KNOX genes influence a gradient of fruit chloroplast development through regulation of GOLDEN2-LIKE expression in tomato



The chlorophyll content of unripe fleshy fruits is positively correlated with the nutrient content and flavor of ripe fruit. In tomato (Solanum lycopersicum) fruit, the uniform ripening (u) locus, which encodes a GOLDEN 2-LIKE transcription factor (SlGLK2), influences a gradient of chloroplast development that extends from the stem end of the fruit surrounding the calyx to the base of the fruit. With the exception of the u locus, the factors that influence the formation of this developmental gradient are unknown. In this study, characterization and positional cloning of the uniform gray-green (ug) locus of tomato reveals a thus far unknown role for the Class I KNOTTED1-LIKE HOMEOBOX (KNOX) gene, TKN4, in specifying the formation of this chloroplast gradient. The involvement of KNOX in fruit chloroplast development was confirmed through characterization of the Curl (Cu) mutant, a dominant gain-of-function mutation of TKN2, which displays ectopic fruit chloroplast development that resembles SlGLK2 over-expression. TKN2 and TKN4 act upstream of SlGLK2 and the related gene ARABIDOPSIS PSEUDO RESPONSE REGULATOR 2-LIKE (SlAPRR2-LIKE) to establish their latitudinal gradient of expression across developing fruit that leads to a gradient of chloroplast development. Class I KNOX genes typically influence plant morphology through maintenance of meristem activity, but this study identifies a role for TKN2 and TKN4 in specifically influencing chloroplast development in fruit but not leaves, suggesting that this fundamental process is differentially regulated in these two organs.


In plants, many biochemical pathways, including those associated with photosynthesis and synthesis of primary and specialized metabolites, are located within the chloroplast. Leaves are typically the primary source of photosynthate, although it is estimated that photosynthesis by fleshy fruits contributes up to 20% of the total fruit carbohydrate (Blanke and Lenz, 1989; Hetherington et al., 1998). Enhanced chloroplast development and chlorophyll content of unripe fruits increases their photosynthetic capacity, leading to greater accumulation of starch (Dinar and Stevens, 1981; Powell et al., 2012; Sagar et al., 2013). Furthermore, the starch content of unripe fruit is positively correlated with the sugar content in ripe fruit, which directly contributes to fruit quality by influencing nutritional value and taste (Davies and Hobson, 1981). Thus, chloroplast development and the carbohydrate status of unripe fruits affect the composition and quality of ripe fruit.

In tomato, several loci influence the development of fruit chloroplasts and the subsequent quality of ripe fruit. Mutations at the high-pigment 1 and 2 (hp-1 and hp-2) loci, which encode negative regulators of photomorphogenesis, possess an increased number of chloroplasts and plastid compartment size, leading to fruits with higher levels of chlorophyll and carotenoids (Mustilli et al., 1999; Cookson et al., 2003; Liu et al., 2004; Kolotilin et al., 2007; Rohrmann et al., 2011). Alteration of phytohormone biosynthesis and responses, including those associated with auxin, ABA, cytokinins and brassinosteroids (BRs), can also have a positive impact on chloroplast development that influences the quality of ripe fruit (Martineau et al., 1994; Galpaz et al., 2008; Sagar et al., 2013; Liu et al., 2014). In contrast, mutation at the uniform ripening (u) locus has a negative impact on the development of fruit chloroplasts, due to loss of the green shoulder that surrounds the calyx end of developing fruit, leading to ripe fruits with a reduced sugar content (Powell et al., 2012). Positional cloning of u revealed a mutation in a tomato homolog of the GOLDEN 2-LIKE (GLK) transcription factor of maize, designated SlGLK2 (Powell et al., 2012). GLKs are required for chloroplast development (Fitter et al., 2002; Waters et al., 2008) and SlGLK2 is preferentially expressed in the shoulder of the fruit, where in U/U genotypes it is involved in enhancing chloroplast development while ectopic expression of SlGLK2 results in homogeneously dark-green fruit (Powell et al., 2012). Similarly, over-expression of ARABIDOPSIS PSEUDO RESPONSE REGULATOR 2-LIKE (SlAPRR2-LIKE) in tomato, which is related to SlGLK2, causes enhanced development of fruit chloroplasts (Pan et al., 2013). How these loci interact to jointly influence the development of fruit chloroplasts is not fully understood, although recent evidence suggests a role for both auxin and BRs in regulating GLK expression (Sagar et al., 2013; Liu et al., 2014). Furthermore, except for the involvement of SlGLK2, the factors that establish the latitudinal gradient of gene expression and chloroplast development in tomato fruit remain undefined.

