Water potential regulation , stomatal behaviour and hydraulic transport under drought : 4 deconstructing the iso / anisohydric concept 5 6

In this review, we address the relationship between stomatal behaviour, water potential regulation and hydraulic transport in plants, focusing on the implications for the iso/anisohydric classification of plant drought responses at seasonal timescales. We first revise the history of the isohydric concept and its possible definitions. Then, we use published data to answer two main questions: (1) is greater stomatal control in response to decreasing water availability associated with a tighter regulation of leaf water potential (ΨL ) across species? and (2) is there an association between tighter ΨL regulation (~isohydric behaviour) and lower leaf conductance over time during a drought event? These two questions are addressed at two levels: across species growing in different sites and comparing only species coexisting at a given site. Our analyses show that, across species, a tight regulation of ΨL is not necessarily associated with greater stomatal control or with more constrained assimilation during drought. Therefore, iso/anisohydry defined in terms of ΨL regulation cannot be used as an indicator of a specific mechanism of drought-induced mortality or as a proxy for overall plant vulnerability to drought.

In this review we address the relationship between stomatal behaviour, water potential regulation 29 and hydraulic transport in plants, focusing on the implications for the iso/anisohidric classification of 30 plant drought responses at seasonal timescales. We first revise the history of the isohydry concept 31 and its possible definitions. Then, we use published data to answer two main questions: (1) is greater 32 stomatal control in response to decreasing water availability associated with a tighter regulation of 33 leaf water potential (Ψ L ) across species? And (2) is there an association between tighter Ψ L regulation 34 (~isohydric behaviour) and lower leaf conductance over time during a drought event? These two 35 questions are addressed at two levels: across species growing in different sites and comparing only 36 species coexisting at a given site. Our analyses show that, across species, a tight regulation of Ψ L is 37 not necessarily associated with greater stomatal control or with more constrained assimilation 38 during drought. Therefore, iso/anisohydry defined in terms of Ψ L regulation cannot be used as an 39 indicator of a specific mechanism of drought-induced mortality or as a proxy for overall plant 40 vulnerability to drought. 41 regulating leaf conductance to water vapour (g L ) and, therefore, transpiration and plant water status, 81 the iso/anisohydric classification is usually interpreted in terms of stomatal behaviour: isohydric 82 species maintain relatively stable Ψ L precisely because of their more strict stomatal control, whereas 83 anisohydric species would show a looser regulation of transpiration (Jones 1998;Tardieu & 84 Simonneau 1998). 85 86 Interpreted in this way, the iso/anisohydric categorization has strong implications for the 87 maintenance of assimilation under varying environmental conditions and, in general, for the carbon 88 economy of plants. This notion was used by McDowell et al. (2008) to distinguish between two 89 interrelated physiological mechanisms leading to plant mortality under drought. Isohydric species 90 would close stomata earlier during drought and, therefore, would depend more heavily on 91 carbohydrate reserves to meet continued carbon demands for respiration, osmoregulation or 92 defense. As a result, they would be more prone to die from carbon starvation. At the other extreme, 93 anisohydric species would close stomata later at the expense of suffering lower water potentials; 94 which would make them more vulnerable to hydraulic failure. There is no doubt that this framework 95 is appealing and has been hugely influential in shaping the research agenda on drought-induced 96 mortality in the last decade (Adams et al. 2009;Mitchell et al. 2012;Hartmann et al. 2013;Sevanto et 97 al. 2014 stomata of isohydric species are more sensitive relative to the water potentials at which they 131 operate, and therefore constrain assimilation further in these species. 132 133 In this review we aim at disentangling the relationship between stomatal behaviour, water potential 134 regulation and hydraulic transport in plants, focusing on seasonal timescales. We first revise the 135 history of the isohydry concept and its possible definitions. Then, we use data retrieved from the 136 literature to address two main questions: (1) is greater stomatal control in response to decreasing 137 water availability associated with a tighter regulation of leaf water potential across species? (2) is 138 there an association between tighter water potential regulation (~isohydric behaviour) and lower 139 leaf conductance (g L ) over time during a drought event? These two questions are addressed at two 140 levels: across species growing in different sites and comparing only species coexisting at a given site. 141 We finish by discussing the mechanisms behind the observed patterns and the ecological 142 implications in terms of characterizing plant responses to drought. 