Research Article

Integration of growth and patterning during vascular tissue formation in Arabidopsis

See allHide authors and affiliations

Science  08 Aug 2014:
Vol. 345, Issue 6197, 1255215
DOI: 10.1126/science.1255215

How to plumb the root is the problem

The vascular system of the plant root is generated from four seemingly similar cells. At some point, though, these cells' decendents need to follow different fates. Combining computational modeling with manipulation of hormone signaling in Arabidopsis, De Rybel et al. discovered the importance of a small bridge connecting two of the four cells (see the Perspective by Mellor and Bishopp). This feature locked in asymmetry in hormone signaling, so that those cells closest to the xylem delivered the maximal response. Two feedback loops in the model function distinctly, with one generating a domain rich in auxin and the other establishing a sharp boundary between domains.

Science, this issue p. 10.1126/science.1255215; see also p. 622


Coordination of cell division and pattern formation is central to tissue and organ development, particularly in plants where walls prevent cell migration. Auxin and cytokinin are both critical for division and patterning, but it is unknown how these hormones converge upon tissue development. We identify a genetic network that reinforces an early embryonic bias in auxin distribution to create a local, nonresponding cytokinin source within the root vascular tissue. Experimental and theoretical evidence shows that these cells act as a tissue organizer by positioning the domain of oriented cell divisions. We further demonstrate that the auxin-cytokinin interaction acts as a spatial incoherent feed-forward loop, which is essential to generate distinct hormonal response zones, thus establishing a stable pattern within a growing vascular tissue.

Because plant cells cannot migrate during development, control of cell division orientation and simultaneous tissue patterning during early growth is vital to create a functional three-dimensional (3D) structure. The basic body plan of the plant is determined during early embryogenesis. The root vascular tissues develop from four provascular initial cells (Fig. 1A) that undergo several rounds of oriented, periclinal cell divisions to create a patterned vascular bundle by the end of embryogenesis (1, 2). The vascular bundle then contains xylem cells marked by high auxin signaling and flanking zones of procambium cells characterized by high cytokinin (CK) signaling (3) from which the phloem tissues will differentiate post-embryonically. Although growth through oriented cell divisions and pattern formation are thus crucial for normal development, it is yet unknown how these intertwined processes are regulated and if this occurs through distinct or overlapping pathways. Here, we combined experimental and theoretical approaches to unravel the integration of vascular growth and patterning during Arabidopsis embryogenesis.

Fig. 1

LOG4 is a direct target of the TMO5/LHW dimer. (A) Growth (blue) and patterning (red: cambium; green: xylem) occur simultaneously in vascular tissue during embryogenesis. (B) Combinatorial microarray analysis, identifying LOG4 as putative target gene of TMO5. (C) ChIP-qPCR experiment shows direct binding of TMO5 and LHW fusion proteins close to a G-box (−326 bp) of the LOG4 promoter. (D) Relative qRT-PCR expression of LOG4 in mutants or TMO5/LHW-OE compared to wild type (WT) (Col-0). (E) Relative expression levels (qRT-PCR) of TMO5 and LOG4 in roots upon 1 μM 2,4-dichlorophenoxyacetic acid (2,4-D) treatment for the indicated time. Error bars in (C) to (E) indicate SE.

LOG4 is a direct TMO5/LHW target gene

The TARGET OF MONOPTEROS5/LONESOME HIGHWAY (TMO5/LHW) basic helix-loop-helix (bHLH) transcription factor dimer is a rate-limiting regulator of periclinal cell divisions (4). Mutants show a reduction in vascular tissue size due to loss of periclinal division frequency, whereas ectopic coexpression of TMO5 and LHW can trigger this type of division in any cell type of the root (4). The expression domains of TMO5 and LHW overlap in young xylem cells that strongly correlate with the zone in which periclinal divisions occur. However, xylem cells do not themselves divide periclinally, which suggests that the TMO5/LHW dimer promotes these divisions by inductive signaling toward neighboring procambium cells. Indeed, TMO5/LHW misexpression induces excess divisions both cell-autonomously and non–cell-autonomously, suggesting that these transcription factors trigger periclinal cell division through a yet unknown diffusible signal (4).