Fruits of the uniform gray-green (ug) mutant of tomato are similar to u and lack the green shoulder surrounding the calyx (Bohn and Scott, 1945). In this study, positional cloning revealed that UG encodes TKN4, a Class I KNOTTED1-LIKE HOMEOBOX (KNOX) transcription factor. KNOX genes typically influence plant morphology, and particularly leaf shape, through maintenance of stem cell populations within apical meristems (Hay and Tsiantis, 2010). However, the present work identifies a previously undefined role for these homeobox proteins in the development of fleshy fruit, and in particular in the establishment of a gradient of chloroplast development through regulation of SlGLK2 expression.


Positional cloning of ug defines a role for KNOX in fruit chloroplast development

The ug mutant lacks the green shoulder surrounding the calyx end of developing fruit (Figure 1a). Chlorophyll levels are reduced in the pericarp of ug fruit (Figure 1b–e). The development of individual chloroplasts in ug fruit is also impaired, with a reduction in size and in the number of thylakoids per granum (Figure 1f–k). The phenotypes of ug are similar to those of fruits with the u mutant allele (Powell et al., 2012) (Figure 1a–d,g,j). Genetic mapping assigned the ug locus to the long arm of chromosome 1, with tight linkage to the marker C2_At1g56345. Fine mapping refined the ug locus to a 14.6-kb interval between the markers Solyc01g100500 and Solyc01g100520 that spans nucleotides SL2.40chr1:82268632–82283258 and contains the single gene Solyc01g100510, which encodes the KNOX transcription factor TKN4. Sequencing of TKN4 from flower cDNA of UG/UG and ug/ug genotypes revealed a single T → A nucleotide change within exon 2 of TKN4 corresponding to a phenylalanine to leucine substitution at residue 137 (Figure 2a). Phenylalanine-137 represents an invariant amino acid within the KNOX2 domain of diverse Class I KNOX proteins (Figure 2b). The KNOX2 domain mediates protein–protein interactions between homeodomain proteins (Bellaoui et al., 2001; Smith et al., 2002). Genotyping of the single nucleotide polymorphism (SNP) associated with the ug phenotype from LA4025 and Purple Prince, which lack the green shoulder and are purported to carry the ug mutation, implied that the F137L substitution co-segregates with the mutant phenotype, indicating that the SNP is conserved in ug/ug genotypes (Figure 2c and Figure S1 in Supporting Information). Attempts to genotype the ug mutant allele in the LA0021 accession, which lacks a green shoulder but is reported to be a ug mutant, failed (Figure S1). Similarly, amplification of TKN4 from LA0021 flower cDNA was unsuccessful (Figure 2d). Subsequent amplification and sequencing of TKN4 genomic fragments from LA0021 revealed the existence of a second mutant allele that contains a 2.4-kb internal deletion spanning nucleotides SL2.40chr1:82280365–82282785 of the tomato genome reference assembly (Figure 2e). Although mutations in Class I KNOX genes typically cause phenotypes associated with altered morphology, and particularly leaf shape (Parnis et al., 1997; Venglat et al., 2002; Hay and Tsiantis, 2010), the effect of ug is subtle and plants carrying the ug mutant allele are indistinguishable from the wild type (WT) except for the lack of a green shoulder in developing fruit. No impact was observed on leaf shape or leaf chlorophyll in ug (Figure S2a–c). TKN4 is expressed in flower pistils and mature green fruits but expression is low in leaves, supporting the lack of an altered phenotype in leaves of the ug mutant (The Tomato Genome Consortium, 2012) (Figure S2d, e). As the ug and u fruit phenotypes are virtually identical (Figure 1a), germplasm lacking a green shoulder was also genotyped for the u mutant allele. While LA4025 is homozygous for both u and ug mutant alleles, LA3539, LA0021 and the Purple Prince cultivar possess the WT (U/U) allele (Figures 2c and S1). These data indicate that the lack of a green shoulder in these accessions is the result of mutations in TKN4 and not mutations at the u locus.

Figure 1.

