Otolith δ¹³C values as a metabolic proxy: approaches and mechanical underpinnings

Additional keywords : bioenergetics, field metabolic rate, isotopic mixing models, oxygen consumption.

Metabolic rate is a proxy of the energy used by individual animals and provides a physiological perspective to interpret behavioural ecology. Laboratory-based measurement of fish standard and maximum metabolic rate (SMR and MMR respectively) is a common approach to investigating fish physiology in response to environmental changes (Killen et al. 2007, 2010; Chabot et al. 2016; Metcalfe et al. 2016), but has limited value for explaining fish behaviours in the field. The relevant trait to measure is the field metabolic rate (FMR). Unfortunately, FMR is challenging to measure with conventional methods for estimating metabolic rates in free-living fish (Treberg et al. 2016). A method of in situ oxygen consumption measurement termed field-based respirometry has been tried to investigate the metabolic rate and swimming activity of wild fish in natural environment (Bailey et al. 2002; Farrell et al. 2003). However, the methods do not allow for monitoring realistic energy demands of, for example, prey–predator interactions or recording a time-integrated total metabolic rate in a wild individual. The stable carbon isotope composition of otolith aragonite (expressed as δ¹³C_oto values) may provide an answer to this problem. δ¹³C_oto values have been studied for decades and show potential as a metabolic proxy because the carbon used to precipitate otolith aragonite is drawn from both metabolic (dietary) and ambient water sources. Therefore, the isotopic composition of carbon in otolith aragonite is a weighted average between the isotope compositions of metabolic carbon released from respiration and the dissolved inorganic carbon from ambient water (Kalish 1991a, 1991b; Iacumin et al. 1992; Gauldie et al. 1994; Gauldie 1996; Thorrold et al. 1997; Wurster and Patterson 2003; Wurster et al. 2005; Solomon et al. 2006; Dufour et al. 2007; Tohse and Mugiya 2008). However, despite a clear theoretical basis backed up by consistent observational data, relating variations in δ¹³C_oto values directly to alternative measurements of metabolic rate have proven challenging, partly due to the difficulty of estimating FMR in aquatic organisms. Over the past 10 years, new knowledge and enhanced analytical techniques have been developed, and the potential for using δ¹³C_oto values as a metabolic proxy should be updated. This short paper briefly overviews the use of δ¹³C_oto values in relation to metabolism and illustrates a way forward to improve the methodology, and thereby provide fish ecologists and physiologists with a strong tool to explore some of the current challenges in fish and fisheries ecology. First, we outline the mechanism underpinning carbon incorporation into otolith aragonite and describe analytical approaches to quantify the otolith metabolic proxy. Second, we summarise efforts to describe the relationship between the otolith metabolic proxy and oxygen consumption. Finally, we show the great potential of using otolith metabolic proxy in combination with other otolith-based analyses to answer physiological questions.

Previously, δ¹³C_oto values have often been used as natural tracers for differences in water or diet source between or within individuals (Nonogaki et al. 2006; Ashford and Jones 2007; Schloesser et al. 2009; Elsdon et al. 2010; von Biela et al. 2015; Fraile et al. 2016) to answer various fishery and ecological questions, such as stock identification (Gao and Beamish 1999; Gao et al. 2001; Bastow et al. 2002; Correia et al. 2011; Shen and Gao 2012) and fish movements and migration (Augley et al. 2007; Kimirei et al. 2013; Currey et al. 2014; Javor and Dorval 2014; Gerard et al. 2015). Other applications have associated δ¹³C_oto values with metabolic rate to reveal variation in fish physiological performance and the factors affecting it (Wurster and Patterson 2003; Wurster et al. 2005; Shephard et al. 2007; Hanson et al. 2013). However, we need to carefully evaluate the drivers behind variation in δ¹³C values of dissolved inorganic carbon (DIC) in the water (δ¹³C_DIC) and δ¹³C values of the diet (δ¹³C_diet) in order to use δ¹³C_oto as an accurate estimate of fish FMR.

