ClCd and ClCf act redundantly at the trans-Golgi network/early endosome and prevent acidification of the Golgi stack Free

Transport through the endomembrane system involves a multitude of interactions between cargo and components of the trafficking machinery. When and where in the endomembrane system reactions including receptor–ligand dissociation, protein processing and protein modification occur is, to a large extent, controlled by their optimum pH (Paroutis et al., 2004). Progressive acidification of the secretory and endocytic routes is mediated by the vacuolar-type H⁺-ATPase (V-ATPase), a nanoscale motor that couples the hydrolysis of ATP to the transmembrane movement of protons (Forgac, 2007; Vasanthakumar and Rubinstein, 2020). Because this activity is electrogenic and generates a transmembrane voltage, for net pumping to occur another ion must move. Cations moving out or anions moving in with protons can serve as counterions that dissipate the voltage (Schumacher, 2014; Stauber and Jentsch, 2013). Differential localization of the V-ATPase is mediated by the largest subunit of the membrane-integral Vo subcomplex (subunit a, ATP6V0a or VHA-a). Isoforms of subunit a are found in almost all eukaryotes and have been shown to target the V-ATPase to the late Golgi (Stv1) and the vacuole (Vph1) in yeast (Kane et al., 2020), and to neurotransmitter vesicles or endosomes (a1), the Golgi (a2), lysosomes (a3) and the plasma membrane (a3, a4) in mammalian cells (Vasanthakumar and Rubinstein, 2020). Within a given compartment, the V-ATPase is often found in the presence of a specific member of the Cl⁻ channel (ClC) family of anion transporters. Gef1, the sole yeast ClC protein, is found in the Golgi complex along with Stv1 (Schwappach et al., 1998), and mammalian ClCs are found in neurotransmitter vesicles (ClC-3; also known as CLCN3), endosomes (ClC-5; also known as CLCN5) and lysosomes (ClC-7; also known as CLCN7) (Jentsch and Pusch, 2018). Given that mutations in either ClCs or V-ATPase subunits lead to similar phenotypes or diseases, it was proposed that ClC proteins might serve as the shunt that enables V-ATPase-mediated acidification (Mindell, 2012). Even after the discovery that many members of the ClC family are secondary active transporters rather than channels (Accardi and Miller, 2004), this hypothesis still seemed plausible, as their 2Cl⁻/H⁺ exchange mode would be compatible with a Cl⁻-dependent shunt. However, loss of ClCs does not always impact acidification of the respective compartment (Kasper et al., 2005), and the use of uncoupling mutations that convert the exchangers into a Cl⁻-permeable pore has clearly demonstrated that the capacity for luminal Cl⁻ accumulation is a critical function of ClC proteins (Novarino et al., 2010; Weinert et al., 2020).

In plants, Cl⁻ is an essential micronutrient required in the oxygen-evolving complex of PSII and its toxicity under salt stress conditions has received considerable attention (Raven, 2017). However, even species that are sensitive to high Cl⁻ concentrations accumulate much more of it than required for efficient photosynthesis, and Cl⁻ can thus also be regarded as a beneficial macronutrient that stimulates growth and biomass accumulation due to its osmotic function (Wege et al., 2017). Owing to their physicochemical similarity, the fate of Cl⁻ and NO₃⁻, the other abundant inorganic anion found in plant cells, is intimately linked. Understanding the specificity of the transporters involved in uptake, storage and translocation of NO₃⁻ and Cl⁻ will be crucial in determining the physiological role of Cl⁻ more precisely. For members of the plant ClC family, a single amino acid inside the pore region defines their transport selectivity (Wege et al., 2010; Zifarelli and Pusch, 2010). Among the seven ClC proteins of Arabidopsis, ClCa and ClCb have been shown to be selective for NO₃⁻ and mediate its uptake into the vacuole (De Angeli et al., 2006; von der Fecht-Bartenbach et al., 2010). In addition, it has recently been shown that, in guard cells, ClCa accounts for cytosolic acidification in response to NO₃⁻ (Demes et al., 2020), demonstrating that, in addition to mediating the accumulation of anions, ClCs that operate as H⁺-coupled exchangers can also contribute to cytosolic pH homeostasis. According to its selectivity motif, ClCc is a tonoplast Cl⁻ transporter predominantly expressed in guard cells and required for turgor changes during stomatal opening (Jossier et al., 2010). The putative Cl⁻ channel ClCg also resides in the tonoplast and has been implicated in salt tolerance (Nguyen et al., 2016), whereas ClCe is found in the thylakoid membrane and has been proposed to be involved in nitrite uptake and Cl⁻ homeostasis (Marmagne et al., 2007; Monachello et al., 2009). ClCd has been shown to reside in the trans-Golgi network/early endosome (TGN/EE), where it colocalizes with V-ATPase complexes containing the subunit a isoform VHA-a1 (Dettmer et al., 2006; von der Fecht-Bartenbach et al., 2007). The TGN/EE is a distinct compartment that can be found in proximity to Golgi stacks but can be distinguished from trans-Golgi cisternae via a unique set of marker proteins including VHA-a1 (Rosquete et al., 2018). It serves as the central hub for protein sorting, and inhibition of the V-ATPase interferes with endocytic and secretory trafficking (Dettmer et al., 2006; Luo et al., 2015; Viotti et al., 2010). The fact that mutants lacking ClCd display hypersensitivity to the V-ATPase inhibitor Concanamycin A (ConcA) indicates that ClC activity might be required for acidification of the TGN/EE (Dettmer et al., 2006; von der Fecht-Bartenbach et al., 2007); however, a more direct involvement of Cl⁻ in the secretory pathway should not be neglected. In both scenarios, the absence of a severe phenotype argues that ClCd might be acting redundantly. Here, we show that ClCf, the only member of the Arabidopsis ClC family so far largely uncharacterized, colocalizes with VHA-a1 and ClCd in the TGN/EE. The combined loss of ClCd and ClCf causes male gametophyte lethality, indicating that ClC activity in the TGN/EE is essential. By using an artificial microRNA (amiR)-based inducible knockdown of ClCf in the clcd null background, we show that reduced ClC activity in the TGN/EE limits cell expansion. In vivo pH measurements reveal that pH in the TGN/EE is unaffected by ClC limitation; however, pH in the Golgi stack is substantially lowered.

