With 29% of wild fish stocks exploited at biologically unsustainable levels, and global capture fisheries almost at same yield in the last 30 years, marine aquaculture has undergone a rapid expansion worldwide, being now the major source of increasing fish food supply, with large potential for economic growth (FAO, 2016). Stinging jellyfish proliferations can affect marine fish farming (Baxter et al., 2011a; Rodger et al., 2011a; Purcell et al., 2013), with negative repercussions on highly productive coastal and offshore systems.

In recent years, facilities from Asia, Australia, North and South America reported sudden fish mortalities as consequences of episodic blooms of different jellyfish species (Palma et al., 2007; Doyle et al., 2008; Willcox et al., 2008; Purcell et al., 2013). Physical contact of sensitive fish tissues (skin and gills) with conspicuous semeostomeae scyphozoan jellyfish Pelagia noctiluca and Aurelia aurita has been identified as the primary agent of repeated mass fish deaths (in 2007, 2010, 2013, and 2014) at salmon facilities in northern Europe (Doyle et al., 2008; FIS, 2014; Marcos-López et al., 2016). Stinging jellies are washed up on fish cages by currents, and can be either broken in pieces or their early stages may be small enough to enter the net cages and get in touch with caged fishes. In parallel, a few studies reported the impact of less conspicuous planktonic stingers on mariculture, with best evidence on the hydrozoan siphonophore Muggiaea atlantica and the microscopic hydromedusae Solmaris corona and Phialella quadrata as determinants of fish mortalities in Irish and Scottish fish farms (Baxter et al., 2011a; Fitridge and Keough, 2013; Purcell et al., 2013).

Additional issues arise for aquaculture from biofouling. Together with the restriction of water flow through net occlusions, the development of large hydrozoan polyp colonies on cage nets eventually lead to seasonal release of free-living propagules inside the cages, such small jellyfish (e.g., Obelia spp.), or the larvae of progenetic hydroids (e.g., the tubulariid Ectopleura larynx; Baxter et al., 2012). Tiny and almost transparent medusae, siphonophores, and hydrozoan larvae may easily remain unnoticed to aquaculture operators, but they can develop high density blooms, be inhaled by fish, and inflict severe damage due to nematocyst discharge and consequent tissues envenomation (Fosså et al., 2003). A range of morphological injuries caused by contacts with cnidarian stinging cells have been described on several cultured fish species (Baxter et al., 2011b; Bosch-Belmar et al., 2016b; Marcos-López et al., 2016). Available evidence shows that a few hours of exposure to stinging cells may lead to severe damage at cellular, tissue and systemic levels, including changes in fish metabolic performances (Bosch-Belmar et al., 2016a).

Gill disorders are considered to be a major problem for fish aquaculture and are usually referred to any of different causes, including jellyfish, phytoplankton, parasites, bacteria, and viruses (Rodger, 2007). To date, several marine fish farms in northern European seas have started plankton monitoring as standard preventative tool against potential impacts. On the contrary, little or no plankton monitoring activities have been launched in Mediterranean mariculture facilities.

Since 2009, two off-shore fish farms located in southern Spain (Almería and Málaga) suffered repeated mortality events of caged European sea bass (Dicentrarchus labrax), for which no causative agent was clearly identified. Specifically, bacterial and known viral diseases, parasites or hypoxic events in the fish cages were excluded by veterinary controls. Injured sea bass showed clinical signs similar to those described by Rodger et al. (2011a) in farmed salmons exposed to jellyfish blooms in the Irish Sea: lethargic behavior, atypical swimming close to the surface, visible respiratory distress, interrupted feeding, and dead or moribund individuals. For this reason, in 2012 a combined plankton vs. fish gill survey was launched at both aquaculture facilities to [I] disclose the spatial and seasonal variability of the plankton community accessing the fish cages; and [II] test the hypothesis of a direct relationship between the occurrence of potentially harmful plankton (i.e., stinging species) and fish health.

To our knowledge, this is the first survey on potentially harmful plankton species carried out to investigate their role as hidden causative agents of farmed fish mortalities in Mediterranean aquaculture facilities. Farmed sea bass mortalities in the two facilities were more clearly related to spring and autumn water temperature and the peaks of abundance of three hydrozoan species, the siphonophores M. atlantica and M. kochii and the actinula larvae of the fouling hydroid colonies of E. larynx.