143 144 A brief history of the isohydry concept 145 The classification of plants based on their capacity to maintain a favorable water balance is a classic 146 theme in environmental plant physiology (e.g., Larcher 2003), which has led to a very rich, and not 147 always consistent terminology. Terrestrial vascular plants are able to maintain their water content 148 relatively stable despite fluctuations in water availability, thanks to a cuticle that minimizes 149 evaporative water losses and large central vacuoles that stabilize the water content in the 150 protoplasm (homoiohydric sensu Walter 1931). It was early realized, however, that vascular plants 151 differ substantially in the degree to which they regulate transpiration to maintain an adequate water 152 balance over diurnal and seasonal timescales. This variability led to the distinction between the 153 hydrostable and hydrolabile types (Stälfelt 1939) and between the isohydric and anisohydric 154 behaviours (Berger-Landefeldt 1936;Stocker 1956). There is a close correspondence between these 155 two classifications, with hydrostable/isohydric species having sensitive stomata and relatively Importantly, these two classifications originated before the water potential concept became widely 158 used in plant physiology (Slatyer & Taylor 1960;Scholander et al. 1965), and therefore focused more 159 (initially) on the ability to regulate transpiration than on the capacity to maintain relatively stable Ψ L 160 per se. The iso-/anisohydry dichotomy also predated the work on xylem water transport showing 161 that hydraulic conductivity was also a function of water potential (Milburn 1966;Zimmermann 1983;162 Tyree & Sperry 1989). All this might explain why a formal definition of the iso-/anisohydric 163 behaviours has remained somewhat elusive and current definitions usually mix the cause (stomatal 164 control) with its expected consequence (water potential regulation) (e.g., Jones 1998;Klein 2014;165 Meinzer et al. 2014;Skelton et al. 2015). 166

167
The current use of the iso-/anisohydry dychotomy stresses the physiological responses, but it is 168 tightly connected to a plethora of more ecological classifications of plant water use under drought 169 . These latter classifications have also a long tradition and oppose plants that tend to 170 reduce transpiration early on during drought development to save water (water savers, pessimistic 171 or drought avoiders) with plants that maintain transpiration rates for longer under drought (water 172 spenders, optimistic or drought tolerant) (Shantz 1927;Turner 1979;Jones 1980;Ludlow 1989). 173 These classifications have obvious implications in terms of competition for limited soil water 174 resources and the coexistence of different plant functional types in water-limited systems (Bunce et 175 al. 1977). More broadly, they reflect the general distinction between conservative/stress tolerant 176 and acquisitive/competitive strategies of plant resource use (Grime 1974;Díaz et al. 2016). 177 178

Towards a definition of iso-and anisohydry 179
It is our view that the iso-/anisohydry concepts will only be operational and useful in advancing our 180 understanding of plant water relations if we are able to define them precisely in terms of measurable 181 quantities. Despite the original focus on the regulation of transpiration (Berger-Landefeldt 1936; Stocker 1956), it seems more consistent with the current use of the terms and their etymology to 183 emphasize the maintenance of relatively constant leaf water potential (Tardieu & Simonneau 1998;184 Sperry et al. 2002). However, even in this more restrictive meaning isohydry can be defined in several 185 ways and at different temporal scales. We focus here on seasonal patterns and advocate for a 186 continuous measure of the degree of isohydry instead of distinguishing only between two idealized 187 extreme behaviours, which would always be somewhat arbitrary (Klein 2014;Martínez-Vilalta et al. 188 2014). 189 190 Figure 1 presents a hydraulic framework where alternative definitions of the degree of isohydry can 191 be mapped. Reduced soil water availability (lower, more negative Ψ soil ) may affect plant conductance 192 in two ways, by lowering its hydraulic conductance (K H ) and/or its leaf conductance (g L ). These 193 reductions, however, have opposite effects on the water potential difference through the plant (∆Ψ 194 = |Ψ L -Ψ soil |): whereas lower K H increases ∆Ψ, lower g L decreases ∆Ψ (everything else being equal). 195 The net change in ∆Ψ will thus depend on the balance between these two processes (i.e., the relative 196 sensitivity of transpiration vs. hydraulic transport to declining Ψ soil ) (Martínez-Vilalta et al. 2014), 197 with the complication that any change in Ψ L through changes in ∆Ψ will feedback onto K H and g L . 198 These feedbacks underlie the tight coordination between hydraulic and water vapor transport at the 199 plant level (Sperry & Love 2015). The dual control of g L by Ψ soil and Ψ L reflects the response of g L to 200 both soil water availability and leaf water status (Tardieu & Simonneau 1998) (Sperry 2013;208 Delzon & Cochard 2014;Trifilò et al. 2014). 209

210