Given that TMO5/LHW is a transcription complex, we sought to identify this unknown signal, or components of its biogenesis, through determining the direct transcriptional targets. We devised a combinatorial transcript profiling approach using independent, rapidly inducible TMO5 versions (see Materials and Methods for detailed information). Only three genes were significantly induced by TMO5 in all independent experiments (P < 0.05, fold change >1.5) (Fig. 1B and table S1), and their regulation by TMO5 was confirmed by quantitative reverse transcription–polymerase chain reaction (qRT-PCR) (fig. S1). One of these genes encodes LONELY GUY 4 (LOG4), an enzyme involved in the final biosynthesis step of the plant hormone CK (5, 6). Given the previously established importance of CK in root vascular tissue patterning (7), we focused our attention on this gene. Chromatin immunoprecipitation (ChIP) confirmed direct binding of both TMO5-3GFP (triple green fluorescent protein) and LHW-YFP (yellow fluorescent protein) to the same upstream fragment of the gene (Fig. 1C). This coincided with a predicted bHLH-binding G-box (CACGTG) located 326 base pairs (bp) upstream of the LOG4 start codon (Fig. 1C). LOG4 transcripts were reduced in tmo5 and lhw mutant roots and increased upon TMO5/LHW misexpression (Fig. 1D), suggesting that TMO5/LHW are indeed required for LOG4 gene expression. TMO5 is a direct auxin response gene (8), and as expected, LOG4 was also induced by auxin, but with delayed induction kinetics (Fig. 1E).

TMO5 expression marks the first four vascular founder cells in the globular embryo (8). Consistent with direct regulation by the TMO5/LHW dimer, LOG4 transcript (Fig. 2A and fig. S2) and a pLOG4-n3GFP reporter (Fig. 2, B and C) were expressed in these same cells. Later, LOG4 expression was confined to the TMO5/LHW domain in the xylem (Fig. 2, G and H). Because the LOG4 expression domain was extended in pRPS5A-TMO5/pRPS5A-LHW misexpression (TMO5/LHW-OX) roots (Fig. 2, I and J) and absent in the xylem domain of lhw roots (Fig. 2, K and L), TMO5/LHW is a key transcriptional input. Nonetheless, expression of LOG4 in xylem pole pericycle and xylem pole endodermis cells (Fig. 2H) suggests additional, TMO5-independent post-embryonic input. Together, these data identify LOG4 as a direct target gene of the TMO5/LHW transcription complex.

Fig. 2

LOG4 expression depends on TMO5/LHW. (A to F) In situ hybridization (A and D) and pLOG-n3GFP reporter expression (B, C, E, and F) of LOG4 and LOG3 during embryogenesis. (G to L) Expression of pLOG4-n3GFP reporter in post-embryonic root in WT (G and H), TMO5/LHW-OE (I and J), and lhw mutant (K and L). Inset in (H) represents a 3D stack cross section showing LOG4 expression in false color scale across the entire xylem axis. Asterisks indicate endodermis; arrows indicate xylem. Roots are counterstained with FM4-64. Scale bars, 10 μm.

TMO5/LHW controls CK biosynthesis through LOG4

LOG proteins catalyze a rate-limiting step in CK biosynthesis (5, 6), and thus the TMO5/LHW dimer may act by promoting CK biosynthesis. Indeed, concentrations of several CK species were reduced in tmo5 t5l1 mutant roots, whereas most CK species were increased in TMO5/LHW-OX roots (Fig. 3A and table S2). In line with these elevated CK concentrations, TMO5/LHW-OX plants showed a strong postembryonic shoot phenotype, including ectopic leaf outgrowths (fig. S3) that resembled CK-overproducing plants (9). We next used gene expression reporters to determine if TMO5/LHW-dependent CK biosynthesis generates a transcriptional response. Whereas the synthetic CK response reporter pTCSn-GFP (10) was active only in the root cap and vascular initial cells in wild-type roots (Fig. 3B), it was ectopically activated throughout TMO5/LHW-OX roots (Fig. 3C). Likewise, the CK-repressed pAHP6-GFP reporter (11) was down-regulated upon TMO5/LHW misexpression (Fig. 3, D and E). Additionally, protoxylem differentiation was inhibited in TMO5/LHW-OX and LOG4 misexpression roots (Fig. 3, H and I), indicative of increased CK activity (Fig. 3, F and G) (11). Hence, the TMO5/LHW complex triggers CK biosynthesis via inducing LOG4 expression.

