Response to nitrogen and salinity in Rhizophora mangle propagules varies by maternal family and population of origin

Many coastal foundation plant species thrive across a range of environmental conditions, often displaying dramatic phenotypic variation in response to environmental variation. We characterized the response of propagules from six populations of the foundation species Rhizophora mangle L. to full factorial combinations of two levels of salt (15 ppt and 45 ppt) reflecting the range of salinity measured in the field populations, and two levels of nitrogen (N; no addition and amended at approximately 3 mg N per pot each week) equivalent to comparing ambient N to a rate of addition of 75 kg per hectare per year. The response to increasing salt included significant plasticity in succulence. Propagules also showed plasticity in maximum photosynthetic rate in response to N amendment, but the responses depended on the level of salt and varied by population of origin. Generally, survival was lower in high salt and high N, but the impact varied among populations. Overall, this study revealed significant phenotypic plasticity in response to salt and N level. Propagules from different populations differed in all traits measured. Variation in phenotypic plasticity and propagule survival in R. mangle may contribute to adaptation to a complex mosaic of environmental conditions and response to climate change.

. This is particularly true in coastal ecosystems that are marked by temporal cycles 31 and spatial variation in tidal inundation, temperature, nutrient availability, and salinity (Pennings and 32 Bertness Foundation species are defined not only as those that dominate a community assemblage numerically 40 or in biomass, but they also determine diversity of associated taxa through a variety of interactions 41 (Ellison, 2019). Further, foundation species modulate fluxes of nutrients and energy in their 42 ecosystem (Ellison, 2019 Bertness, 2020; Qiao et al., 2021). Understanding how these species cope with challenges from 45 anthropogenic impacts is key to preserving the ecosystems they create and define (Guo et al., 2021). 46 Understanding the mechanisms of response in coastal foundation species has become increasingly 47 important for conservation and management strategies as these species must cope with rising sea 48 levels and increased warming due to global change (Gedan et  Global forest resources assessment 2020, 2020). In some regions, mangrove trees are also harvested 55 for wood and charcoal (Ellison et al., 2015), resulting in habitat fragmentation and isolation of 56 existing remnant fragments (Friess et al., 2012;Haddad et al., 2015). The resultant loss of diversity 57 could pose risks for these coastal foundation species in the future, particularly as sea levels are 58 projected to rise between 0.2 and 2 m over the next century due to anthropogenic climate change 59 (Melillo et al., 2014). 60 The vulnerability of coastal foundation plant communities to global change has been debated. Several 61 authors have suggested that the combination of eutrophication and sea-level rise may result in 62 synergistic losses and requires further research ( (Feller, 1995;McKee, 1995;Feller et al., 2003), which could be an important response to 87 anthropogenic activities, such as runoff from agriculture and other types of land use change (Feller et 88 al., 2003;Alongi, 2013). 89 Given spatial differences in salinity, anoxia, and N in the intertidal habitat, plasticity of traits that 90 allow for tolerating such conditions may be adaptive. We expect intertidal plants like mangroves to 91 show plasticity in response to salinity and N conditions. Profitt & Travis (Proffitt and Travis, 2010)  92 found plasticity in growth rate and reproductive output within and among natural Rhizophora mangle 93 mangrove populations in the Tampa Bay region. However, they also found both site of origin and 94 maternal tree of origin affected R. mangle growth and survival, and that these effects varied by 95 intertidal position (significant maternal family by elevation interaction; (Proffitt and Travis, 2010)). 96 On the other hand, we found that nearby populations had low genetic diversity, and little population 97 differentiation. Instead we discovered high epigenetic diversity based on DNA methylation 98 polymorphisms (Mounger et al., 2021). This type of epigenetic diversity has been associated with 99 phenotypic and functional diversity, and could be a mechanism underlying phenotypic plasticity 100 (Zhang et  In this study, we characterized within and among population level variation in putative adaptive traits 108 in response to combinations of salinity and N in a full factorial design. Given the dynamic 109 environment inhabited by R. mangle and the evidence of heritable nongenetic differences among 110 populations, we predicted differences in response to salinity and N amendment treatments among 111 propagules collected from different populations. Our study was designed to test three predictions. 112 First, R. mangle seedlings will be plastic in response to salinity and N amendment in putative 113 adaptive traits that conserve water and adjust allocation of N. Second, response to salinity and N 114 amendment will co-vary as plants shift resources to maintain osmotic balance. Finally, populations 115 will vary in putative adaptive traits, and in plasticity of these traits, due to population differentiation. 116

MATERIALS AND METHODS 117
This is a provisional file, not the final typeset article

