representing 22.2% of the predicted protein-coding genes in this organism (Fig. 3B and Supplemental Tables 3 and 4).
We next looked within this shared set of genes for GO-terms that were significantly enriched (see Methods), and focusing specifically on Biological Process categories, observed the following (Fig. 3, C and D): genes involved in DNA repair (e.g. DNA replication, DNA recombination, and nucleotide-excision repair) and genes involved in protein catabolism (e.g. ubiquitination) were overrepresented among the upregulated genes (Fig. 3C). Approximately one third of the genes in the upregulated group (369/1095, 33.7%), moreover, were predicted to encode proteins with no known function.
Among the downregulated genes (Fig. 3D), biological processes that were significantly enriched were involved in cell growth, including categories such as intracellular trafficking (e.g. microtubule-based movement), ATP synthase-coupled proton transport, and translation (e.g. ribosomal biogenesis) were particularly enriched. Once again, one third of the genes (340/969, 35.1%) were labeled as encoding hypothetical proteins (Supplemental Tables 3 and 4). This is approximately half as numerous as the proportion of unannotated genes within the entire genome (5842/9578, 60.1%), suggesting that the shared IR-responsive gene set may be enriched in previously-characterized and conserved proteins, but it also delineates the further genetic studies and annotation that needs to occur to understand the E. dermatitidis proteome. Taken together, these results suggested that E. dermatitidis cells respond to charged particle irradiation by activating the DNA repair machinery and removing damaged proteins, as well as inhibiting several processes involved in cell growth.
Comparison of the transcriptomic responses of E. dermatitidis to particle and γ-irradiation
The functionally enriched categories observed here, then, were extremely similar to the γ-radiation response. The close relationship between the transcriptomic responses to these two sources could also be seen at a larger level by plotting the Log2fold-change values of genes significantly regulated by particle irradiation for each sample on a heatmap (Fig. 4). However, the heatmap produced by this analysis showed that samples clearly clustered by IR type, demonstrating that there were notable differences between the two responses that needed further analysis to uncover. To address this, we focused on getting more information about the genes were differentially expressed in each group, particularly those that did not have informative annotations. We therefore compiled lists of genes that were differentially regulated more than 5-fold (FDR < 0.05) in both γ-radiation and particle both -irradiation datasets and compared these lists [31]. Notably, we observed less overlap among these two sets than we expected, with only approximately 12% of the genes shared overall (Fig. 5, A and B). We also were surprised to see that, even after compiling the results from six separate experiments (WT and Δpks cells for three particle exposures), there were many more genes that passed the > 5-fold change cutoff in the particle dataset compared with the γ- irradiated dataset (N = 467 vs 166, respectively), indicating that particle irradiation invoked a more dramatic response in the transcriptome.
The lack of overlap between these gene sets caused us to reanalyze these data for patterns of enrichment that were specific to each group. This revealed some interesting features. First, DNA repair genes were still enriched among the shared, upregulated set, but DNA replication and repair genes were also enriched within the γ-specific, upregulated group. These proteins included Rad54B, the NHEJ protein Ku70, DNA damage-binding protein CMR1, as well as DNA polymerase α and ε, replication fork protection complex subunit Swi1/Tof1, and the DNA replication regulator SLD2 (Fig. 5A). Proteins only upregulated by particle irradiation, on the other hand, included autophagy-related proteins (HMPREF1120_02756/ATG8, HMPREF1120_07182/ATG3, and the cysteine protease HMPREF1120_06308/ATG4) and specialized genes potentially involved in sensing and signaling, such as bacteriorhodopsin (HMPREF1120_00264) and a Sir2-like histone deacetylase (HMPREF1120_05820). Additionally, the group of shared, downregulated genes was small (N = 27), and was enriched for proteins involved in cell growth, cell cycle, and intracellular trafficking, but not enriched for those associated with translation and ribosomal structure (Fig. 5B). Translation, rather, was highly enriched among the downregulated genes specific to the particle-irradiated gene set, which is notable given the well-characterized, antagonistic relationship between protein synthesis and autophagy [44–47]. The list of genes downregulated by γ-radiation, on the other hand, was small (N = 18) and was composed of members that had general annotations such as cell surface receptor signaling and not provide much information.
After this analysis many genes still remained without a predicted function, so we further examined each set with information from the EuKaryotic Orthologous Group (KOG) database. Grouping genes into 22 of the KOG categories represented in a substantially decreased the number for which no information was known (from 188/467 to 46/467 for particle irradiation, and from 52/166 to 20/166 for γ-irradiation), and also allowed us to look at the functional composition of the entire E. dermatitidis genome. Figure 5C shows these results. It is immediately apparent that particle-regulated genes embody a greater functional diversity compared to γ-regulated genes. Upregulated genes from the particle irradiation sets include a high proportion annotated as involved in defense mechanisms (e.g. multidrug transporters and general stress-response proteins) while the downregulated gene set from this experiment had a large proportion involved in amino acid transport and chromatin structure compared with the whole genome. The upregulated genes in the γ-IR experiment, however, were dominated by those involved in Replication, Recombination, and Repair, while among downregulated genes there appeared to be an enrichment of those involved in secondary metabolism (Fig. 5C).
