the plant journal (2002) 31 glk gene pairs regulate...

GLK gene pairs regulate chloroplast development in diverseplant species

David W. Fitter, David J. Martin, Martin J. Copley, Robert W. Scotland and Jane A. Langdale*

Department of Plant Sciences, University of Oxford, South Parks Road, Oxford OX1 3RB, UK

Received 9 April 2002; accepted 17 May 2002.*For correspondence (fax +44 1865 275147; e-mail [email protected]).

Summary

Chloroplast biogenesis is a complex process that requires close co-ordination between two genomes.

Many of the proteins that accumulate in the chloroplast are encoded by the nuclear genome, and the

developmental transition from proplastid to chloroplast is regulated by nuclear genes. Here we show

that a pair of Golden 2-like (GLK) genes regulates chloroplast development in Arabidopsis. The GLK

proteins are members of the GARP superfamily of transcription factors, and phylogenetic analysis

demonstrates that the maize, rice and Arabidopsis GLK gene pairs comprise a distinct group within the

GARP superfamily. Further phylogenetic analysis suggests that the gene pairs arose through separate

duplication events in the monocot and dicot lineages. As in rice, AtGLK1 and AtGLK2 are expressed in

partially overlapping domains in photosynthetic tissue. Insertion mutants demonstrate that this

expression pattern re¯ects a degree of functional redundancy as single mutants display normal

phenotypes in most photosynthetic tissues. However, double mutants are pale green in all

photosynthetic tissues and chloroplasts exhibit a reduction in granal thylakoids. Products of several

genes involved in light harvesting also accumulate at reduced levels in double mutant chloroplasts. GLK

genes therefore regulate chloroplast development in diverse plant species.

Keywords: GARP genes, transcription factors, chloroplasts, gene redundancy.

Introduction

Chloroplast differentiation is an essential developmental

process which takes place in all plant cells that ultimately ®x

carbon dioxide. In undifferentiated embryonic and meriste-

matic cells, plastids are present in a proplastid form. These

small organelles, which can develop into chloroplasts,

etioplasts, amyloplasts, leucoplasts or chromoplasts, con-

tain their own genome and are separated from the

cytoplasm by a double envelope membrane (reviewed by

Mullet, 1988). As cells in the shoot apical meristem divide to

form leaf primordia, proplastids within the cells also divide

and initiate a programme of chloroplast development.

Within the chloroplast, the stroma, the thylakoid membrane

and the lumen contained within the folds of the thylakoid

membrane all differentiate (Mullet, 1988). Further develop-

ment distinguishes different types of chloroplasts. For

example, in plants such as rice and Arabidopsis, that utilize

the C3 photosynthetic pathway, chloroplasts in all cell types

exhibit granal (stacked) thylakoids and accumulate Calvin

cycle enzymes (reviewed by Edwards and Walker, 1983). In

these plants, CO2 is ®xed by ribulose bisphosphate

carboxylase (RuBPCase) in the mesophyll cells, whereas

bundle sheath cells act primarily as supporting and con-

ducting tissue. In contrast, C4 plants such as maize

compartmentalize photosynthetic reactions between

mesophyll and bundle sheath cells. RuBPCase activity is

restricted to bundle sheath cells, whereas mesophyll cells

accumulate a set of cell-speci®c enzymes (phosphoenol-

pyruvate carboxylase, pyruvate phosphate dikinase and

malate dehydrogenase) that act to ®x and then shuttle

carbon to the bundle sheath cells (reviewed by Edwards

and Walker, 1983). Thus three types of chloroplasts develop

in C4 plants (reviewed by Langdale and Nelson, 1991). In

addition to the C3-type, which is present at certain devel-

opmental stages and in certain tissues, distinct chloroplasts

differentiate in the bundle sheath and mesophyll cells of the

C4 leaf. These dimorphic chloroplasts accumulate individ-

ual complements of photosynthetic enzymes and can

exhibit distinct arrangements of thylakoid membranes

The Plant Journal (2002) 31(6), 713±727

ã 2002 Blackwell Science Ltd 713

(reviewed by Edwards and Walker, 1983; see also

Kirchanski, 1975; Kirchanski and Park, 1976).

In all chloroplasts the thylakoid membrane contains the

components required for light harvesting and photopho-

sphorylation (reviewed by Staehelin and Arntzen, 1986).

The membrane is composed of a lipid bilayer that is

interspersed with the proteins, pigments and other

components of the ®ve functional complexes that are vital

for photosynthesis. These complexes are the mobile

chlorophyll a/b light-harvesting complex (LHC); photosys-

tem II (PSII); cytochrome f/b6 (cytf/b6); photosystem I (PSI);

and adenosine triphosphate (ATP) synthase (Staehelin and

DeWit, 1984). The photosystems are portioned in a charac-

teristic fashion between the granal and stromal (unstacked)

thylakoids. PSI is localized preferentially to the stromal

thylakoids, PSII to the granal thylakoids (reviewed by Allen

and Forsberg, 2001). Soluble photosynthetic proteins,

including the carboxylating enzyme RuBPCase, accumulate

within the stroma (Andrews and Lorimer, 1987).

Assembly of the photosynthetic apparatus requires co-

ordination between the chloroplast and nucleus because

complexes are comprised of proteins encoded by both

genomes. As such, a large number of nuclear-encoded,

cytosolically synthesized proteins have to be imported into

the chloroplast. All these proteins are transported by a

single pathway that mediates passage through the double

envelope membrane (reviewed by Keegstra and Froehlich,

1999). Once inside the chloroplast, however, proteins are

targeted to either the thylakoid membrane or the thylakoid

lumen by one of four pathways (reviewed by Robinson

et al., 1998). The SecA and DpH pathways transport

lumenal proteins, whereas integral membrane proteins

are imported by either the signal recognition pathway

(SRP) or the spontaneous pathway (Robinson et al., 1998).

Although many of the structural components of the

import pathways and photosynthetic complexes are

known (Allen and Forsberg, 2001; Robinson et al., 1998),

very few genes that regulate chloroplast development

have been characterized. One example is the maize

Golden2 (G2) gene which acts as a general regulator of

chloroplast development in C3-type photosynthetic cells

and as a speci®c regulator of bundle sheath cell chlor-

oplast development in C4 photosynthetic tissue (Hall et al.,

1998; Langdale and Kidner, 1994; Rossini et al., 2001).

Chloroplasts in g2 mutants are smaller than wild type and

develop only rudimentary thylakoid lamellae that exhibit

very few grana (Langdale and Kidner, 1994). A second G2-

like (Glk) gene, ZmGlk1, is thought to regulate mesophyll

cell chloroplast development in C4 tissues (Rossini et al.,

2001). Unlike G2, however, ZmGLK1 function has been

postulated only on the basis of expression patterns and

not on loss-of-function phenotypes.

The GLK proteins are members of the recently categor-

ized GARP superfamily of transcription factors (Riechmann

et al., 2000) de®ned by G2 in maize; the Arabidopsis

RESPONSE REGULATOR-B (ARR-B) proteins (Imamura

et al., 1999); and the PHOSPHATE STARVATION

RESPONSE1 (PSR1) protein of Chlamydomonas (Wykoff

et al., 1999). In the case of G2, three of the four de®ning

features of most transcription factors have been veri®ed

experimentally in heterologous systems. G2 is nuclear-

localized (Hall et al., 1998), is able to transactivate reporter

gene expression, and can both homo-dimerize and hetero-

dimerize with ZmGLK1 (Rossini et al., 2001). DNA-binding

activity of GLK proteins has yet to be demonstrated,

however, the putative DNA-binding domain is highly

conserved with domains in other GARP proteins such as

ARR1 and ARR2 (Riechmann et al., 2000). Notably, ARR1

and ARR2 have been shown to bind DNA (Sakai et al.,

2000), thus it is likely that GLK proteins act as transcrip-

tional regulators of chloroplast development.

The spatial compartmentalization of G2 and ZmGlk1

transcripts in C4 maize tissue may represent a specializa-

tion required for the development of distinct bundle

sheath and mesophyll chloroplasts (Rossini et al., 2001).

This view is supported by the observation that the rice

genes OsGlk1 and OsGlk2, which are orthologous to

ZmGlk1 and G2, respectively, are expressed in overlapping

domains in photosynthetic tissues (Rossini et al., 2001).

Thus the two rice genes may act redundantly.

