The wild-type accession of Arabidopsis thaliana L. used in this study was Col-8 (European Arabidopsis Stock Centre catalogue no. N60000). Besides, the auxin influx AUX1 mutants aux1-T (N657534), aux1-7 (N9583); the auxin efflux mutants pin2/eir1-1 (N8058), pin3-4 (N9363), pin3-5 (N9364), pin4-3 (N9368), and pin7-2 (N9366); and auxin signalling mutants iaa7/axr2-1 (N3077) and tir1-1 (N3798) were obtained from the European Arabidopsis Stock Centre. The ethylene-insensitive mutants etr1-1 (ETHYLENE RESPONSE 1) (Bleecker et al., 1988), ein2-1 (ETHYLENE INSENSITIVE 2) (Guzmán and Ecker, 1990), and ein3-1 (Roman et al., 1995) were kindly provided by Dr. Mike Roberts (Lancaster University, United Kingdom). The auxin reporter line DR5::GFP (Ottenschläger et al., 2003) was a kind gift from Prof. Klaus Palme (University of Freiburg, Germany). All Arabidopsis lines were in the Columbia background.

Surface-sterilised seeds were sown on solid medium containing 0.02 x B5 medium, 1 mM KNO₃, 0.5% (w/v) sucrose and 1% agar in 90 mm diameter Petri dishes (Zhang and Forde, 1998). After stratifying the seed in the dark (4°C) for 2–3 days, the Petri dishes were incubated in a vertical orientation in a growth room at 22°C with a 16 h light period and an irradiance of 100 μmol m^-2 s^-1. Four to five days later, seedlings with similar root length were transferred to fresh plates containing ABA at different concentrations. Five inhibitors were added to the growth medium as required: namely, the ethylene biosynthesis inhibitor AVG (0.3 or 0.5 μM) (A6685, Sigma-Aldrich); the ethylene perception inhibitor silver thiosulfate (STS, 10 μM); and the auxin efflux inhibitors N-1-naphthylphthalamidic acid (NPA, 10 μM) (PS343, Sigma-Aldrich), 2,3,5-triiodobenzoic acid (TIBA, 10 μM) (T5910, Sigma-Aldrich); and the auxin influx inhibitor 3-chloro-4-hydroxyphenylacetic acid (CHPAA, 10 μM) (224529, Sigma-Aldrich). For plates with treatments, 3–6 seedlings were placed on 9 cm diameter plates (25 ml medium), or 7–9 seedlings on 12 cm square plates (50 ml medium). The top one-fifth of the agar medium was excised so that the shoot was not in direct contact with the medium. ABA (A1296, Sigma-Aldrich) stock solutions were made in 10 mM (+ABA) with 0.03 M KOH. A 60 mM STS solution was freshly prepared by mixing 300 mM silver nitrate with 300 mM sodium thiosulphate in a 1:4 (v/v) ratio.

Primary root growth was monitored during the 3–6 days after seedlings were transferred to the treatment plates by marking the position of the root tips on the base of the plate at 24 or 48 h intervals. At the end of each experiment, the plates were imaged on a flat-bed scanner with a ruler as the reference. The images were semi-manually analysed using Optimas Image Analysis software (Version 6.1 Media Cybernetics, Inc., United States) for root length. The primary root lengths at the beginning and the end of an experiment were presented in Supplementary Table 1.

After 3 days of ABA treatments, DR5::GFP seedlings were stained briefly (50 s) with 10 μM propidium iodide. GFP and propidium iodide fluorescence was then detected using a Leica SP2-AOBS confocal laser scanning microscope and the images were electronically superimposed using LCS Lite software (Leica, Germany). Quantification of the GFP fluorescence signal was performed using ImageJ (National Institutes of Health, United States).

The statistical software SPSS 21.0 (IBM, United States) was used to perform one-way or two-way ANOVA with Tukey’s post hoc test at the P < 0.05 level. The effect size of those ANOVA was reported by eta² or partial eta². The criteria for effect size: no effect, eta² = 0; small, eta² = 0.0099; medium, eta² = 0.0588; large, eta² = 0.1379 (Richardson, 2011).

