Two Seven-Transmembrane Domain MILDEW RESISTANCE LOCUS O Proteins Cofunction in Arabidopsis Root Thigmomorphogenesis

Directional root expansion is governed by nutrient gradients, positive gravitropism and hydrotropism, negative phototropism and thigmotropism, as well as endogenous oscillations in the growth trajectory (circumnutation). Null mutations in phylogenetically related Arabidopsis thaliana genes MILDEW RESISTANCE LOCUS O 4 ( MLO4 ) and MLO11 , encoding heptahelical, plasma membrane–localized proteins predominantly expressed in the root tip, result in aberrant root thigmomorphogenesis. mlo4 and mlo11 mutant plants show anisotropic, chiral root expansion manifesting as tightly curled root patterns upon contact with solid surfaces. The defect in mlo4 and mlo11 mutants is nonadditive and dependent on light and nutrients. Genetic epistasis experiments demonstrate that the mutant phenotype is independently modulated by the G b subunit of the heterotrimeric G-protein complex. Analysis of expressed chimeric MLO4/MLO2 proteins revealed that the C-terminal domain of MLO4 is necessary but not sufﬁcient for MLO4 action in root thigmomorphogenesis. The expression of the auxin efﬂux carrier fusion, PIN1-green ﬂuorescent protein, the pattern of auxin-induced gene expression, and acropetal as well as basipetal auxin transport are altered at the root tip of mlo4 mutant seedlings. Moreover, addition of auxin transport inhibitors or the loss of EIR1/AGR1/PIN2 function abolishes root curling of mlo4 , mlo11 , and wild-type seedlings. These results demonstrate that the exaggerated root curling phenotypes of the mlo4 and mlo11 mutants depend on auxin gradients and suggest that MLO4 and MLO11 cofunction as modulators of touch-induced root tropism.


INTRODUCTION
Roots must circumnavigate barriers in soil to reach nutrient patches to optimize root growth and architecture (Osmont et al., 2007). Intense effort now focuses on engineering crop root architecture to maximize yield through improved fitness (de Dorlodot et al., 2007). Because root architecture is flexible, it also serves as a model system to study developmental plasticity, particularly toward understanding intrinsic and extrinsic factors that drive organ form (Malamy, 2005). At the molecular level, a sketch of the molecular pathway linking these signals to formation of root architecture and growth is coming into focus, but many details of these pathways remain obscure.
Root architecture develops out of a combination of cell division and expansion, lateral root formation, and cellular differentiation (e.g., root hair formation). Although roots can grow straight with lateral roots forming at regular intervals, Arabidopsis thaliana roots have more complex growth patterns that are environmentally sensitive. For example, root gravitropic curvature upon reorientation relative to the gravity vector is a well-characterized process (Muday and Rahman, 2007). Root waving, skewing (also termed slanting), and curling are also common growth behaviors when plants are grown on a hard surface that roots cannot penetrate (Okada and Shimura, 1990).
The major increase in root length is due to cell elongation in a zone distal to the tip. The amount and direction of growth is controlled by mechanical (touch) and gravity forces that are sensed in the root tip (Legué et al., 1997;Sack, 1997;Blancaflor et al., 1998;Fasano et al., 2001). Experimental evidence indicates that mechanical and gravity sensing are integrated to drive root growth around barriers in soil over a wide range of impedance to growth, suggesting that mechanosensing feeds back onto gravity perception or response Gilroy, 2003a, 2003b). Mutants with altered auxin response or transport show altered gravitropism, curling, waving, and skewing responses (Okada and Shimura, 1990;Garbers et al., 1996;Luschnig et al., 1998;Rutherford et al., 1998;Marchant et al., 1999;Rashotte et al., 2001;Piconese et al., 2003;Buer and Muday, 2004), indicating a critical role for auxin in regulation of these processes. Additionally, feedback may involve the phytohormone ethylene, as this gas inhibits root waving (Buer et al., 2003), root gravitropism (Buer et al., 2006), and lateral root formation (Negi et al., 2008).
Thigmomorphogenesis is defined as the sum of growth responses of a plant organ in response to a mechanical stimulus (Jaffe et al., 2002). The molecular basis of the touch sensors and signal transduction components that finally give rise to differential organ growth is largely unknown. It is thought that Ca 2+ plays a key role as a second messenger in signal transduction during mechanosensing (e.g., via the activation of mechanosensitive Ca 2+ channels) (Jaffe et al., 2002). Indeed, rapid and transient changes in the concentration of cytosolic free Ca 2+ levels (Ca 2+ spikes) were observed upon mechanical stimulation in roots (Knight et al., 1991;Legué et al., 1997). Cloning of a set of touchinduced genes, of which three encode either calmodulin or calmodulin-related proteins, and identification of a plasma membrane protein that links mechanosensing with Ca 2+ influx provides genetic evidence for the importance of Ca 2+ signaling in this physiological process (Braam and Davis, 1990;Sistrunk et al., 1994;Nakagawa et al., 2007a).
Some mechanosensing pathways in human cells and fungi involve cell surface seven-transmembrane (7TM) domain receptors coupled to heterotrimeric G-proteins that activate phospholipase C, leading to rapid transient calcium increases in the cytoplasm (Chachisvilis et al., 2006;Kumamoto, 2008). Arabidopsis encodes a limited number of predicted 7TM domain proteins (Moriyama et al., 2006;Gookin et al., 2008), and within this set, the largest protein family consists of 15 members. They are designated Mildew Resistance Locus O (MLO) since genetic studies revealed that several of these proteins, including the earliest identified barley (Hordeum vulgare) MLO and its coorthologs, MLO2, MLO6, and MLO12 in Arabidopsis, function in modulating antifungal defense, possibly through negatively regulating cell death and senescence-related physiology (Bü schges et al., 1997;Piffanelli et al., 2002;Consonni et al., 2006). Although MLO proteins are plant specific at the level of amino acid sequences, structural features of these proteins are similar to those of metazoan G-protein-coupled receptors (GPCRs). MLOs are plasma membrane delimited via their 7TM domains with the N terminus positioned extracellularly and the C terminus intracellularly (Devoto et al., 1999). Potential in planta homodimerization of MLO proteins has been revealed by Fö rster/fluorescence resonance energy transfer analysis (Elliott et al., 2005).
Here, we used a reverse genetics approach to identify 7TM domain proteins that play a role in root-associated physiological processes. We discovered two MLO genes, MLO4 and MLO11, that cofunction in root thigmomorphogenesis. By employing a combination of genetic epistasis analysis, chemical genetics, and cell biology, we demonstrate that the touch-induced anisotropic root growth in mlo4 and mlo11 mutants is auxin dependent. We propose that MLO4 and MLO11 operate in a heterooligomeric complex to regulate thigmotropic root growth.