Fruit phenotype and chloroplast development in the uniform gray-green (ug) mutant. (a) Phenotypes of the uniform ripening (u) and ug mutants compared with Ailsa Craig wild type (WT). The arrow highlights the presence of the green shoulder in WT fruit. (b) Chlorophyll content in the pericarp isolated from the shoulder (S) and base (B) of WT, u and ug fruit at 28 days post-anthesis (dpa). Data are presented as the mean ± SEM of = 5 and means with different letters are significantly different (least squares means, < 0.05). (c)–(e) Confocal laser scanning microscopy images of the shoulder pericarp of WT, u and ug fruit showing chlorophyll autofluorescence. Scale bar: 10 μm. (f)–(k) Transmission electron micrographs of 28-dpa fruit chloroplasts of WT, u and ug, Scale bars: (f)–(h), 2 μm; (i)–(k), 0.5 μm).

Figure 2.

Genetic mapping of the uniform gray-green (ug) locus. (a) Genetic and physical map of the ug locus identifying a single candidate gene, Solyc01g100510, which encodes TKN4. The numbers of recombinant individuals between genetic markers are provided. Sequencing of TKN4 from wild type (WT) (UG/UG) and mutant (ug/ug) plants identified a single nucleotide polymorphism that converts phenylalanine 137 to leucine. (b) Alignment of the KNOX2 domain of phylogenetically distinct KNOX proteins. The invariant phenylalanine residue within this domain and its substitution to leucine in ug/ug are shown in red. (c) Correlation of L137 with the ug/ug phenotype in selected accessions of tomato. (d) Reverse transcriptase-PCR amplification of TKN4 from Ailsa Craig (WT) and LA0021 flower cDNA. (e) TKN4 amplification from WT and LA0021 genomic DNA highlighting 2.4-kb deletion in LA0021. This deletion spans the first three exons of TKN4 and is highlighted by the red bar in (a).

KNOX over-expression alters fruit development and has a positive impact on chlorophyll accumulation

Additional support for the role of KNOX genes in regulating the development of fruit chloroplast is evident from characterization of the Curl (Cu) mutant. Cu represents a dominant gain-of-function mutation resulting from ectopic expression of TKN2, leading to altered plant morphology (Parnis et al., 1997). Developing Cu fruit possess elevated chlorophyll levels that extend across the entire surface of the fruit (Figure 3a–e). In addition to the increased chloroplast number and higher chlorophyll content in Cu fruit, the size of the chloroplasts is larger than in WT and there is a general increase in the number of thylakoids per granum (Figure 3f–i). The phenotypes of Cu fruit are reminiscent of SlGLK2 and SlAPRR2-LIKE over-expression lines (Powell et al., 2012; Pan et al., 2013). To assess the potential relationship between Cu and the SlGLKs and SlAPRR2-LIKE, the expression of these genes was examined in the pericarp of WT and Cu fruit. The expression of TKN2, SlGLKs and SlAPRR2-LIKE is elevated in Cu fruit (Figure 3j–m). However, SlGLK1 expression, which is typically lower than SlGLK2 expression in tomato fruit (Powell et al., 2012), was not statistically different between WT and Cu. These data suggest that SlGLK2 and SlAPRR2-LIKE act downstream of KNOX activity to regulate the development of fruit chloroplasts. In support of this hypothesis, fruit of a double mutant constructed from a cross between Cu and u partially suppressed the Cu mutant phenotype (Figure 3n,o). The Cu fruit are smaller in size (WT = 37.61 ± 2.53 g; Cu = 17.24 ± 1.69 g; = 8) with a thinner pericarp and smaller cell size that results in an increased density of cells per unit area (Figure 3d,p,q). Congruent with the phenotype of the ug allele (Figure S2a–c) the Cu allele has no impact on leaf chlorophyll content (Figure S3a), and while the expression of the SlGLKs remains unchanged the expression of SlAPRR2-LIKE is slightly elevated in Cu leaves (Figure S3b–e).

Figure 3.

