The Red Sea has about 135 Km² of mangroves distributed up to 28.207302°N, the northern biogeographical boundary of mangroves. Unlike the coverage of mangroves in other regions, the area covered by mangroves in the Red Sea remains relatively stable (Almahasheer et al., 2016a). Nutrient inputs to the Red Sea are dominated by inputs from the Indian Ocean (Murray and Johns, 1997; Johns and Sofianos, 2012; Churchill et al., 2014; Raitsos et al., 2015), leading to a gradient of oligotrophication toward the north (Ismael, 2015), concurrent with an increase in salinity due to high evaporation losses (Talley, 2002). On the other hand, the sea surface temperature declines from south to north (Rushdi, 2015). The Central Red Sea is an arid environment characterized by high temperatures (Table S1 presents the weather conditions during the experiment), sparse rainfall (Edwards, 1987); the mean annual (sporadic) rainfall in Jeddah is 55 mm), high salinity (Bruckner, 2011) and low nutrient inputs and concentrations (Mandura, 1997). The tidal range on the Central Red Sea is only 20 to 30 cm (Sultan et al., 1996), resulting in mangrove habitats that develop as a narrow (typically one to three trees wide) fringe along the shore, adjacent to sand flats that may occasionally be flooded, with desert further inland. The mangrove fringe supports small halophytes. Recent urban and industrial development of the Saudi shore in the Central Red Sea area has resulted in locally elevated nutrient inputs, providing an opportunity to assess how such inputs may affect the nutrient status of mangroves.

The sampled sites in the study area (Figure 1) were selected to capture the diversity of environments in the region. In particular, we sampled mangrove leaves from four locations in the Central Red Sea, including Thuwal Island, Economic City lagoon, Petro Rabigh and Khor Alkharar (Table 1, Figure 1). Away from anthropogenic sources of nutrients, the mangrove stands at Khor Alkharar, Economic City lagoon and Thuwal Island are pristine: King Abdullah Economic City, about 40 km south of Rabigh, is a new and fast-growing urban development; Khor Alkharar lagoon lies on the coastal plain northwest of Rabigh and is connected to the adjacent Red Sea by a narrow inlet on its northwestern side (Gheith and Abou-ouf, 1996; Al-Farawati, 2011); on Thuwal Island, A. marina grows on shallow soils of weathered coral (Balk et al., 2015). In contrast, the city of Rabigh hosts a large petrochemical complex and the mangrove stand are likely influenced by the industrial environment.

We assessed the nutrient status of mangrove stands in the Red Sea by examining the concentrations of nitrogen, phosphorous and iron, the nutrients most often reported to limit mangrove growth (Feller, 1995; Koch and Snedaker, 1997; Feller et al., 2003a,b; Lovelock et al., 2004, 2006; Naidoo, 2009; Alongi, 2011). We analyzed nutrient concentrations in mature leaves as well as the stoichiometric ratios of the nutrients to carbon. Whereas examination of leaf nutrient concentration alone does not suffice to establish nutrient budgets (Lü et al., 2012), leaf nutrient concentration robustly indicates the nutrient status of plants (Duarte, 1990, 1992), including mangroves (Duarte et al., 1998; Feller et al., 2003b). Moreover, leaf nutrient concentration is an important ecosystem property because it affects the decomposition rates of leaf litter (Enriquez et al., 1993). Deeming that quadrant or transect sampling would not achieve the representativeness of the leaf nutrient concentration required by this study, we used a nested sampling strategy to characterize leaf nutrient concentration at each location. First, we selected two to three disjoint stands at each location. From each of these stands, we randomly selected one tree and sampled 6 to 10 leaves from the main stem of the tree. In total, this yielded 15 to 30 leaves from each location (Table 1). We collected the mangrove leaves in March 2015, following propagule release, which causes significant nutrient loss and may render the plants vulnerable to nutrient limitation.

