Arabidopsis choline transporter-like 1 (CTL1) regulates secretory trafficking of auxin transporters to control seedling growth

Auxin controls a myriad of plant developmental processes and plant response to environmental conditions. Precise trafficking of auxin transporters is essential for auxin homeostasis in plants. Here, we report characterization of Arabidopsis CTL1, which controls seedling growth and apical hook development by regulating intracellular trafficking of PIN-type auxin transporters. The CTL1 gene encodes a choline transporter-like protein with an expression pattern highly correlated with auxin distribution and is enriched in shoot and root apical meristems, lateral root primordia, the vascular system, and the concave side of the apical hook. The choline transporter-like 1 (CTL1) protein is localized to the trans-Golgi network (TGN), prevacuolar compartment (PVC), and plasma membrane (PM). Disruption of CTL1 gene expression alters the trafficking of 2 auxin efflux transporters—Arabidopsis PM-located auxin efflux transporter PIN-formed 1 (PIN1) and Arabidopsis PM-located auxin efflux transporter PIN-formed 3 (PIN3)—to the PM, thereby affecting auxin distribution and plant growth and development. We further found that phospholipids, sphingolipids, and other membrane lipids were significantly altered in the ctl1 mutant, linking CTL1 function to lipid homeostasis. We propose that CTL1 regulates protein sorting from the TGN to the PM through its function in lipid homeostasis.


Introduction
Dynamic endomembrane trafficking delivers proteins and other cargo molecules to a variety of organelles and thus controls almost all aspects of plant development and physiology, including gravitropism, epidermis differentiation, guard cell movement, cell wall remodeling, defense against pathogens, and hormone signaling [1]. In this context, exocytosis and endocytosis determine the abundance and dynamics of signaling receptors and transporters at the plasma membrane (PM), which in turn afford cells the ability to respond to extracellular stimuli. For example, PM-located auxin transporters PIN-formed (PINs) and AUXIN RESIS-TANT 1 (AUX1) have been shown to continuously cycle between PM and endosomal compartments, resulting in dynamic changes in localization and abundance at the PM, where they function to control auxin gradients between cells. Many components have been shown to affect the trafficking and PM abundance of auxin transporters. These include the nucleotide exchange factor for ADP-ribosylation-factor-type GTPases (ARF-GEFs) [2][3][4], retromer proteins [5], adaptor protein complex 3 (AP-3) [6,7], endosomal sorting complexes [8,9], the trans-Golgi network/early endosome (TGN/EE) protein ECHIDNA [10], and several protein kinases [11][12][13]. The trafficking regulators safeguard the polarity and dynamics of auxin transporters, making possible the spatiotemporal variations in auxin distribution that act as an instructive signal for a wide variety of cellular events, ranging from polarity establishment during the earliest phases of plant embryogenesis to other morphogenetic events, including root development and the formation of the apical hook during seedling development [2][3][4][5][10][11][12][13][14]. Because of the dynamics of these auxin transporters in the endomembrane system, they can be used as cargo proteins to study the regulation of the membrane trafficking process in cell biology. In genetic analysis, the mutations in genes encoding these transporters are often trackable because the mutants often show severe phenotypic changes in plant growth and development.
In a screen for mutants defective in phloem development, Dettmer et al. [15] identified choline transporter-like 1 (CTL1) to be essential for sieve tube formation and thus overall plant development. The CTL1 gene encodes a protein with multiple transmembrane domains that has been shown to mediate choline transport [15]. However, the cellular mechanism for CTL1 function in plant development remains unknown. In tracking down the mechanism of CTL1 function, we discovered that CTL1 is closely associated with auxin signaling in the control of plant developmental processes. We further found that disruption of CTL1 function impaired endomembrane trafficking of several auxin transporters and probably some other unidentified cargo proteins as well. CTL1 acts at both the secretory vesicles (SVs) and the clathrin-coated vesicles (CCVs) of the TGN to regulate the trafficking of PIN1 and PIN3, which may explain the phenotypes of the ctl1 mutant in seedling growth and apical hook development. As a choline transporter, CTL1 plays a crucial role in the homeostasis of membrane lipids, including phospholipids and sphingolipids. As lipid compositions of cell membranes influence vesicular transport, CTL1 function provides the mechanistic link between choline transport, lipid remodeling, and vesicular trafficking of PM proteins. We conclude that CTL1 is a previously unrecognized regulator of endomembrane trafficking in plants.

Results
A null mutant lacking CTL1 displays a cell elongation defect A significant portion of plant genomes encode proteins with multiple transmembrane domains that may function as transporters. However, a large majority of these open reading frames (ORFs) are designated as "proteins of unknown function." We took a comprehensive reverse genetic approach to identify the function of these proteins by screening Arabidopsis SALK transfer deoxyribonucleic acid (T-DNA) insertion lines disrupting the expression of corresponding genes. One of the lines, SALK_065853C in Columbia-0 (Col-0) background, was recovered for its severely stunted growth phenotype. This line contains a T-DNA insertion in the At3g15380 gene, previously named CHOLINE-TRANSPORTER-LIKE 1 (CTL1) for its biochemical function as a choline transporter and its function in sieve tube formation [15].
