Many uncontrolled factors can cause variability in results derived from soil toxicity tests with Folsomia candida
], which may affect reproducibility. To verify that the phenanthrene concentrations in soil, taken from the literature [16
], had a significant and reproducible effect on the reproduction of F. candida
in our experiment, we performed a 28 days exposure toxicity test. Nominal phenanthrene concentrations of 24.95 (EC10
) and 45.80 (EC50
) mg kg-1
soil, and also a solvent (acetone) control were tested. Clean LUFA 2.2 soil was used as reference (untreated control). The solvent control did not show a significant effect on reproduction compared to clean reference LUFA 2.2 soil, but phenanthrene concentrations of 24.95 and 45.80 mg kg-1
soil did have significant effects (Figure ). Reproduction was reduced with 27% and 45%, respectively, compared to the reference soil. Actual phenanthrene concentrations in soil were measured using high performance liquid chromatography at the start and the end (28 days) of the experiment. The concentrations of phenanthrene in soil after 28 days were approximately 60% of the initial concentrations, which is expected for non-persistent organic compounds like phenanthrene.
Figure 1 Effect of phenanthrene on reproduction. The average (n = 5) number of Folsomia candida juveniles per jar after 28 days exposure to reference, and solvent (acetone) control soil, and soil containing 24.95 mg kg-1, and 45.80 mg kg-1 phenanthrene. The acetone (more ...)
To identify transcripts that responded to the phenanthrene exposure, we used gene expression microarray analysis. Microarrays were constructed containing 5,069 different 60 nt long DNA probes, printed in triplicate. The probe sequences were taken from an earlier array design, described in a previous study [14
]; each probe represents a different gene cluster from Collembase [12
]. Adult animals were exposed for 2 days to clean LUFA 2.2 soil (reference), and soil spiked with nominal phenanthrene concentrations of 24.95 and 45.80 mg kg-1
soil. No solvent control was used in the gene expression experiment, because it did not show a significant effect on reproduction. However, it cannot be excluded that acetone had a minor effect on gene expression. RNA was extracted from all animals from each experimental jar, constituting one biological replicate sample. We applied a so called replicated reference design [18
]; every treated sample was hybridized to a unique reference (untreated) sample. This resulted in 8 reference, 4 EC10
treated, and 4 EC50
treated samples. To assess significant differential transcripts, linear models and empirical Bayes statistics were used [19
]. Derived p
-values were corrected for multiple testing using Benjamini-Hochberg's (BH) method [20
We identified 405 and 251 differentially expressed transcripts (BH corrected p
< 0.05), following exposure to phenanthrene concentrations of 24.95 and 45.80 mg kg-1
soil, respectively. Remarkably, only 50 transcripts were differentially expressed at both concentrations (Additional file 1
: Figure S1). Many transcripts of F. candida
could not be annotated and have therefore an unknown function. Future sequencing efforts, however, are on their way and hopefully more transcripts can be annotated in the near future.
Exposure to low effect concentration of phenanthrene
Of the 405 transcripts that were significantly differential in response to the low effect concentration of phenanthrene (24.95 mg kg-1
soil), 260 transcripts were upregulated and 145 transcripts were downregulated compared to the reference. All differentially expressed transcripts that responded to the low concentration of phenanthrene are available in Additional file 2
: Table S1. Their putative function is based on sequence homology (e.g., BLAST, interPro). Seven transcripts encoding cytochrome P450s were upregulated. Cytochrome P450 enzymes are most commonly involved in monooxygenase reactions [4
]. In addition to this, we also identified other monooxygenases being up- and downregulated in response to the low concentration of phenanthrene. The upregulated cytochrome P450s and other monooxygenases are most likely involved in phase I of the biotransformation and detoxification of phenanthrene. Furthermore, we found upregulation of aldehyde oxidases, carboxylesterases, and short-chain dehydrogenases, which are probably also involved in phase I reactions with phenanthrene [5
]. Many transferase enzymes are involved in phase II of the biotransformation of xenobiotics. In this phase the reactive metabolites created in phase I are being conjugated with polar groups like glutathione or sugar groups [5
]. We identified 7 transcripts encoding glutathione S-transferases upregulated in response to the low concentration of phenanthrene, and one was downregulated. Also many transcripts encoding proteins that contain a UDP-glucuronosyl/UDP-glucosyltransferase domain were upregulated. Stroomberg et al. showed that the phase II biotransformation of the PAH pyrene in F. candida
produced the metabolite pyrene-1-glucoside, but not pyrene-1-glucuronide [23
]. Therefore we assume that these induced transcripts actually encode UDP-glucosyltransferases, and not UDP-glucuronosyltransferases. Membrane transporters which are involved in phase III were also significantly upregulated; we identified 3 ABC-transporters.
