GLK transcription factors regulate chloroplast development in a cell-autonomous manner


(fax +44 1865 275074; e-mail


In a number of land plants, Golden2-like (GLK) genes encode a pair of partially redundant nuclear transcription factors that are required for the expression of nuclear photosynthetic genes and for chloroplast development. As chloroplast biogenesis depends on close co-operation between the nuclear and plastid compartments, GLK gene function must be dependent on tight intracellular control. However, the extent to which GLK-mediated chloroplast development depends on intercellular communication is not known. Here we used sector analysis to show that GLK proteins operate cell-autonomously in leaf mesophyll cells. To establish whether GLK proteins are able to influence adjacent cell layers, we used tissue-specific promoters to restrict GLK gene expression to the epidermis and to the phloem. GLK genes driven by the Arabidopsis epidermal FIDDLEHEAD (FDH) and MERISTEM LAYER1 (AtML1) promoters failed to rescue the pale-green Atglk1 Atglk2 mutant phenotype, confirming the suggestion that GLK proteins can only influence chloroplast development intracellularly. An exception to this rule was seen in lines in which GLK genes were expressed in the phloem. However, we believe that the partial complementation of the mutant phenotype that was observed resulted from phloem unloading, as opposed to inherent properties of GLK proteins. We conclude that GLK proteins act in a cell-autonomous manner to coordinate and maintain the photosynthetic apparatus within individual cells. Significantly, this suggests that GLK proteins provide a means to fine-tune photosynthesis according to the differential requirements of cells within the leaf.


Chloroplasts originated following an endosymbiotic event between a photosynthetic cyanobacterium and a eukaryotic host. It is generally accepted that this ancestral symbiont gave rise to the three primary groups of extant photosynthetic eukaryotes – the Chlorophyta, Rhodophyta and Glaucophyta (Larkum et al., 2007). During evolution of these groups, most of the original cyanobacterial genes were transferred to the host nucleus (Martin et al., 2002). In land plants, which are most closely related to the Chlorophyta, relatively few photosynthesis-related genes remain in the chloroplast genome. The proteins required for photosynthesis are therefore synthesized in the cytosol and transported across the chloroplast envelope by a protein import complex. Subsequent intraorganellar targeting mechanisms direct proteins to the stroma, thylakoids or thylakoid lumen, whereupon they form functional protein complexes by combining with chloroplast-encoded subunits. Assembly of an operational chloroplast thus requires a balanced contribution from two separate genomes.

To coordinate the expression of nuclear gene products with that of those encoded in the chloroplast, the nucleus responds to well-characterized retrograde signals originating from the plastid during seedling de-etiolation (Woodson and Chory, 2008). After seedling establishment, chloroplast biogenesis is a continual and essential component of leaf organogenesis that requires maintained communication between the plastid and nucleus. Nuclear transcription factors act on multiple promoter regions to synchronize the expression of functionally related gene targets, and a mechanism for nuclear control of plastid development can thus be envisaged. However, very few transcription factors have been shown to positively regulate chloroplast biogenesis. One exception is the Golden2-like (GLK) gene family, which encodes nuclear transcription factors that have been shown to regulate chloroplast development in Physcomitrella patens (moss), Zea mays (maize) and Arabidopsis (Fitter et al., 2002; Hall et al., 1998; Rossini et al., 2001; Yasumura et al., 2005).

In both moss and Arabidopsis, mutant plants lacking GLK gene function are pale green. Mutants exhibit reduced amounts of thylakoid membrane, with a particular defect in granal stacking. Furthermore, the expression levels of nuclear-encoded photosynthesis-related genes are reduced, especially those associated with chlorophyll biosynthesis and photosystem II. Assembly of the photosystems located on the thylakoid membrane is regulated in a stoichiometric fashion, whereby chlorophyll is integrated into light-harvesting chlorophyll-binding (LHCB) proteins concomitantly with protein folding and membrane insertion (Hoober et al., 2007). Due to the complex feedback between the plastid and nucleus, a defect in the expression of chlorophyll biosynthesis genes results in decreased expression of LHCB genes, and vice versa. Similarly, reduced thylakoid content may feed back upon chlorophyll and photosystem abundance, as the thylakoid membrane is the site of the light-harvesting machinery. Due to this confounding of primary and secondary effects of glk mutations, the principal function of GLK genes is unclear. Nevertheless, their functional conservation in anciently diverged plants demonstrates that GLK genes are fundamentally required for chloroplast development.

In Arabidopsis and moss, each of two GLK genes acts redundantly to direct monomorphic chloroplast development: single glk mutants largely resemble wild-type, whereas double glk1 glk2 mutants are pale green (Fitter et al., 2002; Yasumura et al., 2005). In maize, however, GLK genes appear to act in a cell-type-specific manner to direct the development of dimorphic chloroplasts. The golden2 mutant is specifically defective in bundle sheath chloroplast development, but mesophyll chloroplasts are unaffected (Hall et al., 1998; Langdale and Kidner, 1994). GOLDEN2 (G2) is expressed in bundle sheath cells, and its homologue, ZmGLK1, is expressed in mesophyll cells (Rossini et al., 2001). Although a zmglk1 mutant has yet to be identified, the cell-specific expression pattern and phenotype of g2 mutants suggest that GLK genes act cell-autonomously.

