ABCB-mediated shootward auxin transport feeds into the root clock

Although strongly influenced by environmental conditions, lateral root (LR) positioning along the primary root appears to follow obediently an internal spacing mechanism dictated by auxin oscillations that prepattern the primary root, referred to as the root clock. Surprisingly, none of the hitherto characterized PIN- and ABCB-type auxin transporters seem to be involved in this LR prepatterning mechanism. Here, we characterize ABCB 15 , 16 , 17 , 18 , and 22 (ABCB 15 - 22 ) as novel auxin-transporting ABCBs. Knock-down and genome editing of this genetically linked group of ABCBs caused strongly reduced LR densities. These phenotypes were correlated with reduced amplitude, but not reduced frequency of the root clock oscillation. High-resolution auxin transport assays and tissue-specific silencing revealed contributions of ABCB 15 - 22 to shootward auxin transport in the lateral root cap (LRC) and epidermis, thereby explaining the reduced auxin oscillation. Jointly, these data support a model in which LRC-derived auxin contributes to the root clock amplitude.


Introduction
The root system of plants is of vital importance for their growth and survival as it anchors the plant in the soil and is required for the uptake of water and nutrients and symbiotic interactions. The complexity of root systems can be easily expanded by LR branching according to environmentally imposed limitations and stimuli (Motte et al, 2019). LR development is a multistep process occurring over a long time, involving coordinated signaling across several tissues (Stoeckle et al, 2018). The plant hormone auxin is a key regulator of many organogenetic events in plants (Vanneste & Friml, 2009). Its local accumulation triggers dramatic, preprogrammed transcriptional changes that are associated with the progression of the developmental program (Vanneste & Friml, 2009). This is also the case for LR development, where auxin accumulation defines the positioning of prebranch sites (PBS) along the primary root, and thus root architecture complexity (De Smet et al, 2007;Dubrovsky et al, 2008;Moreno-Risueno et al, 2010;Xuan et al, 2016). Therefore, plants have established intricate mechanisms to control auxin distribution within tissues (Rosquete et al, 2012;Adamowski & Friml, 2015), which can be adjusted according to the developmental stage, hormones, and environmental signals (Motte et al, 2019). In the current model, it is proposed that the decision to initiate LR formation is made in a zone close to the meristem (De Smet et al, 2007;De Rybel et al, 2010). In this zone, oscillatory gene expression, also referred to as the root clock, was reported to correlate with the activity of the auxin signaling output reporter DR5:LUC (Moreno-Risueno et al, 2010). This periodic auxin signaling selects a subset of cells to gain a higher competence to form a LR reflected in a maintained expression of the auxin output reporter DR5:LUC. These sites together with the developing LRs display strong DR5:LUC activity and are together referred to as PBS (Moreno-Risueno et al, 2010). Indole-3-butyric acid (IBA) to indole-3-acetic acid (IAA) conversion in the LRC contributes to the amplitude of this oscillation (De Rybel et al, 2012;Xuan et al, 2015), and cyclic programmed cell death of the LRC contributes to the frequency of this oscillation (Xuan et al, 2016). An alternative model is the reflux-and-growth model, which proposes auxin oscillations are an emergent feature associated with meristem cell division and elongation (van den Berg et al, 2021). In this model, cell division and meristem size determine the oscillation frequency, while cell elongation controls oscillation amplitude.
The prevailing model of auxin transport in the root meristem can best be summarized as a reverse fountain of auxin flowing rootward through the vascular tissue and being redirected shootward through the outer layers of the meristem (Grieneisen et al, 2007). This outer shootward auxin flow is thought to rejoin the central rootward auxin flow. Both aforementioned LR prepatterning models are based on the principles outlined by the reverse fountain model. The reverse fountain model for auxin transport is based on cellto-cell transport via a highly coordinated network of auxin uptake and efflux carriers. The uptake of IAA is mainly affected by efficient IAA À /H + symporter via AUX1/Like-AUX1 family members (Swarup & Bhosale, 2019). Multiple members of the PIN and ABCB protein families are known effectors of cellular efflux. All Arabidopsis PINs transport IAA into the endoplasmic reticulum or into the apoplast (Adamowski & Friml, 2015), likely via the proline cross-over-based elevator mechanism of deprotonated, cytosolic IAA À that was recently uncovered (Ung et al, 2022;Yang et al, 2022). In contrast, of 22 full-sized ABCBs (containing two transmembrane domains and two nucleotide-binding domains) (Kang et al, 2011), only ABCB1, ABCB19, ABCB4, ABCB21, ABCB6, ABCB20 have been implicated in auxin transport (Geisler et al, 2017;Zhang et al, 2018;Jenness et al, 2022). ABCB14 was shown to transport malate instead of IAA (Lee et al, 2008). Recently, a group of closely related ABCBs (ABCB15,16,17,18,and 22) were predicted to also transport IAA, based on the presence of a diagnostic D/E-P motif (Hao et al, 2020). Genetic and biochemical studies have indicated a tight interplay between both PINs and ABCBs Blakeslee et al, 2007;Mravec et al, 2008;Deslauriers & Spalding, 2021). A recent simulation of auxin transport in the root meristem identified strong PIN-ABCB co-dependent auxin efflux, in combination with individual auxin transport activities, and fluxes via plasmodesmata as the most likely scenario underpinning auxin transport in the root tip (Mellor et al, 2020(Mellor et al, , 2022. The proposed auxin transport components of the reverse fountain are AUX1, for auxin uptake, in combination with PIN1, PIN2, PIN3, PIN4, PIN7, ABCB1, ABCB19, and ABCB4 for auxin efflux. Of these key auxin transporters, only AUX1 was experimentally proven to be involved in the auxin component of the root clock (Xuan et al, 2016). Despite the strong overlap between the expression domains of PIN2, ABCB1, ABCB4, and ABCB19 efflux carriers, neither the pin2 mutant, nor the pin2abcb1abcb19 triple mutant showed reduced LR densities (Xuan et al, 2016). Moreover, the reported abcb4 root phenotypes seem to be dependent on the growth conditions (Santelia et al, 2005;Terasaka et al, 2005;Lewis et al, 2007;Kube s et al, 2012). This suggests that the auxin efflux component of the root clock is distinct from the currently characterized set of auxin transporters.
Here, we show that five closely related plasma membrane localized ABCBs (ABCB15,16,17,18,and 22) contribute to shootward auxin transport in the root. By characterization of knock-outs and knock-down lines, we found that their reduced LR density correlates with a reduced amplitude of the root clock oscillation. Interestingly, a detailed analysis of the root meristem and LRC of our knock-down lines revealed a strong correlation between LRC cell death rate and root clock oscillation frequency, instead of the predicted reduced oscillation frequency. Jointly, our data expand the repertoire of auxin-transporting ABCBs to improve our understanding of auxin transport mechanisms in the root.