In all, although the transcriptomic response to IR broadly involves upregulation of DNA repair and attenuation of growth and protein production, there are some differences observed in the response to particle irradiation, including an upregulation of genes regulating cell death and a stronger inhibition of ribosomal production in comparison with the γ-radiation transcriptomic response. This result agrees with previous observations that particle irradiation elicits a more complex response due to its ability to cause more serious damage to DNA and to move in more compact tracks through the cell, which may cause unique effects to other cellular structures [42].
Genes most sensitive to particle irradiation exposure
One goal of this study is to identify genetic loci that robustly respond to IR, which could be used to develop strains that can sense IR and produce various outputs tied to gene expression. With this in mind, we searched for the most promising loci using the shared particle gene through MEME analysis, which searches for conserved promoter sequences in gene sets; STRING analysis, which identifies connections between genes within datasets based on published genetic studies; and identification of the genes that showed the strongest change in expression during IR recovery. The output of MEME and STRING analyses are presented in Supplemental Tables 5–8 and Supplemental Fig. 2. Briefly, MEME analysis identified a possible 9-nucleotide sequence marking IR-regulated genes, although it was clear that other unidentified elements also exist (Supplemental Tables 5 and 6). STRING analysis further confirmed the centrality of DNA replication among the upregulated genes, but also identified certain genes that were central to this pathway, such as the Minichromosome Maintenance Protein complex (Supplemental Fig. 2 and Supplemental Tables 7 and 8).
Next, to identify the most prominent among the differentially expressed genes, we averaged expression data across all strains and particle samples. We identified 19 genes that were differentially expressed with a Log2FC > 5, representing a > 32-fold change in expression during recovery from IR exposure (Table 1). The most upregulated gene (Log2FC = 6.44 +/-0.77) among all samples was HMPREF1120_01375. It is annotated as a triacylglycerol lipase and contains acetyl esterase and lipase domains, potentially associated it with fatty acid catabolism in autophagy [48, 49]. Other upregulated genes in this list include HMPREF1120_08154, a homolog of Saccharomyces cerevisiae YTA12, a component of the mitochondrial inner membrane mAAA-protease that could be involved in clearing damaged proteins; HMPREF1120_02698, a homolog of S. cerevisiae UBP15, which is involved in S phase entry [50]; and HMPREF1120_02127, a homolog of S. cerevisiae gene YGL039W, a protein involved in reducing aldehydes, which are present in cells undergoing oxidative stress [51]. Interestingly, HMPREF1120_08448, HMPREF1120_00007, and HMPREF1120_06618 did not contain any predicted functional domains, and which we could identify no orthologues in genomes outside of the Exophiala genus.
Within the group of shared downregulated genes, three genes had no conserved domains or informative orthologues in other species (HMPREF1120_04574, HMPREF1120_03067, HMPREF1120_08344). Of the other six, however, four had functions that could be traced to the IR response, being involved in chromosome dynamics in some way, such as the orthologues of SMC1 and SPC25 (chromosome segregation), the aurora kinase IPL1, and the condensin YCG1. Additionally, the chitin synthase CHS1 and an endoglucanase were the remaining two genes that were on this list of genes repressed during IR exposure recovery. These could be associated with the cessation of growth that we have observed taking place in E. dermatitidis cells while damage is repaired or cleared [31]. Importantly, 16/19 of these genes were regulated similarly in our γ-regulated dataset, so further studies of IR sensitive promoters may be able to focus on the expression of these few genes in different environments.
Particle-specific transcriptomic responses
Another goal of this study is to identify genetic loci that have unique responses to different ionizing particles, so next we looked for patterns specific to certain irradiation datasets. Such loci were not immediately apparent, as the datasets were vastly similar. This is demonstrated by comparing FC values for each significantly regulated gene (FDR < 0.05) shared among the particle groups, which produced R2 values > 0.92 for all pairwise comparisons (Fig. 6A). Moreover, no were observed to be upregulated under one condition and downregulated in another. When these gene groups were minimized to facilitate closer analysis by looking only at genes differentially expressed in each condition with a FC > 5, however, several genes were identified in only one set, and some of these unique gene sets were significantly enriched for certain biological processes (Fig. 6,B and C). For α-particle-irradiated samples, genes involved in DNA repair, including DNA Ligase 1 and the helicase RecQ, were upregulated, while genes predicted to be involved in translation were specifically enriched within the α-downregulated genes. Genes upregulated only after proton-irradiation were enriched in transmembrane transport, including 13 Major Facilitator Superfamily domain-containing proteins, while genes involved in IMP biosynthesis (purine metabolism) and tRNA Aminoacylation were overrepresented among downregulated genes. It is worth noting that protons and α-particles represent the two extremes of particle size, which may account for their higher number of uniquely regulated genes, and the lack of these in the deuteron set. Finally, within the deuteron set, genes involved in transcriptional regulation were enriched among the upregulated transcripts, including RNA-dependent RNA polymerase, which is throught to be involved in RNA-silencing.
