Identification of the novel protein QQS as a component of the starch metabolic network in Arabidopsis leaves


  • Ling Li,

    1. Department of Genetics, Development and Cell Biology, Iowa State University, Ames, IA 50011, USA
    Search for more papers by this author
  • Carol M. Foster,

    1. Department of Genetics, Development and Cell Biology, Iowa State University, Ames, IA 50011, USA
    Search for more papers by this author
    • Present address: Department of Applied Biological Sciences, Arizona State University at the Polytechnic Campus, Mesa, AZ 85212, USA.

  • Qinglei Gan,

    1. Department of Biochemistry, Biophysics, and Molecular Biology, Iowa State University, Ames, IA 50011, USA
    Search for more papers by this author
    • Present address: Department of Agronomy, Iowa State University, Ames, IA 50011, USA.

  • Dan Nettleton,

    1. Department of Statistics, Iowa State University, Ames, IA 50011, USA
    Search for more papers by this author
  • Martha G. James,

    1. Department of Biochemistry, Biophysics, and Molecular Biology, Iowa State University, Ames, IA 50011, USA
    Search for more papers by this author
  • Alan M. Myers,

    1. Department of Biochemistry, Biophysics, and Molecular Biology, Iowa State University, Ames, IA 50011, USA
    Search for more papers by this author
  • Eve Syrkin Wurtele

    Corresponding author
    1. Department of Genetics, Development and Cell Biology, Iowa State University, Ames, IA 50011, USA
      *For correspondence (fax +1 515 294 1337; e-mail
    Search for more papers by this author

*For correspondence (fax +1 515 294 1337; e-mail


Little is known about the role of proteins that lack primary sequence homology with any known motifs (proteins with unknown functions, PUFs); these comprise more than 10% of all proteins. This paper offers a generalized experimental strategy for identifying the functions of such proteins, particularly in relation to metabolism. Using this strategy, we have identified a novel regulatory function for Arabidopsis locus At3g30720 (which we term QQS for qua-quine starch). QQS expression, revealed through global mRNA profiling, is up-regulated in an Arabidopsis Atss3 mutant that lacks starch synthase III and has increased leaf starch content. Analysis of public microarray data using MetaOmGraph (, in combination with transgenic Arabidopsis lines containing QQS promoter–GUS transgenes, indicated that QQS expression responds to a variety of developmental/genetic/environmental perturbations. In addition to the increase in the Atss3 mutant, QQS is up-regulated in the carbohydrate mutants mex1 and sis8. A 586 nt sequence for the QQS mRNA was identified by 5′ and 3′ RACE experiments. The QQS transcript is predicted to encode a protein of 59 amino acids, whose expression was confirmed by immunological Western blot analysis. The QQS gene is recognizable in sequenced Arabidopsis ecotypes, but is not identifiable in any other sequenced species, including the closely related Brassica napus. Transgenic RNA interference lines in which QQS expression is reduced show excess leaf starch content at the end of the illumination phase of a diurnal cycle. Taken together, the data identify QQS as a potential novel regulator of starch biosynthesis.


Starch biosynthesis and degradation play a central role in plant metabolism, with starch acting as a repository for reduced carbon produced in leaves during the day, and as a supply of chemical energy and anabolic source molecules originating from sucrose during the night. In view of this central function, starch production and degradation are highly likely to respond to environmental, metabolic, circadian rhythm and/or hormonal signals (Usadel et al., 2008). Such regulation is evident from the observation that changing the day length affects the rate of starch degradation in Arabidopsis leaves (Lu et al., 2005). When plants are grown under long days (LD), the level of starch accumulation is increased relative to that under short-day (SD) conditions; however, the rate of starch degradation also increases under LD conditions such that, at the end of the night, the basal level of leaf starch is approximately the same regardless of day length. The mechanism of such regulation is not well understood, and is likely to involve complex interactions among transcriptional, translational and metabolic controls (Lu et al., 2005).

The starch biosynthetic enzyme system produces the two glucan polymers that make up starch granules, amylose and amylopectin. Amylose is a primarily linear polymer made up of α-(1→4)-linked glucan chains, whereas amylopectin is composed of relatively highly branched chains of glucose joined via an ordered arrangement of α-(1→4)- and α-(1→6) glycosidic bonds. Starch synthesis involves production of the glucosyl unit donor ADP-glucose (ADPGlc), which is catalyzed by ADPGlc pyrophosphorylase (ADPGPP, encoded by at least six genes in Arabidopsis) (Crevillen et al., 2003; Gibon et al., 2004), elongation of α-(1→4)-linked chains, which is catalyzed by starch synthases (SS, encoded by at least five genes in Arabidopsis) (Ball et al., 1998; Tenorio et al., 2003; Delvalle et al., 2005; Zhang et al., 2005, 2008), and introduction of branch linkages by chain cleavage/transfer reactions that are catalyzed by starch branching enzymes (encoded by three genes in Arabidopsis) (Seo et al., 2002; Tetlow et al., 2004; Dumez et al., 2006). In addition, an editing function involving limited hydrolysis of some α-(1→6) branch linkages catalyzed by starch debranching enzymes (encoded by four genes in Arabidopsis) has been proposed as a component of the starch biosynthetic pathway (Ball et al., 1998; Myers et al., 2000; Delatte et al., 2005, 2006; Wattebled et al., 2005). Starch degradation also involves a large number of enzymes, including starch debranching enzymes, multiple glucan water dikinases, α-amylases, β-amylases, disproportionating enzymes, phosphorylase, and glucose and maltose transporters that export carbon from chloroplasts at night (Smith et al., 2003; Zeeman et al., 2004; Lloyd et al., 2005). It is likely that carbon flux throughout the diurnal cycle is impacted by regulatory networks that precisely coordinate this multitude of enzyme activities. Potential agents affecting starch level regulation, either as targets, signal generators or both, include the biosynthetic and catabolic enzymes themselves.