MLO4 and MLO11 Modulate Asymmetrical Root Growth
We isolated and characterized T-DNA insertion lines of all genes in the 15-member Arabidopsis MLO family, including at least two independent alleles each for MLO4 (mlo4-1, mlo4-3, and mlo4-4), MLO11 (mlo11-3 and mlo11-4), and MLO14 (mlo14-6 and mlo14-7), the subjects of this study ( Figure 1A). As assayed by RT-PCR, these mlo4, mlo11, and mlo14 mutant plants lacked detectable full-length transcripts of the respective MLO genes, suggesting that all are null mutant alleles (see Supplemental Figure 1 online; mlo4-3 not tested). MLO4, MLO11, and MLO14 are the sole members of a defined phylogenetic clade within the Arabidopsis MLO family (Devoto et al., 2003;Chen et al., 2006) and have unknown functions. Prominent expression of MLO4, MLO11, and MLO14 in the root tip  prompted us to investigate root-related growth phenotypes in the respective mutants. In the context of these experiments, we discovered asymmetrical root growth patterns of mlo4 and mlo11 mutants on horizontal surfaces.
Under the conditions used here (minimal medium; see Methods), wild-type roots grew either in a less-consistent growth trajectory, comprising a combination of straight and undulating growth, or in a moderate coiling pattern, also known as root curling (Mirza, 1987). Penetrance of one or the other root growth type was shown to be subject to natural genetic variation (see Supplemental Figure 2A online) and also dependent on various exogenous parameters, such as medium composition and hardness (Buer et al., 2000). Recessive mlo4 and mlo11 alleles, but not mlo14 alleles, exhibited exaggerated root curvature typically resulting in a tight spiral-like root growth pattern that exaggerates the mild root bending of the parental Columbia-0 (Col-0) wildtype plants ( Figure 1B). These root coils were qualitatively distinct from the natural repertoire of root curling observed in a range of Arabidopsis ecotypes that are part of a core set representing accessions with maximal genetic and morphological diversity (see Supplemental Figure 2A online; McKhann et al., 2004). Root coiling of mlo4 and mlo11 mutants was robust on minimal media with varying sucrose content and agar source (see Supplemental Figure 2B online) but was less prominent on medium containing Murashige and Skoog (MS) salts, indicating that one or more ingredient(s) of the MS salts interfere(s) with this phenotype. Formation of mlo4 and mlo11 root coils is not triggered spontaneously but requires a tactile stimulus (e.g., contact of the root tip with a hard surface) ( Figure 1C), suggesting that mlo4 and mlo11 mutants might be defective in thigmomorphogenesis. Besides aberrant root curling, mlo4 mutant seedlings and to a lesser degree also mlo11 seedlings, but not mlo14 seedlings, exhibited irregular root-waving patterns on tilted hard agar containing minimal medium. This phenotype included shorter wavelengths, resulting in compressed root growth tracks (mlo4), and apparently stochastic formation of loops (mlo4 and mlo11; Figure 1D). Moreover, root slanting on inclined hard surfaces was also altered in case of the mlo4 mutant. Compared with the Col-0 wild-type control, which showed the previously reported right-handed (to the left when viewed from the top; Migliaccio et al., 2009) skewing pattern, mlo4 mutants showed less root skewing but exhibited a more prominent waving growth pattern. By contrast, the degree of skewing for mlo11 and mlo14 was indistinguishable from the wild type ( Figure 1E).
Although the qualitative differences in root growth patterns between genotypes were visibly evident, we sought to quantify root curling and skewing of mlo4 and mlo11 mutants. In the case of root curling, two methods were used and the results from both methods concurred (see Supplemental Figure 3 online). One procedure, first described by Mirza (1987), provides the total root angle, which is positively correlated with curling. Typically, this method reports the general direction of curling (handedness relative to the gravity vector). The second approach captures the largest span of the root coil or trajectory, which is negatively correlated to the degree of root curling. We modified the Mirza method by normalizing against root length to eliminate differences due to growth rate and compared it to the root span method to show that both methods adequately quantitated this type of anisotropic growth pattern (see Supplemental Figure 3 online). As shown in Figures 2A and 2B, curling of mlo4 and mlo11 roots was greater than the wild type, whereas disruption of MLO14 alone had no effect on root curling. Root curling of mlo4 mutants occurred more consistently and frequently in tighter spirals than in mlo11 seedlings, which was apparent by the statistically significant difference in variance between the mlo4 and mlo11 mutants (Figure 2A). Quantitative analysis also corroborated sucrose independence of mlo4-conditioned root (A) Genetic map of the T-DNA insertion alleles of mlo4, mlo11, and mlo14 mutants used in this study. MLO exons are represented by dark-gray rectangles and introns by connecting lines; the arrowhead at the end of the last exon indicates 59-39 directionality of the genes. Sites of T-DNA insertion in the various mutant alleles are marked by light-gray triangles. Smaller light-and dark-gray triangles denote the location of translational start and stop codons, respectively. (B) Altered root growth of mlo4 and mlo11 mutants on horizontal plates (minimal medium, 1% sucrose) viewed from the bottom. Seedlings of indicated genotypes were grown under continuous light for 5 d before taking pictures. (C) A tactile stimulus triggers mlo4-mediated root spirals. Col-0 and mlo4-4 seedlings were grown vertically on minimal medium (0.25% sucrose) for 5 d (12-h light/dark cycle) when a glass cover slide (white arrowhead) was positioned a few millimeters under the root tip. Root growth patterns on the cover slide were observed 4 d later. Note the mild coiling or undulating growth of Col-0 roots compared with the tight curls of mlo4 roots. (D) and (E) Root waving (D) and skewing (E) phenotypes of mlo mutant seedlings. Seedlings were vertically grown on hard minimal medium (0.25% sucrose) for 2 d (12-h light/dark cycle); then, plates were slanted at 458 or 358 for observation of waving or skewing phenotypes, respectively. Photographs were taken 4 d later from the top of the agar plate. Mutants tua4 M268I and tua4 S178D , showing either left-handed (to the right when viewed from the top) or exaggerated right-handed slanting, respectively, were included as controls for aberrant skewing phenotypes (Ishida et al., 2007). Bars = 5 mm.
curling (see Supplemental Figure 2C online). Quantification of skewing indicated that the mlo4 roots skewed significantly less than wild-type roots ( Figure 2D).
The fact that disruption of either MLO4 or MLO11 is sufficient to cause detectable alterations in touch-induced asymmetrical root growth supports cofunction of the two genes rather than redundancy. Since mlo4 and mlo11 exhibited subtle yet consistent differences in their exaggerated asymmetrical root growth patterns (Figure 2A), MLO4 and MLO11 may regulate the same process(es), but each protein might play slightly different roles. The mlo4 mlo11 double mutant and the mlo4 mlo11 mlo14 triple mutant did not show an additional reduction in root span ( Figure 2C) or further exaggeration of aberrant root waving or skewing patterns ( Figures 1D, 1E, and 2D), further supporting cofunctionality of the MLO4 and MLO11 polypeptides. Alternatively, the extent of root curling may have a limit for a root of a given length.