Unfortunately, this suggestion has not been con®rmed

because loss-of-function mutations in OsGlk genes are not

yet available. We have therefore examined GLK gene

function in the C3 dicot Arabidopsis with a view to testing

the hypothesis that GLK gene pairs act redundantly to

promote chloroplast development in C3 species. In this

report we demonstrate that two GLK genes are present in

the Arabidopsis genome, and that these genes are struc-

turally similar to the monocot Glk genes. A phylogenetic

analysis con®rms that the GARP gene superfamily is

monophyletic in the context of putative relatives, and

that the GLK gene pairs from Arabidopsis, maize and rice

comprise a monophyletic group within the superfamily.

Through characterization of insertional mutants, we dem-

onstrate that the GLK gene products function in a redun-

dant manner as transcriptional regulators of C3-type

chloroplast development. We propose that this redun-

dancy enabled one GLK gene to play a derived role in the

evolution of distinct bundle sheath and mesophyll chlor-

oplasts in C4 plants.

Results

GLK genes in Arabidopsis

Two Arabidopsis GLK genes were identi®ed when the

ZmG2 sequence was used as a query in BLAST searches of

the Arabidopsis genome database: http://www.arabidopsi-

714 David W. Fitter et al.

ã Blackwell Science Ltd, The Plant Journal, (2002), 31, 713±727

s.org/BLAST/. The AtGLK1 gene encodes a predicted

protein of 420 amino acids and the AtGLK2 gene encodes

a protein of 386 amino acids. Figure 1 illustrates the

sequence similarity between the Arabidopsis GLK proteins

and the GLK proteins of maize and rice. Sequence conser-

vation is particularly high across the putative DNA-binding

domain and across the C-terminal domain, referred to as

the GCT box, which is required to mediate homo- and

hetero-dimerization of the maize proteins (Rossini et al.,

2001). The DNA-binding domain has previously been

postulated to fold as a helix±loop±helix (Rossini et al.,

2001). However, using all six protein sequences, the

consensus output from secondary structure prediction

software is a domain that folds as three a-helices (Figure

1). Immediately downstream of the third helix is an

AREAEAA motif that is conserved exactly in ®ve of the

GLK proteins, and is present as AREVEAA in ZmGLK1.

Both Arabidopsis GLK sequences contain recognizable

nuclear localization signals (NLS). PSORT software predic-

tions (Nakai and Kanehisa, 1992) indicate that AtGLK1 has

an NLS of the Simian Virus 40 (SV40) large T antigen type

(Kalderon et al., 1984) just upstream of the DNA-binding

domain, at amino acid position 130±133. A similar type of

NLS is found in OsGLK2. AtGLK2 is also predicted to have

Figure 1. Alignment of GLK protein sequences from Arabidopsis, maize and rice.Amino acid identity is indicated by black boxes. Solid horizontal bars indicate the predicted a-helices conserved in all six proteins. The GCT-box isdelineated by the last exon. Inverted triangles mark intron positions: white triangle, conserved in all six genes; black, conserved in GLK1 genes only; grey,conserved in OsGlk2 and ZmG2; chequered, conserved in ZmGlk1 and AtGLK1; hatched, OsGlk1 only; vertical stripes, OsGlk2 only; horizontal stripes,AtGLK2 only.

Arabidopsis GLK gene function 715


an SV40 large T antigen-type NLS, downstream of the

DNA-binding domain at position 223±226. In addition,

however, like ZmG2 and ZmGLK1, AtGLK2 has an NLS

upstream of the DNA-binding domain that is of the

Xenopus nucleoplasmin type (Robbins et al., 1991).

GLK gene phylogeny

The GLK DNA-binding domain is characteristic of the

newly de®ned GARP superfamily of transcription factors

(Riechmann et al., 2000). GARP genes have thus far been

identi®ed only in plant genomes, however, the GARP DNA-

binding domain sequence is similar to TEA-binding

domain sequences (BuÈ rglin, 1991) and to the single MYB-

like domain in MYB-related proteins (Baranowskij et al.,

1994). Both TEA- and MYB-related sequences are found in

a range of eukaryotic genomes. To determine whether

GARP gene sequences are phylogenetically distinct from

TEA- and MYB-related sequences, phylogenetic analysis

was carried out with 38 Arabidopsis GARP gene

sequences; Psr1 from Chlamydomonas (Wykoff et al.

(1999); maize and rice Glk gene sequences (Rossini et al.,

2001); six TEA genes; and the MYB-related potato gene

MybSt1. Figure 2(a) demonstrates that there is strong

bootstrap support (86%) for the monophyly of the GARP

superfamily relative to TEA genes and MybSt1. Thus GARP

genes are phylogenetically distinct from both TEA- and

MYB-related genes. Previous reports placing PSR1 and

related Arabidopsis genes in the MYB family are therefore

inaccurate (Rubio et al., 2001), as PSR1 is clearly nested

within the GARP superfamily.

The GARP superfamily is reported to comprise 56 genes

in Arabidopsis (Riechmann et al., 2000). Prior to analysis of

AtGLK gene function, it was therefore important to deter-

mine how many of these GARP genes could be classi®ed

as GLK genes. Using the presence of both the DNA-

binding domain and the GCT box as the de®ning criterion

of GLK gene structure, AtGLK1, AtGLK2 and the

Arabidopsis PSEUDO RESPONSE REGULATOR 2 (APRR2)

gene (Makino et al., 2000) were identi®ed as GLK genes.

APRR2 encodes a protein that comprises a GARP DNA

binding domain, a domain similar to the GCT box, and a

receiver-like domain at the N-terminus which is similar to

that encoded by the ARR-B genes (Imamura et al., 1999;

Makino et al., 2000). Phylogenetic analysis of nucleotide

sequence encoding only the DNA-binding domain dem-

onstrated reasonable bootstrap support for the existence

of a distinct GLK gene group nested within the GARP

superfamily (Figure 2a). Notably, however, AtGLK1 and

AtGLK2 belong to this group but APRR2 does not. APRR2

also lacks the AREAEAA motif. Thus only two GLK genes

are present in the Arabidopsis genome.

Relationship between monocot and dicot GLK genes

Based on synteny analysis, gene sequence, intron/exon

structure and expression patterns, relationships between

Figure 2. GLK gene relationships.(a) Parsimony analysis of the GARP gene superfamily based onnucleotide sequence encoding the putative DNA-binding domain. Theoutgroup consists of six TEA genes and potato MybSt1. Strong bootstrapsupport is shown for monophyly of the GARP superfamily (86) and for aGLK gene subgroup within the GARP superfamily (65).(b) The single most parsimonious tree from a phylogenetic analysis,using an exhaustive search strategy of six GLK genes and APRR2 whichis included to root the tree. Bootstrap percentages are estimated from1000 bootstrap replicates.



the maize and rice Glk genes have been determined

(Rossini et al., 2001). ZmGlk1 is orthologous to OsGlk1

and G2 is orthologous to ZmG2. Thus the gene duplication

occurred prior to the divergence of rice and maize. To

assess the relationship between Arabidopis GLK genes

and the monocot genes, phylogenetic analysis was carried

out using the complete cDNA sequence of each GLK gene

and APRR2 cDNA sequence as an outgroup. Figure 2(b)

demonstrates that the Arabidopsis GLK genes are more

closely related to each other than they are to any of the

monocot genes. Therefore the simplest hypothesis is that

independent duplication events occurred in the monocot

and dicot lineages. However, gene structure data are not

entirely congruent with this suggestion. For example,

AtGLK2 has ®ve exons, like ZmG2, whereas AtGLK1 has

six exons, like ZmGlk1 and OsGlk1. OsGlk2 has six exons,

but the position of the extra intron is not the same as in the

GLK1 genes (Figure 1). To resolve the discrepancy

between the phylogenetic and gene structure data, further

GLK gene sequences need to be analysed.