Muggiaea atlantica was previously identified as a potentially harmful species for marine aquaculture in northern Europe, together with S. corona and P. quadrata (Baxter et al., 2011a). Those species were also identified in samples from the two Mediterranean facilities but statistical analyses showed non-significant correlation of both hydromedusae densities and fish mortalities (Z₁ = −0.01, p = 0.99 and Z₁ = −0.002, p = 0.10 for S. corona and P. quadrata, respectively). High abundances of Muggiaea spp. eudoxid stages were recorded several times in both fish farms. Each calycophoran siphonophore polygastric stage asexually produces several eudoxids, a sexually-reproductive stage that feeds with stinging tentacles, as is the polygastric stage. High abundances of eudoxids could be detrimental for farmed fish health because hundreds of cnidocytes are present in each tentacle and may severely damage sensitive fish tissues. Mortality recorded in November-December 2013 in the Málaga facility coincided with a Muggiaea spp. reproductive event, reaching high abundances of polygastric colonies and tens of eudoxids in the water column.

The appearance of the microscopic actinula larvae of E. larynx were also correlated with fish kill events in the Almería facility. The actinulae were present when temperatures were low, from January to May and from November to December in both facilities. Over the course of the monitoring, several reproductive periods were observed, with the released actinula larvae reaching very high densities in the water during January 2013 (>200 ind m⁻³) when thousands of fish died (Supplementary Figure 1B). This hydroid usually forms part of cage fouling community in the North Sea where is a severe problem for fish farms (Guenther et al., 2010; Carl et al., 2011; Baxter et al., 2012) and also in the Mediterranean Sea aquaculture facilities (facilities staff, pers. obs.). Guenther et al. (2010) demonstrated that after washing to clean the nets, this species regrows and occludes the net apertures rapidly. Underwater cleaning of the net cages resulted in higher numbers of E. larynx actinulae and polyps in the water column that could significantly affect caged fish health by injuring gill tissue after contact (Carl et al., 2011; Baxter et al., 2012).

The scyphozoan mauve stinger P. noctiluca is the most abundant and one of the most venemous jellyfish both in the Mediterranean Sea and in eastern Atlantic waters (Russell, 1970), with the potential to reproduce all year long in some areas, such as the Strait of Messina, Italy (Milisenda et al., in press). Due to its envenomation potential and large population outbreaks, it is renowned as one of the most dangerous species for the European mariculture fish farms (Marcos-López et al., 2016). In zooplankton samples, P. noctiluca ephyra stages, deriving from recent metamorphosis of planula larvae, were observed at densities up to 15 ind m⁻³ several times in both facilities (mainly in February, June, and October). Adult P. noctiluca medusa were not collected by the vertical hauls during the monitoring period, however their occurrence at low density (≤1 ind m⁻³) were visually recorded by the facility operators (unpublished data). In preceding years, the monitored facilities endured large aggregations of P. noctiluca adult medusae, leading in 2011 to severe fish mortalities (10 tons) with associated economic consequences (Bosch-Belmar et al., 2017).

Obelia dichotoma is another common hydrozoan species in Mediterranean coastal areas (Bouillon et al., 2004; González-Duarte et al., 2016). This species was observed in the Almería and Málaga facilities throughout the year, both as hydroids and medusae forming part of biofouling and zooplankton communities. The small hydromedusae were found at high densities in the Almería facility (maximum of 197.35 ind m⁻³) when fish stock mortality was recorded in April 2012. However, despite its high concentrations, a significant relationship of this species with the recorded fish kill events was not found.

Harmful algal blooms produced by dinoflagellates, diatoms, silicoflagellates, raphidophytes, and prymnesiophytes (Granéli and Turner, 2006), are often associated to fish mortality episodes worldwide (Treasurer et al., 2003; Burridge et al., 2010). In our study, some phytoplankton taxa previously related with farmed fish kill events were recorded at high densities in both fish farms (Cembella et al., 2002; Treasurer et al., 2003). Chaetoceros is a colonial diatom genus characterized by long setae, which either can clogging gills causing asphyxia or can penetrate the gill tissues causing histological damages (Smayda, 2006). Fish kills attributed to Chaetoceros occurred in Canada, USA and Scotland where abundances of 10⁵ cells l⁻¹ were recorded (Rensel, 1992; Treasurer et al., 2003). In analyzed samples, we recorded abundances of the same order of magnitude only during one farmed fish mortality event recorded in April 2012 in the Almería fish farm, therefore Chaetoceros should be taken into account as plankton species that may had contributed to fish mortality events.

