Patterning of Leaf Vein Networks by Convergent Auxin Transport Pathways

The formation of leaf vein patterns has fascinated biologists for centuries. Transport of the plant signal auxin has long been implicated in vein patterning, but molecular details have remained unclear. Varied evidence suggests a central role for the plasma-membrane (PM)-localized PIN-FORMED1 (PIN1) intercellular auxin transporter of Arabidopsis thaliana in auxin-transport-dependent vein patterning. However, in contrast to the severe vein-pattern defects induced by auxin transport inhibitors, pin1 mutant leaves have only mild vein-pattern defects. These defects have been interpreted as evidence of redundancy between PIN1 and the other four PM-localized PIN proteins in vein patterning, redundancy that underlies many developmental processes. By contrast, we show here that vein patterning in the Arabidopsis leaf is controlled by two distinct and convergent auxin-transport pathways: intercellular auxin transport mediated by PM-localized PIN1 and intracellular auxin transport mediated by the evolutionarily older, endoplasmic-reticulum-localized PIN6, PIN8, and PIN5. PIN6 and PIN8 are expressed, as PIN1 and PIN5, at sites of vein formation. pin6 synthetically enhances pin1 vein-pattern defects, and pin8 quantitatively enhances pin1pin6 vein-pattern defects. Function of PIN6 is necessary, redundantly with that of PIN8, and sufficient to control auxin response levels, PIN1 expression, and vein network formation; and the vein pattern defects induced by ectopic PIN6 expression are mimicked by ectopic PIN8 expression. Finally, vein patterning functions of PIN6 and PIN8 are antagonized by PIN5 function. Our data define a new level of control of vein patterning, one with repercussions on other patterning processes in the plant, and suggest a mechanism to select cell files specialized for vascular function that predates evolution of PM-localized PIN proteins.


Introduction
Branched structures pervade all levels of organization in living organisms, from molecules to organelles, cells, multicellular organs and entire organisms, and the principles that guide the formation of these branched structures have long been object of interest of biologists and mathematicians. However, few branched structures have historically captured more widespread attention than the vein networks of plant leaves. From a developmental standpoint, such attention seems justified as files of vein precursor cells are selected from within a population of seemingly identical cells [1,2]. Furthermore, in most species the product of this patterning process is both reproducible and variable: reproducible as vein networks supply all areas of the leaf; variable as the exact sites of vein formation are unpredictable [3]. These observations argue against rigid specification of leaf vein patterns and suggest a self-organizing control mechanism that reconciles the plasticity of vein formation with the stringent requirement for organ vascularization [4].
Though the identity of the molecules involved is largely unknown, varied evidence supports a decisive role for the polar, cell-to-cell transport of the plant signal auxin in leaf vein patterning: auxin application induces formation of new veins oriented towards pre-existing veins [3]; the inductive and orienting effect of applied auxin on vein formation is suppressed by auxin transport inhibitors [5]; and auxin transport inhibitors induce characteristic defects in vein patterns [6,7]. During leaf development, expression of the plasma-membrane (PM)-localized PIN-FORMED1 (PIN1) auxin transporter of Arabidopsis thaliana [8,9] is initiated in broad tissue domains and becomes gradually restricted to single files of vascular precursor cells [10,11]. Narrowing of PIN1 expression domains is associated with polarization of PIN1 subcellular localization to the presumed auxin-efflux side of PIN1expressing cells. Auxin levels define the initial size of PIN1 expression domains, and both domain narrowing and PIN1 polarization are obstructed by auxin transport inhibitors. These observations agree with conceptual formulation [12,13] and mathematical modeling (reviewed in [14][15][16][17][18]) of progressive restriction of nondirectional auxin transport across tissues to polar transport in single files of vascular precursor cells by positive feedback between auxin flow through a cell and the cell's auxin conductivity. However, in contrast to the severe vein-pattern defects induced by auxin transport inhibitors [6,7], vein pattern defects in pin1 mutant leaves are mild [6,19,20]. PIN1 is member of a family comprising four other PM-localized PIN proteins and three, evolutionary older, endoplasmic reticulum (ER)-localized PIN proteins [21][22][23][24]. Thus, the mild vein-pattern defects of pin1 have been attributed to redundancy among PM-localized PIN proteins (e.g., [6,10,19,25]), redundancy that has been shown to underlie, to different extents, many other developmental processes ( [20,26,27] and references therein).