AtGLK gene transcript accumulation pro®les

Consistent with a proposed role in chloroplast develop-

ment, maize and rice Glk gene transcripts accumulate

exclusively in photosynthetic tissue (Hall et al., 1998;

Rossini et al., 2001). To determine the spatial expression

pro®les of the Arabidopsis GLK genes, RNA gel-blot

analysis was carried out with RNA extracted from different

tissues. Figure 3 demonstrates that both AtGLK1 (Figure

3a) and AtGLK2 (Figure 3b) transcripts are predominantly

detected in rosette and cauline leaves. In cotyledons,

shoots and ¯owers, AtGLK1 transcript levels are lower

than AtGLK2 levels. Only AtGLK2 transcripts accumulate in

roots and siliques. Thus in the most photosynthetically

Figure 3. AtGLK1 and AtGLK2 transcriptaccumulation patterns.(a,b) Gel-blot analysis of RNA extractedfrom different Arabidopsis tissues. Blotswere hybridized to gene-speci®c fragmentsof AtGLK1 (a) and AtGLK2 (b).The AtGLK1transcript is 1.6 kb and the AtGLK2transcript is 1.5 kb. RNA loading levels weredetermined by quantifying ¯uorescencelevels from the ethidium bromide-stainedgel. Relative intensity of hybridization wasdetermined by densitometric analysis ofautoradiographs and normalization withRNA loading levels.(c,d) Gel-blot analysis of RNA extracted fromrosette leaves of 4-week-old plants harvestedevery 2 h over a 28 h period encompassingan 8 h dark period and a 16 h light period.Blots were hybridized to gene-speci®cfragments of AtGLK1 (c), AtGLK2 (d) and toTUBULIN (c,d). RNA loading levels weredetermined by densitometric analysis ofTUBULIN autoradiographs. Relativehybridization levels of AtGLK1 and AtGLK2were calculated as in (a,b).(e,f) Gel-blot analysis of RNA extracted fromrosette leaves of plants that had beenentrained to a 16 h light/8 h dark cycle for4 weeks. Leaves were harvested every 2 hover a 28 h dark period. Blots werehybridized to gene-speci®c fragments ofAtGLK1 (e), AtGLK2 (f) and to TUBULIN (e,f).RNA loading levels and relative hybridizationlevels were calculated as for (c,d).



Figure 4. Atglk1.1 and Atglk2.1 mutantphenotypes.(a,b) Schematic diagram showing the dSpminsertion site in Atglk1.1 (a)and Atglk2.1 (b)alleles. Horizontal lines representuntranslated regions. Boxes representdifferent domains of the coding region.Black triangles designate intron positionsconserved in both genes; red trianglerepresents position of extra intron inAtGLK1.(c,d) RNA gel-blot analysis of AtGLK1 (c)andAtGLK2 (d)transcript accumulation in wild-type (WT), Atglk1.1, Atglk2.1 and doublemutant plants. The ethidium bromide-stained gel was used to assess RNA loadinglevels and relative hybridization intensitywas determined as for Figure 3(a,b).(e) Wild-type and mutant phenotypes4 weeks after germination.(f) Siliques of wild-type and mutant plants.



active tissues, both AtGLK1 and AtGLK2 transcripts accu-

mulate. In other tissues, AtGLK2 transcripts predominate.

To assess whether light in¯uences AtGLK transcript

accumulation patterns, RNA was extracted at 2 h intervals

through a 28 h diurnal light/dark cycle from rosette leaves

of 4-week-old plants. Figure 3(c) demonstrates that

AtGLK1 transcript levels are noticeably higher in samples

harvested during the 16 h light period than in samples

harvested during the dark period. Light also appeared to

in¯uence AtGLK2 transcript accumulation levels (Figure

3d), however, transcript levels increased in anticipation of

the light period. This observation suggested that AtGLK2

transcript accumulation pro®les are regulated by an

endogenous (possibly circadian) mechanism. To investi-

gate this suggestion further, plants were entrained to a

light/dark cycle for 4 weeks, then kept in the dark for 28 h.

As before, samples were harvested every 2 h through the

28 h period and RNA was extracted and analysed.

Figure 3(e) demonstrates that AtGLK1 transcript levels

remained low throughout the 28 h period, con®rming the

observation that accumulation levels are light-regulated. In

contrast, however, the AtGLK2 transcript accumulation

pro®le (Figure 3f) was similar to that seen previously

(Figure 3d), although absolute levels of transcript were

lower during the ®rst half of the light period. This suggests

that AtGLK2 transcript accumulation patterns are regulated

both by endogenous mechanisms and by light.

Despite the fact that AtGLK1 transcript levels are notice-

ably higher in the light period of a diurnal cycle, only very

low levels were detected during the conversion from

etioplast to chloroplast (data not shown). In contrast,

AtGLK2 transcript levels increased dramatically after dark-

grown plants were exposed to light (data not shown). Thus

it is likely that AtGLK2 functions to facilitate the etioplast-

to-chloroplast conversion and that AtGLK1 plays only a

limited role.

AtGLK1 and AtGLK2 exhibit functional redundancy

To elucidate GLK gene function, insertional mutants were

identi®ed in the Sainsbury Laboratory Arabidopsis

Transposant collection (Tissier et al., 1999). The Atglk1.1

mutant contains a dSpm insertion 112 bp upstream of the

AtGLK1 translation start site (Figure 4a) and the Atglk2.1

mutant contains a dSpm in exon I of AtGLK2, 333 bp

downstream of the translation start site (Figure 4b). To

determine whether the dSpm insertions altered the

expression patterns of the respective genes, RNA gel-blot

analysis was carried out. Figure 4(c) demonstrates that

AtGLK1 transcripts are barely detectable in Atglk1.1

mutants but accumulate to normal levels in Atglk2.1

mutants. Similarly, AtGLK2 transcripts do not accumulate

in Atglk2.1 mutants but are present at normal levels

in Atglk1.1 mutants (Figure 4d). One exception to this

observation is the detection of a truncated AtGLK2 tran-

script in young tissue of Atglk2.1 mutants (data not

shown). This truncated transcript lacks the 3¢ end and is

therefore assumed to be non-functional.

Despite the fact that GLK gene transcripts do not

accumulate, the single mutants are essentially indistin-

guishable from wild type throughout most of development

(Figure 4e). This suggests that the AtGLK genes are

functionally redundant. The exception to this redundancy

is observed in silique tissue. Atglk2.1 mutant plants exhibit

pale green siliques (Figure 4f). Notably, AtGLK1 transcripts

do not accumulate in siliques of wild-type plants (Figure

3a), suggesting that the absence of both gene products

leads to a pale green phenotype. To con®rm this sugges-

tion, Atglk1.1,Atglk2.1 double-mutant plants were gener-

ated. Double mutants accumulated low levels of AtGLK1

transcripts (Figure 4c) and barely detectable levels of

AtGLK2 transcripts (Figure 4d). The presence of AtGLK

transcripts, albeit at low levels, suggests that there may be

some residual GLK function in the double mutants.

However, plants were pale green throughout development

(Figure 4e,f). Although plants were smaller than wild type,

no other phenotypic perturbations were observed. Thus

AtGLK genes are likely to function speci®cally to promote

chloroplast development.

AtGLK gene function is dosage dependent.

Mutant phenotypes suggested that AtGLK1 and AtGLK2

act in a redundant manner. To determine the effect of gene

copy number on GLK function, chlorophyll accumulation

levels were analysed in segregating F1 individuals from a

selfed double heterozygous plant. Two wild-type and four

double mutant plants were preselected, then 18 plants

were selected randomly from the population. Figure 5(a)

shows that chlorophyll levels were markedly reduced in

double mutants as compared to the other 20 plants. In all

cases, chlorophyll levels were higher in rosette leaves than

cauline leaves.

After chlorophyll levels had been determined, individual

plants were genotyped to assess the number of functional

GLK gene copies in each. Figure 5(b) shows that wild-type

and mutant AtGLK1 alleles were distinguished with a

HindIII digest. Wild-type alleles yielded a 6.5 kb fragment

and mutant alleles a 4.3 kb fragment. Similarly, AtGLK2

alleles were distinguished with an EcoRV digest (AtGLK2,

6.5 kb; Atglk2.1, 6.8 kb). Of the 18 plants, ®ve had one,

eight had two and ®ve had three GLK gene copies.

Having determined gene copy number, ANOVA was

carried out to determine whether copy number in¯uenced

chlorophyll levels. Figure 5(c) shows that both AtGLK1 and

AtGLK2 copy numbers affect chlorophyll levels. Notably,

plants with only one AtGLK copy (plants 1 and 13) are

phenotypically closer to wild type in appearance than they



are to the double mutant. Thus a threshold level of GLK

protein (equivalent to or less than one wild-type gene

copy) is required for plants to appear normal (green). As

the interaction term (GLK1*GLK2) is signi®cant, the data

further suggest that loss of both genes has a greater effect

than would be expected were the effects of the Atglk1.1

and Atglk2.1 mutant alleles simply additive (Figure 5c).

Therefore, in addition to functioning on their own, the two

gene products may interact. This suggestion is supported

by the observation that AtGLK1 and AtGLK2 both homo-

and hetero-dimerize in the yeast two-hybrid system (data

not shown).