Fig. 3

TMO5/LHW triggers CK biosynthesis. (A) Abundance of CK species in roots of t5t5l1 double mutant and double misexpression lines compared to WT (Col-0) (t test: *P < 0.05, **P < 0.001). (B to E) Expression of the pTCSn-GFP (B and C) and pAHP6-GFP (D and E) reporters in WT and TMO5/LHW-OX root tips. Right images show false color scales of the left image. (F to I) Basic fuchsin-stained roots of WT (Col-0), WT treated with 0.1 μM benzyladenine (BA), TMO5/LHW-OX, and LOG4 misexpression (m, metaxylem; p, protoxylem). Error bars in (A) and (B) indicate SE. Images in (B) to (E) are counterstained with FM4-64. Scale bars, 10 μm.

We next addressed the biological significance of LOG-dependent CK activity for TMO5/LHW function in promoting periclinal division and vascular tissue development. Consistent with a requirement for CK response, excess TMO5-induced periclinal cell division was suppressed in the wol receptor mutant (Fig. 4, A to F) (12). LOG4 single mutants did not show clear vascular defects (fig. S4, A to C), but this gene is member of a family of nine members (LOG1 to LOG9) that has been shown to act redundantly in CK biosynthesis (5, 6). We found that, indeed, CK-dependent pTCSn-GFP expression was absent in vascular precursors in the log1234578 heptuple mutant roots, but that expression could be restored by CK treatment (Fig. 4, G to J). Hence, LOG function is collectively required for CK biosynthesis in vascular tissue, which is reflected in defective embryonic vascular tissue development (fig. S4, A and B) and patterning (fig. S4C) in this mutant. To determine which of the LOG genes may act redundantly with LOG4, we first determined the expression patterns of all LOG genes. Only LOG3, LOG4, and LOG7 were consistently detected in young vascular tissues (Fig. 2, D to F, and fig. S5) (5). We therefore introduced the log3 log4 log7 triple mutant (13) into the TMO5/LHW-OX line and found this to suppress excess periclinal cell division (Fig. 4K). The partial suppression in this triple mutant suggests contributions of other LOG genes because several were up-regulated in a log4 mutant (fig. S4D), probably through CK-dependent repression (fig. S4E). We next tested if CK activation is not only required but also sufficient for TMO5/LHW-dependent periclinal cell division. Mutations in LHW, as well as in TMO5 and TMO5-LIKE1 (T5L1), reduce vascular cell file numbers and lead to a switch from diarch to monarch pattern (4, 14). Treatment of lhw and t5t5l1 mutants with CK was sufficient to increase the number of periclinal divisions, revert to diarch patterns in both mutants, and even rescue cell number in the lhw mutant to wild-type levels (Fig. 4, L to R). In summary, these data suggest that LOG-derived CK is a major contributor to the vascular function of TMO5/LHW.

Fig. 4

CK activation mediates TMO5/LHW activity. (A to F) Mature roots (top) and primary root meristems (bottom) of WT (Col-0), pRPS5A-TMO5-GR, and pRPS5A-TMO5-GR in wol mutant background; grown on 10 μM dexamethasone (DEX). Arrows indicate ectopic periclinal cell division (m, metaxylem; p, protoxylem). (G to J) pTCSn-GFP expression in log1234578 mutant without (G and H) or with (I and J) BA treatment. (H) and (J) are false color images of (G) and (I), respectively (counterstained with FM4-64). (K) Quantification of the distribution of vascular cell file number in roots of the lines indicated. (L) Number of vascular cell files in WT (Col-0), lhw, and t5t5l1 mutant backgrounds upon BA treatment (concentration as indicated). Error bars indicate SE. (M to R) Histologic cross sections of WT (Col-0), lhw, and t5t5l1 mutant roots grown without (control) or with 0.1 μM BA. Arrows indicate phloem poles. Scale bars, 10 μm.

A model of vascular tissue formation

In addition to its function in vascular periclinal cell division, CK is also essential for patterning the vascular tissue into distinct domains, comprising the xylem axis with high auxin signaling and the flanking cambial domains with high CK signaling (3). Thus, auxin-CK interactions appear to underlie both growth and patterning, and a key question is how these are coordinated. Previously, a phloem source of CK was postulated in the postembryonic root (15), but no functional phloem exists before seed germination (1), and recent modeling suggests that the phloem source may not provide positional information (16). We therefore explored whether the auxin-MP-TMO5/LHW-LOG4-CK module could have a dual role and account for both patterning and growth using a computational approach.