mlo4 Mutants Exhibit Unaltered Root Architecture and Gravitropic Responses
To obtain insight into the potential mechanism of mlo4-conditioned root curling, we studied the morphology of mlo4 and mlo11 mutants in comparison to roots of wild-type plants. Our analyses revealed no obvious difference with respect to size, number, and organization of characteristic cell files of the root tip (see Supplemental Figure 4A online). Additionally, spiral roots of mlo4 and mlo11 mutants did not exhibit twisted epidermal cell (A) Accumulated end-point curvature of 5-d-old light-grown seedling roots on horizontal plates (minimal medium, 1% sucrose). Positive numbers represent counterclockwise coils (as viewed from the bottom) and depict right-handed growth. Negative numbers represent clockwise root curvatures. (B) Seedling root span (the distance from the root tip to the root base) on horizontal plates (minimal medium, 1% sucrose). Data for (A) and (B) were collected from the same 5-d-old light-grown sample population. (C) Root span of 5-d-old, light-grown seedlings on horizontal plates (minimal medium, 1% sucrose). For each experiment in (A) to (C), at least 25 samples were measured per genotype. Shown are data from one replicate each; all experiments were repeated at least once yielding similar results. Columns indicate the average, and error bars represent SD. Asterisks indicate a statistically significant (P < 0.05) difference in comparison to Col-0 according to Bonferroni's multiple comparison test, the # symbol indicates statistically significantly different (P < 0.01) variances between the mlo4-1 and mlo11-3 data sets according to an F-test. Please note that the latter differs from the statistically nonsignificant variances between the mlo4-1 and mlo-11-3 data sets of darkgrown seedlings shown in Figure 4E. (D) Quantification of root skewing phenotypes from seedlings vertically grown on hard minimal medium (0.25% sucrose) as described in Methods. Root slanting angles were measured from the vertical; righthandedness (to the left when viewed from the top; cf. Figure 1D) was scored as negative and left-handedness as positive angle. Columns indicate the mean, and error bars the SD based on three replicates. For each replicate, 12 to 19 samples were measured per genotype. Asterisks indicate a statistically significant (P < 0.05) difference in comparison to Col-0 according to Bonferroni's multiple comparison test. files, excluding touch-triggered organ torsion as a possible cause for the root curling phenomenon (see Supplemental Figure  4B online). To unravel potential alterations in molecular root architecture, we crossed transgenic lines expressing fluorescent marker proteins (pPIN1:PIN1-green fluorescent protein [GFP], pPIN2:PIN2-GFP, and pSCR:GFP) for root cell identity with the mlo4-1 or mlo4-3 mutant. Roots of F2 progeny of these crosses that were homozygous for mlo4 and expressed the respective fluorescent marker protein were either grown vertically (straight roots) or horizontally (curling roots in case of mlo4) and imaged by confocal microscopy. Micrographs revealed similar cellular distribution patterns for the tested polypeptides in both conditions. However, there were potential differences in expression levels and/or protein turnover rates that in part depended on the growth mode. Compared with the wild-type background, PIN1-GFP levels were reduced in the vascular bundle, the pericycle, and the endodermis in horizontally but not in vertically grown roots of the mlo4 mutant ( Figure 3). By contrast, PIN2-GFP levels were higher in the mlo4 mutant background in both growth conditions ( Figure 3A; see Supplemental Figure 5 online). The latter may either be a consequence of the mlo4 mutation or represent the result of variation in transgene expression, thus precluding firm conclusions. In both genotypes (the wild type and mlo4 mutant), GFP fluorescence patterns originating from the pSCR:GFP transgene were too variable and patchy to allow reliable quantification. In sum, these findings indicate that the overall root architecture is unaltered in mlo4 mutants, while steady state levels of the root cell marker PIN1 are conditionally affected in mlo4 roots.
We analyzed root gravitropic responses in mlo4 and mlo11 single mutants and the mlo4 mlo11 double mutant in comparison to wild-type plants to test the hypothesis that lesions in graviperception or graviresponsiveness caused the root behavior phenotypes of the mlo4 and mlo11 mutants. The kinetics of gravity-induced root bending of light-grown seedlings on minimal medium were similar in  (see Supplemental Figure 6 online), indicating that mlo4 and mlo11 mutants are not defective in root gravitropism under these growth conditions. Consistent with this finding, the amount and positioning of starch-containing statoliths in mlo4 and mlo11 mutants were indistinguishable from the wild type (data not shown). Taken together, these data suggest that the altered root growth pattern of mlo4 and mlo11 mutants is not a consequence of globally perturbed root architecture or major defects in graviperception/gravitropism.

Phenotypes of mlo4 and mlo11 Mutants Are Light Dependent
Environmental factors, such as light and gravity, affect the mild root curling of wild-type plants (Mirza, 1987). We observed that Arabidopsis mutants defective in responding to gravity did not show the wild-type level of root curling under darkness (Figures 4A and 4B). This applied to phosphoglycerate mutase1 (pgm1-1) and ADP glucose phosphorylase 1 (adg1-1), two mutants with reduced starch biosynthesis. By contrast, the multidrug resistance protein1/ATP binding cassette transporter b19 (mdr1-100/ abcb19) mutant, which is defective in an auxin efflux transporter (A) Comparison of average GFP fluorescence intensities in projections of z-stacks through the root for tested reporter constructs in Col-0 and mlo4 mutant primary roots. Seedlings were grown for 6 d (12-h light/dark cycle) either horizontally (hor.) or vertically (ver.) on minimal medium (0.25% sucrose). The bar graph depicts mean fluorescence intensities (6SD) based on ;10 root tips per genotype and growth condition. The asterisks indicate a statistically significant (P < 0.01) difference in comparison of GFP signal intensities in mlo4 primary root tips to those in Col-0 primary root tips according to two-tailed Student's t test with unequal variances. Shown are data from one replicate; the experiment was repeated once, yielding similar results. (B) Representative fluorescence micrographs of horizontally grown, 6-dold Col-0 and mlo4-3 seedlings expressing PIN1-GFP under the control of its own promoter. The micrographs show the GFP and FM4-64 fluorescence, respectively, and a merged image of both channels in an optical z-section at the level of the quiescent center of the primary root tips. Seedlings were grown horizontally (minimal medium, 0.25% sucrose; 12-h light/dark cycle) and stained for 10 min with FM4-64. Bars = 20 mm. and shows enhanced gravitropic and/or waving responses (Lin and Wang, 2005;Lewis et al., 2007), curled more tightly under darkness than Col-0 wild-type roots ( Figure 4C).
Light affects the magnitude of curling and the associated variance in the root populations of Col-0, mlo4-1, and mlo11-3. Under the conditions used here, roots of Col-0 wild-type plants curled more in darkness than in light (cf. Figures 4D and 2A to 4E). mlo4-and mlo11-conditioned root curling was significantly greater than the wild type both under constant light of weak intensity (30 mmol m 22 s 21 ) and in 12-h light/dark cycles of higher light intensity (100 mmol m 22 s 21 ). However, for darkgrown roots, differences in curling magnitude were not observed between the wild type and the mlo4-1 and mlo11-3 mutants ( Figure 4E). The fact that the abnormal root curling pattern of the mlo4 and mlo11 mutants is light dependent suggests that the phenotype of these mutants may be associated with a defect in either perceiving or responding to a light-associated or photosynthesis-derived signal.

MLO4 and the Heterotrimeric G-Protein Complex Operate on Different Pathways in Regulating Asymmetrical Root Growth
Since the 7TM domain topology of MLO proteins resembles metazoan/fungal GPCRs, they are candidate plant GPCRs (Devoto et al., 1999;Kim et al., 2002;Moriyama et al., 2006). Therefore, we performed epistasis analysis between mlo4 alleles and null mutations in heterotrimeric G-protein subunit genes. Arabidopsis GPA1 and AGB1 genes encoding aand b-subunits of the G-protein complex (Temple and Jones, 2007), respectively, are expressed in root tips . Roots of the agb1-2 mutant curled less than the wild type ( Figure 5), suggesting an involvement of AGB1 in modulating asymmetrical root growth. Assuming that at least some of the mechanism underlying root waving is shared by root curling, a concept that is supported by previous studies (Simmons et al., 1995) and our own data ( Figures 1B and 1D), it is worth noting that agb1-2 mutant roots also showed reduced root waving (Pandey et al., 2008). Roots of the agb1-2 mlo4-1 double mutant curled less than those of the mlo4-1 single mutant but more than those of the agb1-2 mutant, although this difference was not statistically significant ( Figure 5). These data suggest that the combination of these two mutant alleles is phenotypically additive rather than one mutant allele being clearly epistatic to the other. By contrast, the gpa1-4 mutation had less impact, if any, on root curling ( Figure 5), which is consistent with its wild-type-like root waving phenotype (Pandey et al., 2008). These findings demonstrate that AGB1, but not GPA1, plays a role in both root curling and waving. The data also show that the effect of the null agb1 mutation does not depend on a functional MLO4 and vice versa, (A) and (B) Horizontal root curvature of Arabidopsis mutants defective in responding to gravity. Seedling roots of the wild type, pgm1-1, and adg1-1 were grown on minimal medium (1% sucrose) under darkness for 5 d before recording the image from the bottom (A) and measuring accumulated end-point root curvature (B). (C) Accumulated end-point root curvature of 5-d-old, dark-grown mdr1-100 mutants viewed from the bottom. (D) Comparison of light-and dark-grown wild-type (Col-0) seedling roots. Seedlings were grown on horizontal plates (minimal medium, 1% sucrose) for 5 d either in darkness or under constant low light (25 mM m À2 s À1 ) from above. Digital images were captured from the plate bottom. (E) Accumulated end-point root curvature of 5-d-old, dark-grown (minimal medium, 1% sucrose) wild-type, mlo4-1, and mlo11-3 seedlings. Columns indicate the mean, and error bars represent SD. At least 15 samples were measured for each genotype. Shown are data from one replicate each; all experiments were repeated at least once, yielding similar results. Asterisks indicate a statistically significant (P < 0.05) difference in comparison to Col-0 according to Bonferroni's multiple comparison test; the # symbol indicates a statistically significant (P < 0.05) difference in comparison to Col-0 according to Student's t test. The variance (according to an F-test) between the mlo4-1 and mlo-11-3 data sets of dark-grown seedlings shown is not significant ( Figure 4E) in contrast with the variance observed with light-grown seedlings ( Figure 2A).
[See online article for color version of this figure.] suggesting that the MLO4 activity in root growth patterning is independent of the heterotrimeric G-protein complex.