The low concentration of phenanthrene induced 2 transcripts encoding heat shock proteins (HSPs) and one chaperonin which are likely part of the general stress response. Many transcripts encoding ribosomal proteins and a few translation initiation factors were also upregulated, which indicates increased protein translation. The translation of all the biotransformation enzymes might be a reason for this increase. Interestingly, many transcripts encoding chitin binding proteins and chitinases were also upregulated. These gene products, together with the upregulation of a transcript with homology to the molting fluid carboxypeptidase A [24
], could be involved in the molting process or the formation of the peritrophic envelope. This peritrophic envelope is excreted by the gut epithelial cells in most arthropods, and is a thin membrane which has protective functions against abrasive food particles, invading pathogens, plant toxins, and oxidative damage [25
]. Phase I metabolites of PAHs can often generate reactive oxygen species (ROS) and can cause oxidative damage. In humans, the microbiota in the colon was able to bioactivate PAHs [26
], and we therefore suggest that ingested phenanthrene is being transformed to ROS forming metabolites by the microbiota in F. candida
's gut. The peritrophic envelope could then function as an antioxidant to protect the epithelial gut cells from ROS. However, further research is needed to confirm ROS production by the microbiota in F. candida
's gut. Endogenous transformation of PAHs by cytochrome P450s also generates ROS, and we found transcripts encoding superoxide dismutase (copper/zinc binding) and catalase both upregulated in response to the low concentration of phenanthrene.
Other transcripts that are worth mentioning are vitellogenin and genes containing a vitelline membrane outer layer protein I (VOMI) domain. These transcripts, all upregulated, are involved in egg production. This suggests that phenanthrene is disrupting the reproduction process in F. candida
in a direct manner. Also many transcripts encoding proteins containing a zinc finger domain were significantly affected (up- or downregulated). Zinc finger domains are often involved in DNA binding, like for example in transcription factors [27
]. Most of these transcripts were downregulated, which suggests that many, perhaps less essential, processes were switched off in order to focus on more essential transcripts that cope with phenanthrene detoxification. Furthermore, transcripts involved in post-transcriptional modifications of RNA, or post-translational modifications of proteins were also affected. Post-translational modification is possibly an indication of altered signal transduction. For example, genes from the RAS family were downregulated. It is however difficult to predict exactly which processes (e.g., cell proliferation or apoptosis) are influenced by these signal transduction pathways, but it could suggest carcinogenic potential of phenanthrene in higher animals.
Exposure to high effect concentration of phenanthrene
Compared to the reference, 251 transcripts were significantly differentially expressed in response to the high effect concentration of phenanthrene (45.80 mg kg-1
soil), 122 transcripts were upregulated and 129 transcripts were downregulated. This is clearly less than in the low concentration exposure. A difference with the low concentration exposure is that we here observed relatively more genes downregulated than upregulated. All differentially expressed transcripts that responded to the high concentration of phenanthrene are available in Additional file 3
: Table S2. We can again identify transcripts involved in all three biotransformation and detoxification phases. For phase I, we observed upregulation of 4 cytochrome P450s and one NADPH cytochrome P450 reductase, and many other differentially expressed monooxygenases (up- and downregulated). Furthermore, we observed upregulated carboxylesterases, and short-chain dehydrogenases, which are likely also involved in phase I reactions. Transcripts encoding glutathione S-transferases and UDP-glucuronosyl/UDP-glucosyltransferase domain containing proteins were also upregulated, which are probably involved in phase II reactions. Six transcripts encoding ABC-transporters were also upregulated; they could be involved in phase III. This is twice the amount of transporters compared to the low concentration of phenanthrene exposure.
