Arabidopsis genome contains three RACK1 homologous genes, designated as RACK1A, RACK1B and RACK1C, respectively 15. Within the RACK1 gene family, mutant alleles for only RACK1A have been reported previously 15. We report here the isolation and characterization of rack1b and rack1c mutant alleles. By searching the Salk Institute sequence-indexed insertion mutant collection http://signal.salk.edu/cgi-bin/tdnaexpress, we obtained two independent T-DNA insertional alleles for each RACK1 gene. All alleles are in the Columbia (Col-0) ecotypic background. We designated the two mutant alleles for RACK1B as rack1b-1 and rack1b-2, respectively. In rack1b-1 allele, the T-DNA was inserted in the second exon of RACK1B gene, and in the rack1b-2 allele, the T-DNA was inserted in the first intron (Figure 2A). RT-PCR analysis indicated that the full-length transcript of RACK1B was absent in both alleles (Figure 2B), implying that they are likely loss-of-function alleles. Unlike rack1a mutants, rack1b mutants do not display any apparent developmental defects (Figure 2C). We designated the two mutant alleles for RACK1C as rack1c-1 and rack1c-2, respectively (Figure 2D). In rack1c-1 allele, the T-DNA was inserted in the second exon of RACK1C gene, and in the rack1c-2 allele, the T-DNA was inserted in the 5'-UTR region. RT-PCR analysis indicated that the full-length transcript of RACK1C was absent in both alleles (Figure 2E), implying that they are likely loss-of-function alleles. Similar to rack1b mutants but unlike rack1a mutants, rack1c mutants do not display any apparent defects in plant development (Figure 2C).

Previously, we showed that loss-of-function mutations in one member of Arabidopsis RACK1 gene family, RACK1A, resulted in multiple defects in plant development 15. Because loss-of-function alleles of RACK1B and RACK1C did not display apparent defects in plant development, we wanted to test if mutations in RACK1B or RACK1C can enhance the developmental defects of rack1a mutants. Therefore, we generated rack1a-1 rack1b-2 and rack1a-1 rack1c-1 double mutants. One of the most dramatic phenotypes observed in rack1a single mutants was the reduced number of rosette leaves 15. Therefore, we grew single and double mutants together with wild-type (Col) under identical, short-day conditions with 10/14 h photoperiod, counted the number of rosette leaves in double mutants, and compared it with Col and rack1a-1 single mutant. We found that while rack1b-2 and rack1c-1 single mutants produced wild-type number of rosette leaves, both rack1b-2 and rack1c-1 significantly enhanced the phenotype of reduced number of rosette leaves of rack1a-1 single mutants (Figure 3A, B). When plants were grown under 10/14 h photoperiod for 48 days, wild-type produced approximately 30 rosette leaves, whereas rack1a-1 single mutant produced 22 rosette leaves. Under these conditions, rack1a-1 rack1b-2 and rack1a-1 rack1c-1 double mutants only produced about 16 and 19 rosette leaves, respectively (Figure 3B). The rate of rosette leaf production was reduced approximately 27% and 14%, respectively, in rack1a-1 rack1b-2 and rack1a-1 rack1c-1 double mutants, compared with rack1a-1 single mutant (Figure 3C). We also examined the rosette size by measuring the diameter of rosette of each genotype. Similar to the situation of number of rosette leaves, the diameter of rosette was significantly reduced in rack1a-1 single mutant, compared with wild-type plants, and such reduction was further enhanced in rack1a-1 rack1b-2 and rack1a-1 rack1c-1 double mutants (Figure 3D). Interestingly, no synergistic effect was observed between rack1b-2 and rack1c-1 mutations. Statistically, rack1b-2 rack1c-1 double mutants phenocopied parental single mutants and displayed wild-type traits of these phenotypes (Figure 3A, B).

Subsequently, we generated rack1a-1 rack1b-2 rack1c-1 triple mutant. Very few triple mutants could survive in soil. For those survived, they were extremely slow in growth and development, and produced fewest rosette leaves and smallest rosette size among all genotypes examined (Figure 3A–D). Not surprisingly, the rate of rosette leaf production in the triple mutant was the slowest among all genotypes examined (Figure 3C). Because rack1a-1 rack1b-2 rack1c-1 triple mutants could not survive to maturity to produce seeds, these triple mutants were maintained in plants homozygous for the rack1b-2 and rack1c-1 loci and heterozygous for the rack1a-1 locus. Because rack1b-2 rack1c-1 double mutants had wild-type morphology whereas rack1a-1 rack1b-2 rack1c-1 had extreme pleiotropic phenotype, rack1a-1 rack1b-2 rack1c-1 triple mutants can be readily picked up from the segregating progeny of plants homozygous for the rack1b-2 and rack1c-1 loci and heterozygous for the rack1a-1 locus.