We examine here three possible definitions of the degree of isohydry, assuming that predawn leaf 211 water potential (Ψ PD ) reflects soil water availability as perceived by the plant (but see Donovan et al. 212 2003), and that midday leaf water potential (Ψ MD ) measures Ψ L under maximum daily water demand. 213 Firstly, one could define isohydry simply in terms of the minimum seasonal Ψ MD experienced by a 214 given species or population, with relatively high (close to zero) values implying a more isohydric 215 behaviour. This is the definition adopted in practice by many studies, but it has important limitations, 216 as it is greatly affected by the severity of the drought conditions for which Ψ MD values are reported. 217 In practice, most of the variability in minimum Ψ MD across species is explained by minimum Ψ PD 218 ( Figure 2), and within a site (constant climate) Ψ PD is largely affected by rooting extension and depth 219 (Davis et al. 1998;West et al. 2012;Nardini et al. 2016) (see also section 'What determines minimum 220 operating leaf water potentials across species?' below). A second definition that stresses more the 221 regulation of water potential inside the plant would be the seasonal variability of Ψ MD , with more 222 isohydric species showing less variability. This variability could be measured as the range or standard 223 deviation of seasonal Ψ MD values. This definition reduces, to some extent, the effect of differences in 224 rooting systems across species, but it is still heavily affected by the range of water availability 225 conditions under which water potential measurements are taken. Finally, a third alternative would 226 F o r R e v i e w O n l y 10 We claim here that the relationship between stomatal control and the regulation of leaf water 234 potential, as well as its implications for plant survival under drought, have not been assessed as 235 thoroughly as one might expect for what we feel is one of the foundations of our current 236 understanding of plant drought responses. In order to test whether tighter water potential 237 regulation (~isohydric behaviour) is associated with lower g L under drought, we collected two 238 databases, one across species growing in different sites and a second one focusing on species 239 coexisting at the same sites. These databases, as well as the results of the corresponding analyses, 240 are presented as case studies in the following sub-sections. Note that we use the more general term 241 leaf conductance to water vapour (g L ) throughout the manuscript, but the values we take from the 242 literature are frequently reported as stomatal conductances. These two conductances are considered 243 equivalent unless stomata are nearly or completely closed, in which case other elements (e.g., 244 cuticular conductance) become relevant. 245 246

The relationship between stomatal control and water potential regulation across species 247
For this analysis we used the global water potentials database from Martínez-Vilalta et al. (2014). 248 This database contains Ψ PD and Ψ MD data from 83 articles and includes 102 species growing under 249 Temperate (n = 44, including one Boreal species), Mediterranean (n = 33), Tropical (n = 15) and Dry (n 250 = 10) climates. We first asked whether using different measures of isohydry (cf. above) resulted in 251 different rankings of species. Our results show that this is very clearly the case, particularly when 252 comparing the σ parameter with the two definitions directly based on Ψ MD (Figure 3). The correlation 253 between the minimum Ψ MD and the seasonal range of Ψ MD values across species was high (r = -0.94), 254 whereas it was very low and not even statistically significant when relating σ with minimum Ψ MD 255 values (r = -0.04) or with the Ψ MD range (r = 0.18). As an example, Tamarix chinensis, the second most 256 isohydric species in terms of σ (lowest decline in Ψ MD per unit of reduction in Ψ PD ) reached rather low 257 minimum Ψ MD (-4.6 MPa) and showed considerable seasonal range of Ψ MD values (2.9 MPa). These In a second step, we retrieved seasonal g L data from the papers included in the Martínez-Vilalta et al. 262 (2014) database. A total of 33 papers reported this variable as well as water potentials, including 44 263 species (Table S1). In some species, g L measurements were very abundant at high Ψ PD (well-watered 264 conditions), effectively providing several g L values at a given Ψ PD . To avoid putting excessive weight 265 to these measurements data was previously summarized by calculating the maximum g L by 0.1 MPa 266 intervals whenever more than one measurement was available per bin, and these values were used 267 in all further analyses. Using average instead of maximum g L per bin provided essentially identical 268 results. Mixed linear models were used to fit the (seasonal) relationship between log 10 (g L ) and Ψ PD 269 within and across species. Species and the combination of study by treatment (when present) nested 270 within species were included as random effects on the intercept (both) and slope of the model (only relative to other functional relationships between g L and Ψ PD (linear, power). The resulting model 277 provided an overall good fit to the data (conditional R 2 = 0.95, marginal R 2 = 0.50), and the species-278 level random slopes (γ) were used as an estimate of stomatal sensitivity to Ψ PD for each species. 279 Similarly, the intercept of the relationship was used to calculate maximum g L at Ψ PD = 0 (g L0 ). An 280 exponential relationship between g L and Ψ PD , as assumed here, implies that the relative change in g L 281 per unit change in Ψ PD , determined by the slope, is constant. Consequently, the γ values estimated 282 for each species were used to calculate the water potential decline required to reduce g L by 50% 283 (Ψ gL50 = ln(0.5)/γ). These values ranged between -0.62 MPa (Vicia fava) and -8.7 MPa (Larrea 12 tridentata). The value of Ψ gL50 could not be estimated for Tamarix ramosissima because its almost 285 flat relationship between log 10 (g L ) and Ψ PD resulted in an unrealistically low value (~-71 MPa). 