Phylogenetic analysis of ClC protein sequences of the green lineage (Viridiplantae) revealed that orthologues of Arabidopsis ClCe and ClCf are present in many species including the charophytes Chara braunii and Klebsormidium nitens but are absent in the chlorophytes (subgroup II, Fig. 1A). In contrast, orthologues of ClCd are also present in chlorophytes, whereas a third subclade of the ClC family (ClCv) has representatives from all groups of the Viridiplantae with the exception of vascular plants (subgroups I+III, Fig. 1A). Based on its atypical sequence in the region of the selectivity filter and the proposed absence of the so-called proton glutamate, it has been suggested that ClCf could act as a channel (Zifarelli and Pusch, 2010). However, 3D homology modelling using the cryo-electron microscopy (Cryo-EM) structure of chicken ClC-7 (Schrecker et al., 2020) predicts that E247 serves as the gating glutamate, E300 is found in the position of the proton glutamate and Y463 could be involved in Cl⁻ binding (Fig. 1B). The fact that all three residues are found among all members of subclade II and are embedded in regions of high sequence conservation (Fig. 1C,D) supports our conclusion that ClCe and ClCf are likely to act as antiporters. However, in the absence of structural and functional data for members of subgroup II ClCs, their anion selectivity remains to be determined.

Based on transient expression in onion epidermal cells, ClCf has been reported to be localized at Golgi stacks (Marmagne et al., 2007). To determine the endogenous localization of ClCf, we fused the full-length ClCf coding sequence (CDS) to mRFP and expressed the resulting ClCf-mRFP fusion protein under the control of the constitutive UBQ10 promoter (Grefen et al., 2010). Confocal scanning laser microscopy (CLSM) of stable transgenics co-expressing ClCf-mRFP with the trans-Golgi marker ST-GFP revealed a motile punctate signal pattern for ClCf-mRFP with limited colocalization (Fig. 2A). In contrast, when ClCf-mRFP was co-expressed with the TGN/EE marker VHA-a1-GFP (Fig. 2B) or ClCd-GFP (Fig. 2C) the patterns strongly overlapped, and quantitative analysis revealed a high degree of colocalization between ClCf-mRFP and VHA-a1-GFP (Fig. 2D), comparable to values obtained when VHA-a1-GFP and VHA-a1-mRFP are co-expressed (Fig. 2E). Furthermore, gold particles were observed at the TGN/EE when high-pressure frozen and freeze-substituted seedlings expressing ClCf-mRFP were subjected to immunogold labelling (Fig. 2F). As ClCd and ClCf colocalize in the TGN/EE and ClCs are known to form homo- as well as hetero-dimers (Jentsch and Pusch, 2018; Weinert et al., 2020), we performed Förster resonance energy transfer-fluorescence lifetime imaging microscopy (FRET-FLIM) measurements in seedlings co-expressing ClCf-mRFP as FRET acceptor together with ClCd-GFP or VHA-a1-GFP as FRET donors. pHusion-SYP61, in which GFP and mRFP are directly linked (Luo et al., 2015), was used as a positive control. We did not observe differences in GFP lifetime for ClCd-GFP (Fig. S1A) or VHA-a1-GFP (Fig. S1B) in the presence or absence of ClCf-mRFP. In conclusion, evidence for heterodimerization between ClCd and ClCf or for a direct interaction between ClCf and VHA-a1 could not be obtained using FRET-FLIM.