As a tool to investigate potential regulatory networks, we have made use of a mutation in the Arabidopsis gene SS3 (Zhang et al., 2005). In contrast to the expectation that loss of a starch synthase would cause decreased levels of starch, Atss3 null mutations cause a starch excess in leaves, indicating regulation of the metabolic pathway. This effect occurs through an increased rate of synthesis, as opposed to a decreased rate of degradation, as evidenced by a starch-excess phenotype within a single light phase starting from a basal level of essentially no starch. Although the Atss3 mutant has an increased rate of starch synthesis and altered starch structure, the visual phenotype of this plant is indistinguishable from that of the wild-type (WT) (Zhang et al., 2005). This interaction of SSIII with starch accumulation is complex; the Atss2/Atss3 double mutant (which is deficient in both SSIII and SSII) has decreased starch content (Zhang et al., 2005, 2008).

Regulation of starch biosynthesis by the SSIII protein has been indicated in several other instances. For example, a 14–3–3 protein is thought to bind to SSIII (Sehnke et al., 2000, 2001), and, in maize endosperm, SSIII exists in a multi-subunit complex that is likely to include other starch synthases as well as multiple starch branching enzymes (Hennen-Bierwagen et al., 2008).

Here, global mRNA level profiling of the Atss3 knockout mutant, in which metabolic flux into starch is increased (Zhang et al., 2005), is compared to that of WT plants. The Atss3 mutant is perturbed not only in levels of AtSS3 transcript, but also expression of a subset of other genes. Among the genes with the most significantly altered transcript levels is At3g30720, designated here as QQS (qua-quine starch). QQS may be considered a PUF (protein with unknown functions). Approximately 15–40% of all eukaryotic genes are considered to encode PUFs; little is known about the functions of such genes (Gollery et al., 2006, 2007; Luhua et al., 2008). This RNA profiling suggested that altered starch metabolism might in some way generate a signal that affects expression of the QQS gene, and prompted further investigation of whether the product of QQS is itself involved in regulation of starch metabolism. The experiments presented here indicate that this is in fact the case, as down-regulation of QQS results in increased starch content. Thus, QQS, a protein that consists of only 59 amino acids with no known sequence homologs or predicted structural motifs, appears to provide a previously unidentified function in Arabidopsis starch metabolism.


Global gene expression differences in an Atss3 knockout mutant compared to WT

In order to gain insight into the global changes in plant metabolism associated with increased metabolic flux into starch, an Arabidopsis line homozygous for the mutation Atss3–1 was compared to congenic WT plants. Atss3 mutants accumulate approximately 20% more starch than WT (Zhang et al., 2005). Significantly, the growth rate and appearance of Atss3 mutant plants is normal, so observed transcriptional changes probably result from the alteration in starch metabolism rather than secondary effects of the mutation. The relative steady-state mRNA levels in Atss3 and WT plants were determined during the diurnal cycle using Affymetrix Arabidopsis ATH1 arrays (Affymetrix, A randomized complete block design was used for distribution of the two genotypes in order to minimize the effects of environmental variables. Two independent biological replicate experiments were conducted. At 42 days after imbibition, the 5th–8th rosette leaves were harvested at five time points across the diurnal cycle, specifically 1 and 4 h in the light, and 0.5, 4 and 8 h in the dark.

As expected from the location of the Atss3–1 mutation (Zhang et al., 2005), the AtSS3 transcript level measured by microarray analysis was reduced to an essentially undetectable level (Figure 1). Interestingly, in addition to the expected down-regulation of AtSS3, a number of other genes in the starch metabolic network have altered transcript levels in the Atss3 mutant (Figure 2 and Table S1).

Figure 1.

 Diurnal patterns of accumulation of AtSS3 and QQS transcripts in the Atss3 mutant.
Leaves of plants grown under SD conditions (8 h light/16 h dark) were harvested at intervals during the 24 h cycle, and RNA was analyzed using the Affymetrix Arabidopsis ATH1 chip. The median transformed value of the relative level of the 22 000 transcripts on each chip is set at zero. Standard error bars are shown. QQS shows more than a sevenfold increase in RNA accumulation for all five time points over the diurnal cycle in the Atss3 mutant compared to WT (P value = 0.00003, q value = 0.065).

Figure 2.

 Altered transcript levels in the Atss3 mutant in the starch metabolic network.
The results are based on a genotype comparison, P value <0.05 with corresponding q value <0.2. Genes that are up-regulated in the Atss3 mutant are shown using thick green solid lines; genes that are down-regulated in the mutant are shown using dashed thin orange lines; expression of enzymes with no change in encoding genes in the mutant is shown using thin grey lines. Membrane transporters are indicated using hexagons; metabolites are indicated using ovals. Gene symbols according to the Arabidopsis Genome Information Resource are shown in parentheses next to enzyme names. 14–3–3 ε is thought to repress starch accumulation (Sehnke et al., 2001). Trehalose/trehalose-6-P are thought to modulate starch content (Schluepmann et al., 2003; Kolbe et al., 2005; Lunn et al., 2006).