Genetic analysis indicated that loss-of-function mutations in RACK1A affect the production of rosette leaves, and that the effect of rack1a-1 mutation can be enhanced by the rack1b-2 or rack1c-1 mutation or both (Figure 3). We wanted to extend our analysis to non-aerial organs by examining the impact of these mutations on root development. We measured the length of primary root and counted the number of lateral root and used them as parameters of root development and root architecture. We found that the length of primary root of rack1a-1 mutant was slightly shorter than that of wild-type whereas rack1b-2 and rack1c-1 mutants had wild-type length of primary root (Figure 4A). The length of primary root was further shortened in rack1a-1 rack1b-2 and rack1a-1 rack1c-1 double mutants, compared with that in rack1a-1 single mutant (Figure 4A), indicating that rack1b-2 and rack1c-1 mutations can also enhance the effect of rack1a-1 mutation on primary root growth. Similar to the situation of primary root, rack1a-1 mutant produced fewer lateral roots than wild-type whereas rack1b-2 and rack1c-1 mutants had wild-type number of lateral roots (Figure 4B). As expected, rack1b-2 and rack1c-1 mutations enhanced the lateral root phenotype of rack1a-1 mutant (Figure 4B). Among all genotypes examined, the rack1a-1 rack1b-2 rack1c-1 triple mutant produced the shortest primary root and did not produce any lateral root under our assay conditions (Figure 4A, B).

Genetic analyses indicated that there is unequal genetic redundancy among three Arabidopsis RACK1 genes in regulating rosette leaf production and root development, and that RACK1A is likely a non-dispensable gene in this small gene family. Although RACK1B and RACK1C are likely dispensable, they still contribute significantly to the overall activity of RACK1 genes in regulating plant development, as revealed by the phenotypes of double and triple mutants. We wanted to further explore the mechanism of the unequal genetic redundancy of RACK1 genes. Firstly, because RACK1B and RACK1C are highly similar (about 90% identity) to RACK1A at the amino acid level (Figure 1), we wanted to test if RACK1B and RACK1C are in principle functionally equivalent to RACK1A. We reasoned that if RACK1B and RACK1C are indeed functionally equivalent to RACK1A, one would expect that overexpression of RACK1B or RACK1C complements the developmental defects of rack1a mutants. Therefore, we generated transgenic lines overexpressing RACK1B or RACK1C in the rack1a mutant background using the CaMV 35S promoter. As a control, we generated transgenic plants overexpressing RACK1A in rack1a mutant background. At least two independent transgenic lines were analyzed for each transformation. Overexpression of the transgene in these lines was confirmed by RT-PCR analysis (Figure 5A). We examined the same parameters described above, namely the number of rosette leaves, the length of primary root and the number of lateral roots in the transgenic lines overexpressing each RACK1 gene and compared them with those in Col and rack1a single mutants. As expected, overexpression of RACK1A fully complemented the mutant phenotype of rack1a mutant (Figure 5B–D). Similarly, we found that overexpression of RACK1B or RACK1C fully restored rack1a mutant to wild-type morphology, evident by the wild-type number of rosette leaves, wild-type length of primary root and wild-type number of lateral roots in transgenic lines (Figure 5B–D).

Because constitutive expression of RACK1B or RACK1C could efficiently complement rack1a mutant's developmental defects, these results implied that RACK1B and RACK1C are likely in principle functionally equivalent to RACK1A, and that the unequal genetic redundancy of RACK1 genes is likely due to the difference in their expression patterns or expression levels. Therefore, we sought additional evidence that would shed light on the relationship between RACK1 genes. We examined the expression patterns of RACK1A, RACK1B and RACK1C in various tissues and organs of young seedlings and mature plants by RT-PCR. We found that all three Arabidopsis RACK1 genes were expressed widely in all tissues examined (Figure 6A). These results are largely consistent with the results of analysis of RACK1 gene promoter:β-glucuronidase (GUS) transcriptional reporter lines 15. By using RT-PCR, we noticed that in any given tissues or organs examined, the transcript level of three RACK1 genes were different, with a general trend of RACK1A > RACK1B > RACK1C (Figure 6A).