Chloroplast development is perturbed in Atglk double

mutants

Loss of Glk gene function in maize leads to perturbed

chloroplast development ± plastids are small and thyla-

koids are rudimentary (Langdale and Kidner, 1994). To

Figure 5. Chlorophyll concentration in plants segregating Atglk1.1 andAtglk2.1 alleles.(a) Mean chlorophyll levels in cauline (white) and rosette (black) leaves.Two wild-type (WT) and four double mutant (D) plants were preselected,but the remaining 18 plants were chosen randomly.(b) DNA gel-blot analysis of HindIII (AtGLK1)- and EcoRV (AtGLK2)-digested DNA extracted from individual plants assayed in (a). Blots werehybridized with AtGLK1 and AtGLK2 gene-speci®c fragments.(c) Output from ANOVA. The error term 31.53 is a measure of within-subject variation (the variation within plants). The error term 42.97 is ameasure of the between-subject variation (variation between plants ofthe same genotype), and is the result of nesting variables `Plant number'with `Copies of GLK gene' which is speci®ed as a random variable.Importantly, the between subject error is not signi®cant (1.36) andtherefore the variation between plants of the same genotype is notsigni®cantly greater than that between leaves within a plant. Thus plantswith the same genotype can be treated as identical. The between-plantserror term is used as the denominator to calculate the F values for themain effect variable `Copies GLK1', `Copies GLK2', and the interactionterm `Copies GLK1*GLK2'. The within-plant error is used as thedenominator to calculate the F value for the `Leaf Type' factor.df = degrees of freedom; SeqSS = sequential sums of squares;AdjSS = adjusted sums of squares; AdjMS = adjusted mean squares.

Figure 6. Chloroplast ultrastructure in wild-type and Atglk mutant plants.TEM of bundle sheath (a,c,e,g) and mesophyll (b,d,f,h) chloroplasts inwild-type (a,b); Atglk1.1 (c,d); Atglk2.1 (e,f); and double mutant (g,h)plants. Solid arrows point to granal lamellae; open arrows to stromallamellae. Scale bar = 1 mM.



determine whether the AtGLK genes in¯uence chloroplast

ultrastructure, transmission electron microscopy was car-

ried out on wild-type and mutant rosette leaf tissue.

Figure 6 shows that in both mesophyll and bundle sheath

cells of double mutants, chloroplasts are »50% smaller in

cross-sectional area than the wild type. As chloroplasts

divide by binary ®ssion of existing organelles (reviewed by

Osteryoung, 2000), a negative correlation exists between

chloroplast number and chloroplast size. This is clear from

chloroplast division mutants, such as accumulation and

replication of chloroplasts 6 (arc6), where only two

chloroplasts exist per cell, but chloroplasts are greatly

enlarged as compared to the wild type (Pyke et al., 1994;

Robertson et al., 1995). The arc mutants suggest that

plastid division is distinct from plastid differentiation, as

proliferation is affected, but the chloroplasts appear to

develop normally (Pyke and Leech, 1994; Pyke et al., 1994).

Thus the ARC genes must speci®cally regulate chloroplast

division. In contrast, the GLK genes appear to effect plastid

development and differentiation but not plastid division,

as chloroplast numbers per cell are normal in double

mutants (data not shown).

In addition to a reduction in size, chloroplasts in double

mutants exhibit rudimentary thylakoid lamellae (Figure

6g,h). The absence of grana is striking in that chloroplasts

contain almost entirely stromal lamellae. Notably, this

chloroplast phenotype is very similar to that seen in

chloroplasts of the g2 mutant of maize. This suggests that

the GLK genes in maize and Arabidopsis have similar

functions.

To further characterize the defects in chloroplast devel-

opment that are conditioned by loss of AtGLK gene

function, components of the photosynthetic machinery

were examined by RNA and protein gel-blot analysis. To

assess whether the accumulation of Calvin cycle enzymes

was perturbed in mutant plants, RuBPCase accumulation

patterns were examined. Both the large (LSU) and small

(SSU) subunit of RuBPCase accumulated to normal levels

in both single and double mutant plants (Figure 7a). Thus it

is likely that mutant plants have the capacity to ®x carbon.

The most obvious feature of chloroplast ultrastructure in

double mutant plants is a lack of granal thylakoids. To test

whether Atglk double mutants are fundamentally impaired

in thylakoid membrane formation, levels of the VESICLE-

INDUCING PROTEIN IN PLASTIDS 1 (VIPP1) protein were

compared in wild-type and mutant plants. VIPP1 is a

hydrophilic, integral thylakoid membrane protein that has

been shown to be essential for vesicle formation (Kroll

et al., 2001). Figure 7(a) demonstrates that VIPP1 levels are

normal in double mutants and thus suggests that mem-

brane assembly per se is unaffected by loss of GLK

function.

The reduced levels of chlorophyll seen in double mutant

plants suggested that components of the light-harvesting

Figure 7. Accumulation patterns of photosynthetic gene products in wild-type and Atglk mutant plants.(a) Gel-blot analysis of LHCPII, VIPP1 and RuBPCase LSU and SSUproteins in wild-type and mutant plants. Equal volumes of proteinextracts were compared.(b) Gel-blot analysis of LHCP1, LHCP6, cpSRP43 and cpSRP54 transcriptsin wild-type and mutant plants. Ethidium bromide-stained gels indicateRNA loading levels in each case.(c) Gel-blot analysis of AtCAO and CHLH transcripts in wild-type andmutant plants. Ethidium bromide-stained gels indicate RNA loadinglevels in each case.(d) Gel-blot analysis of representative chloroplast-encoded proteins fromPSI (PsaD); PSII (D1 polypeptide); ATP synthase (CF1a); cytochrome f/b6(Cytf); and NAD(P)H dehydrogenase (NDHH) complexes. Equal volumesof protein extracts were compared.



machinery may be perturbed. The LHCP gene family

encodes the chlorophyll a/b-binding proteins that hold

chlorophyll molecules within the thylakoid membranes.

Figure 7(a) demonstrates that LHCPII protein levels are

reduced in double mutant plants but not in single mutant

plants, in accord with macroscopic observations (Figure

4e). LHCP1 and LHCP6 gene transcript levels are also

greatly reduced in the double mutant (Figure 7b).

The LHCP proteins are inserted into the thylakoid

membrane through the SRP pathway (Schuenemann

et al., 1998) which requires two stromal factors, cpSRP43

and cpSRP54. cpSRP43 dimerizes with cpSRP54 to bind

LHCP (Groves et al., 2001; Jonas-Straube et al., 2001), but

then transport across the membrane is facilitated by

cpSRP43 alone (Klimyuk et al., 1999). SRP43 transcript

levels are reduced in the double mutant, whereas SRP54

transcript levels are unaffected (Figure 7b). However,

SRP43 protein levels are normal in mutant plants (data

not shown), suggesting that the reduced levels of SRP43

transcripts are a consequence of reduced LHCP, rather

than a cause. Other components of the photosynthetic

apparatus that are transported across the thylakoid mem-

brane to the lumen utilize the SecA or DpH pathways

(Mould and Robinson, 1991; Nielsen et al., 1994; Yuan

et al., 1994). Preliminary analysis suggests that other

thylakoid-import pathways are unaffected by loss of GLK

function (data not shown).

Although the reduced chlorophyll levels observed in

double mutant plants may result simply from a reduced

ability to anchor chlorophyll in the membrane, there may

also be a defect in chlorophyll biosynthesis. Levels of

transcripts encoding both subunit H of magnesium chela-

tase (CHLH; an enzyme that functions early in chlorophyll

biosynthesis) and chlorophyll a/b oxygenase (CAO; a key

enzyme in the latter stages of chlorophyll biosynthesis) are

reduced in double mutant plants (Figure 7c).

To assess whether other complexes that accumulate in

the thylakoid membrane are perturbed in double mutants,

representative proteins from PSI (PsaD), PSII (D1 polypep-

tide), ATP synthase (CF1a), cytf/b6 (Cytf) and the NAD(P)H

dehydrogenase (NDHH) complexes were examined by

protein gel-blot analysis. Previous studies have demon-

strated that the stability of ATP synthase and cytochrome f/

b6 complexes is dependent on the accumulation of all the

subunits within a complex (Barkan et al., 1995; Rochaix,

1992). Thus the level of any individual subunit re¯ects the

integrity of the complex as a whole. Consistent with the

observation that LHCPII levels are lower in double mutants

(Figure 7a), Figure 7(d) shows that D1 polypeptide levels are

also lower than in wild-type and single mutant plants. Thus

PSII assembly is impaired. In contrast, all other proteins

assayed are present at normal levels, suggesting that most

of the complexes are unaffected by loss of GLK function.