We conducted two experiments to examine the response of seedlings to the addition of nutrients. The experiments were designed to test whether propagules collected from nutrient-deficient stands on Thuwal Island contained enough nutrients to support seedling growth or if their growth and development could be increased by nutrient input. In addition, we identified the limiting nutrient by using various combinations of treatments of the three nutrients. Two varied treatment experiments (see below) were conducted, the first in 2014 and the second in 2015. The experiments were conducted in a nursery covered with a green mesh to reduce incoming solar radiation by around 90% (Quantum Meter MQ-100, μmol PAR m⁻² s⁻¹). The average radiation in the nursery (mean ± SE = 8.3 ± 2.5% of the incoming radiation) was slightly greater than that received under natural growing conditions in Thuwal (5.5 ± 1.7%, Table 2). The plants were exposed to solar radiation and temperature regimes comparable to those in the Thuwal Island mangrove stand where the propagules were produced. We carefully controlled nutrient conditions in both the sediment and water. In particular, we washed the sediments and diluted the seawater to reduce external nutrient inputs in the control treatment, thereby measuring seedling growth as a response to added nutrients in the different treatments. The control treatment thus captures the growth and development capacity of A. marina seedlings as determined by the nutrients allocated by the mother plant to the propagules. Accordingly, the response of the seedling to various nutrient treatments indicates which nutrients are deficient in the propagules relative to those required to support nutrient-sufficient growth of the resulting seedlings because propagule production is likely to be a major vector for nutrient loss in the mangrove stands studied.

The nursery was fitted with raceways periodically flooded with brackish water using pumps to simulate the tidal cycle (see below). The experiments were initiated in February 2014 and March 2015, using A. marina seedlings germinated from propagules collected from healthy mature trees on Thuwal Island. The collected propagules were fully formed, with a hard pericarp and an average dry weight of 2.98 ± 0.12 g DW (n = 33). Five propagules were dried for nutrient analysis. The raceways were washed with 1% acid followed by nutrient-free water to avoid contamination before the propagules were germinated. The propagules were soaked in brackish water for 3 days until their covers shed naturally. All the propagules used in the experiment germinated on the same day. The sediment used to accommodate the seedlings in the raceway consisted of sand collected from Thuwal that was washed with nutrient-free water to reduce the concentrations of available nutrients. The control treatment was thereby rendered nutrient depleted. The brackish water was a mixture of equal volumes of seawater collected near Thuwal and nutrient-depleted fresh water collected from King Abdullah University of Science and Technology (KAUST), which was found to provide suitable growth conditions (Connor, 1969; Clough, 1984). Previously, between September and November 2013, we conducted a preliminary test of the growth response of seedlings to various salinity regimes (0, 25, 50, 75, 100% seawater) and confirmed that a 50:50 mixure of Red Sea seawater and fresh water provided the best growth conditions. Indeed, this finding is consistent with the occurrence of Red Sea mangroves near groundwater inputs in the Red Sea; many mangrove stands are located near Wadis (dry river beds with associated groundwater discharge). The raceways were flooded with brackish water by pumps for 6 h daily, following (Clarke and Johns, 2002). The excess water then slowly flowed back into the tanks located under the raceways (Figure 2). This flooding simulated the intertidal cycle while also avoiding excess algal growth in the raceways. Furthermore, to avoid water loss due to evaporation and evapotranspiration, we replaced the water in the tanks every month with freshly prepared brackish water as described above. Tanks, valves, and raceways were washed monthly during the experiment.

Propagules were planted at half height in the sediment (Bovell, 2011) in individual pots. The experimental treatments were initiated 4 weeks after planting. By this time, the seedlings were equal in height and had developed 4 leaves each (Clough, 1984; Ball et al., 1987). We used 30 replicated seedlings per treatment in the first experiment and 21 replicated seedlings per treatment in the second experiment, with the treatments applied independently to each replicated seedling.