286 287 Species' estimates of stomatal sensitivity (γ values) were used to test whether stomatal behaviour is 288 associated with different measures of isohydry in terms of water potential regulation. The results of 289 these analyses showed that, as expected, species with less sensitive stomata experience lower 290 minimum Ψ MD values (R 2 = 0.20, P = 0.003) and higher seasonal changes in Ψ MD (R 2 = 0.21, P = 0.002) 291 ( Figure 4a,b). However, stomatal sensitivity was unrelated to σ (R 2 = 0.00, P = 0.94) (Figure 4c), 292 implying that higher stomatal sensitivity did not result in stronger Ψ MD regulation as Ψ PD declined 293 under drought. Very similar relationships were obtained if Ψ gL50 was used instead of γ to characterize 294 stomatal responses (excluding T. ramosissima). Maximum leaf conductance (g L0 ) was unrelated to the 295 three isohydry measures we employed (P > 0.3 in all cases). Overall, these results indicate that 296 species operating at higher (less negative) water potentials tend to close stomata faster with 297 declining Ψ PD than species experiencing lower water potentials, but this does not imply an 298 association between stomatal control and water potential regulation inside the plant (as measured 299 by the σ parameter) across species. 300 301 We can further ask whether different degrees of water potential regulation are associated to greater 302 constrains to gas exchange through stomatal regulation (over time). Leaf conductance (g L ) values 303 were obtained from the same studies from which water potential data had been retrieved (Table S1) 304 and were pre-processed in exactly the same way as explained above (e.g., 0.1 MPa binning). We 305 analyzed the relationship between our three measures of isohydry and two measures of seasonal 306 stomatal behaviour: average g L over the whole study period covered by each study (g L,mean ), as a 307 measure of absolute gas exchange during a drought event; and the ratio of g L,mean to maximum 308 measured g L (g L,ratio ) over the same period, as a relative measure of gas exchange. Our results show 309 that none of our three measures of isohydry was related to g L,mean across species (P > 0.36 in all cases) Figure 5d,e,f). Similarly, the σ parameter and g L,ratio were unrelated (P = 0.58) ( Figure 5c). However, 311 species experiencing lower minimum Ψ MD or wider seasonal ranges of Ψ MD presented lower values of 312 g L,ratio (R 2 = 0.16, P = 0.007; and R 2 = 0.12, P = 0.022; respectively) ( Figure 5a,b). This result implies that 313 species operating at lower water potentials or experiencing wider water potential fluctuations closed 314 stomata more strongly during the period covered by each study than those species operating at less 315 negative water potentials, contrary to the notion that lower g L is associated with maintaining less 316 negative water potentials across species. Interestingly, stomatal sensitivity (measured as γ or Ψ gL50 ) 317 was unrelated to g L,mean or g L,ratio (P > 0.25 in all cases), due to the fact that species with more 318 sensitive stomata tended to operate at higher Ψ L ( Figure 4) and, thus, closed stomata to a similar 319 extent than species with less sensitive stomata but operating at lower Ψ L . 320 321

Stomatal control vs. water potential regulation among coexisting species 322
Assessing the relationship between stomatal control and water potential regulation across species 323 occupying different environments, as done in the previous section, may be problematic because it 324 mixes plants growing under very different conditions, including exposure to drought stress. To 325 overcome this limitation we conducted a similar analysis focusing on the comparison of coexisting 326 species measured concurrently in the same sites, and thus having similar exposure to drought. A new 327 global database was compiled using mostly published sources. We searched the literature for studies 328 fulfilling the following criteria: (1) they compared different species (or cultivars with contrasted 329 stomatal behaviour in the case of Vitis vinifera) growing at the same site under the same 330 environmental conditions; (2) focused on the study of drought effects (including experimental and 331 naturally occurring droughts) over a period of weeks to months; (3) reported multiple measures of 332 Ψ PD , Ψ MD and g L ; (4) these three variables were measured concurrently and could be linked to each 333 other (directly or through third variables such as time); and (5) the range of measured Ψ PD values was 334 > 1 MPa for at least one of the species included in the study, to ensure drought severity was 335 substantial. We also added an unpublished dataset including measures on Phillyrea latifolia and Quercus ilex planted on the ground in a tunnel greenhouse and subjected to a drought-simulation 337 experiment (N. Garcia-Forner et al., unpublished). Altogether, we compiled data from 15 datasets, 338 covering mostly Mediterranean (n = 9) and Dry climates (n = 3) (Table S2). Each study compared 339 between two and seven species growing under the same environmental conditions (33 species in 340 total), except two studies on Vitis vinifera that compared two different cultivars each. 341 342 Using this database we calculated the species-level slopes of the relationships between Ψ MD and Ψ PD 343 (parameter σ) and between log 10 (g L ) and Ψ PD (parameter γ) as explained in the previous section. We 344 fitted a different mixed model for each study, with species as a random effect on the intercept and 345 slope. Model fits were generally