Over the diurnal cycle, the steady-state levels of mRNA from seven genes out of approximately 22 000 analyzed were significantly different in the Atss3 mutant compared to WT plants (Table S2) when significance was defined as a P value of 0.00005 for a test contrasting the expression profiles of WT and mutant plants; this threshold for significance yielded a false discovery rate of approximately 0.065, calculated as described by Storey and Tibshirani (2003).

Identification of QQS

QQS RNA was initially identified as a potential component of starch metabolism based on the observation that it shows a notably large increase in the Atss3 mutant compared to WT (P value = 0.00003) (Figure 1). The abundance of QQS mRNA was high even in WT leaves relative to the median accumulation of transcripts of all genes on the chip. The increase in QQS transcript in the Atss3 mutant relative to the WT at each time point sampled across the diurnal period was at least sevenfold.

In both WT and Atss3 mutant leaves, the abundance of QQS mRNA increased to a slight but significant degree during the dark phase of the diurnal cycle (Figure 1). In WT, accumulation of QQS mRNA was lowest at 4 h in the light. Accumulation increased slowly thereafter in the light, and more rapidly after onset of dark. QQS mRNA accumulation in WT peaked at 4 h in the dark at a level approximately three times higher than its lowest value, and then decreases. In the Atss3 mutant, the diurnal accumulation pattern is somewhat different. In this instance, QQS mRNA accumulation is lowest at 0.5 h in the dark, increases throughout the dark period, and reaches a peak at 8 h in the dark at a level approximately two times higher than its lowest value.

Given that QQS is up-regulated in the starch-excess Atss3 mutant, one possibility (and indeed our working hypothesis) is that this gene plays a role in starch metabolism. At3g30720 (QQS) is annotated in the Arabidopsis Genome Information Resource as an ‘expressed protein’, with cellular component ‘unknown’, molecular function ‘unknown’, and biological process ‘unknown’. Based on its gene model, the transcript had been annotated as beginning approximately 314 bp upstream of the initiation codon, and no 3′ untranslated region (UTR) was identified.

RACE experiments to define the 5′ and 3′ of the QQS mRNA indicated that the 5′ UTR of QQS is 414 nucleotides upstream of the reported translational start site, and that the 3′ UTR is 106 nucleotides downstream of the stop codon (Figure 3a). Five putative transcription factor binding motif sequences are located within the 754 bp intergenic region upstream of the QQS 5′ UTR (Table S3). The QQS transcript encodes a protein comprising 59 amino acids, of molecular weight 7 kDa, which shows no nucleotide or amino acid sequence similarity to any other gene of Arabidopsis, or indeed any genes or sequences in the National Center for Biotechnology Information database. In addition, the QQS protein does not appear to contain any previously described catalytic domains or any structural motifs. Thus, QQS can be characterized as a PUF (Gollery et al., 2006); about 40% of PUFs are species-specific based on their primary sequence (Gollery et al., 2007). QQS is not present in the structural database Immunoblot analysis indicated that the QQS open reading frame is translated into an approximately 7 kDa protein product (data not shown).

Figure 3.

 Structure of QQS and QQS recombinant DNA constructs.
(a) QQS gene locus, as determined by 5′ and 3′ RACE. Black boxes, location of 5′ and 3′ UTRs; white boxes, coding regions; grey boxes, exons of upstream or downstream genes; solid line, intron; dotted line, non-transcribed region.
(b) QQS reporter gene constructs. Diagonally hatched boxes, region of QQS included in each construct. The GUS and GFP reporter sequences are not to scale.
(c) QQS RNAi construct. Diagonally hatched box, region of QQS included; vertically striped grey box, intermittent vector fragment. The 35S promoter and vector are not to scale (see Appendix S1 for details).
Nucleotide positions are in relation to the ATG start codon of QQS.

A notable feature of the QQS chromosomal neighborhood is its multiple pseudogenes/transposon-like genes with homology to the CACTA-like transposase family, the gypsy-like transposon family, and non-LTR retro-element sequences, and its small open reading frames (Figure 4 and Table S4). Apart from QQS, only At3g30725 and At3g30730 have predicted proteins. At3g30730, At3g30740 and QQS are all represented on the Affymetrix ATH1 chip; however, only QQS is expressed over a background level (Figure 5). During evolution, the chromosomal region surrounding QQS may have been a site of high transposon activity, based on the presence of the many transposon-like sequences. Methylation in this region is dense, as might be expected because of the multiple methylation sites in the region of the QQS gene (AnnoJ, Lister et al., 2008).

Figure 4.

 The QQS neighborhood contains small ORFs and putative transposon-like genes/pseudogenes.
Of the 55 kb surrounding the QQS gene, only At3g30720, At3g30725 and At3g30730 are represented by ESTs (details in Table S4). The smallest genes (blue bars) are not drawn in proportion to their sizes.

Figure 5.