One technique for assaying cell autonomy is to mis-express the protein of interest and monitor its activity in cells beyond the expression domain. The Arabidopsis shoot apical meristem (SAM) comprises three cell layers, L1, L2 and L3, which define the subsequent cell layers of developing leaves. L1 forms the leaf epidermis, L2 generates the majority of mesophyll cells, and L3 contributes the inner core mesophyll and the leaf vasculature (Tilney-Bassett, 1986). Epidermis-specific expression can be directed by the FIDDLEHEAD (FDH) and MERISTEM LAYER1 (AtML1) promoters, both of which are active within the L1 layer of the seedling and throughout development (Sessions et al., 1999; Yephremov et al., 1999). The AtML1 promoter has been used to demonstrate the lack of cell autonomy of the transcription factors KNOTTED1 (Kim et al., 2002) and LEAFY (Sessions et al., 2000). Likewise, the FDH promoter has been used to demonstrate that class B MADS box transcription factors can still regulate floral organ identity when expressed only in the epidermis (Efremova et al., 2004). No comparable promoter exists for L2-specific expression, but expression can be restricted to the photosynthetic mesophyll using the RbcS promoter (Kyozuka et al., 1993). Similarly, no reliable promoter exists that is only active in L3-derived cells, but expression can be restricted to phloem companion cells by using the AtSUC2 promoter. AtSUC2 is active primarily during phloem loading in source leaves (Imlau et al., 1999), but also in sink leaves, cotyledons and the stem (Martens et al., 2006). The AtSUC2 promoter has been used to show that phloem-specific expression of CONSTANS is sufficient to induce flowering (An et al., 2004), and that FLOWERING LOCUS T moves from the phloem into the meristem to act as a flowering signal (Corbesier et al., 2007). Thus, gene expression driven by tissue-specific promoters can illustrate the extent to which cellular differentiation in any one cell layer is dependent on signals from neighbouring cell layers.

Knowing the extent to which a transcription factor acts cell-autonomously is critical to understanding the developmental level at which that transcription factor acts. For example, KNOTTED1 (KN1), which is required for the maintenance of undifferentiated stem cells in the SAM (Vollbrecht et al., 2000), can traffic between cells in the SAM via plasmodesmata (Kim et al., 2002). KN1 thus acts non-cell-autonomously across the entire meristem to control cell fate. Conversely, SCARECROW acts cell-autonomously to maintain cell fate in the root by sequestering the mobile transcription factor SHORT-ROOT in the nuclei of endodermal cells. This activity prevents SHORT-ROOT from moving into the next cell file, and thus restricts the endodermis to a single file (Cui et al., 2007; Heidstra et al., 2004). SCARECROW is therefore an example of a transcription factor that operates in a subset of cell types within the root. In the case of GLK genes, it is not known whether control over chloroplast development takes place at the whole-leaf, sub-domain or single-cell level.

To gain insight into the mode of GLK gene action, we investigated the extent to which GLK proteins act cell-autonomously. We first confirmed that AtGLK1 and AtGLK2 are functionally equivalent. Using a combination of sector analysis and layer-specific gene expression, we show that both GLK proteins direct chloroplast development in a cell-autonomous manner. However, when mis-expressed in the phloem, GLK protein and/or mRNA can be unloaded into adjoining photosynthetic tissue, thus influencing adjacent cells in a limited manner. In conclusion, these results suggest that GLK-dependent regulation of chloroplast development is an integrative, continual process that occurs at the single-cell level, providing a means for cell-specific control over photosynthesis.


AtGLK1 and AtGLK2 are functionally equivalent

To study the cell autonomy of GLK proteins, it was first necessary to establish whether the action of either AtGLK1 or AtGLK2 alone is sufficient to promote normal chloroplast development. In wild-type plants, both AtGLK1 and AtGLK2 are expressed in vegetative tissues, and only double mutants have pale-green leaves. This implies that expression of either GLK gene is sufficient for dark-green leaves. However, only AtGLK2 is expressed in siliques, such that Atglk2 mutants can be distinguished from wild-type by their pale siliques (Fitter et al., 2002). This observation may reflect a functional difference between AtGLK1 and AtGLK2. To determine whether AtGLK1 and AtGLK2 proteins are functionally equivalent, full-length cDNAs encoding AtGLK1 or AtGLK2 were constitutively expressed in the Atglk1 Atglk2 mutant background. For each gene, three independent lines that carried either one or two transgene copies were selected for further analysis (Figure 1a). AtGLK1 and AtGLK2 transcript levels were substantially and consistently higher in each transgenic line than in wild-type plants (Figure 1a). Over-expression of either AtGLK1 or AtGLK2 led to complete restoration of leaf chlorophyll (Chl) levels, with most transgenic lines accumulating more chlorophyll than wild-type (Figure 1b,c). Importantly, over-expression of AtGLK1 rescued the pale-green silique phenotype of the double mutant (Figure 1d). This observation confirms that AtGLK1 can functionally substitute for AtGLK2, and implies that the silique phenotype of Atglk2 mutants results from lack of AtGLK1 expression in this tissue. In addition to pale pigmentation, Atglk1 Atglk2 mutants accumulate lower levels of several photosynthesis-related transcripts, such as LHCB6, which encodes a monomeric protein of the photosystem II antenna (Fitter et al., 2002). Constitutive expression of either AtGLK1 or AtGLK2 completely restored the level of LHCB6 transcripts to that of wild-type (Figure 1e), further confirming that AtGLK1 and AtGLK2 are functionally interchangeable.

Figure 1.