Here we show that the vein network of the Arabidopsis leaf is patterned by two distinct and convergent auxin transport pathways: intercellular auxin transport mediated by PM-localized PIN1 and intracellular auxin transport mediated by ER-localized PIN6, PIN8 and PIN5. Our results suggest an ancestral auxintransport-dependent mechanism to specify cell files to vascular function that predates evolution of PM-localized PIN proteins.

Results
Vein patterning functions of Arabidopsis PIN genes WT Arabidopsis grown under normal conditions forms separate leaves, whose vein patterns are defined by reproducible features [28,29] (Figure 1A and 1B): a narrow, central midvein that runs the length of the leaf; lateral veins that branch from the midvein and join distal veins to form closed loops; minor veins that branch from midvein and loops and either end freely or join other veins; and minor veins and loops that curve near the leaf margin, lending a scalloped outline to the vein network. By contrast, WT leaves developed in the presence of auxin transport inhibitors often show separation defects ('fused leaves'), and vein patterns of auxintransport-inhibited leaves deviate from the norm in at least four respects [6,7]. First, the midvein bifurcates near the leaf tip. Second, the vein network consists of more lateral veins. Third, lateral veins do not join the midvein at the centre of the leaf but run parallel to one another to form a wider midvein. Fourth, lateral veins end in a marginal vein that closely parallels the leaf margin, lending a smooth outline to the vein network. Mutation of PIN1 (AT1G73590) approximates these defects, but vein patterns are only mildly affected in pin1 [6,19] (Figure 1). As PIN1 is one of eight PIN genes [21][22][23][24] and other gene families have been implicated in auxin transport (e.g., [30][31][32]), the mild vein-pattern defects of pin1 might reflect contribution of other auxin transporters to vein patterning; here we tested whether other PIN genes contribute to vein patterning.
A PIN6 translational fusion localizes to the ER in tobacco suspension cells [36]. To determine PIN6 subcellular localization in Arabidopsis leaves, we quantified degree of colocalization between expression of PIN6::PIN6:GFP and expression of ER, or PM, markers. Expression of PIN6::PIN6:GFP correlated with expression of ER, but not PM, markers (Figure 2M-2T and Figure  S3), suggesting ER-localization of PIN6 in vein development.

Author Summary
The beautiful and varied patterns of veins in plant leaves have intrigued both artists and scientists since antiquity. The seminal work of some of these scientists has shown that the plant hormone auxin and its transport in plant tissues control leaf vein patterning, but molecular details of auxin-transport-dependent vein patterning are still unknown. Here we find that vein patterning is controlled by the concerted action of two spatially separate auxin transport pathways. One pathway transports auxin from cell to cell, the other to and from different compartments within the cell. The cell-to-cell pathway of auxin transport seems to exist only in land plants with veins. The withincell pathway of auxin transport seems to have appeared earlier in the evolution of plants and exists also in primitive land plants, which form vein-like structures. Our findings suggest an unsuspected level of control of vein patterning, one that regulates patterning of plant features beyond veins and that could even be relevant for the formation of vein-like structures of primitive land plants.
We next asked whether pin6;8 defects in vein network formation were associated with changes in auxin response levels. To address this question, we imaged expression of the DR5rev::YFPnuc auxin response reporter [44]. In subepidermal tissues of WT leaves, DR5rev::YFPnuc was strongly expressed at sites of vein formation ( Figure 5E); in subepidermal tissues of pin6;8 leaves, DR5rev::YFPnuc expression was weaker ( Figure 5F), a defect that was normalized by pin5 ( Figure 5G).
We then asked whether pin6;8 defects in auxin response levels and vein network formation were associated with changes in auxin-responsive PIN1 expression [10,[44][45][46]. To address this question, we imaged expression of PIN1::PIN1:YFP [47] in the leaf area enclosed by the first loop. In both WT and pin6;8,  Figure 5N).