The C-Terminal Cytoplasmic Domain of MLO4 Is Necessary but Not Sufficient for Regulation of Asymmetrical Root Growth
Swapping domains between sequence-related proteins of two different known signaling pathways is a powerful way to map protein function to structure. To identify functional domains of MLO4, we swapped domains between MLO4 and MLO2, two MLO proteins that function in distinct physiological processes. Whereas members of the phylogenetic branch comprising MLO4, MLO11, and MLO14 are involved in regulating asymmetrical root growth, members of the clade comprising MLO2, MLO6, and MLO12 play a role in modulating antifungal defense in leaf epidermal cells (Consonni et al., 2006). Although MLO4 and MLO11 are expressed in leaf tissues , mlo4 and mlo11 T-DNA insertion mutants retain the wild-type level of susceptibility to fungal (powdery mildew) infections (Consonni et al., 2006). Similarly, although MLO2, the gene with the major modulatory role in antifungal defense, is expressed in roots , a powdery mildew resistant mlo2 T-DNA insertion mutant displayed wild-type root growth patterns (see Supplemental Figure 7 online).
Functional specificity of MLO paralogs may reside in amino acid sequences of these proteins that are evolutionarily conserved within each phylogenetic clade (Panstruga, 2005). To elucidate this possibility, we focused on the regions that we identified to be of greatest dissimilarity between MLO2 and MLO4: the first extracellular loop (e1) and the cytoplasmic C terminus (c4). To perform domain swaps between these domains, we dissected the MLO proteins into four segments: the N-terminal fragment before the e1 loop, the e1 loop, the central region between e1 and c4, and the C-terminal c4 fragment. We created binary vectors harboring synthetic genes designed to express all 16 possible combinations of MLO2-MLO4 chimeras ( Figure 6) as fusion with a C-terminal GFP tag to monitor protein expression levels and subcellular localization in the transgenic lines. Expression of the chimeric genes was under the control of the native MLO4 promoter (represented by a 1.05-kb DNA fragment covering the 59 upstream region of MLO4 before the start codon; see Methods). All constructs were transformed into the mlo4-1 null mutant background to assess complementation of the mlo4 root curling phenotype by the chimeric MLO fusion proteins.
Progeny (T2 seedlings) of six independent transgenic lines for each construct were analyzed. All lines yielded detectable GFP signals in the root tip region, although the expression levels and subcellular localization patterns varied between constructs and between transgenic lines (see Supplemental Figure 8 online). Full-length MLO4 cDNA, but not the full-length MLO2 cDNA, rescued the mlo4 root curling phenotype ( Figures 6A and 6B; for definition of criteria for complementation, see Methods). MLO4 harboring the e1 loop or the N terminus plus the e1 loop from MLO2 also complemented the mlo4 mutant phenotype ( Figures  6E and 6K), suggesting that these MLO2-derived domains are either functionally equivalent to the respective MLO4 regions or Seedling root span (A) and accumulated end-point curvature (B) of mlo4-1, G-protein mutants gpa1-4 and agb1-2, and their double mutants. Columns indicate the average, and the error bars represent SD. More than 25 seedlings were measured for each genotype. Seedlings were grown on horizontal plates (minimal medium, 1% sucrose) under constant light for 5 d. Shown are data from one exemplary replicate; the experiment was repeated once, yielding similar results. Asterisks indicate a statistically significant (P < 0.05) difference in comparison to Col-0 according to Bonferroni's multiple comparison test. that the identity of these parts of the protein is subordinate for MLO4 function. By contrast, an MLO4 chimera with the MLO2 c4 domain did not complement ( Figure 6G), indicating the importance of the identity of this domain for functional specificity. Since MLO2 with c4 from MLO4 partially complemented the mutant phenotype ( Figure 6H), functional specificity of MLO proteins seems to require the coordinated action of multiple cytoplasmic domains (Elliott et al., 2005). This hypothesis is further supported by the fact that all other chimeric proteins, with one exception that permitted partial complementation ( Figure  6F), failed to rescue the mlo4 root curling phenotype. Given the variance of this assay, it remains possible that we underestimated the capacity of some chimeric proteins for complementation owing to differential expression, stability, or ectopic subcellular localization of the respective artificial polypeptides.

