A zinc knuckle protein that negatively controls morning-specific growth in Arabidopsis thaliana

Abstract

Growth in plants is modulated by a complex interplay between internal signals and external cues. Although traditional mutagenesis has been a successful approach for the identification of growth regulatory genes, it is likely that many genes involved in growth control remain to be discovered. In this study, we used the phenotypic variation between Bay-0 and Shahdara, two natural strains (accessions) of Arabidopsis thaliana, to map quantitative trait loci (QTL) affecting light- and temperature-regulated growth of the embryonic stem (hypocotyl). Using heterogeneous inbred families (HIFs), the gene underlying one QTL, LIGHT5, was identified as a tandem zinc knuckle/PLU3 domain encoding gene (At5g43630; TZP), which carries a premature stop codon in Bay-0. Hypocotyl growth assays in monochromatic light and microarray analysis demonstrate that TZP controls blue light associated growth in a time-of-day fashion by regulating genes involved in growth, such as peroxidase and cell wall synthesis genes. TZP expression is phased by the circadian clock and light/dark cycles to the beginning of the day, the time of maximal growth in A. thaliana in short-day conditions. Based on its domain structure and localization in the nucleus, we propose that TZP acts downstream of the circadian clock and photoreceptor signaling pathways to directly control genes responsible for growth. The identification of TZP thus provides new insight into how daily synchronization of growth pathways plays a critical role in growth regulation.

The embryonic stem or hypocotyl is an excellent model for studying both internal and external factors controlling growth in plants (1). Genetic screens in common laboratory accessions have yielded direct molecular insight into how light- and hormone-dependent signaling pathways interact with the circadian clock to regulate the final length of the hypocotyl (1). The power of the hypocotyl assay is its simplicity, as well as its obvious meaningfulness. When germinating seeds are exposed to low levels of light, such as those caused by a covering layer of debris, the hypocotyl has to grow for a while. Only after the surface has been broken by the tip of the hypocotyls can the embryonic leaves, the cotyledons, unfold. Conversely, if a seed has fallen on open ground, there is no need for the hypocotyls to be particularly long. Because of the ease and reproducibility with which hypocotyl length can be measured in thousands of individuals, it has also been a powerful model in mapping genes with more subtle effects on light and hormone regulated growth, by using methods of quantitative genetics (2). Multiple light signaling genes controlling hypocotyl length have been characterized in quantitative trait locus (QTL) studies (3–6).

In this study, we use the hypocotyl assay to identify QTL controlling growth in 2 light and 2 temperature conditions. We identified a recessive large effect QTL on chromosome five controlling 40% of the growth variation segregating in Recombinant Inbred Lines (RILs) derived from the Bay-0 and Shahdara accessions of Arabidopsis thaliana. The QTL was fine mapped and the causal factor shown to be a mutation affecting a tandem zinc knuckle/PLUS3 protein (At5g43630; TZP), which is encoded by a single-copy gene in all completely sequenced plant genomes. We show that TZP acts downstream of the circadian clock and light signaling, directly regulating blue light-dependent, morning (dawn)–specific growth, during seedling development and beyond. Based on its nuclear localization and its novel domain structure, we argue that TZP functions at the transcriptional level to control growth-promoting pathways. TZP represents a new component of the growth pathway that was not previously identified by using traditional genetic screens.

Results and Discussion

Mapping QTL for Hypocotyl Elongation in the Bay-0 × Shahdara RIL Population.