Discussion

Chloroplasts are thought to have derived from a cyano-

bacterium ancestor by endosymbiosis (Margulis, 1975).

During evolution, most of the prokaryotic genome of the

ancestral cyanobacterium has been lost or transferred to

the eukaryotic nuclear genome of the host cell (Douglas,

1994). Depending on the plant species, chloroplast gen-

omes now contain between 87 and 183 known genes,

around half of which encode components of the chlor-

oplast translational machinery (Sugiura et al., 1998). The

remaining chloroplast genes encode components of the

ATP-synthesizing complex, the cytochrome b/f complex,

PSI, PSII, various subunits of the respiratory enzyme

NAD(P)H dehydrogenase, and also the LSU of RuBPCase

(Lawlor, 1993). The nucleus therefore encodes many

chloroplast-localized photosynthetic proteins including

®ve subunits of PSI, four subunits of PSII, the SSU of

RuBPCase, and proteins involved in import mechanisms

such as the SecA pathway. Perhaps more importantly, the

nucleus encodes genes that regulate the developmental

programme of chloroplast biogenesis. Of these regulatory

genes, the GLK genes are now among the best character-

ized. There are no GLK gene sequences in the genome of

the cyanobacterium Synechocystis; however, all chloro-

plast-containing organisms examined, from algae to

higher plants, appear to contain GLK sequences in their

genomes. These observations suggest ®rst, that the GLK

genes did not originate in the chloroplast genome, and

second, that GLK gene function is associated speci®cally

with chloroplast development rather than with the ability

to photosynthesize.

On the basis of AtGLK gene transcript accumulation

patterns (Figure 3) and mutant phenotypes (Figure 4), it is

proposed that AtGLK2 functions in photosynthetic tissues

throughout development. The spatial pattern of transcript

accumulation (Figure 3a,b) suggests a role in chloroplast

maintenance, whereas the pro®le during the etioplast to

chloroplast transition (data not shown) suggests a pivotal

role during early chloroplast biogenesis. During vegetative

growth, AtGLK2 transcript pro®les are regulated by

endogenous mechanisms, but accumulation levels are

enhanced by light. In contrast, AtGLK1 accumulation

pro®les are regulated by light ± levels are highest during

the light period of the diurnal cycle. It is therefore

proposed that AtGLK1 acts to boost photosynthetic cap-

ability in tissues with a high photosynthetic workload. This

suggestion is supported by the accumulation of AtGLK1

transcripts exclusively in leaf tissue, and predominantly in

rosette and cauline leaves where the majority of photo-

synthesis takes place.

The defects in Atglk double mutant chloroplasts are

limited to a reduction in granal thylakoids and lower

chlorophyll, LHCPII and D1 polypeptide levels. As VIPP1



levels are normal in mutant plants (Figure 7), we propose

that membrane biosynthesis per se is unaffected by loss of

GLK gene function. It is therefore proposed that thylakoids

are perturbed due to loss of PSII complexes. LHCP gene

transcript levels are reduced in mutant plants (Figure 7),

whereas psbA (which encodes D1) levels are normal (data

not shown). Thus it is likely that the primary defect is loss

of LHCP or chlorophyll, and that D1 protein is degraded

due to defective PSII complex assembly. As GLK proteins

have a proposed DNA-binding domain that is similar in

structure to ARR-B proteins, it is possible that they bind to

a consensus DNA sequence that is similar to the one

recognized by ARR-B proteins. The expected frequency of

any particular 5 bp sequence is one in 1024 bp.

Interestingly, in a region of DNA 1 kb upstream of the

LHCP1 and LHCP6 promoter regions there are three and six

copies of the ARR-B binding site (AGATT), respectively. In

the region upstream of the chlorophyll biosynthetic genes

CAO and CHLH, there are one and two sites, respectively.

Data obtained in Arabidopsis and rice have demon-

strated that in C3 plants the GLK genes play a partially

redundant role in regulating transcription of genes

required for chloroplast development. Redundancy is

widespread in the genomes of higher organisms

(Thomas, 1993), and is thought to arise from gene dupli-

cation events as a result of either tandem duplication of

speci®c chromosomal sections, or whole-genome tetra-

ploidization. Three types of genetic redundancy have been

de®ned, and each is subject to different selection pres-

sures (Nowak et al., 1997). The ®rst is true or complete

redundancy, where the relative ®tness of an organism with

only one functional gene is identical to an organism where

both genes are functional. Only individuals without either

gene suffer a reduction in ®tness. In the second example,

redundancy is not complete in that two genes perform the

same function, but at slightly different ef®ciencies. In such

a case, the less ef®cient gene is fully redundant in that

there is no perturbed phenotype in null mutants, but the

other gene is partially redundant, with mutants exhibiting

a partial loss of ®tness. An example of this is observed

with the Arabidopsis TRYPTOPHAN SYNTHASE B genes

(TSB1 and TSB2) (Last et al., 1991). Both genes are

transcribed but TSB1 produces >90% of the tryptophan

synthase B protein. Despite this observation, under certain

conditions TSB2 is capable of supporting growth in the

absence of TSB1 (Last et al., 1991). The third type of

redundancy is where genes are redundant for one func-

tion, but one or both may be selected for because of

another independent function. The Arabidopsis meristem

identity genes APETALA1 (AP1; Mandel et al., 1992) and

CAULIFLOWER (CAL; Kempin et al., 1995) exhibit redun-

dancy of this type. Both encode MADS-box transcription

factors. The ap1 mutants produce ¯owers that retain

partial in¯orescence meristem identity, whereas cal

mutants do not exhibit a mutant phenotype. However,

ap1 cal double mutants produce presumptive ¯oral

meristems that are completely converted into in¯ores-

cence meristems (Bowman et al., 1993). Hence, CAL activ-

ity appears to be entirely redundant with AP1 activity, but

AP1 has some activities that are not redundant with CAL

(Pickett and Meeks-Wagner, 1995).

The AtGLK genes exhibit partial redundancy as there is a

phenotype speci®c to the Atglk2.1 mutant allele (pale

green siliques) but no phenotype speci®c to the Atglk1.1

allele. This observation may simply re¯ect the fact that the

two genes have different spatial and temporal expression

domains (AtGLK1 is not expressed in siliques), rather than

different functions. Alternatively, the two genes may be

functionally distinct, in which case AtGLK1 is redundant

but AtGLK2 is not. This may be due to weaker ef®ciency of

AtGLK1, as in the case of TSB genes, or may re¯ect the fact

that, like the AP1/CAL interaction, AtGLK2 has some

functions that are independent of AtGLK1. Notably, at

certain stages of development AtGLK1 transcript levels are

higher in Atglk2.1 mutants than in corresponding wild-

type tissues (data not shown). Therefore one of the

additional functions of AtGLK2 may be to regulate

AtGLK1 accumulation levels. A similar regulatory mech-

anism is seen in Antirrhinum with products of the

DEFICIENS (DEF) and GLOBOSA (GLO) genes. In this

case, GLO transcription is independent of a functional DEF

protein during early development; however, after initiation

of organ primordia, GLO and DEF transcription is regu-

lated by a DEF/GLO interaction (Schwarz-Sommer et al.,

1992; Trobner et al., 1992). Perhaps signi®cantly, DEF and

GLO bind DNA as heterodimers (Zachgo et al., 1995).

The functional redundancy (albeit partial) of GLK gene

action in Arabidopsis is a striking contrast to the apparent

specialization of gene function in maize. Compartmentali-

zation of Glk gene products between bundle sheath and

mesophyll cells of maize may re¯ect the fact that ZmG2

and ZmGlk1 have evolved speci®c functions. Alternatively,

the two proteins may be functionally interchangeable but

restricted to spatial domains of action. Either way, it is

possible that compartmentalization of Glk gene products

coincided with the transition to C4 photosynthesis.

However, as ZmG2 promotes the development of C3M

and C4BS type chloroplasts, which are both structurally

and functionally distinct, it is likely that other cell-speci®c

factors were also involved.