Results
Loss of function of cluster ABCB15-22 genes causes defects in root system architecture Of 22 full-sized ABCBs in Arabidopsis (Kang et al, 2011), we selected five closely related, but poorly characterized ABCBs, ABCB15,16,17,18,, for detailed functional characterization.
First, we analyzed the phenotype of T-DNA insertion mutants in these genes. None of these single mutants displayed significant defects in LR density (Appendix Fig S1A-E), indicating functional redundancy among these ABCBs. Their proximity on the chromosome precluded generating higher-order mutants by crossing. Therefore, we attempted to simultaneously target multiple members of this subgroup on via CRISPR/Cas9-mediated genome editing and via a silencing approach.
On the one hand, we designed three multiplex genome editing constructs to target multiple members of the group III ABCBs (Appendix Fig S2A-D). From transformants expressing the respective constructs, we could isolate: a line with mutations in the entire subgroup, named penta CRISPR (Appendix Fig S2A); three lines with different mutant alleles in ABCB16, 17, 18, and 22, named quadri CRISPR (lines F33#1, F33#6 and B64) (Appendix Fig S2C and  D); and a double mutant in ABCB16 and 18, named b16b18 CRISPR (Appendix Fig S2B). An attempt to delete the entire genomic region via genome-editing was not successful.
On the other hand, we identified within a collection of artificial microRNA (amiRNA) lines (Zhang et al, 2018) a homozygous line for pro35S:amiR-2572 (named amiR-2572), overexpressing an amiRNA that is predicted to target ABCB15, ABCB16, ABCB17, ABCB18, and ABCB22 (Appendix Fig S3A and Dataset EV1). Via Q-RT-PCR on dissected root meristems, we confirmed the transcriptional silencing of ABCB15, ABCB16, ABCB18, and ABCB22 (Appendix Fig S3B). Similarly, by crossing amiR-2572 to the corresponding YFP-ABCB reporters, we observed reduced protein levels of most of these ABCBs, but not of the non-target ABCB1 and ABCB19 (Appendix Fig S3C and D).
Assuming limited overlap of potential off-targets in these independent knock-out and silencing lines, we determined the developmental importance of these genes. Strikingly, the amiR-2572 line, the penta CRISPR and the three quadri CRISPR lines all had shorter roots, a reduced LR density (Fig 1A-G) and smaller rosettes in the soil (Appendix Fig S2E and F). Consistently with more genes being mutated, the penta CRISPR phenotypes were more outspoken than quadri CRISPR and amiR-2572, and even showed reduced fertility (Appendix Fig S2E). In contrast, the b16b18 CRISPR double mutant did not display significant root phenotypes compared to WT (Fig 1A-D). The gradient of phenotypic penetrance in penta CRISPR compared to quadri CRISPR and b16b18 CRISPR supports functional redundancy and/or cooperativity among ABCB15,16,17,18,and 22, in the root and shoot.
Together, these analyses suggest a role for members of the ABCB15-22 cluster in root architecture. Additional mutant combinations will be needed to fully dissect the individual contributions of each ABCB to the observed phenotypes.