A detailed comparison of the effects of a range of ABA concentrations on root elongation was performed by transferring 4 day-old Arabidopsis seedlings to vertical agar plates containing 0 (control), 0.1, 1, and 10 μM ABA and measuring the increase in root length at daily intervals over the following 6 days (Figure 1). The results showed that 10 μM ABA inhibited root growth by about 40% while 0.1 μM ABA stimulated growth by almost 20% when measured over the 6-day period (Figure 1A). The stimulatory effect of 0.1 μM ABA persisted over the duration of the treatment and by the 6th day the roots were growing at a rate which was more than 30% faster than the control (Figure 1B). It appears that the intermediate concentration of ABA used (1 μM) is close to the threshold for the transition from stimulation to inhibition as it had little effect on root elongation (Figures 1A,B). In subsequent experiments, concentrations less than 1 μM ABA (usually 0.1 μM ABA) were therefore used for studying the stimulatory effect of low ABA concentrations and concentrations greater than 1 μM ABA (usually 10 μM ABA) were used for studying the inhibitory effect of high ABA concentrations.

It has previously been established that the inhibitory effect of high ABA concentrations on root growth is an ethylene-dependent process (Ghassemian et al., 2000). To confirm these findings under our experimental conditions and to investigate whether the stimulatory effect of low ABA concentrations is also ethylene-dependent, seedlings were treated with different concentrations of ABA in the presence or absence of either AVG (an ethylene biosynthesis inhibitor) or STS (an ethylene perception inhibitor). The primary root elongation rates were determined over a 4-day period of treatment. When 0.3 or 0.5 μM AVG was included along with the 10 μM ABA treatment, the inhibitory effect was relieved as measured after either 1 day (Figure 2A) or 4 days (Figure 2B).

In contrast to AVG’s ability to interfere with the inhibitory effect of ABA, the presence of either 0.3 or 0.5 μM AVG had no influence on the trend of the stimulatory effect after adding 0.1 μM ABA (Figures 2A,B). In addition, AVG treatment tended to promote root growth which might mask the stimulatory effect of ABA at low concentrations (Figures 2A,B and data not shown). Thus while ethylene biosynthesis is required for the inhibitory effect of high ABA concentrations it is not required for the stimulatory effect of low ABA concentrations.

When 10 μM STS was used to interfere with ethylene perception it almost completely overcame the inhibitory effect of 10 μM ABA when measured after the first 2 days of treatment (Figure 2C). This antagonistic effect was lost when root growth was measured over a 4-day period (Figure 2D), which we attribute to the known instability of STS (Ag⁺) when exposed to light. However, when included along with 0.1 μM ABA, the STS did not interfere with the stimulatory effect on root growth as measured after either 2 or 4 days (Figures 2C,D). Therefore, the inhibitory effect of high ABA concentrations, but not the stimulatory effect of low ABA concentrations, could be eliminated by interfering with ethylene perception.

To look further into the role of ethylene signalling in the two components of the root response to ABA, seedlings of three ethylene-insensitive mutants (etr1-1, ein2-1, and ein3-1) were treated with a range of concentrations of ABA. Two-way ANOVA showed that the primary root elongation rate was affected by genotype, ABA treatment and their interaction in the first 24 h and 4 days after treatment (P < 0.0001, Supplementary Table 2). Thus, the four genotypes responded to those ABA treatments differently and the effect of those two factors depend on each other. Figure 3 displays the effect of ABA treatment in each genotype, while Supplementary Figure 1 shows the effect of genotype under each ABA treatment.

All three mutants to varying degrees showed a diminished response to the inhibitory effect of high [ABA] compared to the wild-type (Figure 3). This was particularly evident in etr1-1 and ein2-1 during the 1st day of treatment when even the highest concentration of ABA (30 μM) had no effect on the root elongation rate, and inhibited root elongation by only 14% in etr1-1 and even stimulated root elongation by 6% in ein2-1 compared to 48% inhibition in the wild-type (Figures 3A,C,E). A much less pronounced effect was seen in ein3-1, where 30 μM ABA inhibited root elongation by 35% over the 1st day of treatment (Figure 3G). Over the 4-day period of treatment the inhibitory effect of the high ABA concentrations was stronger in all lines, but the same pattern of decreased sensitivity in the mutants was observed (Figures 3B,D,F,H). The low ABA concentrations (0.1 and 0.2 μM) stimulated root elongation of the wild-type by ∼20% in the 1st day after treatment and by ∼30% over the full 4 days of treatment (Figures 3A,B). Similarly, the low ABA concentrations also stimulated root elongation of the three ethylene-insensitive mutants as seen after either 1 or 4 days (Figures 3C–H). These results confirmed the evidence from the STS treatment (Figures 2C,D) that ethylene signalling is important for the inhibitory effect of high [ABA], but not for the stimulatory effect of low [ABA].