Experimental procedures

Sequence analysis

BLAST searches were performed using appropriate algorithmswith and without low-complexity ®lters on a range of nucleotideand protein databases at http://www.ncbi.nlm.nih.gov/BLAST/and



http://www.arabidopsis.org/BLAST/. Secondary structure predic-tion was performed using three software programs; SSPRO

(promoter.ics.uci.edu/BRNN-PRED/ Baldi et al., 1999), PSIPRED

(insulin.brunel.ac.uk/psipred/ Jones, 1999) and PROF PREDICTION

(http://www.aber.ac.uk/~phiwww/Prof/ Ouali and King, 2000). PHDprotein-folding predictions were made through the PredictProtein e-mail server (maple.bioc.columbia.edu/predictprotein/).NLS prediction was performed using PSORT software (psort.nib-b.ac.jp). Simple pairwise alignments were performed usingGENEJOCKEY software (Biosoft, Cambridge, UK), and theGenetics Computer Group (GCG, Madison, WI) PILEUP packagewas used for multiple sequence alignments. The GCG PRETTYBOX

function was used for highlighting conserved residues and theCANVAS (Deneba Systems, Miami, FL, USA) graphics programmewas used to manipulate and present the alignment.

Determination of gene structure

ESTs were ordered from the Arabidopsis Biological ResourceCentre (ABRC, Ohio State University, Columbus, OH). EST122A10T7 corresponded to AtGLK1. EST 190N23T7 correspondedto AtGLK2. Both ESTs were sequenced and aligned to genomicsequence from BAC, Accession number AC007048(AtGLK1) andP1clone MLN1, Accession number AB005239(AtGLK2). Genestructure was determined by aligning EST sequence and genomicsequences.

Phylogenetic analysis

Phylogenetic analyses were carried out using PAUP* 4.0(Swofford, 1998). For the GARP superfamily phylogenetic analy-sis, 36 Arabidopsis GARP genes, six GLK genes, the Psr1 genefrom Chlamydomonas, six TEA genes, and the potato MYB-related gene MybSt1 were used. The GARP genes were identi®edfollowing BLAST searches of the Arabidopsis genome database(http://www.arabidopsis.org/BLAST/) using exon II of the maizeG2 gene as the query sequence. Individual genes were identi®edfrom within BACs by searching for the highly conserved sequencemotif ASHLQ (last ®ve amino acids encoded by exon II of G2). Foreach gene, nucleotide sequences equivalent to the second exonof G2, plus 12 bp either side of the exon, were aligned usingCLUSTALW (Thompson et al., 1994). This analysis was conductedusing parsimony with all characters unordered and equallyweighted. These results are presented as a majority rule bootstrapconsensus tree (Figure 2a) estimated from 1000 bootstrap repli-cates and a heuristic search strategy comprising 1000 randomaddition replicates with MULPARS on.

For the GLK gene analysis, full-length cDNAs of each GLK geneand APRR2 were aligned by hand according to the amino acidalignment shown in Figure 1. As the number of taxa is relativelysmall, these data were analysed using the exhaustive searchoption. The analysis found one most parsimonious tree (Figure2b). Bootstrap values were estimated from 1000 bootstrap repli-cates analysed using the exhaustive search option. These datawere also analysed using the maximum likelihood method.MODELTEST (Posada and Crandall, 1998) was used to select amodel of DNA substitution, GTR + G + I. The results of themaximum likelihood analysis yielded an identical topology tothe parsimony analysis.

Plant material and growth conditions

Arabidopsis thaliana ecotype Columbia was used in all experi-ments. Plants were grown in a Sanyo Versatile EnvironmentalTest Chamber (Gunma, Japan) with a daily cycle of 16 h light(100 mmol m±2 sec±1) at 22°C and 8 h dark at 18°C (long-dayconditions).

Extraction of DNA

DNA was isolated from plant tissue using a rapid DNA extractionmethod. Grinding sticks were prepared from disposable pipettetips by twisting molten tip ends into 1.5 ml Eppendorf tubes.Tissue was homogenized to a paste in an Eppendorf using agrinding stick, then resuspended immediately in 100 mM Tris±HClpH 8.0, 50 mM EDTA, 500 mM NaCl, 10 mM b-mercaptoethanoland 1.4% SDS. After incubation at 65°C for 10 min, samples wereadjusted to 1 M KOAc and after 5 min the debris was pelleted bycentrifugation. DNA was precipitated from the supernate usingisopropanol and NaOAc.

DNA gel-blot analysis

DNA gel-blot analysis was performed as described by (Langdaleet al., 1991). Hybridizations were performed in 0.45 M NaCl at65°C.

Extraction of RNA and gel-blot analysis

RNA was isolated, electrophoresed, blotted and hybridized asdescribed by Langdale et al. (1988). Densitometry of autoradio-graphs and analysis of ¯uorescence in ethidium bromide-stainedgels were performed using the BioRad Fluor-S MultiImager andBioRad (Hercules, CA, USA) MultiAnalyst software. Gene-speci®cprobes used for RNA gel-blot analysis were as follows: AtGLK1,224 bp fragment corresponding to positions 217±441 of Accessionnumber AY026772; AtGLK2, 261 bp fragment corresponding topositions 1±261 of Accession number AY026773; TUBULIN,1 : 1 : 1 : 1 mix of TUBULIN 1±4 (a gift from Dr M. Knight,Department of Plant Sciences, University of Oxford); LHCP1, ESTAP2L15e09R; LHCP6, EST 23C1T7; SRP43, 71A11T7; SRP54,847 bp fragment corresponding to positions 797± 1844 ofAccession number AF092168; CAO, EST 103D24T7; CHLH, ESTAP2L15fo3R Hybridizations were performed in 0.45 M NaCl at65°C.

Identi®cation of insertional mutants

To identify lines with dSpm insertions in or near to the GLKgenes, the Sainsbury Laboratory Arabidopsis Transposant(Tissier et al., 1999) lines were screened. DNA was obtainedfrom 48 pools of 50 lines and the inverse polymerase chainreaction (iPCR; Ochman et al., 1988), using dSpm primers, wasused to generate products containing DNA sequences ¯ankingeach insertion. iPCR products were spotted onto nylon ®lters byDr S. Rutherford (Department of Plant Sciences, University ofOxford) and ®lters were hybridized with AtGLK1 and AtGLK2cDNA sequences. In each case, a subpool of 50 lines wasidenti®ed. Seed corresponding to the subpool (kindly donatedby Professor Jonathan Jones, Sainsbury Laboratory, John InnesCentre, Norwich, UK) were germinated on peat plugs, and thosecontaining a dSpm element were selected by spraying phosphi-



nothricin (PPT) at 100 mg l±1 after 7 days. DNA was extractedfrom PPT-resistant plants that were pooled into 16 groups of six.PCR using GLK gene-speci®c primers and dSpm-speci®c primerswas then used to identify pools of plants containing an insertionline. PCR conditions were: 35 cycles of 94°C for 30 sec, 55°C for30 sec, 72°C for 2 min. DNA ampli®cation was achieved forAtGLK1 using primers Spm5 (5¢-CGGGATCCGACACTCTTTAATT-AACTGACACTC-3¢) and 2 bgs2 (5¢-AACTGCAGGTTACTGATCCG-ATTGTTCTT-3¢). For AtGLK2, positive ampli®cation was achievedusing primers Spm1 (5¢-CCTATTTCAGTAAGAGTGTGGGGTTTT-GG-3¢) and 5.2 (5¢-CTCCTTGACGTTGTAACTCCG-3¢) or Spm8 (5¢-GTTTTGGCCGACACTCCTTACC-3¢) and ara4 (5¢-TCCGATGTGAC-CTATATTTC-3¢). Having identi®ed a pool of six plants thatcontained an insertion line, the PCR was repeated using DNAfrom individual plants. In each case, copy number of the insertionallele was determined by restriction digestion (HindIII for AtGLK1and EcoRV for AtGLK2) and DNA gel-blot analysis. Both lineswere homozygous for the dSpm insertion and the mutant alleleswere named Atglk1.1 and Atglk2.1, respectively.

Preparation of total leaf protein and gel-blot analysis

Plant tissue (1 g) was harvested, ¯ash frozen in liquid N2 andhomogenized using a pipette grinder in 1.5 ml Eppendorf tubes.Laemmli SDS±PAGE buffer [(1 ml): 4% SDS, 20% glycerol, 10%mercaptoethanol, 0.004% bromophenol blue (BPB), 0.125 M Tris±HCl pH 6.8] was added and equal volumes of samples were thenelectrophoresed on 12% SDS±polyacrylamide gels. Proteins wereblotted onto 0.2 mm nitrocellulose membrane (Schleicher andSchuell, Dassel, Germany) using a Bio-Rad Mini-PROTEAN 3apparatus according to the manufacturer's instructions.Immunoreactions were carried out using horseradish peroxi-dase-conjugated secondary antibody and 4-chloro-1-naphtholsubstrate as described previously (Langdale et al., 1987). LHCPIIantibody was raised against maize protein and was a gift from W.Taylor (CSIRO, Canberra, Australia). RuBPCase antibody wasraised against wheat protein and was a gift from J. Gray(Cambridge University, UK). PsaD, D1 polypeptide, CF1a andCytf antibodies were raised against maize proteins and were giftsfrom A. Barkan (University of Oregon, Eugene, OR). VIPP1antibody was a gift from J. Soll (University of Kiel, Germany)and NDHH antibody was a gift from K. Steinmuller (University ofDusseldorf, Germany).