ABCB15-22 control auxin oscillation amplitude
Given the pronounced phenotypes in the root, we focused on understanding the LR defects in these lines. More detailed phenotyping revealed that amiR-2572 had a strong reduction in the early LRP stages, without accumulating intermediate LRP stages (Appendix Fig S4A and  B), suggesting that the reduced density of emerged LRs in these lines is due to a defect at the level of LR initiation. Lateral root initiation (stage I) is the first anatomical hallmark of LR formation and is preceded by a local maximum of auxin signaling that was installed by the root clock pre-patterning. Such persistent auxin maxima, together with LRP visualized by DR5:LUC, hotspots are jointly referred to as prebranch sites (PBS) (Bustillo-Avendano et al, 2022) (Appendix Fig S4C). Consistently with the reduced LR initiation (Appendix Fig S4A and B), amiR-2572 showed a reduced prebranch site density (Fig 2A and B). Prebranch site formation is instructed by an oscillating auxin signaling (DR5:LUC) maximum (Appendix Fig S4D-F), of which the oscillation period correlates with the rate of programmed cell death in the LRC, and its amplitude is correlated with the concomitant auxin burst derived from the dying cells (Xuan et al, 2016). The root meristem was significantly shorter in amiR-2572 than in WT (Fig 2C and D). Interestingly, the LRC contained similar numbers of cells in both backgrounds ( Fig 2E), with the most distal cell being shorter in amiR-2572 than in WT (Fig 2F), jointly indicating a reduced cell elongation in the LRC to match the reduced root meristem size. The period of the disappearance of DR5:VENUS stripes demarcating cell death in the LRC was also unaffected (Fig 2G and H) and did match with a DR5:LUC oscillation period that was similar in amiR-2572 and wild type ( Fig 2I). In contrast, the DR5:LUC oscillation amplitude in amiR-2572 was significantly smaller than in the WT ( Fig 2J). That amiR-2572 impacts on the root clock amplitude, but not its frequency, correlates with the observed LR defect. According to the current model, a sufficiently intense auxin response is required to translate auxin oscillations into prebranch sites, and thus represents a plausible explanation for the LR defect in amiR-2572. It is however difficult to exclude effects of the overall reduced growth habit of amiR-2572.
ABCB15-22 activities increase cellular IAA efflux ABCB15, 16, 17, 18, and 22 contain the conserved D/E-P motif that was proposed to be diagnostic of their auxin transporting activities (Hao et al, 2020). This, together with the defects in auxin-regulated LR development and root clock amplitude, prompted us to test their auxin transport capacities.
We first determined their subcellular localization. The YFP-ABCB signals co-localized with propidium iodide in Arabidopsis roots ( Fig 3A; Appendix Fig S5A) and were present in Hechtian strands, connecting the retracted cell to the cell wall after plasmolysis ( Fig 3B; Appendix Fig S5B). Similarly, these YFP-ABCBs colocalized with the endocytic tracer dye FM4-64 at the plasma membrane in protoplasts prepared from Agrobacterium-transfected N. benthamiana leaves (Appendix Fig S5C). Jointly, these data show that ABCB15-22 localize to the plasma membrane.
Overexpression of respective YFP-ABCBs increased the IAA export in N. benthamiana mesophyll protoplasts at rates comparable to those seen upon overexpression of the best characterized auxin-transporting ABCB1   (Fig 3C). The presence of intact IAA in the supernatant after protoplast separation demonstrated that the N. benthamiana-expressed ABCBs stimulated the export of intact IAA, and not a catabolic product (Appendix Fig S6A and B). Importantly, the loading of radiolabeled substrates was not indirectly influenced by over-expression of ABCB as exemplified for ABCB17 (Appendix Fig S6C). Mutation of P980 of the conserved D/E-P motif in ABCB17 to glycine entirely reverted IAA export to vector control levels ( Fig 3C), strongly suggesting that enhanced IAA export upon over-expression of members of this ABCB subgroup is most likely driven directly by these ABCBs, rather than indirectly by upregulation of or interaction with tobaccoendogenous transporters. In contrast, none of these ABCBs enhanced the export of the diffusion control, benzoic acid (BA) (Appendix Fig S6D). Moreover, overexpression of ABCB17, as a representative of this group of transporters did not result in significantly altered export of putative ABC transporter substrates other than IAA, including indole-3-butyric acid (IBA) (Ruzicka et al, 2010), abscisic acid (ABA) (Kang et al, 2015) and trans-zeatin (tZ) (Zhang et al, 2014), as well as the ABCB14 substrate, malate (Lee et al, 2008) and diffusion control, benzoic acid (BA) (Fig 3D). This apparent high selectivity is especially remarkable for IBA, which differentiates from IAA solely by an extension of 2 C-atoms in the acid moiety.
Conversely, endogenous auxin transport activities were significantly reduced in leaf mesophyll protoplasts of amiR-2572, but not in abcb T-DNA insertion mutants (Fig 3E), without altering BA export (Appendix Fig S6E). This indicates that the ABCBs that are targeted by amiR-2572 contribute to auxin transport in leaf mesophyll cells.
Jointly, these data highlight ABCB15-22 as a group of ABCBs that stimulate auxin transport.