The δ¹³C_oto value is the weighted average of the isotopic composition of carbon in two main carbon sources, DIC and diet, which are typically very distinct (~1 and –16‰ respectively; Sherwood and Rose 2005; Tagliabue and Bopp 2008). Therefore, the δ¹³C_oto value can be described as the outcome of a two-component mixing model (Schwarcz et al. 1998; Solomon et al. 2006):

where M_oto is the proportion of metabolically derived carbon (from the diet) in otolith carbonate and ϵ_total is the total net isotopic fractionation during carbon exchange between DIC and blood, as well as between the blood and endolymph in which the otolith is formed. DIC uptake is primarily across the gut and gills (Solomon et al. 2006). In contrast, metabolic carbon is released into the blood through the respiration and oxidisation processes, and the rate of oxidation of dietary carbon by definition reflects metabolic rate. Therefore, the weighted average of the isotopic values between DIC and diet is controlled by fish metabolism, and M_oto is viewed and named as a metabolic proxy in the following discussion. Below we review sources of uncertainty in evaluating Eqn 1 and therefore M_oto values.

δ¹³C_DIC values vary between water masses, geographic locations and time, influenced by the release of carbon from the lithosphere, carbon flux exchange within the atmosphere and respiration and photosynthesis from the biosphere. Fresh waters have a wide range of δ¹³C_DIC values among rivers and lakes (for a review, see Bade et al. 2004). For example, δ¹³C_DIC values range from –16 to –8‰ in the Ottawa River basin (Canada; Telmer and Veizer 1999) and from 2.6 to –31‰ among 104 lakes on four different continents (for details, see Bade et al. 2004). The variation of freshwater δ¹³C_DIC values is strongly dependent on geological chemistry, water metabolism and biogeochemical process. By contrast, δ¹³C_DIC values are relatively constant in marine systems, with values that generally vary between 0 and 3‰ on the horizontal spatial scale, and ~1‰ in the vertical gradient (Kroopnick 1985; Tagliabue and Bopp 2008; Schmittner et al. 2013; Becker et al. 2016). In addition to spatial variations, temporal differences in the δ¹³C_DIC values, such as seasonal or annual changes, have been noticed. In areas with strong phytoplankton booms, rates of removal of CO₂ from DIC may exceed diffusion rates of atmospheric CO₂ into surface waters, and preferential uptake of ¹²C into algal cells can cause a temporary increase in δ¹³C_DIC values. Over the past 100 years, oceanic δ¹³C_DIC values have declined continuously because anthropogenic carbon decreases the oceanic δ¹³C_DIC values by CO₂ exchanges between the atmosphere and the ocean. This has been termed the Suess effect and, interestingly, the Suess effect has been recorded in otoliths from Atlantic bluefin tuna (Fraile et al. 2016). According to a recent biogeochemical model, the oceanic δ¹³C_DIC value decreased 0.07‰ per decade from 1860 to 2000, whereas in the recent period from 1970 to 2000 it decreased at a rate of –0.18‰ per decade (Tagliabue and Bopp 2008). The decreasing rate speeds up with time. If we want to use δ¹³C_oto metabolic proxy to compare fish metabolism between decades with an assumed δ¹³C_DIC value but without calibrating the Suess effect, it will overestimate the metabolic rate of fish caught in a recent year. Therefore, we suggest using a model calibration to predict δ¹³C_DIC values or reconstructing δ¹³C_DIC values with given oceanographic parameters for the specific year.

It is possible to acquire δ¹³C_DIC values from the direct measurement of water samples, but it is not always feasible, particularly where studies are based on historical otolith collections or from remote oceanic locations. Nevertheless, there are several ways to acquire δ¹³C_DIC values through modelling predictions. δ¹³C_DIC values can be predicted with a given value of apparent oxygen utilisation (AOU) in the world’s ocean (Kroopnick 1985), and Filipsson et al. (2017) presented a regional relationship between δ¹³C_DIC values and AOU with salinity revision in the Baltic–Skagerrak region at water depths below the halocline:

where S is salinity and AOU is measured in micromoles per kilogram. In addition, a regional multiple linear regression model predicting δ¹³C_DIC values from salinity, temperature and DIC concentrations was used by Becker et al. (2016) to model δ¹³C_DIC values at a depth of more than 1500 m in the North Atlantic Ocean:

where Θ is potential temperature (°C) and DIC is measured in micromoles per kilogram. δ¹³C_DIC values can also be extracted from interpolated spatial models (McMahon et al. 2013) or biogeochemical models (Tagliabue and Bopp 2008; Schmittner et al. 2013). Biogeochemical models take into account both the spatial and temporal factors and yield a global pattern that is necessary for studies on large-scale fish migration and movement.

