Arabidopsis thaliana seeds of Col-0 background and an npf2.5 (SM_3.31001) were obtained from the European Arabidopsis Stock Centre (Nottingham, UK). The npf2.5 knockout mutant contains a transposable dspm-element insertion in the 4th exon of the NPF2.5 ORF. All plants were grown in a growth room with a photoperiod of 10/14 h (light/dark), the photon irradiance at the level of the plant leaves was 120 μmol m⁻² s⁻¹. Growth room temperature was maintained between 21 and 23°C, and humidity was kept between 60 and 75%.

Hydroponically grown plants: Arabidopsis seeds were germinated and grown in growth rooms using the conditions described above and methods described in Conn et al. (2013). Salt stress was applied to the growth solution 4–5 weeks after germination.

MS-plate grown plants: Arabidopsis seeds were surface sterilized using 70% (v/v) ethanol solution and commercial bleach (Unilever Australasia, Australia) for 10 min each. The bleach residue was removed by rinsing the seeds with dH₂O at least 5 times. Seeds were placed on ½ MS plates and sealed. Plates were kept in growth rooms described above.

Soil grown plants: the material and protocol used to grow Arabidopsis in soil was described in Møller et al. (2009). Soil grown plants were treated with 150 mM NaCl solution (500 ml per time, 3 times per week). This was done by applying NaCl solution to the bottom of each planting pot (one plant per pot).

Transcript abundance analysis by qRT-PCR was performed following the method described in Burton et al. (2008) and Jha et al. (2010). The sequences of primers for determining NPF2.5 expression were 5′ TCCTGCTCTTGTGATGGTTG 3′ (Forward) and 5′ CACGAGGTGGTTCTTTGGAT 3′ (Reverse). To test the primer specificity, melt curve analysis was performed and the PCR products were sequenced. Choice of control genes and data normalization followed the protocols described in Burton et al. (2008) and Jha et al. (2010). Four control genes were selected: Cyclophilin (AT2G36130), Tubulin alpha-2 chain (TUA2, AT1G50010), Actin 2 (ACT2, AT3G18780) and Glyceraldehyde 3-phosphate dehydrogenase A (GAPDH, AT3G26650). Four independent biological replicates were used. Two rounds of normalization were performed as shown in Burton et al. (2008) and Jha et al. (2010): (1) in respect to the biological replicates in the same treatments (or of the same transgenic line) using control genes, and (2) in respect to different treatments (or transgenic lines) using control genes. Further normalization relative to control group was performed for Figure 2B.

To express NPF2.5 in X. laevis oocytes, a pGEMHE-DEST expression vector (Li et al., 2016) was used. cRNA was prepared using a mMESSAGE mMACHINE kit (Ambion, CA) following the manufacturer's manual. Injection of cRNA into oocytes was performed using a protocol described in Munns et al. (2012). Two electrode voltage clamping was performed following the protocols in Roy et al. (2008). Whole oocyte currents were recorded in ND50 solution (50 mM NaCl, 2 mM K-gluconate, 1.8 mM Ca-gluconate, 1 mM Mg-gluconate and 5 mM HEPES, pH at 7.5, osmolarity 240–260 mOsmol/kg). Data was obtained using pClamp 8.2 software (Axon Instruments).

MIFE set-up and electrode-manufacture were following the methods described in Shabala (2006). MIFE measurement of oocytes were made 2 days after injection. Each oocyte was transferred from Ringer's solution to antibiotic free ND96 solution (93.5 mM NaCl, 2 mM KCl, 1.8 mM CaCl₂, 2 mM MgCl₂ and 5 mM HEPES, pH at 7.5, osmolarity 240–260 mOsmol/kg) for 5 s and then transferred to ND48 solution (46.7 mM NaCl, 2 mM KCl, 1.8 mM CaCl₂, 2 mM MgCl₂ and 5 mM HEPES, pH at 7.5, osmolarity 240–260 mOsmol/kg) before measurements. Flux measurements were made at animal and vegetal poles for 15 min in both NPF2.5- and water-injected oocytes for 15 min following a 2.5 min period of electrode alignment and solution exchange. Net fluxes of 4 to 7 individual oocytes were averaged for each treatment.