AtGLK1 and AtGLK2 are functionally equivalent.
(a) Identification of transgenic lines over-expressing AtGLK1 or AtGLK2 in an Atglk1 Atglk2 mutant background. Upper panels: DNA blot analysis to determine transgene copy number. Genomic DNA was digested with BglII (35S::AtGLK1, left) or BamHI (35S::AtGLK2, right), each of which each cuts once in the T-DNA of the 35S construct. DNA blots were hybridized with radiolabelled AtGLK1 (left) or AtGLK2 (right) probes. Arrowheads indicate transgene-derived bands; m, mutant allele; wt, wild-type allele. Middle panels: Northern blot analysis to verify transgene expression. Five micrograms of total RNA were loaded per lane. Lower panels: the 25S rRNA band from the ethidium bromide-stained gel was used to verify loading levels for the RNA blot. RNA blots were hybridized with radiolabelled AtGLK1 (left) or AtGLK2 (right) probes.
(b) Mean total chlorophyll content from fully expanded leaves of 3-week-old plants of lines identified in (a). Error bars represent SE (= 9).
(c) Pigmentation and flowering phenotypes of wild-type, Atglk1 Atglk2, 35S::GLK1 1.1 and 35S::GLK2 2.1 (left to right). Plants were grown under long-day conditions and photographed at 4 weeks.
(d) Silique pigmentation phenotype of lines shown in (c).
(e) LHCB6 transcript accumulation in lines shown in (c). Ten micrograms of total RNA was loaded per lane. The 25S rRNA band from the ethidium bromide-stained gel was used to verify loading levels.

Notably, GLK over-expression increased time to flowering compared with wild-type and the Atglk1 Atglk2 mutant background (Figure 1c and Table 1). Under long-day conditions, Atglk1 Atglk2 plants produce fewer vegetative leaves than wild-type but flower within a similar time frame, suggesting accelerated development through the vegetative phase (Table 1). However, this developmental difference becomes negligible under short days, under which Atglk1 Atglk2 plants take longer than wild-type to produce an equivalent number of leaves, perhaps because of reduced photosynthetic capacity. Surprisingly, over-expression of AtGLK1 and AtGLK2 substantially delays flowering, especially under long-day conditions. Given the early flowering phenotype of Atglk1 Atglk2 mutants, this observation suggests that GLK proteins suppress flowering. The similar effects of both AtGLK1 and AtGLK2 on chlorophyll levels and flowering time are consistent with functional equivalence.

Table 1.   Over-expression of AtGLK1 and AtGLK2 delays flowering
 Wild-typeglk1 glk235S::GLK1 1.135S::GLK2 2.1
  1. Time to flowering is shown with respect to both the number of leaves visible and the number of days after planting before the first inflorescence bolt was 1 cm high. n = number of plants scored. Plants were randomized and grown under controlled environmental conditions with either long days (16 h light, 8 h dark) or short days (8 h light, 16 h dark). Lines are those shown in Figure 1(c).

Long days
 Leaves19.5 ± 0.515.0 ± 0.437.1 ± 0.834.9 ± 0.9
 Days44.1 ± 0.745.5 ± 0.868.5 ± 0.962.0 ± 1.4
Short days
 Leaves55.6 ± 1.454.3 ± 0.862.0 ± 1.663.1 ± 1.0
 Days86.3 ± 1.194.8 ± 1.593.3 ± 2.690.7 ± 1.8

Sector analysis suggests cell-autonomous action of GLK proteins in mesophyll tissue

Information regarding how GLK proteins direct chloroplast development within a developing leaf can be obtained through the analysis of wild-type sectors in an otherwise mutant genetic background. To this end, we characterized a sectored plant that was discovered while selecting forAtglk1 Atglk2 mutants that were transformed with a construct encoding a GFP–AtGLK1 fusion protein. This construct typically rescued the mutant phenotype (data not shown), but one primary transformant appeared to be a sectoral chimera of phenotypically wild-type and mutant tissue (Figure 2a). Expression analysis of pale- and dark-green tissue determined that, although both tissue types accumulated GFP–AtGLK1 transcripts, levels were 8–16-fold higher in dark-green tissue as opposed to pale-green tissue (Figure 2c). The sectors persisted throughout development, generating inflorescence shoots of both phenotypes and thus providing a means to propagate each genotype independently (Figure 2b).

Figure 2.

 Sector analysis suggests cell autonomy of GLK proteins in the medio-lateral axis.
(a) Phenotype of a sectored Atglk1 Atglk2 mutant plant transformed with 35S::GFP-AtGLK1. Several leaves (arrows) show regions of both pale- and dark-green tissue phenotypes.
(b) The sector was maintained into the reproductive phase, generating fully pale- and fully dark-green inflorescence shoots.
(c) Semi-quantitative RT-PCR analysis of GFP–AtGLK1 transcript levels in dark- and pale-green sectored tissue. RNA was isolated from non-sectored cauline leaves of each phenotype. RT-PCR was performed using primers spanning part of the GFP and AtGLK1 coding regions (upper panel) for the indicated number of cycles. Equal amounts of cDNA template were confirmed by amplifying the housekeeping gene GAPDH (lower panel). Each primer pair generated a single amplicon of the expected size. Approximately equivalent levels of mGFP6–AtGLK1 product were obtained after 26 and 30 cycles in dark- and pale-green tissue, respectively. A difference of four cycles during the geometric phase of amplification corresponds to a 24- or 16-fold difference in transcript levels. +/− denotes the presence/absence of reverse transcriptase (RT).
(d, e) Confocal laser scanning micrographs of palisade mesophyll cells from a wild-type (d) or Atglk1 Atglk2 (e) leaf visualized by chlorophyll autofluorescence and imaged using identical microscopy parameters.
(f, g) Low- (f) and high- (g) magnification confocal micrographs of a sector boundary; wild-type-like tissue is on the left, mutant-like tissue on the right. Arrows indicate larger, brighter mesophyll chloroplasts; arrowheads indicate smaller epidermal plastids.
Scale bars = 40 μm (c,d,f) and 100 μm (e).