ABCB15-22 contribute to shootward auxin transport
Previously, we inferred from tissue-specific complementation assays, chemical genetics and in silico modeling that shootward auxin transport in the LRC is a critical determinant of auxin oscillation amplitude (De Rybel et al, 2012;Xuan et al, 2015Xuan et al, , 2016. The defect in auxin oscillation amplitude in amiR-2572, together with the observed IAA transport activities of ABCB15-22 indicates that they could represent the elusive efflux component in this model. To test this hypothesis, we determined shootward auxin transport rates in amiR-2572 roots.
Data information: For (B-D, F, G), One-way ANOVA in combination with Tukey's multiple comparisons test, significant differences (P ≤ 0.05) are indicated by different lowercase letters. Central bands in the box plots show the medians; box limits indicate the 25 th and 75 th percentiles as determined by R software; whiskers extend 1.5 times the interquartile range from the 25 th and 75 th percentiles, outliers are represented by dots. Source data are available online for this figure. Indeed, local, exogenous application of radiolabeled IAA to the root tip, revealed a significant reduction in shootward auxin transport in amiR-2572 roots ( Fig 4A). In a complementary approach, we monitored auxin dynamics in the meristem with cellular resolution, using a recently developed estradiol-inducible auxin biosynthesis system that can be activated specifically in the quiescence center (QC) (pWOX5:XVE>>YUC1-2A-TAA1) (Hu et al, 2021). Simultaneous expression of YUC1 and TAA1 results in IAA synthesis from tryptophan Won et al, 2011). In WT (Col-0), estradiol treatment (5 lM) induced a strong ectopic DR5:VENUS expression in the LRC, epidermis and the stele in the elongation zone within 7.5 h. These effects were further enhanced after 9 h estradiol treatment (Fig 4B and C). This is consistent with auxin, which was produced in the QC, being transported via the LRC and epidermis towards the tissues of the elongation zone, where it activates DR5:VENUS expression. In estradiol-treated pWOX5: XVE>>YUC1-2A-TAA1 x amiR-2572, the induction of DR5:VENUS in the elongation zone was at both time-points severely reduced compared to the pWOX5:XVE>>YUC1-2A-TAA1 control (Fig 4B and  C). Additionally, we noted a delay in the emergence of lateral roots in induced pWOX5:XVE>>YUC1-2A-TAA1 x amiR-2572 (Fig 4D), which is consistent with a role in the shootward auxin transport mechanism that contributes to LR formation.  Fig S7B). ABCB18 was almost not expressed in the root meristem, but showed very weak expression in vascular tissues of the hypocotyl and mature root tissues (Appendix Fig S7C). Confocal microscopy and histological sections revealed that the root meristematic expression of ABCB15, B16, B17, and B22 was largely restricted to the epidermis and/or LRC (Fig 5A-C). The expression patterns of the ABCB15-22 cluster define a network with overlapping and/or complementary expression across the outer layers of the root, reflecting that they could exert functionally redundant and/or cooperative functions in these tissues.
The expression pattern of the ABCB15-22 members in the outer layers of the root overlaps with those of ABCB4 and the ABCB1/B19 pair, suggesting they could act in the same pathway. Therefore, we crossed amiR-2572 to abcb4 and abcb1abcb19 double mutants. The amiR-2572 LR and root length phenotypes were epistatic over abcb4 and abcb1abcb19 root phenotypes (Appendix Fig S8A-F), suggesting non-overlapping functions within these ABCBs in the regulation of LR density, despite the significant overlap of their expression domains in the root meristem.
Together with the expression patterns and phenotypes, these data are consistent with the ABCB15-22 cluster contributing to shootward auxin transport mechanism in the outer layers of the meristem that contributes to the DR5:LUC oscillation amplitude and LR density.