MIFE measurements of npf2.5 plants were made using 7-day old seedlings grown on ½ MS media. Plants were transferred to 48 mM NaCl solution for 30 min before a 15-min long measurement. Net fluxes of 6 to 9 individual seedlings were averaged for each genotype.

Web MicroRNA Designer (http://wmd3.weigelworld.org/cgi-bin/webapp.cgi) (Schwab et al., 2006) was used to design the microRNA constructs to specifically knockdown NPF2.5 expression. A 21-bp target sequence was identified from the NPF2.5 sequence. A set of primers (Supplementary Table 1) was used to incorporate the 21-bp amiRNA sequence into the MIR319a vector (Schwab et al., 2006). The amiRNA construct then was cloned into pCR8 and transferred to pTOOL2 (Roy et al., 2013) by a LR reaction (Invitrogen) for constitutive over-expression. The construct was transformed into Arabidopsis Col-0 using an Agrobacterium-mediated floral drip method (Weigel and Glazebrook, 2002). Transgenic lines were selected by application of 120 mg/L herbicide glufosinate (Bayer Crop Science, Australia).

To clone NPF2.5, a cDNA library was prepared by performing a reverse transcription (Superscript III, Invitrogen CA) on total RNA extracted from Arabidopsis roots using TRIzol reagent (Invitrogen, CA). The coding sequence of NPF2.5 was amplified from cDNA with Platinum Taq (Invitrogen, CA) using a forward primer (5′ ATGGCTGATTCAAAATCTGGTG 3′) and a reverse primer (5′ AGGAAAACAACATGTTGTGTCC 3′). After Sanger sequencing to confirm the amplified DNA sequence, NPF2.5 was cloned into the Gateway® enabled pCR8 entry vector (Invitrogen, CA) and transferred from pCR8 to the pTOOL2 destination vector (Qiu et al., 2016) using LR Clonase II (Invitrogen, CA). The pTOOL2 expression vector was transformed into Arabidopsis Col-0 plants using an Agrobacterium-mediated floral dip method (Weigel and Glazebrook, 2002).

To determine the subcellular localisation of NPF2.5, the coding sequence of NPF2.5 was transferred to the destination vector pMDC44 (Curtis and Grossniklaus, 2003) using LR Clonase II (Invitrogen, CA). The vector pMDC44 containing 5′ GFP6 was used for generating a GFP::NPF2.5 construct that was transformed into Col-0 Arabidopsis plants using the floral dip method. Two independent transformations were performed to generate two independent lines. T₂ transgenic plants were geminated on ½ MS media under hygromycin selection (25 μg/ml) and grown for 2 weeks. GFP fluorescence was visualized using an LSM5 PASCAL laser-scanning confocal microscope (Carl Zeiss, Jena, Germany) running PASCAL imaging software (version 3.2 SP2, Carl Zeiss) with an excitation of 488 nm and an emission of 505–530 nm. Plasmolysis was conducted on a slide using 10% (w/v) sucrose solution.

To determine the subcellular localisation of NPF2.5 in Arabidopsis protoplasts, a fusion construct YFP(Yellow Fluorescent Protein)::NPF2.5 was generated using the coding sequence of NPF2.5 and a modified destination vector pattR-YFP (gateway enabled) (Subramanian et al., 2006). Protoplasts were isolated from Arabidopsis mesophyll cells and PEG-mediated transient transformation were conducted using the protocols described in Conn et al. (2013). Transformed protoplasts were examined using a LSM5 PASCAL laser-scanning microscope (Carl Zeiss, Jena, Germany) running PASCAL imaging software (version 3.2 SP2, Carl Zeiss). An excitation of 436 nm and an emission of 470–535 nm were used for cyan fluorescent protein (plasma membrane marker), while an excitation of 514 nm and an emission of 535–610 nm were used for YFP fusion protein.

