Diverse and highly productive epiphytic assemblages composed mainly of microscopic algae are attached to the seagrass and macroalgae leave and benefit from this relationship by gaining a structure on which to grow and by consuming nutrients that the vegetation releases (Hauxwell et al. 2001, Perez et al. 2008).

The host plants play a key role in shaping the composition of the epiphytic community structure (Johnson et al. 2005), since the phenological parameters such as leaf length, leaf area and leaf area index of some magnoliophytes increase during the warm seasons (Mabrouk et al. 2009). The structure of epiphytic communities is also influenced by factors such as the age of the leaf (Mazzella et al. 1994), the seasonal cycle of macroalgae (Gambi et al. 1992) and grazing (Mirella et al. 2012).

The composition and abundance of epiphyte communities can also be influenced by abiotic factors such as irradiance, temperature, salinity and inorganic nutrients. Temperature contributes significantly to the temporal variation of diatom epiphytes (Johnson et al. 2005) and epiphytic dinoflagellates (Armi et al. 2010), and salinity is an important factor in the distribution of the microepiphyte community (Johnson et al. 2005). The influence of other factors such as hydrodynamics and light intensity on the development of epiphytes has also been documented (Nesti et al. 2009).

Epiphytic microalgae may include many toxic species that can damage fisheries and cause human health hazards. In recent years, the proliferation of toxic epibenthic species appears to be expanding on a global scale, probably due to either global climate change (Hallegraeff 2010) or anthropogenic impacts such as eutrophication and transfer of ballast water (Hallegraeff et al. 2010).The increase in studies in various ecosystems all around the world over the past few decades could also explain their apparent global proliferation (Van Dolah 2000, Maso and Garcés 2006).Few studies (Bomber et al. 1989) have focused on the effect of the substrata on the growth of toxic epiphytic species, and their relationships with their hosts. Although the distribution of epiphytes has been shown to depend largely on their host (Cohu et al. 2013, Accoroni et al. 2016a), the question that still needs to be addressed is whether there is an affinity between a toxic species and a given substrate. Furthermore, Aligizaki and Nikolaidis (2006) highlighted correlations between the abundance of these toxic species in the water column and their abundance on macrophytes. Since most environmental monitoring programmes have focused on water column concentration of toxic species, understanding the distribution patterns of epiphytic species on macrophytes should be particularly useful for the design of monitoring programmes.

This study aims to characterize the temporal variability of epiphytic microalgae on different substrates (magnoliophytes and macroalgae) and in the water column, with a special focus on epiphytic toxic dinoflagellates. We particularly wish to examine the following hypotheses:

1) Do the diversity and abundance of epiphytic microalgae vary between substrata and environmental conditions?
2) Is there a relationship between the concentrations of the toxic species present on specific substrata and in the water column?
3) Do toxic epiphytic species show different distribution patterns on the leaves of Posidonia oceanica?

Detailed nutrient mean concentrations are reported in Table 1. These mean concentrations were calculated taking into account the number of samples of each substrate during the study period. The results showed that nutrient concentrations varied from one substrate to another. Zostera noltii and Asparagopsis armata exhibited more variability than the other substrates.

The concentrations of nitrite and nitrate were high in magnoliophytes, with a maximum recorded for P. oceanica, whereas in macroalgae the maximum was recorded for Cystoseira mediterranea. Ammonium concentrations showed a different trend, with the highest values being recorded in macroalgae (Table 1). The highest concentrations of phosphorus and silicate were recorded for Dictyopteris membranacea (Table 1).

During the study period, a total of 31 microalgal taxa were identified (Table 2). Three algal groups were represented in our study area: namely, Baccillariophyceae (19 species), Dinophyceae (9 species) and Cyanophyceae (3 species). The species number differed according to the host species (Table 2). The highest species diversity was recorded on P. oceanica and A. armata. Only a few species were recorded on macrophytes (Table 2).

On magnoliophytes, the highest abundances of diatoms were recorded on P. oceanica leaves during spring. This class was also rather important in spring and summer, with abundance exceeding 60% and reaching over 96%of total epiphyte microalgae on macroalgae, especially on Asparagopsis armata (Table 3).