Intraspecific variation provides a fertile source of genetic combinations that can be used to map new genes (or new alleles) involved in complex traits such as growth. To investigate natural variation for hypocotyl elongation response to light and temperature, we phenotyped a core set of 164 RILs from the Bay-0 × Shahdara cross in 4 different environments combining 2 white-light [17 μmol/m⁻²s⁻¹ (L1) and 10 μmol/m⁻²s⁻¹ (L2)] and 2 temperature (22°C and 26°C) conditions [supporting information (SI) Fig. S1A]. The parental phenotypes reveal that in our conditions Shahdara responds poorly to temperatures above 22°C or light below 17 μmol/m⁻²s⁻¹, or a combination of both. In contrast, Bay-0 responds strongly to both temperature and light, with a synergistic interaction between both factors. Variation among the RILs seems to follow parental variation with signs of bimodality [especially at 26°C/L1 and 26°C/L2] suggesting the segregation of some large-effect QTL. Transgression was also prevalent in all conditions and in both directions. Overall, genotypic variation was significant in each environment and broad-sense heritability of the trait was accordingly high, greater than 70% (www.inra.fr/vast/Files/Loudet_PNAS_SITables.xls). Although RIL response to contrasted temperature and light treatments was significant (P < 0.001), RIL × light interactions were not significant at either 22°C or 26°C, and the RIL × temperature interaction was significant only under L1 (data not shown). This indicates that most of the phenotypic variation between RILs is stable and conserved across environments.

The genetic architecture of variation in hypocotyl elongation under these environmental conditions is presented in Fig. S1B. Two loci with major effects are detected across all environments and called LIGHT1 and LIGHT5, whereas the remaining loci called HYP are specific to a single environment with more subtle phenotypic contributions (only 4% each). The identification of these two major effect QTL is in accordance with the meager RIL × environment interactions found. The LIGHT5 locus explains over 40% of the variance, with no LIGHT5 × environment interaction found in any condition (data not shown). Its negative allelic effect is predicted to represent a combined 2.1 to 2.5 mm increase in hypocotyl elongation contributed by Shahdara alleles (Sha) relative to the Bay-0 alleles (Bay). In contrast, Sha alleles at LIGHT1 are responsible for a decrease in hypocotyl elongation compared with Bay alleles, but with a relatively smaller phenotypic contribution (explaining 25–30% of the variance). The opposite allelic effects of LIGHT1 and LIGHT5 are responsible for most of the transgression observed in each environment. LIGHT1 interacts with temperature under either L1 or L2 conditions and with light at either 22 °C or 26 °C (data not shown). There is no significant epistatic relationship between LIGHT1 and LIGHT5, or among any other pair of loci.

Confirmation and Fine-Mapping of LIGHT5 to 3 Candidate Genes.

Identifying the Quantitative Trait Gene (QTG) underlying a QTL is a challenging task that requires several independent lines of proof that a gene is linked to a trait of interest and that variation in this gene explains the trait (7). The most general approach is fine-mapping to a very small physical candidate interval, which in the best case allows immediate identification of candidate polymorphisms [quantitative trait nucleotide (QTN)] within the causal gene or regulatory regions (8). For most QTL and situations in A. thaliana, phenotyping for a QTL effect remains much more limiting than genotyping many individuals. Therefore, an efficient strategy for fine-mapping is to first isolate recombinants within a segregating nearly isogenic line based on genotype alone and then to phenotypically interrogate only the informative ones (in successive rounds) to reduce the candidate interval to the gene level (9).

We followed the HIF strategy to build nearly isogenic lines from a RIL (RIL350) that was segregating solely for the LIGHT5 region. Comparing plants homozygous for the Sha allele with plants homozygous for the Bay allele at the QTL region (in an otherwise identical genetic background) confirmed the phenotypic impact of LIGHT5 on hypocotyl elongation (Fig. 1A). HIF350-Sha hypocotyls are consistently 1.6 mm longer than those of HIF350-Bay (slightly less than predicted by the QTL analysis). The analysis of heterozygous plants showed that the Sha allele of LIGHT5 is fully dominant over the Bay allele (Fig. 1A). The phenotypic effect observed was identical when first fixing alternate genotypes at the QTL region and then comparing the phenotypes of the descendants produced by those homozygous plants (“fixed progeny”) or when directly studying the segregating descendants of a heterozygous plant (“progeny testing”). This precludes any maternal phenotypic effect and demonstrates a direct control of the phenotype expressed in seedlings by the LIGHT5 alleles.