The complete 1.65 kb intergenic region between NPF2.5 and NPF2.4 (the last gene upstream) was amplified from gDNA with Platinum Taq (Invitrogen, CA) using a forward primer (5′ GACACTAACGTGTTCTGTCCTCGTTTTC 3′) and a reverse primer (5′ GGTAGAGAACAAGATGAACCAGGAGGGCAA 3′). The PCR fragment was cloned into the Gateway® enabled entry vector pCR8. The proNPF2.5 was transferred into pMDC162 (Curtis and Grossniklaus, 2003) to drive uidA expression using LR Clonase II (Invitrogen, CA). The expression vector was transformed into Arabidopsis Col-0 plants using the floral dip method. Two-week old T₂ proNPF2.5:uidA plants were GUS-stained for 1 h (seedlings) or 2 h (flowers and true leaves) following the protocol described in Weigel and Glazebrook (2002) using 0.5 mg/ml X-gluc (5-bromo-4-chloro-3-indolyl glucuronide). To prepare cross-sections of roots of proNPF2.5:uidA plants, stained material was embedded, dehydrated and sectioned following a method described in Li et al. (2016). Images were taken of two independent transformation events using a Leica ASLMD compound microscope equipped with a DFC480 CCD camera (Leica microsystem, Wetzlar, Germany).

NPF2.5 was cloned into the yeast expression vector pYES-DEST52 using LR Clonase II (Invitrogen, CA) and transformed to a S. cerevisiae strain InvSc2 (Invitrogen, CA) using a LiAc/SS DNA/PEG method (Gietz and Schiestl, 2007). A growth inhibition assay was performed using solid SD media containing 2% (w/v) D-galactose, 2% (w/v) agar and NaBr/KBr at concentrations as indicated. Yeast cells grew to an OD₆₀₀ value of 3.0 (using a UV-160A spectrophotometer, Shimaduzu, Japan) were pelleted and resuspended in MQ water. Liquid cultures were serially diluted in sterile MQ water by four successive 10-fold dilutions. From each dilution, 10 μl was placed on SD media (no uracil). Results are presented for one of two independent transformation events performed, each with three technical replicates. Both experiments had consistent results. Plates were incubated at 28°C for 2–3 days and growth phenotypes were recorded.

Whole shoots were harvested, weighed, freeze-dried and ground into a powder. Approximately 10–20 mg of shoot material was digested in 2 ml of 1% nitric acid at 80°C overnight. A Chloride Analyser (Sherwood Scientific model 926, Cambridge, UK) was used to determine Cl⁻ content in prepared samples following the manufacturer's instructions. The NO3− assay described in Kamphake et al. (1967) and a micro plate-reader (BMG LABTECH, Germany) were used to determine NO3− content in prepared sample. A model 420 Flame Photometer (Sherwood Scientific Ltd, Cambridge, United Kingdom) was used to determine Na⁺ and K⁺ content in the samples.

In all hydroponic experiments, the plants were randomly distributed. Anion accumulation and transcript abundance in hydroponically grown plants were analyzed by one-way ANOVA and Tukey tests (n ≥ 4). Two-tailed t-tests were used to compare results from the MIFE and TEVC experiments, see figure legends for number of replicates. All data was plotted and analyzed in Graphpad Prism 6.

The Arabidopsis NAXT sub-family members are clustered on chromosome 3 and have high sequence similarity to each other (Segonzac et al., 2007; Tsay et al., 2007). Protein sequence alignment of NPF2.5 to NAXT family members with known functions (NPF2.4, NPF2.3 and NPF2.7/NAXT1) was performed. The NPF2.5 of Col-0 background possesses 83.2% identity (467/561) and 88.6% similarity (497/561) in protein sequence to the Cl⁻ transporter NPF2.4 (Col-0) (Figure 1A). NPF2.5 is predicted to have 560 amino acids, which are predicted to form 12 trans-membrane domains (Figure 1B)-a signature structure of NRT1/PTR proteins (Tsay et al., 2007). A central hydrophilic loop was predicted for NPF2.5, between the trans-membrane domain 6 and 7 (Figure 1B).