mlo4-Conditioned Root Curling Requires Normal Auxin Transport
An unbiased chemical genetic screen of 2300 tested substances from three compound libraries (for details, see Methods) yielded 14 low molecular weight compounds that attenuated the exaggerated root curling phenotype of mlo4 and mlo11 mutants in a dosage-dependent manner (see Supplemental Figure 9 online). Among these 14 compounds were the auxin transport inhibitors 1-N-naphthylphthalamic acid (NPA), difluenzopyr and quinchlorac, the vesicle transport inhibitor Brefeldin A (BFA), four benzopyrans (methylesculetin, karanjin, robustic acid, and prunetin), and six structurally related compounds of unknown chemical identity (see Supplemental Figure 9B online). Three of the four detected benzopyrans are flavonoids, a group of plant secondary metabolites that were implicated in the regulation of auxin transport (Brown et al., 2001). Likewise, BFA was previously shown to indirectly affect auxin transport by impairing subcellular auxin efflux carrier cycling and positioning (Geldner et al., 2001(Geldner et al., , 2003. Identification of three known auxin transport inhibitors, BFA, and three flavonoids from this screen suggests an essential contribution of auxin transport in mlo4-and mlo11-conditioned root curling. Owing to the space constraints associated with the microtiter-based compound screen, conclusions on the magnitude of impact of these substances on wild-type roots relative to the mutants were difficult to draw. We therefore focused our study on the effect of the well-characterized auxin transport inhibitor NPA on Col-0 wild-type plants using agar plates. Consistent with a general role of auxin transport in asymmetrical root growth, NPA interfered not only with mlo4-and mlo11-conditioned root coiling but also with the moderate root curling of Col-0 wild-type plants ( Figure 7A). Taken together, these data suggest that MLO4 and MLO11 function upstream of the establishment of auxin distribution patterns and/or auxin transport.
PIN2 plays an essential role in the basipetal transport of auxin to the elongation zone and has been shown to regulate the asymmetrical root growth under gravity stimuli (Chen et al., 1998;Rashotte et al., 2000Rashotte et al., , 2001. Disruption of MLO4 seems to have a constitutive positive effect on the expression levels of a pPIN2: PIN2-GFP construct (Figure 3; see Supplemental Figure 5 online). Mutation of PIN2 (eir1-1 mutant) caused significantly decreased root curling as a single mutant ( Figure 7B), indicating the importance of PIN2 for this asymmetrical growth pattern. The eir1-1 mlo4-1 double mutant root exhibited the eir1 phenotype of reduced root curling ( Figure 7B), indicating that the root curling phenotype of mlo4 also depends on a functional PIN2.
Differential auxin responsiveness is often observed by following auxin-induced reporters in the asymmetrical growth of gravitropism and phototropism Friml et al., 2002;Paciorek et al., 2005). The involvement of auxin in root curling suggests that the primary vertical stimuli from gravity and light may be interpreted and the response is manifested through asymmetric auxin distribution or signaling in the root tip and/or elongation zone. To examine whether asymmetric auxindependent gene expression occurs during root curling, we examined transgenic seedlings expressing the auxin-inducible synthetic DR5rev:GFP reporter (Ottenschlä ger et al., 2003) on horizontal surfaces. In a population of 400 etiolated seedlings, >65% exhibited differential distribution of the GFP-based auxin reporter. Representatives of the four specific types of differential GFP distribution observed at the root tip are shown in Figures 8A  to 8D. Regardless of the handedness of root growth, roots with a relatively higher amount of DR5-regulated GFP expression on the concave side were more prevalent than roots with a GFP expression on the convex side. A relatively higher amount of auxin at the concave side of roots was observed during gravity responses as well, and others concluded that this differential auxin distribution drives the orientation of root tip growth Paciorek et al., 2005). Considering that the steady state level of GFP requires time for biosynthesis upon increase of local auxin concentrations and time for degradation when the auxin level decreases, the observed DR5-driven GFP pattern is unlikely to reflect the real-time dynamics of auxin redistribution. To explain the unexpected 19.5% (sum of types shown in Figures 8C and 8D) root tips showing more GFP fluorescence at the convex side of curling roots, we speculate that the GFP signal in these roots represents the auxin The MLO4-MLO2 chimeras ([A] to [P]) were fused with a C-terminal GFP, and expression in transgenic lines was controlled by the MLO4 promoter. Black-colored fragments were from MLO4; dark-gray fragments were from MLO2, and light-gray boxes represent transmembrane domains. For each chimera, 20 to 40 light-grown T2 seedlings of six independent transgenic lines were tested for the ability to rescue the mlo4-1 phenotype, measured as root span, using Col-0 wild-type and mlo4-1 plants on the same plate (minimal medium, 1% sucrose) as controls. Columns indicate the average, and error bars represent SD. Asterisks indicate a statistically significant (P < 0.05) difference in comparison to mlo4 but no significant difference from Col-0 (evidence for full complementation), while the # symbol indicates a statistically significant difference in comparison to both mlo4 and Col-0 (evidence for partial complementation) according to Bonferroni's multiple comparison test. distribution pattern established prior to a change in growth direction. Our data demonstrate that lateral auxin asymmetry accompanies naturally occurring asymmetrical root growth.
Since asymmetrical root growth in wild-type plants correlated with characteristic auxin distribution patterns at the root tip, we compared the pattern of auxin-induced gene expression, as an indirect measure of auxin distribution, in wild-type and mlo4 mutant plants. Compared with the Col-0 wild type, we observed a broader yet symmetrical distribution of DR5 promoterregulated GFP expression in the root tips of horizontally grown (curling) mlo4-3 mutant plants (Figures 8E and 8F). This pattern is reminiscent of DR5 promoter-driven GFP expression in the root tip of the mdr1/abcb19 mutant, which is defective in an auxin transporter-associated ABC transporter . Vertically grown roots of mdr1/abcb19 seedlings exhibit sporadic root curvature and reduced acropetal auxin transport Wu et al., 2007). In Col-0 and mlo4-3 root tips grown vertically on minimal medium, the DR5:GFP expression pattern was similar to the one observed in Col-0 seedlings grown horizontally. However, accumulation of the GFP signal in cells of the quiescent center was always observed in Col-0 root tips, whereas it was consistently reduced in mlo4-3 root tips ( Figures 8G and 8H). Notably, these differences in DR5:GFP pattern and accumulation were not seen in roots of mlo4-3 and wild-type plants vertically grown on MS medium (see Supplemental Figures 10A and 10B online), indicating that the observed effects are not simply due to variation in transgene expression in the two lines but are conditional on the growth direction and medium composition. Taken together, these findings suggest that moderately altered auxin allocation in the root tip might either be the cause for or a consequence of mlo4-conditioned asymmetrical root growth.
To further assess the role of auxin in root coiling, we determined basipetal and acropetal auxin transport in mechanically nonstimulated wild type and mlo4-4 mutant roots. We observed a minor (1.5-fold) but statistically significant difference in basipetal auxin transport from the root tip into the elongation zone and beyond between Col-0 wild-type and mlo4-4 mutant plants ( Figure 9A). However, acropetal auxin transport in roots of mlo4-4 was strongly (eightfold) increased compared with Col-0 wild-type plants ( Figure 9B). Notably, mlo4 and mlo11 mutant seedlings exhibited wild-type-like root growth inhibition in response to the exogenous application of the synthetic auxin analogs 2,4-D and naphthalene acetic acid (see Supplemental Figure 11 online). We conclude that altered auxin distribution at the root tip is associated with the enhanced curling of the mlo4 mutant, which is possibly a consequence of mildly enhanced basipetal and strongly increased acropetal auxin transport.

MLO4-GFP Localizes to the Plasma Membrane and Endomembranes
The altered PIN1 levels (Figure 3), the requirement of auxin transport for root curling (Figure 7), the aberrant DR5:GFP expression pattern (Figure 8), and the increased acropetal and basipetal auxin transport ( Figure 9B) in mlo4 mutants suggest an involvement of MLO4 in auxin-related processes. Several proteins involved in polar auxin transport/distribution are known to   (Swarup et al., 2001;Wisniewska et al., 2006;Titapiwatanakun et al., 2009). To determine whether MLO4 also exhibits polarized subcellular localization, we imaged Arabidopsis cells expressing the functional MLO4-GFP fusion protein ( Figure 6A) under control of the MLO4 59 regulatory sequences in the mlo4-1 mutant background. We detected the MLO4-GFP fusion protein at the cell periphery in cells of the root cap, epidermis, and cortex ( Figure  10A), consistent with both the gene expression profiles of these root cell types (see Supplemental Figure 12 online) and our previous promoter-GUS fusion studies . This expression pattern is reminiscent of PIN2 but distinct from the PIN1 gene (see Supplemental Figure 12 online; Birnbaum et al., 2003). Confocal imaging of MLO4-GFP in intact roots and root cross sections did not reveal evidence for a potential baso-apical or lateral polarization of the fusion protein ( Figures 10A and 10B). Besides localization at the cell periphery, we also observed a prominent intracellular MLO4-GFP pool that largely colocalized with the FM4-64 membrane dye, indicating that the fusion protein is likely subject to continuous cycling between the plasma membrane and endomembrane compartments ( Figure  10C). Lack of detectable GFP fluorescence in the nucleus suggests that the majority of the intracellular GFP signal does not result from cytoplasmically localized MLO4-GFP degradation fragments. Cotreatment with FM4-64 and the vesicle trafficking inhibitor BFA led to accumulation of the majority of the intracellular MLO4-GFP signal in spherical intracellular compartments that resemble the previously described Brefeldin or BFA bodies (Orci et al., 1993;Geldner et al., 2001), further supporting that the MLO4-GFP fusion protein is subject to cycling within the endomembrane system ( Figure 10D). Similar results were obtained upon coaddition (together with FM4-64 and BFA) of the translation inhibitor cycloheximide, suggesting that it is preexisting and not de novo biosynthesized MLO4-GFP protein that accumulates in the BFA compartments (Geldner et al., 2001). Plasmolysis resulted in retraction of the GFP signal from the cell periphery and visualized fluorescent cell wall-plasma membrane connections (Hechtian strands), corroborating authentic plasma membrane localization of the MLO4-GFP fusion protein ( Figure 10E, arrowhead). Taken together, our analysis indicates plasma membrane-and endomembrane-associated MLO4-GFP pools as well as steady endocytotic cycling of the fusion protein, which is reminiscent of the cycling of PIN auxin efflux carriers (Geldner et al., 2001).