To determine the in vivo function of ClCf, we identified a homozygous transfer DNA (T-DNA) insertion line from the SALK collection [SALK_112962 (Alonso et al., 2003)]. PCR-based genotyping and sequencing confirmed a T-DNA insertion in intron 6 of ClCf, and PCR on complementary DNA (cDNA) using primers flanking the insertion site revealed that the T-DNA was not spliced out and full-length transcript could not be detected (Fig. 3A). Similar to clcd, clcf mutant plants are indistinguishable from wild type under standard growth conditions (Fig. 3B). To test whether clcf, like clcd (von der Fecht-Bartenbach et al., 2007), displays hypersensitivity to ConcA, we measured hypocotyl length of etiolated seedlings in the presence of the V-ATPase inhibitor ConcA. Whereas on control medium hypocotyl length in all genotypes was comparable, clcf and clcd were found to be 15% shorter than wild type when grown in the presence of 100 µM ConcA (Fig. 3C). Lines expressing ClCf-mRFP in the clcf background behaved like wild type, indicating that ClCf-mRFP is functional (Fig. 3C). Root growth defects on alkaline medium (von der Fecht-Bartenbach et al., 2007) could not be confirmed (Fig. S2A). We next attempted to identify the clcd clcf double mutant by crossing both T-DNA insertion lines but failed to identify homozygous double mutants. We thus performed reciprocal crosses to test whether gametophyte development was affected. Via genotyping, we detected the segregating T-DNA in ∼33% of individuals when clcd clcf/+ was used as the female parent. However, when clcd clcf/+ was used as pollen donor, the respective T-DNA was present in only 2% of the offspring (Fig. 3D). In line with the reduced female transmission, seed set in mature siliques of clcf clcd/+ plants was reduced and 25% degenerated ovules were present, suggesting that knockout of both genes results in lethality (Fig. 3E). By introduction of the UBQ10:ClCf-mRFP construct, we were able to rescue the clcd clcf mutant. Nevertheless, plants were reduced in growth compared to wild type, indicating that ClCf-mRFP is not fully functional (Fig. S2B,C).

To study the function of ClCd and ClCf during vegetative growth, we employed inducible expression of an amiR to knock down ClCf in the clcd background. An amiR targeting ClCf was designed and cloned under control of the 6xOP dexamethasone (Dex)-inducible promoter/LhGR system (Craft et al., 2005). The amiR construct (Fig. 4A) was introduced into the clcd background expressing ClCf-mRFP, and two independent homozygous lines were established. Induction of the amiR by 72 h treatment with 60 µM Dex reduced ClCf-mRFP protein levels to ∼50% in clcd amiR-ClCf #1 (Fig. 4B) and less than 5% in clcd amiR-ClCf #2 (Fig. 4C). When a construct in which the amiR recognition site was mutated (mClCf-GFP) was introduced, no reduction in GFP levels was observed after Dex induction, indicating that amiR expression is causative for the observed reductions in ClCf protein levels (Fig. 4A,D). Upon Dex induction, root length was reduced to ∼85% in clcd amiR-ClCf #1 and to ∼50% in clcd amiR-ClCf #2 (Fig. 4E,F). In line with these results, we observed stunted cells in the elongation and differentiation zone as well as arrested root hair bulges in both induced clcd amiR-ClCf lines (Fig. 4B,C). Hypocotyl length of etiolated clcd amiR-ClCf #2 seedlings was also reduced upon induction. In line with the incomplete rescue of clcd clcf by ClCf-mRFP, hypocotyl growth was not fully restored but significantly less reduced in the presence of mClCf-GFP, indicating that reduced cell expansion is not due to off-target effects (Fig. S3). Taken together, these results suggest that the function of TGN/EE-localized ClCs, like the TGN/EE V-ATPase, is required for cell expansion.

To determine the effects of reduced ClC activity on the cellular level, we introduced VHA-a1-GFP (Dettmer et al., 2006) into clcd amiR-ClCf and quantified colocalization with the endocytic tracer FM 4-64. Upon application, FM 4-64 immediately stains the plasma membrane and within minutes is then delivered to the TGN/EE, from where it is either recycled back to the PM or reaches the tonoplast via late endosomes (Rigal et al., 2015). Taking the dynamic distribution of FM 4-64 into account, we quantified colocalization of VHA-a1 with FM 4-64 (Manders’ overlap coefficient M1) as well as colocalization of FM 4-64 and VHA-a1 (Manders’ overlap coefficient M2) and measured the distribution of VHA-a1 particle sizes. Whereas colocalization of VHA-a1-GFP and FM 4-64 as well as particle size distribution did not change after 3 days of Dex treatment in wild type, Col-0 (Fig. 5A,B,E,F), an increased number of larger particles was found in induced clcd amiR-ClCf seedlings (Fig. 5E). Surprisingly, in uninduced clcd amiR-ClCf, as well as in clcd single mutant seedlings, M1 and M2 were found to be significantly different from wild type (Fig. 5C,D,F). Whereas the lower M1 coefficient indicates that VHA-a1 is partially shifted to a compartment that is not stained by FM 4-64, the higher M2 coefficient indicates that more FM 4-64 accumulates in the TGN/EE and this shift is enhanced upon induction in clcd amiR-ClCf (Fig. 5F). Taken together, these results suggest that ClC function is required to maintain the functional boundaries between post-Golgi compartments and could therefore affect both secretory and endocytic trafficking.

To assess whether reduced ClC activity affects post-Golgi trafficking, we crossed plants expressing a secreted version of mRFP (secRFP; Batoko et al., 2000; Zheng et al., 2005) or the plasma membrane-localized auxin efflux carrier PIN1 (PIN-FORMED1)-GFP (Benková et al., 2003) with both inducible clcd amiR-ClCf lines. Both markers were imaged 72 h after amiR induction via Dex. However, neither intracellular accumulation of secRFP (Fig. 6A,C,E) nor mislocalization of PIN1-GFP (Fig. 6B,D,F) were observed in clcd amiR-ClCf after induction. Furthermore, we assessed whether post-Golgi trafficking to the vacuole is affected by a FM 4-64 chase experiment. We detected tonoplast labelling in Col-0 wild type and clcd amiR-ClCf lines after 100 min FM 4-64 uptake (Fig. 6G–J). In summary, we could not detect major effects on secretory and vacuolar trafficking in clcd amiR-ClCf, and it thus remains to be determined how the knockdown of ClC function in the TGN/EE leads to the observed inhibition of cell expansion.