In summary, pin6;8 had defects in expression of DR5 and PIN1 and in formation of vein networks, and such defects were normalized by pin5.
We then asked whether ectopic expression of PIN8 or PIN5 had any effect on vein network formation. To address this question, we analyzed vein patterns of plants expressing the cDNA of PIN8 or PIN5 by the MP promoter (MP::PIN8 or MP::PIN5, respectively). We found that MP::PIN8 leaves had simpler, open vein networks and that MP::PIN5 leaves had a more complex vein network ( Figure 6I and 6N-6Q).
In summary, ectopic PIN6 expression was sufficient to control expression of DR5 and PIN1. Furthermore, ectopic expression of PIN6 or PIN8, on the one hand, and of PIN5, on the other, had opposite effects on vein network formation.

Simulation of pin1;6 defects by reduction of auxin levels in pin1
Weaker expression of auxin responsive markers in pin6;8 leaves and stronger expression of auxin responsive markers in leaves of RPS5A::PIN6 and MP::PIN6 suggest that PIN6-mediated auxin  transport increases intracellular levels of biologically active auxin (see Discussion). If this were the case, pin1;6 phenotype spectrum might be mimicked by expressing in pin1 the bacterial gene iaaL, which decreases levels of free auxin by its conjugation to lysine [50,51]. Expression of PIN6::iaaL (PIN6 promoter driving expression of iaaL) in pin1 mimicked the phenotype spectrum of pin1;6 leaves ( Figure 7) and cotyledons ( Figure S1), a finding that is consistent with the hypothesis that PIN6-mediated intracellular auxin transport increases auxin levels.

Discussion
The mechanisms that control the patterning of vein networks of plant leaves have long fascinated biologists and mathematicians. Varied evidence has increasingly been accumulating that supports an inductive and orienting role for the transport of the plant signal auxin in leaf vein patterning [3,5-7,10-13], but molecular details have remained unclear. Here we show that the vein network of the Arabidopsis leaf is patterned by two distinct and convergent auxintransport pathways: an intercellular pathway mediated at the plasma membrane (PM) by the PIN1 auxin transporter and an intracellular pathway mediated at the endoplasmic reticulum (ER) by the PIN6, PIN8 and PIN5 auxin transporters. While a role for other families of auxin transporters (e.g., [30][31][32]) in vein patterning is by no means precluded, our results suggest a new, unsuspected level of control of vein patterning by PIN-mediated auxin transport, one with repercussions on patterning of other plant features.
The localization of PIN1 to the PM [8] and of PIN6 to the ER together with the appearance in pin1;6 of vein pattern defects that exceed the sum of the single-mutant defects suggest that PIN1 and PIN6 act in distinct auxin-transport pathways whose functions converge on vein patterning. By contrast, the overlapping subcellular localizations of PIN6 and PIN8 [36][37][38][39] together with the purely quantitative enhancement of pin1;6 vein pattern defects by pin8 suggest redundant function of PIN6 and PIN8 in PIN1dependent vein patterning. Further, the overlapping subcellular localization of PIN8 [36][37][38][39] and PIN5 [36,37,39] together with the partial normalization of pin1;6;8 vein pattern defects by pin5 suggest antagonistic functions of PIN8 and PIN5 in PIN1/PIN6dependent vein patterning. The interaction between the intercellular auxin-transport pathway controlled by PIN1 and the intracellular auxin-transport pathway controlled by PIN6, PIN8 and PIN5 is relevant for at least two other patterning events, separation of cotyledons and of leaves, and might have implications for other developmental processes. The redundancy between PIN6 and PIN8 and the antagonism between PIN8 and PIN5 are, however, independent of PIN1-mediated intercellular auxin transport: in the presence of PIN1 function, PIN6 and PIN8 redundantly control vein network formation, and such redundant functions are antagonized by PIN5. Redundant functions of PIN6 and PIN8 and antagonistic functions of PIN5, as inferred from loss-of-function data, are also consistent with gain-of-function evidence: ectopic