To search for more subtle patterns within this data, we again performed STRING analysis to identify pathways that were specific to a certain particle. To achieve this in a manageable way with this large dataset, we only analyzed genes that possessed highly-confident links to at least two other genes (probability > 0.9) among all significantly upregulated genes (FDR < 0.05) for each condition (Supplemental Tables 9 and 10). This provided a detailed list of the specific complexes that were included in the DNA repair and protein catabolism pathways induced during the response to particle exposure, including the polymerases involved in DNA replication (α, δ, and ε), the proteins involved in regulating the cell cycle (Cdc6, 45, and 48), and those involved in DNA repair (for HR – Rad50, 51, 52, MRE11, and Rhp54; for UV excision repair – Rad32; for mismatch repair – MLH1/MSH3/PMS2; and translesion synthesis - Rev1). It also revealed proteins involved in transcriptional regulation, including two histone acetyltransferase-associated categories, the histone chaperone ASF1, and a Jumonji domain-containing protein, which are involved in transcriptional repression through modulating histone methylation [52]. Finally, it identified other distinct pathways not directly related to DNA or protein damage, including the Myosin ATPase, involved in intracellular trafficking, and nitrogen (glutamine synthase and glutamate dehydrogenase) and carbon (phosphoenolpyruvate carboxykinase) metabolism gene nodes.
Supplemental Tables 11 and 12 present the genes represented as highly confident nodes present in only one or two conditions. This analysis provided the most detailed picture of the different ways each source affects E. dermatitidis. For example, eighteen different categories had two or more genes that were related with high confidence between the α- and deuteron-irradiated samples, which included two unique categories of DNA-associated proteins: the RecQ repair protein and the cohesin complex. In the proton and deuteron datasets, moreover, DNA topoisomerase (DNA replication), the FACT complex (chromatin structure), the RNA helicase (splicing), and Tyrosinase (melanin production) nodes were present only in deuteron and proton datasets. Even more nodes were discovered that were unique to individual irradiation sources. α-particle samples had subunits of the actin-related protein 2/3 complex (which is involved in endocytosis and morphogenesis [53]) and synaptobrevin, which is involved in intercellular trafficking. Similarly, the deuteron set contained proteins linked to fimbrin, which is involved in actin cytoskeleton regulation, endocytosis, and morphogenesis [54], and the MYST family of histone acetyltransferases. The proton data set had the most unique categories in this analysis (N = 30), and included interesting links such as cystathionine synthase and glutaredoxin genes, which may be involved in the response to increased ROS levels, and several genes involved in mRNA splicing, including the splicing factors slt11 and ini1. This may warrant further exploration of alternative splicing and isoform analysis of E. dermatitidis cells recovering from specific sources of IR.
Lastly, we attempted to discover distinct loci that robustly responded to one specific source, which was more challenging. In all, only 15 genes were upregulated > 10-fold in response to one type of particle. Of these genes, moreover, only three had predicted functions (Table 2). Proton-irradiation again had a larger set of genes than α- or deuteron-irradiation, of which one interesting member was the receptor for Mating Pheromone A, but the overall pattern still held with this sample, with 9/11 genes having no identifiable functional domains. With the transcriptome being dramatically altered in irradiated compared with non-irradiated samples, this lack of highly-regulated loci unique to each particle reiterates that there is substantial overlap in the type of cellular damage that occurs after acute exposure to each of these IR sources, but this dataset can be used to focus on specific loci for the development of IR detection and discrimination strains in the future, and for functional analyses.
Effect of melanin on the response to charged particle radiation exposure
The final analyses we performed was a comparison of the responses of melanized and non-melanized cultures to these particle sources. Because the particle-associated transcriptome was distinct from our previous γ-radiation study, we were prepared to see some difference between the two strains. However, in agreement with the survival data we collected, the responses had substantial overlap, with comparative plots of Log FC between each sample exhibiting R2 values of ~ 0.9 and greatet with no genes significantly regulated in opposite directions between any samples (Fig. 7A-D), and no differences in highly confident STRING nodes (data not shown). However, we did observe some broad differences between the two responses. Specifically, non-melanized cultures had more enriched biological processes compared to the melanized gene set, even though the former contained a smaller group of genes (Fig. 7, E and F). Enriched categories from this analysis included transcription, mismatch repair, protein catabolism, and autophagy in the upregulated, non-melanized set, while additional genes involved in translation (protein folding, ribosomal biogenesis, RNA processing) and growth and energy production (cell cycle, glycolysis) were enriched among the downregulated genes unique to the non-melanized strain. Interestingly, we have observed several other times that the regulation of translation is affected by melanization status, with ribosomal genes preferentially downregulated, under normal conditions, in melanized strains [31, 38]. These enrichment analyses provide some further credence to the idea that a lack of melanin mediates a complex response in E. dermatitidis, but the response is subtle and, with regard to IR recovery, indirect.