In our mathematical model (Fig. 5A), the auxin-MP-TMO5/LHW-LOG4-CK module is represented as a set of ordinary differential equations mapped to a growing cellular grid using the VirtualLeaf software (17) (see Supplementary Mathematical Modeling). Because of the highly linear pathway, MP, TMO5/LHW, and LOG4 intermediates are not explicitly modeled, and thus in the model, auxin directly promotes CK production. The simulation starts with a heart stage embryo template where the four vascular founder cells are surrounded by pericycle and endodermis layers (cortex and epidermis are not included in the model). These surrounding layers do not produce CK and have a fixed growth rate, which is not affected by CK levels. These layers thus do not contribute to the model itself. Besides this, the entire model (Fig. 5A) continuously runs in all cells.

Fig. 5

Connectivity of vascular initial cells. (A) Schematic representation of the identified genetic network (left) in an early heart stage embryo, and simplified ordinary differential equations reflecting these interactions in the model (right). (B to E) 3D reconstructed heart stage embryo showing the association of the connected vascular initial cells to the forming cotyledons. (F to I) pTMO5-n3GFP expression at different locations in the heart stage embryo relative to the location of the cotyledons. Scale bars, 10 μm.

Because CK-dependent cell division (18) is an output of the model, the cellular grid evolves during simulations. Cell growth is modeled using turgor pressure uniformly exerted on all walls of a cell that can counteract, irreversibly expand, or yield in response (fig. S9). Fluctuations around the optimal energy configuration lead to nonidentical energy minimizations paths, which reflects the nondeterministic nature of cell division. Thus, different runs of the same simulation produce similar, but not identical, outputs (fig. S8).

In our model, we incorporate the following assumptions (Fig. 5A; Supplementary Mathematical Modeling), each of which is supported by experimental evidence: (i) Auxin levels in cells are determined by basal synthesis and degradation activity, as well as by passive diffusion and PIN-mediated active transport. (ii) PIN levels are subject to regulation by auxin and CK in addition to fixed synthesis and degradation. Auxin in neighboring cells promotes PIN localization toward the membranes (19). In contrast, CK cell-autonomously inhibits PIN localization at the membrane (20), as has been shown for PIN1, the dominant PIN expressed in embryonic vasculature (21). We thus simplify the redundant PIN gene family (21, 22) by a representative “general PIN” that has PIN1-like properties. (iii) CK levels are promoted by auxin, and further depend on degradation and passive diffusion across membranes. (iv) Even though xylem cells produce CK, only the neighboring cambial cells respond by undergoing periclinal divisions. Thus, the CK-producing tissues do not respond to CK themselves. To capture this, we distinguish between CK and CK response. CK response is inhibited by auxin, as has been shown experimentally (11, 23). (v) CK promotes periclinal division, which is represented in the model by lowering the cell size at which cells undergo division.

In addition, we specified two “source cells” that contain elevated auxin concentration as a consequence of auxin transport from overlying tissues. During vascular tissue initiation, the cotyledon primordia become specified, which generates PIN1 convergence points (24) and hence local auxin sources in the cell layers overlying the vascular initials. As a consequence, auxin is nonuniformly provided to the underlying vascular initials that are in closest proximity to the cotyledon primordia (Fig. 5, B to E). Indeed, 3D reconstructions showed that auxin-dependent pTMO5 (Fig. 5, F to I), pDR5, and pLOG4 expression (fig. S6, A and B) was stronger in the two vascular initials that subtend the cotyledon primordia, and cotyledon number has previously been correlated with vascular tissue patterns (25).

Early geometric constraints bias vascular patterning

We simulated this network (Fig. 5A) and monitored growth and patterning, where auxin accumulation is a proxy for xylem identity and CK response reflects the cambial domain. Because exact values are unknown for most parameters, we performed a survey to find parameter sets for which the model generates a bisymmetric vascular bundle with a central high-auxin domain (Supplementary Mathematical Modeling; fig. S11 and table S4). All following model analysis is based on the identified well-performing parameters sets.