PIN8 expression induces vein pattern defects similar to those induced by ectopic PIN6 expression and opposite to those of pin6;8, suggesting that PIN8 can supply vein patterning functions similar to those of PIN6; and ectopic PIN5 expression induces vein pattern defects opposite to those induced by ectopic expression of PIN6 or PIN8 and similar to those of pin6;8, suggesting that PIN5 can supply vein patterning functions opposite to those of PIN6 and PIN8. The internally consistent relationship between the effects of loss of function of PIN6, PIN8 and PIN5 and the effects of gain of function of these genes is not limited to vein patterning but extends to auxin response: auxin response levels are decreased by simultaneous mutation of PIN6 and PIN8 and by ectopic expression of PIN5 [36] and increased by ectopic expression of PIN6 or PIN8 [39] and by mutation of PIN5 [36]. Thus, the genetic interaction between PIN6, PIN8 and PIN5 might reflect general properties of the mechanism with which these proteins function, a conclusion that is also supported by the antagonistic functions of PIN8 and PIN5 in a process unrelated to vein patterning such as pollen development [39]. The most parsimonious explanation for the antagonistic interaction between PIN6/PIN8 and PIN5, which transports auxin from the cytoplasm to the ER lumen [36], is that PIN6 and PIN8 transport auxin from the ER lumen to the cytoplasm or to the nucleus, whose envelope is continuous with the ER membrane [52,53] (Figure 8C). This scenario is also supported by the observation that reducing auxin levels in pin1 leads to pin1;6 characteristic defects, and is consistent with higher levels of auxin measured in PIN8 overexpressors [38,39] and pin5 mutants [36]. Alternatively, or in addition, PIN6, PIN8 and PIN5 could all transport in the same direction but have different affinities for auxins or auxin conjugates with different, even opposing, developmental functions (reviewed in [54]) ( Figure 8D).
Our data suggest that auxin transported by PIN6 and PIN8 increases levels of auxin-responsive PIN1 expression during vein formation ( Figure 8A). In pin6;8, low auxin levels at sites of vein formation could accelerate formation of narrow vein-associated domains of PIN1 expression, thus resulting in high-complexity vein networks [13,[55][56][57][58]. Conversely, in leaves ectopically expressing PIN6, high auxin levels at sites of vein formation could hinder or prevent formation and connection of vein-associated PIN1-expression domains, thus resulting in simple, open vein networks [13,56,57,[59][60][61]. Alternatively, PIN6/PIN8-mediated auxin transport could control vein network formation exclusively through PIN1-independent pathways; defects in auxin response or PIN1 expression would then be consequence, rather than cause, of defective vein-network formation ( Figure 8B). In any case, the synthetic enhancement of pin1 defects by pin6 and pin6;8, as opposed to the epistasis of pin6;8 to pin1, suggests vein patterning functions of PIN6/PIN8-mediated intracellular auxin transport beyond regulation of PIN1 expression or of intercellular auxin transport mediated by PM-localized PIN proteins ( Figure 8A and 8B). We can probably exclude that such functions of PIN6 and PIN8 require physical interaction with PIN1 because PIN1 and PIN6/PIN8 localize to different cellular compartments and because PIN1/pin1;pin6/pin6 lacks defects in cotyledons and vein patterns (n.50), defects that would be expected for interacting proteins (reviewed in [62]). Because the localization of PIN6 and PIN8 is maintained in pin1 backgrounds, it seems unlikely that these two proteins have auxin transport functions that are redundant and homologous with those of PIN1. Nevertheless, PIN6/PIN8/PIN5mediated intracellular auxin-transport could contribute to intercellular auxin transport by a mechanism nonhomologous to that used by PMlocalized PIN proteins [63] (Figure 8A and 8B), a function of ERlocalized PIN proteins that would be particularly relevant in primitive land plants like mosses, which seem to lack PM-localized PIN proteins [36] but do select cell files for vascular function [64].