To determine whether the boundary between pale- and dark-green tissue was discrete, we used confocal microscopy to view the palisade mesophyll cells immediately below the adaxial (upper) surface of the leaf. Chlorophyll autofluorescence was used as a proxy for chlorophyll content. In wild-type tissue, mesophyll chloroplasts are larger and emit a stronger chlorophyll signal than those in glk1 glk2 mutant tissue (Figure 2d,e). When a sector boundary was examined, a clear, sharp distinction was observed between chloroplasts in the two halves (Figure 2f,g). Cells at the very edge were either fully wild-type or fully mutant in appearance, with no intermediate phenotypes or mixed populations of chloroplasts within a single cell. This observation implies that chloroplasts within any particular cell develop independently from those in an adjacent cell, and that this effect is a result of the cell-autonomous action of GLK proteins in the medio-lateral leaf axis.

Tissue-specific expression of GLK genes leads to variable degrees of mutant complementation

To determine whether GLK proteins can act cell-autonomously in the abaxial–adaxial leaf axis, we expressed GLK1 and GLK2 in specific cell layers of Atglk1 Atglk2 mutant leaves (Figure 3a). As leaf chloroplasts develop primarily in mesophyll cells, we used chlorophyll levels and mesophyll chloroplast morphology to detect GLK activity. To observe potential trafficking of GLK proteins, all constructs encoded C-terminal translational fusions with GFP, and, to confirm that the GLK–GFP fusions complemented the mutant phenotype, we also expressed them constitutively under the control of the 35S promoter (Figure 3a).

Figure 3.

 Tissue-specific expression of GLK–GFP fusion proteins.
(a) Diagrammatic representation of DNA constructs and their expression domains in a leaf cross-section. Constructs encoded either AtGLK1 or AtGLK2 followed by GFP in a translational fusion, driven by various tissue-specific promoters. The FDH promoter is active in the epidermis (blue); the Nicotiana tabaccum RbcS promoter is active in the photosynthetic mesophyll and bundle sheath (dark green); the AtSUC2 promoter is active in the phloem companion cells (red) associated with phloem sieve cells (pink) within the vascular bundle (grey).
(b) Whole-plant phenotypes of representative lines at 3 weeks. Genotypes: (1) wild-type; (2) Atglk1 Atglk2; (3, 5, 7, 9) GLK1–GFP and (4, 6, 8, 10) GLK2–GFP driven by FDH (3,4), RbcS (5,6) AtSUC2 (7,8) or 35S (9,10) promoters.
(c) Mean total chlorophyll content the genotypes shown in (b). For (3)–(10), each bar corresponds to pooled measurements from at least two independent lines. Error bars represent SE ( 10).
(d) LHCB6 transcript accumulation in individual lines as shown in (b). Ten micrograms of total RNA were loaded per lane. Transcript levels (bottom) are expressed relative to wild-type after standardization to the 25S rRNA band from the ethidium bromide-stained gel.

Atglk1 Atglk2 mutants carrying layer-specific expression constructs exhibited varying phenotypes. In terms of gross morphology, all transgenic lines resembled the wild-type (Figure 3b). However, whereas both 35S and RbcS lines were dark green, FDH and AtSUC2 lines were paler than wild-type. Chlorophyll assays showed wild-type levels in 35S and RbcS lines, intermediate levels in AtSUC2 lines and mutant levels in FDH lines (Figure 3c). As an additional indicator of phenotypic rescue, we assessed LHCB6 transcript levels in representative lines containing each construct. The levels of LHCB6 transcripts broadly correlated with the degree of pigmentation, with the levels in RbcS and 35S GLK–GFP lines approaching those of wild-type (Figure 3d). Consistent with the known functional equivalencies of AtGLK1 and AtGLK2, there were no discernible phenotypic differences between plants expressing GLK1–GFP and those expressing GLK2–GFP. We were unable to detect GFP fluorescence in any line. This was particularly surprising in the case of 35S::GLK-GFP lines, as complementation of the Atglk1 Atglk2 mutant phenotype indicates that functional protein is produced in these plants. We were only able to detect truncated proteins from plant extracts, and fusion proteins synthesized in vitro tended to degrade (Supplementary Figure S1). Despite our failure to detect full-length GLK–GFP fusions in the transgenic lines, these observations suggest that GLK expression in the leaf mesophyll fully rescues the mutant phenotype, but expression in the epidermis and phloem does not.

GLK–GFP expression in the epidermis cannot drive chloroplast development in the mesophyll

To investigate more thoroughly the effect of epidermis-specific GLK expression on chloroplast development in the underlying mesophyll, we characterized three independently transformed lines of FDH::GLK1-GFP and FDH::GLK2-GFP. In all cases, both fully expanded and developing leaves (Figure 4a) and seedlings (data not shown) were visually indistinguishable from the mutant background. Likewise, chlorophyll levels and LHCB6 transcript levels were essentially the same as in the mutant (Figure 4b,c). To verify that the GLK genes were transcribed from the construct, total RNA was extracted from expanded leaves of 21-day-old plants. Transcript levels from an epidermis-specific promoter are expected to be relatively low because of the small proportion of RNA that the epidermis contributes to the whole-leaf extract, but transcripts were detected in all lines. There was some between-line variability in transcript levels, but this did not correlate with chlorophyll content or LHCB6 transcript levels (Figure 4d). Thus, epidermis-specific accumulation of GLK–GFP mRNA does not have an impact upon chloroplast development in the underlying mesophyll layer.

Figure 4.

 Epidermally expressed GLK–GFP does not rescue the Atglk1 Atglk2 mutant phenotype.
(a) Fifth (left) and ninth (right) leaves taken from 3-week-old individuals of representative FDH::GLK-GFP lines.
(b) Mean total chlorophyll levels in fully expanded fifth leaves of independent FDH::GLK-GFP lines. One sample was taken per leaf. Error bars represent SE (= 7).
(c) Mean total chlorophyll levels of whole 9-day-old seedlings of each line. Error bars represent SE (= five batches of eight seedlings).
(d) Transcript accumulation of GLK1–GFP and GLK2–GFP (upper panel) and LHCB6 (middle panel) in each independent line. Note that GLK1–GFP and GLK2–GFP are detected using the same GFP probe but produce transcripts of different sizes. The 25S rRNA band from the ethidium bromide-stained gel was used to verify loading levels (lower panel).