We initially started from a stylized cross section of the embryonic root at early heart stage (1) (Fig. 6A). However, this simple geometry did not lead to a continuous central xylem axis flanked by cambial domains (Fig. 6A). Further exploration of the starting template indicated that the initial geometry strongly biases final model output. Only a configuration in which two of the four founder cells are connected by a small “bridge” (Fig. 6B) yielded stable realistic patterns (Fig. 6, E and F, I and J, and movies S1 and S2). We found that the existence of a bridge is required, but its exact size or orientation did not change model output (Supplementary Mathematical Modeling; fig. S10). We next evaluated the presence of a cellular connection between source cells, and found this geometry in nearly all embryos analyzed in 3D up to globular stage (21 out of 26; Fig. 6D and fig. S6C) and in all embryos at postglobular stages (11 out of 11; Fig. 6D and fig. S6C). Tracing the origin of the bridge revealed that it is a consequence of division planes at the two- to four-cell transition (fig. S6C). Thus, this particular geometry in the center of the embryo follows from the improbability of generating exact four-way junctions (26) during earlier cell divisions, followed by cell expansion (2).

Fig. 6

A growing model of vascular tissue formation. (A to C) Different initial geometries tested in the model. Left: four-way junction; middle: connection between source cells (SC); right: connection between non-source cells. Auxin signaling is shown in green, and PIN protein quantity at the membrane in red. (D) 3D reconstructed globular stage embryo showing connection between two vascular founder cells. (E to L) Model simulations using a realistic heart stage embryo template before (E and I) and after (F to H and J to L) growth showing auxin signaling (green) (E to H) and CK signaling (red) (I to L) in a WT situation (E and F, I and J), with reduced auxin signaling (representing mp: G and K) and with reduced CK response (representing wol: H and L). Scale bars, 10 μm (D); those in (E) to (L) are relative to each other.

Intriguingly, the bridge needs to connect the source cells subtending the future cotyledon primordia (Fig. 6, B and C), and hence receive increased auxin input (Fig. 5, F to I, and fig. S6, A and B). We therefore analyzed whether the bridge serves a function in intercellular signaling. In our simulation, we blocked transport of auxin and CK across this bridge and found that the model output remained unchanged (fig. S10). We therefore conclude that the bridge imposes a geometric constraint to cell division.

We next tested robustness of our model toward parameter variation by performing sensitivity analysis (Supplementary Mathematical Modeling; fig. S11). This showed that model performance depended most strongly on parameters connected to CK biosynthesis, CK response, and effect on PIN1 localization and cell division.

Finally, to validate this model, we determined whether it could recapitulate the developmental consequences of reduced auxin response [monopteros (mp) mutant; parameter k1 set to 0] (27) or reduced CK response (wol mutant, parameter k8 set to 0) (12). In both cases, the model correctly predicted the cellular pattern and hormonal responses (Fig. 6, G and K, H and L).

Local CK biosynthesis is crucial for normal development

In our model, the same genetic network runs in all cells, but spatial bias imposed by local auxin sources is propagated to limit CK activation to TMO5/LHW-expressing xylem cells. To determine if local CK activation is required for normal growth and patterning, we first simulated the effects of uniform TMO5/LHW or LOG4 expression by increasing the CK activation rate in all cells. This increased periclinal cell division and generated a large disorganized vascular bundle lacking a central xylem axis (Fig. 7, A and B), resembling the TMO5/LHW-OX phenotype (4). We next tested the model prediction that local CK activation is important for vascular development by complementing the log1234578 mutant with either local (pTMO5) or ubiquitous (pRPS5A) LOG4 expression. pTMO5-driven LOG4 expression completely complemented the log1234578 mutant phenotype, whereas in contrast, pRPS5A-driven LOG4 expression induced supernumerary vascular cell files and loss of protoxylem differentiation (Fig. 7, C to F, and fig. S7). Hence, local LOG4 expression is sufficient for normal vascular development, and limitation to this domain is required to constrain cell number and patterning.

Fig. 7

Local CK activation integrates patterning and growth. (A and B) Model of ubiquitous CK signaling (representing the TMO5/LHW-OE phenotype). (C to E) Xylem patterns in log1234578 mutant roots (C) and log1234578 complemented with pTMO5-LOG4 (D) or pRPS5A-LOG4 (E) constructs (m, metaxylem; p, protoxylem). (F) Vascular cell file number in WT (Col-0), log1234578 mutant, and log1234578 complemented with pTMO5-LOG4 or pRPS5A-LOG4 constructs. Error bars indicate SE as determined by two-sided t test (**P < 0.01; n.s., not significant). (G) A relative CK response plot (in log scale) in model output of Fig. 3Q shows a steep gradient emanating from central xylem cells. Note that CK response in xylem cells is very low. (H to K) pTCSn-GFP (H and I) and pARR5-nYFP (J and K) expression in root meristems show differential response. Scale bars, 10 μm (C to E and H to K); those in (A) and (B) are relative to each other.