The first experiment aimed at assessing the roles of nitrogen, phosphorus and iron, alone and in combination, in promoting seedling growth beyond the growth capacity corresponding to the nutrients received from the mother plant, whereas the second experiment aimed at separating the possible effects of iron and the chelant used in combination in the first experiment, which was found to significantly enhance seedling growth (see Results). A chelant is an organic ligand incorporating metal ions in a structure that facilitates metal uptake by plants (Salt et al., 1998; Hart, 2000). Specifically, Ethylenediaminetetraacetic acid (EDTA) has been found to facilitate Fe uptake. We used the nutrient concentrations of (Ball et al., 1987) for nitrogen and phosphorus and (Steiner and van Winden, 1970) for Fe-EDTA. The first experiment, initiated in March 2014, included the following final nutrient concentrations in six treatments applied to the brackish water used to flood the raceways: (1) nitrogen (as 2 mmol NH₄NO₃L⁻¹); (2) phosphorus (0. 2 mmol NaH₂PO₄L⁻¹); (3) iron (0. 2 mmol Fe-EDTA L⁻¹); (4) nitrogen, phosphorous, and iron, each supplied at the preceding concentrations; and (5) a complete fertilizer, including 4 mmol Ca (NO₃)₂L⁻¹, 0. 2 mmol L⁻¹ NaH₂PO₄, 1 mmol L⁻¹ MgSO₄, 0. 2 mmol Fe-EDTA L⁻¹, 2 mmol NH₄NO₃L⁻¹, 0.025 mmol H₃BO₃L⁻¹, 0.002 mmol MnSO₄L⁻¹, 0.002 mmol ZnSO₄L⁻¹, 0.0005 mmol CuSO₄L⁻¹, 0.0005 mmol H₂MoO₄L⁻¹, and 0.025 mmol MgCl₂L⁻¹; and (6) a control without nutrient addition. In the second experiment, the following four treatments were tested: (1) iron (0. 2 mmol FeCl_2, L⁻¹); (2) EDTA (0. 2 mmol EDTA L⁻¹); (3) Fe + EDTA (0.2 mmol Fe-EDTA L⁻¹); and (4) the control without iron or EDTA addition. Nutrient concentrations in the brackish water were adjusted monthly. The nutrient concentrations used in the treatments were higher than in the water in a lagoon near Thuwal, where the (mean ± SE) concentrations of NO₂ were 0.062 ± 0.006 μmol L⁻¹ and NO₂+NO₃ were 0.181 ± 0.017 μmol L⁻¹ (Banguera-Hinestroza et al., 2016). They were also higher than values found in coastal water from Duba (north) and Jazan (south) along the Saudi coast of the Red Sea (NO₃ = 2.351 ± 0.386 μmol L⁻¹ and PO₄ = 0.239 ± 0.039μmol L⁻¹; (Pearman et al., 2016).

The following parameters were measured once per month in the first experiment and every month and a half in the second experiment: (a) C, N, P, Fe. and Chl a concentrations in the first pair of fully developed apical leaves of three seedlings per treatment; and (b) growth rate, as determined by the elongation rate along the main meristem, number of nodes, number of leaves and root development for each seedling in the experiment. Chl a was extracted from 1-cm diameter discs cut from fresh leaves using a QIAGEN TissueLyser II and extracted overnight in 10 ml 80% acetone. The Chl a concentration in the extract was measured spectrophotometrically based on absorbance at 663 nm following (Wellburn, 1994). The other apical leaf was washed with nutrient-free water, dried at 60°C in a drying oven and ground for carbon and nutrient analyses. Carbon and nitrogen concentrations were determined on 2 mg samples of powderized leaf material wrapped in a pre-combusted aluminum capsule using a FLASH 2000 CHNS Analyzer (Zimmermann et al., 1997). Iron and phosphorous concentrations were determined after digesting 0.50 mg of the same powderized leaf with 5 ml concentrated HCL and a few drops of H₂O₂ in a Digi PREP digestion system, following three temperature steps: 30°C for 30 min, 50°C for 30 min. and 75°C for 45 min. The samples were allowed to cool, diluted to 25 ml with Mili-Q water and analyzed by Inductively Coupled Plasma-Optical Emission Spectrometry (Varian Inc. model 720-ES).

Iron and phosphorous analyses were run in duplicate. Two different standards (Inorganic Ventures and PerkinElmer's Pure Plus) were used to assess the accuracy of the measurements, with recovery rates of 100.1% for iron and 99.1% for phosphorous (n = 10) in the first experiment (2014) and 98.8% for iron (n = 10) in the second (2015). The recovery was, on average, 98.7% for phosphorous and 104.9% for iron (n = 20) in the analyses of leaves collected in the survey (n = 30). We also used a standard reference material from The National Institute of Standards and Technology (NIST) to verify the accuracy of the phosphorous and iron concentrations, resulting in 94.6% accuracy in apple leaves and 81.9% accuracy in peach leaves for the second experiment and 73.3% accuracy in apple leaves and 79.2% accuracy in peach leaves in the analyses run for the leaves collected in the survey. To analyze carbon and nitrogen concentrations, we used a standard reference material (BBOT) and a calibration standard (Sulfanilamide) to obtain an average recovery of 99% for nitrogen and 102% for carbon (n = 14) for the first experiment and an average of 101.1% for nitrogen and 100.7% for carbon for the leaves collected in the survey (n = 30).