Chlorophyll assays

0.5 cm2 leaf sections were sampled under green safe light using abore with an 8 mm diameter. Discs were ¯ash frozen in liquid N2,ground in an Eppendorf using a grinding tip, then suspended in1 ml 80% acetone. Following centrifugation to remove debris, theOD of the supernate was measured at 645 and 663 nm.Chlorophyll concentration was calculated using the formula:

chlorophyll concentration (mg ml±1) = OD645 3 20.2 + OD663 3 8.0

(Arnon, 1949). This value was then divided by 0.5 to obtain thetotal chlorophyll concentration per cm2 tissue.

ANOVA

The experimental design of the chlorophyll assay resulted in anunbalanced data set, with each genotype represented by differentnumbers of plants. A general linear model (GLM) was applied

using the model `Chlorophyll concentration' = `LeafType' +`Copies GLK1' + `Copies GLK2' + `Copies GLK1*Copies GLK2'.This examined the variation attributed to differences in leaf type(cauline or rosette) and to the number of copies of the GLK genes.The `Copies GLK1*Copies GLK2' term assesses the variation inchlorophyll levels attributable to an interaction between the twoGLK genes. That is, variation that cannot be attributed to eitherGLK gene individually is explained when both GLK genes areincluded at the same time. Analysis was carried out using MINITAB

software (Coventry, UK).

Transmission electron microscopy

TEM analysis was carried out essentially according to Langdaleand Kidner (1994), with the following exceptions. Tissue sampleswere cut under and vacuum in®ltrated with Trumps ®xative (3%formaldehyde, 3% gluteraldehyde, 0.025 M phosphate bufferpH 7.2) for 2 h at room temperature. After sectioning, sampleswere mounted on Butvar B98 slots (Agar Aids, Essex, UK) andstained using a 2168 Ultrostainer Carlsberg System (Leica,Heerbrugg, Switzerland) in Ultrostain1 for 2 h and Ultrostain2for 10 min according to the manufacturer's instructions. Sectionswere photographed using AGFA Scientia EM ®lm 23 D 56(Leverkusen, Germany).

Acknowledgements

We thank Laura Rossini (University of Milan, Italy) and membersof the laboratory for stimulating discussions throughout thecourse of this work, and Cledwyn Merriman for technical assist-ance. This work was supported by grants from the BBSRC toJ.A.L.; M.J.C. was the recipient of a BBSRC postgraduatestudentship.

References

Allen, J.F. and Forsberg, J. (2001) Molecular recognition inthylakoid structure and function. Trends Plant Sci. 6, 317±326.

Andrews, T.J. and Lorimer, G.H. (1987) Rubisco: structure,mechanism and prospects for improvement. In TheBiochemistry of Plants (Hatch, M. D. and Boardman, N. K.,eds). Orlando, FL: Academic Press, pp. 131±218.

Arnon, D.L. (1949) Copper enzymes in isolated chloroplasts.Polyphenoloxidase in Beta vulgaris. Plant Physiol. 24, 1±15.

Baldi, P., Brunak, S., Frasconi, P., Soda, G. and Pollastri, G. (1999)Exploiting the past and the future in protein secondarystructure prediction. Bioinformatics, 15, 937±946.

Baranowskij, N., Frohberg, C., Prat, S. and Willmitzer, L. (1994) Anovel DNA binding protein with homology to Myboncoproteins containing only one repreat can function as atranscriptional activiator. EMBO J. 13, 5383±5392.

Barkan, A., Voelker, R., Mendel-Hartvig, J., Johnson, D. andWalker, M. (1995) Genetic analysis of chloroplast biolgenesis inhigher plants. Physiologia Plantarum, 93, 163±170.

Bowman, J.L., Alvarez, J., Weigel, D., Meyerowitz, E.M. andSmyth, D.R. (1993) Control of ¯ower development inArabidopsis thaliana by APETALA1 and interacting genes.Development, 112, 721±743.

BuÈ rglin, T.R. (1991) The TEA domain: a novel highly conservedDNA binding motif. Cell, 66, 11±12.

Douglas, S. (1994) Chloroplast origins and evolution. In TheMolecular Biology of Cyanobacteria (Bryant, D. A., ed.).Dordrecht, the Netherlands: Kluwer, pp. 91±118.



Edwards, G.E. and Walker, D.A. (1983) C3, C4: Mechanisms andCellular and Environmental Regulation of Photosynthesis.Oxford, UK: Blackwell Scienti®c Publications.

Groves, M.R., Mant, A., Kuhn, A., Koch, J., Dubel, S., Robinson, C.and Sinning, I. (2001) Functional characterization ofrecombinant chloroplast signal recognition particle. J. Biol.Chem, 276, 27778±27786.

Hall, L.N., Rossini, L., Cribb, L. and Langdale, J.A. (1998)GOLDEN2: a novel transcriptional regulator of cellulardifferentiation in the maize leaf. Plant Cell, 10, 925±936.

Imamura, A., Hanaki, N., Nakamura, A., Suzuki, T., Taniguchi, M.,Kiba, T., Ueguchi, C., Sugiyama, T. and Mizuno, T. (1999)Compilation and characterization of Arabidopsis thalianaresponse regulators implicated in His±Asp phosphorelaysignal transduction. Plant Cell Physiol. 40, 733±742.

Jonas-Straube, E., Hutin, C., Hoffman, N.E. and Schunemann, D.(2001) Functional analysis of the protein-interacting domains ofchloroplast SRP43. J. Biol. Chem, 276, 24654±24660.

Jones, D.T. (1999) Protein secondary structure prediction basedon position-speci®c scoring matrices. J. Mol. Biol. 292, 195±202.

Kalderon, D., Roberts, B.L., Richardson, W.D. and Smith, A.E.

(1984) A short amino acid sequence able to specify nuclearlocation. Cell, 39, 499±509.

Keegstra, K. and Froehlich, J.E. (1999) Protein import intochloroplasts. Curr. Opin. Plant Biol. 2, 471±476.

Kempin, S.A., Savidge, B. and Yanovsky, M.F. (1995) Molecularbasis of the cauli¯ower phenotype in Arabidopsis. Science, 267,522±525.

Kirchanski, S.J. (1975) The ultrastructural development of thedimorphic plastids of Zea mays L. Am. J. Bot. 62, 695±705.

Kirchanski, S.J. and Park, R.B. (1976) Comparative studies of thethylakoid proteins of mesophyll and bundle sheath plastids ofZea mays. Plant Physiol. 58, 345±349.

Klimyuk, V.I., Persello-Cartieaux, F., Havaux, M. et al. (1999) Achromodomain protein encoded by the Arabidopsis CAO geneis a plant-speci®c component of the chloroplast signalrecognition particle pathway that is involved in LHCPtargeting. Plant Cell, 11, 87±99.

Kroll, D., Meierhoff, K., Bechtold, N., Kinoshita, M., Westphal, S.,Vothknecht, U.C., Soll, J. and Westhoff, P. (2001) VIPP1, anuclear gene of Arabidopsis thaliana essential for thylakoidmembrane formation. Proc. Natl Acad. Sci. USA, 98, 4238±4242.

Langdale, J.A. and Kidner, C.A. (1994) bundle sheath defective, amutation that disrupts cellular differentiation in maize leaves.Development, 120, 673±681.

Langdale, J.A. and Nelson, T. (1991) Spatial regulation ofphotosynthetic development in C4 plants. Trends Genet. 7,191±196.

Langdale, J.A., Metzler, M.C. and Nelson, T. (1987) The argentiamutation delays normal development of photosynthetic cell-types in Zea mays. Dev. Biol. 122, 243±255.

Langdale, J.A., Rothermel, B. and Nelson, T. (1988) Cellularpatterns of photosynthetic gene expression in developingmaize leaves. Genes Dev. 2, 106±115.