The insertion in CTL1 occurred in the first intron (281 bp from the start codon) to generate a transcriptional knockout (S1A-S1C Fig). We backcrossed the ctl1 mutant with the wild type for 2 generations to segregate other possible T-DNA insertions and found that the isolated homozygous ctl1 mutant lacking other insertions consistently showed the typical ctl1 phenotype observed with the original SALK line. We further confirmed that the phenotype of the ctl1 mutant resulted from lack of CTL1 function by complementation using a genomic fragment of the CTL1 gene and a CTL1-enhanced green fluorescent protein (EGFP) fusion construct driven by the CTL1 promoter, respectively (S1D Fig and S2A Fig). The transgenic plants expressing CTL1 genomic fragment or CTL1-EGFP fusion protein in the ctl1 mutant background displayed a similar growth phenotype to that of the wild-type plants (Fig 1A, S1E Fig  and S2D Fig), supporting the conclusion that disruption of the CTL1 gene rendered the dwarf phenotype of the mutant plants.
We examined the ctl1 phenotype in detail and found that 30-day-old ctl1 seedlings grown in the soil displayed smaller, round, and dark green epinastic leaves with short petioles (Fig 1A  and 1B and S1F-S1H Fig) and that 2-month-old adult ctl1 mutants exhibited a dwarfed phenotype ( Fig 1C). To examine root growth, we planted seeds on half-strength Murashige and Skoog (MS) agar medium and found that 5-day-old ctl1 seedlings had shorter roots as compared to the wild type ( Fig 1D and 1E), consistent with previous observations [15].
We further examined the cell size in both wild-type and mutant plants and found that the stunted growth phenotype in the ctl1 mutant resulted from a defect in cell elongation. To determine cell size, we expressed AUX1-GFP in wild-type, ctl1, and complemented lines (ctl1C) so that the cell periphery lit up with the GFP for confocal imaging. The images collected from confocal laser scanning microscopy revealed that the size of leaf epidermal cells in ctl1 was smaller than that of those in the wild type and ctl1C (Fig 1F-1H). Typically, the cell size in the ctl1 mutant plants was only 49% of the wild type (S1 Table). As hypocotyl cells elongate quickly without cell division, hypocotyls grown in the dark are excellent models for examining cell elongation [16]. We thus analyzed the hypocotyl length of etiolated seedlings and found that the hypocotyls of ctl1 mutant seedlings elongated more slowly compared with the wild type and ctl1C (Fig 1I and 1J).
Moreover, the epidermal cell length of ctl1 hypocotyls was only about half of that of the wild type ( Fig 1K). Taken together, these results demonstrate that CTL1 is an essential regulator of cell elongation in Arabidopsis. This finding explains an overall dwarf phenotype of the ctl1 mutant seedlings beyond the defect in phloem development observed earlier [15].
CTL1 is localized to the PM, TGN, and prevacuolar compartment (PVC) To further study the cellular mechanism of CTL1 function, we determined the subcellular localization of CTL1 in root cortex cells of Arabidopsis seedlings by using a CTL1-EGFP transgenic line in the ctl1 mutant background. Real-time quantitative reverse transcription PCR The growth phenotype of wild-type (WT), ctl1 mutant, and complementation lines transformed with At3g15380 genomic DNA (ctl1C). After being grown on half-strength Murashige and Skoog (MS) medium for 7 days, the seedlings were transferred to the soil and grown for 23 days before photographs were taken. Bar = 1 cm. (B) The rosette leaves from 30-dayold ctl1 mutant as compared to those of the WT. Bar = 1 cm. (C) The dwarfed phenotype of 2-month-old adult ctl1 plants compared to the WT and ctl1C. Bar = 2 cm. (D) Five-day-old WT and ctl1 mutant grown on halfstrength MS medium. Bar = 3 mm. (E) The root length of 5-day-old WT, ctl1, and ctl1C. Three independent experiments were performed. Data are mean ± SD. n = 6 for each genotype. Significant differences are indicated as *P < 0.05 (Student t test). (F-H) Comparison of epidermal cells in the cotyledon of 5-day-old WT (F), ctl1 (G), and ctl1C seedlings (H). Representative cells were highlighted with yellow asterisks. Scale bars in panels F-H are 20 μm. (I) Growth phenotype of 5-day-old etiolated WT and ctl1 seedlings. Bar = 2 mm. (J) Hypocotyl length of etiolated WT, ctl1, and ctl1C seedlings. Data are mean ± SD. n = 5 for each genotype. (K) The hypocotyl cell length of 5-day-old etiolated WT, ctl1, and ctl1C seedlings. Data are mean ± SD, and significant differences are indicated as *P < 0.05 (n = 30, Student t test). The raw data for panels E, J, and K can be found