The making of a chloroplast

The biogenesis of chloroplasts requires substantial protein import from the cytosol. Most chloroplast proteins are imported through the Toc/Tic complex, which both recognizes and transports nascent proteins across both envelope membranes (for a review, see Soll and Schleiff, 2004). Major components of the Toc/Tic complex are upregulated by light and even provide substrate specificity. For example, the Arabidopsis Toc33 knockout mutant, ppi1, is defective in the import and accumulation of photosynthetic proteins, but not of most non‐photosynthetic proteins, and AtTOC33 is most strongly expressed in young, light‐grown seedlings (Kubis et al, 2003). Toc159, a GTP‐dependent molecular motor that drives translocation, is also required for precursor protein recognition. The Toc159 subunits are encoded by four genes in Arabidopsis: AtTOC159, AtTOC132, AtTOC120 and AtTOC90. The atToc159 mutant is albino and does not survive past the cotyledon stage, implying that the other Toc159 family members cannot compensate for this defect (Bauer et al, 2000). Furthermore, overexpression of AtTOC159 is unable to complement the pale green atToc132 atToc120 phenotype (Kubis et al, 2004). Together, these findings imply that each Toc159 isoform exhibits substrate selectivity. Expression of such different isoforms may provide an efficient strategy for enhancing the rate of photosynthetic protein import over that of non‐photosynthetic proteins during early chloroplast development.

Thylakoid membranes are rich in galactolipids, which are synthesized in the chloroplast envelope membranes (Kelly and Dörmann, 2004), and galactolipid biosynthesis is essential for thylakoid formation (Kobayashi et al, 2007). Proplastids contain a limited amount of internal membranes, called prothylakoids, which form the starting point for the biogenesis of bona fide thylakoids. Many of the enzymes in the later stages of carotenoid and chlorophyll biosynthesis are also present on the plastid envelope, as these lipid‐soluble pigments must be incorporated into light‐harvesting chlorophyll (Lhc)‐binding proteins that are being inserted into the inner envelope membrane as a continuation of the protein import process (Hoober et al, 2007). These hydrophobic components must reach the prothylakoids by crossing the aqueous stroma. Several lines of evidence suggest that vesicles bud from the inner envelope membrane, most likely carrying a cargo of chlorophyll, enzymes and photosynthetic proteins, and migrate across the stroma to fuse with the developing thylakoids. First, when leaves are cooled to 12°C, vesicle‐like structures contiguous with the inner envelope membrane accumulate in the chloroplast stroma (Morre et al, 1991). Second, direct connections between the inner envelope and thylakoid membranes have been reported, implying that the two compartments represent a partly contiguous, dynamic continuum (Shimoni et al, 2005). Third, the vipp1 mutant is defective in thylakoid formation and does not form cold‐induced vesicles (Kroll et al, 2001; Aseeva et al, 2007). Another mutant, thf1, exhibits a variegated phenotype, and affected chloroplasts contain profuse vesicles with no thylakoid membrane (Wang et al, 2004). VIPP1 is associated with both the thylakoids and inner envelope, whereas THF1 is found in the stroma and thylakoids; the presence of two suborganellar locations is consistent with a trafficking function for these proteins. Finally, chloroplast bioinformatics has revealed the presence of homologues of small GTPases with putative membrane fusion functions similar to those in the eukaryotic secretory pathway, such as ARF1 and Sar1 (Andersson and Sandelius, 2004). Recently, a dynamin‐like GTPase called FZL has been identified that specifically affects thylakoid membrane structure in Arabidopsis. Again, FZL is localized to both the inner envelope and the thylakoid membranes (Gao et al, 2006). Although fzl mutant plants are not deficient in thylakoid formation per se, fzl chloroplasts are large and unusually shaped, they contain abnormal proportions of stromal and granal lamellae and they frequently accumulate small vesicles (Gao et al, 2006). These findings imply that FZL is a membrane‐remodelling factor that is required for maintaining a dynamic thylakoid network, but the basis for abnormal chloroplast division is unclear.