Study species 118
The red mangrove, Rhizophora mangle L. 1753 (Malpighiales, Rhizophoraceae), is an evergreen 119 shrub or tree found along tropical and subtropical coastlines across the Americas, East Africa, 120 Bermuda, and on a number of outlying islands across the South Pacific (Proffitt and Travis, 2014;  121 Tomlinson, 2016; DeYoe et al., 2020) that can grow to heights of twenty-four meters (Bowman, 122 1917). Poleward expansion of the species is limited by freezing events (its current northern range 123 limit is roughly 29º N latitude; ( mixtures of two other mangrove species that are common in Florida: Laguncularia racemosa L. and 158 Avicennia germinans L.. We refer to plants from different sites as members of different populations 159 based on our previous work which found differences among sites based on molecular markers 160 (Mounger et al., 2021). At each population, we collected 20 propagules directly from each of 10 161 maternal trees separated by at least 10 m from each other to maximize the range of genetic variation 162 sampled within each population (Albrecht et al., 2013). Propagules from each maternal tree were at 163 least half-siblings but they could be more closely related due to the high selfing rate in the study area 164 (Proffitt and Travis, 2005) . 165 We refrigerated the propagules at 4 ºC for up to 14 days until we planted them in the greenhouse at 166 the University of South Florida Botanical Gardens. In the greenhouse, propagules from four of the 167 maternal trees at AC and nine of the maternal trees at FD failed to establish, so we returned to sample 168 propagules and maternal tissue from 8 new maternal trees at FD on August 12 and 29, and from the 169 same original maternal trees at AC on October 17. Hence, while most of the propagules were in the 170 greenhouse from the end of June until mid-October before they were exposed to treatments, 171 propagules from these maternal families had less acclimation time before treatments began. 172

Experimental treatments 173
We measured the length of each propagule and planted each in a 0.5-liter pot with a 50:50 mixture of 174 sand and peat soil. We watered the plants each day with tap water until we started applying the 175 salinity and N amendment in mid-October. We set up the experiment in five spatial blocks. succulence, maximum photosynthetic rate, and LMA. We included leaves that were attached, but not 222 50% green or fully developed in dry above ground biomass. We measured the biomass of above and 223 below ground tissues of all harvested plants after the tissues were dried at 60°C until they maintained 224 constant mass. Finally, we measured the total dry mass of leaves after drying in silica desiccant beads 225 for a minimum of seven days to constant mass. 226

Statistical analysis 227
We performed all statistical analyses in R , version 4.0.3 (R Core Team, 2020). All analyses reported 228 here used either the General Linear Model (GLM) or Generalized Linear-Mixed-Model (GLMM) 229 frameworks. We checked the residuals to assess normality on traits as appropriate; we did not 230 transform height growth or photosynthesis (lmer and glm), but we used the log link function (glmer) 231 for analysis of succulence, LMA, root to shoot ratio, and total biomass. We used the function lmer or 232 glmer implemented within the lme4 package (Bates et al., 2015) to fit a series of models and identify 233 the best fit model for each trait (Table 1). For each trait and for survival we began with a saturated 234 model that included as fixed factors salt, N, and their interaction. For change in height, we included 235 the covariate of height at the beginning of treatments in October. The saturated models also included 236 random effects of block, population and maternal family nested within population. In several cases, 237 models with random terms for block or population failed to converge (Table 1), most likely because  238 there were relatively few blocks and populations. In these cases, we began by treating these as fixed 239 effects. We used Anova (type III) in the package car (Fox and Weisberg, 2019) to evaluate the 240 significance of fixed effects when they were included in the best model. 241 242 To gain insight about variance explained by models, we calculated the R 2 approximations proposed 243 by (Nakagawa and Schielzeth, 2013). As these authors explain, in the context of GLMMs this leads 244 to two different sorts of R 2 , a marginal R 2 that reflects variance explained by fixed factors only, and a 245 conditional R 2 that reflects variance explained by both fixed and random factors. We report each of 246 these as appropriate, e.g., where the best model includes only fixed factors we report only the 247 marginal R 2 . 248 249 We assessed three survival states that were coded as 0 for live plants, 1 for dormant plants, and 2 for 250 plants that died during the experiment. We modeled survival using random effects logistic regression. 251 In one set of models, we included dormant plants as alive, and in another we excluded them. The 252 results were qualitatively similar, so we report only the case where dormant plants were treated as alive.

