All chemicals were sourced from Sigma Aldrich. Capped complementary RNA (cRNA) production, X. laevis oocytes preparation, and whole oocyte two-electrode voltage clamping (TEVC) recording were performed as our previous work (Heng et al., 2018; Long et al., 2020). 46 nL of 200 mM Na₂-malate (pH 7.2) was pre-loaded before TEVC recording. The bath solution for malate current recording consisted of 80 mM Na-gluconate, 1 mM Ca-gluconate₂, 1 mM K-gluconate, 1 mM Mg-gluconate₂, 25 mM malic acid, 0.1 mM LaCl₃, and 10 mM MES/Tris (pH 5.8). And the bath solution for different anion permeability contained 25 mM NaNO₃, 25 mM Na₂-malate, 25 mM NaCl, and 25 mM Na₂SO₄ respectively. The method for creating inside-out patch clamp in X. laevis oocytes followed our previous protocols (Long et al., 2020) with the solution modified, 20 mM Na₂-malate in the bath solution (cytosolic side) and 10 mM Na₂-malate in the pipette solution (outside) and the pH of both was adjusted to 7.2 (Hepes/Tris).

The coding regions of OsALMT7, paab1-t1, paab1-t2, and TaALMT1 were cloned into pSPYCE and pSPYNE vectors (Waadt et al., 2008; Waadt and Kudla, 2008). The BiFC assays were performed as described previously (Walter et al., 2004; Waadt and Kudla, 2008). The tested construct pairs were expressed in leaves of Nicotiana benthamiana for 3 d before microscopy observation. The YFP fluorescence in the transformed leaves was imaged using a confocal laser scanning microscope (ZEISS LSM710).

Heng et al. (2018) found two transcripts in a panicle apical abortion (paab1) mutant rice, paab1-t1 and paab1-t2. Figure 1A depicts the full-length OsALMT7 protein and the paab-t1 and paab1-t2 variants in which the last 2-3 transmembrane α-helices (depending on topology predictions) and the cytosolic C-terminus are absent. Both paab1 truncations encode malate-transport competent proteins but with a much-reduced capacity compared to full-length OsALMT7 when expressed in X. laevis oocytes (Heng et al., 2018). Here, we carried out an in-depth analysis of the behaviour of paab1 proteins in X. laevis oocytes. Our results confirmed that the paab1-t1 and paab1-t2 proteins conducted significantly greater inward currents than water-injected oocytes at the polarization range of membrane potential (more negative than -100 mV), but these currents were lower in magnitude than those of wild-type OsALMT7 (malate efflux; Figure 1B ; Supporting Figure 1 ). Channel conductance (G) analysis of the paab1 mutant truncated proteins found that the maximal conductance was lower, and voltage-dependent activation occurred at a more negative potentials than that of the wild-type channel (OsALMT7). The G_max values of OsALMT7, paab1-t1, and paab1-t2 were 112.2 ± 13.1 μS, 20.7 ± 5.8 μS, and 16.7 ± 5.2 μS, and the V_1/2 values of OsALMT7, paab1-t1, and paab1-t2 were -118.4 ± 11.9 mV, -136.9 ± 16.0 mV, and -138.1 ± 18.0 mV, respectively ( Figure 1C ). Isolated inside-out membrane patches (cytosolic side facing the bath solution) from X. laevis oocytes injected with OsALMT7, paab1-t1, or paab1-t2 cRNA showed, compared to water-injected controls, weak activated currents with short open times reminiscent of TaALMT1 (≈1 pA; Long et al., 2020). paab1 mutants showed both a smaller magnitude of activated currents and less open probability (P_open), with single channel currents at -160 mV of 1.03 ± 0.03 pA, 0.42 ± 0.05, and 0.40 ± 0.04 pA and P_open of 16.9 ± 3.8%, 9.3 ± 1.3%, and 10.5 ± 1.2% for OsALMT7, paab1-t1, and paab1-t2, respectively ( Figures 1D, E ).