Fig. 1.

Confirmation and fine-mapping of the LIGHT5 QTL to 3 genes, among which At5g43630 is highly polymorphic between Bay-0 and Shahdara. (A) Confirmation of LIGHT5 using HIFs. The Sha allele is fully dominant over the Bay allele. Two rounds of recombinant screening from HIF350 followed (B). Horizontal marks on the chromosomes are markers with physical position in Mb indicated to the right. Red arrows indicate the approximate position of different recombination events that were individually tested in rHIF to establish the QTL position. Fine-mapping identified a 7kb-region containing three genes (C). Gray boxes along the vertical axes represent exons from three genes highlighted by vertical arrows. Horizontal red arrows indicate the exact physical position of the last five recombinants defining the QTL candidate region. Amino acid changes between Bay-0 and Shahdara within the interval are presented in the table and linked to their respective physical position along the gene model. (D) Model of the protein structure of TZP (At5g43630).

Screening for recombinants by genotyping 600 plants descended from the initial RIL350 individual (heterozygous over the whole QTL region) allowed us to identify 80 recombinants (recombined HIF, or rHIF) over the ≈1.9 Mb heterozygous region. Of these rHIF, 26 were individually investigated in successive rounds of progeny testing to score for the presence (or absence) of the QTL effect caused by the remaining interval. This first screen allowed us to narrow the candidate region to less than 300 kb between markers at 17.405 and 17.692 Mb (Fig. 1B). Screening 4,000 descendants from one of the positive rHIFs found in the previous step allowed us to identify 73 new rHIFs within the 300 kb candidate interval; 25 of these were individually progeny tested to define a smaller interval containing the QTL (Fig. 1B). Results were consistent among rHIFs, and the last 5 recombinants delimited a 7 kb interval, from 17.545 Mb (1 recombinant) to 17.552 Mb (4 independent recombinants; Fig. 1C). Three predicted genes, At5g43630, At5g43640, and At5g43650, are at least partially included in the 7-kb candidate region. We crossed two independent rHIFs with appropriate genotypes (see Fig. S2) to generate a line that was segregating only for the candidate region (and fixed to the north and south), following a strategy suggested by Kroymann and Mitchell-Olds (10) that we named “advanced rHIF (arHIF) cross”. This approach confirmed that the 17.545–17.552-Mb interval was sufficient to recapitulate the LIGHT5 phenotype (Fig. S2).

Sequencing the 7 kb interval in Bay-0 and Shahdara revealed dozens of SNPs and several indels, many of which resulted in nonsilent changes in the coding regions of the three candidate genes. Eight, one, and five SNPs caused amino acid changed in At5g43630, At5g43640, and At5g43650, respectively. Two single amino acid deletions and one larger deletion were discovered in At5g43630. Finally, an 8-bp insertion in Bay-0 caused a frameshift and a premature stop within the coding region for the predicted PLUS3 domain of At5g43630 (Fig. 1 C and D).

Identification of the LIGHT5 Causal Gene as Tandem Zinc Knuckle/PLUS3 (TZP).

The 3 genes in the LIGHT5 interval are annotated as encoding a tandem zinc knuckle/PLUS3 (TZP, At5g43630), a 40S ribosomal protein (RPS15E, At5g43640), and a basic helix–loop–helix protein (bHLH092, At5g43650). TZP has no close homologs, but does share similarity with other Arabidopsis proteins that only have the zinc knuckle or PLUS3 domains. bHLH092 is part of the large bHLH family of transcription factors (11), and a close homolog is TRANSPARENT TESTA 8 (TT8; At4g09820). There are 4 closely related RPS15E genes (At1g04270, At5g09500, At5g09510, and At5g09490) (12). The transcript/expression of all 3 annotated genes in the Columbia background was confirmed by tiling array data and full length cDNAs. Quantitative RT-PCR (qPCR) revealed that all 3 genes are expressed in Bay-0 and Shahdara (data not shown).