ABCB15-22 activities in the outer layers of the root are involved in PBS formation
The amiR-2572 line had prominent seedling phenotypes, not only in the shoot, but also had a reduced root meristem size (Fig 2C and D), which complexifies the interpretation of its LR phenotype. To specifically assess the contribution of ABCB15-22 activities in the outer tissues of the root to LR density, we pursued a tissue-specific silencing strategy. Therefore, we used synthetic trans-acting smallinterfering RNAs (syn-tasiRNAs) in the AtTAS1c backbone (Carbonell et al, 2014), that are predicted to target ABCB15, 16, 17, 18, and 22 (syn-tasi-1522). Two independent and distinct syn-tasiRNAs, indicated as "syn-tasi-1522A" and "syn-tasi-1522B" (Dataset EV1; Appendix Fig S3A), were expressed in the outer layers of the root (LRC, epidermis, cortex) via the PIN2 promoter (Marques-Bueno et al, 2016). For both syn-tasi-1522A and syn-tasi-1522B, we characterized two independent lines. Each line displayed significant reductions in LR density, but not in root length (Appendix Fig S9A-F).
Ó 2023 The Authors EMBO reports 24: e56271 | 2023 silencing activity was largely restricted to the outer layers of the root and was more efficient in these outer root meristem tissues than in amiR-2572 (Fig 6A-D; Appendix Fig S10). The absence of silencing in the leaf of syn-tasi-1522A#1, in comparison to amiR-2572 ( Fig 6A  and E), matched a lack of auxin transport defects in leaf mesophyll cells (Fig 3E, Appendix Fig S6E). Therefore, syn-tasi-1522A#1 phenotypes should mainly derive from gene silencing in the outer tissues of the root. The lack of overt root length, meristem or LRC defects in syn-tasi-1522A#1 (Fig 2C-F) compared to amiR-2572, therefore demonstrates that ABCB15-22 have additional roles in overall plant growth and development. Importantly, a significant reduction in LR initiation (Appendix Fig S9G and H), prebranch site formation (Fig 6F), and reduced DR5:LUC oscillation amplitude (Fig 6G and H) could still be detected. These data suggest that the network of ABCB15-22 activities in the outer layers of the root modulates prebranch site formation via contributions to auxin oscillation amplitude.
Recently, the repertoire of auxin-transporting ABCB was further expanded with the ABCB6/20 pair (Zhang et al, 2018), raising the question if more ABCBs could be classified as auxin transporters. Among all thus far characterized auxin-transporting ABCBs, a conserved D/E-P motif was identified that was not only essential for auxin transport activities, but was also sufficient to introduce a significant auxin transport capacity to the malate-transporting ABCB14 (Hao et al, 2020). ABCB15, 16, 17, 18, and 22 contained this motif, highlighting them as putative auxin transporters. Indeed, we found that overexpression of these ABCBs increases the auxin efflux from tobacco protoplasts and that this activity depends on the presence of an intact D/E-P motif that is was proposed as diagnostic for auxin transporting capacity in ABCBs (Hao et al, 2020). Moreover, the amiR-2572 displayed defects in shootward auxin transport in the root. To control artifacts related to developmental changes in amiR-2572, we used tissue-specific silencing lines. The syn-tasi-522A#1 displayed root clock amplitude and PBS defects similar to those observed in amiR-2572. Using transient overexpression assays, we found increased IAA export from plant cells, while we did not detect a change in the export of the structurally related IBA, malate or other structurally unrelated molecules, suggesting that this group of five ABCBs have a high IAA specificity compared to the PLEIOTROPIC DRUG RESISTANCE (PDR) subclade of ABCGs. In example ABCG36 has been implicated in the transport of auxin analogs (2,4-D) and precursors (IBA), auxin transport inhibitor (NPA), but could not transport IAA (Ito & Gray, 2006;Ruzicka et al, 2010). While we cannot exclude that this group of five ABCBs have additional substrates, our data could only detect effects on IAA transport, suggesting specificity. More work is needed to establish the auxin transport characteristics of these ABCBs, for instance via biochemical assays and heterologous systems. The availability of 11 auxin-transporting ABCBs implies a large potential for functional Data information: For (C-E), One-way ANOVA in combination with Tukey's multiple comparisons tests; significant differences (P ≤ 0.01) are indicated by different lowercase letters; For (F-H), Unpaired two-tailed Student's t-test with Welch's correction, P < 0.001 (***). Central bands in the box plots show the medians; box limits indicate the 25 th and 75 th percentiles as determined by R software; whiskers extend 1.5 times the interquartile range from the 25 th and 75 th percentiles, outliers are represented by dots. Source data are available online for this figure. redundancy, making it difficult to uncover the contributions of ABCB-mediated auxin transport to developmental processes. This is illustrated by the appearance of increasingly severe phenotypes in double and higher order mutants in ABCB1, 4, 6, 19, and 20 (Geisler et al, 2003;Zhang et al, 2018;Jenness et al, 2022). Similarly, we found increasingly penetrant phenotypes with increasing numbers of ABCB15-22 genes being mutated. While most of them are expressed in the same region, their expression pattern is not fully overlapping, and in some case mainly complementary, resulting in a complex genetic interaction. A full dissection of the genetic interactions among these ABCBs will require additional mutant combinations, tissue-specific complementation, and modeling.
Interestingly, all auxin-transporting ABCBs characterized to date have been shown to be expressed in roots and seem to contribute to auxin transport in the root. Recent models of auxin transport in the root meristem are based only on ABCB1, 4 and 19, and could be used to simulate realistic auxin distribution patterns (Mellor et al, 2022). However, the auxin distribution changes predicted by simulating the abcb4 mutant did not match the experimental findings, a finding that was proposed to be attributable to uncharacterized ABCBs. Our analyses show that ABCB15-22 are expressed in tissues of the root meristem overlapping with the modeled ABCB4 activity and seem contribute to shootward auxin transport. Therefore, it will be interesting to evaluate these ABCBs to improve models of auxin transport in the root.
At its core, the spacing of LRs in Arabidopsis is determined by prepatterning along the root that is instructed by the periodic activation of auxin signaling in the pericycle (De Smet et al, 2007;Moreno-Risueno et al, 2010;Xuan et al, 2015). This model of LR prepatterning assumes a local build-up or oscillation of auxin that triggers LR initiation when an auxin signaling threshold is surpassed (Xuan et al, 2015).
Two models based on the reverse fountain auxin transport model have been proposed to explain this oscillation in auxin signaling. In the first model, periodic cell death of the lateral root cap releases auxin into the shootward auxin flow, resulting in a peak of auxin in the stele tissues (Xuan et al, 2016). Consequently, the rate of LRC cell death set the frequency of the oscillation, and the auxin transport rates set the amplitude of the oscillation. In the reflux-and-growth model, cell division and elongation dynamics in a growing root model automatically generate auxin oscillations (van den Berg et al, 2021). In this model, cell division rates and meristem size set the oscillation frequency, while elongation rates determine the priming amplitude. According to the reflux-and-growth model, the smaller meristem in amiR-2572 should translate into a reduced oscillation frequency and downstream a reduced LR density. Instead, we found that oscillation frequency was unaffected in amiR-2572 and that this frequency did match with LRC cell death rates. In this background, LRC cell death rates thus seem to have a greater contribution to the auxin oscillation frequency than meristem size and proliferation rates.
The pWOX5:XVE>>YUC1-2A-TAA1 construct was generated by cloning the YUC1-2A-TAA1 cassette into XhoI and SpeI sites of the pER8 vector (Zuo et al, 2000). The full-length cDNA of YUC1 was cloned into the BamHI site and the full-length cDNA of TAA1 into the BglII site of the pM2A vector containing 2A peptides (Kim et al, 2011). For QC-specific activation of the YUC1-2A-TAA1 cassette, the genomic DNA of WOX5 promoter (WOX5pF: CAATATATCCTGTCAAACaaagacttttatctaccaacttcaa; WOX5pR: GCCGTTAACGCTTTCATcgttcagatgtaaagtcctcaactgt) was used.