Heng et al., 2018 showed that transgenic plants expressing both OsALMT7 and paab1 had a similar phenotype to paab1-1 plants; we propose that this may be the result of paab1 mutant channels inhibiting OsALMT7 channel function. To investigate the impact of paab1 on OsALMT7 channel activity, we coexpressed OsALMT7 and either of the two paab1 transcripts. The TEVC recording showed that co-expression of either of the paab1-t1 and paab1-t2 proteins inhibited the channel activity of OsALMT7. Current magnitudes mediated by OsALMT7 and paab1 co-injection with identical amounts of cRNA fell between those obtained with sole injection of OsALMT7 or paab1 alone ( Figures 2A, B ). Interestingly, we found that OsALMT7 and paab1 mutant channels showed different time dependence, with OsALMT7 exhibiting instantaneous currents while paab1 currents had time-dependent activation at negative membrane potentials ( Figure 2A ). What about the current curve of paab1 and OsALMT7 coexpressing situation? As shown in Figures 2A, C , OsALMT7 and paab1 co-injection resulted in currents possessing both instantaneous and time-dependent components, taking on hybrid characteristics of both parent channels.

To confirm that paab1 inhibited OsALMT7, we injected different proportions of cRNA, increasing the cRNA amount of the paab1 mutant to the same amount of OsALMT7 cRNA. As paab1 cRNA was increased, the inhibition of OsALMT7 increased ( Figures 2D, E ). These TEVC results in X. laevis oocytes are consistent with the inhibition of OsALMT7 by paab1 mutant channels.

In Heng et al, we reported that OsALMT7 had a high permeability to NO₃ ^- and malate, and a low permeability to Cl^- and SO₄ ^2-. In this study, we examined the anion permeability following paab1 and OsALMT7 co-injection by substitution of the anions in the bath solution. TEVC showed that paab1 mutant channels had no effect on anion permeability to OsALMT7; the currents following paab1-OsALMT7 co-injection shared the same anion selectivity as OsALMT7 ( Supplemental Figure 1 ). We also investigated the effect of pH on OsALMT7, paab1-t1, paab1-t2, paab1-t1-OsALMT7, and paab1-t2-OsALMT7 channels, with the external bath pH setting to 7.2, 5.8, and 4.2, representing alkali, neutral, and acidic soil conditions. The TEVC recording showed that all these channels shared no dependence on external pH ( Supplemental Figure 2 ), unlike wheat ALMT1 (Delhaize et al., 2004; Sasaki et al., 2004). In summary, paab1 did not affect the anion selectivity or pH dependence of OsALMT7.

We hypothesized that paab1 mutant channels inhibited OsALMT7 by combining into heteromers. To test our hypothesis, we first examined the physical interactions between OsALMT7 itself, OsALMT7 and paab1 mutant channels and the two paab1 mutant channels. BiFC experimental results were consistent with that OsALMT7 interacting with itself, paab1-t1, and paab1-t2 proteins in tobacco leaves and the two paab1 channels interacting with each other as well ( Figure 3 ). TaALMT1 was used as a PM localized control and did not interact with OsALMT7, but interacted with itself ( Figure 3 ).