in S1 Data. (qRT-PCR) assay showed that this transgenic line had levels of CTL1 transcript that were similar to those in the wild-type plants ( S2B Fig). Furthermore, the CTL1-EGFP construct in this transgenic line complemented the dwarfed phenotype of the ctl1 mutant (S2D Fig), indicating that this line expressed functional CTL1-EGFP fusion proteins in the same compartments as the native CTL1 protein. Therefore, we used this line to track down the subcellular location of CTL1-EGFP protein using confocal microscopy. The fluorescence signal generated by CTL1-EGFP fusion appeared at the PM and also displayed a punctate pattern in the cytosol (Fig 2A  and S2E Fig). To identify intracellular punctate structures at which CTL1-EGFP is localized, we crossed this CTL1-EGFP transgenic line with the transgenic lines that stably express fluorescent markers for several organelles. We found that most of the punctate signals of CTL1-EGFP colocalized with vacuolar proton ATPase subunit VHA-a isoform 1 monomer red fluorescent protein (VHA-a1-mRFP) (Fig 2B), a TGN/EE marker [17]. The linear Pearson correlation coefficient (r p ) for the colocalization was 0.87 ( Fig 2I). We did not observe colocalization of (G) were captured after the roots were treated with 100 μM BFA for 60 minutes. Bars = 2 μm. (H) CTL1-EGFP signals in root cells after they were treated with 33 μM wortmannin (Wm) for 60 minutes. White arrow shows the ring-like structure. Bars = 2 μm. (I-K) Colocalization analysis of CTL1-EGFP with VHA-a1-mRFP (I), SYP32-mCherry (J), or Rha1-mCherry (K) using the ImageJ software with the colocalization finder plugin. The linear Pearson correlation coefficient (r p ) indicates the percentage of signal overlap, and an r p value of 1.0 represents 100% colocalization. (L) The colocalization percentage of CTL1-EGFP with VHA-a1-mRFP, SYP32-mCherry, or Rha1-mCherry. Data are mean ± SD. Six images acquired from 3 roots were used. The raw data for panel L can be found in S1 Data. CTL1-EGFP fluorescence signals with Syntaxin of plants 32 (SYP32)-mCherry (r p = 0) ( Fig 2C  and 2J), the Golgi apparatus marker [18]. In addition, some CTL1-EGFP fluorescence signals overlapped with those of Arabidopsis Rab homolog F2A (Rha1)-mCherry (r p = 0.16) (Fig 2D  and 2K), the PVC marker [18]. We conducted more experiments to further examine the subcellular localization of CTL1. First, we stained root tips with FM4-64 for a short time to label the PM [19] and found that the CTL1-EGFP fluorescence colocalized with the FM4-64 signal ( Fig 2E). Secondly, we treated root tips with fungal toxin brefeldin A (BFA), which blocks trafficking from the TGN to the PM, leading to the formation of enlarged endosomal compartments referred to as BFA bodies [20]. We detected both CTL1-EGFP and VHA-a1-mRFP fluorescence in the BFA bodies ( Fig 2F). However, the CTL1-EGFP and SYP32-mCherry signals did not overlap in the presence of BFA ( Fig 2G). Finally, we treated root tips with wortmannin (Wm), a drug that preferentially triggers the swelling of PVC but has little effect on the morphology of the Golgi apparatus and the TGN/EE [21]. After 1 hour of treatment with 33 μM Wm, the morphology of some organelles labeled by CTL1-EGFP became swollen and showed a PVC-specific ring-like structure (Fig 2H), indicating that CTL1 is also sorted to the PVC.

Postembryonic expression of CTL1
We used the CTL1-EGFP transgenic line to study the expression pattern of CTL1, since the expression of CTL1-EGFP was driven by the CTL1 promoter region (S2 Fig). The GFP fluorescence was ubiquitously distributed in the meristem and elongation zones but restricted to vascular tissues in the mature zone of roots ( Fig 3A). Cross sections through the root meristem zone displayed the distribution of CTL1-EGFP signal in all cell layers (Fig 3B and 3C). Furthermore, GFP was also highly expressed in the newly initiated lateral roots (Fig 3D and 3E).
In dark-grown seedlings, CTL1 was expressed in the whole cotyledon with enrichment at the blade tips, the vasculature of the hypocotyl, and the concave side of the apical hook ( Fig  3F). When grown on the half-strength MS medium under dark conditions, Arabidopsis apical hook development proceeds through 3 phases: the hook formation phase, the maintenance phase, and the opening phase [10,23,24]. CTL1-EGFP signals appeared at both the convex and the concave sides of the hook, with the maximum in the concave side (S4G and S4H Fig), and the expression at the concave side gradually disappeared during the opening phase (S4A-S4F and S4J Fig), suggesting that CTL1 function may be associated with apical hook development.