T-DNA insertions in RPS15E or bHLH092 did not result in hypocotyl elongation phenotypes (data not shown). No insertion mutants were available for TZP (an insertion GABI-KAT line could not be recovered). To determine the association of the stop codon in TZP with the hypocotyl phenotype, we sequenced the PLUS3 domain in other A. thaliana accessions. Although we identified 10 synonymous changes and 13 nonsynonymous changes (including 2 changes affecting residues at least partially conserved in other species), we could not detect the Bay-0 premature stop codon polymorphism in a panel of ≈300 additional accessions (data not shown). We discovered that a Bay-0 single seed descent (SSD) line (Bay-0[41AV]), which was derived independently of the parent for the RIL set, did not have the 8-bp insertion in TZP. Sequencing the entire 7-kb candidate region revealed that this was the only sequence difference between Bay-0 and Bay-0[41AV] at LIGHT5. Genotyping additional markers in Bay-0[41AV] (as well as a careful phenotypic observation of the lines) showed that it is from the same genetic background as Bay-0 (data not shown). We crossed the 2 lines, with and without the TZP stop codon, to determine whether the mutation was responsible for the LIGHT5 phenotype. As shown in Fig. S2, phenotypes of F₂ plants homozygous for either allele were very similar to the phenotypes of the arHIF fixed for either the Bay (stop) or Sha allele, respectively. This demonstrates that the 8-bp insertion in TZP is sufficient to explain the LIGHT5 phenotype, and allows us to conclude that TZP controls hypocotyl growth. Bay-0[41AV] is the only Bay-0 stock we have which does not carry the 8-bp insertion. The question remains as to when this causative polymorphism appeared in the Bay-0 lineage (before or after collection) and whether it really exists in nature. Unfortunately, indications about the exact collection site of Bay-0 are very poor and make it nearly impossible to locate the original natural population and answer this question.

In addition, we used a transgenic approach to confirm the role of TZP in hypocotyl growth. The Sha alleles of all three genes were overexpressed in the rHIF containing the Bay allele (rHIF138–8). Overexpression of TZP (TZP-OX) caused plants to have very long hypocotyls (Fig. 2 B and C), with increased growth throughout development and extended duration of the reproductive phase (Fig. 2 D and E). All linearly elongating organs were more extended, including petioles, internodes, and pedoncules and the main floral stem of TZP-OX plants was usually twice as long as in its background. In contrast, RPS15E or bHLH092 overexpressing plants were indistinguishable from the parental line in terms of growth (data not shown). Overexpression of bHLH092 resulted in white seeds, similar to a phenotype found in mutants of its closest homolog TT8 (13). From these data we conclude that TZP is the QTG explaining the LIGHT5 QTL, and that the 8-bp indel leading to a premature stop codon in the Bay-0 allele is the causative polymorphism.

Fig. 2.

LIGHT5/TZP controls growth throughout development. (A) Increased TZP activity results in longer hypocotyls under blue light. Seedlings were grown at the fluence indicated and measured on day 6. TZP-OX(3) and TZP-OX(5) are two independent transgenic lines overexpressing TZP in rHIF138-8 background. rHIF138-8 contains the Bay allele with the premature STOP in TZP and rHIF138-13 contains the functional Sha allele. (B) Sha allele of TZP or overexpression result in longer hypocotyls. Plants were grown under light/dark cycles (12 hours/12 hours) at 22 °C and hypocotyls were measured 7 DAG. Representative seedlings were used to make the images in (C). Measurements represent three independent experiments of 20 seedlings each. (C) TZP-OX plants have long hypocotyls 7 DAG under light/dark cycles (12 hours/12 hours) compared with its background (rHIF138-8). (D) TZP-OX petioles are elongated 24 DAG compared with rHIF138-8. (E) Mature TZP-OX plants are almost twice as tall as background plants 50 DAG. Plants were grown under long days (light/dark: 16 hours/8 hours) for 50 days. All plants pictured are representative of at least 2 independent transgenic lines (lines 3 and 5) and 2 independent experiments. Vertical bar represents 20 cm. (F) TZP::YFP is localized to speckles in the nucleus. T2 plants carrying the 35S::TZP:YFP fusion were imaged to detect TZP localization. TZP:YFP localizes to the nucleus in guard cells of the stomata, in addition to all other tissues tested. Gray lines highlight the cell walls.