 Accumulation of QQS RNA.
Each point on the x axis represents a publicly available mRNA transcriptomics profiling data sample (from 956 Affymetrix ATH1 chips) for a given experimental condition (Li et al., 2007; Mentzen et al., 2008). The y axis represents the normalized expression level for selected genes. The mean transcript accumulation level for each chip is normalized to a value of 100 (marked with an arrow). The data were visualized using MetaOmGraph software (
(a) QQS (At3g30720, red line) has a complex pattern of RNA accumulation across a wide variety of experimental conditions. Only two other genes in the QQS neighborhood (Figure 4) are represented in the ATH1 chip. Expression of these two small putative ORFs (At3g30730 and At3g30740, blue and green lines respectively) located near QQS on chromosome 3 is low or non-existent.
(b) QQS RNA accumulation is increased during pollen maturation (microarray data from Honys and Twell, 2004 and Schmid et al., 2005). UM, uni-nucleate microspores; BP, bicellular pollen; TP, tricellular pollen; MP, mature pollen.

QQS transcript accumulation is responsive to developmental, environmental and genetic signals

QQS mRNA accumulates in many plant organs under a wide variety of conditions in an environmentally responsive manner as visualized using MetaOmGraph (Li et al., 2007; Wurtele et al., 2007; Mentzen et al., 2008) (Figure 5). For example, QQS mRNA accumulation is high in cell cultures (Figure 5a), and doubles in cell suspensions in which there is no sucrose in the medium compared to cultures with added sucrose (microarray data from Contento et al., 2004). In addition, QQS mRNA accumulation increases as pollen develops, and reaches a peak when pollen matures (Figure 5b, microarray data from Honys and Twell, 2004 and Schmid et al., 2005). The QQS transcript level is also increased in response to numerous genetic alterations. Among these are loss-of-function mutations in the gene ETR1, which codes for an ethylene receptor (Gamble et al., 2002), and over-expression of nahG, which codes for salicylate hydroxylase (Heck et al., 2003). QQS is up-regulated in a mutant with knockout mutations of PEN3, which codes for an ATP binding cassette transporter that is implicated in the salicylic acid response (Stein et al., 2006), in the hog1 mutant, which has a defect in DNA methylation (microarray data from Jordan et al., 2007), in AtrbohB mutant seedlings, which have reduced tolerance to heat (Larkindale et al., 2005), and in a mutant over-expressing CKX1, which codes for a cytokinin oxidase/dehydrogenase (Bilyeu et al., 2001). QQS is down-regulated under cold growth conditions.

QQS transcript levels are increased in several mutants that over-express microRNA, as visualized in Genevestigator (Zimmermann et al., 2005). In particular, QQS is up-regulated in the miR159a over-expression mutant (Achard et al., 2004), indicating a possible role of these genes in QQS processing.

This complex expression pattern of QQS is fairly distinct relative to the patterns of expression of other genes in the Arabidopsis genome. Indeed, no genes are correlated with QQS at a Pearson correlation above 0.6, and only 14 other loci are somewhat correlated with QQS at Pearson correlations of 0.5–0.6 (MetaOmGraph; Table S5). QQS is among the 25% of Arabidopsis genes that are not included in any of the regulons (co-expressed gene clusters) of Arabidopsis (Mentzen and Wurtele, 2008).

Spatial and temporal QQS expression at the level of translation initiation

The spatial and temporal localization of QQS transcription in plant tissues was detected using a reporter gene fusion, QQSpro, in which GFP/GUS coding sequences were expressed under the control of the QQS promoter (Figure 3). Seven independent transformant lines were stained for GUS activity throughout the plant life cycle.

Activity of the QQS promoter is evident at 2 days after imbibition in hypocotyls and root tips (Figure 6d). As seedlings grow, QQS expression expands to the vasculature, mesophyll cells, hydathodes, and trichomes of leaf blades and petioles. Microscopic dissection indicated no expression in the shoot meristem; the dark GUS staining in the shoot tip is associated with the adjacent vasculature. GUS activity is higher in mature leaves compared to young emerging leaves; it consistently appears somewhat unevenly distributed, and is predominantly located in the vasculature (Figure 6a,e–g, Table S6 and Figure S1a–h); this pattern is maintained throughout development.

Figure 6.

 Spatial patterns of expression of QQS in WT, Atss3 and Atss2/Atss3 genetic backgrounds.
Transgenic lines containing a QQS promoter–GFP/GUS fusion construct (QQSpro, detailed in Figure 3) were evaluated for GUS activity. Eight independent lines were analyzed for each genetic background. For each independent line, five plants or more were evaluated at each developmental stage.
(a–c) QQS expression in WT lines during development. Expression of QQS is highest in leaves (a), root vasculature and tips (b), and filaments and sepals (c).
(d–i) Comparison of QQS expression in WT, Atss3 and Atss2/Atss3 backgrounds in rosettes (d–g) and flowers and buds (h,i). QQS expression in the Atss3 mutant has a generally similar pattern to that of the WT. However, accumulation of QQS in the Atss3 background is greater for many developmental conditions. The Atss2/Atss3 mutant does not maintain this increased QQS expression. In petals, QQS is expressed only in the Atss3 background. an, anther; ct, cotyledon; elr, emerging lateral root; fi, filament; hc, hypocotyl; hy, hydathode; pd, pedicel; pe, petiole; po, pollen; pt, petal; rt, root tip; rv, root vasculature; sm, shoot meristem; sp, sepal; sti, stigma. Red scale bars = 1 mm; blue scale bars = 0.1 mm.

QQS expression is low in flower buds; however, by the time of flower opening, QQS expression is evident in pedicels, sepals, filaments, mature pollen, stigma papillae and styles, but not petals (Figure 6c,h,i and Figure S1a–h). During silique development, QQS expression increases in the stigma papillae and style, and becomes apparent throughout the maternal tissues of the silique wall and receptacle.