Dinoflagellates accounted for the highest abundances on P. oceanica in winter. This class also showed high abundances on Cymodocea nodosa (Table 3). On macroalgae, Cystoseira mediterraneas howed the highest abundances of dinoflagellates. Comparing to diatoms, dinoflagellates showed a low abundance on Padina pavonica and this group was almost absent on Asparagopsis armata.

The diversity index (H′) of epibenthic species was very high on P. oceanica, P. pavonica, H. stipulacea and D. dichotoma (Table 3). The highest diversity index (H’) was recorded on P. oceanica for magnoliophyte (H’=3.656) and on D. dichotoma for macroalgae (H′=3.835) during spring, and the lowest was recorded on Z. noltii during the same season (H′=1.825) (Table 3).

The co-inertia plot (Fig. 2A) illustrated close relationships between the composition of phytoplankton communities and the water properties above the ten sampling substrates. The overall model explained 33% of the total variation (permutation test, p=0.02, 1000 replicates). This variation was due to microphytoplankton taxa (20%) and to physical and chemical variability (18.52%) (Fig. 2B). Posidonia, Cystoseira and Halophila substrates showed close links between nitrite and nitrate and the phytoplankton species, as was illustrated by the position of Ostreopsis, P. lima and total dinoflagellates (Fig. 2A). In contrast, Dictyota dichotoma, Zostera noltii and Asparagopsis armata substrates were surrounded by the numerically dominant Coolia monotis, total phytoplankton and diatoms (Fig. 2A).

The toxic and potentially toxic dinoflagellates were mostly concentrated on P. oceanica, where they represented about 65% of the total epiphyte microalgae, followed by Cystoseira mediterranea and Cymodocea nodosa.

On P. oceanica, the occurrence frequency of toxic and potentially toxic dinoflagellates was high both on substrate and in the water column (Table 4). The highest occurrence of toxic dinoflagellates was observed for P. lima, with 66.83% on macrophytes and 54.26% in the water column (Table 4). The frequency was low for Ostreopsis, with only 1.84% on macrophytes and 9.30% in the water column (Table 4). C. monotis did not exceed 10% in the water column and was about 2.76% on macrophytes (Table 4). The other epiphytic species were barely observed within the water column and on macrophyte leaves, except for the epiphytic Polykrikos kofoidii (27.76%) (Table 4).

The abundance of the epiphytic dinoflagellates Ostreopsis cf. ovata was higher on magnoliophytes than on macroalgae, especially for Cymodocea nodosa, on which it reached 22.73% of the total dinoflagellates in autumn (Table 5). This toxic species did not show a significant difference in concentrations between the studied substrata, although the concentrations reached 0.5 10³ cells g^–1 FW on Cymodocea nodosa, 10³ cells g^–1 FW on Posidonia leaves in February and September (Fig. 3A), and relatively high abundances on Padina pavonica in September and November (Fig. 3C). A significant positive correlation (P<0.05, R²=0.30) was observed between the species concentrations on Padina pavonica and in the water column above this macroalga (Fig. 3C, D).

The abundance of P. lima on Posidonia leaves accounted for 62.59% of the total dinoflagellates (Table 5). During the sampling period, P. lima was the most dominant and frequent species on magnoliophytes as well as on macroalgae. This species significantly accumulated on P. oceanica (P<0.05), with the highest concentrations, exceeding 10⁴ cells g^–1 FW, being observed from February to May (Fig. 3A). On macroalgae, this species did not show significant variations between substrata. It exceeded 8.3 10³ cells g^–1 FW during March on Cystoseira mediterranea (Fig. 3E), whereas on Padina pavonica and on Halophila stipulacea, concentrations did not exceed 2 10³ cells g^–1 FW (Fig. 3C, G). In contrast to other toxic species, P. lima showed significant variations in its abundance on P. oceanica over time (P<0.05). This species was also present in the water column above the P. oceanica bed, with a concentration reaching over 10³ cells L^–1 (Fig. 3B). A significant correlation (P<0.005, R²=0.80) was pointed out between P. lima concentrations on different substrata and in the water column.

