• Open Access

A synthetic de-greening gene circuit provides a reporting system that is remotely detectable and has a re-set capacity

Authors


* Correspondence (fax: 970-491-0649; e-mail: June.Medford@colostate.edu)

Summary

Plants have evolved elegant mechanisms to continuously sense and respond to their environment, suggesting that these properties can be adapted to make inexpensive and widely used biological monitors, or sentinels, for human threats. For a plant to be a sentinel, a reporting system is needed for large areas and widespread monitoring. The reporter or readout mechanism must be easily detectable, allow remote monitoring and provide a re-set capacity; all current gene reporting technologies fall short of these requirements. Chlorophyll is one of the best-recognized plant pigments with an already well-developed remote imaging technology. However, chlorophyll is very abundant, with levels regulated by both genetic and environmental factors. We designed a synthetic de-greening circuit that produced rapid chlorophyll loss on perception of a specific input. With induction of the de-greening circuit, changes were remotely detected within 2 h. Analyses of multiple de-greening circuits suggested that the de-greening circuit functioned, in part, via light-dependent damage to photosystem cores and the production of reactive oxygen species. Within 24–48 h of induction, an easily recognized white phenotype resulted. Microarray analysis showed that the synthetic de-greening initiated a process largely distinct from normal chlorophyll loss in senescence. Remarkably, synthetically de-greened white plants re-greened after removal of the inducer, providing the first easily re-settable reporter system for plants and the capacity to make re-settable biosensors. Our results showed that the de-greening circuit allowed chlorophyll to be employed as a simple but powerful reporter system useful for widespread areas.

Introduction

It is currently impractical to monitor large civilian areas, such as cities, transportation hubs and sports arenas, for surprise chemical/biological attacks or the accidental release of pollutants (Huppert, 2004). Hence, there is an urgent need for simple and inexpensive biological and chemical monitors. Because plants have evolved elegant mechanisms to continuously sense and respond to their environment, they could make inexpensive and widely used biological monitors, or sentinels, for human threats. For plants to function as sentinels in cities or large areas, two traits must be developed: a sensing (or input) system and a readout (or output) system. Our laboratory is working on developing specific sensing systems in plants using computationally designed receptors (Looger et al., 2003). In plants, typical readout systems include the reporter genes green fluorescent protein (GFP) and its variants, β-glucuronidase (GUS), luciferase or anthocyanin accumulation (Lloyd et al., 1992). However, to monitor large areas, such as cities or shopping malls, a distinct reporter or readout system is needed. A plant sentinel readout system must include four key characteristics: (i) ease of recognition by the public; (ii) a means for remote detection; (iii) a rapid response; and (iv) a re-set capacity so that multiple or repeated threats can be detected. No currently available reporter system satisfies these requirements.

The green pigment chlorophyll has many of these characteristics. Changes in chlorophyll are easily recognized by the public because of the foliage colours displayed by plants each autumn. Chlorophyll fluorescence can be measured remotely with either satellites or inexpensive, hand-held instruments, and has been used to collect information about the physiological state of field-grown plants (Carter and Knapp, 2001; Zarco-Tejada et al., 2002; West et al., 2003). Chlorophyll systems have responses so sensitive and rapid that a simple cloud moving between the sun and the surface of the earth leads to a change in chlorophyll fluorescence (Nedbal and Brezina, 2002; Nedbal et al., 2003). However, developing chlorophyll as a readout or reporter system presents considerable challenges because chlorophyll is an abundant pigment, with levels regulated by both genetic and environmental factors.

Chlorophyll loss in plants is normally a slow process that typically occurs during the complex mechanism of senescence. The half-life of chlorophyll has been estimated to be 2–5 days for relatively mature and fully greened leaves (Stobart and Hendry, 1984), but visual perception of chlorophyll loss in leaves can take longer. For example, a time of 6–9 days was necessary for a photobleaching phenotype to become evident in Arabidopsis plants when phytoene desaturase, an enzyme involved in carotenoid biosynthesis, was inhibited by RNA interference (RNAi) (Guo et al., 2003). A substantially faster loss is needed if chlorophyll is to be used as a reporter system for a plant sentinel.

Steady-state chlorophyll levels result from the combined effects of synthesis and breakdown. Because chlorophyll catabolites have never been found in photosynthetically active leaves, the chlorophyll degradation pathway found in senescing tissues is thought to be distinct from that involved in maintaining steady-state levels (Matile et al., 1999). Constitutive over-expression of genes encoding two distinct enzymes in the senescence-related chlorophyll degradation pathway has been reported (Mach et al., 2001; Benedetti and Arruda, 2002). Plants over-expressing chlorophyllase (CHLASE) showed substantial accumulation of the enzyme's product, chlorophyllide, indicating that the enzyme was functional (Benedetti and Arruda, 2002), but the plants still retained a green colour. These results suggest that manipulation of the genes involved in the chlorophyll degradation pathway alone will not allow the use of chlorophyll as a plant sentinel reporter. One explanation for the retention of the green colour in these transgenic lines is that plants compensate for degradation by enhancing biosynthesis. If so, chlorophyll levels might be subjected to rapid manipulation by co-regulating chlorophyll biosynthesis and degradation.

Synthetic biology aims to create novel behaviour and to provide insight into biological processes by designing ‘gene circuits’ with specific functions (McDaniel and Weiss, 2005). To remove chlorophyll from genetic and environmental input, we designed a synthetic ‘de-greening circuit’ so that chlorophyll biosynthesis and degradation could be simultaneously regulated in response to a specific input. Our test input system used the detection of a steroid hormone, transcriptionally linked to the simultaneous inhibition of chlorophyll biosynthetic genes and induction of chlorophyll breakdown genes. We showed that the de-greening circuit uncoupled chlorophyll levels from genetic and environmental input, producing rapid chlorophyll loss. Changes were remotely detected within 2 h, and easily recognizable white plants were produced within 24–48 h. The synthetic de-greening process was largely independent of normal chlorophyll loss in senescence, allowing plants to re-green and providing the needed re-set capacity in a biological sensor.

Results

Assembly and testing of de-greening gene circuits

In many biological responses, sensing of a specific substance leads to a transcriptional response. The synthetic sensing system that we are developing for a plant sentinel links input to transcriptional output (Looger et al., 2003); hence, we sought to develop our readout system triggered by a transcriptional response. Numerous transcriptional induction systems are available which can provide a surrogate means to test the chlorophyll reporter system. We modified a synthetically designed, steroid inducible system to function in plants. In the presence of a synthetic steroid (4-hydroxytamoxifen, 4-OHT), a chimeric transcriptional regulator relocates to the nucleus and induces the expression of a promoter made up of specific response elements and the −46 region of the cauliflower mosaic virus (CaMV) 35S promoter, designated 10XN1P (see ‘Experimental procedures’). When we started this work, few inducible transcriptional systems were available for plants. However, the 4-OHT induction system is essentially analogous to other inducible transcriptional systems (Zuo et al., 2000), and complete details will be described elsewhere (manuscript in preparation).