To determine NPF2.5 expression in the root and shoot, and its regulation by NaCl salt, quantitative RT-PCR was performed using cDNA prepared from each tissue of 4-week old hydroponically grown Arabidopsis (Col-0) with and without salt treatments. The transcript abundance of NPF2.5 was 50 times higher in the root than in the shoot when the plants were grown under the low salt condition (2 mM NaCl) (Figure 2A). There was a statistical significant increase in NPF2.5 transcript abundance in the root when the plants were grown in 50 mM or 100 mM NaCl for 72 h (Figure 2B). At 100 mM NaCl, NPF2.5 expression increased by 50% compared to that observed in plants grown in 2 mM NaCl (Figure 2B).

The localisation of NPF2.5 expression within tissues was determined using the promoter-GUS reporter system. The complete 1.65 kb intergenic region between NPF2.5 and NPF2.4 (the last gene upstream) was used to drive the expression of uidA in Col-0 Arabidopsis. GUS activity driven by proNPF2.5 in 2-week old transgenic plants was detected predominantly in the outer cells of the root after 1 h of incubation with X-Gluc (Figures 3B–D). To further determine the specific cell types of the root where NPF2.5 was expressed, root cross sections of proNPF2.5:uidA plants were prepared and revealed that NPF2.5 was expressed in the root cortex (Figure 3F). To determine if NPF2.5 was expressed in other tissues at a different developmental stage, flowering proNPF2.5:uidA plants (12-week old) were stained for 2 h. GUS activity was detected within stigma, sepals and trichomes of flowering plants (Supplementary Figure 1).

To determine the sub-cellular localisation of NPF2.5, GFP was fused to the 5′ end of the NPF2.5 coding sequence, and was stably transformed into Col-0 Arabidopsis. GFP::NPF2.5 plants (T₂) were grown on ½ Murashige & Skoog (MS) plates for 2 weeks. GFP was detected on the periphery of the root cells of GFP::NPF2.5 plants (Figures 4A,B). Plasmolysis was used to detach the plasma membrane from the cell wall and revealed GFP fluorescence on the Hechtian strands (Figures 4C–E).

In order to gain further evidence for the membrane localisation of NPF2.5 in planta, YFP was fused to the 5′ end of NPF2.5 coding sequence, and transiently co-transformed with a plasma membrane marker, eCFP::ROP11 (Molendijk et al., 2008), to isolated Arabidopsis mesophyll protoplasts. Confocal images showed the NPF2.5-associated YFP fluorescence overlapping with the cyan fluorescence of the plasma membrane marker (Figures 4F–I). The position of the tonoplast in the transformed protoplasts is indicated in the bright field image distinct from the plasma membrane (Figure 4H). Therefore, both transformation systems indicated that NPF2.5 resides on the plasma membrane.

To test if the NPF2.5 protein was permeable to anions, NPF2.5 was expressed in yeast (Saccharomyces cerevisiae). The transformants were grown on both low and high concentrations of Br⁻, a toxic analog of Cl⁻ (MacRobbie, 1975, 1982). Under low salt conditions (5 mM KBr or NaBr), the growth rate of yeast expressing NPF2.5 was similar to that of empty vector controls (Figures 5A,B). Under high salt condition (400 mM KBr or 300 mM NaBr), the growth of the NPF2.5-transformed yeast was inhibited when compared to the empty vector controls. The inhibition in growth observed was not specific to either K⁺ or Na⁺ (Figures 5C,D). Similar results were obtained across all technical replicates in yeast derived 2 independent transformation events.

To gain further insight into the transport properties of NPF2.5, NPF2.5 was expressed in X. laevis oocytes and two Electrode Voltage Clamp was performed. Significantly higher inward currents were elicited from oocytes pre-injected with NPF2.5 cRNA when compared with water-injected control oocytes when the membrane potential was clamped at -120 mV (similar to that of plant cells) (Figure 6A); full current-voltage relationships can be found in Supplementary Figure 2. The net Cl⁻ flux across the plasma membrane of X. laevis oocytes injected with either NPF2.5 cRNA or water were compared using non-invasive Microelectrode Ion Flux Estimation (MIFE) (Shabala, 2006). A greater Cl⁻ efflux was observed in NPF2.5 cRNA injected oocytes than water-injected controls when the [Cl⁻] was lowered from 96 to 48 mM in isosmotic solutions (Figure 6B).