To measure pH in the TGN/EE of clcd amiR-ClCf #2, we used an improved version of SYP61-pHusion in which we replaced mRFP by mCherry as it combines improved photostability and brightness with identical pH stable emission properties (Shaner et al., 2005). In vivo calibration of Arabidopsis lines expressing SYP61-pHusion2 showed that dynamic range and emission properties are comparable to those of the original version. However, signals were less prone to photobleaching and less excitation energy was required (Fig. S4). As expected, pH in the TGN/EE of clcd and clcf single mutants was found to be similar to wild type (Fig. 7B). To address whether ClC knockdown affects pH in the TGN/EE pH, we introduced UBQ10:SYP61-pHusion2 into the strong clcd amiR-ClCf #2 knockdown line. In vivo pH measurements and calibrations were performed in all genetic backgrounds to exclude possible quenching effects of Cl⁻ on pHusion2 (Fig. S4). Neither average pH (Fig. 7C) nor the distribution of values for individual TGN/EEs (Fig. 7D) were found to be affected in clcd amiR-ClCf plants. However, we noted that the SYP61-pHusion2 signal in clcd amiR-ClCf #2 appeared more diffuse after induction, indicating that the sensor localization might be affected (Fig. 7E). We thus next used ST-pHusion to measure pH in the neighbouring trans-Golgi cisternae (Fig. 7A) and found that it was significantly lower in the clcd amiR-ClCf #2 line after induction (Fig. 7F). Given that the V-ATPase transits through the Golgi stack on its way to the TGN/EE, hyperacidification of trans-Golgi cisternae could indicate a shift in V-ATPase distribution. However, we could not detect increased colocalization between ST-mCherry and VHA-a1-GFP in clcd amiR-ClCf upon induction (Fig. S5). How hyperacidification of trans-Golgi cisternae causes reduced cell expansion remains to be determined.

The TGN/EE is a highly dynamic compartment that many proteins pass en route to their destination. Among the residents that stay behind proteins that maintain pH and ion homeostasis are dominant. As the primary proton pump, the V-ATPase plays a central role; however, the presence of several other transport proteins indicates that their concerted action is required (Schumacher, 2014; Sze and Chanroj, 2018). NHX5 and NHX6, two K⁺/H⁺-antiporters, reside in the TGN/EE and in multivesicular bodies; however, it remains to be determined whether their function as proton leaks to alkalize the endosomal lumen during maturation or the luminal potassium concentration is responsible for the severe phenotype of nhx5 nhx6 mutants (Bassil et al., 2011; Dragwidge et al., 2019). Similarly, ClCd is localized at the TGN/EE and could contribute to the shunt conductance required to balance the electrogenicity of the V-ATPase or maintain the concentration of Cl⁻ or other anions in the lumen of the TGN/EE (von der Fecht-Bartenbach et al., 2007). However, compared to mutants with reduced V-ATPase activity (Brüx et al., 2008; Schumacher et al., 1999), the clcd mutant has a rather mild phenotype, and we thus reasoned that it might act redundantly with ClCf, the only member of the Arabidopsis ClC family that remained largely uncharacterized. The evolutionary origin of all ClCs can be traced back to prokaryotes, but ClCe and ClCf belong to a particular subgroup that is closely related to cyanobacterial ClCs, indicating that they might be of plastidial rather than mitochondrial origin. Indeed, subgroup II is confined to the green lineage, and ClCe is found in thylakoid membranes (Marmagne et al., 2007; Monachello et al., 2009). Members of the ClCf clade are characterized by the presence of an N-terminal extension of variable length and low sequence conservation that does not contain features predictive of chloroplast localization. Based on biochemical approaches, ClCf has been reported to be localized in the chloroplast outer envelope (Teardo et al., 2005). However, based on transient expression of a GFP fusion protein, ClCf has been localized to Golgi membranes, and the fact that ClCf rescues the gef1 mutant that lacks the Golgi-localized sole yeast ClC independently confirms that ClCf is not a plastidial protein (Marmagne et al., 2007). We have shown here that ClCf colocalizes with ClCd and VHA-a1 in the TGN/EE, and it will be of interest to determine whether localization of the two only distantly related ClCs is mediated by a similar mechanism. Not only is knowledge regarding TGN/EE retention or retrieval of integral membrane proteins limiting, functional analysis of ClCd and ClCf will require their rerouting to membranes amenable to patch-clamp analysis such as the tonoplast. Our attempts at rerouting have so far not been successful, and we thus have to rely on predictions based on homology modelling. Whereas ClCd belongs to the well-characterized subgroup I and has all the hallmarks of an H⁺-coupled anion transporter with a strong preference for Cl⁻ over NO₃⁻, ClCf has been suggested to have channel-like properties, as the characteristic proton glutamate seemed to be missing (Zifarelli and Pusch, 2010). Based on homology modelling, we have shown here that E300, a residue conserved in all subgroup II ClCs, could occupy the same position as E312, the proton glutamate of chicken ClC-7 (Schrecker et al., 2020), indicating that ClCf and members of subgroup II in general are proton-coupled antiporters. Despite the absence of experimental evidence regarding selectivity, Cl⁻ seems to be the most likely substrate, yet other anions of course cannot be excluded. In any case, the results of our genetic analysis are in line with ClCd and ClCf having redundant functions as H⁺/Cl⁻ antiporters in the TGN/EE. Both single mutants can be distinguished from wild type based on their increased sensitivity to the V-ATPase inhibitor ConcA, a phenotype that could be indicative of their role as shunt conductances that counter the charge imbalance created by H⁺ pumping (Sze and Chanroj, 2018). ClCd has been reported to negatively regulate pathogen-triggered immunity in response to flg22 (Guo et al., 2014), whereas reduced V-ATPase activity dampens the flg22 response (Keinath et al., 2010), indicating that the functional relationship between ClCs and the V-ATPase might be more complex and a more detailed characterization of the single mutants might be required to reveal non-redundant functions. However, the fact that the combined loss of ClCd and ClCf causes male gametophyte lethality, a phenotype also observed in V-ATPase null mutants (Dettmer et al., 2005; Lupanga et al., 2020), also supports a role for ClC activity in supporting acidification of the TGN/EE. Further analysis is required to determine whether the underlying pollen defects are indeed similar, yet developing pollen is not the most suitable experimental system to study ClC function in the TGN/EE in more detail. To investigate the role of ClCs in TGN/EE acidification, we established an inducible amiR-based knockdown system for ClCf in a clcd null background and used an improved version of the TGN-localized pH sensor SYP61-pHusion to measure pH. To our surprise, reduced ClC activity did not affect pH in the TGN/EE but instead caused increased acidification of trans-Golgi cisternae. It remains to be determined whether impaired cell elongation is the consequence of improper TGN/EE anion homeostasis or whether the altered trans-Golgi pH is responsible.