We hypothesized that rapid de-greening would require the inhibition of chlorophyll synthesis concurrent with the induction of degradation. To test this hypothesis, we first assembled each function separately under the control of our transcriptionally inducible system. Two potential means to inhibit chlorophyll synthesis (‘stop synthesis’ circuits) were used: directly, through its biosynthetic pathway, and indirectly, through a precursor/trafficking pathway. The rate-limiting enzyme in chlorophyll biosynthesis is NADPH:protochlorophyllide oxidoreductase. In Arabidopsis, this enzyme is encoded by three POR genes (PORA, PORB, PORC) (Oosawa et al., 2000; Frick et al., 2003; Masuda et al., 2003). POR is a light-dependent enzyme that catalyses the conversion of protochlorophyllide a to chlorophyllide a. GUN4, GENOMES UNCOUPLED 4, is a single copy gene that functions in regulating chlorophyll biosynthesis and precursor trafficking, and may have a role in photoprotection (Larkin et al., 2003; Verdecia et al., 2005). GUN4 has been shown to bind and activate magnesium chelatase, an enzyme complex that produces magnesium protoporphyrin IX. A diRNA (double-stranded interfering RNA) construct was prepared to a conserved POR gene region, with its expression placed under the control of 10XN1P; the construct was introduced into Arabidopsis plants. A diRNA construct to GUN4 was also prepared, and likewise placed under the control of the 10XN1P promoter. Figure 1(a) shows that, even after 48 h of induction of either the POR or GUN4 diRNA constructs (‘stop synthesis’), the plants remained largely green, with only older leaves showing some chlorophyll loss. Wild-type Columbia plants used as a control showed only minimal chlorophyll loss with and without the 4-OHT inducer (Figure 1c).

Figure 1.

Regulatory circuits for de-greening. (a) Stop synthesis circuit; induction of diRNA to protochlorophyllide oxidoreductase (POR) or diRNA to GENOMES UNCOUPLED 4 (GUN4). (b) Initiate breakdown circuit; induction of chlorophyllase (CHLASE) and red chlorophyll catabolite reductase (RCCR) or CHLASE and pheophorbide a oxygenase (PAO). Panels show appearance of plants at 0, 24 and 48 h after induction. (c) Control plants. Wild-type Columbia plants (w.t.) with and without the inducer at 0, 24 and 48 h.

Genes involved in key steps in chlorophyll breakdown have also been identified (Eckhardt et al., 2004). Chlorophyll breakdown involves a series of enzymatic steps, with key processes being hydrophobic tail removal by CHLASE (Tsuchiya et al., 1999; Benedetti and Arruda, 2002), porphyrin ring cleavage by pheophorbide a oxygenase (PAO), closely followed by red chlorophyll catabolite reductase (RCCR) (Wuthrich et al., 2000; Pruzinska et al., 2003, 2005). Two distinct gene circuits designed to initiate chlorophyll degradation were assembled (‘initiate breakdown’ circuits), containing CHLASE and either PAO or RCCR under the control of the 10XN1P promoter. Figure 1(b) shows that, after 48 h of induction of the chlorophyll ‘initiate breakdown’ gene circuit, the plants also remained green, even with over-expression of the two major genes involved in chlorophyll metabolism. These results are consistent with our hypothesis that plants are able to partially compensate for changes in chlorophyll levels by regulating overall metabolism.

We combined the ‘stop synthesis’ constructs with the ‘initiate breakdown’ genes in one T-DNA to test whether a rapid, regulated chlorophyll loss system could be developed. The constructs/genes were brought together in various combinations to produce five different ‘de-greening gene circuits’ (Table 1). Each ‘de-greening circuit’ consisted of an inducible diRNA to either the POR or GUN4 gene, CHLASE inducible expression, combined with PAO and/or RCCR inducible expression. De-greening circuits were also obtained by crossing plants containing the separate gene circuits (e.g. plants from Figure 1a,b), and produced comparable results (data not shown).

Table 1.  Genes used in the construction of each of the complete de-greening circuits
De-greening circuitStop synthesisInitiate breakdown
  1. CHLASE, chlorophyllase; GUN4, GENOMES UNCOUPLED 4; PAO, pheophorbide a oxygenase; POR, protochlorophyllide oxidoreductase; RCCR, red chlorophyll catabolite reductase.

1PORCHLASE, RCCR
2GUN4CHLASE, RCCR
3PORCHLASE, PAO
4GUN4CHLASE, PAO
5GUN4CHLASE, RCCR, PAO

Figure 2 shows that induction of the de-greening circuits caused a loss of chlorophyll from all regions of the plant, with less effect early on in the shoot apex. Chlorophyll loss was extensive, with plants becoming white within 24–48 h after induction (Figure 2). In senescing Arabidopsis, loss of chlorophyll typically reveals yellow carotenoid pigments; the white phenotype observed after induction of the synthetic de-greening circuits suggests that carotenoid pigments were also lost.

Figure 2.

De-greening circuits. Left panel shows the appearance of two plants from each de-greening circuit (numbers 1–5) at 0, 24 and 48 h after induction. Right panel shows plants induced to fully de-green that re-greened for 3 days. None of the transgenic lines containing the de-greening circuit or its parts showed an accelerated cell death phenotype, which has been reported in plants that accumulate intermediates in the chlorophyll breakdown pathway.

A notable feature of chlorophyll loss by the de-greening circuits was that the rational design of the synthetic circuit (simultaneous regulation of biosynthesis and breakdown) led to a similar white phenotype regardless of the specific gene combination. Only a slight difference was seen between stopping synthesis with diRNA to POR and diRNA to GUN4 (compare circuits #1 and #2, or #3 and #4). Inclusion of the RCCR gene with CHLASE to initiate breakdown produced slightly better chlorophyll reduction than breakdown initiation involving PAO and CHLASE (compare circuits #1 and #2 with #3 and #4). De-greening circuit #5 differed from the others as it contained the PAO and RCCR (as well as CHLASE) genes to initiate breakdown. Plants containing circuit #5 lost chlorophyll within 24 h; however, the plants were light green prior to induction.

We verified the changes in the expression of de-greening circuit genes using semi-quantitative reverse transcriptase-polymerase chain reaction (RT-PCR) (Figure 3). As predicted, induction of diRNA constructs led to a decline in the respective mRNAs: POR mRNA declined in plants with circuits #1 and #3; GUN4 mRNAs declined in circuit #2 and, to a lesser extent, in circuits #4 and #5. mRNAs for genes that were induced were found to increase dramatically: CHLASE and RCCR increased in plants containing circuit #1 or #2. For plants containing circuit #3 or #4, CHLASE levels increased and PAO induction showed a strong increase in circuit #3, with a more modest induction in plants containing circuit #4. Plants containing circuit #5, which were light green prior to induction, appeared to have lost the regulation of the CHLASE and RCCR genes. The level of cyclophilin mRNA was used as a control and, in some lines, the level declined, probably reflecting the decrease in total RNA at 24 and 48 h of induction (Figure S1, see ‘Supplementary material’). Consequently, the RT-PCR results probably under-estimate the decrease/increase in the levels of de-greening circuit gene mRNAs. This was confirmed by the analysis of CHLASE in circuit #3; RT-PCR results indicated a threefold induction, whereas microarray analysis showed an induction of more than eightfold (see below).