Distinguishing and measuring the isotopic values of metabolically derived carbon from DIC in blood and endolymph is difficult. Hence, using the δ¹³C values in various tissues, such as muscle, liver and heart, is an alternative approach to estimating the δ¹³C values of metabolically derived carbon, δ¹³C_diet. Tissue δ¹³C values represent a weekly to monthly average of diet signals, depending on the tissue turnover rates among species, the types of tissue and diet preferences (Ankjærø et al. 2012). Isotopic enrichment from diets to tissues is also influenced by various biological and environmental factors, such as growth rate, metabolism and temperature, with typical isotopic offsets between diet and muscle tissue ranging from –1.75 to 3.7‰ (Sweeting et al. 2007). Post (2002) reported an average value of 0.4‰ in carbon isotope enrichment, but Sweeting et al. (2007) suggested out that 1.5‰ is a more appropriate value. These reported values are species averages, but in reality tissue–diet isotopic spacing is a dynamic variable rather than a fixed trait, varying within and among individuals and species depending on physiological status, life history traits and feeding histories. Despite the variation found between species and studies, muscle δ¹³C values provide a reasonable approximation of δ¹³C_diet in the M_oto estimation, because a 1‰ variation of δ¹³C_diet values only contributes a maximum of ~0.005 to the uncertainty in the M_oto term (see details in the following sections). A drawback of using soft tissue is that individual trophic history cannot be reconstructed from these tissues because their δ¹³C values are continuously changing due to variable diet and their metabolic turnover.

δ¹³C values recorded in otolith organic matters have been recently used to indicate diet signals and trophic information in wild fish (Sirot et al. 2017). Compared with muscle tissue, otolith organic materials have the advantage that their δ¹³C values appear close to those of the diet (i.e. show little trophic enrichment; Grønkjær et al. 2013). Moreover, otoliths grow continuously and record ontogenetic information and, in theory, if we can extract the organic materials from otolith aragonite formed at different periods or life stages of an individual, it would be possible to reconstruct that individual’s trophic history. This would allow estimates of FMR through the life of a single individual. However, the proportion of organic material in otoliths is extremely small (<10%), and analysis of individual trophic history is at present only feasible with fish species possessing large otoliths.

Physiology controls carbon isotope incorporation into otoliths and it directly affects the isotopic fractionation factor, ϵ_total. There are three different settings of ϵ_total that have been used in previous studies. Schwarcz et al. (1998) used a value of 2, which was based on the findings of carbon isotope enrichment from ambient fluids (HCO₃^–) to biogenic aragonite carbonates at 5°C (Grossman and Ku 1986). Høie et al. (2003) and Wurster and Patterson (2003) adopted a value of 2.7, which was derived from the inorganic precipitation of aragonite carbonate where the enrichment factor was temperature independent (Romanek et al. 1992). Solomon et al. (2006) used rainbow trout (Oncorhynchus mykiss) and conducted a controlled laboratory experiment with ¹³C-enriched diets and a ¹³C bicarbonate spike in water, finding that ϵ_total was slightly negative (–1.8), but not significantly different from zero. The determination of ϵ_total is still unresolved and remains a source of uncertainty in M_oto measurements (Dufour et al. 2007). Further research is needed to investigate the specific ϵ_total values among species and minimise the bias of M_oto estimations.

Two notable studies have conducted controlled laboratory experiments to estimate the proportion of metabolic carbon in fish otoliths. Solomon et al. (2006) reared juvenile rainbow trout (O. mykiss) in water with different δ¹³C_DIC values and fed them food with different δ¹³C_diet values, and reported a M_oto value of 0.17. Tohse and Mugiya (2008) used the isotope labelling technique on goldfish (Carassius auratus) to estimate the proportion of metabolically derived carbon, which they found to account for 25% of overall otolith carbon composition (M_oto value of 0.25). The percentage of metabolically derived carbon was higher (28%; M_oto value of 0.28) during the day and lower (13–20%; M_oto value of 0.13–0.20) during the night. In most other previous studies, M_oto values estimated from the two-component mixing model fell in the range 0–0.5 (Table 1). High values over 0.5 suggested by Wurster and Patterson (2003) and Hanson et al. (2013) reflect consideration of a range of possible δ¹³C_DIC and δ¹³C_diet values and associated the uncertainty in M_oto estimations.