good, with conditional R 2 = 0.3 -0.98 for the regressions between 346 Ψ MD and Ψ PD and conditional R 2 = 0.32 -0.93 for the regressions between log 10 (g L ) and Ψ PD . As 347 before, species' estimates of stomatal sensitivity (γ and Ψ gL50 values) were used to test whether 348 stomatal behaviour is associated with different measures of isohydry (minimum Ψ MD , range of Ψ MD 349 values and σ) using mixed models with site as a random factor. In all cases, model fit in terms of AIC 350 was best when random effects were included on the intercept only. The overall relationships were 351 similar to those obtained in the previous section using the global database (compare Figure 6a,c,e 352 with Figure 4), with γ being positively related to minimum Ψ MD (P = 0.001), negatively related to Ψ MD 353 range (P = 0.002), and unrelated to σ (P = 0.17). However, since we were interested in the 354 comparison within sites and the previous analysis mixes the effect of stomatal sensitivity within and 355 between sites, we also used mixed linear models to fit the relationships between the three measures 356 of isohydry and centered γ (or Ψ gL50 ) values. Centering was achieved by subtracting the average γ (or 357 Ψ gL50 ) for the corresponding site to each species' γ (or Ψ gL50 ) value, and ensured that fixed effects 358 were evaluated only within sites. The relationship between centered stomatal sensitivity and the 359 three isohydry measures was not significant (P > 0.05 in all cases) (Figure 6b,d,f), although the 360 (negative) effect of γ on σ was close to significant (P = 0.06). Overall, these results indicate that We used a similar approach to explore whether different degrees of water potential regulation are 365 associated to greater constrains on gas exchange through stomatal regulation when comparing 366 coexisting species. As before, mixed linear models with site as random factor were used to fit the 367 relationships between g L,mean and g L,ratio (response variables) and the three measures of isohydry 368 (centered minimum Ψ MD , centered Ψ MD range and centered σ). In all cases, the best fitting model in 369 terms of AIC included the random effect of site on the intercept but no effect of the fixed 370 explanatory variable (P > 0.35 for all model comparisons), indicating that our three measures of 371 isohydry were unrelated to stomatal behaviour when comparing different species measured within a 372 site (Figure 7). The corresponding plots using non-centered explanatory variables instead of centered 373 values are provided in Figure S1. 374 375 Why are water potential regulation and stomatal behaviour decoupled across species? 376 The results reported in the previous sections have to be considered with caution, as they come from 377 a synthesis of different data sources, each covering different time periods and using potentially 378 different experimental protocols. However, our analyses at two different levels (across species and 379 within sites) suggest that water potential regulation and stomatal control are largely unrelated across 380 species. Of course, this is not to mean that these variables are not related in general. It is very well 381 established both theoretically and empirically that, for a given plant, stomatal closure reduces 382 transpiration and hence limits the water potential difference between the soil and the leaves and the 383 risk of hydraulic failure (Tyree & Sperry 1988;Jones & Sutherland 1991;Cochard 384 et al. 2002). However, the situation becomes more complex when we compare different species, 385 even if they grow at the same site. This is because water potential dynamics are affected by several 386 plant attributes that are coordinated across species, including stomatal behaviour but also hydraulic  (Figure 1). At longer time scales, stomatal conductance and 394 transpiration (including the effects of vapour pressure deficit and leaf area dynamics) will determine 395 the rate of water extraction from the soil and, therefore, will feed back into Ψ PD : species showing 396 higher transpiration rates will deplete soil water faster and hence experience also faster reductions 397 in Ψ PD over time. In addition, once the hydraulic system of the plant is disconnected from the soil 398 (complete loss of hydraulic conductivity somewhere in the hydraulic pathway) Ψ PD will cease to track 399 fluctuations in soil water potential. 400 401 Arguably, the ultimate minimum water potential a plant can withstand is determined by the 402 vulnerability of its hydraulic system (Brodribb & Cochard 2009;Nardini et al. 2013;Urli et al. 2013;403 Brodribb et al. 2014). The high degree of phylogenetic conservatism in vulnerability to xylem 404 embolism (Maherali et al. 2004) supports the notion that hydraulic vulnerability may have driven 405 differences in water potential regulation over evolutionary time scales. Relatively high hydraulic 406 vulnerability (e.g., low resistance to xylem embolism) tends to be associated with tight stomatal 407 control across species (Brodribb et al. 2003;Arango-Velez et al. 2011;Klein 2014) and also within 408 species (e.g., when comparing Vitis vinifera cultivars with contrasted stomatal behaviour, Tombesi et 409 al. 2014). This association is also supported by the positive relationship between the water potential 410 causing 50% loss of hydraulic conductivity in stem xylem (Ψ PLC50 ), obtained from the Choat et al. 411 (2012) database, and stomatal sensitivity as obtained from our analysis (cf. '1. The relationship 412 between stomatal control and water