To further confirm the expression pattern of CTL1, we generated transgenic plants harboring the CTL1pro:CTL1:GUS construct in which the EGFP reporter was replaced by the β-glucuronidase (GUS) reporter. In the apical hook region, CTL1-GUS shows the same expression pattern as the CTL1-EGFP in the concave side of the apical hook (S4I Fig). In the roots of 5-day-old seedlings, GUS activity was detected in the similar tissues labeled by EGFP signals (Fig 3G-3J). In shoots of 7-day-old seedlings, the expression pattern of CTL1 was ubiquitous, but a relatively higher level of GUS activity was detected in the apical meristem, at the tips of cotyledon and true leaves, and in vascular tissues (Fig 3K-3N), largely overlapping with the GFP signals detected earlier ( Fig 3F).

CTL1 loss-of-function mutant shows defects associated with auxin
Phenotypic analyses of the ctl1 mutant in detail provided insights into CTL1 function in growth and development. The ctl1 mutant showed reduced primary root growth, whereas the lateral roots reached approximately the same length as the primary root, apparently resulting from lack of apical dominance [15]. In addition, we found that the ctl1 showed a severe defect in apical hook development, consistent with high levels of expression of CTL1 in the apical hook ( Fig  3F and S4 Fig). In particular, etiolated ctl1 seedlings never formed the fully closed apical hook, and the hook quickly opened without an apparent maintenance phase (Fig 4A and 4B).
ctl1 defects in cell elongation (described earlier) and defects in primary and lateral root growth all point to relevance with the auxin response. Furthermore, apical hook development is the result of differential cell elongation mediated by uneven auxin distribution [23]. These auxin-related phenotypes of ctl1 prompted us to investigate the connections between CTL1 and auxin. We went on exploring whether there are other auxin-related phenotypes altered in the ctl1 mutant. One process we studied was phototropism, a well-characterized light response mediated by asymmetric auxin distribution in plants [4]. As shown in Fig 4, ctl1 had a reduced hypocotyl phototropic response to unilateral light as compared with the wild type and the complemented line (Fig 4C and 4D). In addition, we also found that ctl1 seedlings were less sensitive to high levels of 1-naphthylacetic acid (NAA) that inhibit root elongation. Compared to the wild type, ctl1 roots were resistant to auxin inhibition over a range of auxin concentrations (0, 0.1, and 0.3 μM) (Fig 4E and 4F). We also tested the sensitivity to salicylic acid (SA) and found that both the wild type (WT) and the ctl1 mutant were sensitive to SA (S5 Fig), unlike the situation with auxin. All these findings support the hypothesis that CTL1 is involved in auxin-related processes. Auxin distribution is altered in the ctl1 mutant To study auxin distribution in the ctl1 mutant, we crossed it with a transgenic line harboring DR5::GUS, a reporter of auxin levels [5]. In the ctl1 mutant background, the DIRECT REPEAT5 (DR5)-driven GUS expression was greatly enhanced in the cotyledons, especially at the tips, but was dramatically reduced in the vascular tissues of the hypocotyls and roots, as compared to the pattern of DR5::GUS in the WT background (Fig 5A-5H and S6A Fig). In the roots, the DR5::GFP signal in the WT and the ctl1 mutant showed a similar distribution pattern, with a condensed signal in the primary root tips. However, after being treated with NAA (0.1, 0.3, 0.5, and 1 μM) for 12 hours, the DR5::GFP signal was enhanced in the meristem and elongation zones of WT roots with increasing NAA concentrations, but the level of this response was reduced in ctl1, suggesting reduced transport or reduced response of auxin ( Fig  5I and S6B Fig). Moreover, a maximum DR5::GFP signal was detected in the concave side of the apical hooks of the WT, but this was largely diminished in the mutant (Fig 5J and S6C Fig).