Although these results do not seem to support a role for ClCs as shunt conductances, it needs to be considered that the knockdown is likely incomplete so that residual ClC activity might still be sufficient to maintain V-ATPase activity in the TGN/EE. Hyperacidification of the trans-Golgi could affect different aspects of cell wall biosynthesis. Cellulose synthase complexes (CSCs) are assembled in the Golgi stack (Zhang et al., 2016), and suboptimal pH could affect CSC assembly or structure of individual CesA subunits. The major pectin component hemigalacturonan (HG) is synthesized in the Golgi complex and methyl-esterification occurs in the trans-Golgi. The more acidic trans-Golgi pH could alter pectin methyltransferase activity, leading to overall lower HG methyl-esterification, and it been suggested that enzymes responsible for pectin stability and degradation are co-processed with its substrate in the Golgi stack (Anderson, 2016). However, less obvious connections also need to be considered. In yeast mutants lacking Gef1, the only ClC protein, the multicopper oxidase Fet3 fails to acquire copper ion cofactors and, by the introduction of an uncoupling mutation that allows acidification but abolishes antiport activity, it was shown that Gef1 has marked effects on cellular glutathione homeostasis (Braun et al., 2010). Also in mammalian cells, uncoupling mutations that convert ClCs from antiporters to channels have revealed that the capacity for luminal Cl⁻ accumulation is a critical function of ClC proteins (Novarino et al., 2010; Weinert et al., 2020), and it has been suggested that endomembrane compartments exploit Cl⁻ for volume regulation (Stauber and Jentsch, 2013). We still know very little about the spatiotemporal aspects of TGN/EE maturation and its functional separation from the Golgi stack. However, the effects of the monovalent cation ionophore monensin that induces rapid osmotic swelling of the TGN/EE followed by trans-Golgi cisternae (Dragwidge et al., 2019; Zhang et al., 1993) underline that volume regulation is a critical aspect. Recently, the Cation Chloride Cotransporter 1 (CCC1) of Arabidopsis has been shown to be localized at the TGN/EE and to affect pH and ion homeostasis. By mediating both cation and anion efflux, CCC1 complements the ion transport circuit of the TGN/EE (McKay et al., 2020 preprint). However, the circuit might not yet be complete. In mammalian cells, the proton-activated Cl⁻ (PAC; also known as DUSP) channel releases Cl⁻ from the endosomal lumen. In contrast to CCC1, deletion of PAC leads to hyperacidification of endosomes (Osei-Owusu et al., 2021; Ullrich et al., 2019). It remains to be determined whether a PAC1-like protein is present in plants, but, in any case, determining the functional dependencies between the different players is likely to provide further insight into the physiology of the TGN/EE.

ClC protein sequences (Table S1) were pairwise aligned (global alignment with free end gaps with gap open penalty and gap extension penalty of 3) using the Blosum 45 cost matrix, and a neighbour-joining tree was built using the Jukes–Cantor genetic distance model in Geneious 10.1.3.

The annotated, full-length amino acid sequence of ClCf was submitted as a query without reference structure to the online protein structure prediction tool ‘I-Tasser’ (Yang and Zhang, 2015). The top-ranking model among the structural analogues in PDB was identified to be 7JM6A (ClC-7), with a TM-score of 0.772. Images of the resulting structure were generated using ChimeraX-1.1.1. (Pettersen et al., 2021).