To further analyse the putative molecular function of NPF2.5, a npf2.5 knockout line was obtained. Quantitative RT-PCR was used to confirm the elimination of NPF2.5 expression in the knockout line (Supplementary Figure 3). MIFE was performed on the npf2.5 knockout mutant pre-treated with 48 mM NaCl for 5 days. A significantly lower efflux of Cl⁻ from npf2.5 roots was observed when compared with Col-0 plant roots (Figure 6C).

The npf2.5 knockout line was also used to determine whether abolishing NPF2.5 expression could alter Cl⁻ accumulation in the shoot. When npf2.5 and Col-0 plants were grown in soil for 4 weeks and treated with either 75 mM or 100 mM NaCl solution for 5 days, shoot Cl⁻ concentration was found to be higher in npf2.5 plants compared with the Col-0 plants (Figure 6D, Supplementary Figure 4A). There was no difference in accumulation of NO3−, K⁺, or Na⁺ in the shoot of npf2.5 plants when compared to Col-0 plants (Supplementary Figures 4B–D).

To further test the effect of reduced expression of NPF2.5 on accumulation of Cl⁻ in the shoot, artificial microRNA (amiRNA) was used to knockdown the expression of NPF2.5 in the root of Col-0 Arabidopsis. A total of five independent amiRNA-NPF2.5 knockdown lines were generated (KD-NPF2.5-1, KD-NPF2.5-2, KD-NPF2.5-3, KD-NPF2.5-4, and KD-NPF2.5-5). Transgenic plants were grown hydroponically for 4 weeks and treated with 75 mM NaCl for 5 days. Following salt treatment, the transcriptional levels of NPF2.5 in the knockdowns were significantly lower than in the null segregants (Figure 7A). Knockdown lines tended to accumulate more Cl⁻ in the shoot compared with the null segregants (P = 0.014 for KD-NPF2.5-5) (Figure 7B). Shoot accumulation of NO3−, K⁺ and Na⁺ in the knockdown plants was similar to null segregants (Figures 7C–E). To test if the expression of NPF2.4, a stelar Cl⁻ transporter with high homology to NPF2.5, was affected in the amiRNA NPF2.5 knockdowns, quantitative RT-PCR was performed on the roots of amiRNA-NPF2.5 knockdown lines grown in low salt conditions (2 mM NaCl). NPF2.4 expression of the amiRNA mutants was at the same level as the null segregants (Supplementary Figure 5).

To determine if constitutive over-expression of NPF2.5 had an effect on Cl⁻ accumulation, NPF2.5 was over-expressed in Col-0 Arabidopsis under the control of the 35S promoter. Two independent T₃ NPF2.5 over-expression lines (OEX-NPF2.5-1 and OEX-NPF2.5-2) were grown hydroponically for 4 weeks and treated with 2 mM (low salt) or 75 mM (high salt) NaCl for 5 days. Quantitative RT-PCR revealed that both lines had significantly higher abundance of NPF2.5 transcript when compared with null segregants, with OEX-NPF2.5-1 plants being approximately 40-fold higher (Figure 8A). Following a low salt treatment, Cl⁻ concentration was higher in the shoot of both over-expression lines when compared with null segregants, with OEX-NPF2.5-1 being statistically significant (Figure 8B). Shoot NO3− concentration in both over-expression lines was shown to be at the similar levels as the null segregants (Figure 8C). After a high salt treatment, shoot accumulation of Cl⁻ and NO3− increased and decreased respectively in all plants tested (Figures 8B–E). However, under these conditions, over-expression lines had similar anion concentrations when compared with null segregants (Figures 8D,E).

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