In order to quantify the difference in transcript level of RACK1A, RACK1B and RACK1C genes, we used quantitative real-time PCR to more accurately compare the transcript level of three RACK1 genes in different tissues and organs of wild-type Col plants. We selected the samples of shoots and roots of 4 d- and 7 d-old light-grown seedlings and rosette leaves and roots of mature plants for quantitative real-time PCR analysis. We found that consistent with the result of RT-PCR analysis, the transcript level of RACK1C was the lowest and that of RACK1A was the highest among three RACK1 genes, with a trend of RACK1A > RACK1B > RACK1C in all samples examined (Figure 6B). For example, the transcript level of RACK1A was about 5-fold higher than that of RACK1C in the roots of 4 d-old, light-grown seedlings (Figure 6B). In this sample, the transcript level of RACK1B was approximately 2-fold higher than that of RACK1C.

The analysis of the expression patterns and transcript level of three RACK1 genes in various tissues and organs supported the view that the unequal genetic redundancy of RACK1 genes is likely due to the difference in the gene expression level. However, other possibilities may also exist. For example, as reviewed by Briggs et al. (2006), cross-regulation is another mechanism that attributes to the unequal genetic redundancy of some homologous genes 16. Because RACK1A, RACK1B and RACK1C are approximately 90% identical to each other at the amino acid level, we were unable to obtain antibodies that can specifically recognize each RACK1 protein. Therefore, in this study, we examined the impact of loss-of-function mutations of each RACK1 gene on the transcription of the other two RACK1 genes. Further, we examined the impact of combination of loss-of-function mutations of two RACK1 genes on the transcription of the other RACK1 gene. Specifically, we examined the transcript level of RACK1A in rack1b and rack1c single and double mutants, the transcript level of RACK1B in rack1a and rack1c single and double mutants, and the transcript level of RACK1C in rack1a and rack1b single and double mutants, and compared with their transcript levels in wild-type. For this analysis, we used the 4.5 d-old, light-grown whole seedlings. By using RT-PCR, we noticed that the transcript level of RACK1B was reduced in rack1a and rack1c single and double mutants (Figure 7A). Similarly, the transcript level of RACK1C was reduced in rack1a and rack1b single and double mutants (Figure 7A). However, we did not observe a dramatic reduction of the transcript level of RACK1A in rack1b and rack1c single and double mutants, compared with that in wild-type (Figure 7A). Because the transcript level of RACK1A is the most abundant among three RACK1 homologous genes and the conditions used for RT-PCR (e.g. PCR at 28 cycles) may not allow us to visualize any differences in RACK1A transcript level among different samples, subsequently we used quantitative real-time PCR to more accurately compare the transcript level of three RACK1 genes in wild-type and mutants. We found that the transcript level of any given RACK1 gene was reduced in the loss-of-function alleles of each and both of the other two RACK1 genes (Figure 7B).

RACK1 gene is evolutionarily conserved in diverse organisms. Although the research interest in RACK1 has grown exponentially since its discovery 1 and RACK1 is now viewed as a multi-functional, versatile scaffold protein in mammals and in yeasts (reviewed in 34), the function of RACK1 in plants remains poorly understood. We are just starting to have some hints about its potential functions in plants. Preliminary analysis suggested RACK1 may mediate multiple hormone responses and developmental processes in Arabidopsis 15. In this study, we focused on the two characteristic developmental defects of rack1a mutants, namely the reduction in rosette leaf production and the reduction in primary root growth and lateral root formation, to study the function of RACK1B and RACK1C and the genetic relationship between RACK1 homologous genes in plant development. We demonstrated that RACK1 genes are critical regulators of plant development and are essential for plant survival. Simultaneous disruption of the function of all three RACK1 genes results in lethality. Thanks to the unequal genetic redundancy of RACK1 genes, we are still able to study the role of RACK1 genes in plant development. The rack1a single mutants, rack1a rack1b and rack1a rack1c double mutants all display developmental defects and are viable. Therefore, these mutants can be treated as "weak alleles" of rack1 mutants. Now that we have identified RACK1 genes as critical regulators of plant development and all "weak alleles" of rack1 mutants are available, future studies should focus on the elucidation of the molecular mechanism by which RACK1 genes regulate plant development, including rosette leaf production, root growth and lateral root formation. Because rack1a mutants have also been shown to display altered responses to hormones 15, it remains unclear if the developmental defects observed in rack1 mutants are due to the altered responses to multiple hormones and if there is also unequal genetic redundancy of RACK1 genes in mediating hormone responses. This is a fertile area that is worth further investigation.