AtML1-driven expression leads to chloroplast development in the cotyledon margin

To check whether the failure of epidermis-specific GLK expression to rescue the mutant phenotype was not due to a specific feature of the FDH promoter, we conducted the same experiment using a second epidermis-specific promoter, AtML1. Mature leaves were pale green (data not shown), but close inspection of AtML1::GLK-GFP lines revealed a dark-green edge to cotyledons (Figure 5). This dark-green margin was sufficient to lead to a measurable increase in total chlorophyll levels in the seedlings of five independent lines expressing AtML1::GLK-GFP (Figure 5g). Chlorophyll autofluorescence under low-magnification confocal microscopy allowed better visualization of this chlorophyllous margin (Figure 5a,b,f). As epidermal cells do not contain fully developed chloroplasts (Dupree et al., 1991), we postulated that the observed chlorophyll was located in mesophyll cells. To confirm this suggestion, transmission electron microscopy (TEM) was used to examine thylakoid morphology within various regions of the cotyledon. Mesophyll chloroplasts located at the edge of the cotyledon showed grana resembling those of wild-type chloroplasts (Figure 5c,h), whereas chloroplasts closer to the centre contained rudimentary thylakoids that are essentially the same as seen in Atglk1 Atglk2 chloroplasts (Figure 5d,i). The cotyledon-margin phenotype was specific to the AtML1 promoter, and was not observed with the FDH promoter. This difference is unlikely to be a consequence of promoter strength, as transcript levels in cotyledons of AtML1 lines were not consistently higher than in FDH lines – indeed, the strongest expressing line was FDH::GLK2-GFP 2.3 (Supplementary Figure S2). It is possible that, in contrast to previous reports using in situ hybridizations (Lu et al., 1996; Sessions et al., 1999), the AtML1 promoter is active at some stage in embryogenesis in sub-epidermal tissues. Indeed, a sensitive multiple-GFP reporter gene revealed AtML1 expression in sub-epidermal layers up to the 32-cell stage embryo (Takada and Jürgens, 2007). Alternatively, the margin could represent a novel domain in cotyledons, in which the L1 meristem layer contributes cells to the underlying layers that later adopt mesophyll cell fate. In either case, given that the FDH promoter contains an L1-box element recognized by AtML1 (Abe et al., 2001, 2003), FDH may act downstream of AtML1 to define a more precise epidermal expression domain. Considering all possibilities, we favour the interpretation that GLK proteins act cell-autonomously in the L1 meristem layer and in the derived epidermal leaf layers.

Figure 5.

AtML1-driven GLK–GFP leads to chloroplast development in the mesophyll cells of cotyledon margins.
(a, b) Low-magnification confocal micrographs of wild-type (a) and Atglk1 Atglk2 (b) cotyledons from 7-day-old seedlings visualized by chlorophyll autofluorescence and imaged under identical microscopy parameters. Scale bar = 400 μm. Insets show equivalent whole seedlings.
(c, d) Thylakoid ultrastructure of mesophyll chloroplasts from wild-type (c) and Atglk1 Atglk2 (d) cotyledons. Brackets indicate grana. Insets show lower-magnification views of the same chloroplast. Scale bars = 200 nm (main image) and 500 nm (insets).
(f) Cotyledon of a seedling expressing AtML1::GLK2-GFP, imaged as in (a,b).
(g) Mean total chlorophyll levels of whole 9-day-old seedlings of each line. Error bars represent SE (= five batches of eight seedlings).
(h, i) Thylakoid ultrastructure of mesophyll chloroplasts located in the margin (h) and more central region (i) of a cotyledon of a seedling expressing AtML1::GLK1-GFP. Scale bars and insets as in (c,d).

GLK–GFP expression in phloem companion cells induces chloroplast development in adjacent mesophyll cells

Chlorophyll analysis indicated that expression of GLK–GFP in mesophyll tissue using the RbcS promoter fully complemented the Atglk1 Atglk2 mutant phenotype (Figure 3). To confirm this, we examined chloroplast ultrastructure. Wild-type leaf mesophyll cell chloroplasts contain grana with a mean of seven appressed thylakoids per stack, whereas Atglk1 Atglk2 chloroplasts contain two to three (Figure 6a,b,i). When RbcS::GLK1-GFP is expressed in the mutant background, the granal phenotype is restored to wild-type. Thus, mesophyll expression of GLK–GFP leads to normal chloroplast development (Figure 6c,i). To establish whether GLK–GFP expression in non-photosynthetic, L3-derived layers of the leaf could influence the chloroplast ultrastructure in mesophyll cells, we characterized two independent lines each of AtSUC2::GLK1-GFP and AtSUC2::GLK2-GFP. All four lines accumulated GLK–GFP transcripts (Figure 6d). In AtSUC2::GLK-GFP plants, mesophyll chloroplasts at least four cells from a vein develop granal stacks that are intermediate in size between wild-type and Atglk1 Atglk2 plants. Quantification of the number of thylakoids per stack confirmed the intermediate nature of phenotypic rescue (Figure 6i), consistent with chlorophyll accumulation levels and LHCB6 transcript levels in these lines (Figure 3). These observations suggest that GLK transcription factors can act non-cell-autonomously when expressed in the phloem, but only to a limited extent.

Figure 6.