DISCUSSION
As in the case of root skewing (Rutherford and Masson, 1996), the extent of root curling is subject to natural genetic variation among Arabidopsis ecotypes (see Supplemental Figure 2 online), providing an additional example of genetically determined developmental plasticity within a plant species. A continuum of Arabidopsis mutants display aberrant root skewing, waving, and/ or curling phenotypes (e.g., Okada and Shimura, 1990;Garbers et al., 1996;Rutherford and Masson, 1996;Mullen et al., 1998;Ferrari et al., 2000;Furutani et al., 2000;Yuen et al., 2003;Santner and Watson, 2006;Korolev et al., 2007;Fortunati et al., 2008;Pandey et al., 2008;reviewed in Migliaccio et al., 2009). Consistent with a role for polar auxin transport in tropic organ curvature, several of these mutants are defective in genes i.e., mlo4-3 roots exhibit the root curling phenotype) or vertically (G) on minimal medium (0.25% sucrose, 12-h light/dark cycle). Information about the fluorescence intensities was extracted from the confocal images along the dotted white line using the Zeiss AIM imaging software and plotted as a function of the distance for horizontally grown (F) and vertically grown (H) roots of the two genotypes. Shown are representative images with corresponding graphs from one series of experiments. This analysis was repeated twice with similar results. Bars = 20 mm. Basipetal (A) and acropetal (B) auxin transport were measured by applying 3 H-IAA to the root apex and to the root-shoot junction, respectively. The amount of radioactivity was determined in distal segments of the roots at 5 to 7 h (basipetal transport) or 18 h (acropetal transport) after application of the radiolabeled synthetic auxin. For basipetal auxin transport, data presented are pooled results (obtained in three independent experiments) from 50 wild-type (Col-0) and 24 mlo4-4 seedlings. In the case of acropetal transport, three independent assays were performed, and a similar pattern was observed in each experiment. Data were pooled from two independent experiments involving 34 wild-type (Col-0) and 20 mlo4-4 seedlings. Data are shown relative to the Col-0 wild type (set as 1). Columns indicate the mean, and error bars represent SD. Asterisks indicate a statistically significant (P < 0.05) difference in comparison to Col-0 according to a Student's t test.
encoding components of the auxin biosynthesis/transport pathway, such as trp2, trp3, trp5, aux1, and pin2 (Okada and Shimura, 1990;Rutherford et al., 1998;Piconese et al., 2003). One of the most curious phenotypes is of the Arabidopsis rcn1 mutant that lacks one out of three phosphatase 2A regulatory A subunits. rcn1 roots curl more than the wild type in the presence of the auxin transport inhibitor NPA but less than the wild type in the absence of NPA (Garbers et al., 1996). The mutant has enhanced auxin transport that can be reversed to wild-type levels by treatment with NPA , consistent with the root curling phenotype of rcn1 plants.
The microtubular cytoskeleton is a critical determinant in the axial directionality of organ expansion. The handedness of organ curvature is dictated by an endogenous default structural asymmetry that, at the molecular level, is based on the spatial arrangement of transverse cortical microtubular arrays. Consequently, these arrays shape the arrangement of cellulose microfibrils, thus translating into a preferential directionality of anisotropic organ growth (Furutani et al., 2000;Thitamadee et al., 2002). Consistently, mutants affected in components of the microtubular cytoskeleton, such as map70-5, spr1, spr2, tua, tub, and wvd2, exhibit alterations in the handedness of root skewing (Furutani et al., 2000;Thitamadee et al., 2002;Ishida et al., 2007;Korolev et al., 2007;Perrin et al., 2007).
The most dramatic root curling phenotype described to date is displayed by null alleles of the MLO4 and MLO11 genes in the Col-0 ecotype (Figure 1; cf. Supplemental Figure 2 and cited literature). The dramatic root coiling of recessive mlo4 and mlo11 mutants requires a tactile stimulus ( Figure 1C), is dependent on light and nutrients (Figure 4), and requires an intact auxin transport machinery (Figures 4 and 7) but is independent of medium sucrose content (see Supplemental Figures 2B and 2C online) and is not associated with aberrant gravitropic responses (see Supplemental Figure 6 online). Additionally, mlo4 and mlo11 roots are morphologically indistinguishable from wild-type roots, do not show altered cell file arrangement or identity (see Supplemental Figure 4 online), and do not exhibit altered responsiveness to exogenous application of synthetic auxins (see Supplemental Figure 11 online). Consistent with the hypothesis that root curling, waving, and skewing are mechanistically interconnected (Simmons et al., 1995), mlo4 and, in part, mlo11 mutants also exhibited aberrant root waving and root skewing patterns (Figures 1D, 1E, and 2D). Thigmotropic reorientation of root growth directionality involves interplay between the touch and gravity sensing/response systems (Massa and Gilroy, 2003a). It was suggested that both cellular systems share common mechanosensing/transduction elements and crosstalk (Fasano et al., 2002) that might, at least in part, involve a contribution of auxin. This explains why the thigmomorphogenetic defect of mlo4 and mlo11 mutants requires intact auxin transport machinery and correlates with an altered auxin distribution pattern (Figure 8) possibly as a result of altered synthesis, transport (Figure 9), and metabolism. The light dependency of mlo4-conditioned root curling suggests interplay between light signaling and auxin transport and/or signaling, which is consistent with previous results suggesting crosstalk between light and auxin (Jensen et al., 1998;Rashotte et al., 2003;Cluis et al., 2004).
In wild-type plants, contact of root tips with physical barriers gives rise to unpredictable alternation of growth direction ( Figure  1B). By contrast, roots of mlo4 and mlo11 mutant plants, once mechanostimulated, initiate genetically predefined and stable growth direction, resulting in curling ( Figures 1B and 1C). We speculate that mlo4 and mlo11 mutants are defective in resetting the cellular machinery for subsequent perception or interpretation of growth-determining exogenous stimuli, such as touch and/or gravity. In fact, wild-type plants exhibit constant gravitropic resetting of the root tip in response to touch stimulation from encountered physical barriers (Massa and Gilroy, 2003a;Weerasinghe et al., in press). At the molecular level, this reset machinery may comprise plasma membrane-localized mechanosensors and/or ion channels and possibly include heptahelical MLO proteins for which currently there is no known biochemical function.
In an ecophysiological context, root architecture is vital to physically support the growing shoot and to maximize nutrient and water access in the typically nutrient-and water-poor environment of soil substrates. Spiral root growth (manifested as curling on a hard surface) is a morphogenetic process thought to allow roots to penetrate soil, to support circumnavigation of barriers along the main growth trajectory, to advance the establishment of contact to beneficial (growth-promoting) microorganisms, and to offer increased structural support of aerial organs. It presumably also maximizes the exploration of nutrient pools by locally increasing root tissue density within nutrient patches. Simulation of plant growth success for three different degrees of root curling is consistent with the latter concept (see Supplemental Figure 13 online). Besides nutrient and water gradients, positive gravitropism, negative thigmotropism, and root circumnutation are further driving forces decisively shaping root growth patterns. To some extent, the touch-induced root curling pattern of mlo4 and mlo11 mutant plants is reminiscent of plant organ circumnutation (Minorsky, 2003). It was suggested that root waving and curling are essentially the consequence of circumnutation, though this view is under debate (Simmons et al., 1995;reviewed in Migliaccio and Piconese, 2001;Oliva and Dunand, 2007). Regardless of the outcome of this debate, in contrast with genuine nutatory organ movements, the anisotropic root growth pattern of mlo4 and mlo11 mutant plants is not triggered spontaneously but requires a tactile stimulus. This suggests that these mutants do not show authentic circumnutation but are rather defective in part of the touch-response machinery (see above).
We conclude that MLO4 and MLO11 function as rheostats of root thigmomorphogenesis, representing novel genetically defined, regulatory components of this phenomenon. Like Arabidopsis MLO2, MLO6, and MLO12, modulators of antifungal defenses (Consonni et al., 2006), MLO4 and MLO11 reside in a discrete phylogenetic clade of the MLO family (Devoto et al., 2003) and operate in a common physiological process. However, unlike MLO2, MLO6, and MLO12, which function in part redundantly, our genetic data indicate that MLO4 and MLO11 cofunction as regulatory components since mlo4 and mlo11 single mutants both exhibit a similar (though not identical) defect in root thigmomorphogenesis (Figures 2A to 2C) that is not further exaggerated in mlo4 mlo11 double mutants ( Figure 2C). Notably, the third member of the MLO4/MLO11/MLO14 phylogenetic clade, MLO14, does not appear to contribute to this phenomenon since we observed no aberrant root growth pattern in the mlo14 single mutant and no epistatic effects of mlo14 on the phenotype of mlo4 and mlo11 mutant alleles ( Figures 1D, 1E, and 2A to 2D). It would be interesting to learn whether the variability in root curling among Arabidopsis ecotypes (see Supplemental Figure 2 online) results, at least in part, from natural genetic variation in MLO gene structure and/or expression levels.
Cofunctionality of two sequence-related genes in a common physiological process can be best rationalized by the formation of a heterooligomeric complex of MLO4 and MLO11. Initial evidence for the homooligomerization of MLO proteins results from previous Fö rster/fluorescence resonance energy transfer experiments demonstrating low levels of energy transfer between yellow fluorescent protein-and cyan fluorescent proteintagged variants of the barley MLO protein (Elliott et al., 2005). In this scenario, where MLO4 and MLO11 form a heterooligomeric protein assembly, lack of either component (such as in mlo4 or mlo11 mutants) would cause identical phenotypes owing to a qualitatively similar impairment of the functional polypeptide complex. The apparent discrepancy between the suggested model and the slight differences in the mlo4 and mlo11 root growth patterns (Figure 2A) can be best explained by assuming residual, although qualitatively different, activities of MLO11 and MLO4 homooligomeric protein complexes in mlo4 and mlo11 mutants, respectively. It is also possible that loss of either MLO4 or MLO11 would create respective MLO11 or MLO4 homodimers that confer different neomorphic effects. Functionally distinct heterooligomeric protein complexes may likewise account for the observed unequal genetic redundancy of MLO2, MLO6, and MLO12 in the context of modulating antifungal defense (Consonni et al., 2006). The formation of heterooligomeric polypeptide assemblies could thus represent a universal feature of MLO proteins. Extensive biochemical work will be required to provide unequivocal experimental evidence for the in planta existence of the proposed heterooligomeric complexes. It will also be intriguing to learn whether additional polypeptides associate with the hypothesized MLO heterooligomers.
Owing to the topological resemblance of fungal and metazoan GPCRs, MLO proteins were suggested to represent plant surface receptors that may transduce signals via the heterotrimeric G-protein complex (Devoto et al., 1999). Pharmacological and genetic experiments based on the application of G-protein activators and the expression of constitutive active and dominant negative Ga variants in barley epidermal cells did not reveal any evidence for a contribution of heterotrimeric G-protein(s) in MLO-regulated antifungal defense (Kim et al., 2002). Likewise, despite a general modulation of root coiling patterns by the gene encoding the Gb subunit ( Figure 5), genetic analysis also does not support a function of Arabidopsis MLO4 and MLO11 as GPCRs in the context of thigmomorphogenesis. Taken together, the data obtained in barley and Arabidopsis suggest that the biochemical function of MLO proteins is unrelated to the role of canonical GPCRs. While the findings do not per se exclude a potential role for MLO polypeptides as cell surface receptors, they challenge a cytoplasmic transduction cascade via the heterotrimeric G-protein complex. It also remains possible that MLO proteins exert a primary biochemical function that is distinct from ligand binding and signal transduction.
Previous studies revealed that Triticum aestivum MLO-B1, one of the three hexaploid wheat orthologs of barley Mlo, fully complements powdery mildew resistance of barley mlo mutants, suggesting that the closely related monocot ortholog is functionally equivalent to barley Mlo (Elliott et al., , 2005. Surprisingly, domain swap experiments between the two orthologous and highly sequence-related genes showed that chimeric proteins that harbor cytoplasmic domains of the respective counterpart are only partially functional, suggesting cooperative loop-loop interplay either within or between MLO polypeptides (Elliott et al., 2005). In this study, we performed domain-swap experiments between two paralogous Arabidopsis MLO members, MLO2 and MLO4, and tested functional specificity of the resulting chimeras in the context of controlling root thigmomorphogenesis. We focused on two regions of the heptahelical proteins that are known to be most divergent: the first extracellular loop and the cytoplasmic C terminus (Devoto et al., 2003). Though potential variation in transgene expression, protein stability, and/or subcellular localization of some chimeras may limit the interpretation of the data, our results suggest that MLO paralogs have a functional identity (Figures 6A and 6B). The findings further indicate that sequence characteristics of the first extracellular loop are dispensable for functional specificity, whereas the C terminus is necessary but not sufficient to confer functional specificity (Figures 6C to 6P). The latter discovery corroborates previous findings demonstrating that both identity and integrity of the cytoplasmic C terminus are critical for MLO functionality and that loop-loop cooperativity is decisive for MLO activity (Elliott et al., 2005). It is intriguing that four Cys residues in the extracellular loops, which are invariant throughout the MLO family, are required for MLO function (Elliott et al., 2005), although at least the sequence context in which three of them are imbedded is irrelevant ( Figure 6K). This finding supports the notion that the four Cys residues mainly serve a structural role, most likely by generating two disulphide bridges that stabilize a higher order structure of MLO proteins and/or their homo-and/or heterooligomeric complexes.
Intriguingly, all substances with known biological activity identified in our screen for compounds that interfere with mlo4conditioned root curling directly or indirectly affect auxin transport (see Supplemental Figure 9 online). This finding suggests that polar auxin transport is indeed the primary physiological process that is required for appropriate thigmomorphogenesis. However, we do not rule out other pathways that also contribute to this phenomenon but are either less or not amenable to pharmacological interference. Our data further indicate that the enhanced root curling in mlo4 is accompanied by subtle changes in auxin-induced gene expression patterns at the root tip ( Figure  8) together with altered expression of the auxin efflux carrier fusion proteins PIN2-GFP (see Supplemental Figure 5 online), PIN1-GFP (Figure 3), and enhanced acropetal auxin transport ( Figure 9B). Given that the biochemical functions of MLO proteins remain enigmatic, it is currently unclear how heptahelical MLO proteins connect to the root auxin transport/distribution machinery. Unlike proteins directly mediating auxin transport, such as members of the PIN family, MLO4 does not exhibit any obvious subcellular polarization ( Figures 10A to 10D).
Although inhibition of basipetal indole-3-acetic acid (IAA) transport by NPA treatment or the eir1/pin2 mutation abolishes root curling in wild-type and mlo4/mlo11 mutants, we were able to detect only small changes in root basipetal IAA transport. In comparison, the sporadic spontaneous root curvature and enhanced root waving in the mdr1/abcb19 mutant, previously shown to have wild-type levels of root basipetal transport, but reduced root acropetal transport, suggests a complex interplay between the two polar auxin transport streams that control mechanically directed root orientation Wu et al., 2007). In addition, the observation that acropetal auxin transport in the mlo4 mutant was much higher than the wild type while the auxin-induced gene expression patterns were only subtly changed indicates feedback mechanisms between auxin synthesis, movement, response, and breakdown.