Table 1.  Estimations of the proportion of metabolically derived carbon (M_oto) in the literature
Information before 2006 is extracted from table 2 in Solomon et al. (2006)

Uncertainty in the δ¹³C_DIC and δ¹³C_diet values determine the precision of M_oto estimations. As an example, we performed a sensitivity test considering the effect on estimates of M_oto rising from the sources of variation in Eqn 1.

We calculated M_oto values corresponding to simulated values of δ¹³C_oto ranging between 0 and –6‰. We allowed δ¹³C_DIC values to vary by 1‰, capturing the likely uncertainty in most marine applications (Kroopnick 1985; Tagliabue and Bopp 2008; Schmittner et al. 2013; Becker et al. 2016). We varied δ¹³C_diet values in a range from –16 to –22‰, reflecting typical isotope values of dietary items for benthic to pelagic fish species in temperate latitudes. The ϵ_total term was assumed to be 0 based on the observations by Solomon et al. (2006). Varying the δ¹³C_DIC term across a range of 1‰ resulted in M_oto values ranging between ~0.05 and 0.35, depending on the δ¹³C_diet and δ¹³C_oto values used in the calculation (Fig. 1a). The s.d. of the M_oto term varied between 0.01 and 0.02, and systematically changed with δ¹³C_oto and δ¹³C_diet values. This suggests variation in the precision of M_oto within the fish functional groups. Fish with more positive δ¹³C_diet values, such as benthic fishes, usually also have higher δ¹³C_oto values (Sherwood and Rose 2003). Higher δ¹³C_oto values mean that the difference between δ¹³C_oto and δ¹³C_DIC values is smaller, and therefore the uncertainty associated with the M_oto term will increase. This is seen in our sensitivity tests, because benthic fish have a systematically higher uncertainty (s.d.) in the M_oto estimations than pelagic fish (with more negative δ¹³C_diet and δ¹³C_oto values; Fig. 1b). Therefore, uncertainty in δ¹³C_DIC values contributes more to estimated M_oto values of benthic or low metabolic rate fishes than to pelagic or higher metabolic rate fishes.

Fig. 1.  Sensitivity tests on the proportion of metabolically derived carbon (M_oto) estimations by 1‰ variations in (a, b) δ¹³C values of dissolved inorganic carbon (DIC) in the water (δ¹³C_DIC) and (c, d) δ¹³C values of the diet (δ¹³C_diet). The mean (a, c) and s.d. (b, d) of the M_oto term were estimated using 1000 Monte Carlo simulations. The 1‰ variation of δ¹³C_DIC is set from 0 to ~1‰ and, each run, a δ¹³C_DIC value is randomly chosen from a uniform distribution U[0,1]. One thousand δ¹³C_DIC vectors were produced, and each δ¹³C_DIC vector was used to estimate M_oto with a given value of δ¹³C_diet and δ¹³C_oto (δ¹³C values of otoliths). As a result, 1000 values of M_oto were generated and calculated as a mean and s.d. (a, b) Across a range from 0 to –6‰ for δ¹³C_oto and from –16 to –22‰ for δ¹³C_diet, we produced a contour plot with a resolution of 100 × 100 grids for M_oto mean (a) and s.d. (b). Similarly, δ¹³C_diet was set from –18 to –19‰ for uniform distribution U[–18,–19]. We followed the same procedure of simulation to estimate M_oto values and make contour plots showing the mean (c) and s.d. (d) with a resolution of 100 × 100 grids across a range from 0 to –6‰ for δ¹³C_oto and from 0 to 3‰ for δ¹³C_DIC.

Similarly, we conducted sensitivity tests on the effect of δ¹³C_diet uncertainty, which was set as a 1‰ variation between –18 and –19‰. δ¹³C_DIC values were set to range between 0 and 3‰, which basically covers δ¹³C_DIC values in the surface ocean around the world (Tagliabue and Bopp 2008). δ¹³C_oto values ranged from 0 to –6‰. As expected, higher M_oto values were accompanied by a higher uncertainty (Fig. 1c, d). However, compared with DIC, a smaller s.d. was observed from the diet sensitivity test even with the same setting of 1‰ variation. The range of δ¹³C_diet values is comparable to those commonly seen in temperate and subtropical marine environments, but in coastal or freshwater ecosystems the uncertainty will be amplified according to a wider range of δ¹³C baseline changes. Moreover, higher uncertainty is expected in the migratory species with habitat changes (δ¹³C_DIC variation), especially if these habitat changes infer diet shifts (δ¹³C_diet variation).