Seedling growth rates were calculated from the slopes of the fitted linear regressions on the relationship between the development metrics and time. Statistical analysis, including descriptive statistics, linear regression analyses, general linear models to test for the effects of the different experimental treatments or differences among stands, and Tukey's HSD (honest significant difference) post-hoc test to assess pairwise differences were carried out using JMP, a computer program for statistical analyses developed by SAS Institute.

Mangrove stands in the Central Red Sea are characterized by low leaf nutrient concentrations and low carbon-to-nutrient stoichiometric ratios, with an overall mean of 1918 C: 36 N:1 P: 0.5 Fe, indicative of severe nutrient depletion, particularly P and Fe, across stands and pointing to the likelihood of nutrient limitation of Central Red Sea mangroves. However, there were important differences among stands, with mangroves receiving nutrient inputs from industrial and urban sources at Petro Rabigh showing the highest concentrations of nitrogen, phosphorus and iron across stands. Fe deficiency was particularly acute in the propagules collected to initiate the seedling experiments, which had P:Fe ratios that were 10 to 20-fold higher than those of leaves, indicative of extreme Fe deficiency in these propagules. Unfortunately, C: N: P: Fe ratios for balanced mangrove growth have not yet been defined. Nevertheless, C: N: P ratios for nutrient sufficient A. marina growth can be inferred from those in the fertilized treatments reported by Alongi (Alongi, 2011), which correspond to 346 C: 14 N: 1 P, similar to the treatments receiving P in our experiments, which had the lowest C:N:P ratio. Even the control plants, which did not receive any nutrient additions, in the stand examined by Alongi (Alongi, 2011) had a C:N:P ratio of 711:30:1, well below the ratios found in leaves of Central Red Sea mangroves. The stoichiometric ratios in A. marina seedlings receiving Fe+N+P in our experiments can be considered to be nutrient-sufficient plants. These ratios are quite similar to those for nutrient-sufficient macroalgae and N-fertilized A. marina plants in Northern Queensland, Australia, but are higher than those for nutrient-sufficient seagrass and phytoplankton (Table 4).

The experimental results allow a first approximation of the critical level of nutrient concentrations required to support nutrient-sufficient growth of A. marina seedlings. These were defined as the nutrient concentrations below which carbon concentrations declined steeply (cf. Figure 6). The inferred thresholds are 2 mmol N g DW⁻¹, 0.04 mmol P g DW⁻¹, and 0.001 mmol Fe g DW⁻¹. These critical thresholds are higher than the thresholds of nitrogen and phosphorus reported for seagrass (N < 1.2 mmol N g⁻¹, P < 0.06 mmol P g⁻¹, recalculated from Duarte, 1990).

Our results provide compelling evidence that the growth of A. marina seedlings in the oligotrophic Central Red Sea requires environmental Fe inputs, as iron addition alone led to a growth response and an increase in Chl a concentrations comparable with those in the treatments also receiving nitrogen and phosphorous. Although, our experiment showed iron to be the main limiting factor for seedling growth, nitrogen addition significantly increased the number of leaves while addition of phosphorous and particularly iron led to increased root development. Yet, the experimental treatment that combined N, P, and Fe addition did not increase growth beyond that achieved by Fe fertilization alone, substantiating Liebig's law of the minimum, with iron as the limiting nutrient. The results from the second experiment confirmed that it was indeed iron, and not the ligand added in the first experiment (EDTA), which elicited the observed growth responses.

The finding of iron-limited growth of Central Red Sea seedlings is consistent with (Alongi, 2010), who reported high iron requirements in the transition stage from seedling to sapling in five mangrove species growing in north Queensland, Australia. Iron limitation in mangrove growth in the Central Red Sea is expected, as iron limitation is characteristic of marine coastal habitats dominated by biogenic carbonate sediments, which are typically iron depleted (Duarte et al., 1995). Moreover, the lack of riverine inputs, which are the main vector of iron to coastal marine sediments (Poulton and Raiswell, 2002; Raiswell, 2006), also renders the Red Sea prone to iron limitation. Iron-limited seagrass growth has also been reported in the Mexican Caribbean (Duarte et al., 1995), where the karstic landscape lacks riverine input and where mangroves are also stunted. Similar to our findings in the Red Sea, iron addition stimulated leaf Chl a concentration in Caribbean seagrasses receiving iron (Duarte et al., 1995).