Figure 3.

Semi-quantitative reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of de-greening circuit transcripts at 0, 24 and 48 h after induction and after re-greening for 3 days (R). CHLASE, chlorophyllase; CYC, cyclophilins used as a sample loading control; GUN4, GENOMES UNCOUPLED 4; –ind, samples incubated in the absence of the inducer (10 µm 4-hydroxytamoxifen); +ind, samples incubated in the presence of the inducer (10 µm 4-hydroxytamoxifen); PAO, pheophorbide a oxygenase; POR, protochlorophyllide oxidoreductase; RCCR, red chlorophyll catabolite reductase.

To determine whether decreased levels of POR and GUN4 transcripts produced a corresponding decrease in proteins, Western blots using antibodies to POR and GUN4 were performed. Induction of the de-greening circuits led to a dramatic decrease in POR and GUN4 proteins, with levels decreasing by 70%−90% within 24 h (Figure S2, see ‘Supplementary material’).

Re-greening of plants

A re-set capacity is an essential feature in both a plant sentinel for terrorist threats (to allow multiple or repeated threats to be detected) and a plant environmental monitor (for long-term pollutant monitoring). The use of the de-greening circuit as a reporter system provides an endogenous means to re-set the system. Plants that had lost their chlorophyll from induction of the de-greening circuit re-developed their green colour, or re-greened, after the inducer was removed (Figure 2). The re-greening process was enhanced with cytokinin treatment, providing a simple and readily available means to re-set the reporter system. Furthermore, re-greened plants could be induced to de-green again (data not shown). Re-greening was most apparent in rapidly expanding leaves. However, partially de-greened leaves and tissues were also capable of re-greening. Older leaves that completely de-greened early in the process, and lost turgor, did not re-green. In re-greening plants, the levels of POR and GUN4 transcripts increased, whereas the levels of CHLASE, PAO and RCCR mRNAs decreased to levels approaching normal (Figure 3).

Effect of light on de-greening

Because light plays an essential role in chlorophyll metabolism, we determined whether light was required for synthetic de-greening. Plants grown under different light intensities (50–350 µE/m2/s) and different temperatures (14 to 30 °C) were not affected in their ability to de-green (data not shown). However, when plants were grown under standard light conditions and were induced to de-green in complete darkness, the effects of the de-greening circuit were muted (Figure 4a). After 24, 48 or 72 h of induction in complete darkness, the plants eventually became paler, but were still a light green, indicating that light was required for rapid chlorophyll loss. If plants induced in complete darkness were subsequently transferred to the light, de-greening proceeded at an enhanced rate.

Figure 4.

Rapid loss of chlorophyll after induction of the de-greening circuit requires light in both whole plants and detached leaves. (a) Effect of light on de-greening of whole plants. Fourteen-day-old plants containing de-greening circuit #3 were induced to de-green by placing the leaves on Murashige–Skoog (MS) medium without sucrose, supplemented with 10 µm 4-hydroxytamoxifen, and transferred to complete darkness. After 24, 48 or 72 h, samples were removed from the dark, photographed and transferred to normal light conditions (approximately 100 µE/m2/s; 16 h light/8 h dark). Plants were re-photographed after 24, 48 or 72 h in the light. Dark panels, plants placed in complete darkness. Arrows indicate samples moved to light. (b) De-greening of detached leaves in the light from plants containing different circuits. Top panels, leaves at 0 h of induction; bottom panels, leaves at 48 h of induction; C, Columbia control plants; 1, 2 and 3, specific de-greening circuits. (c) Effect of light on de-greening of detached leaves. Leaves were excised from 14-day-old plants containing circuit #2, and the de-greening circuit was induced as described above. Left panels show leaves incubated in the light and right panels show leaves incubated in the dark; right bottom panel shows leaves incubated in the dark after they had been transferred to light. Numbers next to panels indicate the time of incubation in hours.

In Arabidopsis, detached leaves and whole plants show distinct responses to dark-induced senescence (Weaver and Amasino, 2001). Darkness is known to induce senescence in detached leaves but not in whole plants. To determine whether the de-greening circuit functioned similarly to senescence, we induced de-greening in detached leaves. As expected, under standard light conditions, de-greening induction caused detached leaves to fully de-green within 48 h (Figure 4b). However, darkness by itself failed to induce full de-greening in detached leaves even after 72 h (Figure 4c). We conclude that the de-greening circuit induces a pathway that is distinct from chlorophyll loss in senescence.

Ultrastructural changes during de-greening

Because the de-greening circuit produced chlorophyll loss in an unprecedented timeframe, we investigated how cell and plastid structures were affected. Ultrastructural changes in plastids and cells that take place during normal senescence-based chlorophyll loss proceed in a predictable and orderly manner (Biswal et al., 2003). We examined the ultrastructure of uninduced plants containing the de-greening circuit and found that the cellular and chloroplast ultrastructure appeared normal (Figure 5a). In transgenic plants containing de-greening circuit #3, 24 h after induction, nuclei, mitochondria and vacuoles appeared normal, but the chloroplast showed significant ultrastructural disruptions (Figure 5b–h). A gradient of severity was observed, in which chloroplasts in cells closest to the outside of the leaf showed the most severe disruptions, whereas chloroplasts in more interior cells were least disrupted. Yet, even the least disrupted chloroplasts were swollen and had disoriented thylakoid membranes (Figure 5b). In more severely affected plastids, thylakoid membranes separated and dilated, but the nuclei and mitochondria appeared normal (Figure 5c). Chloroplasts with more advanced effects had severely dilated thylakoids, such that small vesicles ( 0.1–0.2 µm) were formed (Figure 5e,f). Disrupted thylakoid vesicles (Figure 5g) and small plastoglobuli were seen. In some cases, cells containing relatively normal chloroplasts were found adjacent to cells with chloroplasts in which the exterior plastid membrane had disintegrated (Figure 5h). Disintegration of the exterior chloroplast membrane was more common after 48 h, with the observation of plastid membrane remnants and small plastoglobuli (Figure 5I–k). Remarkably, some dilated thylakoid membranes remained even after the exterior chloroplast membrane had disintegrated (Figure 5j). With very severe cases, cell membranes and vacuoles were disrupted leaving only cellular debris (Figure 5k). In re-greening plants, chloroplasts with normal grana and stroma were found (Figure 5l). These chloroplasts could arise from proplastid differentiation and/or from recovery of damaged plastids.

Figure 5.