If CK is indeed locally activated in the xylem axis and diffuses outward, a gradient of CK response should be observed in procambium cells with maximal intensity close to the xylem (Fig. 6J). We tested this prediction by analyzing reporters for CK response. Indeed, both pTCSn-GFP and pARR5-nYFP markers displayed this gradient (Fig. 7, H to K) (3), suggesting that our model accurately predicts CK activity in the growing vascular tissue.

Network architecture during vascular tissue patterning

To understand how the genetic network is able to generate growth and patterning, we analyzed its modules in more detail. The conceptualized version of this network contains two interconnected incoherent feed-forward loop (IFFL) motifs (Fig. 8) (28). IFFL-1 entails the opposing effects of auxin and auxin-dependent CK on PIN levels, whereas IFFL-2 describes the effects of auxin on CK biosynthesis and response. Although the temporal dynamics of IFFLs have been previously discussed (28), their spatial properties may be more relevant for correct patterning in our model, given the intercellular signaling within a growing multicellular structure. We analyzed the individual contributions of these two IFFLs to spatiotemporal tissue patterning (see Supplementary Mathematical Modeling for more details) by separating the two subnetworks. We found that IFFL-1 can generate a high-auxin domain between the two source cells in the growing tissue (Fig. 8), but fails to from a sharp bisymmetric pattern. IFFL-2, on the other hand, can generate sharp boundaries between the high-auxin domain and the neighboring CK response domains (Fig. 8). Integration of both motifs thus generates stable and distinct hormonal response zones within the growing vascular tissue. As such, the high-auxin domain in the xylem axis acts as an organizer for the entire vascular bundle.

Fig. 8

The genetic network comprises two incoherent feed-forward loops. Overview of the two connected IFFLs in the genetic network. IFFL-1 generates a high-auxin domain, whereas IFFL-2 creates sharp boundaries, as indicated by a discretized 1D model (along the white arrow) representing CK concentration (dashed line) and CK response (in cambium, red bars) according to the distance from the auxin domain (xylem axis, green bar).


For decades, classical tissue culture experiments have been used to study the interaction between auxin and CK during tissue growth (29). These phytohormones were suggested to act in a mutual inhibitory fashion in vascular tissue patterning (3, 11). Here, we identified an interaction in which auxin promotes local CK activity in a manner central to both growth and patterning.

In parallel to promoting CK biosynthesis through TMO5/LHW as shown here, auxin was previously revealed to suppress CK response in the xylem (11). Hence, auxin triggers the formation of a nonresponding CK source. Modeling showed that this network, representing an IFFL, could account for generating a sharp boundary between high-auxin and high-CK domains. Furthermore, provided that auxin-dependent CK response suppression is cell-autonomous, CK diffusion will displace the domain of response to neighboring cells. An important question is what the nature of CK response inhibition is. Auxin activates expression of the CK response inhibitor AHP6 (3), but mutant phenotypes suggest that parallel or redundant functions must exist (11). A recent study proposed local activity of a CK oxidase in xylem cells as a potential redundant mechanism (16). However, it is questionable if local catabolism is compatible with the xylem acting as a CK source. The presence of a CK response gradient with the highest levels close to the xylem axis (3, 16) also renders regulation at the level of catabolism unlikely. Thus, inhibition likely acts at the level of CK response, and it will be important to identify its mediators.

Our simulations also revealed a surprising contribution of initial geometry to vascular tissue patterning. The bridge that connects the two auxin-accumulating vascular founder cells can be traced to the second round of divisions of the apical cell in the embryo. In addition, we have previously shown that the orientation of the first division is biased relative to the axes of the seed (2). Thus, symmetry breaking in the vascular tissue could occur much earlier than previously thought. Likewise, lineages that generate the two cotyledons, the likely sources of auxin for vascular initiation (3, 30), can also be traced to these very early divisions (31). Hence, both the geometric constraints and the signaling input that promote vascular tissue patterning are biased by the same cues. Whereas these findings show a clear correlation between geometry and pattern, it will be challenging to test causality because cell arrangements cannot easily be manipulated. Another critical question emerging from this is whether vascular tissue development in other plant species and during postembryonic organogenesis is similarly influenced by tissue geometry.

In conclusion, here we have identified a genetic network that reinforces an early developmental bias in auxin distribution to create a local, nonresponding source of CK, which drives growth and patterning of the embryonic vascular tissues.