Genetic redundancy of homologous genes is thought to be due to gene duplication events during the evolution of the organism. Between homologous genes, genetic redundancy can be classified as full redundancy, partial redundancy, and unequal redundancy 16. While full redundancy and partial redundancy have been documented in numerous cases, unequal genetic redundancy has just begun to be recognized as a common phenomenon of genetic relationship of homologous genes 16. Unlike non-plant organisms whose genomes contain only a single RACK1 gene, some plant genomes contain more than one RACK1 genes (Figure 1). In particular, the Arabidopsis genome contains three RACK1 genes, which share the similar gene structure with two exons and one intron, and encode three highly similar proteins with approximately 90% identity at the amino acid level 15. However, the relationship between three Arabidopsis RACK1 homologous genes has been unknown. Previously, we showed that loss-of-function mutation in one member of Arabidopsis RACK1A genes, RACK1A, conferred multiple defects in plant development 15. Here we show that loss-of-function mutations in RACK1B or RACK1C do not confer apparent developmental defects (Figure 2). These results suggested that RACK1B and RACK1C are likely dispensable in plant development. However, we found that although rack1b and rack1c mutants displayed wild-type morphology, rack1b and rack1c can strongly enhance the developmental defects of rack1a mutants (Figure 3, Figure 4). These results suggested that RACK1B and RACK1C still contribute significantly to the overall activity of RACK1 genes. Because the significance of the RACK1B and RACK1C is determined via the mutants, not directly in the wild-type plants, it is also possible that in the wild-type plants, all the function of RACK1 genes is explicated by RACK1A with no contribution from RACK1B or RACK1C and the these latter can play a role only if RACK1A is not present (e.g. in the rack1a mutant). Nonetheless, the behaviors and relationship of rack1 mutants satisfy the key criteria for RACK1 genes being unequally redundant homologous genes 16.

The unequal genetic redundancy is caused by many factors. Among them, the difference in gene expression pattern, expression level and cross-regulation of homologous genes have been recognized as major determinants 16. The unequal genetic redundancy of some homologous genes is mainly due to the difference in expression pattern and/or expression level. For example, CAULIFLOWER (CAL) is closely related in sequence to APETALA1 (AP1), but AP1 and CAL regulate the formation of floral meristem in an unequally redundant manner because AP1 is expressed at much higher level than CAL throughout sepal and petal development 17. The unequal genetic redundancy of homologous genes can also be primarily due to the cross-regulation. For example,LONG HYPOCOTYL 5 (HY5) and its close homolog HY5 HOMOLOG (HYH), both of which are regulators of photomorphogenesis, are a pair of unequally redundant genes with similar expression patterns and levels 1618, but a normal protein expression and activity of HYH was dependent on the presence of a functional HY5 18.

In order to get insight into the mechanism of unequal genetic redundancy of three RACK1 genes, we examined each of these possibilities. Firstly, we showed that RACK1B and RACK1C are likely in principle functionally equivalent to RACK1A, because overexpression of either RACK1B or RACK1C under the constitutive CaMV 35S promoter fully complemented the developmental defects of rack1a mutants (Figure 5). Ideally, it would be advantageous to use the native RACK1A promoter to assess the extent of functional equivalency. Nonetheless, results from our complementation studies indicated that overexpression of RACK1B or RACK1C can restore rack1a mutant to wild-type equally well as overexpression of RACK1A, supporting the view that RACK1B and RACK1C likely function similarly as RACK1A. These results implied that the unequal genetic redundancy of RACK1 genes is likely due to the difference in gene expression pattern and/or expression level, rather than the difference in protein sequence or activity. To examine this possibility directly, we found that three RACK1 genes are widely expressed in various tissues and organs in young seedlings and in mature plants (Figure 6). However, RACK1 genes are expressed at different levels with a general trend of RACK1A > RACK1B > RACK1C in all tissues and organs examined (Figure 6). These results supported the view that the difference in gene expression level attributes to the unequal genetic redundancy of RACK1 genes in plant development. However, these results cannot rule out the possibility that the expression of each RACK1 gene may also be restricted to certain cell types. For example, BRL1 and BRL3 are homologous to BRI1, a receptor for brassinosteroid (BR), and function as BR receptors in vascular differentiation in Arabidopsis 19. It was found that BRI1 is ubiquitously expressed in growing cells, but the expression of BRL1 and BRL3 is restricted to non-overlapping subsets of vascular cells. Future expression analysis at cell level (e.g. by in situ hybridization and reporter GFP analyses) may help address the possibility of cell type-specific expression of RACK1 genes.

