This study was performed at Gran Canaria Island, a subtropical island in the centre of the Canary Islands (eastern Atlantic, 28°N). Sea surface temperature typically varies between 18°C in winter and 24ºC in summer (Tuya et al. 2014). Vegetative fragments of C. nodosa (horizontal rhizomes ca. 10-15 cm long, including 5-7 shoots and associated leaves) and ramets of C. prolifera, including stolons, rhizoids and fronds (ca. 20-25 cm long), were collected by hand by SCUBA divers at 5 and 20 m depth at Gando Bay (27°55′49″N, 15°22′1″W). Neither species is typically found in shallower water in the study region, which is exposed to large oceanic swells that prevent long-term persistence. Collections of both macrophytes were immediately used to assess photochemical responses, as a way to infer their physiological status (see below). On the day of collection, we additionally carried out a reciprocal transplant experiment, in which 10-15 C. nodosa and C. prolifera vegetative fragments collected at 5 m were transplanted at 20 m depth and vegetative fragments collected at 20 m were transplanted at 5 m depth (hereafter called ‘allochthonous’ transplants). To control for potential artefacts during manipulations, C. nodosa and C. prolifera fragments from both depths were re-transplanted at the same depth (i.e. procedural controls, hereafter called ‘autochthonous’ transplants). All transplants were then subjected to the same manipulations. During transplants, the below-ground compartments of both macrophytes were gently buried inside the sediment by hand; to facilitate retention of below-ground tissues, plastic meshes covering horizontal rhizomes and stolons were secured to the bottom by metal stakes (Fig. 1). In all cases, the erect parts of both macrophytes (seagrass leaves and C. prolifera fronds) come out from the mesh to capture light with minimum perturbation. This experimental approach was repeated twice to evaluate the temporal consistency of results; the first started on 6 February 2014 and the second started on 15 May 2014. The duration of each transplant set varied slightly as a result of logistical constraints; on the first occasion (February 2014), transplants were maintained for 11 days; on the second occasion (May 2014), transplants were maintained for 14 days. During both periods, the amount of PAR reaching the bottom was measured by using Onset HOBO U12 4-channel external data loggers at 5 and 20 m depth, respectively.

To facilitate further interpretation of results, we described changes in the biomass of both C. nodosa and C. prolifera with depth at the study site. Samples (n=5) were collected through a 0.0125 m^–2 core at 5, 10, 15 and 20 m in Gando Bay at both sampling times. All vegetation was dried (24 h at 70°C) and the dry weight was calculated. Additionally, sediment samples (n=2 cores, ca. 100 g) were collected, in February 2014, at 5, 10, 15 and 20 m to work out differences in the mean particle diameter (D₅₀), the total N and P content and the organic matter content of sediments with depth. After thawing, sediment samples were sieved through a 0.5-mm sieve, and the fraction <0.5 mm was oven-dried at 90°C for 24 h. This fraction was then dry-sieved at 0.5 ϕ intervals, down to 1.0 ϕ (0.5 mm). The <0.5 mm fraction was freeze-dried and analysed on a Coulter LS130 laser sizer. The laser sizer results were combined with the dry sieve results to give the full particle size distribution. The method of Walkley and Black (1934) was used to calculate the organic matter content of sediments via rapid dichromate oxidation. Total N was determined following the Kjeldahl method (Bradstreet 1965) and total P was determined using a spectrophotometric method (Murphy and Riley 1962).

At the start (initial collections) and at the end of the transplantation experiments, between n=6 and 8 C. nodosa leaves (12-15 cm long with no necrosis signs) and n=6 and 8 C. prolifera fronds (5-8 cm long) from each depth and manipulation treatment were used to estimate their photochemical status. In vivo chlorophyll a fluorescence of Photosystem II (PSII) was assessed through a portable pulse amplitude modulation fluorometer (Diving-PAM, Waltz GmbH, Germany). After 15 min of dark adaptation (‘relaxed state’), the minimum (basal) fluorescence was measured (F_o) and the maximum fluorescence (F_m) was obtained immediately after applying a saturated pulse of actinic light (2,350 µmol photons m^–2 s^–1, 0.8 s). The optimum quantum yield was therefore calculated as F_v/F_m=F_m–F_o/F_m, which is an indicator of physiological stress (Maxwell and Johnson 2000). A rapid light curve (RLC) was initiated, involving 20 s of exposure to 9 successive irradiances, including 20, 66, 137, 224, 337, 469, 693, 942 and 1418 µmol photons m^–2 s^–1. RLCs were then obtained by calculating the electron transport rate (ETR) through the PSII for each level of actinic light through the formula:

ETR (µmol electrons m^–2 s^–1) = (ΔF/F′m) × E × A × FII

where ‘E’ is the irradiance; ‘A’ is the absorptance of each macrophyte (0.88±0.02 for C. nodosa and 0.93±0.03 for C. prolifera, calculated using the method of García-Sánchez et al. 2012); and ‘FII’ is the fraction of chlorophyll associated with the PSII (0.5 for green seaweeds, according to Grzymski et al. 1997). RLCs were fitted through the model provided by Jassby and Platt (1976) to obtain the initial slope of the curve (α_ETR, i.e. the photosynthetic efficiency), the saturation irradiance (E_k) and the maximal ETR (ETR_max); the model of Platt and Gallegos (1980) was applied when photo-inhibition was detected.