To investigate the role of auxin transport in the root responses to ABA, two auxin efflux inhibitors (NPA and TIBA) and an auxin influx inhibitor (CHPAA) were firstly employed in this study. In this experiment, the stimulatory effect of the low ABA concentration (0.1 μM) was only seen after 4 days treatment and not after the 1st day (Figure 4). However, some other repeated experiments showed clear effect of low ABA concentration (data not shown). This discrepancy during the first 24 h may be due to changes of new batches of chemicals during this study and some other unknown factors. When seedlings were grown for 4 days in the presence of either of the auxin efflux inhibitors, the stimulatory effect of 0.1 μM ABA was no longer observed (Figure 4B). However, in the presence of CHPAA this concentration of ABA still had a large positive effect (28% stimulation over CHPAA alone, compared to 34% in the control). Thus, it can be concluded that auxin efflux is necessary for the stimulatory effect of low ABA concentrations but that there is no evidence of a role for auxin influx.

Looking at the inhibitory effect of a high ABA concentration, this was surprisingly accentuated in the presence of either of the auxin efflux inhibitors, leading to an 86–89% inhibition of root elongation after 4 days, compared to 48% inhibition with 10 μM ABA alone (Figure 4B). By contrast, the auxin influx inhibitor CHPAA had the effect of reducing the inhibitory effect of 10 μM ABA to 6 and 23% of CHPAA alone after 1 and 4 days of treatment respectively (Figure 4). These results indicated that auxin influx is important for the root response to high ABA concentrations and that auxin efflux may play a negative role in the mechanism by which high ABA concentrations inhibit root elongation.

A genetic approach was used to investigate the respective roles of auxin efflux and influx in the root responses to ABA. The allelic auxin influx mutants aux1-7 and aux1-T and five auxin efflux mutants (pin2/eir1-1, pin3-4, pin3-5, pin4-3, and pin7-2) were treated with a range of concentrations of ABA, and their root elongation rates were compared with that of wild-type after the 1st day and over a 4-day period. The results of three separate experiments are shown in Figure 5. Two-way ANOVA was performed for each of those three experiments to test the impact of genotype, ABA treatment and their interaction. In all experiments, irrespective of whether measurements were made in the first 24 h after treatment or 4 days after treatment, there were significant effects of genotype and ABA treatment (P < 0.05, Supplementary Table 3). In the first experiment (wild-type, pin2/eir1-1, aux1-T, and iaa7/axr2-1), the results showed that there was significant genotype × ABA treatment interaction effect on the primary root elongation rate in the first 24 h and 4 days after treatment (P < 0.0001, Supplementary Table 3). In the second experiment (wild-type, pin4-3, pin7-2, and tir1-1), the interaction between genotype and ABA treatment significantly affected the average primary root elongation rate after 4 days treatment (P < 0.0001), but not in the first 24 h after treatment (P = 0.42, Supplementary Table 3). In contrast, the results of the third experiment (wild-type, aux1-7, pin3-4, and pin 3-5) suggested that the interaction between genotype and ABA treatment significantly affected the primary root elongation rate in the first 24 h of ABA treatment (P < 0.0001), but not the average primary root elongation rate during the 4 days treatment (P = 0.11, Supplementary Table 3). Overall, the statistical analysis confirmed that different genotypes responded differently to ABA treatment.

In the first one of these experiments (Figures 5A,B), it was found that the aux1-T knockout mutant was insensitive to both low and high concentrations of ABA in the 1st day and to the higher concentration of ABA when measured over 4 days, but that a slight positive effect of low concentration of ABA could be detected after 4 days. However, the aux1-7 missense mutant showed a weaker phenotype, being unaffected in its sensitivity to low [ABA] over either 1 or 4 days (Figures 5E,F) and insensitive to high [ABA] during the first 24 h of treatment (Figure 5E) but not during the subsequent 3 days (Figure 5F). These results are consistent with a role for AUX1-mediated auxin influx in the inhibitory effect of high [ABA], confirming the results obtained with CHPAA (Figure 4). An additional role of AUX1 in the stimulatory effect of low [ABA] cannot be ruled out but was only detectable in the early stages of treatment and only in the knockout mutant.