TZP Is a Large, Nuclear-Localized Protein Encoded by a Single Copy Gene.

TZP is a single copy gene in A. thaliana. TZP orthologs are also single copy in the fully sequenced plant genomes (Fig. S3). Interestingly, TZP is not found in the moss Physcomitrella patens, or in Chlamydomonas reinhardtii, or any earlier algal lineages queried, suggesting that it is specific to vascular plants.

The zinc knuckle (znkn; CX₂CX₄HX₄C) domain has been shown to be important for protein–protein interactions, as well as for binding single-stranded DNA (14). There are at least 24 znkn-containing proteins in Arabidopsis, five of which are closely related to each other but not to TZP (www.inra.fr/vast/Files/Loudet_PNAS_SITables.xls). Mutations in one of them, SWELLMAP 1 (SMP1), result in small plants with small cells because of a dysfunction in the commitment to the cell cycle (15). Another znkn protein, RSZ33, regulates interactions between splicing factors (16, 17). In yeast, the znkn of MPE1 has been found to control 3′ end processing of premRNA, and its human ortholog RBBP6 has been shown to interact with tumor suppressor pRB1 (14). Znkn domains are also found in retroviral gag proteins (nucleocapsid), including that of HIV (18).

The PLUS3 domain is thought to play a role in nucleic acid binding through 3 conserved positive amino acids (19). There are 5 proteins in Arabidopsis with the PLUS3 domain (Fig. S3). Three of these also have a SWIB domain, two have either a CCCH or a C3H3C4 zinc finger, and VERNALIZATION INDEPENDENT 5 (VIP5) contains only the PLUS3 domain (20). VIP5 is a homolog of the yeast RTF1 protein, which is part of the yeast Paf1 complex that regulates histone H2B ubiquitination, histone H3 methylation, RNA polymerase II carboxy-terminal Ser-2 phosphorylation, and RNA 3′ end processing (19). The SWIB domain-containing proteins are part of the SWI/SNF chromatin-remodeling complex (21). The combination of tandem zinc knuckles and a PLUS3 domain is unique to TZP-type genes in plants. Based on GFP fusions, TZP is localized to the nucleus in small punctate structures (Fig. 2F). Considering the domain structure and the localization, it seems most likely that TZP has a role in transcriptional control, perhaps at the level of chromatin remodeling.

TZP Regulates Light Quality-Dependent Growth.

Our initial QTL study suggested that LIGHT5 is involved in light-regulated hypocotyl growth. To determine whether the growth defects in LIGHT5 are specific to certain light environments, we measured hypocotyl lengths for both the rHIF and TZP-OX lines under different fluences of monochromatic red, blue, and far-red light, and in continuous dark. We found that both loss and gain of TZP activity had a significant effect on hypocotyl length under a range of blue or white fluences, but not in red or far-red light, or in the dark (Fig. 2A and 2B; data not shown). These results are consistent with TZP playing a role in light-dependent hypocotyl elongation, and suggest that TZP is involved in blue-light signaling.

We measured transcript abundance in Shahdara, Bay-0, rHIF138–8, rHIF138–13, arHIF47–2, arHIF47–5, and TZP-OX line #3 seedlings grown in constant blue light (15 μmol/m²s) for 5 days and then harvested at subjective dawn (relative to the time when they were moved from stratification to light). The extent of growth paralleled TZP expression levels (Fig. S4), consistent with TZP having a direct effect on growth.