Lebaudy et al. provided strong evidence that KAT1 and KAT2 formed heteromers by constructing KAT1-KAT1, KAT1-KAT2, KAT2-KAT2, and KAT2-KAT1 tandems (translationally fused proteins) and expressing them in X. laevis oocytes. To confirm that paab1 mutant channels and OsALMT7 combined and formed a heteromer, tandems of OsALMT7-OsALMT7, OsALMT7-paab1-t1, paab1-t1-paab1-t1, and paab1-t1-OsALMT7 were constructed and assayed for channel activity in X. laevis oocytes ( Figure 4A ). As shown in Figure 4 , all the constructs were functional (had currents in excess of those of the water-injected controls); the OsALMT7-OsALMT7 tandem construct mediated the strongest malate currents, and the paab1-t1-paab1-t1 tandem construct mediated the weakest malate current. For the combination constructs, the tandem with OsALMT7 fused at the N-terminus had the weaker channel activity, while the tandem with paab1-t1 in the N-terminus showed stronger channel activity with an intermediate value between the OsALMT7-OsALMT7 and paab1-t1-paab-t1 constructs ( Figure 4 ). Furthermore, paab1-OsALMT7 showed stronger current in magnitude, while OsALMT7-paab1 showed similar current shape in time dependent with OsALMT7-OsALMT7. We propose that these performances were caused with the artificial multimer with the way of tandem constructing. The TEVC recordings of these tandem constructs expressed in X. laevis oocytes provided further evidence that paab1 mutant channels and OsALMT7 assemble to form homo- or heteromultimers.

The paab1 mutant terminates transcription in the middle of the 5^th transmembrane α-helices, causing a lack of the last 2 transmembrane α-helices and C-terminal cytosolic domains ( Figure 1A , Heng et al., 2018). Truncated mutants of OsALMT7 with different numbers of transmembrane α-helices were constructed to examine their contribution to channel activity in the context of the whole protein ( Figure 5A ). These truncations were named OsALMT7-M1 to OsALMT7-M6 and contained 2 to 7 transmembrane α-helices respectively. Surprisingly, we found that OsALMT7-M2 with just 3 transmembrane α-helices mediated malate efflux, while OsALMT7-M6 with all 7 helices showed no channel activity ( Figures 5B, C and Supplemental Figure 3 ).

To investigate other truncated mutant could form heteromer with OsALMT7, we detected the inhibitory effects of OsALMT7-M1 and OsALMT7-M6 to OsALMT7 channel activity. We found that OsALMT7-M1 with 2 helices did not inhibit OsALMT7, while OsALMT7-M6 did ( Figures 6A, B ). These data confirmed that OsALMT7 functions as a multimer and found that at least 3 helices were necessary for the multimer formation.

Furthermore, we truncated TaALMT1 to TaALMT1-M1 (1-148 amino acid residues with 3 and half transmembrane α-helices), which corresponds to paab1-t1 in rice, TaALMT1-M2 with 4 transmembrane α-helices, and TaALMT1-M3 with all 6 predicted transmembrane α-helices (corresponding to TaALMT1^Δ219-459 in Ligaba et al., 2013) ( Supplemental Figure 4A ). TEVC recordings showed that, in contrast to the wild-type TaALMT1, all the truncations showed no channel activity at pH 4.5 with Al³⁺ similar to the result in Ligaba et al., 2013. However, at pH 7.2, these truncations showed malate permeability, and the current increased with the number of transmembrane α-helices increasing ( Supplemental Figures 4B, C ).

To investigate the broader implications of our observations in other ALMTs, we examined the effect of combinations between TaALMT1 and its truncated mutants. When exposed to either pH 4.5 with Al³⁺ or pH 7.2 in the bath, the truncated mutants TaALMT1-M1 and TaALMT-M3 inhibited the channel activity of TaALMT1, indicating that TaALMT1 functions as a homomultimer like OsALMT7 ( Supplemental Figures 4D, E ).

According to the studies in OsALMT7 and TaALMT1, we summarized that mediating anion flux with lacking of transmembrane helices was special for OsALMT7, and the dominant deactive proformance of its truncations to wild-type channels was common to the ALMT family as they function as dimer.

Ligaba et al. (2013) removed the C-terminal domain of TaALMT1 to various degrees and retained the full complement of transmembrane domains (TaALMT1^Δ219-459) and found amino acid residues in C- terminal important for Al³⁺ sensitivity and the retention of malate transport ability, they also truncated transmembrane domains and found no activity of the protein. Zhang et al. (2013) found that key sites in the transmembrane helices of AtALMT9 affected channel activity. Recently, Li et al. (2020) found that truncating apple ALMT9 at the C-terminus affected channel activity. Our work is the first to show that truncating the transmembrane domains of ALMTs can still result in a transport competent protein and that such truncation results in an in planta phenotype (Heng et al., 2018).