To deal with the variation in both δ¹³C_DIC and δ¹³C_diet values, as well as uncertainty in ϵ_total, a Bayesian framework for isotopic mixing models offers an attractive statistical solution. This method provides the likelihood of a given M_oto term determined using Bayesian methods and considering the uncertainty of the two sources in terms of δ¹³C_DIC and δ¹³C_diet variations. It also facilitates comparing metabolic performance (M_oto term) between fish populations, and is easy to conduct within the well-established R software package MixSIAR (see https://github.com/brianstock/MixSIAR, accessed 21 March 2019; Stock et al. 2018).

Although a δ¹³C_oto metabolic proxy corresponding to fish mass-specific metabolism has been described (Dufour et al. 2007; Trueman et al. 2013, 2016; Chung 2015), there are limited studies describing the scaling of M_oto values with mass-specific oxygen consumption. Here, we introduce a standard bioenergetics model to evaluate the likely relationship between M_oto values and oxygen consumption. The model allocates energy intake into three compartments: metabolism, growth and waste (Treberg et al. 2016; Deslauriers et al. 2017):

where SDA is specific dynamic action. SMR can be predicted by measuring experienced temperature and body mass of the fish according to the metabolic theory of ecology (MTE; Brown et al. 2004):

where the B₀ is the normalised constant, BM is the body mass and T is temperature in kelvin; α is the allometric scaling exponent of body mass, which follows the three-quarters power law in MTE (as –0.25 for mass-specific metabolism; Brown et al. 2004) but was found to be 0.79 for teleost fishes (Clarke and Johnston 1999; Clarke 2006).

For wild-caught fishes, experienced temperature can be estimated from otolith δ¹⁸O values (e.g. Shirai et al. 2018 and references therein). Second, otolith increment analysis provides a chronological record of body mass. A lifelong history of body mass can be reconstructed from von Bertalanffy growth curves with given age inferred by the otolith increment numbers. Otherwise, it is possible to back-calculate fish body mass from fish length, obtained from otolith back-calculations (Campana 1990). Using these methods, several previous studies present expected relationships between δ¹³C_oto or M_oto values and temperature (Kalish 1991a; Høie et al. 2004a; Gao et al. 2010) and body mass (Trueman et al. 2013; Chung 2015).

The M_oto value is regarded as a proxy of FMR, corresponding to the sum of SMR, activity and SDA. To examine the relationship, we obtained M_oto values as well as fish length data and the otolith δ¹⁸O values of Atlantic cod (Gadus morhua) extracted from Jamieson (2001) and Jamieson et al. (2004). Fish lengths and otolith δ¹⁸O values were used to reconstruct body mass and experienced temperature of fish, which are critical for metabolic rate estimations (Table 2). The three metabolic compartments (i.e. SMR, activity and SDA) are estimated theoretically with the body mass and temperature by Fish Bioenergetics (ver. 4.0, see http://fishbioenergetics.org, accessed 21 March 2019), a package in R programming software (Deslauriers et al. 2017). The metabolic rate of the sum of the three metabolic compartments is expressed as the mass-specific oxygen consumption rate. The M_oto term increased significantly with mass-specific oxygen consumption (Fig. 2). Our regression trend indicated a positive and linear relationship between the M_oto term and oxygen consumption, but gave an unrealistic M_oto value (0.20) when the oxygen consumption was close to zero (Fig. 2). Considering that the M_oto term is constrained by both upper (~0.5) (Table 1) and lower boundaries (0), this may imply that the relationship is not a simple linear regression (Kalish 1991a), but an exponential decay model in increasing form (Chung et al. 2019). It is critical that the relationship between M_oto values and oxygen consumption should be widely investigated, especially across species. The functional form, including the upper limit of M_oto values, may vary between species according to their life history traits and physiological regulations. Nevertheless, it is believed that the relationship between M_oto and oxygen consumption rate among species will provide valuable information that will enhance progress in the research field of fish physiological ecology.