Of the five auxin efflux mutants tested, only pin2/eir1-1 behaved differently to the wild-type, showing less sensitivity to low [ABA], but normal sensitivity to high [ABA] (Figures 5A,B). There was one exception that in one of the three repetitions of this experiment, low [ABA] (0.1 μM) showed similar and weak stimulatory effect in root elongation in pin2/eir1-1 as in wild-type in the 4 days treatment (by 15% vs. 17% in wild-type, data not shown). However, pin3-4, pin3-5, pin4-3, pin7-2 all showed similar ABA responses to the wild-type (Figures 5C–F). These results are consistent with the evidence from the auxin efflux inhibitors (NPA and TIBA) that blocking auxin efflux did not alleviate the inhibitory effect of high [ABA] (Figure 4). It also suggested that the role for auxin efflux in the stimulatory effect of low [ABA] indicated by use of these inhibitors might involve PIN2/EIR1.

Two auxin insensitive mutants (tir1-1 and iaa7/axr2-1) were used to investigate the role of auxin signalling in the root responses to ABA. While the iaa7/axr2-1 mutant showed reduced sensitivity to both low and high [ABA] (Figures 5A,B), the tir1-1 mutant did not respond differently from the wild-type (Figures 5C,D).

Previous studies identified the importance of ethylene signalling and ethylene biosynthesis for the inhibition of primary root growth by ABA (Ghassemian et al., 2000; Luo et al., 2014). In the present study the objective was to try to understand how different concentrations of exogenous ABA can have opposing effects on root growth and to compare the signalling mechanisms responsible for the positive and negative responses. Use of the ethylene perception inhibitor STS (Figures 2C,D) along with three ethylene-insensitive mutants (etr1-1, ein2-1, and ein3-1) (Figure 3) confirmed that ethylene signalling was important for the inhibitory effect of high [ABA] under our experimental conditions.

In this study, the ability of the ethylene biosynthesis inhibitor AVG to completely suppress the inhibitory effect of high [ABA] (Figures 2A,B) is consistent with recent evidence that ABA inhibits root growth in Arabidopsis by promoting ethylene biosynthesis (Luo et al., 2014). The discrepancy between the present results and an earlier finding (Ghassemian et al., 2000) that AVG did not overcome the inhibitory effect of high [ABA], and even increased the degree of inhibition, could be attributable to the earlier authors’ use of higher concentrations of AVG than those used here. These higher concentrations may themselves have been inhibitory to root growth through AVG’s reported effects on auxin biosynthesis (Soeno et al., 2010). A recent report by Rowe et al. (2016) suggested that severe osmotic stress (-1.2 MPa) inhibited Arabidopsis root growth independently of either ABA or ethylene. However, in their experimental system, they did not observe the promotion of root growth by either low exogenous ABA or moderate osmotic stress (-0.37 MPa) that have been reported here and in other systems (Creelman et al., 1990; Xu et al., 2013; Rowe et al., 2016). Some other studies have also suggested that ABA and ethylene can antagonise each other and that ethylene can prevent ABA accumulation or modulate cellular sensitivity to ABA under stress (Wilkinson and Davies, 2010; Wilkinson et al., 2012; Chen et al., 2013). Therefore, the interaction between ABA and ethylene in stressed plants can be complex.

In contrast to the ethylene-dependence of the inhibitory effect of high [ABA], there was no evidence of any involvement of ethylene biosynthesis or signalling in the stimulatory effect of low [ABA]: neither AVG nor STS blocked the stimulatory effect (Figure 2) and the ethylene-insensitive mutants still responded positively to low [ABA] (Figure 3). These results indicate that the opposing effects of high and low ABA concentrations on root elongation operate through different signalling pathways.