The carotene content was determined spectro-photometrically. The analyses were carried out by extracting pigments from plants (ca. 20 mg dry weight [DW], n=3) using 1 ml of saturation solution of acetone 90% + C₄Mg₄O₁₂ and maintaining them in darkness at 4°C for 12 h. After centrifugation at 4000 rpm for 20 min, each supernatant was used to measure carotenes in a spectrophotometer at an absorption of 480 and 750 nm. The carotene concentration, expressed as mg g^–1 DW, was calculated using the equation provided by Parsons and Strickland (1963).

Thalli of C. nodosa and C. prolifera (ca. 0.25 g FW, n=3) were ground with a mortar and a pestle in sand at 4°C and extracted overnight in centrifuge tubes with 2.5 ml of 80% (v/v) methanol. The mixture was centrifuged at 4000 rpm (30 min) and the supernatants were collected (Sigma 2-16PK, Göttingen, Germany). The DPPH (2,2-diphenyl-1-picrylhydrasyl) free-radical scavenging assay was carried out in triplicate, according to the method of Blois (1958). Briefly, 150 µL of each methanolic extract was mixed with 1.5 mL of 90% methanol and 150 µL of DPPH solution prepared daily at 1.27 mM. The reaction was complete after 30 min in darkness at room temperature, and the absorbance was registered at 517 nm. The calibration curve made with DPPH was used to calculate the remaining concentration of DPPH in the reaction mixture after incubation. Values of DPPH concentration (µM) were plotted against plant extract concentration (mg DW mL^–1) to obtain the EC₅₀ value (oxidation index), which represents the concentration of the extract (mg mL^–1) required to scavenge 50% of the DPPH in the reaction mixture. Ascorbic acid was used as a positive control.

A three-way ANOVA tested for differences in the biomass of C. nodosa and C. prolifera across depth; the model included the factors ‘Species’ (fixed factor), depth (fixed factor) and ‘Time’ (random factor). Linear regression models tested whether the mean particle diameter (D₅₀), the mean total N and P content and the mean organic matter content of sediments varied with depth. To test for initial differences (‘start’ collections) in physiological responses between macrophytes (C. nodosa vs C. prolifera), depths (5 vs 20 m) and times (February vs May 2014), three-way ANOVAs were carried out. Each model incorporated the factors ‘Species’ (fixed factor), depth (fixed factor) and ‘Time’ (random factor). A three-way ANOVA tested separately for each macrophyte for significant differences in physiological responses between depths (two levels: 5 vs 20 m), times (two levels: February vs May) and the origin of the transplant (autochthonous = procedural control vs. allochthonous) at the end of the transplants. The model incorporated the same factors as above, in addition to ‘Origin’ (fixed factor with two levels). For all ANOVAs, data were transformed to avoid heterogeneous variances (see Tables 1, 2 and 3 for specific transformations); Cochran’s C test was used in this sense. If no transformation rendered homogeneous variances, the significance level was set at 0.01 rather than the 0.05 level to avoid increasing a type I error (Underwood 1997). Pairwise comparisons were carried out via Student-Newman-Keul (SNK) tests wherever necessary.

In February 2014, integrated mean PAR values at 5 m were ca. 2.57 times larger than at 20 m (Fig. 2A; means =124.50 vs 48.40 μ mol photons m^–2 s^–1), whereas in May 2014 they were ca. 2.24 times larger (Fig. 2B; means =251.72 vs 111.89 μ mol photons m^–2 s^–1). In general, maximum daily peaks in PAR increased from February to May 2014; for example, at 5 m, peaks increased from <1300 µmol photons m^–2 s^–1 to ~2000 µmol photons m^–2 s^–1. Sediments were dominated by fine sands at all depths (Appendix 1).