CTL1 controls the abundance of PIN1 and PIN3 in the root and the apical hook
We showed that disruption of CTL1 function appears to affect auxin transport and distribution (Figs 4 and 5). Together with the finding of CTL1 subcellular localization to the PM and  (Fig 2), we speculated that CTL1 may regulate the trafficking of PM-localized proteins so that distribution of auxin transporters may be altered in the ctl1 mutant. In order to test this hypothesis, we examined the expression and localization of several auxin transporters in the ctl1 mutant by using auxin transporter-GFP markers. We crossed the ctl1 mutant with transgenic lines containing PIN1pro:PIN1:GFP, PIN2pro:PIN2:GFP, PIN3pro:PIN3: GFP, AUX1pro:AUX1:GFP, or LAX3pro:LAX3:GFP and monitored the GFP signals by confocal microscopy. We found a dramatically reduced accumulation of PIN1-GFP and PIN3-GFP in the ctl1 mutant as compared to the WT. More specifically, the abundance of PIN1-GFP in the mutant was about half (56% ± 14.2%) of the WT, and for PIN3 it was only a quarter (26% ± 19.3%) of the WT (Fig 6A, 6C and 6F). However, other auxin transporters, including PIN2, AUX1, and Like AUX1 protein 3 (LAX3), were not affected by the CTL1 mutation (Fig 6B and  6D-6F). We also found that the brassinosteroid receptor BRI1 showed no difference in the WT and the ctl1 mutant (S8 Fig). Because PIN-mediated auxin efflux also plays a major role during apical hook development [24,25], we also examined GFP signal in the apical hook and again found reduced levels of PIN1-GFP and PIN3-GFP, but not of AUX1-GFP, in the ctl1 mutant as compared to the WT (S7 Fig). The loss of PIN3 can cause an enhanced DR5 signal in the cotyledons of Arabidopsis [25]. As we also detected an enhanced DR5 signal in the cotyledons of ctl1 mutant, we proposed that reduced abundance of PIN1 and PIN3 may lead to changed auxin transport and distribution (shown by DR5::GUS/GFP lines earlier) that in turn causes changes in auxin-related phenotypes in the ctl1 mutant. To determine whether the reduced abundance of PIN1 and PIN3 proteins in the ctl1 mutant results from a decrease in the transcript level, we examined the RNA levels of auxin transporter genes in the WT and the ctl1 mutant. The qRT-PCR analysis showed that the mRNA levels of PIN1 and PIN3 in ctl1 were slightly higher than those in the WT (S9 Fig), suggesting that downregulation of PIN1 and PIN3 proteins in ctl1 occurs at the post-transcriptional level.
CTL1 regulates trafficking of PIN1 and PIN3 to the PM The reduced abundance of PIN1 and PIN3 in the ctl1 mutant, together with CTL1 localization to the TGN, suggests that CTL1 protein may affect the recruitment of PINs from the TGN to the PM. In the study of protein trafficking, BFA is often used to block protein sorting from the TGN to the PM and to induce protein internalization from the PM to the TGN. Thus, BFA treatment of cells results in the formation of large endosomes called BFA bodies that contain PM-destined proteins. These BFA bodies are rapidly disintegrated, and the proteins are targeted back to the PM again after BFA washout [26][27][28]. Therefore, BFA washout has been widely used to measure the speed of trafficking from the TGN back to the PM [26][27][28][29]. We analyzed the trafficking of PIN1-GFP and PIN3-GFP using a similar approach. Upon treatment of WT and ctl1 roots with 50 μM BFA for 60 minutes, similar percentages of cells with BFA bodies were detected in the WT and the ctl1 mutant (Fig 7A, 7B and 7I). However, after BFA washout for 90 minutes, we detected 9.7% and 21.6% of root cells containing PIN1-GFPdecorated BFA bodies in the WT and ctl1, respectively (Fig 7C, 7D and 7I). Similarly, about 7.6% and 23.5% of cells contained PIN3-GFP-labeled BFA bodies in the WT and ctl1, respectively (Fig 7E-7H and 7J). We also did the BFA washout experiment using the PIN2-GFP/WT and PIN2-GFP/ctl1 transgenic lines and found no difference between the WT and the ctl1 mutant (S10 Fig). These data indicate that loss of CTL1 causes a delay of trafficking in PIN1 and PIN3 proteins, which may in turn lead to the auxin-related developmental defects we observed in the ctl1 mutant plants.

CTL1 plays a vital role in the homeostasis of membrane lipids in Arabidopsis
It has been reported that sphingolipids are involved in the trafficking of auxin transporters [30][31][32]. Choline is predominantly utilized for the synthesis of essential lipid components of the cell membranes, including sphingolipid [33]. A defect in choline transport may impair the homeostasis of these lipids, thereby affecting vesicle trafficking. To test if CTL1 functions in lipid homeostasis, we first monitored the lipid content in situ in the WT and mutant seedlings using Nile Red staining [32,34]. The results showed that both polar and nonpolar lipids were reduced in the apical hook region of the ctl1 mutant as compared to the WT (S11 Fig). To provide a more detailed analysis of lipid content, we used ultra-performance liquid chromatography coupled with mass spectrometry (UPLC-MS) to survey the lipid profiles of WT and ctl1 plants. In the ctl1 mutant, the total lipids were reduced to about 84% of the WT level (Fig 8A). In further analysis, we found that the glycerophospholipids (lysophosphatidylethanolamine, phosphatidylcholine, and phosphatidylinositol) in the ctl1 mutant were reduced (Fig 8B). The content of total glycerophospholipids in ctl1 mutant was about 77% of the level in the WT ( Fig  8C). Although the total sphingolipid contents were similar between the WT and the ctl1 mutant (Fig 8C), we found that the content of some specific types of sphingolipids were significantly altered in the ctl1 mutant (Fig 8D). In particular, the abundance of sphingolipids containing very-long-acyl-chain fatty acids (C > 18 carbons) was altered in the ctl1 mutant (S12 Fig), consistent with the earlier finding [31] that these sphingolipids are required for the targeting of specific auxin carriers to the PM.