Enhanced chloroplast development in Curl (Cu) fruit. (a), (b) Variation in fruit phenotype between Ailsa Craig wild type (WT) and Cu at 16 days post-anthesis (dpa). (c), (d) Confocal laser scanning microscopy images of WT and Cu fruit shoulder pericarp showing chlorophyll autofluorescence. Scale bar: 10 μm. (e) Chlorophyll content of WT and Cu fruit pericarp (n = 5). (f)–(i) Transmission electron micrograph of 28-dpa fruit chloroplasts of WT and Cu. Scale bars: (f), (g), 2 μm; (h) and (i), 0.5 μm. (j)–(m) Quantitative RT-PCR analysis of gene expression in WT and Cu fruit pericarp. (n) Fruit phenotype of Cu and Cu/u. (o) Chlorophyll content of Cu and Cu/u fruit pericarp (= 3). (p), (q) Pericarp thickness of WT and Cu fruit (= 6). Data are presented as the mean ± SEM. Asterisks denote significant differences (***< 0.001; **< 0.01) as determined by Student's t-tests. ns, not significant.

KNOX loci establish gradients of SlGLK2 expression that affect chlorophyll accumulation in fruit

The presence of a green shoulder in immature tomato fruit highlights the existence of a latitudinal gradient of chloroplast development that correlates with the increased expression of SlGLK2 at the calyx end of the fruit relative to the stylar end (Powell et al., 2012). To assess the role of KNOX genes in regulating the formation of this gene expression gradient and its associated differential pattern of chlorophyll accumulation, fruit pericarp was divided into whole, shoulder, middle and base, and the expression of SlGLK2, SlAPRR2-LIKE, TKN2 and TKN4 investigated (Figure 4a). In WT fruit, chlorophyll levels are higher in the fruit shoulder than in the middle and base of the fruit (Figure 4b), and the expression of each gene matches the gradient in chlorophyll accumulation with higher transcript levels observed in the shoulder compared with the middle and base, although this pattern is more pronounced for SlGLK2, TKN2 and TKN4 (Figures 4c–f and S4). The Cu and ug mutations disrupt both the chlorophyll gradient (Figure 4b) and the expression gradient that is present in WT fruit. For example, the expression level of SlGLK2 is positively affected by Cu and negatively affected by ug, and the typical gradient of expression is disrupted (Figure 4c). In contrast, the ug locus has no impact on the expression of SlAPRR2-LIKE, and while the expression of this gene is elevated in Cu the expression gradient is not disrupted (Figure 4d). Together, these data not only support the role of KNOX genes in regulating both chloroplast development and gene expression gradients but also highlight that while SlGLK2 and SlAPRR2-LIKE both have a positive impact on fruit chloroplast development (Powell et al., 2012; Pan et al., 2013) they are regulated by overlapping, yet distinct, mechanisms.

Figure 4.

Mutations at knox loci disrupt the gradient of SlGLK2 expression and affect fruit chloroplast development. (a) Depiction of the sampling strategy. (b) Chlorophyll content in the pericarp isolated from the whole (W), shoulder (S), middle (M) and base (B) of Ailsa Craig wild-type (WT), Curl (Cu) and uniform gray-green (ug) mutant fruit at 28 days post-anthesis (= 5). (c)–(f) Quantitative RT-PCR analysis of expression of SlGLK2, SlAPRR2-LIKE, TKN2 and TKN4 in the same tissue samples as described in (b) relative to the expression level in the base of WT fruits (= 3). Data are presented as the mean ± SEM and means with different letters are significantly different (least squares means, < 0.05).

Several studies implicate phytohormones in mediating fruit chloroplast development in tomato, and in the case of auxin and BRs enhanced fruit chloroplast development is positively correlated with increased SlGLK expression (Martineau et al., 1994; Galpaz et al., 2008; Sagar et al., 2013; Liu et al., 2014). In addition, auxin gradients are involved in several developmental processes, and disruption of auxin signaling in the arf6/arf8 double mutant results in ectopic KNOX expression and altered floral morphology (Benkova et al., 2003; Petersson et al., 2009; Tabata et al., 2010). The potential role of auxin in regulating the KNOX-mediated chloroplast gradient in tomato fruit was investigated utilizing SlARF4-silenced lines, which possess increased starch content, enhanced chloroplast development and elevated SlGLK1 expression (Sagar et al., 2013). The SlARF4-silenced lines were developed in the Micro-Tom background (Kobayashi et al., 2014), which is homozygous for the u mutation and lacks a functional SlGLK2. Unexpectedly, and unlike the u mutant allele in the Ailsa Craig background (Figure 1a and b), a decreasing gradient of chlorophyll content from the shoulder to the base of the fruit was observed in both Micro-Tom and a SlARF4-silenced line, with the latter displaying the expected elevated chlorophyll content (Figure 5a). The expression of SlGLK2 and SlAPRR2-LIKE was elevated in the SlARF4-silenced line, particularly in the shoulder, and whereas the expression of TKN2 and TKN4 was also enriched in the fruit shoulder, the difference in expression levels between Micro-Tom and the SlARF4-silenced line are minor (Figure 5b–e). These data highlight the existence of a gradient of chlorophyll accumulation in Micro-Tom that occurs independently of SlGLK2 and suggest that SlARF4 negatively regulates the expression of SlGLK2 and SlAPRR2-LIKE.