The biomass of both macrophytes varied between depth strata following different trends (‘Sp×De’, F_3,64= 66.89, P<0.0001, Fig. 3). The biomass of the seagrass C. nodosa was larger at 5 and 10 m than at 15 and 20 m (SNK tests, Fig. 3). The biomass of C. prolifera showed a minimum at 5 m and similar values at 10, 15 and 20 m (SNK tests, Fig. 3). These patterns were consistent through times (‘Species×Time’, F_1,64=0.19, P=0.6657, ‘De×Ti’, F_3,64=0.11, P=0.9556), reinforcing the generality of these observations.

The seagrass C. nodosa showed a larger optimum quantum yield (F_v/F_m) than C. prolifera, particularly in May 2014 (‘Ti×Sp’, P=0.001, Table 1; Fig. 4), independently of depth (all terms involving ‘De’, P>0.05, Table 1; Fig. 4). Similarly, C. nodosa showed a significantly larger ETR_max than C. prolifera in May 2014 (‘Ti×Sp’, P=0.0008, Table 1; Fig. 4), regardless of depth (all terms involving ‘De’, P>0.05, Table 1; Fig. 5). The seagrass C. nodosa also had a larger α_ETR than C. prolifera (‘Sp’, P<0.00001, Table 1; Fig. 6), which was particularly accentuated in May 2014 (‘Ti×Sp’, P=0.001, Table 1; Fig. 6). Both macrophytes were more photosynthetically efficient, i.e. larger α_ETR at 20 than 5 m depth (‘De’, P=0.0230, Table 1; Fig. 6). Appendix 2 includes means and SE of means of all photosynthetic parameters. The carotenoid content was significantly larger in C. prolifera than in C. nodosa (‘Sp’, P=0.001, Table 1; Fig. 7). While C. nodosa increased its carotene content from 5 to 20 m depth at both times, C. prolifera decreased it (‘De×Sp’, P<0.00001, Table 1; Fig. 7). The antioxidant activity of C. nodosa was significantly greater than that of C. prolifera (‘Sp’, P=0.0001, Table 1; Fig. 8).

The optimum quantum yield of C. nodosa did not differ either between depth strata or between the origin of the vegetative fragment (all terms involving ‘De’ and ‘Or’, P>0.05, Table 2; Fig. 9A, B); significant differences only occurred between times (‘Ti’, P<0.00001, Table 2; Fig. 9A, B), including a decay in the optimum quantum yield in May in comparison with February (2014). The optimum quantum yield of C. prolifera varied between depths inconsistently from time to time (‘Ti×De’, P=0.0011, Table 3; Fig. 9C, D): in February 2014, the optimum quantum yield was larger for specimens collected at 20 than at 5 m, whereas in May 2014 no significant differences were detected (SNK tests, Fig. 9C, D). Thalli of C. prolifera increased the optimum quantum yield when transplanted from 5 to 20 m, while the opposite pattern was detected when thalli from 20 m were transplanted at 5 m (Fig. 9C, D). The ETR_max of both macrophytes was larger in May than in February 2014 (‘Ti’, P<0.01, Tables 2 and 3; Fig. 10), regardless of depth and the origin of the vegetative fragments. The photosynthetic efficiency of C. nodosa (α_ETR) varied exclusively between times; larger α_ETR values were observed in May (2014) (‘Ti’, P<0.00001, Table 2; Fig. 11A, B). Irrespective of times, however, C. prolifera showed a larger photosynthetic efficiency at 20 than at 5 m (‘De’, P=0.0066, Table 3; Fig. 11C, D). Thalli of C. prolifera increased their photosynthetic efficiency when transplanted from 5 to 20 m, while the opposite pattern was observed for thalli transplanted from 20 to 5 m (Fig. 11C, D); however, this pattern was not detected statistically (all terms involving ‘De’ and ‘Or’, P>0.05, Table 3; Fig. 11C, D). Appendix 3 includes means and SE of means of all photosynthetic parameters. The carotene content of C. nodosa differed between depths according to the origin of vegetative fragments from time to time (‘Ti×De×Or’, P=0.0004, Table 3; Fig. 12A, B). The magnitude of differences between depth levels depended on the origin (‘De×Or’, P=0.08, Table 3; Fig. 12C, D); differences in the carotene content between allochthonous and autochthonous fragments were more accentuated at 5 than at 20 m depth (SNK tests, Fig. 12C, D). The carotene content of C. prolifera differed exclusively between times (‘Ti’, P=0.00025, Table 3; Fig. 12C, D); a decrease in the carotene content was observed with increasing depth, but it was not statistically significant (‘De’, P=0.2084, Table 3; Fig. 12C, D). The antioxidant activity of both macrophytes varied between depths according to times, regardless of the origin of the vegetative fragments (‘Ti×De’, P<0.00001, Tables 2 and 3; Fig. 13); differences in the antioxidant activity between 5 and 20 m depth were significant in February but not in May 2014 (SNK tests, Fig. 13).