We also explored the possibility of cross-regulation by examining the transcript level of each RACK1 gene in the loss-of-function alleles of each or both of the other two RACK1 genes. We found that the transcript level of any given RACK1 gene was reduced in the single or double mutants for the other two RACK1 genes (Figure 7). Therefore, both the difference in gene expression level and the cross-regulation contribute to the unequal genetic redundancy of RACK1 genes. Unlike HY5 and HYH, for which the expression of the duplicate gene (HYH) depends on the presence of the ancestral gene (HY5) 18, RACK1 homologous genes mutually depend on each other for reaching full expression, adding another level of complexity for the unequal genetic redundancy. The molecular basis of such mutual cross-regulation of RACK1 genes is presently unknown. It would be interesting to test if RACK1 proteins can work together in a complex, for instance, through homo- and hetero-dimerization.

All mutants are in the Arabidopsis Columbia (Col-0) ecotype background. The rack1a-1 and rack1a-2 mutants have been reported previously 15. Plants were grown in 5 × 5 cm pots containing moistened 1 : 3 mixture of Sunshine Mix #1 (Sun Gro Horticulture Canada Ltd., http://www.sungro.com) and Metro-Mix 220 (W.R. Grace & Co., http://www.grace.com) with 10/14 h (short-day conditions) or 14/10 h (long-day conditions) photoperiod at approximately 120 μmol m^-2 s^-1 at 23°C.

The T-DNA insertion mutants of RACK1B (At1g48630), rack1b-1 (SALK_117422) and rack1b-2 (SALK_145920), and the T-DNA insertion mutants of RACK1C (At3g18130), rack1c-1 (SAIL_199_A04) and rack1c-2 (SALK_017913), were identified from the SALK T-DNA Express database http://signal.salk.edu/cgi-bin/tdnaexpress. For the SALK T-DNA insertional mutants 20, the insertion was confirmed by PCR using RACK1B-specific primers (5'-TCTCGACCTCAAACCCTG-3' and 5'-GAGAAGACTTTAGAGTCGATGGA-3') or RACK1C-specific primers (5'-ATCTCTCGCTCTGTTACGC-3' and 5'-ACAATACTGACGCAGTCTGG-3') and a T-DNA left border-specific primer JMLB1 (5'-GGCAATCAGCTGTTGCCCGTCTCACTGGTG-3'). For the SAIL T-DNA insertion mutants 21, a different T-DNA left border-specific primer, GarlicLB3 (5'-TAGCATCTGAATTTCATAACCAATCTCGATACAC-3'), was used. The absence of full-length transcript of RACK1B or RACK1C in these alleles was confirmed by RT-PCR.

Double mutants between rack1a-1 and rack1b-2 or rack1c-1 were generated by crossing rack1b-2 or rack1c-1 into rack1a-1 single mutant and isolated in the F2 progeny by PCR genotyping. Similarly, double mutants between rack1b-2 and rack1c-1 were generated by crossing rack1c-1 into rack1b-2 single mutant and isolated in the F2 progeny by PCR genotyping. For simplicity, the rack1a rack1b, rack1a rack1c and rack1b rack1c double mutant nomenclatures in this report refer specifically to the rack1a-1 rack1b-2, rack1a-1 rack1c-1 and rack1b-2 rack1c-1 mutants, respectively.

