Transcriptional biomarkers of toxicity - powerful tools or random noise? An applied perspective from studies on bivalves

Aquatic organisms are constantly at risk of being exposed to potentially harmful chemical compounds of natural or anthropogenic origin. Biological life can for instance respond to chemical stressors by changes in gene expression, and thus, certain gene transcripts can potentially function as biomarkers, i.e. early warnings, of toxicity and chemical stress. A major challenge for biomarker application is the extrapolation of transcriptional data to potential effects at the organism level or above. Importantly, successful biomarker use also requires basal understanding of how to distinguish actual responses from background noise. The aim of this thesis is, based on response magnitude and variation, to evaluate the biomarker potential in a set of putative transcriptional biomarkers of general toxicity and chemical stress.
Specifically, I addressed a selection of six transcripts involved in cytoprotection and oxidative stress: catalase (cat), glutathione-S-transferase (gst), heat shock proteins 70 and 90 (hsp70, hsp90), metallothionein (mt) and superoxide dismutase (sod). Moreover, I used metal exposures to serve as a proxy for general chemical stress, and due to their ecological relevance and nature as sedentary filter-feeders, I used bivalves as study organisms.
In a series of experiments, I tested transcriptional responses in the freshwater duck mussel, Anodonta anatina, exposed to copper or an industrial waste-water effluent, to address response robustness and sensitivity, and potential controlled (e.g. exposure concentration) and random (e.g. gravidness) sources of variation. In addition, I performed a systematic review and meta-analysis on transcriptional responses in metal exposed bivalves to (1) evaluate what responses to expect from arbitrary metal exposures, (2) assess the influence from metal concentration (expressed as toxic unit), exposure time and analyzed tissue, and (3) address potential impacts from publication bias in the scientific literature.
Response magnitudes were generally small in relationship to the observed variation, both for A. anatina and bivalves in general. The expected response to an arbitrary metal exposure would generally be close to zero, based on both experimental observations and on the estimated impact from publication bias. Although many of the transcripts demonstrated concentration-response relationships, large background noise might in practice obscure the small responses even at relatively high exposures. As demonstrated in A. anatina under copper exposure, this can be the case already for single species under high resolution exposures to single pollutants. As demonstrated by the meta-regression, this problem can only be expected to increase further upon extrapolation between different species and exposure scenarios, due to increasing heterogeneity and random variation. Similar patterns can also be expected for time-dependent response variation, although the meta-regression revealed a general trend of slightly increasing response magnitude with increasing exposure times.
In A. anatina, gravidness was identified as a source of random variability that can potentially affect the baseline of most assessed biomarkers, particularly when quantified in gills. Response magnitudes and variability in this species were generally similar for selected transcripts as for two biochemical biomarkers included for comparison (AChE, GST), suggesting that the transcripts might not capture early warnings more efficiently than other molecular endpoints that are more toxicologically relevant. Overall, high concentrations and long exposure durations presumably increase the likelihood of a detectable transcriptional response, but not to an extent that justifies universal application as biomarkers of general toxicity and chemical stress. Consequently, without a strictly defined and validated application, this approach on its own appears unlikely to be successful for future environmental risk assessment and monitoring. Ultimately, efficient use of transcriptional biomarkers might require additional implementation of complementary approaches offered by current molecular techniques.

Organisms in the environment constantly encounter various natural and man-made chemicals. Many of the regular encounters are more or less safe, but depending on the intensity of the exposure, all chemicals have the potential to be harmful in different ways. Thus, in the research field of environmental toxicology, there are different ways to address questions concerning potential negative effects from chemicals. One such approach is the use of so-called biomarkers. In short, a biomarker is a selected biological feature or measure that (1) can be measured in or on an organism, and (2) change either upon encounters with chemicals, or from the harmful effects that can arise. Since many unwanted effects from chemicals originate at the cellular level, various molecular responses are often considered as potential biomarkers to anticipate harmful effects on the organism.
Genes are pieces of biological information carried by all organisms, and commonly contain instructions to produce proteins with functions in the cell. The first step of gene expression is known as transcription, and occurs when a gene, i.e. a certain sequence in the organism’s DNA, is transcribed to temporary working copies of itself, i.e. gene transcripts made of RNA. These are in turn used as templates for cellular production of the protein that corresponds to the particular gene. Although the DNA itself is constant, the levels of different gene transcripts will vary over time depending on the current needs of the organism. For instance, certain genes are known for being involved in protective actions against chemical stressors, and are expected to be expressed at higher levels in organisms under chemical stress. By responding to exposures that might eventually be harmful, transcripts of these genes are believed to serve as early warnings before actual harm arises.
In this thesis, I address six selected stress genes and evaluate whether their gene transcripts can be used as biomarkers of toxic chemicals in general. Specifically, I used the freshwater duck mussel (Anodonta anatina) in a series of laboratory experiments, to test how gene transcription was affected by copper and an industrial wastewater effluent. In these studies, I addressed how sensitive and how robust the gene transcripts were for consideration as general biomarkers. In addition, I performed a study on published scientific literature, a so-called systematic review and meta-analysis, to address similar questions on a larger scale, i.e. on bivalves and metals in general.
The transcript response signals were in general both weak and variable, which limits the potential use as biomarkers. The experiments and meta-analysis together suggested that we cannot necessarily expect detectable biomarker signals simply because mussels are exposed to toxic metals. I could show in duck mussels that responses commonly increased with increasing concentrations of copper, i.e. increasing chemical stress, but such relationships were generally not detected across species and metals in the meta-analysis. In general, the background noise was so large that it risked obscuring the biomarker signal, even in cases where gene expression actually changed as a response to the chemicals. Overall, random variation both within and between species and exposures will likely limit the ability to detect biomarker signals.
One of many potential sources of variability in duck mussels was gravidness. When gravid, as occurs in nature from late summer to early spring, the background noise can be expected to increase for most of my tested biomarkers. This was however not limited to gene transcripts. To put the transcripts into a wider perspective, I also tested two additional enzymatic biomarkers. These showed similar variability and response signals as the transcripts, both in general and with regards to gravidness. On one hand, this could mean that the selected transcripts are neither substantially better nor worse than established biomarkers at detecting chemical stress. On the other hand, there would then be no advantage in using transcripts as biomarkers, as compared to molecular responses that are easier to interpret from a toxicological perspective.
In summary, clear biomarker signals could possibly be expected under specific and validated test conditions (e.g. avoiding gravidness, targeting specific species and chemicals), and in particular at high concentrations and/or after long exposure durations. In practice, conditions for environmental monitoring and risk assessment are however rarely optimal, which can critically limit the universality of these biomarkers based on general stress genes. Overall, the best way for practical implementation of transcripts in future environmental risk assessment might be a complementary approach of validated biomarkers, models that link transcripts to harmful effects on the organism, and techniques that early on can detect general changes in the organism’s transcriptional patterns.