QQS is expressed in roots throughout development (Figure 6b,d. and Figure S1a–h). Expression is highest in the root tip, specifically the root cap, columella cells and peripheral cap, lower in the root meristem region, and absent in the epidermis. QQS is expressed at the site of lateral root initiation, and in the root tip and vasculature during its emergence; as the lateral root matures, expression remains detectable throughout the root cortex vasculature.

QQS expression under conditions of altered starch biosynthesis

To expand on the microarray data, which show that the QQS transcript level is increased by the absence of AtSS3 gene expression, the QQS promoter–GFP/GUS fusion gene was introduced into the Atss3 background. GUS activity driven by the QQS promoter was higher in the Atss3 single mutant than in WT under virtually all conditions (Figure 6d–i). Expression was detectable throughout the entire seedling at 2 days after imbibition (Figure 6d), as well as later in development, in particular in leaves, flowers and roots (Figure 6c–i and Figure S1i–n). Although the general pattern of expression is similar in the Atss3 mutant and WT, QQS is expressed ectopically in petals in the Atss3 mutant (Figure 6h,i). These results are consistent with, and an expansion of, the transcriptomics data. Furthermore, this analysis indicates that induction of QQS in the Atss3 mutant involves regulation of transcription initiation frequency and/or translation initiation frequency.

In addition, to evaluate whether it is the absence of SSIII that affects the QQS expression pattern, the QQS promoter–GFP/GUS fusion gene was introduced into an Atss2/Atss3 double mutant background. In this background, although SSIII is eliminated, starch accumulation is decreased (Zhang et al., 2008). Expression of QQS promoter–GUS in the Atss2/Atss3 double mutant background was more nuanced, but was in general similar to or somewhat lower than that in WT throughout leaf development (Figure 6).

Subcellular localization of QQS protein

If there is an interaction between QQS and SSIII, one possibility is that QQS might interact with starch metabolic enzymes in the plastid. To examine the subcellular localization of QQS protein, a second QQS–GFP/GUS gene fusion, termed QQSprotar, was generated, in this instance expressing a translational fusion in which the QQS promoter and the nucleotides encoding the full QQS amino acid sequence were fused to the GFP/GUS reporter genes (Figure 3). Thus, any protein target signal present in QQS would be expected to direct the fusion protein to the same location as the native protein.

Leaves were observed by confocal fluorescence microscopy in six independent QQSprotar lines. The QQS promoter-coding sequence–GFP fusion protein does not appear to be targeted to the plastids, nor is it concentrated in the nucleus, or vacuole, nor does it appear to be located in any other compartment (Figure 7). This evidence that QQS is cytosolic is consistent with the fact that QQS has not been reported in any protein localization database characterizing plastids or other organelles, including PPDB (Friso et al., 2004), Plprot (Kleffmann et al., 2006), AMPDB (Heazlewood and Millar, 2005), AraPerox (Reumann et al., 2004) and AtNoPDB (Brown et al., 2005). It is also consistent with the absence of any obvious trafficking signals in QQS as evaluated by PSORT (, ChloroP and TargetP ( This result does not support a physical interaction occurring within the plastid between QQS protein and SSIII or any other starch metabolic enzymes.

Figure 7.

 QQS–GFP subcellular localization.
Mesophyll cells of fully expanded leaves from seedlings at 15 days after imbibition were analyzed by confocal microscopy. Thirty leaves from six independently transformed lines (five leaves from each) were evaluated. Scale bars = 20 μm. R, red signal from the autofluorescence; G, green signal from the GFP; M, merged autofluorescence and GFP. Signal was not detected in the plastids or nuclei, and appears to be predominantly cytosolic.

Effects of QQS deficiency

In the Atss3 line, QQS expression could either be responding to changes in the levels of starch, or a related metabolite, or it could be causing the increase in starch biosynthesis. Alternatively, QQS might not have any functional relationship to starch metabolism. To examine whether QQS participates in the regulation of starch metabolism, Arabidopsis lines were developed in which the QQS protein concentration was reduced relative to WT (Figure 8a). RNA interference (RNAi) repression of QQS expression was used, because T-DNA insertion mutations have not been identified in this small genetic element in any public strain collection. QQS RNAi mutant lines were constructed by transformation with a gene construct expressing a 150 bp sense fragment of QQS coding sequence (nucleotides 24–173 from the native translational initiation codon) under the control of the CaMV 35S promoter (Figure 3c). Four independent T2-stage QQS RNAi lines were selected for phenotypic evaluation. Similar to the Atss3 mutants, QQS RNAi mutant plants are indistinguishable from WT with regard to gross plant morphology, at all stages from seedling to senescence during growth under each of three light regimes (SD, LD and continuous light) (Figure 8b and data not shown).

Figure 8.

 Seedlings of the QQS RNAi T2-2 line show decreased QQS protein levels and a starch-excess phenotype at the end of the light period.
QQS RNAi (four independent lines) and WT plants were grown in a random block design under an LD regime. At 21 days after imbibition, five seedlings were harvested at the end of the light period and stained for starch, or pooled and extracted for Western blotting.
(a) Western blot analysis of total leaf protein extracts. Fifty micrograms of protein were loaded per lane; blots were probed with anti-QQS serum; the arrow indicates the location of QQS protein.
(b) WT, QQS, RNAi, T2-2 and Atss3 plants before harvest.
(c) Iodine staining of leaf starch in WT, QQS RNAi T2-2 and Atss3 seedlings. Green scale bar = 1 cm; red scale bar = 1 mm.