Transmission electron micrographs showing ultrastructural changes in mesophyll cells of plants containing de-greening circuit #3. (a) Uninduced control. Cells and chloroplast appear normal with thylakoids organized into grana and stroma. Plants after 24 h of induction (b–h), after 48 h of induction (i–k) and after 3 days of re-greening (l). (b) Normal nucleus adjacent to chloroplast showing incipient effects, i.e. slight enlargement of the plastid and some disorientation of thylakoid membranes (arrowhead). (c) Cell showing chloroplast with thylakoid membranes beginning to dilate (arrowhead) with normal-appearing nucleus and mitochondria. (d) Close-up showing dilation and separation of stroma and grana thylakoid membranes (arrowhead). (e, f) Chloroplasts showing more advanced thylakoid dilation to the point that small vesicles are formed (arrowhead). Chloroplast envelope and other cell organelles appear normal. (g) Close-up showing dilation, vesicle formation and disruption of thylakoid membranes (arrowheads) with small plastoglobuli appearing as small dark-staining punctuate elements. (h) Juxtaposition of a cell with a disrupted chloroplast (arrowhead) to a cell with chloroplasts still containing organized grana and stroma at 24 h of induction. (i) Disrupted chloroplasts with remnants of thylakoid membranes and large vesicles; large number of plastoglobuli are also present. (j) Close-up of disrupted chloroplasts with plastoglobuli and remnants of thylakoid membranes. (k) Cells with disrupted chloroplast; other cells contain debris and lack identifiable organelles. (l) Re-greening plants have chloroplast with normal grana and stroma. CE, chloroplast envelope; CW, cell wall; G, grana; IC, intact chloroplast; M, mitochondria; N, nucleus; PG, plastoglobuli; S, stroma. Bar, 1 µm.

Remote measurement of chlorophyll fluorescence

Induced chlorophyll loss provided an easily recognizable phenotype (white plants), distinct from stressed plants, and a re-set capacity. The two remaining characteristics needed for a readout system in a plant sentinel or wide area reporter are rapid response and remote imaging capacity.

Light energy absorbed by chlorophyll follows several competing paths: photosynthetic electron transport heat dissipation through the xanthophyll cycle, re-emission as fluorescence or formation of triplet chlorophyll (Maxwell and Johnson, 2000). Because the pathways are competing, and the sum of the rate constants is unvarying, information about the plant's physiological state can be inferred by remote analysis of chlorophyll fluorescence (Zarco-Tejada et al., 2002; West et al., 2003). In addition, because plants must quickly respond to changing light conditions, chlorophyll responses are rapid.

To determine whether chlorophyll fluorescence measurements would provide rapid diagnostic and remote imaging capacities to the de-greening circuit system, a commercially available chlorophyll monitor and software (FluorCam, Photon Systems Instruments, Brno, Czech Republic) were used to measure how chlorophyll fluorescence behaved before and after induction of the de-greening circuit. Figure 6 depicts the changes in three parameters derived from chlorophyll fluorescence measurements in control (Columbia) and transgenic plants containing circuit #1. Control plants, both induced and uninduced, showed only slight changes in all parameters measured. Fv/Fm measures the maximum quantum efficiency of photosystem II (PSII) in dark-adapted plants, and is often used as a measurement of plant stress (Maxwell and Johnson, 2000). Uninduced plants containing the de-greening circuit had an initial Fv/Fm value of 0.8 (Figure 6a). An Fv/Fm value of 0.8 is close to the maximum possible PSII efficiency observed in normal, non-stressed plants (Adams et al., 1990). These results indicate that there are few physiological effects of the de-greening circuit in uninduced plants.

Figure 6.

Remote monitoring of de-greening circuit by effects on chlorophyll fluorescence. (a) Time course changes of total chlorophyll and maximum chlorophyll fluorescence in dark-adapted plants (Fm) (top panel), operating efficiency of photosystem II (PSII) (ΦPSII) (middle panel) and maximum quantum efficiency of PSII (Fv/Fm) (bottom panel) in transgenic plants containing de-greening circuit #1 and in wild-type control (Columbia). Full blue line with filled squares, control plants; broken orange line with open squares, induced plants; broken green line with open triangles, total chlorophyll levels of induced plants. F.W., fresh weight. (b) Spatial changes in Fm, ΦPSII and Fv/Fm in plants containing de-greening circuit #1. Plants were incubated in liquid Murashige–Skoog (MS) medium in the presence (induced) or absence (control) of 10 µm 4-hydroxytamoxifen to induce the de-greening circuit. Images of plants were false coloured to depict the spatial changes in intensity of each of the parameters measured or calculated. Scale is to the right of each panel, where the blue colour represents the lowest values and the red colour represents the highest values.

With induction of the de-greening circuit, all measured parameters showed substantial changes. An initial decrease in Fv/Fm was seen within 2 h, and more substantial reductions were observed at 6 h. Fm is a measure of maximum chlorophyll fluorescence and an indirect measurement of total chlorophyll in dark-adapted plants. After de-greening circuit induction, Fm showed some initial variability, with stable reductions seen after 12 h (Figure 6). These initial variations in Fm were similar to those found when total chlorophyll levels were measured spectrophotometrically (Figure 6a; Figure S3, see ‘Supplementary material’), and may reflect mechanisms in the plant to attempt to compensate for enhanced chlorophyll degradation.

One of the most robust parameters derived from chlorophyll fluorescence measurements is ΦPSII, representing the portion of light absorbed by chlorophyll in PSII used in photochemistry. ΦPSII can also be measured in light-grown plants under natural conditions (Maxwell and Johnson, 2000). Figure 6 shows that plants induced to de-green exhibited a large (nearly 60%) decrease in ΦPSII within 2 h, the first time point measured. Other de-greening circuits showed similar changes in ΦPSII and Fv/Fm (Figure S4, see ‘Supplementary material’). Unlike ΦPSII, Fv/Fm showed more initial variability (Figure 6a), indicating that ΦPSII is a more precise parameter for following the initial induction of the de-greening circuit. Spatial analysis showed that the decrease in ΦPSII was seen throughout the plant, with the effects seen last in the shoot apex (Figure 6b). This was consistent with the reversal of the process, re-greening, which was first evident in the shoot apex (Figure 2). The spatial changes in Fv/Fm were similar to those of ΦPSII (Figure 6b). These results show that measurement of ΦPSII provides a straightforward means to rapidly and remotely detect induction of the de-greening circuit reporting system.

Global changes in gene expression during de-greening are distinct from those of senescence