On this respect, it is worth noting fish mortalities in the two Spanish facilities were detected in months characterized by either large peaks of abundance of renowned harmful species (Muggiaea spp., E. larynx) or by the coincident abundance of different plankters not always recognized as harmful taxa (including phytoplankters, P. noctiluca ephyrae, small hydromedusae) (Supplementary Figures 1, 2). This evidence may lead to the hypothesis that concurrent planktonic cnidarians (and even organisms with long stiff bristles, such as diatom phytoplankton cells) may also play a potential impact as guild (Piraino et al., 2002; Boero et al., 2008) because of their cumulative stinging potential, rather than as single species. Overall, low-medium densities of high number of mild stingers might produce similar effects as high concentrations of a single powerful stinger, both in open water and in fish farms. What makes the difference between low, medium, or highly harmful plankters? For cnidarians, it has been suggested that jellyfish stings of some species maybe more harmful than others not only because of chemical differences among venom types but also due to differential cnidocyst structure, with particular reference to the length of different stinging filaments, which may penetrate prey tissues at different depths (Purcell, 1997; Kitatani et al., 2015). If this would be confirmed as a general rule, it might be possible to predict the potential toxicity and harm for fish farms of cnidarian plankters (jellyfish, larvae, hydroid propagules) by investigating their cnidome composition, abundance, and cnidocyst morphology.

Total cnidarian densities were related with temperatures, with reduced densities during summer period, especially for species with a pelago-benthic life cycle. Previous observations revealed a typical summer impoverishment of the Mediterranean Sea filter feeders, including the disappearance of small hydrozoan colonies, entering in a resting phase as dormant creeping stolons (Boero and Fresi, 1986; Bavestrello et al., 2006; González-Duarte et al., 2013). In the Almería farm, the fish mortalities were significantly related to cold months (Figure 5). In the Málaga facility, even though it was not possible to perform regression analyses, the unique recorded mortality event occurred at December 2013, coincident with high cnidarian densities (up to 30 ind m⁻³).

Gill scoring presented a significant positive relationship with cnidarian abundances, and demonstrated the existence of severe gill disorders related with high cnidarian densities even when fish mortalities were not recorded. For example, in April 2014 in Almería and May 2014 in the Málaga fish farm, gill scores were 6 ± 1.10 and 9.20 ± 0.60, respectively, and jellyfish occurred in medium-high densities (19.71 ± 3.07 ind m⁻³ in the Almería pen and 48.80 ± 10.79 ind m⁻³ in the Málaga facility) but no fish mortalities were reported in these periods. Experimental studies with the scyphozoan A. aurita showed gill epithelium recovery required 2 weeks after even brief (10 h) contact between jellyfish and salmonids (Baxter et al., 2011b). Equally, laboratory experiments carried out by Bosch-Belmar et al. (2016b), demonstrated partial recovery of sea bream gill tissue after 3 weeks from fish exposure to medium densities of P. noctiluca. These gill disorders used to pass unnoticed, but the consequences of this stressful situation (repeated contact with jellyfish) could be severe, and even if fish gill epithelium could recovered, fish metabolic performances may be affected and in the long term also fish growth could be compromised (Rodger et al., 2011a,b; Bosch-Belmar et al., 2016a).

This study allowed to identify some blooming and potentially injuring hydrozoan jellyfish species occurring in the monitored mariculture facilities, and to determine the range of temperatures boosting jellyfish proliferations. The optimal thermal values are typical of spring-autumn months, when small sea bass juveniles (15 g in weight) are usually transferred in cages at sea. As a consequence, most mortalities were recorded in different stocks with fish weights between 20 and 70 g. The interaction of biotic and abiotic factors together with the changing fish metabolism adapting to new conditions could create severe stressful situation for fingerling fish. Reduction of fish stock numbers or variation in juveniles' cages introduction times could represent some of the preventive actions to minimize the impact of jellyfish blooms on farmed fish health. Due to the growth of the aquaculture sector and the increased frequency of jellyfish blooms in some coastal waters, the negative interactions of stinging jellyfish on caged finfish is expected to become a substantial issue producing highly relevant economic losses (Purcell et al., 2013). Regular plankton monitoring will be crucial to gather site-specific information on potential harmful species, including their seasonal occurrence and densities, and predict potential impacts of harmful plankters in fish gill disorders and mortalities.

MB, VF, and SP conceived the study. Plankton and fish samplings were performed by MB and fish farms staff. Histological analysis of gill samples was carried out by AG. Zooplankton quantification and identification was performed by MB and VT, while phytoplankton screening and identification by CT and SA. MB and GM performed all statistical analysis and figures. MB, VF, and SP wrote the manuscript and all co-authors commented on and provided edits to the original manuscript.