Synthetic conversion of leaf chloroplasts into carotenoid-rich plastids reveals mechanistic basis of natural chromoplast development

The first committed step of the carotenoid pathway is the conversion of geranylgeranyl diphosphate (GGPP) to phytoene, catalyzed by phytoene synthase (referred to as PSY in plants and crtB in bacteria). We previously found that the virus-mediated expression of a bacterial crtB gene in tobacco (Nicotiana tabacum and Nicotiana benthamiana), tomato (Solanum lycopersicum), Arabidopsis thaliana, and several other plants caused leaf yellowing due to increased accumulation of colored endogenous carotenoids downstream of phytoene (10). When the production of crtB was optimized using appropriate viral vectors for specific target hosts, an intense and widespread yellow phenotype was achieved in edible leaves such as those of lettuce (Lactuca sativa) (Fig. 1A) and green vegetables such as zucchini (Cucurbita pepo) (Fig. 1B). These results illustrate the potential of this approach to boost the nutritional value of green (i.e., chloroplast-containing) plant tissues, which are particularly recalcitrant to carotenoid enrichment. To further investigate the mechanism underlying the characteristic yellow phenotype of crtB-expressing tissues, we used transmission electron microscopy (TEM) to analyze plastid ultrastructure in leaves from N. tabacum and Arabidopsis plants treated with crtB-harboring viral vectors. Yellow sectors in infected leaves contained plastids with a distinctive ultrastructure that were absent in empty vector controls (Fig. 2 A and B). Very similar plastids were observed when the crtB gene was transiently expressed from Agrobacterium tumefaciens-delivered vectors in agroinfiltrated N. benthamiana leaves (Fig. 2C). These plastids were devoid of the organized photosynthetic thylakoids and grana found in typical chloroplasts but contained electron-dense (i.e., lipid-containing) membrane stacks much more tightly appressed than grana (Fig. 2D). The stacks were connected by what appeared to be remnants of thylakoid membranes. Plastids of crtB sectors also showed a proliferation of small electron-dense round vesicles identified as plastoglobules (Fig. 2). Loss of thylakoid and grana integrity as well as proliferation of plastoglobules and new membrane systems are features typically observed when chloroplasts differentiate into chromoplasts (3, 8, 11). TEM examination of dark-incubated senescent N. benthamiana leaves showed that the plastids found in crtB-producing cells were completely different from gerontoplasts (Fig. 2E).

To further substantiate the identity of the chromoplast-like plastids that developed in crtB-producing leaves, we analyzed chloroplast and chromoplast marker proteins by immunoblot analysis (Fig. 2F). Virus- or A. tumefaciens-mediated expression of crtB in N. tabacum or N. benthamiana leaves, respectively, resulted in increased levels of fibrillin, a protein associated with chromoplast development (12, 13). In contrast, the levels of D1 (also known as PsbA), a core component of photosystem II (PSII) that is highly down-regulated during chloroplast-to-chromoplast differentiation (14, 15), decreased in crtB-producing leaves (Fig. 2F). These results together suggest that expressing the bacterial crtB gene in leaf cells is sufficient to differentiate chloroplasts into chromoplast-like plastids.

We next used Arabidopsis double mutants defective in OR chaperones (AtOR and AtOR-LIKE) to test whether the differentiation process triggered by crtB involved pathways depending on these well-characterized promoters of chromoplast development (5, 16). Similarly, double mutants lacking cytosolic and plastidial carotenoid cleavage dioxygenases (CCD1 and CCD4, respectively) were used to investigate the possible contribution of signaling molecules derived from enzymatic degradation of carotenoids in the differentiation mechanism (17–19). Virus-mediated expression of crtB in these mutants resulted in leaves showing the characteristic yellow phenotype and carotenoid overaccumulation observed in the WT (Fig. 1C), suggesting that OR activity or enzymatic cleavage of carotenoids are not required for the process. Besides leaves, yellowing was widespread in all other green tissues, including cauline leaves, stems, sepals, and siliques, where it remained stable until plants died (Fig. 1D).