To identify potential downstream transcriptional targets of TZP, we analyzed genome-wide expression patterns of 5 genotypes (rHIF138–8 and rHIF138-13; arHIF47-2 and arHIF47-5; TZP-OX line#3) by using Affymetrix Arabidopsis ATH1 GeneChip arrays. In pair-wise comparisons, we found 135, 12, and 769 genes to be differentially expressed between lines with contrasting alleles, rHIF138-8 vs. rHIF138-13, arHIF47-2 vs. arHIF47-5, and TZP-OX vs. rHIF138–8, respectively (P < 0.01; SI Materials and Methods) (www.inra.fr/vast/Files/Loudet_PNAS_SITables.xls). The top four significant gene ontology categories in the comparison of TZP-OX and rHIF138-8 (769 genes) are cytosol, ribosome, structural molecule activity, and cell wall (www.inra.fr/vast/Files/Loudet_PNAS_SITables.xls), consistent with TZP playing a specific role in modulating growth. We found similar results with the rHIF138-8 vs. rHIF138-13 comparison (data not shown).

TZP Controls Morning-Specific Growth Through an Auxin-Related Pathway.

It is well established that hypocotyl elongation is controlled by the circadian clock (1, 22). Therefore, we tested whether TZP is clock regulated and whether the circadian clock is altered in TZP-OX plants and the rHIF carrying the Bay-0 allele. We found that TZP cycles under both diurnal and circadian conditions with peak expression at dawn (transition from dark to light; Fig. 3A; Fig. S5). Hypocotyl growth is maximal at dawn and many genes that regulate growth, such as cell wall and phytohormone genes, have dawn-specific transcript abundance (1, 23). The dawn-specific TZP transcript peak suggests that it may be part of a circadian-controlled growth mechanism. To confirm that the clock controls TZP, we asked whether circadian mutants disrupt expression of TZP. Indeed, TZP expression was changed in both early flowering3 (elf3) and late elongated hypocotyl (lhy) mutants (Fig. S6). However, the circadian clock is not disrupted in either the rHIF or TZP-OX under light/dark cycles (Fig. S7), consistent with no feedback of TZP into the circadian clock. Based on these results we propose that TZP functions to control growth downstream of the circadian clock.

Fig. 3.

LIGHT5/TZP controls morning-specific growth pathways. (A) TZP displays dawn-specific transcript abundance under light/dark cycles (12 hours/12 hours) and constant 22 °C (six time points). The second day of data are copied from the first (double plotted) for visualization purposes. Expression was determined by qPCR with primers specific to TZP and SyberGreen. (B) TZP transcript abundance is overexpressed in TZP-OX. Five independent lines overexpressing TZP were characterized. Two lines are shown here. Data were collected and plotted as in (A). (C) The genes that are misexpressed (P < 0.01) in long hypocotyl genotypes (TZP-OX or rHIF138-13 vs. rHIF138-8) under blue light short day photocycles are expressed at dawn as determined with PHASER. (D) Long hypocotyls of the TZP-OX mutant are due in part to overexpression of cell wall genes. As an example of the expression pattern of the cell wall genes that are overexpressed in the TZP-OX mutant, IRX1 continues to cycle with peak expression at dawn, but its peak expression is 3-fold higher in TZP-OX (www.inra.fr/vast/Files/Loudet_PNAS_SITables.xls). Results are from array data.