To clone UBQ10:ClCf-mRFP, the full-length CDS of ClCf (At1g55620) was amplified without stop codon from Arabidopsis thaliana Col-0 cDNA using primers ClCf-AatII-Fw and ClCf-noSTOP-PvuI-Rv. The 2355-bp PCR fragment was subcloned into pJET1.2blunt and sequenced. The CDS of mRFP was amplified from SYP61-pHusion (Luo et al., 2015) using mRFP-PvuI-Fw and mRFP-SalI-Rv. The resulting 693-bp fragment was subcloned into pJETblunt1.2. The respective fragments were released from pJET1.2blunt using restriction enzymes AatII, PvuI and SalI, respectively, and ligated with pUTKan (Krebs et al., 2012) that has been opened with AatII and SalI to finally obtain UBQ10:ClCf-mRFP.

The UBQ10:ST-GFP construct was generated and assembled using the GreenGate (GG) cloning system (Lampropoulos et al., 2013). The CDS of rat sialyltransferase (ST) was amplified from P16:ST-pHusion (Luo et al., 2015) with GG-ST-C-Fw and GG-ST-C-Rv primers, followed by subcloning into pGGC000 to receive pGGC-ST. After sequencing, pGGC-ST was used in a GG reaction with the modules listed in Table S2 to generate UBQ10:ST-GFP.

UBQ10:ClCd-GFP was generated using GG cloning. To generate the entry module pGGC-ClCd, two internal Eco31I F sites were removed from the wild-type ClCd sequence by introduction of silent mutations. Furthermore, the entire sequence of ClCd intron 1 was added due to toxic effects of the gene product for bacteria. For that purpose, four separate PCR reactions were performed. First, a 86-bp fragment of the 5′ region of ClCd was amplified from cDNA with primers GG-ClCd-C-Fw and GG-ClCd-Int1-Rv, thereby mutating nucleotide at position 54 from T to A, followed by a second PCR, in which a 518-bp fragment was amplified from Arabidopsis genomic DNA with primers GG-ClCd-Int2-Fw and GG-ClCd-Intron1-Rv. In a third PCR reaction, using primers GG-ClCd-Intron1-Fw and GG-ClCd-Int2-Rv, an 881-bp fragment was amplified, thereby mutating A to G at position 1731. The last PCR reaction using primers ClCd-Int3-Fw and ClCd-noSTOP-Rv amplified a 677-bp product of the ClCd CDS. All four PCR products were digested with Eco31I and ligated in pre-opened pGGC000 to receive pGGC-ClCd. The correct assembly of the entry clone was verified by sequencing. pGGC-ClCd was then used in a standard GG reaction with the modules listed in Table S2 to generate UBQ10P:ClCd-GFP.

The Dex:amiR-ClCf construct was generated using GG cloning. An amiR against ClCf (5′-AGCGAACCCTGACCATAATA-3′) was designed using the WMD3 microRNA design tool (Schwab et al., 2006). The amiR targets the ClCf CDS between position 2928 and 2947 and was introduced into pRS300 using oligonucleotides amiR-ClCf-I, amiR-ClCf-II, amiR-ClCf-III and amiR-ClCf-IV. The 439-bp amiR cassette was amplified using primers GG-amiR-FW-B-overhang-EcoRI and GG-amiR-RV-C-overhang-BamHI digested with Eco31I, and integrated into pGGI000 (Lampropoulos et al., 2013) to obtain pGGI-amiR-ClCf. The amiR-ClCf cassette was verified by sequencing and subsequently used to generate the intermediate vector pGGN-6xOP:amiR-ClCf via a GG reaction with the modules listed in Table S2. The final GG reaction was performed with pGGN-6xOP:amiR-ClCf, pGGM-UBQ10:Lh4GR (Lupanga et al., 2020) and the destination vector pGGZ003 (Lampropoulos et al., 2013) to receive Dex:amiR-ClCf.

VHP1:mClCf-GFP was created using GG cloning. To generate an amiR-resistant version of ClCf (mClCf) several bases within the 20-bp amiR recognition site differing from the wild-type ClCf sequence were mutated without changing the amino acid sequence (5′-CGAAAATCCAGATCACAACT-3′). For this, the full-length mClCf CDS without stop codon was amplified from UBQ10:ClCf-mRFP in two separate PCR reactions using the primers GG-ClCf-C-Fw and GG-mClCf-Rv and mClCf-Fw and GG-ClCf-C-Rv, which resulted in 1814-bp and 572-bp fragments, respectively. Both fragments were Eco31I digested and inserted into pGGC000 to obtain pGGC-mClCf. A correct clone was confirmed by sequencing. pGGC-mClCf was then used in a GG reaction with the modules listed in Table S2 to obtain VHP1:mClCf-GFP.

For UBQ10:SYP61-pHusion2, the bleaching-sensitive mRFP of pHusion was exchanged for the more bleaching-resistant mCherry. For this, the CDS of mCherry was amplified from pJET-mCherry with primers mCherry-HindIII-Fw and mCherry-noSTOP-Eco31I-Rv, introducing HindIII sites at the 5′ and a unique 4-bp overhang (TAAC) at the 3′ end. EGFP was amplified from P16:SYP61-pHusion (Luo et al., 2015) using primers EGFP-Eco31I-Fw and pHusion-SalI-Rv, which spans the full-length CDS of EGFP plus a five-amino acid linker (AVNAS). A fragment encoding SYP61 was released with BamHI and HindIII from P16:SYP61-pHusion (Luo et al., 2015). The three fragments obtained, SYP61 (BamHI/HindII), mCherry (HindIII/TAAC) and EGFP (TAAC/SalI), were ligated with BamHI/SalI-digested pUTkan (Krebs et al., 2012).