Materials and methods

Plant material and cloning

All seeds were surface-sterilized, sown on solid MS plates, and vernalized for 2 days before growing at a constant temperature of 22°C in a growth room. The log mutants (13) and TMO5/LHW misexpression constructs were genotyped using the primers listed in table S3. All cloning was performed using the LIC cloning system (32) and the vectors described therein. For transcriptional fusions of the LOG genes, 3-kb fragments upstream of the ATG were PCR-amplified from genomic DNA using Q5 polymerase (NEB). To generate pRPS5A-driven misexpression, the coding sequences of all genes were amplified from complementary DNA (cDNA) clones. All constructs were completely sequenced. The primers used are listed in table S3.

Microscopic analysis

Differential interference contrast, fluorescence, and confocal microscopy were performed as described previously (33). For histological sections, roots were fixed overnight and embedded as described previously (34). 3D imaging of embryos was performed according to (35). Confocal image stacks were reconstructed, and segmentation was performed in MorphoGraphX software (36). Confocal imaging was performed on a Leica SP5-II system (HyD detector).

In situ hybridization

In situ hybridization was performed as described in (37). Briefly, siliques were harvested and fixed in phosphate-buffered saline (PBS) containing 4% paraformaldehyde by vacuum infiltration for 30 min. Samples were washed in PBS and subsequently dehydrated in consecutive ethanol solutions up to 100%. Samples were embedded in paraffin, and 8-μm-thick sections were prepared. Sections were pretreated with proteinase K (1 μg/ml) for 30 min at 37°C, and digoxigenin (DIG)–labeled RNA probes were hybridized at a concentration of 50 ng−1 ml−1 kb−1 for 16 hours at 45°C. The sections were incubated with anti-DIG antibody (Roche Applied Science) for 1 hour at room temperature. The sections were subsequently incubated in detection buffer [100 mM tris-HCl (pH 9.5), 100 mM NaCl, and 50 mM MgCl2] containing nitro blue tetrazolium (0.5 mg/ml) and bromochloroindolyl phosphate (0.125 mg/ml) (Roche Applied Science) for 1 to 16 hours. Detection was stopped in TE buffer [10 mM tris-HCl (pH 8.0), 1 mM EDTA]. DIG-labeled RNA probes of LOG1, LOG3, and LOG4 were generated from the entire cDNA sequence (TAIR) using T3 RNA polymerase and DIG Labeling Mix (Roche Applied Science).

CK measurements

CK quantification by ultrahigh performance liquid chromatography–electrospray tandem mass spectrometry (LC-MS/MS) was performed according to the method described previously (38). Briefly, 25 to 60 mg fresh weight of 4-day-old Arabidopsis seedling roots were collected and extracted in ice-cold modified Bieleski buffer (methanol/water/formic acid, 15:4:1, v/v/v) (39). To each extract, stable isotope-labeled CK internal standards (0.5 pmol of CK bases, ribosides, N-glucosides and 1 pmol of O-glucosides, nucleotides) were added to validate the quantification. For the purification of free CKs, two solid-phase extraction columns were used: the octadecylsilica-based column (500 mg of C18 sorbent, Applied Separations) and the MCX column (30 mg of mixed-mode sorbent, Waters) (40). Analytes were eluted by two-step elution using a 0.35 M NH4OH aqueous solution and 0.35 M NH4OH in 60% (v/v) methanol solution. Samples were then evaporated under vacuum at 37°C to dryness in vacuo. Purified samples were analyzed by the LC-MS/MS system consisting of an ACQUITY UPLC System (Waters) and a Xevo TQ (Waters) triple quadrupole mass spectrometer. Quantification was obtained using the multiple reaction monitoring mode of selected precursor ions and the appropriate product ions. For each mutant line, four independent biological replicates were analyzed.

ChIP and qRT-PCR analysis

ChIP was performed as previously described (41). Six hundred milligrams of 5-day-old pTMO5-TMO5-3GFP or pLHW-LHW-sYFP seedlings grown in long-day conditions were used. Anti-GFP antibody (5 μl per sample) (cat. no. 632592, Clontech) was used. Real-time PCR was performed using Power SYBR Green PCR Master Mix (Applied Biosystems). TMO5 and LHW occupancy on genomic DNA was calculated by computing the enrichment over the respective input and normalized over wild type. The primers used for ChIP-qPCR are listed in table S3.