Among the three ethylene-insensitive mutants tested here, the order of the effect on the root response to high [ABA] is ein2-1 > etr1-1 > ein3-1. These results were consistent with the report that ein2-1 has stronger impairment in ethylene responsiveness than etr1-1 does (van Loon et al., 2006). In addition, the mutant alleles of ein3 were previously reported to be less insensitive to ethylene than the strong alleles of etr1 and ein2 (Roman et al., 1995; Chao et al., 1997). Six members of the EIN3 family have been identified, with EIL1 being one the most closely related to EIN3 (Alonso et al., 2003b). A completely ethylene-insensitive phenotype has been reported in an ein3 eil1 double mutant, while the ein3 and eil1 single mutants have incomplete ethylene insensitivity (Alonso et al., 2003a,b). This indicates that there is partial redundancy of function between EIN3 and EIL1, which would provide an explanation for the weaker phenotype of the ein3-1 mutant in our experiments.

A role for auxin in the inhibitory effect of high [ABA] on Arabidopsis root growth has already been established from a number of studies using mutants defective in auxin transport and signalling (Belin et al., 2009; Wang et al., 2011; Thole et al., 2014; Zhao et al., 2015). Additionally, the distribution of GFP in the root tip of the DR5::GFP auxin reporter line was altered by adding either low or high [ABA] compared with the control (Supplementary Figure 2), which is consistent with the idea that auxin is involved in regulating root growth responses to both low and high [ABA]. However, similar distribution of GFP was observed for both low and high [ABA] treatments. These results suggest that auxin may not involve in root growth responses to low and high [ABA] through changing auxin maximum pattern in root tip but other means, e.g., plant sensitivity to auxin. In addition, the GFP expression patterns observed in this study were 3 days after the start of treatment, so we cannot rule out the possibility that there were short-term differences in the effects of the high and low [ABA] treatments on DR5::GFP expression that were missed in these experiments.

To look further into the involvement of auxin signalling in the root responses to both low and high [ABA], two mutants defective in components of the auxin signalling were used. The finding that iaa7/axr2-1 had reduced sensitivity to both the inhibitory effect of 10 μM ABA and the stimulatory effect of 0.1 μM ABA (Figures 5A,B) indicates that both the low and high [ABA] responses of root growth depend on auxin signalling. It has previously been shown that ABA represses the expression of the IAA7/AXR2 gene independent of auxin, leading to the suggestion that IAA7/AXR2 is at the nexus of crosstalk between ABA and auxin signalling pathways by acting as a negative regulator of both pathways (Belin et al., 2009). On the other hand, the gain-of-function mutant iaa7/axr2-1 has an amino acid substitution in the conserved domain II of IAA7/AXR2, which results in the accumulation of this AUX/IAA family protein and leads to repression of a series of downstream auxin responses (Nagpal et al., 2000; Gray et al., 2001). Thus, it may be that downstream components of the auxin signalling pathway, rather than IAA7/AXR2 per se, are required for both low and high [ABA] effects on root growth. The lack of a similar phenotype in another auxin signalling mutant tir1-1 (Figures 5C,D) is consistent with an earlier report that the tir1-1 mutant showed normal repression of embryonic axis elongation in response to ABA (Belin et al., 2009). This could indicate that either other F-box proteins are involved or it could be explained by genetic redundancy amongst members of this small family of auxin receptors (Dharmasiri et al., 2005; Parry et al., 2009).

Two previous studies reported that aux1 auxin influx mutants were less sensitive to high concentrations of ABA than wild-type (Belin et al., 2009; Thole et al., 2014) and a pin2 auxin efflux mutant was also found to be insensitive to ABA-dependent repression of both hypocotyl and radicle elongation (Belin et al., 2009). The insensitive phenotype with respect to high [ABA] in aux1 was confirmed in the present study (Figures 5A,B,E,F) and the role for auxin influx in ABA’s inhibitory effect on root growth was further supported by the ability of the auxin influx inhibitor CHPAA to antagonise this response to high [ABA] (Figure 4). Previously, Rowe et al. (2016) found that root growth in the aux1-7 mutant responded similarly to the wild-type under moderate and severe osmotic stresses. However, they collected root length data only after 4 days of osmotic stress (between 5 and 9 days after germination), which would not necessarily reveal the very early ABA response of aux1-7 as shown in Figures 5E,F.