Low scores for plastidial targeting were obtained for crtB with TargetP and ChloroP servers, indicating a nonreliable prediction. However, the 12-aa-long noncatalytic N-terminal region of this protein contains some features found in plastid-targeting peptides, including five hydrophobic and two hydroxylated residues. A C-terminal fusion of the full-length crtB enzyme to the green fluorescent protein (crtB-GFP) mainly localized to the cytosol in N. benthamiana leaf cells, but some fluorescence was also detected in chloroplasts (10). By optimizing the transient expression conditions (using only young leaves and adding the helper component protease of the Watermelon mosaic virus to prevent gene silencing), crtB-GFP fluorescence was detected at a much higher intensity. This allowed to confirm that the crtB-GFP protein was present in both cytosol and chloroplasts of agroinfiltrated cells (SI Appendix, Fig. S1A). When GFP was fused to the N terminus of crtB, the resulting protein (GFP-crtB) was completely excluded from chloroplasts (SI Appendix, Fig. S1A). It is therefore likely that the bacterial crtB enzyme harbors a cryptic plastid-targeting signal in its N terminus that becomes blocked and, hence, inactivated in the GFP-crtB protein. To unambiguously target crtB to the chloroplast, we next added the plastid-targeting sequence of the Arabidopsis enzyme HDS (20) to the crtB-GFP reporter. As expected, the resulting (p)crtB-GFP protein was only found in chloroplasts (SI Appendix, Fig. S1A). Agroinfiltrated leaf tissues expressing either crtB or (p)crtB developed the characteristic yellow phenotype at 4–5 d postinoculation (dpi), whereas tissues expressing the cytosolic GFP-crtB version—renamed as (c)crtB—remained green as the controls expressing GFP (Fig. 3A). Analysis of carotenoid contents showed identical profiles for leaf sections agroinfiltrated with GFP and (c)crtB and confirmed that, similarly to crtB, (p)crtB triggered carotenoid overaccumulation (Fig. 3B). Agroinfiltration of N. benthamiana leaves with crtB or (p)crtB constructs did not initially reduce chlorophyll levels compared to leaf tissues agroinfiltrated with GFP or (c)crtB (Fig. 3B). However, estimation of photosynthesis-related parameters such as effective quantum yield of PSII (ɸPSII) and nonphotochemical quenching (NPQ) showed that both crtB and (p)crtB, but not (c)crtB or GFP, had a dramatic impact on chloroplast function (Fig. 3C). A plastid-targeted version of GFP did not cause any yellowing or ɸPSII defect (SI Appendix, Fig. S1B), confirming that disturbance of chloroplast photosynthesis is not caused by the accumulation of a foreign protein in chloroplasts but specifically by crtB. TEM analyses confirmed that (p)crtB induced the differentiation of chromoplast-like plastids very similar to those found in leaf tissues expressing the untargeted crtB enzyme, whereas only chloroplasts were present in leaves producing either (c)crtB or GFP (Fig. 2 C and D). These results confirm that crtB elicits a synthetic (i.e., nonnatural) differentiation of chromoplasts only when localized in plastids, where carotenoids are made.

Differentiated leaf chromoplasts (Fig. 2C) and associated features such as high carotenoid levels and low ɸPSII (Fig. 3D) were maintained for weeks after agroinfiltration, and no reversions to chloroplasts or green color were ever observed. About a week after agroinfiltration with crtB, chlorophylls started to decrease. Carotenoids also decreased but not as much as chlorophylls, eventually resulting in a higher carotenoid-to-chlorophyll ratio and, as a consequence, a stronger yellow color as leaves became older (Fig. 3D). The strong yellow phenotype associated with chromoplast differentiation and carotenoid enrichment was also stably maintained when the crtB gene was expressed from a viral vector in Arabidopsis, lettuce, and zucchini (Fig. 1), strongly suggesting that the mechanism by which crtB overexpression eventually results in chromoplastogenesis is conserved in plants.