Table 2.  Individual information used to construct Fig. 2
Fish length, δ¹³C_oto (δ¹³C values of otoliths), δ¹⁸O_oto (δ¹⁸O values of otoliths) and the proportion of metabolically derived carbon (M_oto) values are extracted from Jamieson (2001). Fish weight is derived from the length–weight relationship from FishBase (R. Froese and D. Pauly, see www.fishbase.org): Weight = 0.0071 × Length^3.08. Temperature is reconstructed following the δ¹⁸O_oto–temperature equation given by Høie et al. (2004b) and seawater δ¹⁸O is set as –2‰. Oxygen consumption is estimated with body mass and temperature according to Deslauriers et al. (2017)

Fig. 2.  Relationship between the proportion of metabolically derived carbon (M_oto) and reconstructed oxygen consumption rate of Atlantic cod (Gadus morhua; y = 0.20+3.03 × 10^–4x; n = 39, t = 11.17, P < 0.01). M_oto values were extracted from Jamieson (2001). Oxygen consumption is theoretically estimated with known body mass and environmental temperature according to a bioenergetics model, and the calculations and parameter values follow Deslauriers et al. (2017). Individual information and data used in the relationship are given in Table 2.

Knowledge of fish energy allocation between metabolic compartments (SMR, SDA and activity) may increase our understanding of their behavioural adaptation to environmental changes. The use of the otolith metabolic proxy could be instrumental in gaining this knowledge. For example, Sherwood and Rose (2003) found that δ¹³C_oto values relate to the aspect ratios of the caudal fin of fish, which is associated with swimming form and activity. Solomon et al. (2006) further analysed these data to provide a regressed trend of M_oto values with the aspect ratios of the caudal fin of fish:

where K_caud is the aspect ratio of the caudal fin. The relationship revealed the potential of using M_oto values to evaluate activity but without a link to fish swimming speeds and oxygen consumption rate. Thus, an experimental design in which the activity level of fish is manipulated (e.g. by enforcing different swimming speeds) may give direct evidence of the effect of activity metabolism on M_oto variations.

Otolith accretion and opacity are regulated by metabolic processes. The otolith annual pattern with alternating opaque and translucent bands is likely synchronised with energy acquisition and usage (Grønkjær 2016). At the microstructural level, increment widths have been found to relate linearly to SDA (Armstrong et al. 2004). As a general assumption, SDA is proportional to energy intake, and corresponds to 0.1- to 0.4-fold the total assimilated energy (Jobling 1981; Soofiani and Hawkins 1982; Kiørboe et al. 1987; Wieser and Medgyesy 1990). However, in wild fishes, it is difficult to determine SDA owing to uncertainties in meal size, feeding frequencies and postprandial durations. As an alternative, otolith increment analysis combined with the otolith metabolic proxy may make SDA determination possible. Furthermore, a modelling framework based on Dynamic Energy Budget (DEB) theory can be used to try to reconstruct individual and otolith growth history with known temperature and otolith opacity patterns (Fablet et al. 2011; Pecquerie et al. 2012). In this modelling framework, otolith growth and opacity are defined by two energy fluxes in the metabolism (i.e. maintenance and fish growth; Fablet et al. 2011). This means that the metabolic information of SMR and SDA, which is associated with maintenance and growth energy fluxes in the DEB model, can be acquired by analysing the optical properties of the otolith microstructure. Multiple approaches combining the δ¹³C_oto metabolic proxy, otolith δ¹⁸O analyses, microstructure analyses and the DEB model hold great potential when it comes to investigating and reconstructing individual life history in response to environmental changes.

In this paper we have illustrated three perspectives on otolith δ¹³C metabolic proxy: (1) how to obtain the parameters used to estimate M_oto values according to a two-component mixing model; (2) the several unanswered questions that should be considered when using the otolith metabolic proxy; and (3) the great potential of using the otolith δ¹³C metabolic proxy to study fish physiological ecology in combination with other valuable and well-established approaches. Despite the considerable efforts needed to acquire the necessary parameter values across species, the novel approach of the δ¹³C_oto metabolic proxy shows great promise with regard to allowing us to track the ontogenetic and environmental effects on individual fish physiology, and thereby removes a major obstacle to understanding and predicting the performance of free-ranging wild fish.

The authors declare that they have no conflicts of interest.

This review paper is supported by the European Commission Marie Skłodowska-Curie Individual Fellowships (‘OTOLOG’ project, 707481) and the European Regional Development Fund (Interreg IVa, ‘MarGen’ project).