Agrobacterium and Arabidopsis transformation
Agrobacterium tumefaciens strain GV3101 was transformed with the relevant binary plasmids via the freeze-thaw procedure (Chen et al, 1994). An individual PCR confirmed Agrobacterium colony was used for floral dip (Clough & Bent, 1998). Transformants were selected, and the segregation of the T2 was analyzed using appropriate antibiotics.

Phenotyping and LR staging
To quantify the LR phenotype in wild-type plants and mutants, emerged LR of whole seedlings were counted under a dissecting microscope, 8 days after germination. Root lengths were measured via Fiji (ImageJ 1.52n) (Schindelin et al, 2015) using digital images obtained by scanning the Petri dishes. To analyze the LR primordium stages, root samples were cleared as described previously (Malamy & Benfey, 1997). All samples were analyzed by differential interference contrast microscopy (Olympus BX51).

Oscillation analysis and prebranch site
The Luciferase imaging of whole seedlings and oscillation expression analysis was performed as described (Xuan et al, 2018). A Lumazon FA imaging system (Nippon Roper) carrying a CCD camera from Princeton Instruments Ltd. (Trenton, NJ, USA) or NightSHADE LB985 in vivo plant imaging system (BERTHOLD TECHNOLOGIES) carrying a deep-cooled slow scan CCD camera from Andor Instruments Ltd. (Belfast, UK) were used for luciferase imaging.
To monitor the pre-branch site numbers, we used 8-day-old DR5: LUC seedlings for pre-branch site quantification. The D-luciferin solution (1 mM) was sprayed gently on the seedlings, kept for 10 min in the dark and imaged in the Lumazon system with a 15min exposure time. Static luminescence signals that were visible along the primary root outside the OZ were counted as pre-branch sites.
To monitor the periodic time and amplitude of DR5 oscillations. Three-day-old DR5:LUC seedlings were transferred to plate spraying with 1 mM D-luciferin solution. The sequential images of the root tip are taken every 15 min with 7 min exposure time. The luciferase signal was quantified by measuring the analog digital units per pixel with the Fiji software. A square region was selected where a prebranch site is formed, and this region should cover DR5 oscillation that occurred prior to pre-branch site formation. The signal intensity changes in this region overall images of the movie were measured (Appendix Fig S7B and C). The difference between the highest value and lowest value of DR5:LUC in the OZ defines the amplitude of DR5 oscillations (Appendix Fig S7C). The period of the DR5 oscillations was determined based on the number of frames that space a DR5:LUC maximum in the OZ (Xuan et al, 2018).

Macroview stereo microscope
To monitor the DR5:VENUS signal over time, an Olympus MVX10 macroview stereo microscope was applied to image the fluorescence signal from vertical growing Arabidopsis roots as described (Xuan et al, 2016). Three-day-old seedlings were imaged every 10 min with 1 s exposure time to visualize the DR5:VENUS stripes. To determine the time between the consecutive disappearances of the nuclear-localized fluorescence signals in the most distal lateral root cap cell files. The time-lapse pictures were saved as tiff files and further analyzed with FIJI software. The number of frames up to the frame with a complete absence of the fluorescence signal in the most distal lateral root cap, were counted. The time between two consecutive events was calculated based on the number of frames counted.