 Phloem companion cell-specific expression of GLK–GFP partially complements the Atglk1 Atglk2 mutant phenotype.
(a–c) Thylakoid ultrastructure of leaf mesophyll chloroplasts of wild-type (a), Atglk1 Atglk2 (b) and RbcS:GLK1-GFP (c) lines. Brackets indicate grana. Insets show lower-magnification views of the same chloroplast.
(d) RNA blot analysis of independent lines expressing GLK1–GFP and GLK2–GFP from the AtSUC2 promoter, probed with GFP.
(e–h) Electron micrographs from a plant expressing AtSUC2::GLK2-GFP. (e) Cross-section of a leaf minor vein. bs, bundle sheath; cc, phloem companion cell; pp, phloem parenchyma; se, phloem sieve element; x, xylem vessel. (f) Chloroplast ultrastructure in a spongy mesophyll cell immediately adjacent to the bundle sheath, labelled with a star in (e). (g, h) Chloroplast ultrastructure in (g) spongy and (h) palisade mesophyll cells 3–4 cells away from the nearest visible vein.
(i) Quantification of granal stacking in mesophyll cells. A thylakoid is defined as a pair of membranes enclosing a single lumenal space. Thylakoids were counted in five grana each in at least six chloroplasts, using one chloroplast per cell and two cells per leaf ( 6 chloroplasts; error bars represent SE). A single representative line was used for RbcS and AtSUC2.
Scale bars = 200 nm (main image, a–c,f–h), 500 nm (insets, a–c,f–h) and 5 μm (e).


We have shown that, with the exception of some limited phloem unloading, GLK transcription factors act in a cell-autonomous manner to regulate chloroplast development. An Arabidopsis mesophyll cell contains up to 150 chloroplasts – the precise number varies but is closely correlated with cell size (Pyke and Leech, 1994). The nucleus receives retrograde signals from multiple chloroplasts but must integrate them into a single response that applies equally to each individual chloroplast. The single cell is one scale at which the nucleus can influence chloroplast development, and indeed such control may be advantageous considering the heterogeneity between photosynthetic cells across the leaf. However, it is notable that in C4 members of the Chenopodiaceae, two types of chloroplast can exist in a single cell (Voznesenskaya et al., 2002), suggesting that control over chloroplast development may exist at an even finer scale. In the case of C4 plants with Kranz anatomy, discrete biochemical and structural compartments across the leaf are essential for operation of the C4 cycle. The cell-autonomous function of GLK proteins is fully consistent with their predicted role in directing cell-specific C4 chloroplast development (Rossini et al., 2001). In Arabidopsis, however, there is no evidence for a cell-specific role for each GLK protein except in siliques, suggesting that GLK1 and GLK2 are largely redundant in C3 photosynthetic species.

During cell expansion, chloroplasts are formed by binary fission from pre-existing chloroplasts. The formation of new chloroplasts requires the synthesis of new proteins and pigments, and this synthesis relies on constant forward input from the nucleus. Two observations suggest that GLK genes play a continual role in promoting this process. First, random silencing in lines over-expressing GLK genes in the mutant background leads to localized pale regions on otherwise dark leaves (Figure 2, and unpublished results). Second, unstable (g2-bsd1-m1 and g2-pg14) mutant alleles of maize G2 generate dark-green revertant sectors of varying sizes, suggesting that GLK genes are necessary for normal chloroplast biogenesis at all stages of leaf development (Hall et al., 1998). These observations imply that GLK genes act relatively late in the transition from a proplastid to a fully photosynthetic chloroplast. This suggestion is supported by the fact that Atglk1 Atglk2 mutants are principally defective in pigmentation and thylakoid stacking, but are otherwise morphologically similar to wild-type and exhibit no defects in cell or leaf development. Intriguingly, epidermal chloroplasts remain small and relatively under-developed compared to mesophyll chloroplasts even in plants over-expressing GLK proteins (Figure 2, and unpublished results), suggesting that other cell-specific factors inhibit chloroplast development in these cells. Both AtGLK1 and AtGLK2 are expressed at all stages of leaf development, both are light-induced and AtGLK2 is circadian-regulated (Fitter et al., 2002), further implying a constant requirement for GLK proteins even when chloroplasts are fully developed. The light and circadian regulation of GLK proteins is reflected in the expression patterns of putative GLK targets. Using inducible expression constructs in a mutant background, we have obtained preliminary evidence showing that expression of key chlorophyll biosynthetic genes is rapidly increased following the induction of AtGLK1 and AtGLK2 (M.T.W. and J.A.L., unpublished data). Furthermore, induction of GLK expression disrupts the normal diurnal regulation of LHCB1 and LHCB6, by maintaining high levels of LHCB transcripts at times when normally they would decrease. These observations suggest that both chlorophyll biosynthetic genes and light-harvesting components are regulated by GLK proteins. Thus, we conclude that GLK proteins act cell-autonomously as part of a continual process to promote and maintain an efficient photosynthetic system.

Expression of GLK–GFP in phloem companion cells led to partial complementation of the mutant phenotype (Figures 3 and 6). How can this be explained in light of the data indicating cell autonomy in the mesophyll (Figure 2)? The mature phloem sieve element is metabolically supported by the adjoining phloem companion cell, and the two cells are cytoplasmically coupled through specialized plasmodesmata (Oparka and Turgeon, 1999). The phloem is functionally divided into collection phloem, transport phloem and release phloem. The AtSUC2 H+-sucrose symporter loads sucrose synthesized in source leaves into the collection phloem sieve element, and sucrose is then unloaded in sink tissues via the release phloem (Van Bel, 2003). When free GFP is expressed in companion cells under the control of the AtSUC2 promoter, it is transported into sieve elements and subsequently released into surrounding tissues in sink leaves where it undergoes extensive post-phloem transport (Imlau et al., 1999). The phloem can thus transport metabolites and large macromolecules from source tissues and release them in sink tissues. An investigation into phloem transport has shown that cytoplasmic GFP fusions of at least 67 kDa can move between companion cells and sieve elements, and thus from source to sink tissues (Stadler et al., 2005). GLK1–GFP and GLK2–GFP are 74 and 70 kDa, respectively, but, given that molecular mass may not accurately reflect the physical dimensions of globular proteins, it is reasonable to suppose that these proteins may also be loaded into the phloem sieve element. However, we observed no difference in phenotype between source leaves and developing sink leaves in AtSUC2::GLK-GFP lines. As the AtSUC2 promoter is active in transport phloem (Martens et al., 2006), which includes the mid-rib of sink leaves, it is therefore likely that GLK–GFP is unloaded from the sieve element directly in sink leaves rather than originating in source leaves and subsequently transported to sink leaves. Thus, we conclude that the apparent movement of GLK–GFP from phloem companion cells probably reflects the special phenomenon of phloem unloading of macromolecules, as opposed to inherent properties of GLK proteins.