Figure 5.

Silencing of SlARF4 affects fruit chlorophyll and expression of SlGLK2 and SlAPRR2-LIKE. (a) Chlorophyll content in whole (W), shoulder (S) and base (B) pericarp of Micro-Tom and SlARF4 antisense lines at 21 days post-anthesis (= 5). (b)–(e) Relative expression level of SlGLK2, SlAPRR2-LIKE, TKN2 and TKN4 in same samples described in (a). Data are presented relative to the expression level in base of Micro-Tom fruits (= 3). Means ± SEM with different letters are significantly different (least squares means, < 0.05).

The fruit phenotype of hp-1 is partially suppressed by u and ug

Fruits of the hp-1 mutant have an increased number of chloroplasts, an increased plastid compartment size and a higher chlorophyll content (Cookson et al., 2003). The contribution of KNOX and SlGLK2 to the hp-1 phenotype was assessed using a combination of single and double mutants in both WT and the hp-1 background. Both u and ug partially suppress the hp-1 phenotype, although suppression by u is greater than that achieved by ug and a slightly additive phenotype is apparent when u and ug are combined (Figure 6a,b). Congruent with the elevated levels of chlorophyll, expression of SlGLK2, SlAPRR2-LIKE, TKN2 and TKN4 is higher in hp-1 fruit pericarp than WT (Figure 6c–f). Together, these data indicate that both SlGLK2 and TKN4 contribute to the hp-1 mutant phenotype and are also suggestive of a role for SlAPRR2-LIKE in determining the elevated chlorophyll content of hp-1 fruit, possibly via regulation by TKN2. The reduced suppression of hp-1 by ug compared with u may be due to the residual expression of SlGLK2 observed in the ug mutant (Figure 4c), which may indicate that the ug allele (accession LA3416), which carries the F137L substitution, is not a complete null or that additional unknown factors regulate SlGLK2 expression. Furthermore, evidence for the involvement of SlAPRR2-LIKE or additional transcription factors in determining the elevated chlorophyll content of hp-1 fruit is supported by the observation that while chlorophyll levels of the hp1/u mutant are reduced compared with those observed in WT fruit, they are not reduced to the levels observed in the u mutant (Figure 6b).

Figure 6.

The uniform ripening (u) and uniform gray-green (ug) loci partially suppress high-pigment 1 (hp-1). (a) Variation in fruit phenotype between single, double and triple mutant combinations of hp-1, u and ug at 21 days post-anthesis (dpa). (b) Chlorophyll content in the pericarp of genotypes in (a) (= 5). (c)–(f) Relative expression level of SlGLK2, SlAPRR2-LIKE, TKN2 and TKN4 in wild type (WT) and hp-1 pericarp at 21 dpa. Data are presented as the mean ± SEM. Means with different letters are significantly different (least squares means, < 0.05) and asterisks denote significant differences (**< 0.01) as determined by Student's t-test.