To link the sphingolipids metabolism to CTL1 function further, we used the sphingolipid synthesis inhibitor fumonisin B1 (FB1) to treat the WT seedlings and found that this treatment induced apical hook defects as observed in the ctl1 mutant (S13 Fig). Taken together, CTL1, as a choline transport protein, plays a role in homeostasis of lipids, including sphingolipids. The disturbed homeostasis of sphingolipids and other lipids in ctl1 mutant may affect the delivery of PIN1, PIN3, and other cargos from the TGN to the PM.

CTL1 is essential for auxin-regulated plant development
Auxin is the most investigated phytohormone and plays a pivotal role in virtually every aspect of plant growth and development, and such function often depends on its differential distribution within plant tissues [35]. At the whole plant level, auxin controls organogenesis, apical dominance, lateral root emergence, apical hook development and tropism, vascular development, and many other developmental processes [23,[36][37][38][39]. At the cellular level, auxin regulates cell elongation, cell division, and cell differentiation [40].
We showed that Arabidopsis CTL1 plays a key role in a number of auxin-regulated processes, including cell elongation, apical hook development, root morphology, and hypocotyl phototropic response to unilateral light in Arabidopsis. CTL1 expression corresponds to the auxin maxima in the shoot apical meristems, the tips of leaves, the lateral root primordia, the vascular system, and the concave side of apical hook. The dynamic changes in the CTL1 expression pattern strongly resemble the changes in auxin distribution during apical hook development [24]. Moreover, CTL1 is required for maintaining a normal pattern of DR5 expression, a marker for auxin distribution. Loss of function of CTL1 results in enhanced DR5 signals in the tips of cotyledons and reduced signals in the vascular system of hypocotyls, roots, and the concave side of the apical hook. The auxin-related phenotypes and altered auxin Fucα1-2GalNAcβ1-4(NeuAcα2-8NeuAcα2-3) Galβ1-4Glcβ-Cer; Galβ1-Cer: Galβ1-3GalNAcβ1-3Galα1-3Galβ1-4Glcβ-Cer; GalNAcβ1-Cer: GalNAcβ1-3Galα1-3Galα1-3Galα1-3Galα1-4Galβ1-4Glcβ-Cer; GlcNAc β1-Cer: GlcNAcβ1-4Manβ1-4Glcβ-Cer; KDNα2-Cer: KDNα2-6Galβ1-4GlcNAcβ1-3Galβ1-4Glcβ-Cer; Manα1-Cer: Manα1-3Manβ1-4Glcβ-Cer; and NeuAcα2-Cer: NeuAcα2-8NeuAcα2-3Galβ1-4Glcβ-Cer. The classification of these SPs was from the method provided by the website LIPID Metabolites and Pathways Strategy(LIPID MAPS) (http://www.lipidmaps.org/). Data are mean ± SD. Four independent experiments were performed. Red asterisks indicate a significant difference (*P < 0.05, Student t test). The raw data for panels A-D can be found in S1 Data.
https://doi.org/10.1371/journal.pbio.2004310.g008 distribution observed in the ctl1 mutant provide insights on the cellular and physiological processes that underpin the role of CTL1 in plant development.

CTL1 affects the trafficking of auxin transporters to the PM
Auxin action depends on its differential distribution in plant tissues through a process called polar auxin transport [41]. A number of PM auxin transporters, such as PINs, AUX1, LIKE AUX1 (LAX), and subfamily B of the ATP-binding cassette proteins (ABCB), contribute to the influx and efflux of auxin and subsequently determine the distribution of auxin among different tissues and cells. Therefore, auxin homeostasis requires accurate subcellular targeting of these auxin transporters through secretory trafficking processes. In plant cells, the TGN is a central traffic hub responsible for protein trafficking in both biosynthetic and endocytic pathways. The endocytosed cargos from the PM were received by the TGN, and also through the TGN, biosynthetic cargos were sorted to the PM, cell wall, cell plate, PVC, tonoplast, and vacuole [42]. Protein composition and localization at the PM are regulated by exocytosis and endocytosis in response to developmental and environmental cues.