Iodine staining of leaf starch was used as an initial test of whether starch metabolism was altered in the QQS RNAi lines. All four lines have a starch excess phenotype at the end of the light period (Figure 8c and data not shown), but not at the end of dark phase of the diurnal cycle (data not shown). By this iodine staining assay, the starch-excess phenotype of the QQS RNAi lines appeared similar in magnitude to that of the Atss3 mutants (Figure 8c). Starch levels were then quantified by enzymatic assay at the end of the light and dark phases of the LD cycle. Starch content was increased by 20–30% in leaves of QQS RNAi mutants compared to WT, and this starch-excess phenotype was observed in each of the independent transformant lines (Table 1). The degree of starch increase in the QQS RNAi lines was similar to that previously observed for the Atss3 mutant lines (Zhang et al., 2005). Neither the QQS RNAi lines nor the Atss3 mutant displayed any significant difference from WT with regard to starch content at the end of the dark phase (Table 1).

Table 1.   Leaf starch content in WT and QQS RNAi plants
GenotypeStarch content (mg g−1 fresh weight)a
  1. aStarch content measurements are the mean ± SE of three independent biological replicates, with three plants per replicate. Four QQS RNAi T2 lines were analyzed.

  2. bStarch contents of QQS RNAi mutants that are significantly different from that of WT are marked with an asterisk (Student’s test, < 0.01).

  3. cEOL indicates the end of the light period in a LD cycle, and EOD indicates the end of the dark period in a LD cycle.

WT4.85 ± 0.150.50 ± 0.009
QQS RNAi T2-1b5.89 ± 0.15*0.56 ± 0.011
QQS RNAi T2-2b6.38 ± 0.21*0.48 ± 0.010
QQS RNAi T2-3b6.32 ± 0.33*0.50 ± 0.008
QQS RNAi T2-4b5.83 ± 0.14*0.52 ± 0.009

Starch content was quantified at 4 h intervals over the course of the diurnal cycle in a homozygous QQS RNAi line (T2–2). In this line, QQS protein accumulation is reduced (Figure 8a). The starch level was notably higher in the QQS RNAi line than in WT at the end of the light phase (Figure 9a). By 4 h after the onset of darkness, the amount of starch in the QQS RNAi mutant leaves had decreased to that of the WT level. These data suggest that the increased accumulation of starch in QQS RNAi mutants is due to increased starch synthesis, rather than decreased starch hydrolysis. To test this point further, plants were depleted of starch by incubation in the dark for an extended 18 h period (Zhang et al., 2005). The plants were then exposed to one complete LD cycle, and starch content was analyzed at the end of the light and the end of the dark phase (Figure 9b). No difference in starch accumulation between the QQS RNAi line and WT was observed at the end of either the extended dark period (18 h) or the subsequent normal dark period (8 h). In contrast, the starch level in the RNAi line was increased by about 30% at the end of the first light period. These data demonstrate definitively that reduction of QQS gene expression results in an increased rate of starch synthesis.

Figure 9.

 Leaf starch levels.
(a) Starch levels during an LD cycle. WT and QQS RNAi plants were grown under a 16 h light/8 h dark cycle, and starch content was determined in leaves harvested at 4 h intervals. Data points are the mean ± SE of three biological replicates, with three plants per replicate. Where SE is not shown, the value is either less than the width of the symbol, or less than the SE value for the overlapping time point.
(b) Starch accumulation during a single light period. Starch levels were measured as in (a) after an extended dark period of 18 h (D18), after the subsequent light period of 16 h (L16), and again after the next 8 h dark period (D8).


This study sought to discover mechanisms by which starch accumulation is regulated in leaves. We postulated that, under conditions of an increase in starch accumulation, some of the molecules associated with that shift might be evident. To enrich for molecular-genetic alterations specifically associated with changes in starch metabolism, we used the Atss3 knockout line, which is morphologically and developmentally indistinguishable from WT but has a greater starch content (Zhang et al., 2005). In these mutants, the transcriptome of the starch (and trehalose) metabolic network is selectively perturbed. These shifts are consistent with an increased capacity for transport of substrate into the plastid and an increase in the starch biosynthetic machinery as a whole. QQS, a gene of previously of unknown function, was identified based on its increased mRNA levels in the Atss3 mutant, and genetic and biochemical evidence is presented demonstrating that this novel gene functions in the control of starch content.

The spatial expression and temporal patterns of QQS expression, as determined by the QQS promoter–GUS expression analysis combined with global analysis of public microarray data using MetaOmGraph, reveal a complex expression pattern that is sensitive to developmental stage, organ, cell and tissue type, genetic background, and environmental changes. The QQS promoter has few identifiable regulatory motifs; however, there are indications of other possible factors that might control QQS transcript levels. Several lines of evidence indicate that QQS regulation might be influenced by methylation. QQS is up-regulated in the hog1 mutant, which is hypomethylated (Jordan et al., 2007). Also, mRNA sequencing data predict that QQS transcript abundance will be very low in immature flowers; however, when DNA methylation levels in immature flowers are reduced, as in met1 or ddc DNA methyltransferase mutants, the QQS transcript level is substantially higher (Lister et al., 2008). This is consistent with expression analysis using microarrays and data from the QQS promoter–GUS transgenic lines, both of which show that little QQS transcript is present in immature flowers.