Once chloroplast biogenesis is underway, the chloroplasts must proliferate to match cell division and expansion: Arabidopsis mesophyll (M) cells can contain over 100 individual chloroplasts and the final count is tightly correlated with cell size (Pyke and Leech, 1994). The molecular nature of chloroplast division has been covered extensively in recent reviews (Maple and Moller, 2007; Yang et al, 2008), but one particular development is worth discussing here. As leaf development progresses, chloroplasts become progressively larger and dumb‐bell‐shaped plastids become less common, suggesting that division occurs early in chloroplast biogenesis (Pyke, 1999; Okazaki et al, 2009). Chloroplasts divide by binary fission, driven by two contractile protein rings that form on each side of the chloroplast envelope. The inner division ring forms first and is composed of the FtsZ1 and FtsZ2 proteins, which are homologous to bacterial fission proteins (Osteryoung and McAndrew, 2001). The constituents of the outer ring are not fully known, but the plastid division1 (PDV1) and PDV2 proteins in the outer envelope membrane recruit a cytosolic dynamin‐like component, DRP5B, around the chloroplast exterior in alignment with the inner ring (Miyagishima et al, 2006). It has recently been shown that PDV1 and PDV2 are determinants of the rate and extent of chloroplast division, a question that has remained open for some time (Okazaki et al, 2009). pdv1 and pdv2 mutants had earlier been shown to contain large, deformed chloroplasts (Miyagishima et al, 2006), but when both PDV1 and PDV2 are overexpressed together, Arabidopsis M cells contain small chloroplasts that are twice as numerous as in wild type (Okazaki et al, 2009). PDV promoter activity is highest around the meristem, in which proplastids are differentiating into chloroplasts. Crucially, the levels of PDV protein decrease in concert with the rates of chloroplast division as leaves aged, but FtsZ2 and DRPB5 levels remain at similar levels throughout development (Okazaki et al, 2009), tying in neatly with observed developmental patterns of chloroplast division and size. Constitutive expression of the cytokinin responsive transcription factor CRF2 and application of exogenous cytokinins both increase the activity of PDV2, linking cell division and chloroplast division and implying that the PDV proteins are primary mechanistic components in determining the cell's chloroplast complement. This PDV‐dependent mechanism seems to be evolutionarily conserved, holding true in the moss Physcomitrella patens, in common with other components of the PDV machinery such as FtsZ (Okazaki et al, 2009).

It is clear that molecular‐genetic approaches have been incredibly powerful tools in establishing what events are critical for chloroplast biogenesis. A notable point is that mutations in genes required for any one particular molecular process, such as chloroplast RNA processing, severely hamper the establishment of photosynthetic competence in general, as many mutants are pale green, albino or even embryo lethal (Table I). As such, many chloroplast processes are in some way interdependent: for example, the light‐harvesting complex of PSII (LHCII) is comprised of several Lhc‐binding proteins, which are only imported into the chloroplast and properly folded in the presence of chlorophyll synthesized on the inner envelope membrane (Espineda et al, 1999; Reinbothe et al, 2006). Similarly, defects in lipid biosynthesis severely compromise chlorophyll levels and PET complex assembly because thylakoid membranes cannot be generated (Kroll et al, 2001; Kobayashi et al, 2007). Such tight mutual dependence requires exquisite co‐ordination between the chloroplast, in which the events are happening, and the nucleus, in which many of the protein components are encoded.

Among all the original gun mutants, gun1 is unique because it is the only one to respond similarly to both norflurazon, which inhibits carotenoid biosynthesis and induces photo‐oxidative damage, and lincomycin, an inhibitor of plastid protein synthesis (Gray et al, 2003; Nott et al, 2006). Double mutant analyses have shown that GUN1 and GUN2–GUN5 define two distinct, but partially redundant signalling pathways that regulate overlapping groups of nuclear genes (Mochizuki et al, 2001; Strand et al, 2003). GUN1 is, therefore, required for a second signal that is triggered by defects in plastid gene expression (PGE). GUN1 was recently identified as a plastid‐localized pentatricopeptide repeat protein that is associated with nucleoids, which are transcriptionally active complexes of plastid DNA, RNA and ribsomes (Koussevitzky et al, 2007). An abscisic acid‐insensitive mutant, abi4, also exhibits a gun phenotype, showing that ABI4 is a further component of the PGE pathway (Koussevitzky et al, 2007). ABI4 is a nuclear transcription factor that binds to a sequence adjacent to or overlapping the G‐box motif. GUN1 is also required for transmitting the ‘MgProtoIX’ signal described above, and for glucose‐mediated repression of photosynthetic gene expression. As such, GUN1 acts as an integrator of several signals within the plastid (Koussevitzky et al, 2007). A model has, therefore, been proposed in which GUN1 is a master switch that generates or transmits an unknown signal, which in turn induces ABI4 to bind to promoter sequences and block photosynthetic gene expression in the nucleus, perhaps by inhibiting access of transcription factors such as HY5 to the G‐box (Koussevitzky et al, 2007; Larkin and Ruckle, 2008). Key questions that remain are the mechanism by which GUN1 integrates multiple signals, one of which does not seem to be a chlorophyll intermediate after all, and the nature of the secondary signal that is transmitted subsequently.