The de-greening circuit caused chlorophyll loss in an unprecedented time, with response of leaves to dark treatment and ultrastructural aspects distinct from those of normal chlorophyll loss observed during senescence. To further test the hypothesis that the de-greening circuit initiates a synthetic process distinct from normal chlorophyll loss in senescence, we employed microarray analysis. If the de-greening circuit initiates a synthetic process, we predicted that the genes involved in senescence should be relatively unaffected, whereas the genes encoding photosystem components might be affected. Figure 7 shows plots of the number of genes grouped by annotated function that were up- or down-regulated at each time point tested (24 h, 48 h and re-greening). The majority of genes known to be up-regulated during senescence were not induced by the de-greening circuit (Figure 7; Table S1, see ‘Supplementary material’). Genes that were strongly induced during senescence, but not during de-greening, included enzymes for degradation of macromolecules (e.g. proteases, nucleases, lipases), transcription factors, kinases/phosphatases, defence-related genes and flavonoid/anthocyanin biosynthetic genes (Buchanan-Wollaston et al., 2005). Within certain categories, small subsets of genes were commonly induced by both senescence and de-greening, including chaperones, redox, autophagy and alkaloid-like biosynthetic genes. Genes down-regulated in senescence were likewise largely not down-regulated by the de-greening circuit. The prominent exception, in which down-regulated senescence genes were down-regulated in de-greened plants, included most of the nuclear encoded components of the photosynthetic machinery, e.g. PSII, PSI and Calvin cycle genes. The majority of annotated PSII- and PSI-related genes (e.g. light-harvesting chlorophyll a/b binding and oxygen-evolving complex proteins) were down-regulated within 24 h of induction, the first time point measured (Table S1). This included PSII subunits (PSO1, PSO2, PSP1, PSP2, PSQ1, PSQ2) and PSI subunit precursors (Kieselbach and Schroder, 2003). Other notable down-regulated genes included DegP2, encoding a protease responsible for the initial repair of damaged PSII proteins (Haussuhl et al., 2001), and FtsH6, a chloroplast light-harvesting complex II (LHCII) protease (Zelisko et al., 2005). After 48 h, these genes were also down-regulated. Also notable amongst the group down-regulated by the de-greening circuit was PsbS (NPQ4), which allows excessive energy to be dissipated by photosynthetic antennae through the xanthophyll cycle (Li et al., 2004; Holt et al., 2005). Although most photosystem genes were down-regulated, a few known photosystem genes were found to be up-regulated, including PsbP and genes encoding lumen proteins of unknown function (e.g. At1g03600, At3g09490). In addition, microarray analysis showed that genes involved in the production of reactive oxygen species (ROS) were up-regulated by the de-greening circuit (Table S1). Genes involved in antioxidative processes (Mittler et al., 2004) and in the detoxification of products of lipid peroxidation (Loeffler et al., 2005) were induced by de-greening (redox regulation and oxidative stress), including glutathione transferase, microsomal glutathione S-transferase, type 2 peroxiredoxin, NADP-dependent oxidoreductase, glutaredoxin, thioredoxin, peroxiredoxin, alternative oxidase, ferritin, blue copper protein, glutathione peroxidase and several other peroxidases.

Figure 7.

Transcriptome responses to de-greening and re-greening. Genes are classified by annotated function. Transgenic plants containing de-greening circuit #3 were used for induction. All samples were compared with transgenic plants without induction. Black bars, transgenic plants induced for 24 h; light grey bars, transgenic plants induced for 48 h; dark grey bars, plants de-greened for 48 h and re-greened for 3 days.

De-greening induces ROS

The effects detected by chlorophyll fluorescence measurements, the requirement for light and the differences in global gene expression patterns from those of normal chlorophyll loss suggested that the de-greening circuit brought about rapid chlorophyll loss in a unique manner. The loss of other pigments, such as carotenoids, the rapid change in ΦPSII and the induced genes suggested that ROS were produced with induction of the de-greening circuit. To test whether this was the case, we examined induced and uninduced plants for ROS generation. Figure 8 shows that the induction of de-greening caused extensive ROS production. ROS were found throughout the mesophyll, with detection first noted at 8 h.

Figure 8.

Reactive oxygen species (ROS) (green) accumulate in the mesophyll cells of plants containing de-greening circuit #3 (30 h after the induction of de-greening). Chloroplasts (red) in de-greening (induced) plants have an abnormal shape.

Discussion

In order to use plants to monitor large areas for environmental contaminants or terrorist agents, a reporter or readout system is needed. Current gene reporter systems were developed for laboratory use and do not provide the characteristics needed for a plant sentinel. We developed a synthetic de-greening circuit that allows the green pigment chlorophyll to be used as a biosensor readout system. Induction of the de-greening circuit allows remote detection, displays a rapid response, provides a re-set capacity, and results in a phenotype readily recognized by the general public. Because the de-greening circuit produces a white phenotype, it is easy to distinguish it from plants stressed from biotic or abiotic conditions, which produce yellow phenotypes via senescence-related pathways. The inability to re-set biosensors has been the major limitation to their use. The de-greening circuit provides a simple capacity to be re-set. Plants re-greened after removal of the inducer, and re-greening could be enhanced by a brief cytokinin treatment. Because the transcriptional inducer used (4-OHT) is relatively stable, the de-greening circuit may not fully switch to an ‘off’ position immediately following removal of the inducer, and the re-greening process may not start until the inducer within the plant degrades. Hence, it should be possible to substantially reduce the time needed for re-greening, which is currently 3 days.

It should also be possible to reduce the response time from less than 2 h to minutes. Our initial time point at 2 h detected a substantial decrease in ΦPSII, one of the most robust parameters in chlorophyll imaging. The rapid decrease seen at 2 h suggests that we may be able to detect changes earlier; indeed, detailed statistical analysis of the fluorescence parameters (L. Nedbal, personal communication) should allow us to accurately determine when the first significant changes can be detected. In addition to improving remote detection ability, the genetic circuitry could be further enhanced by rationally applying principles developed for synthetic gene circuitry (McDaniel and Weiss, 2005). For example, gene circuitry could be designed to be activated and remain active with a single exposure to a small amount of inducer (4-OHT or via a sensing pathway) (Looger et al., 2003).

The de-greening circuit, combining ‘stop synthesis’ with an ‘initiate breakdown’ function, caused the loss of chlorophyll in an unprecedented timeframe. When each function was introduced separately, plants did not visibly de-green in the 48-h timeframe, except in the cotyledons. The fact that expression of the ‘initiate degradation’ circuits (CHLASE and PAO, or CHLASE and RCCR) failed to produce rapid de-greening suggests that plants have a means to enhance chlorophyll biosynthesis when needed. Conversely, the fact that the ‘stop synthesis’ circuits (diRNA to POR or GUN4) likewise failed to produce rapid de-greening supports the concept of a large amount of stable chlorophyll being present within the plant (Eckhardt et al., 2004). The rational combination of these two functions in one T-DNA produced a synthetic ‘de-greening circuit’. The synthetic nature of the designed gene circuit is indicated by three types of data: (i) the response of excised leaves to dark-induced senescence; (ii) the distinctive ultrastructural changes; and (iii) the microarray data showing a difference in the genes regulated by the de-greening circuit and by normal chlorophyll loss in senescence.

Light was shown to be important for the occurrence of the rapid de-greening process, as induced plants incubated in the dark failed to turn white, even after 72 h of induction. When induced plants were transferred to light, de-greening proceeded at an enhanced rate (Figure 4). These results suggest that the de-greening circuit is poised to respond in darkness, but not able to initiate rapid de-greening without light. Chlorophyll biosynthesis and breakdown intermediates are potentially phototoxic (Matile et al., 1999). Because the de-greening circuit interferes with the normal balance of chlorophyll, and probably its intermediates, it is possible that, with light exposure, these molecules cause photo-oxidation. A similar light requirement for de-greening was observed for detached leaves. Under standard light conditions, de-greening induction caused detached leaves to fully de-green within 48 h. However, darkness failed to induce full de-greening in detached leaves, even after 72 h of induction. Because darkness is known to induce senescence in detached leaves of Arabidopsis (Weaver and Amasino, 2001), these results support the difference between chlorophyll loss from the de-greening circuit and that from senescence.