In an evolutionary context, movement of the multiple repeats/transposons that are replete in the QQS neighborhood in earlier times, may have contributed to control of its transcription and the uniqueness of its sequence. For example, CACTA transposons have been reported to be able to translocate proximal genetic material (Lister et al., 2008). Currently, no ESTs exist for any of the transposons-like sequences in the neighborhood of QQS, indicating these sequences may now be pseudogenes that are presumably no longer able to affect QQS structure.

Two lines of evidence indicate that QQS may mediate starch accumulation. The first line of evidence involves direct manipulation of QQS levels. Decreases in QQS expression increase starch accumulation, as shown by the finding that a genetically engineered decrease in QQS RNA, using QQS RNAi lines, results in an increased rate of starch synthesis. Second, QQS expression positively correlates with high starch accumulation in the Atss3 mutants, and also in mex1 mutants (M. Stettler and S.C. Zeeman, Institute of Plant Sciences, ETH, Zurich, personal communication), which have a high starch content, presumably owing to a block in starch catabolism (Niittyla et al., 2004). Consistent with this interpretation that QQS expression is influenced by starch metabolism, QQS expression is not up-regulated in the Atss2/Atss3 double mutant (which has a somewhat decreased starch content relative to WT; Zhang et al., 2008). Thus, expression of AtSS3 is eliminated in both the Atss3 single mutant and the Atss2/Atss3 double mutant. However, QQS is up-regulated only in the Atss3 mutant, which accumulates high levels of starch. These data indicate that up-regulation of QQS is not specifically associated with a mutation in AtSS3.

QQS induction could be caused by some aspect of the high-starch phenotype, or it could itself cause the abnormally high starch accumulation. The increased starch level in QQS RNAi mutants, in which QQS expression is reduced, is seemingly inconsistent with induction of QQS being responsible for the starch-excess phenotype. Thus, a more likely explanation of the data is that QQS expression is shifted in response to some feature of starch metabolism (Figure 10).

Figure 10.

 Proposed working model of QQS with respect to the multi-compartment starch metabolic and regulatory network (yellow cloud).
(a) In the QQS RNAi lines, a decrease in QQS expression induces starch biosynthesis via an unknown mechanism (dashed arrow), and starch accumulation is increased.
(b) In the Atss3 knockout mutant, starch accumulation is increased due to increased starch synthesis (Zhang et al., 2005). In addition, starch biosynthetic transcripts are increased. It is not clear how the Atss3 mutation mediates starch increase. For example, an altered form of the sucrose synthase complex could lead to signaling that increases starch biosynthetic transcript accumulation (dashed arrows). QQS transcript is also increased in Atss3, through an unknown mechanism, possibly involving changes in starch synthesis. The up-regulation of QQS is not associated specifically with the mutation in AtSS3; Atss3/Atss2 double mutants, which have starch levels that are slightly lower than normal, do not show a similar increase in QQS.

The RNAi data further indicate that QQS may function as a negative regulator of starch accumulation (Figure 10a). The simplest explanation of the high QQS expression in the Atss3 and mex1 high-starch mutants is that QQS is induced by some metabolic factor associated with increased starch content, and QQS itself provides a homeostatic function with respect to starch metabolism, responding to some aspect of starch metabolic signaling (Figure 10b). Our working hypothesis is thus that QQS may respond to some aspect of carbohydrate metabolic signals, and act by signaling the modulation of starch synthesis, thereby altering starch accumulation.

These data by themselves do not reveal what mechanistic role QQS plays in starch metabolism. Nor does the very little known about function of other plant PUFs help to elucidate QQS function – in the only example we are aware of, an intriguing role has been described for three PUFs that have an impact on the ability of Arabidopsis to withstand oxidative stress through some as yet unknown mechanism (Luhua et al., 2008).

The QQS protein contains no predicted functional motifs that would provide clues as to the molecular mechanism by which it regulates starch accumulation. One possibility is that the QQS protein directly interacts with a starch anabolic protein(s) to affect starch synthesis. This suggestion implies that QQS would be located within plastids; however, QQS–GFP localization indicates QQS is probably a cytosolic protein. Although it is possible that a small proportion of QQS is present in an organelle, or that the GFP tag somehow interferes with targeting, a cytosolic location is also consistent with the absence of any known targeting motif on the QQS protein. Thus, QQS probably does not act by physically modulating starch synthesis within the plastid. A second possibility is that QQS functions as a transcriptional factor that has an impact on expression of the network of starch metabolic genes; however, its apparent absence from the nucleus obviates a direct role in control of transcription. A third possibility, most consistent with the localization data, is that QQS exerts its regulatory function via interaction with cytosolic (or cytosolic-facing membrane-bound) factors that have an indirect impact on starch metabolism. Indeed, the level of the QQS transcript is increased more than fourfold in the sugar-insensitive sis8 mutant (S. Gibson, Department of Plant Biology, University of Minnesota, personal communication); this is consistent with a possible involvement in some aspect affecting the balance of carbon flow to sucrose. When QQS is reduced, as in the QQS RNAi mutants, more carbon could be diverted to starch; conversely, when starch accumulation is high, QQS expression is induced.