Triple mutant among rack1a-1, rack1b-2 and rack1c-1 was generated by crossing rack1b-2 rack1c-1 into rack1a-1 rack1b-2 double mutants. Because rack1a-1 rack1b-2 rack1c-1 triple mutants cannot survive in soil to maturity, they are maintained in plants homozygous for the rack1b-2 and rack1c-1 loci and heterozygous for the rack1a-1 locus. The status of triple mutant was confirmed by PCR genotyping.

The full-length open-reading frames of RACK1A (At1g18080), RACK1B and RACK1C were amplified from a cDNA library made from seedlings grown in light for 10 d, cloned into the pENTR/D-TOPO vector (Invitrogen, http://www.invitrogen.com), and then subcloned into Gateway plant transformation destination binary vector pB2GW7 22 by LR recombination reactions. In these constructs, the expression of RACK1A, RACK1B or RACK1C was driven by the 35S promoter of the Cauliflower mosaic virus. Binary vectors were transformed into rack1a-1 or rack1a-2 mutants by Agrobacterium-mediated transformation 23. At least 16 independent transgenic lines were selected from each transformation, and two to four representative lines were used for further studies. The expression of transgene was examined by RT-PCR.

For tissue/organ expression pattern analysis, total RNA was isolated from different parts of seedlings or mature plants, using the TRIzol reagent (Invitrogen). cDNA was synthesized from 1 μg total RNA by oligo(dT)₂₀-primed reverse transcription, using THERMOSCRIPT RT (Invitrogen). RACK1A-specific primers (5'-GGCATCTCCAGACACCGAAA-3' and 5'-GCAGAGAGCAACGACAGC-3'), RACK1B-specific primers (5'-TCTCGACCTCAAACCCTG-3' and 5'-GAGAAGACTTTAGAGTCGATGGA-3'), and RACK1C-specific primers (5'-ATCTCTCGCTCTGTTACGC-3' and 5'-ACAATACTGACGCAGTCTGG-3') were used to amplify the transcripts of these three genes, respectively. The expression of ACTIN2 (amplified by primers 5'-GTTGGGATGAACCAGAAGGA-3' and 5'-GAACCACCGATCCAGACACT-3') was used as a control in PCR reactions. For the examination of the transcript level of RACK1A, RACK1B and RACK1C in the T-DNA insertional mutants or in the transgenic lines, 10 d-old, light-grown seedlings were used for total RNA isolation.

For the quantitative analysis of RACK1A, RACK1B and RACK1C transcript levels in the different tissues/organs of wild-type Col plants or in the rack1a-1, rack1b-2 and rack1c-1 single and double mutants, real-time PCR was performed. RACK1A-specific real-time PCR primers (5'-CTGAGGCTGAAAAGGCTGACAACAG-3' and 5'-CTAGTAACGACCAATACCCCAAACTC-3'), RACK1B-specific real-time PCR primers (5'-GGTTCTACTGGAATCGGAAACAAGACC-3' and 5'-CTAGTAACGACCAATACCCCAGACCC-3'), and RACK1C-specific real-time PCR primers (5'-GCAGAGAAGAATGAAGGTGGTGT-3' and 5'-CTAGTAACGACCAATACCCCAGACCC-3') were used. The expression of ACTIN2 (amplified by real-time PCR primers 5'-CCAGAAGGATGCATATGTTGGTGA-3'and 5'-GAGGAGCCTCGGTAAGAAGA-3') was used to normalize the expression of each gene. The quantitative real-time PCR was performed using the MJ MiniOpticon real-time PCR system (Bio-Rad, http://www.biorad.com) and IQ SYBR Green Supermix (Bio-Rad).

The number of rosette leaves was collected from wild-type Col and mutant plants grown under 10/14 h or 14/10 h photoperiod with approximately 120 μmol m^-2 s^-1 at 23°C. At least four plants from each genotype were used in each experiment, and the experiment was repeated twice. The rate of rosette leaf production was expressed as the number of rosette leaves divided by the age of plant.

Seedlings were grown on MS/G plates consisting of 1/2 Murashige & Skoog (MS) basal medium supplemented with vitamins (Plantmedia, http://www.plantmedia.com), 1% (w/v) sucrose and 0.6% (w/v) phytoagar (Plantmedia), with pH adjusted to 5.7 with 1N KOH. The plates were placed under 14/10 h photoperiod with approximately 120 μmol m^-2 s^-1 at 23°C with a vertical orientation for monitoring root growth. The length of primary and the number of lateral roots were collected from at least 15 seedlings each genotype.