A role for KNOX genes in the development of fleshy fruit

KNOX genes influence multiple aspects of plant morphology and are typically expressed within apical meristems, where they maintain pluripotent cell populations required for organ initiation (Jackson et al., 1994; Long et al., 1996; Byrne et al., 2002; Belles-Boix et al., 2006; Hay and Tsiantis, 2010). Loss of KNOX activity can lead to inhibition of shoot formation, whereas ectopic KNOX expression results in pleiotropic phenotypes including increased lobe formation in simple leaves and increased complexity of compound leaves (Lincoln et al., 1994; Hareven et al., 1996; Long et al., 1996; Hay and Tsiantis, 2006). In addition to roles in shoot apical meristem formation and leaf development, KNOX genes influence plant height, disrupt apical dominance and alter patterns of cell division and elongation (Douglas et al., 2002; Venglat et al., 2002). The Arabidopsis KNOX genes, SHOOT MERISTEMLESS (STM), BREVIPEDICELLUS (BP), KNAT2 and KNAT6, also influence development of the inflorescence and carpels together with the architecture of abscission zones (Scofield et al., 2007; Ragni et al., 2008; Shi et al., 2011). The role of individual KNOX genes in tomato remains obscure, but identification of the underlying basis of the ug mutant has revealed an atypical role for TKN4 in influencing fruit chloroplast development and specifically the formation of the green shoulder (Figure 1). This role is supported through characterization of the fruit phenotypes of Cu that result from ectopic expression of TKN2 (Figure 3). TKN2 is closely related to the STM gene of Arabidopsis, which is required for both vegetative and reproductive meristem development (Long et al., 1996; Scofield et al., 2007; Hay and Tsiantis, 2010). As observed by TKN2 over-expression in the Cu mutant, ectopic expression of STM also leads to aberrant meristem activity and altered plant morphology (Parnis et al., 1997; Gallois et al., 2002; Scofield et al., 2013). Within the shoot apical meristem, both TKN2 and STM function in part through modulation of hormone activity, and particularly by influencing the levels of cytokinins and gibberellins (Hay et al., 2002; Jasinski et al., 2005; Yanai et al., 2005; Hay and Tsiantis, 2010; Shani et al., 2010). At present the role of these hormones in contributing to the altered phenotypes observed in the fruit of the Cu mutant (Figure 3) remains to be defined. TKN4 does not possess an apparent ortholog in Arabidopsis, although previous phylogenetic analysis suggested that it is closely related to FaKNOX1 from strawberry and NTH1 in tobacco (Chatterjee et al., 2011). Manipulation of FaKNOX1 expression in transgenic strawberry and Arabidopsis alters plant morphology, including leaf, flower and meristem architecture. However, altered expression of additional KNOX family members was also apparent in FaKNOX1-silenced lines, raising the possibility that the observed phenotypes could, in part, be due to altered expression of other strawberry KNOX genes (Chatterjee et al., 2011). Similarly, over-expression of NTH1 in tobacco altered plant morphology, manifested by crinkling of the leaves and altered phyllotaxy, phenotypes that are typical of KNOX over-expression (Tamaoki et al., 1999). Interestingly, the ug mutant does not display any obvious phenotypes related to organ morphology (Figure S2), which may be the result of partial redundancy, a phenomenon often observed between KNOX genes (Byrne et al., 2002; Belles-Boix et al., 2006) and which is supported by the overlapping expression patterns of the four tomato Class I KNOX genes TKN1–4 in different tissues and organs (The Tomato Genome Consortium, 2012). However, functional specificity also exists between the tomato KNOX genes, as although TKN2 and TKN4 display increased expression in the shoulder of the fruit (Figures 4 and S4), TKN2 does not substitute for reduced TKN4 activity in the ug mutant. Furthermore, characterization of Cu revealed a broader role for KNOX genes in fleshy fruit development as TKN2 over-expression not only promoted chloroplast development but also affected fruit size, pericarp thickness and pericarp cell density (Figure 3). Identification of a loss-of-function allele of TKN2 or fruit-specific gene silencing of this gene will facilitate improved understanding of its role in fruit development, including chloroplast development.

Differential regulation of chloroplast development in tomato leaves and fruits

Both u and ug affect the gradient of chloroplast development in tomato fruit, although neither locus influences chloroplast development in leaves (Figures 1 and S2a–c). In the case of the u locus the lack of impact on leaf chloroplast development is likely to be due to redundancy and the overlapping expression of SlGLK1 and SlGLK2 in tomato leaf (Powell et al., 2012), whereas the lack of impact of ug on leaf chlorophyll content may be due to the absence of TKN4 expression in leaves (The Tomato Genome Consortium, 2012). However, over-expression of TKN2 in the Cu mutant does not increase chlorophyll content in leaves but affects chloroplast development and the accumulation of chlorophyll in fruit (Figures 3 and S3). These data suggest differential regulation of chloroplast development in fruits and leaves of tomato, a hypothesis supported by data indicating that over-expression of GLKs, and SlAPRR2-LIKE, together with suppression of SlARF4, has a positive impact on chloroplast development in fruits but not leaves (Powell et al., 2012; Pan et al., 2013; Sagar et al., 2013). The underlying basis for this differential regulation of chloroplast development in tomato is unknown but may be due to saturation of chloroplast development in tomato leaves, as these contain levels of chlorophyll that are approximately 20 times higher than in fruit (Figures 1b and S2c) and therefore possibly less amenable to further enrichment. In addition to the apparent differential regulation of chloroplast development in tomato leaves and fruits, parallel pathways regulate fruit chloroplast development. For example, the impact of SlAPRR2-LIKE over-expression on fruit chloroplast development is independent of SlGLK2, as experiments were performed in Micro-Tom which is homozygous for the u mutant allele (Pan et al., 2013; Kobayashi et al., 2014). Similarly, although u and ug suppress the hp-1 mutant phenotype, they do not restore chlorophyll to levels detected in the single mutants in the absence of hp-1 (Figure 6a,b). This discrepancy may be due in part to the elevated expression of SlAPRR2-LIKE in the hp-1 background (Figure 6d). The availability of SlAPRR2-LIKE mutant alleles or silencing of SlAPRR2-LIKE in WT and mutant backgrounds will facilitate greater understanding of the role of this gene in chloroplast development, together with its relationship to KNOX genes and SlGLK2.