Plant Growth Conditions
We used a medium composition similar to the one reported by Mirza (1987). Briefly, sterilized seeds (Turk et al., 2003) were sown on buffered (0.05% MES, pH 5.7) agar (0.5% [w/v] Phyto or Bacto agar; Research Products International) medium containing either 0.25 or 1% sucrose (in this study referred to as minimal medium). This medium mimics the poor nutrient environment of soil conditions and allows rapid growth of primary roots without initiating lateral roots and therefore is suitable for examining root curvature. For some experiments, where indicated, we used MS medium (13 MS, 0.9% [w/v] Phyto agar, and 1% [w/v] sucrose, pH adjusted to 5.8 with NaOH). Samples for comparisons were placed on the same plate to minimize variations owing to exogenous factors. Plates were incubated at 48C for 2 or 3 d and then transferred to 21 to 228C, either wrapped with aluminum foil, or placed under a light source. Considering that roots are underground organs, most of our experiments were performed under constant white light of relatively weaker intensity (30 mmol m 22 s 21 ) compared with the light conditions (160 mmol m 22 s 21 ) used by Mirza (1987). Some experiments were performed in a 12-h or 16h/8-h light/dark cycle (100 mmol m 22 s 21 ), and some experiments were performed in darkness as indicated. Most of the phenotype characterization and quantitation was performed at both the University of North Carolina at Chapel Hill and at the Max-Planck Institute in Cologne. The acropetal auxin transport assays were performed at both the University of North Carolina at Chapel Hill and at Wake Forest University.