Because light was required for rapid chlorophyll loss, we looked at how light was handled by plants induced to de-green. The use of remote chlorophyll fluorescence measurements provides an easy detection system for plant sentinels. ΦPSII measures the proportion of light absorbed by chlorophyll associated with PSII that is used in photochemistry, and is an indication of overall photosynthesis (Maxwell and Johnson, 2000). De-greening plants showed a rapid decline in ΦPSII within 2 h of induction. Because ΦPSII is the most robust parameter obtained from chlorophyll fluorescence measurements, it provides a simple means for rapidly detecting induction of the de-greening circuit. Moreover, the disruption provides an insight into how the de-greening circuit is functioning (below). Fv/Fm, a widely used parameter for assessing a plant's level of stress, had an initial value of 0.8 in uninduced plants, indicating that uninduced plants were not stressed. Induction of the de-greening circuit caused a decrease in Fv/Fm values, when compared with controls, also providing an indication that the photosynthetic ability of de-greening plants was disrupted.

A decrease in Fv/Fm typically results from a combination of two processes: an increase in the rate constant for thermal dissipation and/or a decrease in the rate constant for photochemistry. Because light is required and the plants lose chlorophyll and yellow pigments, one possibility is that excitation energy is dissipated from chlorophyll by interaction with xanthophylls. De-excitation of chlorophyll has recently been shown to occur by a rapidly reversible electron exchange between chlorophyll and zeaxanthin (Holt et al., 2005). If chlorophyll dissipates energy through zeaxanthin, we predict a substantial change in non-photochemical quenching (NPQ). However, although NPQ measurements changed with induction, these changes varied in time and between the de-greening circuits (Figure S5, see ‘Supplementary material’); thus, NPQ is not the primary means through which the de-greening circuit functions. Hence, the decrease seen in Fv/Fm is not primarily caused by enhanced thermal dissipation. Because ΦPSII and Fv/Fm decreased prior to a substantial decrease in chlorophyll level, our data suggest that the de-greening circuit functions by the inactivation or removal of PSII cores, which precedes substantial removal of chlorophyll. If the de-greening circuit functions through the inactivation or removal of PSII cores, there should be a large production of ROS, which was observed (Figure 8). A mechanism proceeding via the action of ROS on photosystem cores would also account for the observation that de-greening circuits with varying gene compositions all produce a similar phenotype. The various de-greening circuits all produced a similar phenotype and changes in ΦPSII and Fv/Fm. Collectively, these data lead to the hypothesis that the rapid changes observed with induction of the de-greening circuit proceed via damage to the photosystems, resulting in the generation of ROS. In support of this hypothesis, DegP2, encoding a protease that is responsible for the initial repair of damaged PSII proteins (Haussuhl et al., 2001), was down-regulated. In addition, FtsH6, a chloroplast LHCII protease (Zelisko et al., 2005), was likewise down-regulated. Further analysis of microarray data suggested that various PSII- and PSI-related genes were down-regulated, whereas ROS-related genes were simultaneously up-regulated, indicating a process largely distinct from normal chlorophyll loss in senescence.

Collectively, these data indicate that the de-greening circuit provides a synthetic means to control chlorophyll levels in plants. The trigger for the de-greening circuit is transcriptional input. The steroid inducible 10XN1P promoter used here could be replaced with other promoter elements, allowing the readout system to be linked to various inputs.

Native plants have been used to monitor airborne fluorides and explosives (Jefferson, 1993). Specific inputs could be combined with the de-greening circuit, such as for the monitoring of metals or phosphorus (Hammond et al., 2003; Krizek et al., 2003). If input could be provided by designed receptors (Looger et al., 2003), the sensitivity and specificity would be substantially enhanced.

Experimental procedures

Plant materials, growth conditions and transgenic plant production

Arabidopsis thaliana, ecotype Columbia (Col-0), plants were used for analysis and transgenic plant production. Standard growing conditions were as follows: 25 °C and approximately 100 µE/m2/s light in either a Percival AR75L growth chamber (25° ± 1 °C) or light shelf (25° ± 2 °C); day/night cycle (16 h light, 8 h dark); Metro Mix 200 growth medium (Scotts, Marysville, OH, USA), supplemented with Miracle Grow fertilizer. Transgenic plants were produced using the floral dip method (Clough and Bent, 1998). Plasmids were assembled as described below, transferred into Agrobacterium tumefaciens strain GV3101 by electroporation, and the selected Agrobacterium was used for Arabidopsis transformation. Transgenic lines were selected on Murashige and Skoog (MS) medium (Murashige and Skoog, 1962) containing 50 mg/L kanamycin or 5 mg/L glufosinate (Crescent Chemical Co., Islandia, NY, USA), depending on the vector, and 100 mg/L cefotaxime. T0 lines were selected and allowed to self-pollinate, and only lines segregating for one T-DNA insert (3 : 1), and not showing any obvious phenotypic lesion, were analysed further. When possible, homozygous transgenic lines were obtained and used for analysis (de-greening circuits #1, #2 and #3). With two de-greening circuits (#4 and #5), homozygous lines could not be obtained, or the homozygous lines set seed too poorly for detailed analysis. For plants containing these circuits, heterozygous lines were used.

Chemicals and enzymes

MS salts, vitamins and 4-OHT were all purchased from Sigma-Aldrich (St. Louis, MO, USA). 4-OHT stock solutions (5 mm) were prepared in ethanol and stored at −20 °C. Enzymes used for cloning and DNA amplification were obtained from commercial suppliers and were used according to the manufacturer's instructions.

Plasmid construction

To test and develop the de-greening gene circuits, a chemically inducible transcription system was assembled for plants using the synthetic zinc finger proteins and DNA binding elements described by Barbas and colleagues (Beerli et al., 2000; Segal et al., 2003). This provides a steroid-regulated transcription system in plants that is very similar to the oestrogen inducible systems described previously (Zuo et al., 2000); the use of these components in plants has been described previously (Stege et al., 2002). A complete characterization of this inducible transcription system will be described elsewhere. The regulatory elements for the inducible system were first assembled. A steroid binding chimeric transcription factor (NEV) was placed under the control of a strong constitutive promoter (figwort mosaic virus, FMV) (Sanger et al., 1990; Bhattacharyya et al., 2002), translation was enhanced with the omega enhancer at the 5′ end (Gallie et al., 1987), and expression was terminated with the 3′ end of the nopaline synthase (Nos) gene. NEV is a fusion protein, containing the synthetic N1 zinc finger DNA binding domain (N), an oestrogen receptor domain (E) and four copies of the herpes simplex virus (VP16) transcriptional activation domain (V) (Beerli et al., 2000). In animal cells, the NEV protein translocates to the nucleus and binds the synthetic N1 DNA element in the presence of the inducer. To make this system functional in plant systems, we synthesized 10 copies of the N1 DNA element upstream of the minimal −46 CaMV 35S promoter (Odell et al., 1985) (Egea Biosciences, Inc., San Diego, CA, USA), creating a 223-bp promoter fragment referred to as 10XN1P. Both the chimeric transcription factor (FMV::Ω-NEV-Nos) and the synthetic N1 promoter (10XN1P) were cloned into the plant transformation vectors pCAMBIA2300 and pMLBART.