We truncated different transmembrane helices of OsALMT7 and investigated the activity of them. However, only OsALMT7-M2, with 3 transmembrane helices, and paab1 showed channel activity, suggesting that the first 3 transmembrane helices are important for the pore formation on PM for malate permeability, and OsALMT7-M2 and paab1 could form a pore but other truncations with more transmembrane helices could not. Previous studies showed that the last 2 transmembrane helices were essential for pore formation and channel activity (Ligaba et al., 2013; Zhang et al., 2013), yet we found that OsALMT7-M2 and paab1 were functional for transport. These results are novel and will inform ALMT structural studies.

For TaALMT1, we got none channel activity for all the truncations at both pH 4.5 with Al³⁺ or pH 7.2. However, Ligaba et al. (2013) found that TaALMT1-M3 (TaALMT1^Δ219-459) showed channel activity but lost Al sensitivity at pH 4.5, while our study failed to obtain TaALMT1-M3 channel activity for at least 5 times TEVC experiments. We think that this was caused by the different conditions of the X. laevis oocytes in the two labs.

In this study, we have shown that the PM-localized channels OsALMT7 and TaALMT1 function as multimers by TEVC recording in X. laevis oocytes and physical interaction analysis in tobacco leaves. Although structural biology evidence is lacking, in light of the cases of OsALMT7, we propose that it functions as multimers. Zhang et al., 2013 proposed that the vacuolar ALMT channel AtALMT9 functions as a multimer. Similar to our study, they coexpressed point mutations and wild-type AtALMT9 channels in tobacco mesophyll protoplasts, detecting the inhibition of channel activity, and they further showed the multimer formation by immunoblot analysis. Recently, Wang et al., 2022 and Qin et al., 2022 reported that ALMT1 and ALMT12 functions as a dimer. Our study started from the clew in Heng et al. (2018). when wild-type plants were transformed with a genomic fragment containing the paab1 base substitution causing both OsALMT7 and paab1 expressing and resulting in a panicle abortion phenotype. That dominant-negative phenotype implied that paab1 might inhibit OsALMT7 channel activity and the following experiment approved that. Although we cannot predict the OsALMT7 forming a dimer or a trimer, it is credible that the geometric symmetry structure formed with monomer unit is necessary for the ALMT anion channels. Furthermore, we propose that introduction of mutant ALMT channels to wildtype plants would be a method to alter ALMT function as a tool to manipulate plant phenotype.

In the BiFC experiment, we detected interaction fluorescence only when YFP truncations were fused at the N-terminus of OsALMT7 or paab1 ( Figure 3 ) but no fluorescence when YFP truncations were fused at the C-terminus (data not shown). One reason is that unlike TaALMT1, OsALMT7 was predicted to have 7 transmembrane helices, and the C-termini of OsALMT7 and paab1 might face different side of the PM respectively. Another is that Mumm et al. (2013) showed that fusing a YFP at the N-terminus had no effect on channel characteristics and PM localization for ALMT12, while the C-terminus fused YPF affected function and PM localization of ALMT12. So, we proposed that YFP truncations fusing in the C-terminus might inhibit the interaction between OsALMT7 and paab1. These might be the reason that the tandem construction causes the malate conductance reducing comparing to WT OsALMT7 channel ( Figure 4 ). Furthermore, for the case of the tandems of paab1-t1-OsALMT7 and OsALMT7-paab1-t1, which protein was designed at the N-terminal did affect the malate permeability of the tandem. According to the TEVC data, we proposed that the paab1-t1-OsALMT7 provided more complete pore than OsALMT7-paab1-t1. Moreover, the tandem with paab1 in the N-terminus had stronger channel activity than in C-terminus ( Figure 4 ) suggesting that paab1 mutants and OsALMT7 channel need a especial combination to mediate malate transporting.