The vast majority of plastidial proteins are encoded by nuclear genes (2). We therefore reasoned that the drastic remodeling of plastidial ultrastructure associated with crtB-triggered chromoplast differentiation would require changes in nuclear gene expression. RNA-sequencing (seq) analyses of N. benthamiana leaf samples at 96 h postinfiltration (hpi) showed that about 5,000 genes were differentially expressed in yellow (p)crtB sections compared to green GFP controls (Dataset S1). Such a massive reprogramming of gene expression included the up-regulation of 3,183 genes and the down-regulation of 1,803 genes in chromoplast-containing samples. Gene Ontology (GO) term enrichment analyses (Dataset S2) showed overrepresentation of genes involved in protein folding and binding to RNA and ribosomes among those induced by (p)crtB (SI Appendix, Fig. S2). Enrichment of genes with roles in transmembrane transport, cell signaling (protein phosphorylation, calcium binding), and nuclear gene expression (transcription factors) was observed among those repressed when chromoplast biogenesis was induced (SI Appendix, Fig. S2). This profile was strikingly similar to that of ripening tomato fruits (where chromoplasts naturally differentiate from chloroplasts) but very different from that of senescent Arabidopsis leaves (SI Appendix, Fig. S2). Many of the genes potentially involved in carotenoid biosynthesis were up-regulated during crtB-triggered transformation of leaf chloroplasts into chromoplasts (SI Appendix, Fig. S3). However, the changes in carotenoid-related gene expression detected during chromoplastogenesis in N. benthamiana leaves were modest compared to those taking place in tomato ripening fruit (SI Appendix, Fig. S3 and Dataset S3).

Energy and carbon required for carotenoid biosynthesis rely on photosynthesis (i.e., the Calvin–Benson cycle) in chloroplasts. Given that leaf chromoplast differentiation was associated with impairment of photosynthesis (Fig. 3), we asked whether primary cell metabolism might also be reprogrammed. Of 52 metabolites detected by gas chromatography/time-of-flight/mass spectrometry analysis in N. benthamiana leaves (Dataset S4), 13 displayed statistically significant changes in (p)crtB sections compared to GFP controls (SI Appendix, Fig. S4 and Table S1). We observed reductions in the levels of ascorbate and hexoses (glucose and fructose, the main soluble carbohydrate stores and respiration substrates). This, together with increments in tricarboxylic acid (TCA) cycle intermediates (citrate, 2-oxoglutarate, and malate) and amino acids (valine, isoleucine, aspartate, and glutamate), suggested that sugars were used to produce ATP through the TCA cycle to sustain amino acid and carotenoid biosynthesis (and likely other cellular functions). Indeed, respiration rate—determined as total oxygen consumption in the dark—was higher in chromoplast-containing leaf tissues (SI Appendix, Fig. S4). While an increased respiration is also associated with the onset of carotenoid overproduction in tomato and other climacteric fruits, the metabolic changes that we observed in (p)crtB-producing N. benthamiana leaves are often opposite to those occurring during chromoplast differentiation in tomato (11, 21). In particular, hexoses and ascorbate do not decrease but increase. We speculate that this might be because leaf metabolism is devoted to produce and export photoassimilates, whereas tomatoes are sink organs that have been selected to accumulate sugars and acids as positive taste attributes. In any case, our data together show that the activity of crtB in leaf chloroplasts is sufficient to trigger a deep reprogramming of nuclear gene expression and whole-cell metabolism associated with the differentiation of chromoplasts, a plastid type that is not naturally found in tobacco or Arabidopsis leaves.

To investigate the dynamics of the crtB-dependent chromoplast differentiation process, we next followed the time course of crtB expression, phytoene production, and downstream carotenoid accumulation after agroinfiltration of N. benthamiana leaves with the (p)crtB construct. Transcripts encoding (p)crtB were reliably detected at 24 hpi and peaked at 48 hpi (Fig. 4A), whereas phytoene started to accumulate between 24 and 36 hpi and suddenly increased at 48 hpi (Fig. 4B). Downstream carotenoids began to increase at 48 hpi (Fig. 4B). To follow chloroplast membrane remodeling dynamics, we also monitored ɸPSII, NPQ, and D1 protein levels as estimators of photosynthesis, photoprotection, and photodamage, respectively. Both ɸPSII (Fig. 4C) and NPQ (SI Appendix, Fig. S5) remained unchanged up to 36 hpi, and then decreased as the levels of both phytoene and downstream carotenoids increased (Fig. 4). The levels of D1 started to decrease later, between 48 and 60 hpi (Fig. 4D), likely as a result of photodamage. A higher temporal resolution analysis of both ɸPSII and carotenoid levels between 25 and 40 hpi showed that phytoene levels increased before ɸPSII decreased, whereas downstream carotenoids took a bit longer to accumulate (Fig. 4E). Taken together, these results suggest that the crtB-mediated production of phytoene causes a disruption of the chloroplast photosynthetic functionality before carotenoids start to overaccumulate.