Other qRT-PCR analyses were performed as described previously (42). RNA was extracted with the RNeasy kit (Qiagen). Poly(dT) cDNA was prepared from 1 μg of total RNA with an iScript cDNA Synthesis Kit (Bio-Rad) and analyzed on a CFX384 real-time PCR detection system (Bio-Rad) with iQ SYBR Green Supermix (Bio-Rad) according to the manufacturer’s instructions. Primer pairs were designed with Beacon Designer 8.0 (Premier Biosoft International). All individual reactions were done in triplicate with two or three biological replicates. Data were analyzed with qBase (43). Expression levels were normalized to those of EEF1α4, CDKA1;1, and ACTIN2. The primer sequences are listed in table S3.

Combinatorial transcriptome profiling, cell sorting experiments, and data analysis

To identify targets of the TMO5/LHW dimer, transcriptional changes were analyzed after a brief (1 hour) induction of ubiquitously expressed TMO5-GR protein (pRPS5A::TMO5-GR) (4) by dexamethasone in root tips (Fig. 1B). This induction was also performed in the presence of the protein synthesis inhibitor cycloheximide to exclude activation of indirect target genes. Because ectopic TMO5 expression can induce PD outside of the vascular domain, in an independent approach, TMO5-GR was ectopically induced in the ground tissue cells, where LHW is already present, thus allowing ectopic TMO5/LHW dimers to form. To enrich for TMO5-GR-expressing cells, we analyzed transcriptional changes in GFP-positive ground tissue cells after dexamethasone treatment in J0571>>TMO5::GR/GFP root tips (4) (Fig. 1B). These cells were sorted by fluorescence-activated cell sorting using fluorescence of the J0571 GAL4 enhancer trap line. Protoplasting and cell sorting was done as reported previously (44). Total RNA (100 ng) was labeled using an Ambion WT Expression Kit (Life Technologies) and hybridized to Arabidopsis Gene 1.0 ST arrays (Affymetrix), which probes the expression of 27,827 unique genes. Sample labeling, hybridization to chips, and image scanning were performed according to the manufacturer’s instructions. Microarray analysis was performed using MADMAX pipeline for statistical analysis of microarray data (45). Expression values were calculated using the robust multichip average (RMA) method, which includes quantile normalization (46, 47). Probe sets on the array were redefined using current genome information (48). Here, probes were reorganized on the basis of the gene definitions as available in the TAIR10 database. Differentially expressed probe sets (genes) were identified by linear models and an intensity-based moderated t statistic, taking into account the paired design (49). P values were corrected for multiple testing by a false discovery rate method (50), and probe sets that satisfied the criterion of P < 0.05 were considered to be significantly regulated. Combining these three experiments yielded a small set of 143 genes that were significantly up-regulated (>1.5-fold; P < 0.05) in at least one experiment (see table S1). The transcriptomics data files are submitted to Gene Expression Omnibus (GEO) (accession no. GSE56868).

Supplementary Materials

Figs. S1 to S12

Tables S1 to S4

Movies S1 and S2

Mathematical Modeling

References (5153)

References and Notes

  1. Acknowledgments: We would like to thank H. Sakakibara, Y. Helariutta, T. Laux, and B. Müller for sharing plant materials; J. Jansen for technical assistance with microarray hybridizations; and M. Glosová and H. Martínková for their help with cytokinin analyses. B.D.R. was supported by long-term FEBS and Marie Curie Fellowships (IEF-2009-252503), and D. Weijers received funding from the Netherlands Organization for Scientific Research (NWO; ALW-VIDI-864.06.012 and ALW-820.02.019 and the ERA-CAPS project EURO-PEC; 849.13.006) and the European Research Council (Starting Grant “CELLPATTERN,” contract no. 281573). M.A. was funded by the SysmedIBD project funded under the European Union 7th framework project (contract no. 305564). O.N. acknowledges the Ministry of Education, Youth, and Sports of the Czech Republic (The program “Návrat” for Research, Development, and Innovations, grant LK21306; the National Program for Sustainability I, grant LO1204) and grant funding no. 14-34792S provided by the grant agency of the Czech Republic. K.L. acknowledges the Swedish Governmental Agency for Innovations Systems (VINNOVA) and the Swedish Research Council (V.R.). D. Wagner was supported by NSF grant (IOS 1257111). The transcriptomics data files are available at GEO (accession no. GSE56868).
View Abstract

Navigate This Article