TKN4 influences the gradient of chloroplast development in tomato fruit through regulation of SlGLK2

Transcription factor gradients play important roles in the development of multicellular organisms but remain poorly understood in plants (Stathopoulos and Levine, 2002; Sozzani et al., 2010; Chen et al., 2012). In grasses, leaf differentiation proceeds basipetally and is accompanied by a gradient of differential gene expression, chloroplast development and photosynthetic competency (Leech et al., 1973; Li et al., 2010). Similarly, the present study has investigated a tissue-level macro-gradient of chloroplast formation that occurs during the development of tomato fruit that leads to the formation of a green shoulder. The physiological role of this chloroplast gradient is unknown, but its formation is increased under shade conditions and the shoulder possesses higher photosynthetic capacity than the remaining pericarp tissues, suggesting that this may be an adaptive trait that contributes to fruit photosynthesis (Smillie et al., 1999). Manipulation of KNOX activity, either through mutation in TKN4 in ug or over-expression of TKN2 in Cu, disrupts the gradient of chloroplast development in tomato fruit by altering the spatial expression of SlGLK2 and SlAPRR2-LIKE (Figure 4). However, TKN2 and TKN4 possess overlapping and distinct activities, and while over-expression of TKN2 influences the expression of both SlGLK2 and SlAPRR2-LIKE, reduced TKN4 activity only disrupts SlGLK2 expression (Figures 4 and 7). These data promote a model (Figure 7) suggesting the existence of distinct regulatory modules that influence chloroplast development in tomato fruit, and it will be of interest to determine whether TKN2 and TKN4 bind directly to regulatory regions of SlGLK2 and SlAPRR2-LIKE. Similarly, while suppression of SlARF4 did not dramatically alter the expression of TKN2 and TKN4, expression of both SlGLK2 and SlAPRR2-LIKE was elevated (Figure 5). These data suggest multiple regulatory inputs that influence the expression of SlGLK2 and SlAPRR2-LIKE, a hypothesis supported by the recent report indicating that activation of BR responses leads to elevated expression of SlGLK2 and enhanced chloroplast development (Liu et al., 2014). Although TKN4 influences the latitudinal gradient of SlGLK2 expression in tomato fruit, TKN2 and TKN4 also display a gradient-dependent expression pattern in developing fruit, suggesting that KNOX genes themselves do not establish the gene expression gradient that exists in this tissue and that as yet unidentified signaling pathways are required. Furthermore, KNOX proteins and their mRNAs are also trafficked through plasmodesmata (Lucas et al., 1995), raising the possibility that the fruit chloroplast gradient may, at least in part, also be regulated through cell to cell movement of these proteins as well as their gradient-dependent transcript accumulation. The impact of gene expression gradients in plants is not fully understood, but this study has identified a role for KNOX genes, whereby they regulate the gradient-dependent expression pattern of transcription factors required for the development of fruit chloroplasts.

Figure 7.

A proposed model for KNOX activity in developing tomato fruit. The KNOX genes TKN2 and TKN4 positively influence SlGLK2 and SlAPRR2-LIKE expression to promote fruit chloroplast development. Auxin, via SlARF4, represses the expression of SlGLK2 and SlAPRR2-LIKE whereas brassinosteroids (BRs), via expression of BZR1-1D, promotes SlGLK2 expression (Liu et al., 2014). Data indicate that TKN2 influences the expression of both SlGLK2 and SlAPRR2-LIKE, whereas TKN4 only influences the expression of SlGLK2. HP-1 acts upstream of these genes to repress their expression.