Plants
Origin and nature of lines, genotyping strategies and oligonucleotide sequences are in Tables S1, S2, S3, respectively. Seeds were sterilized and germinated, and plants were grown and transformed as in [65].

Imaging
For plasma-membrane stainings, seedlings were incubated in 10 mg ml 21 FM 4-64 (Invitrogen) for ,4 min under vacuum before mounting. For endoplasmic-reticulum stainings, dissected leaves were incubated in 10 mM ER-Tracker Red (Invitrogen/ Life Technologies) or ER-Tracker Blue-White DPX (Invitrogen/Life Technologies) for ,30 min under vacuum before mounting. Leaf primordia were mounted in water under 0.17mm-thick coverslips (Fischer Scientific) and imaged with the 206/0.8 Plan-Apochromat or 406/1.2 W C-Apochromat objective of an Axio Imager.M1/LSM 510 META confocal microscope (Carl Zeiss). 5126512-pixel frames were scanned unidirectionally at 8-bit depth with 1.6-msec pixel dwell time  Figure 1). Difference between pin1 and WT, between pin1;6 and pin1, and between PIN6::iaaL;pin1 and pin1 was significant at P,0.001 (***) by and 8-fold averaging. Scanning zoom was adjusted to set pixel size to half the spacing of the features to be resolved, with minimum pixel size equal to half the objective lateral resolving power. Emission was collected from ,5-mm-thick (singlefluorophore imaging) or ,1-mm-thick (multi-fluorophore imaging) optical slices. Amplifier gain was set at 1; detector gain at ,50-60%. Laser transmission and offset value were adjusted to match signal to detector's input range. Marker-line-specific imaging parameters are in Tables S4 and S5. For multifluorophore imaging, sequential excitation and collection of emission were performed in line-by-line channel-switching mode, which did not result in signal crossover under our imaging conditions. Signal levels in images acquired at identical settings were visualized as in [65]. Signal colocalization was visualized as in [66] and quantified as in [67]. Mature leaves were imaged as in [49]. Images were analyzed and processed with ImageJ (National Institutes of Health), and figures were generated with Canvas (ACD Systems International Inc.). Figure S1 Cotyledon patterns of pin mutants. (A-G) Dark-field illumination of 4-day-old seedlings illustrating phenotype classes: two separate cotyledons (A); fused cotyledons and separate single cotyledon (B); three separate cotyledons (C); fused cotyledons (D); single cotyledon (E); partially fused cup-shaped cotyledon, side view; inset: top view (F); completely fused cup-shaped cotyledon, side view; inset: top view (G). (H) Percentages of seedlings in phenotype classes. Difference between pin1 and WT, between pin1;6 and pin1, between UBQ10::amiPIN6-10;pin1 and pin1, between pin1;6;8 and pin1;6, and between pin1;5;6;8 and pin1;6;8, and between PIN6::iaaL;pin1 and pin1 was significant at P,0.001 (***) by Kruskal  PIN1-independent functions of PIN5/PIN6/PIN8-mediated intracellular auxin transport in vein patterning could include intercellular auxin transport by a mechanism nonhomologous to that used by PM-localized PIN proteins [63] (grey arrows). (C,D) Localization of PIN1 (blue) at the basal plasma membrane and of PIN5 (orange), PIN6 (green) and PIN8 (yellow) at the endoplasmic reticulum/nuclear envelope (ER/NE) of vascular cells. For simplicity, PIN5, PIN6 and PIN8 are shown to be expressed in the same cell, only PIN6 is shown to be also localized at the NE, and localization of AUX/ LAX [30], ABCB/PGP/MDR [32] and PILS [31] auxin transporters is not shown. Arrows indicate presumed directions of auxin transport. Antagonism between pin5 and pin6;8 in vein patterning could reflect opposite directions in which PIN5 and PIN6/PIN8 transport auxin (C), different abilities of PIN5 and PIN6/PIN8 to transport different auxins or auxin conjugates (orange, green and yellow arrows) (D; for simplicity, the mirror-image scenario, i.e. transport to the ER/NE lumen, is not shown), or varied combinations of the two. See text for additional details.