potential regulation across species' section above) ( Figure S2). The proximal mechanism underlying the coordination between vapour and liquid phase water 415 transport in plants is provided by the response of stomata to hydraulic signals (Meinzer 2002;Sperry 416 et al. 2002;Buckley 2005;Meinzer et al. 2009). This response is complex and has several potentially 417 problematic aspects, including the fact that Ψ L , the obvious integrator of leaf water status to which 418 stomata may respond through its effect on guard cell turgor (Figure 1), is also the same variable that 419 is maintained relatively constant as a result of stomatal control. There is indeed plenty of evidence 420 showing stomatal responses to hydraulic signals without significant changes in bulk Ψ L (Sperry & 421 Pockman 1993;Saliendra et al. 1995;Salleo et al. 2000;Hubbard et al. 2001). However, this is still 422 consistent with a regulation of stomatal conductance through a negative feedback with Ψ L if we 423 consider that embolism itself may provide the amplification required to achieve nearly homeostatic 424 regulation of leaf water potential (Buckley 2005) (Figure 1). Hormonal signals play also a prominent 425 role in modulating stomatal responses, particularly through abscisic acid (ABA) synthesis in roots and 426 leaves and its subsequent accumulation in leaves (Mittelheuser & Van Steveninck 1969;Zhang & 427 Davies 1989;Bauer et al. 2013;Tombesi et al. 2015). Recent evidence suggests that stomatal closure 428 under drought stress evolved from a passive, purely hydraulic process, to the more complex 429 mechanism involving hormonal signalling and active ion exchange between guard and epidermal cells 430 currently characterizing most angiosperms, with stomatal regulation in conifers being intermediate 431 between these two modes (Brodribb & McAdam 2010;McAdam & Brodribb 2014. 432 433 Even within conifers, stomatal closure seems to be induced by two contrasted mechanisms. Whereas 434 some species show fast stomatal closure under drought in response to sustained high levels of ABA in 435 leaves, a second group of species show slower stomatal responses at lower Ψ L (Brodribb et al. 2014). 436 Importantly, these two contrasting modes of stomatal regulation are associated to differences in 437 vulnerability to xylem embolism, with the first mode of stomatal regulation described above being 438 characteristic of species with more vulnerable xylem (Brodribb et al. 2014). This association has been what plant attributes are associated with maintaining relatively high and stable (as opposed to low 452 and declining) leaf water potentials under drought? The first one is obviously deep rooting, 453 particularly considering the tight relationship between Ψ MD (~Ψ L ) and Ψ PD (~Ψ soil ) reported in Figure  454 2. Species with more extensive and deeper root systems are able to access more stable water 455 sources, thus buffering changes in hydrological conditions (Jackson et al. 2000;Oliveira et al. 2005). 456 Accordingly, deep-rooted species should be able to maintain less negative and more stable water 457 potentials (particularly Ψ PD ), everything else being equal. Although there are many cases in which this 458 is the case (Bucci et al. 2009;West et al. 2012), there appear to be also counterexamples in which 459 species known to be relatively shallow-rooted operate at higher water potentials than coexisting 460 deep-rooted species (West et al. 2007;Plaut et al. 2012;Aguadé et al. 2015). 461 462 Minimum Ψ L is also associated to the vulnerability to xylem embolism, both at the local and global 463 scales (Pockman & Sperry 2000;Choat et al. 2012), with more resistant species being able to operate 464 at lower Ψ L . Globally, a positive relationship between resistance to xylem embolism and rooting depth is to be expected, as these two characteristics tend to occur under similar environmental 466 conditions (Schenk & Jackson 2002;Maherali et al. 2004;Choat et al. 2012). Within a given site, 467 however, species that are hydraulically more vulnerable and cannot sustain very low water potentials 468 may require deeper root systems. Accordingly, a number of studies report negative relationships 469 between resistance to xylem embolism and rooting depth (Hacke et al. 2000;470 Lopez et al. 2005), although exceptions occur (Pivovaroff et al. 2016;Nardini et al. 2016). Species 471 that are shallow-rooted and relatively vulnerable to xylem embolism may disconnect their hydraulic 472 system from the soil early during drought development. This disconnection may be purely hydraulic 473 or physical, involving fine root mortality (Bauerle et al. 2008;Espeleta et al. 2009), and it is frequently 474 associated to drought deciduousness (Kolb & Davis 1994;Miranda et al. 2010;Hoffmann et al. 2011). 475 Dynamic aspects related to vertical water redistribution in the soil may also be important in 476 explaining differences in Ψ L dynamics between coexisting species with different root distributions 477 (Meinzer et al. 2004;Neumann & Cardon 2012). 