Treatment validation 256
To ensure that our N amendment treatments were not flushed out during the once weekly flow 257 through watering we measured the total dissolved N in leachate from a subsample of the seedlings. 258 We found that N was not significantly different between the low salt-no N and the high salt-high N 259 amended plants and, therefore, confirmed that we did not lose the N due to watering between 260 treatments (Mean Square = 0.11, F ndf 3/ddf 48 = 0.23, Pr(>F) = 0.88). 261

Trait responses to treatment 262
The only useful predictor among the fixed effects for the change in height is the height at the start of 263 treatments in October (  Figure S1). It's not obvious that any population has more among-270 family variability, but our design is limited to determine this. The best model for total biomass included only the random effects of populations and maternal 286 families within populations. The conditional R 2 = 0.15. The terms for populations and families nested 287 within them account for 14% and 23% of the random variance, respectively ( Figure S5). 288 289 Root to shoot biomass ratio was the only variable for which the data supported the saturated model as 290 best model (Figure 2). The conditional R 2 = 0.134, while the marginal R 2 = 0.007. An Anova to 291 evaluate these fixed effects revealed that the main effects of salt and N were not individually 292 significant but the interaction was (chisq = 7.24; p= 0.007). However, the small size of the marginal 293 R 2 suggests that these effects are mainly meaningful when conditioned on the random terms. The 294 random terms population, population and block account for 12%, 18%, and 10% of the random 295 variance, respectively. For this trait, population WI was different from all the others ( Figure S4). 296 297 The photosynthesis data were most limited in sample size since we were only able to assess one 298 individual with at least two healthy leaves for each surviving maternal line for each treatment 299 This is a provisional file, not the final typeset article (N=118). The best model was one including the fixed effects salt, N, and their interaction, but no 300 random effects (Table 1). As with root to shoot biomass ratio, an Anova to evaluate these fixed 301 effects showed the main effects of salt and N were not significant but the interaction was (chisq = 302 4.46; p = 0.035). At ambient N levels (no N added), maximum photosynthetic rate and root-to-shoot 303 ratio both declined with increasing salt concentration, but this negative impact of salt was absent or 304 reversed upon the addition of N fertilizer ( Figure 2B and 2C). 305 306 In summary, we found that some combination of salt and N treatments were included in the best 307 models for three of the six traits we measured. Succulence decreased in response to higher salt 308 (Figure 2A), but this response varied largely by family ( Figure S3). For root to shoot biomass ratio 309 and maximum photosynthetic rate the responses to experimental treatments depended on changes in 310 both salt and N ( Figure 2B and 2C). Root to shoot ratio also varied by family and population ( Figure  311 S4). 312

Survival 313
Of 1149 plants, 76 were unequivocally dead in April, 1073 unequivocally alive, and 166 dormant. 314 The number of plants that showed active growth ranged from 63% in HI to 91% in WI (Figure 3). 315 The number of plants that didn't show growth, but also didn't appear to be dead ranged from 3% in 316 WB and WI to 7% in UTB. We modeled survival by including these dormant plants alternatively as 317 either alive or dead. In both cases the best-supported model was one including a fixed effect for 318 population and a random effect for maternal family ( Figure S7) plastic in response to salinity and N amendment in putative adaptive traits that conserve water and 332 adjust allocation of N. Our study showed that only succulence was plastic in response to salt, 333 regardless of N treatment. Allocation to root and shoot biomass and maximum photosynthetic rate 334 were also plastic, but response of both traits to N amendment depended on the level of salt. This 335 supported our second prediction that response to salinity and N amendment will co-vary as plants 336 shift resources to maintain osmotic balance. We also found support for our third prediction that 337 populations would vary in putative adaptive traits, and in plasticity of these traits, due to population 338 differentiation. Importantly, every trait except for photosynthesis varied among population and 339 maternal families within populations. This was also true of survival. In fact, maternal family and 340 population were the most consistent predictors for variation in traits and survival. 341