To next confirm whether impairment of chloroplast functionality was due to phytoene overaccumulation, we used norflurazon (NF) to prevent phytoene conversion into downstream carotenoids (22). N. benthamiana leaves were agroinfiltrated with constructs to produce GFP, (p)crtB, and PAR1, an Arabidopsis transcription cofactor that promotes total carotenoid biosynthesis but not phytoene accumulation in photosynthetic tissues (23). At 24 hpi, some agroinfiltrated leaves were treated with NF (Fig. 5). Untreated leaves transiently expressing the Arabidopsis PAR1 gene accumulated higher levels of carotenoids downstream of phytoene but did not exhibit changes in ɸPSII compared to GFP controls. NF treatment resulted in phytoene accumulation, reduced levels of downstream carotenoids, and decreased ɸPSII in all of the samples (Fig. 5). The reduction in ɸPSII in these NF-treated samples correlated with the accumulation of phytoene (the higher the levels of phytoene, the stronger the reduction of ɸPSII). However, the reduction in ɸPSII was modest compared to that observed in (p)crtB samples lacking NF (Fig. 5A), in which phytoene conversion into downstream carotenoids led to the differentiation of chromoplasts. We concluded that phytoene overaccumulation by itself disrupts photosynthesis but does not abolish chloroplast identity, resulting in a relatively modest reduction in ɸPSII. The much stronger decrease in ɸPSII that takes place in (p)crtB samples would not be directly due to the effect of phytoene on photosynthetic activity but resulted from the dismantling of the photosynthetic apparatus as chloroplasts accumulate downstream carotenoids and differentiate into chromoplasts. Interestingly, PAR1-mediated accumulation of carotenoids but not phytoene (Fig. 5A) was unable to trigger the massive drop in ɸPSII and the yellow leaf phenotype characteristic of chromoplast differentiation (Fig. 5B). We therefore conclude that the accumulation of phytoene causes a concentration-dependent disruption of the photosynthetic identity of chloroplasts, a condition that might be necessary for them to become chromoplasts upon the subsequent production and accumulation of downstream carotenoids.

In contrast with the results using crtB but similar to those with PAR1, overexpression of PSY-encoding genes from Arabidopsis and tomato could not elicit the characteristic yellow leaf phenotype associated with chromoplast differentiation (Fig. 5 and SI Appendix, Fig. S6) (11, 24, 25). Interestingly, the plant enzymes yielded significantly lower levels of phytoene compared to crtB and did not substantially impact photosynthesis as deduced from ɸPSII values (Fig. 5 and SI Appendix, Fig. S6). We hence speculated that phytoene might act as a metabolic threshold switch that only alters the photosynthetic performance of chloroplasts when exceeding a certain level. To overcome this putative threshold, we treated N. benthamiana leaves with DCMU (3-(3,4-dichlorophenyl)-1,1-dimethylurea, diuron), a widely used inhibitor of photosynthesis that interrupts the photosynthetic electron transport chain. The next day, treated and untreated control leaves were agroinfiltrated with constructs encoding either GFP or the Arabidopsis PSY enzyme. At 96 hpi, DCMU-treated GFP leaf sections showed a decrease in ɸPSII but unchanged carotenoid levels compared to untreated samples (Fig. 5). By contrast, DCMU-treated PSY sections showed a more dramatic drop in ɸPSII and much higher levels of carotenoids than untreated PSY or GFP controls (Fig. 5A). As a consequence, PSY leaf sections treated with DCMU turned yellow (Fig. 5B), similar to that observed when chloroplast-to-chromoplast differentiation was triggered by crtB in the absence of inhibitor. In summary, our results are consistent with the existence of a two-step process responsible for the crtB-mediated transformation of leaf chloroplasts into chromoplasts (Fig. 6). First, chloroplast identity is weakened by overaccumulation of phytoene and, second, increased production of carotenoids in preconditioned chloroplasts allows the differentiation of chromoplasts. Without preconditioning, carotenoid levels can increase but chromoplasts do not differentiate, as shown in untreated leaves producing PAR1 or PSY. Moreover, if carotenoids downstream of phytoene are not produced, preconditioned chloroplasts remain unchanged, as shown in NF-treated (p)crtB leaves (Fig. 5).