Establishing a molecular function for GLK genes has been hampered by protein instability and low protein abundance in planta. Whilst it is possible that the inability to detect GLK–GFP protein resulted from the fusion with GFP, this is unlikely because (i) fusions with GFP at the N- or C-terminus are both undetectable, thus ruling out artefacts relating to the position of the GFP; (ii) the proteins complement the mutant phenotype; (iii) unambiguously detecting native or epitope-tagged GLK1 protein even in over-expressing lines is difficult; and (iv) full-length fragments of GLK proteins expressed in heterologous systems tend to degrade (M.T.W. and J.A.L., unpublished data). Therefore, we believe that GLK proteins are inherently unstable. Some transcription factors are rapidly turned over by the proteolytic machinery as a gene regulatory mechanism (Jensen and Winge, 1998;Kornitzer et al., 1994). This may be expected for proteins like GLKs that are regulated at the transcriptional level and are involved in directing the expression of genes related to photosynthesis, itself a tightly controlled process. Analysis of the AtGLK1 sequence indicates the presence of an N-terminal PEST domain, which is associated with proteolytic protein turnover (Rechsteiner and Rogers, 1996). Experiments to confirm regions of GLK proteins that are responsible for degradation are underway. Regardless of the precise cause, unstable regulatory proteins are highly responsive to changing inputs, and thus suitable for adapting photosynthesis to diurnal variations in light quality and quantity. Furthermore, cells across the leaf, such as spongy mesophyll on the underside compared to palisade mesophyll on the upper side, are subject to subtly different light regimes. When spinach leaves are illuminated from the upper surface, less than 10% of blue light reaches the spongy mesophyll, compared to 25% of red light (Vogelmann and Evans, 2002). Non-cell-autonomous, supra-cellular control of photosynthesis across the entire leaf would make such disparities difficult to accommodate. Analysis of publically available datasets suggests that both AtGLK1 and AtGLK2 are up-regulated by red and blue light. Transcription of AtGLK2 increases fivefold within 1 h of exposure of etiolated seedlings to red light, through a pathway that is dependent on phyA and phyB; AtGLK1 is less responsive over the same time scale (Tepperman et al., 2006; GEO dataset accession number GSE3811, available at The response of both genes to blue light shows a similar pattern (AtGenExpress, accessible at Such light-dependent responses provide a mechanism for GLK proteins to regulate transcription on a single-cell basis, and to adjust photosynthesis to suit the particular needs of different cells within the leaf.

Experimental procedures

Plant material and growth conditions

Arabidopsis thaliana ecotype Columbia was used in all experiments. The Atglk1 Atglk2 mutant is available from the European Arabidopsis Stock Centre (NASC ID N9807). Plants were grown on 42 mm peat plugs (Jiffy Products International, in the greenhouse with supplementary lighting providing a 16 h light/8 h dark photoperiod. Seeds were stratified for 3 days at 4°C. Seedlings used for chlorophyll analysis were germinated on nutrient agar containing 1× MS salts and 1× Gamborg’s vitamins (Sigma, For flowering time analysis, plants were grown in a Sanyo versatile environmental test chamber (Sanyo Biomedical,, with day and night temperatures of 22 and 16°C, respectively.

Generation of transgenic Arabidopsis plants

Binary vectors were transferred into Agrobacterium tumefaciens GV3101, and Atglk1 Atglk2 plants were transformed using the floral dip method as described previously (Clough and Bent, 1998). Primary transformants (T1 generation) were selected on the basis of antibiotic resistance, and at least seven lines per construct were taken through to the T2 generation, where lines were screened for transgene segregation. At least two homozygous lines per construct were maintained through to the T3 and T4 generations, which were used for experimental analysis.

Generation of expression constructs

Over-expression of AtGLK1 and AtGLK2  Full-length cDNAs were amplified by RT-PCR, and the coding sequence was subcloned into pGEM-T Easy (Promega, and sequenced. The coding fragment was cloned downstream of the CaMV 35S promoter in pDH51 (Pietrzak et al., 1986). This expression cassette was transferred to binary vector pBIN+ (van Engelen et al., 1995).