Here we provide strong evidence that CTL1 is located at multiple subcellular sites associated with secretory trafficking, including the TGN, PVC, and PM. As an endosomal protein, CTL1 may contribute to the trafficking of proteins from the TGN to the PM, especially those PM-localized auxin transporters, as supported by the finding that the abundance of some PINtype auxin transporters was reduced in the ctl1 mutant, which displays a number of auxinrelated phenotypes. In particular, CTL1 is required for the efficient trafficking of PIN1 and PIN3 from the TGN to the PM. As a consequence of losing CTL1 function, the abundance of PIN1 and PIN3 is reduced, and auxin distribution is altered in the root and the apical hook. Further, the CTL1 expression pattern largely overlaps with those of PIN1 and PIN3 in roots and apical hooks [43] [44]. Loss of function of PIN1 or PIN3 causes defects in apical hook development, primary and lateral root growth, and response to unilateral light, which are also the defects found in the ctl1 mutant [4,45,46]. Together, results on expression pattern, CTL1 contribution to PIN1 and PIN3 trafficking, and similar mutant phenotypes all support the conclusion that CTL1 regulates the trafficking of PIN1 and PIN3, thereby controlling auxinrelated plant developmental processes.

The mechanistic insights of CTL1 function in plant cells
CTL1 has been recently identified as a choline transporter required for sieve plate development [15], although the cellular mechanism underlying CTL1 function is not clear. Our discovery that CTL1 functions in promoting the trafficking of PIN1 and PIN3 may provide a mechanism for CTL1 function in auxin-related developmental processes, including previously described sieve plate biogenesis. In eukaryotic cells, lipid-dependent protein sorting is a major delivery mechanism for cargo proteins from the TGN to cell surface [47]. However, the mechanism by which lipids control vesicular trafficking remains unclear. As membrane lipids provide the microenvironment for protein actions, each membrane system consists of different domains with a specific composition of lipids, such as the sphingolipid-rich microdomains and lipid rafts. Recent studies also reveal various TGN subdomains with different sphingolipid and sterol compositions [48]. Choline is one of the precursors for the synthesis of membrane lipids, including phospholipids and sphingolipids. It is conceivable that CTL1, as a choline transport protein, would play a role in choline delivery to various cell membranes and thus modulate lipid composition of multiple organelles in plant cells. The defect in the homeostasis of membrane lipids in the ctl1 mutant would cause the trafficking defects of some cargo proteins such as PIN1 and PIN3.
Although we cannot detect any differences between the WT and the ctl1 mutant on the morphology, size, and number of the compartments labelled by different fluorescence-tagged organelle markers using confocal microscopy (S14 Fig), there might be other defects on the endomembrane system in ctl1 that we were not able to identify. It remains unclear whether CTL1 functions as a transporter-like protein that prefers choline over other substrates [15]. Further investigations will be required to determine whether CTL1 transports choline specifically or can also transport other substrates. CTL1 is localized to different cellular compartments, including the TGN, PM, and PVC, and its trafficking is sensitive to BFA, suggesting that CTL1 itself is a cargo protein that goes through the secretory pathway. We found that CTL1 affects the trafficking of some special cargo proteins such as PIN1 and PIN3, but not PIN2. This indicates that CTL1 is not required for the proper function of global trafficking machinery; instead, it affects subgroups of proteins. Further survey on the cargo proteins that are affected by CTL1 and identification of proteins that interact with CTL1 will help further understanding of the mechanisms of CTL1 function.

Pharmacological treatments
The seedlings were incubated in liquid half-strength MS medium containing a final concentration of 50 μM BFA (Sigma-Aldrich) or 33 μM wortmannin (Sigma-Aldrich) for 60 minutes, unless described otherwise in the figure legends. Labeling with FM4-64 (Molecular Probes) was carried out as described [29]. For BFA washouts, the seedlings were treated with 50 μM BFA for 1 hour, rinsed 3 times with water, and then incubated in half-strength MS liquid medium for 90 minutes before the imaging procedure.
Root length, hypocotyl length, and apical hook curvature measurements Images of individual seedlings were acquired using the OLYMPUS stereoscopic microscope equipped with a DP27 camera. The root length, hypocotyl length, and hook angle were then measured using the ImageJ software. The bending angles of the apical hook were scored as previously described [24]. For each time-lapse experiment, 6 seedlings were analyzed.