478 479 If the hydraulic system of the plant remains connected to the soil, the rate of transpiration and water 480 uptake will affect Ψ soil dynamics in the rooting zone and, hence, will contribute to explain differences 481 in Ψ L dynamics across species (Mitchell et al. 2012) (Figure 1). Maintenance of relatively high 482 transpiration rates under drought (high g L ) will deplete soil water resources faster and will result in 483 steeper declines in Ψ soil and Ψ PD over time. An important additional aspect is that in general this 484 effect will be driven not only by the water uptake of the target plant but also by all individuals with 485 roots within its belowground neighborhood (Casper & Jackson 1997;Zavala & Bravo de la Parra 486 2005). Our results suggest that the positive relationship between stomatal sensitivity and minimum 487 Ψ MD across species (Figures 4 and 6a) may be more associated to the effect of water uptake (or to 488 the covariation with rooting depth) than to the role of stomatal control on water potential regulation 489 inside the plant (as measured by the σ parameter). 490 Assuming steady state, the water potential gradient within the plant will be determined by the 492 maximum transpiration rate per unit of hydraulic transport capacity, which defines the leaf water 493 potential at Ψ soil ≈ 0; and by the relative sensitivity of transpiration and the plant hydraulic system to 494 declining Ψ PD (σ; cf. Figure 1) (Martínez-Vilalta et al. 2014). An important result of recent data 495 syntheses is that the vulnerability of stem xylem to embolism (measured as Ψ PLC50 ) appears to be 496 more variable than stomatal sensitivity across species, with Ψ gL50 rarely falling below -4 MPa represents an adaptation for coping with low and fluctuating water potentials (Meinzer et al. 1986(Meinzer et al. , 508 2014. Substantial reductions in Ψ tlp with declining Ψ L (together with high hydraulic 509 compartmentalization in the leaf, Buckley et al. 2015) may help explain the puzzling result that many 510 species from dry habitats appear to operate largely below their Ψ tlp as determined on fully 511 rehydrated samples (Meinzer et al. 2014). 512

513
In most field situations non-steady state conditions prevail, implying that the water content in the 514 plant is not constant. Hydraulic capacitance, the water content change per unit of variation in water 515 potential, allows the plant to (partially) uncouple the changes in transpiration from water potential 516 dynamics, effectively dampening the temporal fluctuations in Ψ L (Meinzer et al. 2003(Meinzer et al. , 2009Sperry et  other hydraulic traits. In particular, higher sapwood capacitance seems to be associated with higher 519 water potentials, lower resistance to xylem embolism and narrower hydraulic safety margins (Pratt 520 et al. 2007;Sperry et al. 2007;Meinzer et al. 2009;Mcculloh et al. 2014). Clearly, capacitance and 521 water storage need to be considered as additional elements, together with changes in stomatal and 522 hydraulic conductance, determining the water potential regulation inside the plant, and hence Ψ L at 523 a given Ψ soil (Matheny et al. 2015). At very low water potentials stomata close completely and plant 524 water losses are driven by leaf cuticular conductance. If severe embolism has not yet developed, 525 cuticular conductance will determine the time needed to reach hydraulic failure and thus low 526 cuticular conductance can confer substantial drought tolerance (Scoffoni et al. 2011;Blackman et al. 527 2016). However, our knowledge on the determinants and implications of the variability of cuticular 528 conductance across species is limited (the last review we are aware of was written 20 years ago by 529 Kerstiens (1996)) and requires further research. 530 531

Implications for drought-induced mortality mechanisms 532
An important implication of our results is that isohydric species in terms of water potential regulation 533 are not necessarily more carbon limited than anisohydric species. When comparing species 534 coexisting within a given site there is no relationship between any of the three measures of isohydry 535 used in this study and average g L (either in absolute terms or relative to the seasonal maximum g L ; 536 Figure 7). When this relationship is assessed across species growing at different sites, species 537 experiencing lower minimum Ψ MD or wider seasonal Ψ MD range tend to have lower average g L 538 (relative to its maximum) ( Figure 5), despite also having lower stomatal sensitivity (Figure 4). These 539 results simply reflect that the range of minimum Ψ MD across species and sites is wider than the range 540 of stomatal sensitivities, which appears to be relatively constrained across species ( In this review we have shown that, contrary to what is usually assumed, a tight regulation of Ψ L is not 575 necessarily associated with greater stomatal control across species. Therefore, we advocate for a 576 clear and quantitative definition of iso/anisohydry that separates these two concepts. This distinction 577 is important, as iso/anisohydry defined in terms of Ψ L regulation tells us little by itself about leaf gas 578 exchange dynamics or the degree of hydraulic or carbon limitations under drought. Therefore, it 579 cannot be used as an indicator of a specific mechanism of drought-induced mortality (sensu 580 McDowell et al. 2008) or as a proxy for overall vulnerability to drought. The way we understand and 581 define the iso/anisohydryc behaviours has important implications for the modelling of drought 582 responses at scales that range from the individual to the ecosystem and the Biosphere (Roman et al. 583 2015;Combe et al. 2016). 