Langdale, J.A., Taylor, W.C. and Nelson, T. (1991) The cell-speci®caccumulation of maize phosphoenolpyruvate carboxylase iscorrelated with demethylation at a speci®c site 3kb upstream ofthe gene. Mol. Gen. Genet. 225, 49±55.

Last, R.L., Bissinger, P.H., Mahoney, D.J., Radwanski, E.R. andFink, G.R. (1991) Tryptophan mutants in Arabidopsis: theconsequences of duplicated tryptophan synthase beta genes.Plant Cell, 3, 345±358.

Lawlor, D.W. (1993) Photosynthesis ± Molecular, Physiological

and Environmental Processes. Hong Kong: Longman Scienti®cand Technical.

Makino, S., Kiba, T., Imamura, A. et al. (2000) Genes encodingpseudo-response regulators: insight into His-to-Aspphosphorelay and circadian rhythm in Arabidopsis thaliana.Plant Cell Physiol. 41, 791±803.

Mandel, M.A., Gustafson-Brown, C., Savidge, B. and Yanofsky,M.F. (1992) Molecular characterization of the Arabidopsis ¯oralhomeotic gene Apetala1. Nature, 360, 273±277.

Margulis, L. (1975) Symbiotic theory of the origin of eukaryoticorganelles; criteria for proof. Symp. Soc. Exp. Biol. 29, 21±38.

Mould, R.M. and Robinson, C. (1991) A proton gradient is requiredfor the transport of two lumenal oxygen- evolving proteinsacross the thylakoid membrane. J. Biol. Chem. 266, 12189±12193.

Mullet, J.E. (1988) Chloroplast development and gene expression.Ann. Rev. Plant Physiol. Plant Mol. Biol. 39, 475±502.

Nakai, K. and Kanehisa, M. (1992) A knowledge base forpredicting protein localization sites in eukaryotic cells.Genomics, 14, 897±911.

Nielsen, V.S., Mant, A., Knoetzel, J., Moller, B.L. and Robinson, C.(1994) Import of barley photosystem I subunit N into thethylakoid lumen is mediated by a bipartite presequence lackingan intermediate processing site. Role of the delta pH intranslocation across the thylakoid membrane. J. Biol. Chem.269, 3762±3766.

Nowak, M.A., Boerlijst, M.C., Cooke, J. and Smith, J.M. (1997)Evolution of genetic redundancy. Nature, 388, 167±171.

Ochman, H., Gerber, A.S. and Hartl, D.L. (1988) Geneticapplications of an inverse polymerase chain reaction.Genetics, 120, 621±623.

Osteryoung, K.W. (2000) Organelle ®ssion. Crossing theevolutionary divide. Plant Physiol. 123, 1213±1216.

Ouali, M. and King, R.D. (2000) Cascaded multiple classi®ers forsecondary structure prediction. Protein Sci. 9, 1162±1176.

Pickett, F.B. and Meeks-Wagner, D.R. (1995) Seeing double:appreciating genetic redundancy. Plant Cell, 7, 1347±1356.

Posada, D. and Crandall, K.A. (1998) Modeltest: testing the modelof DNA substitution. Bioinformatics, 14, 618±622.

Pyke, K.A. and Leech, R.M. (1994) A genetic analysis ofchloroplast division in Arabidopsis thaliana. Plant Physiol.104, 201±207.

Pyke, K.A., Rutherford, S.M., Robertson, E.J. and Leech, R.M.(1994) arc6, a fertile Arabidopsis mutant with only twomesophyll cell chloroplasts. Plant Physiol. 106, 1169±1177.

Riechmann, J.L., Heard, J., Martin, G. et al. (2000) Arabidopsistranscription factors: genome-wide comparative analysisamong eukaryotes. Science, 290, 2105±2110.

Robbins, J., Dilworth, S.M., Laskey, R.A. and Dingwall, C. (1991)Two interdependent basic domains in nucleoplasmin nucleartargeting sequence: identi®ciation of a class of bipartite nucleartargeting sequence. Cell, 64, 615±623.

Robertson, E.J., Pyke, K.A. and Leech, R.M. (1995) arc6, anextreme chloroplast division mutant of Arabidopsis also altersproplastid proliferation and morphology in shoot and rootapices. J. Cell Sci. 108, 2937±2944.

Robinson, C., Hynds, P.J., Robinson, D. and Mant, A. (1998)Multiple pathways for the targeting of thylakoid proteins inchloroplasts. Plant Mol. Biol. 38, 209±221.

Rochaix, J.D. (1992) Post-transcriptional steps in the expression ofchloroplast genes. Annu. Rev. Cell Biol. 8, 1±28.

Rossini, L., Cribb, L., Martin, D.J. and Langdale, J.A. (2001) Themaize Golden2 gene de®nes a novel class of transcriptionalregulators in plants. Plant Cell, 13, 1231±1244.



Rubio, V., Linhares, F., Solano, R., Martin, A.C., Iglesias, J., Leyva,A. and Paz-Ares, J. (2001) A conserved MYB transcription factorinvolved in phosphate starvation signaling both in vascularplants and in unicellular algae. Genes Dev. 15, 2122±2133.

Sakai, H., Aoyama, T. and Oka, A. (2000) Arabidopsis ARR1 andARR2 response regulators operate as transcriptional activators.Plant J. 24, 703±711.

Schuenemann, D., Gupta, S., Persello-Cartieaux, F., Klimyuk,V.V., Jones, J.D.G., Nussaume, L. and Hoffman, N.E. (1998) Anovel signal recognition particle targets light-harvestingproteins to the thylakoid membranes. Proc. Natl Acad. Sci.USA, 95, 10312±10316.

Schwarz-Sommer, Z., Hue, I., Huijser, P., Flor, P.J., Hansen, R.,Tetens, F., Lonnig, W.E., Saedler, H. and Sommer, H. (1992)Characterization of the Antirrhinum ¯oral homeotic MADS-boxgene DEFICIENS: evidence for DNA binding and autoregulationof its persistent expression throughout ¯ower development.EMBO J. 11, 251±263.

Staehelin, L.A. and Arntzen, C.J. (1986) Photosynthesis III. Berlin,Germany: Springer Verlag.

Staehelin, L.A. and DeWit, M. (1984) Correlation of structure andfunction of chloroplast membranes at the supramolecular level.J. Cell Biochem. 24, 261±269.

Sugiura, M., Hirose, T. and Sugita, M. (1998) Evolution andmechanism of translation in chloroplasts. Annu. Rev. Genet, 32,437±459.

Swofford, D.L. (1998) Paup* Phylogenetic Analysis UsingParsimony (*and Other Methods), Version 4. Sunderland, MA:Sinauer Associates.

Thomas, J.H. (1993) Thinking about genetic redundancy. TrendsGenet. 9, 395±399.

Thompson, J.D., Higgins, D.J. and Gibson, T.J. (1994) CLUSTALW: improving the sensitivity of progressive multiple sequencealignment through sequence weighting, position-speci®c gappenalties and weight matrix choice. Nucl. Acids Res. 22, 4673±4680.

Tissier, A.F., Marillonnet, S., Klimyuk, V., Patel, K., Torres, M.A.,Murphy, G. and Jones, J.D. (1999) Multiple independentdefective suppressor-mutator transposon insertions inArabidopsis: a tool for functional genomics. Plant Cell, 11,1841±1852.

Trobner, W., Ramirez, L., Motte, P., Hue, I., Huijser, P., Lonnig,W.E., Saedler, H., Sommer, H. and Schwarz-Sommer, Z. (1992)GLOBOSA: a homeotic gene which interacts with DEFICIENS inthe control of Antirrhinum ¯oral organogenesis. EMBO J. 11,4693±4704.

Wykoff, D.D., Grossman, A.R., Weeks, D.P., Usuda, H. andShimogawara, K. (1999) Psr1, a nuclear localized protein thatregulates phosphorus metabolism in Chlamydomonas. Proc.Natl Acad. Sci. USA, 96, 15336±15341.

Yuan, J., Henry, R., McCaffery, M. and Cline, K. (1994) SecAhomolog in protein transport within chloroplasts: evidence forendosymbiont-derived sorting. Science, 266, 796±798.

Zachgo, S., Silva, E., Motte, P., Trobner, W., Saedler, H. andSchwarz-Sommer, Z. (1995) Functional analysis of theAntirrhinum ¯oral homeotic DEFICIENS gene in vivo andin vitro by using a temperature-sensitive mutant.Development, 121, 2861±2875.



the plant journal (2002) 31 glk gene pairs regulate...

Documents