Experimental Procedures

Plant material, growth conditions and chlorophyll extraction

Seeds of tomato (Solanum lycopersicum) cultivars, mutants and mutant combinations at the u, ug, Cu and hp-1 loci (Table S1), together with Solanum pimpinellifolium accession LA1589, were obtained from the Tomato Genetics Resource Center, UC Davis, CA, USA or commercial vendors. Plants were grown in peat-based compost supplemented with fertilizer in greenhouses at Michigan State University, East Lansing, MI, USA under conditions of 16-h day (25°C) and 8-h night (20°C). Micro-Tom and the SlARF4-silenced line (Sagar et al., 2013) were grown in a growth room under similar conditions. Flowers were tagged at anthesis and fruits subsequently harvested at defined ages based on days post-anthesis. The chlorophyll content was sampled from fruit pericarp and the terminal leaflet of expanding leaves of 4-week-old plants as previously described (Nadakuduti et al., 2012).

Genetic mapping of the ug locus and genotyping

An interspecific F2 population segregating for the ug locus was generated from a cross between S. lycopersicum (ug/ug) (LA3539) and S. pimpinellifolium (UG/UG) (LA1589). Fruits from 60 F2 individuals were phenotyped for the presence or absence of a green shoulder and plants were genotyped with genetic markers spanning the tomato genome. Subsequent fine mapping was performed on further 1600 F2 individuals using chromosome 1 genetic markers (Table S2). Genomic DNA was extracted as previously described (Barry et al., 2005). Details of genetic and physical maps together with associated molecular markers are available through the Sol Genomics Network. Genotyping tomato accessions (Table S1) for the ug mutant allele was achieved using a cleaved amplified polymorphic sequence marker (Table S3) that was resolved by digestion with MseI. Genotyping for the u mutant allele was achieved through sequencing of PCR fragments amplified using primers (Table S3) that span the single nucleotide insertion site (Powell et al., 2012).

Molecular cloning and sequence analysis

Full-length cDNA clones of TKN4 were amplified and cloned from flower cDNA of WT and the ug mutant as previously described (Nadakuduti et al., 2012). Five independent clones from each genotype were selected for sequencing. The entire TKN4 gene was sequenced from LA0021 genomic DNA. Sequences from the region with an unexpected band size containing the deletion were cloned, and four independent clones were sequenced and assembled using sequencher®version 4.7. Amino acid alignments were generated using muscle (Edgar, 2004). Primers are listed in Table S3 and additional sequences utilized in multiple sequence alignments are provided in Table S4.

Quantitative RT-PCR analysis

Expression analysis was performed according to Nadakuduti et al. (2012) with the exception that 10-μl PCR reactions were assembled using a Biomek 3000 liquid handler ( and amplified using an Applied Biosystems ABI Prism 7900HT Real-Time PCR System ( with the following sequence: 2 min at 50°C and 10 min at 95°C, followed by 40 cycles of 15 sec at 95°C and 1 min at 60°C. Primers are listed in Table S3.


Chloroplast imaging using confocal laser-scanning microscopy (CLSM) and transmission electron microscopy (TEM) was performed as previously described (Barry et al., 2012) with minor modifications. The CLSM images were captured using an objective with a numerical aperture of 0.5 and maximum intensity projection images were generated from a confocal z-series acquired through an average thickness of 170 μm in 10-μm increments. Samples for TEM imaging were embedded in Spurr resin.

Statistical analyses

Statistical analyses were performed using sas. The genotypic constituents were evaluated by Student's t-test and least squares means.


This research was supported by a Discretionary Funding Initiative award from Michigan State University (MSU), a Strategic Partnership Grant from the MSU Foundation and by USDA National Institute of Food and Agriculture, hatch project number MICL02265. WLH was supported by the MSU College of Agriculture and Natural Resources Undergraduate Research Program, the Larry D. Fowler Endowment Undergraduate Research Stipend and an Undergraduate Research Support Program Grant from MSU College of Natural Sciences. We thank Professor Mondher Bouzayen, University of Toulouse, France for providing seeds of the SlARF4 antisense line, Drs Alicia Pastor and Melinda Frame of the Center for Advanced Microscopy at MSU for their assistance and Dr Rebecca Grumet for comments on the manuscript.