It is clear that during photomorphogenesis, developing chloroplasts are subject to a combination of positive and negative signals resulting from light and plastid status. How might these conflicting signals be integrated into an appropriate gene expression response? It has recently been shown that light and plastid signals can modulate one another. In a genetic screen to identify new components in plastid signalling, Ruckle et al (2007) recovered four mutants with a subtle gun phenotype, all of which turned out to be cry1 mutant alleles. This is surprising, because cry1 is usually considered to be a positive regulator of Lhcb expression. An even more surprising result was that cry1 gun1 double mutants showed much stronger derepression of Lhcb when grown on lincomycin than either single mutant. This suggests that cry1 and GUN1 act synergistically to effect most, if not all, of Lhcb repression under blue light when chloroplast biogenesis is blocked, and that a plastid signal can convert cry1 from a positive into a negative regulator of Lhcb. This observation is true in blue light and white light, but not in red light, suggesting that when chloroplast biogenesis is blocked, maximum repression of Lhcb expression requires photo‐activated cry1. Consistent with the function of cry1 acting through COP1 to regulate HY5, cop1‐4 is epistatic to cry1 (i.e. cry1 cop1‐4 double mutants do not exhibit derepression of Lhcb expression when treated with lincomycin), and hy5 is a subtle gun mutant (Ruckle et al, 2007). Furthermore, a hy5 cry1 double mutant is indistinguishable from either single mutant in blue light, suggesting that both cry1 and HY5 operate in the same pathway. This implies that in healthy seedlings, cry1 acts through HY5 to promote Lhcb expression, but in the presence of dysfunctional plastids, a GUN1‐independent signal converts HY5 into a negative regulator (Larkin and Ruckle, 2008). This is in agreement with additional data suggesting that HY5 alone is insufficient to regulate transcription, and that HY5 can act as both a positive and a negative regulator (Lee et al, 2007).

Curiously, RbcS expression is not derepressed in lincomycin‐treated cry1 or cry1 gun1 mutants, suggesting that cry1 only induces RbcS expression and cannot repress it (Ruckle et al, 2007). Furthermore, under the conditions used by Ruckle et al (2007), HY5 does not induce RbcS in blue light because hy5 mutants accumulate similar levels of RbcS transcripts as wild type. This implies the existence of another cry1/COP1‐regulated transcription factor that does promote RbcS transcription. However, HY5 can still repress RbcS when chloroplast biogenesis is blocked as it does for Lhcb, provided the GUN1‐mediated plastid signal is inactivated (i.e. in a gun1 mutant background) (Ruckle et al, 2007). This observation fits neatly with the model in which the GUN1 signal induces ABI4 to bind immediately upstream of the G‐box of photosynthesis‐related genes, thus preventing access by HY5 (Koussevitzky et al, 2007). There is also good evidence that phyB contributes to the repression of Lhcb, but not RbcS, when chloroplast development is blocked (Ruckle et al, 2007). Overall, these data imply that genes such as Lhcb and RbcS are regulated by complex and distinct mechanisms, incorporating a GUN1‐independent plastid signal that can convert positive regulators into negative ones when plastids are damaged. Importantly, it is clear that this signalling network is crucial for efficient chloroplast biogenesis. gun1, cry1 and hy5 mutants are all more susceptible than wild type to photo‐oxidative damage induced by high intensity light, with gun1 cry1 and gun1 hy5 double mutants being particularly badly affected (Ruckle et al, 2007). Finally, plastid signals dependent on cry1 and GUN1 influence several other aspects of seedling development, including anthocyanin biosynthesis, cotyledon expansion and inhibition of hypocotyl elongation (Ruckle and Larkin, 2009).