Layer-specific expression  The coding sequences of AtGLK1 and AtGLK2 were amplified by PCR using primers containing 3′KpnI restriction sites. These were cloned into the KpnI site of two pBIB-based binary vectors containing the mGFP5 coding sequence downstream of either a 1.6 kb fragment of the FDH promoter or a 3.4 kb fragment of the AtML1 promoter (both gifts from G. Ingram, Institute of Molecular Plant Sciences, University of Edinburgh, UK). The KpnI sites maintained the reading frame, generating FDH::GLK1-GFP and FDH::GLK2-GFP, and AtML1::GLK1-GFP and AtML1::GLK2-GFP. The GLK1–GFP and GLK2–GFP coding sequences were then amplified by PCR and cloned into pGEM-T Easy, yielding pMTW4 and pMTW5, respectively. The subsequent cloning strategies for GLK2–GFP are identical to those for GLK1–GFP. To generate RbcS::GLK1-GFP, an XhoI/SalI fragment from pMTW4 was cloned downstream of the RbcS promoter into the SalI/SacI sites of the binary vector pROK8 (Barnes et al., 1994). To generate AtSUC2::GLK1-GFP, a GLK1–GFP fragment was cloned behind a 949 bp fragment of the AtSUC2 promoter in pEPS1 (a gift from N. Sauer, Department of Molecular Plant Sciences, University of Erlangen-Nuremburg, Erlangen, Germany). Finally, a HindIII/EcoRI fragment comprising AtSUC2::GLK1-GFP was excised and cloned into pBIN+. To generate 35S::GLK1-GFP, an XhoI/HindIII fragment from pMTW4 was cloned downstream of the 35S promoter in pART7 (Gleave, 1992). This expression cassette was then excised using NotI and transferred to pART27 to generate the final binary vector.

GFP–AtGLK1 fusion construct  The full-length AtGLK1 coding sequence was amplified by PCR from pMTW4 using primers containing 5′attB sites to facilitate cloning via site-directed recombination. The PCR product was cloned into pDONR207 (Invitrogen, and sequenced. The resulting entry clone was recombined with the destination vector pMDC43 (Curtis and Grossniklaus, 2003), producing the final binary vector.

DNA and RNA analysis

DNA was isolated using a rapid DNA extraction method as described previously (Fitter et al., 2002). RNA was isolated by guanidinium thiocyanate–phenol–chloroform extraction (Chomczynski and Sacchi, 2006). DNA and RNA gel blots were prepared and hybridized in 0.45 m NaCl at 65°C as described previously (Langdale et al., 1998) using gene-specific probes as follows: AtGLK1 (At2g20570), a fragment from bases 345–569; AtGLK2 (At5g44190), a fragment from bases 67–327; LHCB6 (At1g15820), a fragment from EST 23C1T7 corresponding to bases 39–474; mGFP5, a full-length fragment covering the mGFP5 coding sequence from pMTW4 (base designations refer to cDNA sequences of each locus in brackets as defined by The Arabidopsis Information Resource (TAIR); Northern blots were quantified using a Molecular FX phosphorimager (Bio-Rad,


Total RNA (2 μg) was treated with TURBO DNAse I (Ambion, and subjected to reverse transcription using oligo(dT) and RevertAid reverse transcriptase (Fermentas, in a 20 μl reaction. PCRs contained 1 μl of cDNA, 0.4 μm of each primer, 2 mm MgCl2 and 0.625 units of GoTaq DNA polymerase (Promega). Transcripts of GFP–AtGLK1 were detected by PCR using a forward primer corresponding to GFP (5′-GATCCCAACGAAAAGAGAGAC-3′) and a reverse primer corresponding to AtGLK1 (5′-ATCCTAGGTTAGGTGTCTACTCCATAGATATG-3′). Transcripts of GAPDH (At3g04120) were detected using the forward primer 5′-CACTTGAAGGGTGGTGCCAAG-3′ and the reverse primer 5′-CCTGTTGTCGCCAACGAAGTC-3′.

Chlorophyll analysis

Leaf discs (0.4 cm2) were taken from fully expanded leaves of randomized, greenhouse-grown, 3-week-old plants, transferred to a 1.5 ml tube and frozen in liquid nitrogen. One leaf disc represents one replicate; at least seven replicates from different plants were taken per line. For seedlings, seeds were germinated on MS medium and eight 9-day-old seedlings were pooled per replicate. Frozen tissue samples were ground to a powder in the tube, and 1 ml of 80% v/v acetone was added. The tube was vortexed and incubated in the dark at 4°C overnight to ensure complete chlorophyll extraction. The cell debris was pelleted by centrifugation for 1 minute at 15 000 g, and the absorbance of the supernatant was measured at 646.6, 663.6 and 750 nm as described previously (Porra et al., 1989). Chlorophyll content was expressed as micrograms per square centimetre of leaf. For seedlings, chlorophyll content was expressed as micrograms per seedling.

Confocal microscopy

Leaf segments and whole cotyledons were mounted adaxial (upper) side uppermost in 20% glycerol, and visualized using a Zeiss LSM 510 confocal microscope ( at an excitation wavelength of 488 nm. Chlorophyll autofluorescence was detected between 631 and 729 nm. Images shown are maximum projections of a series of images taken in the z plane.

Transmission electron microscopy

Whole cotyledons and 1 mm-wide hand-cut sections of leaves were vacuum-infiltrated and fixed overnight in 4% paraformaldehyde, 3% gluteraldehyde in 0.05 m potassium phosphate buffer, pH 7. Samples were stained sequentially with 2% w/v OsO4 and 0.5% w/v uranyl acetate before dehydration through an acetone series and embedding in TAAB 812 resin (TAAB Laboratory Equipment, Sections 0.1 μm thick were cut with a diamond blade, stained with 0.2% w/v lead citrate for 2 min and then rinsed in deionized water. Sections were examined using a Zeiss (LEO) Omega 912 electron microscope. Digital images were captured using the SIS package (Soft Imaging Software GmbH,


We thank Gwyneth Ingram (University of Edinburgh) for the gift of FDH and AtML1 promoters and for helpful discussion and advice. We also thank John Gray (University of Cambridge) for the pROK8 plasmid, and Norbert Sauer (University of Erlangen-Nuremburg) for the pEPS1 plasmid. We thank Gaspar Giner and Jayne Davis for invaluable technical assistance, Muris Korkaric for database analysis, and members of the laboratory for stimulating discussion. This work was funded by the Biotechnology and Biological Sciences Research Council.