584 585 Several issues remain that limit our understanding of plant water relations and our capacity to 586 predict vegetation responses under ongoing climate change. Among other aspects, significant 587 advances could be achieved by: 588 • Improving our understanding of how relevant traits scale up from the tissue to the whiole-589 plant levels (Sperry et al. 2007;Meinzer et al. 2010;Petit & Anfodillo 2011) and, in particular, 590 resolving where the hydraulic bottleneck is in the soil-plant-atmosphere continuum and how 591 this bottleneck might change during drought. Candidates include the rhizosphere, the xylem 592 of different organs and extraxylary tissues. This is a long-standing issue in plant hydraulics 593     water availability (higher absolute value of Ψ soil ) may affect plant conductance in two ways, by 1162 lowering its hydraulic conductance (K H ) and its leaf conductance (g L ). These reductions have opposite 1163 effects on the water potential gradient through the plant (∆Ψ), so that the net change in ∆Ψ will 1164 depend on the balance between these two processes, with the complication that changes in leaf 1165 water potential (Ψ L ) will feedback onto K H and g L . High transpiration rates (through high g L ) cause 1166 faster reductions in Ψ soil , unless the hydraulic system of the plant becomes disconnected from the 1167 soil. Important plant attributes and processes (rooting depth, capacitance, osmoregulation) have 1168 been omitted for simplicity. See text for further details. between Ψ MD and Ψ PD (σ, MPa MPa -1 ) for 44 species (see text for details). Stomatal sensitivity was 1185 estimated as the slope of the (seasonal) relationship between log 10 (g L ) and Ψ PD (see text for further 1186 details). Solid and dashed lines indicate significant and non-significant relationships between 1187 variables. Grey-vertical lines show the equivalence of γ in terms of the water potential required to 1188 reduce leaf conductance to water vapour by 50 % (Ψ gL50 , in MPa). Species abbreviations are given in 1189 Table S1. the slope of the relationship between Ψ MD and Ψ PD (σ, MPa MPa -1 ). Stomatal behaviour over time is 1194 characterized using two variables: the ratio of average g L to maximum measured g L (g L,ratio ; panels a, 1195 b, c) and average g L (g L,mean ; panels d, e, f). Solid and dashed lines indicate significant and non-1196 significant relationships between variables. Species abbreviations are given in Table S1. 1197 1198 Figure 6. Relationship between stomatal sensitivity to decreasing predawn leaf water potential and 1199 three different measures of isohydry for species coexisting at a given site (see text for details). 1200 Isohydry measures include minimum midday water potential (minimum Ψ MD , MPa), seasonal range 1201 of Ψ MD (MPa), and the slope of the relationship between Ψ MD and Ψ PD (σ, MPa MPa -1 ). Stomatal 1202 sensitivity was estimated as the slope of the (seasonal) relationship between log 10 (g L ) and Ψ PD (see 1203 text for further details), and it is expressed in two different ways: as absolute γ values (log(mmol m -2 1204 s -1 ) MPa -1 ; panels a, c, e) and as centered γ values (log(mmol m -2 s -1 ) MPa -1 ; panels b, d, f). Each dot 1205 indicates a species and colors designate studies. Species measured in the same study are linked by 1206 colored lines, to facilitate assessing the relationships within sites. Grey-vertical lines show the 1207 equivalence of γ in terms of the water potential required to reduce leaf conductance to water vapor 1208 by 50 % (Ψ gL50 , in MPa). Study codes are given in Table S2. (MPa) and the slope of the relationship between Ψ MD and Ψ PD (σ, MPa MPa -1 ). Stomatal behaviour 1214 over time is characterized using two variables: the ratio of average g L to maximum measured g L 1215 (g L,ratio ; panels a, b, c) and average g L (g L,mean ; panels d, e, f). Each dot indicates a species and colours 1216 designate studies. Species measured in the same study are linked by coloured lines, to facilitate 1217 assessing the relationships within sites. Study codes are given in Table S2.   Figure S1. Relationship between stomatal behaviour over time and three different measures of isohydry for species coexisting at a given site (see text for details). Isohydry measures include minimum midday water potential (minimum Ψ MD , MPa), seasonal range of Ψ MD (MPa) and the slope of the relationship between Ψ MD and Ψ PD (σ, MPa MPa -1 ). Stomatal behaviour over time is characterized using two variables: the ratio of average g L to maximum measured g L (g L,ratio ; panels a, b, c) and average g L (g L,mean ; panels d, e, f). Each dot indicates a species and colours designate studies. Species measured in the same study are linked by coloured lines. Study codes are given in Table S2. Figure S2. Relationship between stomatal sensitivity to decreasing predawn leaf water potential (γ, in log(mmol m -2 s -1 ) MPa -1 ) and the water potential at 50 % loss of hydraulic conductivity in the stem (Ψ 50PLC ) across species. Separate linear regressions are depicted for angiosperms and gymnosperms in red and blue, respectively. Grey shadows around lines indicating 95% confidence intervals. A linear model accounting for the differences between angiosperms and gymnosperms in the intercept of the relationship resulted in a highly significant Ψ PLC50 effect (R 2 = 0.33, P = 0.007). Overall model fit increased substantially if Tamarix ramosissima, a clear outlier of the relationship, was excluded from the analysis (R 2 = 0.44, P < 0.001). Species abbreviations are given in Table S1. Ψ 50PLC data was obtained from  Table S1. Characteristics of the species considered in this study and list of data sources used to build the database of leaf water potentials and stomatal responses (cf. '1. The relationship between stomatal control and water potential regulation across species' section in main text).