Root Curling and Waving/Skewing Assays
Root curling was measured as described by Mirza (1987) under a dissecting microscope: 5 d after germination on minimal medium, plates were scanned, and seedling root span was measured; roots were then pulled out from the medium and lengths of the roots that produced given angles were measured using a ruler. To eliminate differences introduced by different root growth rates between genotypes, we normalized the degree of root curling by the root length. Root span was assessed by measuring the distance from the root tip to the root base on scanned micrographs. Root waving and skewing were determined on plates containing hard minimal medium (0.05% MES, pH 5.7, 0.25% [w/v] sucrose, and 1.5% [w/v] Phyto agar), held vertically for 48 h, then tilted at 458 (waving) or 358 (skewing) for 4 d. Root skewing was quantified based on the angle between a vertical line drawn downwards from the hypocotyl and the root tip. For all assays, plates were sealed with porous tape to prevent ethylene accumulation. Digital images were captured and processed using either ImageJ (http://rsb.info.nih.gov/ij/) or Adobe Photoshop software.

Chemical Genetics Screen
Chemical compound libraries were provided by Mike Robinson from Syngenta, MicroSource Discovery System (The NatProd Collection; Vogt et al., 2005), and AnalytiCon Discovery (Megabolite Natural Product Library, Berlin, Germany). Surface-sterilized seeds were stratified for 2 to 3 d at 48C. Seedlings were grown in 96-well microtiter plates (NUNC) containing approximately three to four seeds per well on 100 mL minimal medium (buffered [0.05% MES, pH 5.7] agar [0.5% w/v Bacto agar], and 0.25% sucrose). The microtiter plates were incubated at 218C in a growth chamber (Percival AR-75L) under 12-h light/dark cycles. After 3 d, chemicals dissolved in DMSO were added with a final concentration of 10 ppm or 10 mM for each compound. Root curling phenotypes were scored 4 d later.

Generation of Transgenic Plants Expressing Chimeric Proteins of MLO2 and MLO4
A 1.05-kb DNA fragment covering the 59 upstream region of MLO4 before the start codon was amplified from genomic DNA of wild-type Arabidopsis of Col-0 ecotype using Phusion enzyme (Finnzymes). Similarly, fragments of MLO2 and MLO4 were amplified from cDNA clones of MLO2 and MLO4. Primers used are listed in Supplemental Methods 1 online. Combinations of overlapping PCR fragments were fused by sequential PCR reactions, so that all chimeric genes were formed under the control of the MLO4 promoter. These PCR products were cloned into the Gateway pENTR/D-TOPO vector (Invitrogen), sequenced, and moved into the Gateway destination binary vector pGWB4 (Nakagawa et al., 2007b) by Gateway LR recombination reactions. All constructs were transformed into mlo4-1 plants via floral dip as described by Clough and Bent (1998). The transgenic plants were selected for resistance to kanamycin and hygromycin (marker genes for both resistances are present in pGWB4, ensuring selection of transgenic plants with full-size T-DNAs), and the presence of the corresponding transgene was confirmed by PCR. For each construct, 20 to 40 seedlings of six independent transgenic lines were tested at the T2 generation for root span and compared with wild-type and mlo4-1 mutant seedlings. We defined the criteria for complementation of the mlo4 root curling phenotype as follows: At least three transgenic T2 lines had to exhibit a root span that in statistical analysis is significantly different from the mlo4 mutant, but undistinguishable from the Col-0 wild type (full complementation), or at least three transgenic T2 lines had to exhibit a root span that in statistical analysis is significantly different from the mlo4 mutant and from the Col-0 wild type (partial complementation). Representative fluorescence micrographs depicting expression patterns and subcellular localization of some of the chimeric proteins were taken in the T3 generation.

Confocal Microscopy
Roots of seedlings expressing fluorescent reporter constructs grown on either minimal or MS media were stained in a solution of 4.4 mM FM4-64 (Invitrogen) in water for analysis of endocytic events for the time mentioned with the experiments or with 1 mM FM4-64 in water for >10 min as root contrasting agent. For investigation of MLO4-GFP trafficking, seedlings were incubated in 50 mM cycloheximide (Sigma-Aldrich) and/or 50 mM BFA (Invitrogen) for the times indicated in the figure legends.
Subsequently, the roots were investigated by confocal laser scanning microscopy using a Zeiss LSM 510 META-CONFOCOR2 combination setup. A 340 C-Apochromat water immersion objective (Zeiss) with a numerical aperture of 1.2 was used. GFP was excited with the 488-nm argon laser line, and excitation and emission light were separated by a HTF488 dichroic filter. GFP fluorescence was filtered through a 505-to 530-nm band-pass filter before detection. FM 4-64 was excited by the 543-nm laser line of the helium laser. Excitation and emission light were separated by a 543-nm dichroic filter. FM4-64 fluorescence was filtered by a 650-to 710-nm band-pass filter before detection. For GFP signal quantification, confocal images of roots expressing the fluorescent markers were taken as a consecutive series along the z axis. Microscope settings were kept the same for all measurements. To assess differences in signal intensity in an unbiased manner, a maximum intensity projection along the z axis was generated of the series of confocal images using the Zeiss LSM imaging software (version 3.2 SP2) and Zeiss LSM image examiner (version 3.1). Subsequently, fluorescence intensities in the projection were analyzed using ImageJ 1.37 (http://rsbweb.nih.gov/ij/). Significance levels of observed differences in fluorescence intensities were assessed with a Student's t test.

Basipetal and Acropetal Auxin Transport Assays
Basipetal and acropetal auxin transport measurements were performed in roots of Col-0 wild-type and mlo4-4 mutant seedlings as previously described (Lewis and Muday, 2009). Seedlings were grown vertically under long-day conditions (16-h/8-h light/dark cycle; 100 mmol m 22 s 21 ) for 5 d on minimal medium (0.25% sucrose). For each experiment, 10 to 15 seedlings were moved and aligned on a fresh minimal medium plate. For basipetal auxin transport assays, 3 H-IAA agar lines (final concentration of 200 nM) were applied to the aligned root tips of the seedling. Plates were incubated vertically in the dark for 5 to 7 h. Subsequently, the first 2 mm of the root tip touching the radioactive agar was discarded and the 5-mm section of the root above the site of cut was assayed for radioactivity by scintillation counting. For acropetal auxin transport assays, 3 H-IAA agar lines (final concentration of 200 nM) were applied to the root region just below the root-shoot junction. Plates were inverted upside down and vertically incubated in the dark for 18 h. Subsequently, agar lines were removed, roots straightened (as they would have responded to gravity), and 5-mm sections from the root tip were cut and measured for radioactivity by scintillation counting.

Study of Bending Direction and Differential Auxin Distribution at the Root Tip
To study bending direction and visualize differential auxin distribution in root tips, DR5rev:GFP transgenic seedlings  were grown on cover slides overlaid with a thin layer of solidified root growth minimal medium in a sterile Petri dish placed horizontally under dark for 3 d after incubation at 48C for 2 or 3 d. GFP and curvature were imaged with an Olympus IX81 microscope.

Statistical Analyses
We performed Bonferroni's multiple comparison test (GraphPad Prism; http://www.graphpad.com/prism/Prism.htm) when analyzing multiple data sets in comparison with a control and employed a two-tailed Student's t test with unequal variances when analyzing single data sets in comparison with a control. An F-test was used to comparatively analyze sample variance.

Computer Simulation of Plant Growth Success
We used the software MATLAB 7.3.0 to simulate the growth of roots in matrices containing, at the indicated density, randomly distributed re-source spots having a diameter equivalent to the width of the root or in matrices containing randomly distributed resource spots of different sizes at a given density. Three roots of the same length but different in degrees of curling tightness were generated for these tests. Summation of the number of intersections between a resource spot (counted only once) over the length of each root equals an estimate for plant growth success. The original code is provided in Supplemental Figure 13E online and additionally as a text file in Supplemental Methods online.

Supplemental Data
The following materials are available in the online version of this article. technical assistance and Haiyang Wang, Takashi Hashimoto, Renze Heidstra, Ben Scheres, and Jiri Friml for providing genetic material as well as Mike Robinson for making available a core set of the Syngenta chemical library.