To produce a synthetic de-greening circuit, genes regulating chlorophyll biosynthesis and breakdown were placed under the control of the inducible 10XN1P promoter and introduced into plant transformation vectors. Because multiple genes (up to four) were introduced, a transcription block was placed between each gene, or the genes were placed in a manner so that expression would not interfere (Padidam and Cao, 2001). The complete de-greening circuit combines two types of genetic circuit: a gene circuit to inhibit the biosynthesis of new chlorophyll and a gene circuit to stimulate chlorophyll breakdown. All chlorophyll regulatory genes in the different constructs were placed under the control of the 10XN1P promoter, allowing coordinated expression with addition of the 4-OHT inducer.

To inhibit the biosynthesis of new chlorophyll, diRNA constructs were assembled to the GUN4 gene (Larkin et al., 2003) or a conserved region found in the POR gene (Oosawa et al., 2000; Masuda et al., 2003). To generate the diRNA constructs, POR sense-intron-POR antisense or GUN4 sense-intron-GUN4 antisense was cloned into pBluescript KS(+), including an approximately 500-bp GFP coding sequence used as intron sequence at NotI/EcoRV. POR sense/GUN4 sense was cloned in via SacI/NotI and POR antisense/GUN4 antisense was cloned in via EcoRV/XhoI (XmaI/XhoI for the GUN4 diRNA construct), and then subcloned downstream of 10XN1P in p2300-FMV::Ω-NEV-Nos-10XN1P at the AvrII/MluI sites. An octopine synthase (ocs) terminator was added at the MluI site, resulting in p2300-FMV::Ω-NEV-Nos-10XN1P::POR diRNA-ocs or p2300-FMV::Ω-NEV-Nos-10XN1P::GUN4 diRNA-ocs.

The gene circuit to initiate the breakdown of chlorophyll contains the CHLASE (Tsuchiya et al., 1999) gene to remove chlorophyll's hydrophobic tail, and a gene or gene(s) to open the porphyrin ring, RCCR (Wuthrich et al., 2000) and/or PAO (Pruzinska et al., 2003). Chlorophyll degradation genes were added as a combination of either CHLASE and RCCR, or CHLASE and PAO, or, in its final form, as CHLASE, PAO and RCCR. These gene combinations, all under the control of the 10XN1P promoter, were first assembled in pBluescript KS(+). CHLASE was placed downstream of 10XN1P via BstXI/NotI, and the Nos terminator was added at the SpeI/SmaI sites; RCCR was fused with 10XN1P via PstI/EcoRI, and the Nos terminator (EcoRI/ApaI). Transcription blocks were included between the Nos terminator and 10XN1P via SmaI/PstI. The RCCR gene was replaced with PAO using the same restriction sites described above to generate the CHLASE–PAO combination. Assembled 10XN1P::CHLASE-Nos-TB-10XN1P:: RCCR-Nos or 10XN1P::CHLASE-Nos-TB-10XN1P::PAO-Nos was added to the 3′ end of either POR or GUN4 diRNA in p2300-FMV::Ω-NEV-Nos-10XN1P::POR diRNA-ocs (or p2300-FMV::Ω-NEV-Nos-10XN1P::GUN4 diRNA-ocs), as an ApaI fragment, to complete the de-greening circuits. The ‘initiate breakdown’ constructs (CHLASE + RCCR and CHLASE + PAO) were assembled by first PCR amplifying FMV::Ω-NEV-Nos-TB from p2300-FMV::Ω-NEV-Nos-TB using primers FMVFwd (5′-ATTTAGCAGCATTCCAGATTGGGTTC-3′) and TBRev (5′-AGAGAAATGTTCTGGCACCTGCACTTG-3′). The PCR product was cloned as a blunt fragment into SpeI-digested and Klenow-treated pART27-based vector pMLBART (Gleave, 1992), resulting in pMLBART-FMV::Ω-NEV-Nos-TB. The CHLASE and RCCR genes, as well as the CHLASE and PAO genes, all under the control of the 10XN1P promoter, were excised from vectors containing de-greening circuits #1 and #3, respectively, as ApaI fragments. After flushing the ends, these fragments were ligated into NotI-digested, Klenow-treated pMLBART-FMV::Ω-NEV-Nos-TB. All gene fusions and all chlorophyll regulatory genes were verified by sequencing (Macrogen Inc., Seoul, South Korea) before final assembly into pCAMBIA2300 and pMLBART vectors.

Induction of plant de-greening and re-greening

For the induction of de-greening, 14-day-old transgenic plants containing a single copy of the specific gene circuit were grown aseptically on MS medium with 50 mg/L kanamycin, but without sucrose. Individual plants were incubated in 24-well Cellstar culture plates (Greiner Bio-one, Longwood, FL, USA), each well containing 2 mL of liquid MS medium without sucrose, supplemented (induced) or not (control) with the inducer, 10 µm 4-OHT. Induction of the de-greening circuit typically started 1 h into the 16-h light period, but results were the same regardless of what time of the day the induction was started. Plants were returned to the growth conditions described above and incubation was continued for the different time periods or the conditions described in the text were applied. After de-greening, plants were induced to re-green by incubation in 1 µm t-zeatin for 6 h, and then transferred to plates containing MS medium, and allowed to re-green for up to 7 days. Plants re-greened without use of the cytokinin; however, cytokinin treatment enhanced the process.

Semi-quantitative RT-PCR analysis

Total RNA was isolated from whole plants using the Aurum™ Total RNA Mini Kit (Bio-Rad Laboratories, Hercules, CA, USA), according to the manufacturer's instructions. cDNA synthesis and PCR amplification were performed with 200 ng of total RNA, using the AccessQuick™ RT-PCR System (Promega, Madison, WI, USA) and gene-specific primers. Although the RT-PCR system uses a DNase step, primers were designed to span an intron–exon junction, except for GUN4 which lacks introns. The following primers were used: cyclophilins, 5′-GCGTTCCCTAAGGTATACTTCGAC-3′ and 5′-CCCATGAGAACACACACCAAAC-3′; GUN4, 5′-ACGCAAAATCTGGTTAAAAGTGAA-3′ and 5′-TTGTGAGCGGTAAGTGTCCTAAAG-3′; POR, 5′-TTGACCATCAAGGAACAGAGAA-3′ and 5′-TATTTGTGTTTCCTGTTATAGA-3′; CHLASE, 5′-TAGCCCCACAGTTGTGCAAATT-3′ and 5′-AAGTCCGTTGGTGCGCATGGTG-3′; RCCR, 5′-AATCTTCTCCGATTGATTTTGT-3′ and 5′-CTAGAGAACACCGAAAGCTTCT-3′; PAO, 5′-TCTATGAACAAAATTGAGTTAG-3′ and 5′-CTACTCGATTTCAGAATGTACA-3′. PCR amplification consisted of one cycle at 95 °C for 2 min, followed by 25 cycles each of 95 °C for 40 s, 52 °C for 30 s and 72 °C for 1 min, and a final extension step at 72 °C for 5 min. Amplification products were separated on 2% agarose gels, and photographed under UV light using a Scion Image Capture System (Frederick, MD).

