*2.2. Lipid Trajectories*

Principal Component Analyses (PCA) were performed using fatty acids relative proportions (in %) for each salinity treatment (Figure 2a–c). The effect of the temperature gradient on fatty acids profiles is displayed for each salinity separately. The PCA allowed to track fatty acids trajectories which corresponded to the path of gradual change in whole fatty acid composition according to T °C. At 70 PSU, no clear trajectories were observed and the PCA ordination explained 55% of the whole inertia in comparison to 61 and 63% for 23.5 and 50 PSU (Figure 2).

Lipid trajectories showed clear opposite directions at 23.5 and 50 PSU between host-algae (LD and SL) meaning that the elevation of temperature induced opposite responses in *P. salina* according to the algal host. Further investigation of lipid trajectories showed an opposition between 18:1*n*-9 and 18:2*n*-6. A fatty acid index (*FAI*18-*C*) was calculated using major C18 unsaturated fatty acids relative concentrations (%) :

$$FAI\_{18\text{-C}} = \frac{\left[18.2n\text{-6}\right]}{\left[18.1n\text{-9}\right]}\tag{1}$$

*FAI*18-*<sup>C</sup>* was significantly affected by host-algae, temperature and salinity effects (Table 2) but only host-algae:temperature and temperature:salinity interactions were significant.

**Table 2.** Thee-way analysis of variance (ANOVA) of *FAI*18-*<sup>C</sup>* index as a function of host-algae, temperature and salinity. Normality assumption by group was tested using Shapiro–Wilk. In total, 14 out of 18 groups showed *p* > 0.05. Homogeneity of variance was tested using a Levene's test (*d f*<sup>1</sup> = 17, *d f*<sup>2</sup> = 36, statistic = 1.19, *p* = 0.321). n.s. = not significant, \* *p* < 0.05, and \*\*\* *p* < 0.001.


It also showed clear opposed linear relationships with temperature according to the algal host (Figure 2d). The index increased with temperature in LD but decreased in SL. At salinities 23.5 and 50 PSU the effect of host algae on *FAI*18-*<sup>C</sup>* was significant as the assumption of homogeneity of slope regression was not met (ANCOVA, interaction host-algae:temperature, *F* = 35.509, *d fn* = 1, *d fd* = 14, *p* = 3.49 × 10−<sup>5</sup> and *F* = 34.483, *d fn* = 1, *d fd* = 14, *p* = 4.06 × 10−<sup>5</sup> respectively). This indicated that slopes of the regression lines were significantly different. At the salinity of 70 PSU, the homogeneity of slope regression was validated (ANCOVA, interaction host-algae:temperature, *F* = 2.818, *p* = 0.115) but the effect of host algae was not significant (ANCOVA–temperature effect: *F* = 4.124, *p* = 0.06, host-algae effect: *F* = 0.355, *p* = 0.56). When regression lines were significantly different, the angle between the two regression lines was calculated (hereafter named *α*-value). It represented the degree of influence of host-algae in the adaptation of *P. salina* to temperature (Figure 2). This degree of influence was significantly decreasing with salinity (Supplementary Materials Figure S1).

**Figure 2.** Fatty acid trajectories of endophytic *P. arenaria* isolated from *L. digitata* (LD) or *S. latissima* (SL) grown at different salinities (23.5, 50 and 70 PSU) and different temperatures (10, 18 and 25 °C): (**a**–**c**) PCA scores. (**d**) *FAI*18-*<sup>C</sup>* index calculated with relative contributions (%) of major C18 fatty acids showing opposite trends in the acclimation of *P. salina* to temperature between host algae. *α* = angle between the two regression lines.