The HSP genes that were upregulated by the low concentration of phenanthrene were also upregulated by the high concentration of phenanthrene, which is indicative for general stress. Also here one chaperonin was induced, but interestingly, it was not the same transcript that was induced in response to the low concentration. Furthermore, one transcript containing a DnaJ (HSP40) domain was downregulated in response to the high concentration. A few transcripts encoding ribosomal proteins were all downregulated, which indicate a suppression of protein translation, although a translation initiation factor and a tRNA synthetase were both upregulated. The suppression of these ribosomal proteins is a clear difference between the low and high concentration phenanthrene exposure, because the low concentration induced several other ribosomal proteins. The synthesis of proteins is an energy costly process, and is therefore often suppressed in stressful situations like in detoxification of xenobiotics, in order to reallocate energy budgets [28
]. Furthermore, we identified upregulated transcripts involved in the synthesis of the antibiotic compound penicillin, and downregulated C-type lectins. This suggests that the high concentration of phenanthrene is evoking an immune response, and thus increases susceptibility to pathogens. We observed upregulation of a superoxide dismutase (copper/zinc binding), indicating oxidative stress. This gene was, however, not the same superoxide dismutase that was upregulated in response to the low concentration of phenanthrene. Nevertheless, other transcripts, like glutaredoxins and thioredoxins, that play a role in oxidative stress, were downregulated. More transcripts that were affected in response to the high effect concentration of phenanthrene noteworthy to mention were involved in: transcription and chromatin remodeling, DNA replication, post-transcriptional and post-translational processes, and signal transduction.
Comparison between low and high exposures
Only 50 transcripts were differentially expressed in response to both exposure concentrations. We used hierarchical clustering to group the similarly expressed transcripts (Figure ). The transcripts can roughly be divided into three separate groups. The first group contains transcripts that were highly upregulated in response to both phenanthrene concentrations (most upper [purple] group in Figure ). This group contains cytochrome P450s and unknown genes (with no significant homology). The second group created by hierarchical clustering contains transcripts that were moderately upregulated in response to both phenanthrene concentrations (lower [orange] group in Figure ). In this group we indentified transcripts encoding e.g., heat shock proteins, glutathione S-transferases, and ABC-transporters. The third group (middle [green] group in Figure ) contains transcripts which were all downregulated in response to the high concentration of phenanthrene. In response to the low concentration, some transcripts were also downregulated in this group, but most transcripts, including C-type lectins, were slightly upregulated.
Figure 2 Heatmap for transcripts differentially expressed in response to each of the phenanthrene concentrations. Four microarrays were used for each exposure concentration. Hierarchical clustering was performed using log2 fold change values (treatment/reference). (more ...)
Of the 50 transcripts, we wanted to identify which responded different between the two phenanthrene concentrations. The transcripts that differ significantly could help to explain the different effects on reproduction after 28 days exposure. These transcripts that responded differently between the two concentrations were identified using a t-test, and 16 were found to respond significantly different (p
-value < 0.05) between the two concentrations (see Table ). Many of the genes have an unknown or unclear function. Interestingly, the expression of genes encoding proteins with possible detoxifying and biotransformational functions, (carboxylesterase, monooxygenase, short-chain dehydrogenase, and cytochrome P450), all increased with increasing phenanthrene concentration. This implies that these genes responded in a concentration dependent manner, and are likely good quantitative biomarkers for the relevant phenanthrene concentrations in soil. Two C-type lectins, which are assumed to play a role in the invertebrate immune system [29
], are upregulated in response to the low, but downregulated in response to the high concentration of phenanthrene. This suggests again, like already mentioned above, that the high concentration of phenanthrene had a negative impact on the immune system. This seems to be likely, because more genes involved in the production of antibiotics responded to the high concentration. An impaired immune response was observed in an earlier toxicogenomic study with F. candida
exposed to cadmium [14
]. This implies that the immune response is correlated with suppression of reproduction, no matter the mode-of-action of the toxicant. Interestingly, a recent toxicogenomic study with earthworms, wherein Eisenia fetida
was exposed to soil containing 2,4,6-trinitrotoluene [30
], also showed an impaired immune response.
Transcripts with a significant difference in gene expression between two phenanthrene exposures
Many transcripts were differentially expressed only in response to one phenanthrene concentration, but shared similar putative functions. For instance, we found ABC-transporters that were only significantly upregulated in response to the low concentration, but not to the high concentration of phenanthrene, and vice versa. Other examples are the already above mentioned superoxide dismutases, and chaperonins. Interestingly, the transcripts involved in all three phases of the detoxifying and biotransformation of xenobiotics were upregulated in response to both phenanthrene concentrations. However, only a few were differentially expressed in response to both concentrations, and then only a few of them were regulated in a concentration responsive manner. This demonstrates the complexity of the transcriptional regulation of these biotransformation enzymes. The transcripts involved in chitin metabolism were mostly upregulated in response to the low concentration of phenanthrene. An explanation might be that the high phenanthrene concentration could kill or inhibit the ROS forming microorganisms in the gut. Therefore, the synthesis of a peritrophic envelope would be less necessary.