Generation of plasmid constructs and transgenic Arabidopsis lines
For genetic complementation, a 6,107-bp genomic DNA fragment containing a 1,918-bp promoter region, a 3,117-bp coding region, and a 1,072-bp terminator region after the stop codon of the CTL1 gene were amplified from WT genomic DNA and then cloned into the binary vector pCAMBIA-1300. To generate the CTL1pro:CTL1:EGFP construct, we fused the EGFP and NOS terminator sequence with the 5,032-bp genomic DNA of CTL1, including the 1,918-bp promotor region and the 3,114-bp coding region, and then cloned this recombinant DNA into the binary vector pCAMBIA-1300. A 10-alanine (Ala) linker sequence (GCT GCT GCC GCT GCC GCT GCG GCA GCG GCC) was also inserted between CTL1 and EGFP. To generate the CTL1pro:CTL1:GUS construct, we fused the GUS and NOS terminator sequence to the CTL1 gene as described for the CTL1pro:CTL1:EGFP construct. The organization of these 2 constructs is shown in S1D and S2A Figs, respectively. To generate the LAX3pro:LAX3:EGFP construct, a 3,007-bp genomic DNA of LAX3 including the promotor region and the coding region was fused with the EGFP and NOS terminator sequence and then cloned into pCAM-BIA1300. The constructs were introduced into the Agrobacterium tumefaciens GV3101 strain for transformation into Arabidopsis by the floral dipping method [55]. The primers used for plasmids constructions are listed in S2 Table. Transgenic lines expressing various markers including DR5::GFP, DR5::GUS, PIN1pro: PIN1:GFP, PIN2pro:PIN2:GFP, PIN3pro:PIN3:GFP, AUX1pro:AUX1:GFP, LAX3pro:LAX3: GFP, BRI1pro:BRI1:GFP, VHA-a1-mRFP, SYP32-mCherry, and Rha1-mCherry were individually crossed into the ctl1 mutant background. Homozygous ctl1 plants harboring various markers were isolated from F2 populations by scoring the specific phenotype of the mutant. The F3 and later generations were used for analyses. For colocalization analysis, transgenic lines expressing VHA-a1-mRFP, SYP32-mCherry, Rha1-mCherry, and mCherry-CLC2 were individually crossed with transgenic lines expressing CTL1pro:CTL1: EGFP.

Confocal microscopy analysis
Imaging was performed on an LSM-710 confocal microscope (Zeiss) equipped with an argon/ krypton laser. For the quantitative fluorescence intensity, confocal pictures were acquired using strictly identical acquisition parameters (laser power, photomultiplier, offset, zoom factor, and resolution) among the experimental seedlings. The excitation wave lengths for the GFP, FM4-64, mRFP, and mCherry signals were 488, 514, 543, and 587 nm, respectively. For colocalization analysis, the Colocalization Finder plugin of ImageJ was used. The linear Pearson correlation coefficient (r p ) was used to indicate the extent of colocalization, with the value of +1.0 as complete colocalization. Detection of nonpolar lipids (excitation at 514 nm, emission at 520-560 nm) and polar lipids (excitation at 534 nm, emission at 600-700 nm) after Nile red staining was performed according to the published protocol [56].

Quantitative real-time PCR analysis
Total RNA was extracted from plant samples using TRIzol reagent (Invitrogen), followed by synthesis of Poly (dT) complementary DNA using the M-MLV Reverse Transcriptase (Promega). qRT-PCR was performed using the SYBR Green I Master kit (Roche Diagnostics) according to the manufacturer's instructions on a CFX Connect Real-Time System (Bio-Rad). All individual reactions were done in triplicate. The primers used for qRT-PCR are listed in S2 Table. Lipidomics analysis by UPLC-MS Plant lipids from 5-day-old WT and ctl1 mutant seedlings were extracted according to the published method [58], and the extracts were directly subjected to LC-MS analysis. The lipidomic data were recorded on an ESI-QTOF/MS (Xevo G2-S Q-TOF, Waters) coupled with a UPLC (ACQUITY UPLC I-Class system, Waters). Parameters for mass spectrometry were as follows: scan range, m/z 50-1,500; ion source, ESI; loop time, 0.2 seconds; cone voltages, 25 KV; capillary voltage, 3 KV; collision energy, 15-60 V; source temperature, 120˚C; desolvation temperature, 400˚C; desolvation gas flow, 500 L/h; and cone gas flow, 25 L/h. Raw data were imported into the commercial software Progenesis QI (Version 2.3) for data processing, including peak picking and acquiring compounds' associated information such as m/z, retention time, and intensity. Next, data filtering was performed to delete low-quality data. R project was also applied in further data processing and statistical analysis. hypocotyl, and root parts of the WT and the ctl1 mutants. Data are mean ± SD calculated from 5 seedlings (Student t test, Ã P < 0.05). (B) Relative intensity of DR5-green fluorescent protein (DR5-GFP) signals in the root tips of the WT and ctl1 mutants when treated with 0, 0.1, 0.3, 0.5, or 1 μM 1-naphthylacetic acid (NAA). Data are mean ± SD calculated from 5 seedlings. (C) Relative intensity of DR5-GFP signals in the concave sides of the apical hook of the WT and the ctl1 mutant at 1 and 3 days after germination. Data are mean ± SD calculated from 5 seedlings (Student t test, Ã P < 0.05). The raw data for panels A-C can be found in S1 Data. (E) Percentage of cells with PIN2-GFP labeled BFA bodies before and after BFA washout in the WT and the ctl1 mutant. Data are mean ± SD. Three independent experiments were performed. Five roots were used for each experiment (Student t test, Ã P < 0.05). The raw data for panel E can be found in S1 Data. The organelle size and intensity were measured using the ImageJ software. Data are mean ± SD, and the significant difference was analyzed by Student t test. Six images from 3 roots were used. The raw data for panels G and H can be found in S1 Data. (TIF) S1 Table. The