For UBQ10:ST-pHusion, the CDS of rat ST was amplified from 35S:ST-pHluorin (Martinière et al., 2013) using primers ST-AatII-Fw and ST-noSTOP-PvuI-Rv, which removed the stop codon. The CDS of pHusion was amplified from P16:SYP61-pHusion with primers pHusion-PvuI-Fw and pHusion-SalI-Rv. The ST and pHusion PCR fragments were subcloned into pJETblunt1.2. Correct clones were verified by sequencing. ST and pHusion fragments were released from their respective pJET backbones using AatII/PvuI and PvuI/SalI, respectively, and finally ligated into the AatII/SalI-digested vector backbone of the binary vector pUTBar. All primers used for cloning of the constructs are listed in Table S3. Cloning of VHP1:ST-mCherry has been described previously (Stührwohldt et al., 2020).

If not stated otherwise, A. thaliana ecotype Col-0 seeds were grown on media containing 0.5×Murashige and Skoog (MS) medium, 0.5% sucrose, with pH adjusted to 5.8 with KOH. Plates were solidified using 0.6% phyto agar. Agar and MS basal salt mixture were purchased from Duchefa Biochemie. Seeds were surface sterilized with EtOH followed by stratification for 48 h at 4°C. Seedlings were grown at 22°C with cycles of 16 h light and 8 h darkness.

For hypocotyl measurements, seeds were surface sterilized and plated on medium consisting of 1% phyto agar, 5 mM MES-KOH, pH 5.8. After stratification at 4°C for 48 h, seeds were exposed to 5 h light (120 μmol m⁻² s⁻¹) before being wrapped with a double layer of aluminium foil. Seedlings were then grown in the dark for 4 days at 22°C. To measure hypocotyl length, seedlings were sandwiched between two layers of acetate and scanned. The digitized images were analysed using the segmented line tool in Fiji (Schindelin et al., 2012). For the inhibitor treatment, 100 nM ConcA (Santa Cruz Biotechnology) or the equivalent volume of DMSO was added to the plant growth medium. For each condition, hypocotyls from 30 seedlings were measured.

For root length measurements, seedlings were grown for 7 days on vertical plates containing 0.5×MS medium, pH 5.8, 1% phyto agar in long-day conditions. Plates were supplied with 30 μM Dex (Sigma-Aldrich) or equivalent amounts of DMSO.

Mature siliques of A. thaliana ecotype Col-0 wild type and the ClCd/+ ClCf mutant, harbouring one copy of the T-DNA insertion for ClCd (ClCd-1) and two copies of the T-DNA insertion for ClCf (ClCf-1), grown under long-day conditions, were opened along the seed valve from base to tip. Phenotype of the seed set was documented using a SteREO Discovery V20 (Carl Zeiss Microscopy).

To analyse whether transmission through either male or female gametes is dependent on ClCf, reciprocal crosses between wild type and ClCd/+ ClCf were performed. The resulting F1 offspring was analysed for presence of ClCf wild type and ClCf-1 T-DNA insertion allele via genotyping as stated above. Fifty plants were analysed for each reciprocal cross.

UBQ10:ClCf-mRFP and UBQ10:SYP61-pHusion2 were introduced into Agrobacterium tumefaciens strain GV3101:pMP90. UBQ10:ST-GFP, UBQ10:ClCd-GFP, VHP1:mClCf-GFP and VHP1:ST-mCherry were introduced into Agrobacterium tumefaciens strain ASE containing the pSOUP helper plasmid. Arabidopsis thaliana ecotype Col-0 wild-type plants were used for transformation via floral dip using standard procedures. Transgenic lines were isolated on plates containing 0.5×MS medium, 0.5% sucrose, pH 5.8 (KOH) and 0.55% phyto agar supplemented either with 50 µg/ml Kanamycin, 10 mg/ml BASTA, 11.25 µg/ml Sulfadiazine or 25 µg/ml HygromycinB.

To study the function of ClCf, a homozygous T-DNA insertion allele (clcf-1) from the SALK collection [SALK_112962 (Alonso et al., 2003)] was characterized. The position of the T-DNA insertion in intron 6 was verified by PCR using primers listed in Table S3 followed by confirmation of the insertion site by sequencing. Gene-specific primers ClCf-fwd and ClCf-rev1 were used to detect exon 6, which is situated before the T-DNA insertion. A gene-specific primer pair (ClCf-fwd, ClCf-rev2) spanning the T-DNA insertion and a gene-/T-DNA-specific primer pair (ClCf-fwd, TL1-LB) were used to detect the T-DNA insertion in the ClCf locus. PCR amplification of an ACT2 fragment with gene-specific primers was used as a control.

Stable transgenic Arabidopsis lines expressing VHA-a1-GFP (Dettmer et al., 2006), P16:SYP61-pHusion (Luo et al., 2015), secRFP (Zheng et al., 2005) and PIN1-GFP (Benková et al., 2003), as well as the T-DNA insertion allele for clcd (clcd-1; von der Fecht-Bartenbach et al., 2007), have been described previously.

CLSM of Arabidopsis roots was performed with an SP5II system equipped with an inverted DMI6000 microscope stand (Leica Microsystems). Images were recorded using a HCX PL APO lambda blue 63.0×1.20 WATER UV or a HCX PL APO CS 20.0×0.70 IMM UV objective (Leica Microsystems). GFP was excited with 488 nm; mRFP was excited with 561 nm. Fluorescence was detected at 500–545 nm for GFP and 620–670 nm for mRFP using hybrid detectors (Leica Microsystems). Images for non-quantitative purposes were adjusted in brightness and contrast using Fiji (Schindelin et al., 2012).