The larger number of differentially expressed genes in response to the low concentration of phenanthrene, compared with the high concentration, was likely caused by the reallocation and distribution of the animal's energy budget. A higher concentration of phenanthrene in the organism would switch priorities to the production of biotransformation enzymes, and would leave less energy left for regulation of other less essential cellular processes and e.g., reproduction. Especially the genes encoding the biotransformation enzymes that responded in a concentration dependent manner indicate the importance of this process in order to cope with phenanthrene toxicity.
Comparison with the cadmium induced transcriptome
Compared to our previous transcriptomic study [14
], (data available under GEO [31
] accession number GSE11122), wherein F. candida
was also exposed for two days to cadmium polluted soil (also EC50
on reproduction after 28 days), the number of genes that responded are by far less. After 2 days of exposure to cadmium (EC50
) polluted soil, a total of 964 differentially expressed transcripts were identified compared to 251 for phenanthrene (EC50
) polluted soil. One hundred and twelve transcripts are significantly differentially expressed in response to both compounds, which is almost half of the phenanthrene responsive transcripts. Although the experimental conditions were similar in both exposure experiments, we have to be cautious comparing these two datasets, because the two microarray designs were technically slightly different. In the present study a custom Agilent microarray with an 8 × 15 k format was used and in the previous (cadmium) study we used a custom Agilent microarray with a 2 × 11 k format. Consequently, the spots on the two different microarrays varied in diameter. Furthermore, the microarray in this study contained 5,069 different probes printed in triplicate and the microarray in the previous study contained 5,131 different probes printed in duplicate. However, Shi et al. (2006) showed high consistency across different microarray platforms [32
], supporting data comparison between different platforms to some extent. Thus, if we compare the results of the present phenanthrene exposure study with the results of our previous cadmium exposure study, we can see that many transcripts respond in the same manner and other transcripts respond differently. The transcripts that respond similar to both compounds with similar effect concentrations are potential biomarkers for level (degree) of soil toxicity, and transcripts that respond differently are potential compound specific biomarkers. In accordance with our results, this was also shown in previous toxicogenomic studies in e.g., yeast [33
], where a general environmental stress response can be distinguished from a treatment specific stress response. In Additional file 4
: Figure S2 a heat map is shown of hierarchical clustering of the expression of all significant transcripts in response to both compounds. Transcripts encoding monooxygenase (Fcc01289) or short-chain dehydrogenase (Fcc02784) were upregulated in response to both compounds. These transcripts were significantly lower expressed in response to the low concentration of phenanthrene. It seems that their transcription is more regulated by the toxic effect concentration, but independent of which xenobiotic compound is used. Such biomarkers would be very useful for fast toxicity screening of potentially polluted sites. Further research is, however, needed to validate the usefulness of these transcripts as biomarkers.
Quantitative RT PCR (qPCR) validation
To validate our microarray platform we selected 6 genes with different functions and performed quantitative RT PCR (qPCR). We used the same RNA samples (treated and untreated) that were used in the high effect concentration phenanthrene exposure microarray experiment. YWHAZ
was used as reference gene for normalization, because it was shown to be one of the most stable endogenous genes in F. candida
(de Boer et al., in press [34
]). First, the log2
transformed fold change was calculated for the same pairs of RNA samples (treated vs. untreated) that were hybridized on the microarrays. Then, the average log2
fold change was calculated for each transcript and a significant correlation of 0.943 (Spearman's Rho, p
< 0.01) was found between the qPCR and microarray platforms (Figure ). We can observe that the log2
transformed fold change values are reduced in the microarray data compared to the qPCR (Figure ). This is often observed between microarray and qPCR data and is probably due to loess normalization [35
Figure 3 Correlation between gene expressions measured in Folsomia candida using microarray analysis and qPCR. The average log2 fold change values were used, and each point represents a differentially expressed gene. A significant correlation of 0.943 (Spearman's (more ...)
Figure 4 Comparison between gene expression derived from microarray analysis and qPCR. In the histogram the average log2 fold change is shown for 6 different genes, in response to the high effect concentration of phenanthrene (45.80 mg kg-1 soil). The genes used (more ...)