C. monotis was also present on different substrates, with a maximum of 37.5% recorded on Dictyota dichotoma (Table 5). The monthly abundance of C. monotis showed no significant difference between the studied substrata (P>0.05) (Fig. 3). This species showed generally low concentrations in the water column sampled near Posidonia, Cystoseira and Halophila (Fig. 3F, H). P. micans showed the highest concentrations on P. oceanica but its abundances were rather low on other substrates (Table 5). Other substrata were barely observed in our study area. P. lima and C. monotis were the main species present on these substrates, where the maximum concentration was approximately 1.7 10³ cells g^–1 FW on Asparagopsis armata during April.

In spring, when the maximum of substrates were available, the clustering analysis of epiphytic species similarity between different substrates showed four groups (Fig. 4). The first cluster was composed of C. mediterranea, which hosted C. monotis, P. lima and P. micans (Fig. 4). The second cluster was composed of Z. noltii, H. stipulacea, A. armata and D. dichotoma, which hosted only two species, P. lima and C. monotis (Fig. 4). The third cluster was composed of P. pavonica, hosting mainly P. lima and C. monotis (Fig. 4). Finally, the last cluster was composed of P. oceanica, which showed high dissimilarity to the other substrates, hosting the different epiphytic dinoflagellates with a dominance of P. lima (Fig. 4).

According to the SNK test results, the distribution of Ostreopsis cf. ovata on Posidonia leaf revealed three groups (a homogeneous subset) (Table 6). The highest abundance was marked on the inner face of the apical and the middle parts of the leaf. This toxic species was also present with a relatively high abundance on the inner face of the basal part. On the other hand, it was particularly absent on the outer face of Posidonia leaf (Fig. 5). As regards P. lima, there were only two identified groups of the distribution of this species on the leaf of P. oceanica (Table 6). The first group was formed on the inner face of the middle part and the outer and inner face of the apical part, where the abundance was the highest (Fig. 5). The second group formed the outer face of the middle and basal parts of the leaf, where the abundance was lower (Fig. 5).

The results of the ANOVA showed that the abundance of Ostreopsis cf. ovata and P. lima showed variability on the different parts of the Posidonia leaf (apical, middle and basal) (Fig. 5) and according to their position on the inner and outer faces of the leaves (P_{Ostreopsis cf. ovata}<0.0001; P_P.lima<0,005) (Table 7). Coolia monotis was only represented in a single subset that was recorded in low abundances and only on inner faces of Posidonia leaf (Fig. 5). The distribution showed no differences according to the face or part of the leaf (P_C.monotis>0.05) (Table 7).

Posidonia seagrass beds, in contrast to macrophytes, which are generally rather scattered with a high inter-annual variability, cover large areas of the Gulf of Gabès and are structured in valleys (Mabrouk et al. 2009, 2011, Ben Brahim 2013). During the year, Posidonia was by far the substrate hosting the greatest biomass and diversity of epiphytes (Table 3). This result could be explained by the diverse conditions that Posidonia offers for the success of epiphytic species: (i) the amount of physical structure usable as living space, as Posidonia provides both a shading effect and high microhabitat diversity because of its large leaf areas (Kikuchi and Pérès 1977); (ii) coexistence of Posidonia seagrass material, dead or alive, suspended particulate organic matter and leaf epiphytes as potential food sources within the ecosystem (Dauby 1989); (iii) protection from predators thanks to a dense rhizome mat; and (iv) the reduction of hydrodynamic forces (Lewis 1984). The P. oceanica canopy tends to mitigate currents and waves, thereby reducing the forces exerted on individual shoots (Koch et al. 2006).

Posidonia offers a greater surface for epiphytes than macroalgae such as Padina pavonica and Cystoseira mediterranea. However, the latter two host a relatively abundant population of epiphytes. Though they do not have the highest biomass of epiphytic species, marine macroalgae hosted the highest species diversity (Table 3), probably because they showed spatial complexity and could modulate the availability of resources, therefore affecting assemblages of associated epibiota (Gestoso et al. 2010). In particular, host algae with a branched structure like Cystoseira mediterranea or with a filamentous structure like Dictyota dichotoma usually have a high degree of structural complexity, which may make them more suitable as habitats for epibiota (Totti et al. 2009).