Average fluorescence intensities were determined in roots of 7-day-old Arabidopsis seedlings expressing UBQ10:ClCf-mRFP or VHP1:mClCf-GFP in the clcd amiR-ClCf background 3 days after amiR-ClCf induction with 60 μM Dex. Plants were pre-grown for 4 days on 0.5×MS medium, pH 5.8, 0.55% phyto agar and afterwards transferred to identical control medium or medium supplied with 60 μM Dex. Plants were further grown for 3 days in identical light conditions before imaging was performed. Imaging of mRFP and GFP was performed as described above. Average fluorescence intensity values were quantified with Fiji, a distribution of ImageJ (Schindelin et al., 2012), using a custom macro that is available from the authors upon request. Per genotype and condition, 15 images were taken. The experiment was repeated three times.

For colocalization analysis, images were acquired using the HCX PL APO lambda blue 63.0×1.20 water immersion objective. The imaging parameters were as follows: 1024×1024 pixel resolution with a zoom of 2.5 or 512×512 pixel resolution with a zoom of 5.3 was chosen to adjust the voxel size to 96 nm. All microscope settings remained identical for each sample. All images were processed with Fiji (Schindelin et al., 2012) using a Gaussian blur filter with a sigma radius of 1. Pearson and Spearman correlation coefficients as well as scatter plots were calculated using the PSC Colocalization plugin (French et al., 2008) with a threshold level of 10. Manders' overlap coefficient between VHA-a1-GFP- and FM 4-64-positive endosomes or ST-mCherry (Stührwohldt et al., 2020) was determined in 7-day-old Arabidopsis seedlings stably expressing VHA-a1-GFP in wild type, in clcd or in amiR-ClCf clcd. Prior to induction, seedlings were grown for 4 days on 0.5×MS medium, pH 5.8, 0.5% phyto agar followed by transfer to identical control medium or medium supplied with 60 μM Dex. Seedlings were imaged 72 h after Dex induction and after 15 min of staining with 1 µM FM 4-64 in liquid medium (0.5×MS medium, pH 5.8 with KOH). To determine Manders’ overlap coefficient, images of ten cells from eight to ten different seedlings within the root elongation zone were acquired using a HCX PL APO lambda blue 63.0×1.20 water immersion objective. The imaging parameters were as follows: 512×512 pixel resolution with a zoom of 5.3 was chosen to adjust the voxel size to 96 nm. All microscope settings remained identical for each sample. All images were processed with Fiji (Schindelin et al., 2012) using a Gaussian blur filter with a sigma radius of 1. Regions of interests were drawn manually to exclude the plasma membrane signal of FM 4-64. M1 and M2 Manders’ overlap coefficients were calculated using the JACoP plugin in Fiji (Bolte and Cordelières, 2006) with a threshold of 40 for both channels.

Surface area measurements of TGN/EEs and TGN/EE clusters were performed in root epidermal cells of 7-day-old Arabidopsis seedlings stably expressing UBQ10:SYP61-pHusion2 in wild type, in clcd and in clcd amiR-ClCf background. Plants were pre-grown for 4 days and then transferred to identical control medium or medium supplemented with 60 μM Dex. Plants were grown for 3 days further on the respective medium prior to imaging. Images were acquired using the HCX PL APO lambda blue 63.0×1.20 water immersion objective with pinhole size of 0.75 arbitrary units. The fluorescence signals of the GFP channel were used for size calculation in Fiji (Schindelin et al., 2012). Average background values were determined and subtracted from each image. Afterwards, the Gaussian blur filter with a sigma radius of 1 was applied, and images were thresholded (30/255) to remove nonspecific background signals. The particle size of single or clusters of TGN/EEs was determined using the Particle analyzer with particle sizes of 0.1–3.0 μm² and a circularity of 0.05–1.00. Values were grouped according to area.

pH measurements in the TGN/EE using P16:SYP61-pHusion, UBQ10:SYP61-pHusion2 and UBQ10:ST-pHusion were performed as described previously (Luo et al., 2015).

Entire root tips of 6-day-old Arabidopsis seedlings stably expressing UBQ10:ClCf-mRFP were harvested and processed as previously described (Scheuring et al., 2011) Freeze substitution was performed in an electron microscope AFS2 freeze substitution unit (Leica Microsystems) in dry acetone supplemented with 0.3% uranyl acetate as previously described (Hillmer et al., 2012). Ultrathin sections of 80–90 nm were obtained using an Ultracut S (Leica Microsystems) ultra microtome with a diamond knife (Diatome). Grid-mounted sections were incubated with an anti-α-DsRed antibody at 1:200 dilution, followed by incubation with a goat anti-rabbit (1:50) antibody linked to 10 nm colloidal gold particles. Immunolabelled sections were observed with a JEM-1400 (JEOL, Tokyo, Japan) electron microscope operating at 80 kV. Micrographs were taken with a FastScan F214 digital camera (TVIPS). Brightness and contrast was adjusted using Fiji (Schindelin et al., 2012).

We thank Fabian Fink for technical assistance and Beate Schöfer for plant care.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.258807