In the absence of rivers and negligible precipitation, such as in the Red Sea, nutrient inputs from the Indian Ocean (Murray and Johns, 1997; Johns and Sofianos, 2012; Churchill et al., 2014; Raitsos et al., 2015), sub-surface mixing (Triantafyllou et al., 2014; Yao et al., 2014), and atmospheric sources are likely to be the most significant path for new nutrient input (Brindley et al., 2015), including nitrogen fixation and inputs of phosphorus and iron with dust deposition (Aerts and Chapin, 1999; Schroth et al., 2009). The Red Sea receives one of the highest dust inputs of all the oceans (Jish Prakash et al., 2015), and mangrove canopies may intercept dust during periods of dust storms, thereby increasing iron supply above that corresponding to passive dust deposition. The solubility of atmospheric iron varies between 0.01 and 80% depending on acid deposition, with 95% of atmospheric iron derived from desert dust and the rest from combustion processes (Mahowald et al., 2009). Although, data from the study area are lacking, seabird guano e.g., brown booby birds, Red-billed tropic-birds and Bridled terns may also be an important source of nutrients to Red Sea mangroves, which support abundant bird communities (Evans, 1987). The results presented also suggest anthropogenic inputs to be significant in mangrove stands near developed areas, as exemplified by the mangroves near Petro Rabigh, which exhibited elevated concentrations of iron as well as nitrogen and phosphorous.

A range of mangrove fertilization experiments were previously conducted, mostly focused on nitrogen and phosphorous addition, but those exploring iron limitation have been scarce. The outcome of these experiments depend on location, with some mangrove stands shown to be limited by phosphorous (Feller, 1995; Koch and Snedaker, 1997; Feller et al., 2003a; Lovelock et al., 2004, 2006) and others to be limited by nitrogen (Feller et al., 2003b; Naidoo, 2009). Prior to the present study, iron limitation in mangrove growth had been reported, on the basis of experimental addition, only in mangroves in north Queensland, Australia (Alongi, 2010). The finding that mangroves in the Central Red Sea are nutrient-limited, as evidenced by low nutrient concentrations in their leaves and very low iron levels in the propagules they form, allows us to begin to understand the functioning and drivers of these important components of the Red Sea ecosystem. Nutrient-limited plants are likely to re-absorb significant amounts of nutrients before shedding their leaves (Reef et al., 2010), suggesting that nutrient-limited mangroves may act as a sink rather than a source of nutrients to the ecosystem. Moreover, the decomposition rate of plant litter decreases with increasing C:N ratios (Enriquez et al., 1993). Nutrient-limited Central Red Sea mangroves are thus likely to produce relatively recalcitrant litter, which may act as a carbon sink in mangrove sediments. Lastly, our results show that nutrient-deficient mangroves are likely to produce nutrient-poor propagules. The resulting seedlings require additional nutrient inputs to support high growth and survival. The absence of significant nutrient inputs may, thus, act as a bottle-neck in the stability of the population (Selvam, 2007). Nutrient-deficient propagules may also occur in other nutrient-limited mangrove stands. This suggestion requires further testing. Increased anthropogenic nutrient emissions in the Central Red Sea are likely to lead to increased mangrove growth and improved nutrient status with consequences for the ecosystem. Nutrient limitation may also affect the capacity of mangroves to respond to other stresses affecting the Red Sea, such as the high temperature and lack of fresh water supply, as nutrient limitation would impose limits to the capacity of the plants to synthesize proteins and other molecules involved in building resilience to these stresses. Hence, nutrient limitation likely renders the plants more vulnerable to the extreme environmental conditions in the Central Red Sea. However, responses to multiple stresses have not yet been investigated as we lack information on the thermal thresholds of A. marina in the Central Red Sea as well as on the consequences of a low fresh water supply for the performance of the plants. These responses, along with the implications of nutrient limitation in modulating the resistance to stress, should be targeted by future research.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.