In epigeous seedlings, germination takes place beneath the soil surface, and hypocotyl elongation pushes the cotyledons into the light. The cotyledons initially act as storage organs to support seedling growth, and only later become photoautotrophic. Although true leaves are generated post‐embryonically from the shoot apex, development of the cotyledons largely takes place during embryogenesis when tissue types and growth axes are specified (Aida et al, 1999; Stoynova‐Bakalova et al, 2004). Likewise, the chloroplasts in leaves develop from meristematic proplastids as the leaf primordia emerge, but chloroplasts in cotyledons develop from etioplasts that are already present in M tissue within the embryo. These etioplasts are primed for rapid conversion to chloroplasts on light exposure. The prolamellar body, a crystalline agglomeration of Pchlide, Pchlide oxidoreductase and fragments of prothylakoid membranes, provides the structural framework for the incipient photosynthetic apparatus (Sundqvist and Dahlin, 1997). The etioplast state does not normally occur in leaves because the shoot apex and leaf primordia are routinely exposed to light. Consistent with their different cytological origins, several lines of genetic evidence suggest that chloroplasts in cotyledons develop through distinct mechanisms from those in leaves.

Genetic screens have revealed mutants in which the greening of cotyledons and leaves are differentially affected. The snowy cotyledon (sco) mutants exhibit pale green or white cotyledons, but normal, green leaves. In sco1‐1 mutants, germination is delayed and the cotyledons are initially white, but occasionally the cotyledons eventually are green and the seedlings survive if provided with exogenous sucrose (Albrecht et al, 2006; Ruppel and Hangarter, 2007). Immature Arabidopsis embryos contain photosynthetically active chloroplasts (Ruuska et al, 2004), but during seed dehydration and maturation, the chlorophyll and thylakoid membranes are lost, resulting in white embryos. Immature green sco1‐1 embryos dissected from siliques germinate normally and do not exhibit white cotyledons, suggesting that sco1‐1 mutants are defective in the re‐greening process (Ruppel and Hangarter, 2007). However, it is unknown if etioplast development is normal in sco1‐1 mutants, so the defect could occur before or during seed maturation. Indeed, the sco1‐2 and sco1‐3 null alleles confer embryo lethality, suggesting that SCO1 is essential for early embryogenesis as well as de‐etiolation (Ruppel and Hangarter, 2007). SCO1 encodes a chloroplast‐localized elongation factor G (EF‐G) and is thought to bind to the ribosomal complex to support plastid translation in a manner similar to that in Escherichia coli (Mohr et al, 2002). The Arabidopsis genome encodes just two other EF‐Gs, which appear to be weakly expressed compared with SCO1 and may be dual targeted to the plastid and mitochondrion (Ruppel and Hangarter, 2007), and it is possible that these EF‐Gs support plastid translation in other tissue types. The sco1 embryo‐lethal phenotype emphasizes the importance of plastid translation on all stages of plant development (Ahlert et al, 2003).

Another mutant specifically defective in cotyledon chloroplast biogenesis is cyo1. This mutant has recently been characterized in detail and is allelic to sco2 (Shimada et al, 2007; Albrecht et al, 2008). Unlike sco1, null cyo1 mutants complete embryogenesis and germinate normally, and etioplasts resemble those of wild type (Shimada et al, 2007). However, similar to sco1, precocious germination of cyo1 mutants rescues the pale green cotyledon phenotype (Albrecht et al, 2008). Thus, the defect is specific to the generation of chloroplasts from etioplasts in cotyledons. The CYO1 protein includes a predicted zinc‐finger domain similar to E. coli DnaJ, possesses disulphide isomerase activity and is localized to the thylakoid membrane (Shimada et al, 2007). CYO1 may be required for the folding of cysteine‐rich thylakoid‐resident proteins, such as those comprising the photosystems, during de‐etiolation when rapid arrangement of the photosynthetic apparatus is critical. Presumably, this function is either unnecessary in leaf chloroplasts or is performed by another unidentified chaperone. It is notable that DnaJ‐like proteins have been shown to have several functions in plastid biogenesis: in maize, BSD2 is required for the assembly of the Rubisco holoenzyme in bundle sheath (BS) cells (Brutnell et al, 1999), and in cauliflower, the gain‐of‐function or mutation triggers the accumulation of carotenoids in the plastids of otherwise colourless tissues (Lu et al, 2006).