How aux1 mutations affect the stimulatory effect of low [ABA] in our study was less clear-cut: an absence of stimulation of root growth by low [ABA] was only observed in the aux1-T knockout mutant during the first 24 h of treatment (Figures 5A,B). The wild-type phenotype was seen in the aux1-7 missense mutant with low [ABA] treatments (Figures 5E,F). Nevertheless, CHPAA failed to block the stimulatory effect of low [ABA] (Figure 4), indicating that there are differences between the positive and negative responses to ABA in their requirement for auxin influx.

When the positive and negative responses to ABA were compared for their requirement for auxin efflux, a distinct difference was found. No evidence of a positive role for auxin efflux in the inhibitory effect of ABA was obtained, based on the phenotypes of the pin2/eir1-1, pin3-4, pin3-5, pin4-3, and pin7-2 mutants (Figures 5A-F) and the inability of two auxin efflux inhibitors (NPA and TIBA) to overcome the inhibitory effect (Figure 4). On the other hand, the enhanced degree of inhibition by 10 μM ABA that was seen in the presence of either NPA or TIBA in the latter experiment suggests that auxin efflux may play a role in counteracting the inhibitory effect of high [ABA]. PIN1 is an important auxin efflux carrier that regulates polar auxin transport and supports auxin transport from shoot to root (Gälweiler et al., 1998; Růžička et al., 2007). Reduced PIN1 expression can normally cause auxin deficiency in roots, which is accompanied by growth inhibition (Blilou et al., 2005; Fernández-Marcos et al., 2011). Application of high [ABA] (10 μM) to well-watered plants reduced PIN1 expression in the root tip (Rowe et al., 2016). Thus, the application of an auxin efflux inhibitor (NPA or TIBA) in combination with the high [ABA] treatment under well-watered conditions is likely to result in even lower auxin levels in root tips and much stronger inhibition of root growth. By contrast, both NPA and TIBA were successful in blocking the stimulatory effect of low [ABA] and the pin2/eir1-1 mutant (but not the other tested pin mutants) was also defective in its response to low [ABA] (Figures 4, 5A,B). This evidence of the importance of auxin efflux in the response to low [ABA] agrees with a previous report that TIBA was able to partially suppress the positive effect of a low concentration of ABA on root growth in rice (Zhao et al., 2015).

Of the four PIN genes of which the role in ABA responses was tested (PIN2, PIN3, PIN4, and PIN7), it is notable that PIN2 is the only one that is expressed in the lateral root cap (Blilou et al., 2005; Kleine-Vehn et al., 2010; Band et al., 2014). The pin2/eir1-1 mutant also shows an altered pattern of distribution of the auxin maximum in the root tip compared to other pin mutants (Ottenschläger et al., 2003; Blilou et al., 2005). It is possible that the reduced sensitivity to low [ABA] that is seen in the pin2/eir1-1 mutant might be related to specific alterations in auxin distribution that arise from loss of PIN2/EIR1’s contribution to auxin efflux in the lateral root cap. On the other hand, auxin influx and efflux carriers act redundantly and a single mutation may have limited effect to ABA treatments. For example, PIN3 in the columella cells is proved to be important for gravity-sensing, but pin3 mutant showed marginal defect in gravity response, which is related to the redundancy of PIN3, PIN4, and PIN7 (Blilou et al., 2005; Kleine-Vehn et al., 2010).

The results in this study provide evidence that the stimulatory effect of low ABA concentrations on root growth operates through an ethylene-independent pathway, and requires auxin signalling and auxin transport by the PIN2/EIR1 auxin efflux carrier (Figure 6). However, the inhibitory effect seen at high ABA concentrations is through an ethylene-dependent pathway that requires auxin signalling and auxin influx through AUX1 (Figure 6). Růžička et al. (2007) found ethylene inhibits root elongation by stimulating auxin biosynthesis and auxin transport toward the root elongation zone. Ethylene failed to activate auxin response and inhibit root growth in mutants that are defective in auxin perception or basipetal auxin transport (i.e., tir1-1, aux1-T) (Růžička et al., 2007). In future it will be important to test the hypothesis that the same distinct pathways also account for the analogous biphasic response of root growth to different degrees of water deficit (Sharp and Davies, 1979; van der Weele et al., 2000; Xu et al., 2013). The resulting insight into how root growth responds to different degrees of soil drying could have agronomic significance in helping to manage crop development and to develop crop varieties whose root responses to soil drying are more beneficial to crop productivity in a changing climate and an increasingly water-scarce world.