Regulation via protein–protein interaction has been experimentally shown to be a function of several non-plant PUF, and has been postulated as a general PUF characteristic (Gollery et al., 2006, 2007). The absence of homology between QQS and any other protein from other species, including Brassica napus, provides an additional intriguing quandary as to its function. One possibility is that Arabidopsis has evolved a unique mechanism to regulate metabolism. A second possibility is that other species contain functionally homologous proteins that provide a similar control mechanism, but that the primary sequences of these proteins are highly divergent. This could imply that their function may be conserved via higher orders of structure, i.e. structural homologs of QQS may be present in other plant species and act in a similar capacity on the basis of their three-dimensional similarity. Clarifying the function of the QQS gene will shed light on the function of PUFs in plants, and deepen the understanding of the mechanisms that provide a balance between homeostasis and metabolism in plants.

Experimental procedures

Plant material

Wild-type Arabidopsis thaliana ecotype Columbia (Col–0), Atss3 (SALK_065732) and Atss2/Atss3 mutant lines (crossed double mutant of SALK_065639 and SALK_065732) (Zhang et al., 2005, 2008) were used in this study.

Microarray data collection and analysis

The Atss3 mutant and WT plants were arranged according to a randomized complete block design. The plants were planted in pots in rows, seven rows in each flat; two plants of the same genotype per pot. Plants were grown under a SD photoperiod (8 h light/16 h dark) in a growth chamber. Eight rows from different flats were randomly selected and harvested for each time point at five time points across a diurnal cycle. The resultant samples (each sample consisting of leaf numbers 5–8 from 16 plants) were stored in liquid nitrogen. This growth and harvest process was then repeated in the same growth chamber to obtain a second sample for each genotype at each time point. Independent randomizations for plant growth and harvest were used for each of the two biological replicates. Each sample is considered as a biological replicate, and was analyzed using a single microarray chip.

The microarray data were transformed by log (MAS 5.0 signal + 1) (Affymetrix Microarray Suite Version 5, and median-centered. A linear mixed model was fitted to the data for each gene. Each model included fixed effects for replications, genotypes, times and genotype-by-time interaction, and random effects for replication-by-time interaction to allow for correlation among observations obtained from a single harvest occasion. To identify genes with expression profiles that differed between genotypes at one or more of the five time points, a single F-test was performed for each gene as part of each mixed linear model analysis. Based on the P values obtained using these F-tests, the method described by Storey and Tibshirani (2003) was used to estimate the false discovery rate (FDR).

5′ and 3′ RACE

5′ and 3′ RACE experiments were performed to define the 5′ and 3′ UTRs of the QQS gene using the GeneRacer kit (Invitrogen Life Technologies, Total RNA was extracted from the leaf tissue of WT and Atss3 plants grown under LD conditions. The RNA was decapped and ligated to the GeneRacer RNA oligo, and GeneRacer Oligo dT primer was used to reverse transcribe this RNA (for 5′ RACE) or total RNA (for 3′ RACE), using SuperScript™ III reverse transcriptase. GeneRacer reactions were performed using the touchdown PCR conditions recommended by the manufacturer and an annealing temperature of 60°C. The primers used were 5′-CGACTGGAGCACGAGGACACTGA-3′ and 5′-GTAGAACTGAAGCCCGACCCATGA-3′ for 5′ RACE, and 5′-CATTGAAGAAGCCTCCTCTCATTACC-3′ and 5′-GCTGTCAACGATACGCTACGTAACG-3′ for 3′ RACE. High-fidelity platinum Taq DNA polymerase (Invitrogen Life Technologies) was used for all reactions. The PCR products were identified on 1% agarose gels, and sequenced.

Starch content

For I2/KI staining of plants, the method used was as previously described (Li et al., 2007). At least ten plants from each of six independent T2 lines were analyzed, and this experiment was repeated twice.

Starch was quantified as previously described (Zhang et al., 2005) using amyloglucosidase and GOPOD (Megazyme, Three plants from each of four independent T2 lines and WT were analyzed, and this experiment was repeated three times.

Molecular and biochemical methods

Vector construction, Arabidopsis transformation and selection, microscopy and histochemical analysis, and protein blots were conducted by standard methods as specified in Appendix S1.

Bioinformatics analysis

MetaOmGraph was used to analyze the expression patterns of starch debranching enzymes and starch-related genes using the normalized experimental data and metadata (metadata includes gene, experiment and sample annotations) from 70 experiments comprising 956 Affymetrix ATH1 microarray slides (Wurtele et al., 2007), available online (

cis-acting motifs in QQS were evaluated as previously described (Li et al., 2007) and using Motif Search (

Accession numbers

Sequence data from this article are available in the Arabidopsis Genome Information Resource under the following accession numbers: QQS, At3g30720; AtSS2, At3g01180; AtSS3, At1g11720. The GenBank accession number for QQS (defined by 5′ and 3′ RACE) is EU805808. The GEO accession number ( for the microarray data discussed here is GSE11708.


We are grateful to Xiaoli Zhang and Christophe Colleoni for help in harvesting leaves for the microarray experiments, and Basil Nikolau for discussion on RACE. We thank Ryan Lister for thoughtful insight on the QQS chromosome region, and Drena Dobbs for her valuable input on the QQS structural motif. We particularly thank Nick Ransom for rapid addition of functions that expand the functionality within the MetaOmGraph software. This work was supported by the National Science Foundation Arabidopsis 2010 program: grants DBI 0520267 and MCB 0209789.

The GenBank accession number for the QQS sequence is EU805808.