Diatoms were the dominant group and prevailed throughout the sampling period. This dominance could be attributed to their successful behaviour in attaching to the algae and establishing a mutualistic relationship with their host (Romagnoli et al. 2007). Indeed, pennate diatoms have the ability to cling to seaweeds by mucilage stalks and sheaths or gelatinous pads or by the attachment of the cell along its entire valve face. The centric forms are often trapped by the thallus of seaweeds or held in the tangle of attached forms (Totti et al. 2009).

A high diversity and abundance of confirmed toxic and potentially toxic dinoflagellate species hosted in vegetated habitats were recorded, especially on P. oceanica leaves (Table 3). Particularly P. lima, the most abundant species (Fig. 3), seems to affect P. oceanica leaves. This species has been reported as a widespread dinoflagellate in many coastal waters and estuaries around the world, generally in summer and autumn (Levasseur et al. 2003), in the Fleet lagoon in the UK (Foden et al. 2005), in Greek coastal waters (Aligizaki et al. 2009), along the coast and inside the harbours of the Abruzzo region in the Adriatic Sea (Ingarao et al. 2009), and on the northern coasts of Tunisia (Aissaoui et al. 2014). In the study area, it reached about 25000 cells g^–1 FW on Posidonia leaves, which is higher than the 70 cells g^–1 FW found in the same area by Mabrouk et al. (2011). However, these concentrations were lower than those reported for Cymodocea nodosa in Greece, where the abundance reached 133000 cells g^–1 FW (Aligizaki et al. 2009). The high concentrations of P. lima on Posidonia raise the problem of its sampling representativeness, since most monitoring programmes focused on the water column, which might lead to an underestimation of the species abundance (Marr et al. 1992). P. lima concentrations showed a significant relationship between P. oceanica and the water column (R²=0.79), suggesting that the species, being a weekly swimming dinoflagellate that can even be affected by low water motion conditions (Richlen and Lobel 2011), might move from one compartment to another. The establishment of a direct relationship between the concentration of this species in the water column and on macrophytes allows us to assess the concentration in one compartment by referring to the concentration in the other one. Moreover, to the best of our knowledge, the toxicity threshold used for this species in the monitoring programmes was only established for the water column (Abdennadher 2014), so the use of this relationship to extrapolate to the substrata needs to be further investigated.

Ostreopsis cf. ovata had no preference for a given substratum, as indicated by the absence of a significant difference in concentrations between the substrata studied. The significant relationship found between the species on Padina and in the water column above this alga could be explained by the fact that this species is loosely attached to hard substrates and seaweeds with mucilaginous strands (Tindall and Morton 1998). Water motion could cause leaf agitation, allowing the shift of epiphytic species into the water column. A particularly low abundance, with a maximum of 1.85 10³ cells g^–1 FW recorded on Posidonia leaves, was observed during this survey compared with the high species abundances observed in the western Mediterranean, where 7.2 10⁶ cells g^–1 FW was reported in Catalonia (Mangialajo et al. 2011), 2.5 10⁶ cells g^–1 FW on the Genoa coasts (Mangialajo et al. 2008) and 1.7 10⁶ cells g^–1 FW in the Adriatic Sea (Totti et al. 2010). In the eastern Mediterranean (Greece), a maximum abundance of 0.41 10⁶ cells g^–1 FW was observed (Aligizaki and Nikolaidis 2006). These findings suggest that the study area might have some constraints preventing the accumulation of this toxic species, known to cause serious health concerns in other ecosystems and particularly in the Mediterranean (Totti et al. 2010, Cohu et al. 2011).