The plastid genome is transcribed by both a nuclear‐encoded RNA polymerase and a plastid‐encoded, eubacterial‐like RNA polymerase (PEP). Promoter specificity of PEP is mediated by nuclear‐encoded sigma factors, of which there are six in Arabidopsis (Allison, 2000). Of these, SIG2 and SIG6 are thought to have a specific function in the de‐etiolation of cotyledon chloroplasts. Although the cotyledons of SIG2 antisense plants are chlorophyll deficient, the leaves are dark green (Privat et al, 2003); similarly, sig6‐1 null mutants exhibit delayed greening of cotyledons, but otherwise normal leaves (Ishizaki et al, 2005). In vitro, SIG2 strongly binds to rbcL and PsbA promoters, whereas SIG1 does not, even though both sigma factors are expressed in young seedlings (Privat et al, 2003). This observation suggests that differential promoter recognition may be partly responsible for different paths of chloroplast biogenesis. The degree of redundancy in the SIG family has yet to be determined; complementation tests, promoter swaps and analysis of mutant combinations will help elucidate tissue specificities and functional divergence.

In C₄ plants such as maize, photosynthesis is spatially divided between two cell types in the leaf: the M and the BS. Carbon is initially fixed in M cells before being shuttled to the BS cells in which the Calvin cycle operates. The chloroplasts of M and BS cells of maize are morphologically and biochemically distinct: M cell chloroplasts contain grana, accumulate PSII and lack starch, whereas BS cell chloroplasts lack grana, accumulate Rubisco and contain multiple starch granules (Nelson and Langdale, 1992). A proteomic analysis of the stromal proteins of each chloroplast subtype has revealed in detail the degree to which each chloroplast type is specialized (Majeran et al, 2005). One of the more interesting findings of this study was that homologues within gene families are differentially expressed in each chloroplast type, consistent with models predicting functional divergence after selection at gene regulatory regions (Sage, 2004). Nevertheless, relatively few maize mutants have been characterized with defects specific to M and BS cells. One is BS defective2, which lacks the BSD2 DnaJ‐like protein discussed above and, therefore, exhibits perturbed BS cell chloroplasts (Brutnell et al, 1999). Another mutation, golden2 (g2), also leads to aberrant BS chloroplasts, but in this case the mutated gene encodes an Myb family transcription factor that now defines a family of golden2‐like (GLK) genes present in diverse groups of land plants (Rossini et al, 2001; Yasumura et al, 2005). Each species contains at least two GLK genes, and in Arabidopsis and moss, each gene acts largely redundantly to promote nuclear photosynthetic gene expression in all photosynthetic cell types (Fitter et al, 2002; Yasumura et al, 2005; Waters et al, 2009). This redundancy also seems to be true in rice, as a pale green phenotype only results once the activity of both GLK genes is knocked down (P Wang and JAL, 2009, unpublished data). In maize, however, g2 specifically disrupts photosynthetic development in BS cells, leaving M cell chloroplasts unaffected (Langdale and Kidner, 1994). Accordingly, G2 is expressed in BS cells, whereas its homologue ZmGLK1 is expressed most strongly in M cells (Rossini et al, 2001). This has led to the intriguing speculation that, as transcriptional activators, G2 and ZmGLK1 each has a central function in the series of events that leads to BS and M cell differentiation. However, attempts to recover a zmglk1 mutant to test this hypothesis have been unfruitful, which may indicate an essential function for ZmGLK1 in early photosynthetic development of the maize seedling.

A recently characterized maize mutant that is specifically deficient in M cell function has provided some insight into how differential gene expression across the two cell types might be achieved. Zmhcf136 is a seedling lethal mutation, which leads to loss of PSII activity and the absence of grana in M cell chloroplasts, whereas those in BS cells are unaffected (Covshoff et al, 2008). This phenotype is consistent with the earlier assigned function of HCF136 in Arabidopsis as a PSII stability or assembly factor (Plücken et al, 2002). Accordingly, ZmHCF136 transcripts accumulate only in M cells. Analysis of BS and M transcriptomes revealed that each cell type responded independently to the Zmhcf136 mutation: generally, genes that are normally differentially expressed in the BS became less so, and M‐enriched transcripts became more abundant than in wild type (Covshoff et al, 2008). Such transcripts include those encoding the C₄ enzymes carbonic anhydrase (relatively more M‐enriched in Zmhcf136 compared with wild type) and phosphoenolpyruvate carboxykinase (relatively less BS‐enriched). In the light of the extensive feedback signalling between the plastid and nucleus as described above, Covshoff et al (2008) suggest that much of the differential gene expression associated with the C₄ state may result from modifications of the cell environment—such as plastid redox profiles and energy metabolite gradients—without invoking the need for extensive innovation at multiple gene regulatory regions. This might help explain how C₄ photosynthesis has evolved independently from the basal C₃ state at least 45 times in the angiosperms (Kellogg, 1999).