Confocal microscopy
For reporter lines and translational fusion, seedlings were imaged on a Zeiss 710 confocal microscope. For the propidium iodide (PI)treated root images, seedlings were stained with 2 lg/ml PI for 3 min, washed with water, and used for confocal imaging. For root imaging, GFP was excited at 488 nm and acquired at 500-530 nm. YFP was excited at 514 and the emission between 519-564 nm was collected for YFP and between 614-735 nm for PI. Confocal settings were kept constant between WT (Col-0), syn-tasi-1522A#1 and amiR-2572 F1 progeny with either pro35S:YFP-ABCB15, B16, B17, B18, and B22. Fluorescence intensities were measured by FIJI software (https://imagej.net/Fiji). The average intensity was measured in a fixed-size dashed line box for all seedlings by the "rectangle tool".
For the pWOX5:XVE>>YUC1-2A-TAA1 experiments seeds were sown on MS plates, stratified at 4°C for 2 days, and grown vertically in the growth chamber for 4 days at 21°C. Four-day-old seedlings of the pWOX5:XVE>>YUC1-2A-TAA1, DR5:VENUS in Col-0 and amiR-2572 background were treated with 5 lM estradiol for the indicated time-points. Seedlings were stained in 10 mg/l propidium iodide for 2 min and rinsed in water for 30 s. Confocal microscopy was performed using a Zeiss LSM780 inverted confocal microscope equipped with a 20×/0.8 M27 objective lens. VENUS and propidium iodide were excited using an argon-ion laser and a diode laser, respectively. VENUS was excited at 514 nm and detected at 518-588 nm, and propidium iodide was excited at 561 nm and detected at 588-718 nm.

GUS staining and root sectioning
The GUS assay was performed as previously described (Beeckman & Viane, 2000). For Arabidopsis cross-section root specimens, GUSstained seedlings were subjected to fixation, dehydration, and embedding as previously described (De Smet et al, 2004). GUS-stained tissues were imaged using a Leica Bino, Olympus BX51 microscope, and a Keyence VHX-7000 microscope.
Genotyping T-DNA lines for the ABCB single mutants were ordered from The Arabidopsis Information Resource (https://www.arabidopsis.org/), and genotyping primers for T-DNA insertion were designed using the T-DNA Primer Design Tool powered by Genome Express Browser Server (GEBD) (http://signal.salk.edu/tdnaprimers.2. html). Homozygous mutants were selected by PCR performed with primers listed in Appendix Table S1.

RNA extraction and RT-Q-PCR
Total RNA was extracted with the ReliaPrep TM RNA tissue Miniprep System (Promega) from plants grown in vitro. cDNA synthesis was performed with a random and Oligo-d(T) primer mix (Quanta qScript cDNA SuperMix). RT-qPCR was done on a LightCycler 480 (Roche Diagnostics) on 384-well plates with LightCycler 480 SYBR Green I Master (Roche). qPCR data were processed and analyzed with qbase + software. qPCR was performed with primers listed in Appendix Table S1.
For validating the integrity of exported IAA, [2,4,5,6,7-2 H5]IAA (D5-IAA) was used in vector control tobacco export experiments as described above. Export supernatants were acidified using formic acid to pH < 3 and desalted on self-packed C 18 -SPE columns. After elution with 0.1% formic acid in acetonitrile, samples were dried and resuspended in HPLC mobile phase. LC-MS measurements were performed on a QExactive Plus mass spectrometer (Thermo Fisher) coupled to an EasyLC 1000 nanoflow-HPLC. The mass spectrometer was operated in positive ion mode (ESI) with an electron spray voltage of 2.3 kV at 250°C of the heated capillary temperature. Fragmentation was induced by a normalized collision energy of 30%.

Box plots
Box plots were generated by a web tool BoxPlotR (http://shiny. chemgrid.org/boxplotr/) (Spitzer et al, 2014), center lines show the medians; box limits indicate the 25 th and 75 th percentiles as determined by R software; whiskers extend 1.5 times the interquartile range from the 25 th and 75 th percentiles, outliers are represented by dots.

Experimental study design and statistics
No statistical methods were used to estimate the sample size. No blinding was done.

Data availability
This study includes no data deposited in external repositories.
Expanded View for this article is available online.