Phenotypic plasticity in response to treatments
We expected that R. mangle propagule traits and survival would respond to salinity by decreasing 343 growth and respond to N fertilization by increasing growth. We also expected that N could alleviate 344 some of the effects of salinity as indicated by an interaction of the two conditions. However, we 345 found no response to treatments in height growth, succulence or total dry biomass. This may be due 346 to the propagules being supported by resources provided by the maternal tree, which in R. mangle 347 can support growth for at least a year (Ball, 2002;Proffitt and Travis, 2010). If the seedlings were 348 supported by these maternal reserves, height growth would likely be more correlated to propagule 349 length at collection which would be corrected for in the start-of-experiment (time zero) height 350 measurements that we included as a covariate. Because our treatment duration was only six months, 351 the lack of growth response to treatments is consistent with dependence on maternal reserves. 352 However, we did see seedling response to treatment in succulence, maximum photosynthetic rate and 353 allocation to root and shoot biomass. 354 Increased succulence is a common response to water deficiency under high salinity conditions 355 (Rosenthal et  1984; Kao et al., 2001). In fact, with N fertilization K. candel decreased leaf thickness when salinity 363 was increased (Kao et al., 2001). Thus, one possible explanation for our results is that the N fertilized 364 seedlings were able to reallocate resources and still maintain turgor and water uptake in the high salt 365 condition without increased succulence. 366 Although we saw significant plasticity for three of the six traits in response to our treatments, 367 response in only R:S and maximum photosynthetic rate depended on the interaction between salt and 368 N fertilization. We expected maximum photosynthetic rate to increase in response to N fertilization 369 because the enzyme RuBisCO, which catalyzes the dark reactions in photosynthesis, requires a large 370 amount of N (Sage et al., 1987;Andersson, 2008). Further, a meta-analysis across different species 371 and biomes showed increased maximum photosynthetic rate with increased N (Walker et al., 2014). 372 This could lead to increased shoot biomass which is supported by our analysis in the absence of salt 373 (i.e. decreased R:S in response to N). Despite this expectation, there was no overall response to high 374 N level independent of salt treatment. One reason might be that photosynthesis was limited by other 375 nutrients, not just N, and thus increasing N alone might not have been enough to elicit a response. In 376 a field study, dwarf R. mangle did not respond to N alone, but did increase biomass in response to 377 fertilization with N, phosphorus (P), and potassium ions (K + ), potentially because they were P limited 378 (Feller, 1995). We also expected that response to N amendment would depend on salinity. We found 379 that in plants treated with high salt, maximum photosynthetic rate was slightly enhanced by high N. 380 Possibly, the additional N enabled the plants to synthesize N-rich compatible solutes for osmotic 381 regulation and continue photosynthetic gain of carbon (Flowers and Colmer, 2008). Plants in high 382 salt also responded differently in allocation with increased N; instead of increasing shoot biomass, 383 they increased root biomass on average. 384

Variation within and among sites 385
Phenotypic variation that is maintained in common garden from within and among populations 386 would indicate R. mangle has heritable trait diversity to adapt to changing environmental conditions. This is a provisional file, not the final typeset article We found variation in height growth, succulence, R:S, LMA, and total dry biomass was largely 388 determined by maternal families within populations. Seedling survival depended on population and 389 varied among maternal families for all of the six populations. Proffitt and Travis (Proffitt and Travis,  390 2010) also found seedling survival varied among maternal families, as well as by location in the 391 intertidal zone. But in their study after three years, growth and survival did not reflect initial 392 propagule size (Proffitt and Travis, 2010). Our results support these previous findings that propagule 393 length is positively correlated to short term performance, which suggests that maternal reserves in the 394 R. mangle propagule can help the seedling survive, and larger propagules contain more maternal 395 reserves than smaller propagules (Ball, 2002;Proffitt and Travis, 2010). Because our study was a 396 short-term, controlled greenhouse study, maximum photosynthetic rate might be the best indicator for 397 an immediate response. Variation in maximum photosynthetic rate can ultimately manifest as 398 variation in growth and allocation of resources, particularly once the seedling has depleted maternal 399 reserves. The seedlings did not show significant differences in height growth or total dry biomass in 400 response to treatments. But given the plasticity we saw in maximum photosynthetic rate, and the one-401 year and three-year growth results found in a previous study of nearby populations ( The amount of heritable phenotypic diversity and differentiation we discovered in this study is an 408 important indicator of the potential for this species to respond to changing conditions, which may be 409 surprising. We previously reported low genetic diversity among these plants based on molecular 410 markers, which is expected to limit the potential for different responses among individuals. On the 411 other hand, we discovered high epigenetic diversity (Mounger et al., 2021), which could contribute to 412 phenotypic and functional diversity, and could be a mechanism underlying the type of phenotypic 413 differences and plasticity we found here (Zhang et  shoot biomass allocation respond to salinity, N level, or the combination of these conditions, but also 435 the magnitude of responses varies among populations and even maternal families within populations. 436 In addition, R. mangle seedling survival depended on maternal family for all of the six sites. 437 438 This variation in important traits and survival among families and among populations is particularly 439 interesting given our previous work with genetic markers that showed that these populations had low 440 genetic diversity, and little population differentiation. Considering the importance of this foundation 441 species for the functioning of the coastal ecosystem, the lack of genetic diversity might be alarming. 442 However, accumulating studies provide important evidence that genetic variation must be interpreted 443 with caution (Hufford and Mazer, 2003) and that the emphasis on only variation in DNA sequence 444 can be misguided (Keller, 2002(Keller, , 2014 changing environmental conditions and contribute to future adaptation to a complex mosaic of 447 environmental conditions. 448

DATA AVAILABILITY STATEMENT 449
All data and scripts will be submitted on zenodo and updated here as soon as possible. 450

Conflict of Interest 451
The authors declare that the research was conducted in the absence of any commercial or financial 452 relationships that could be construed as a potential conflict of interest. 453