N. tabacum Xanthi nc, N. benthamiana RDR6i, and A. thaliana Columbia-0 (Col) and Landsberg erecta (Ler) plants were grown under standard conditions as described previously (10, 40). Growth of double mutants ccd1 ccd4 (16) and ator atorl (17), both in the Col background, was facilitated by transferring them to low light (40 μmol photons·m⁻²·s⁻¹) days after germination. For generating dark-induced leaf senescence, detached leaves were maintained inside dark, humid chambers until visible yellowing occurred. Collected plant material was frozen in liquid nitrogen, lyophilized, and then homogenized to a fine powder using a TissueLyser system (Qiagen) for further analyses.

Transmission electron microscopy of plant leaves was performed as previously described (41).

See SI Appendix, Materials and Methods for details.

For transient expression studies using viral vectors, leaves of 4- to 6-wk-old N. tabacum and Arabidopsis plants were mechanically inoculated with crude extracts from frozen-stored infected plant tissue and collected upon the appearance of the yellowing phenotype as described previously (10). For agroinfiltration experiments, the second or third youngest leaves of 4- to 6-wk-old N. benthamiana plants were infiltrated with A. tumefaciens strain GV3101 carrying vectors of interest following the procedure described previously (42). Gene silencing was prevented by coagroinfiltration with a strain carrying the helper component protease (HC-Pro) of the Watermelon mosaic virus (WMV) in vector HcProWMV-pGWB702 (kindly provided by Juan José López-Moya and Maria Luisa Domingo-Calap, CRAG, Barcelona, Spain). Infiltration cultures were grown on Luria–Bertani medium at 28 °C and used at optical density at 600 nm (OD₆₀₀) of 0.5. For pharmacological treatments, NF or diuron (DCMU) were diluted in water and 0.05% Tween 20. The treatments with NF (2 µM) were performed by infiltration with a syringe of leaf areas that had been agroinfiltrated with different constructs 24 h earlier. DCMU (10 µM) was applied on the leaf surface with a fine paintbrush 24 h before agroinfiltration.

See SI Appendix, Materials and Methods for details.

Protein extraction, quantification, and immunoblot analyses were performed as described (43) using anti-fibrillin (44) or anti-PsbA serum (Agrisera).

Carotenoids were analyzed as previously described (40). Phytoene was quantified using a concentration curve with a commercial standard (Sigma). Primary metabolites were extracted and analyzed as described in SI Appendix, Materials and Methods.

See SI Appendix, Materials and Methods for details.

Differentially expressed genes (DEGs) were identified by comparing crtB and GFP RNA-seq datasets with the DESeq2 statistical method in the AIR platform. The resulting crtB/GFP list was filtered using cut-offs of false discovery rate <0.05 and log₂-transformed fold change (log₂FC) >0.585 for up-regulated genes and <−0.599 for down-regulated genes. GO enrichments were identified by the Parametric Analysis of Gene Set Enrichment (PAGE) function of the AgriGO v2.0 web-based tool (bioinfo.cau.edu.cn/agriGO/) after transforming the gene IDs to the Niben v04.4 annotation. For the comparison of different biological systems, we selected the significantly enriched gene ontologies from our N. benthamiana RNA-seq experiment (P and q values <0.05) and compared their Z-score values with those obtained from the analysis of published RNA-seq data of tomato fruit ripening and Arabidopsis leaf senescence. In particular, we used the reads per million mapped reads values of the total pericarp at mature green (MG), light ripe (LR), and red ripe (RR) stages (45) and the fragments per kilobase million mapped reads values of the 16D and 30D senescence stages (46). DEGs resulting from LR/MG, RR/MG, and 30D/16D comparisons were filtered as described above for N. benthamiana crtB/GFP. Student’s t tests were used for the rest of statistical analyses using GraphPad Prism 5.0a (GraphPad Software). Carotenoid-related genes in N. benthamiana were identified by BLASTp using tomato sequences. Heatmaps were made using the pheatmap package in R (https://cran.r-project.org/web/packages/pheatmap/index.html).