Because TZP is regulated by light/dark cycles and the circadian clock, we asked whether the genes disrupted by TZP-OX were time-of-day specific. We used the PHASER time-of-day analysis tool (http://phaser.cgrb.oregonstate.edu), which determines whether there is a pattern of time-of-day co-expression in a given gene list compared with a background model. The peak expression of genes that were disrupted in TZP-OX, and differentially expressed in rHIF138–8 vs. rHIF138-13 is biased toward dawn (Fig. 3C). This is consistent with the dawn-specific expression of TZP. In addition, morning-specific response elements such as the morning element (CCACA), Gbox (CACGTG) and HUD (CACATG) (23, 24) (www.inra.fr/vast/Files/Loudet_PNAS_SITables.xls) were overrepresented in the promoters (500 bp) of these genes as determined by using the ELEMENT motif-searching tool (25). These results support a role of TZP in the transcriptional activity of dawn-specific genes.

In an effort to capture the entire effect of TZP-OX, we carried out a time course under light/dark cycles (12 hours white light/12 hours dark) in 7-day-old seedlings, sampling every 4 hours over 1 day in the same genotypes as above. We validated the overexpression of TZP by using qPCR, which revealed that TZP continued to cycle despite overexpression (Fig. 3B), suggesting that TZP is also controlled posttranscriptionally. Global expression changes were assessed by using Affymetrix ATH1 GeneChip arrays in TZP-OX, rHIF138-8, and rHIF138-13. The resulting time courses were analyzed for differentially expressed genes and for cycling genes [(24); SI Materials and Methods] (www.inra.fr/vast/Files/Loudet_PNAS_SITables.xls). Using the time points as replicates, we identified 117 up-regulated and 40 down-regulated genes in TZP-OX (P < 0.01).

Similar to the results in blue light, the genes that were up-regulated in TZP-OX were dawn specific (Fig. S8) and cell wall genes were overrepresented (www.inra.fr/vast/Files/Loudet_PNAS_SITables.xls). IRX1 is one example of the eight cell wall genes that were up-regulated in TZP-OX (Fig. 3D). Like IRX1, many of the up-regulated genes were specifically overexpressed at dawn, similar to the overexpression pattern of TZP itself (Fig. 3A; Fig. S9). Peroxidases, which can function to polymerize cell wall compounds, were also up-regulated in TZP-OX, consistent with their role in growth and cell wall expansion (26). Peroxidases PER27, PER30, and PER64 have been shown to be part of the cell wall proteome (27) (six PER genes total; Fig. S9A). However, CATALASE 3 (CAT3; SEN2) was one of the most down-regulated genes (Fig. S9D). CAT3 cycles under all diurnal and circadian conditions and may play a role in senescence and stress responses (28, 29). Two additional genes of note were down-regulated in TZP-OX: PW9 (encoding a MATH/TRAF domain protein) and one of the PLUS3 homologs that encodes also a SWI/SNF domain (Fig. S9 E and F). In general, the genes that were misregulated in TZP-OX are involved in cell expansion, consistent with TZP being intimately related to regulation of hypocotyl elongation and downstream of the circadian clock.

Auxin-response genes, including WES1 [GH3.5; (30)], DFL1 [GH3.6; (31)] (Fig. S9B), and AXR5 [IAA1; (32)], all of which have been shown to control hypocotyl growth, were also up-regulated upon TZP overexpression. It has been shown recently that auxin controls growth in a time-of-day fashion (33), which fits with TZP controlling morning-specific growth through these genes. In addition, 2 homeobox-leucine zipper genes induced in TZP-OX (HAT4 and HAT52) are up-regulated under low light conditions and in response to auxin (Fig. S9C). Mutations in these genes lead to growth defects (34, 35). Together, these results support the notion that TZP plays a broad role in the regulation of phytohormone-dependent gene transcription (23).

A similar number of genes (≈7,700 genes) were found to cycle across all 3 genotypes as reported for this condition previously (24), and there was no significant difference between the number of genes cycling in rHIF lines and TZP-OX, although there were genes specific to each genotype. However, the peak transcript abundance of 362 genes was shifted by 6 hours or more in TZP-OX compared with rHIF138-8 (only genes that cycled in both genotypes were considered). Phytohormone-related (SIR1, ETO1, GH3.3, RGA1, ABF4), homeobox-leucine zipper (HAT2, HAT3), chromatin remodeling (HD2B, HDT4, SUVH9, CHR4, TAF1), leaf polarity (KAN3, AS1), and ribosomal genes were misphased in TZP-OX. PHYTOCHROME D (PHYD) was also phased 8 hours earlier in TZP-OX (Fig. S9L). The first phyD mutant was originally identified as a natural allele, and subsequently was shown to affect shade avoidance associated growth and flowering (3, 36).