Leaves of the seagrass P. oceanica hosted the highest population, especially of P. lima, whereas Ostreopsis cf. ovata and other species were very scarce or planktonic (Fig. 3). This opposing pattern between Ostreopsis cf. ovata and P. lima was also illustrated in the divergence of these species in the cluster analysis (Fig. 4). A habitat separation between Ostreopsis spp. and Prorocentrum spp. has already been reported in the Pacific Ocean (Richlen and Lobel 2011). This behaviour could be attributed to allelopathic effects between dinoflagellates leading to possible niche separation. Indeed, some phytoplankton species, including P. lima (Sugg and VanDolah 1999), produce and release secondary metabolites that negatively affect the growth of other organisms (Rizvi and Rizvi 1992). These species quickly cause cell lyses of most competitors within minutes, when the latter are exposed to either certain amounts of the allelochemicals or to certain cell densities of the allelopathic algae. Such allelopathy is thought to reduce competition for nutrients, vitamins, etc. (Fistarol et al. 2004a). Indeed, the co-inertia plot showed that the distribution of Ostreopsis cf. ovata, and to a lesser degree P. lima, was explained by nitrogen, mainly nitrate and nitrite, which might suggest competition between these species for nitrogen availability. Both species were documented to be positively correlated with nutrient availability (nitrate, nitrite, phosphate, and silicate) concentrations in the waters surrounding Hawaii (Parsons and Preskitt 2007). Cohu et al. (2013) reported that phosphate concentration, rather than nitrogen or silicate concentration, was positively associated with Ostreopsis cf. ovata abundances in the north western Mediterranean Sea. Furthermore, many studies have shown that nutrient limitation decreases Ostreopsis cf. ovata growth, an effect that is more accentuated under N-limitation (Accoroni et al. 2014).

For P. oceanica, there is an increase in the cover of most epiphytic species in the apical and middle regions of the leaves (Fig. 5). This result had already been reported in previous studies (Alcoverro et al. 2004) and explained by the fact that the apical part of the leaves, and to a lesser degree the middle part, expose their epiphytes to high light intensities and water movement. This would promote photosynthetic organisms such as epiphytic macroalgae, which increase the nutrient intake from water and remove inhibitory substances (Trautman and Borowitzka 1999). Moreover, the epiphytic species zonation on leaves of Posidonia was reported to be related to the concentration of phenolic compounds produced in abundant quantities, depending on the state of stress caused by environmental conditions (Dumay et al. 2004). However, the use of artificial leaves made of plastic tape showed the same apico-basal distribution of epiphytic algae (Trautman and Borowitzka 1999), supporting the hypothesis that epiphyte settlement was unlikely to be the result of changes in the surface chemistry of the leaves (Borowitzka et al. 2006). These variations were likely due to differences in hydrodynamic or light intensity related to the shape and orientation of the leaves. The inner surface of adult and intermediate leaves seemed to be the most exposed (Borowitzka and Lethbridge 1989).

The concentration of P. lima on the outer surface of the Posidonia leaf, explained by the behaviour it uses to escape predators (Ben Brahim et al. 2010), is in opposition to the general behaviour of other epiphytic species, particularly Ostreopsis cf. ovata, which has been shown to prefer the inner face of Posidonia leaves (Alcoverro et al. 2004, Peirano et al. 2011). This would suggest competition for space between Ostreopsis cf. ovata and P. lima, and might support their apparently opposed distribution pattern.

This study has highlighted the diversity of epiphytic microorganisms on vegetated ecosystems, particularly on macroalgae, and has confirmed the previous finding on the potential of P. oceanica to accumulate epiphytic biomass. This finding suggests that more attention should be paid to the protection of the P. oceanica meadows and their associated epiphytes.

P. lima, by far the most abundant epiphytic toxic species on all vegetated substrates, showed a preference for P. oceanica. A significant correlation was found between the species concentration on that substrate and in the water column. More effort should be made to accurately determine this relationship under different hydrological conditions. One of the practical implications of this result is the recommendation to include the sampling of P. lima on Posidonia leaves in HAB monitoring programme and to set up the toxicity threshold of this species on P. oceanica leaves.

P. lima showed an opposed distribution pattern to that of Ostreopsis cf. ovata on Posidonia leaves, suggesting that competition for space and nutrient between the two species is likely. This hypothesis needs to be investigated in order to assess and apprehend the proliferation mechanisms of the two species.