Consistent with TZP controlling blue-light dependent growth, LONG HYPOCOTYL IN FARRED 1 (HFR1) is overexpressed at dawn more than 100-fold with the same pattern as IRX1 (Fig. S9I). HFR1 encodes a bHLH transcription factor that is required for both phytochrome A-mediated far-red and cryptochrome 1-mediated blue light signaling (37). HFR1 expression is high under low light conditions such as shade and continuous dark conditions (35), and is elevated in circadian and light signaling mutants, much like in TZP-OX [Fig. S9I; (23)]. Recently it has been shown that two other bHLH transcription factor genes related to HFR1—PIF4 and PIF5—control morning-specific hypocotyl growth. Disruption of the circadian clock gene CCA1 results in the overexpression of PIF4 and PIF5 leading to uncontrolled elongation (1). However, PIF4 and PIF5 are expressed at control or slightly lower levels in TZP-OX (Fig. S9 G and H). These results support TZP acting in parallel with (or downstream of) PIF4 and PIF5 in growth control.

Conclusions

Despite extensive forward genetics screens in A. thaliana, natural variation has recently made important contributions to the identification of genes not previously known to impact several different traits (38–41). Apart from being able to exploit allelic variation (in multiple genetic backgrounds) that cannot be generated by conventional mutagenesis, the success of these studies has often been because of the use of quantitative phenotyping, as opposed to the qualitative gauges used in typical mutant screens. We have demonstrated here the power of QTL analysis to reveal a new component of the hypocotyl growth pathway in A. thaliana, TZP, a unique, tandem zinc knuckle/PLUS3 domain protein encoded by a single copy gene in the vascular plant lineage. TZP provides a direct link between light signaling and the pathways that control growth in an environmentally independent fashion.

Materials and Methods

A detailed and referenced version of this section is available online (SI Materials and Methods).

Plant Material and Phenotyping.

The core population of 164 RILs from the Bay-0 × Shahdara set was phenotyped in four different light and temperature environments to map QTL affecting hypocotyl elongation (www.inra.fr/vast/Files/Loudet_PNAS_SITables.xls). HIF350 was developed from an F7 line (RIL350) that still segregated for a single and limited genomic region around LIGHT5 locus. Plants still heterozygous for the QTL region were screened with adequate markers to isolate recombinants (rHIF) used in the fine-mapping process. Advanced rHIF crosses were generated from two different rHIFs recombined immediately to the north or immediately to the south of the LIGHT5 interval giving rise to lines arHIF47. Distinct Bay-0 lines from the stock center were used to find variants at LIGHT5. rHIF138-8[Bay] was complemented by over-expressing each of the three positional candidate genes cloned from rHIF138-13[Sha].

QTL Mapping.

Analyses used hypocotyl length mean values of an average of 16 seedlings (from 2 distinct experiments) per genotype per environment. QTL analyses were performed by using QTL Cartographer, with classical parameters for interval mapping and composite interval mapping.

Microarray Analysis.

Microarray experiments were carried out per Affymetrix protocols (ATH1 GeneChip), on 7-day-old tissue harvested under either continuous blue at subjective dawn or every 4 hours (starting at dawn) under 12 hours white light/12 hours dark cycles over 1 day (six time points). Hybridization intensities from all microarrays were normalized together by using gcRMA implemented in the R statistical package. The blue dataset was then separated and differentially expressed genes were identified by using linear modeling with the limma bioconductor package in R.