Total chlorophyll and protein measurements

Chlorophyll was extracted in 2.5 mm sodium phosphate-buffered (pH 7.8) 80% acetone. The absorbance of the resulting solution was measured at 646.6 and 663.6 nm on a Shimadzu UV-1201 spectrophotometer (Columbia, MD), and the total chlorophyll content (µg/mL) was calculated using the formula: 17.76A646.6 + 7.34A663.6 (Porra et al., 1989). Total proteins were estimated using the Bradford reagent (Bio-Rad Laboratories) using bovine serum albumin (BSA) as a standard (Bradford, 1976).

Western blot analysis

Plants (three to five per time point) were ground in liquid nitrogen and resuspended in 250 µL of 12.5 mm sodium phosphate buffer (pH 7.8). Protein concentration was determined using the Bradford reagent (Bradford, 1976). Protein samples were loaded, on the basis of equal fresh weight, on 12% sodium dodecylsulphate-polyacrylamide gel electrophoresis (SDS-PAGE) gels, electrophoresed for 40 min at 200 V, and transferred to Hybond-P membranes (Amersham Biosciences, Piscataway, NJ). Western hybridizations were performed with the ECL plus Western blotting detection system (Amersham Biosciences). The GUN4 antibody was a kind gift from Dr R.M. Larkin (Michigan State University), and the conditions used were as recommended. Briefly, this consisted of blocking with 5% Blotting Grade Blocker non-fat dry milk (Bio-Rad Laboratories) in phosphate-buffered saline (PBS) (pH 7.5), followed by incubation with the GUN4 antibody at a 1 : 1600 dilution and the secondary antibody (horseradish peroxidase-conjugated anti-rabbit immunoglobulin G; Pierce Biotechnology, Rockford, IL, USA) at a 1 : 20 000 dilution. The POR antibody was a kind gift from Dr G.A. Armstrong (Ohio State University), and the blocking conditions were similar to those described for GUN4, followed by incubation with the POR antibody at 1 : 500 dilution and secondary antibody at 1 : 10 000 dilution. Western blots were scanned and quantified using a Molecular Dynamics Storm 840 system (Sunnyvale, CA).

Microarrays

Fourteen-day-old transgenic plants bearing de-greening circuit #3 were harvested before induction, 24 and 48 h post-induction with 10 µm 4-OHT, and after being allowed to re-green for 3 days. Treatments were replicated in two biologically independent experiments. Total RNA was isolated from each sample (eight plants per treatment) using an Aurum™ Total RNA Mini Kit (Bio-Rad Laboratories). Paired 2.5-µg aliquots from each treated and control RNA sample were reciprocally labelled with either Cy3 or Cy5 dendrimers using the Array 900™ system (Genisphere, Hatfield, PA, USA), such that RNA from each treatment was labelled and hybridized (in parallel with a contrastingly labelled control sample) four separate times, for a total of 12 microarrays balanced with respect to treatment, dye and biological replication. Microarrays were spotted at high density on SuperAmine™ slides (Telechem, Sunnyvale, CA, USA) using amine-modified 70-mer oligonucleotide probes (ATH1 version 1, Operon, Huntsville, AL, USA), representing essentially every predicted gene in the Arabidopsis genome. Microarray signal intensities were quantified and analysed using the TM4 software suite (Saeed et al., 2003). Raw signal intensities were normalized within array blocks using the Lowess function, and normalized log2-transformed signal intensities of > 10 units (on a scale of 0–26) were subjected to statistical tests (t-test, one-way analysis of variance, anova) to identify expression ratios that differed significantly from the mean. The complete microarray data are deposited in the ArrayExpress public database at the European Bioinformatics Institute (EMBL-EBI) under accession numbers A-MEXP-294 and E-MEXP-505.

Electron microscopy

Leaves from 14-day-old plants that had been incubated with 4-OHT for 0, 24 and 48 h were fixed in 2.5% and 5% glutaraldehyde in 0.1 m sodium cacodylate buffer (pH 7.2) for 1.5 h, rinsed three times in 0.1 m sodium cacodylate buffer and post-fixed in 2% osmium tetroxide in 0.1 m sodium cacodylate buffer for 2.5 h. Specimens were dehydrated in a graded acetone series (30, 50, 70, 90 and 100%) for 10–15 min. Both the fixation and dehydration steps, with the exception of the 100% dehydration step, were performed at 4 °C. The specimens were embedded over 2 days by adding Spurr's epoxy resin, and polymerized at 70 °C. Thin sections were obtained with a Porter Blum MT-2 ultramicrotome (Sorvall Inc., Norwalk, CT), stained for 7 min in 1% (w/v) uranyl acetate and for 2 min in 0.2% (w/v) lead acetate, and examined under an AEI electron microscope (Harlow, Essex, UK).

Chlorophyll fluorescence imaging

To visualize early changes in photosynthetic efficiency as a result of the de-greening process, images of chlorophyll fluorescence were obtained using a FluorCam instrument (Photon Systems Instruments). Plants were induced to de-green as described above, and were compared with non-induced and wild-type Columbia plants as a control. Plants were dark-adapted for 30 min prior to fluorescence measurements. Data were obtained using the default fluorescence quenching analysis protocol, with the following modifications: a 30-s dark pause was used after the Fm measurement; pulse fluorescences were subtracted for Fm measurement of the ‘dark’ level and for Fm measurement during Kautsky induction. Data analysis was performed using the manufacturer's software (Fluorcam version 5.0). Time course plots of the relevant parameters were generated using Microsoft Excel.

ROS detection

To investigate the production of ROS during the de-greening process, the probe CM-H2DCFDA (Molecular Probes, Eugene, OR, USA) was used. DCFDA is a cell-permeant indicator which is non-fluorescent until the acetate groups are removed by intracellular esterases and oxidation occurs within the cell. ROS production was visualized under an Olympus FVX-IHRT Fluoview confocal laser scanning microscope (Center Valley, PA) using an argon (488-nm) laser. Chlorophyll autofluorescence was visualized following excitation with an HeNe (543-nm) laser.

Acknowledgements

Funding for this work was provided by the US Department of Defense (Defense Advanced Research Projects Agency and Office of Naval Research) and Canadian Defence Research & Development Canada. We thank Drs Eric Eisenstadt and A.S.N. Reddy for helpful comments, Dr Marinus Pilon for reading the manuscript and for use of his FluorCam instrument, Chris Cohu for help with the FluorCam, and Drs Barbara Demmig-Adams, William Adams and Ladislav Nedbal for reviewing the FluorCam data and providing helpful comments. We thank Aaron Donahue for microarray technical assistance. We thank Drs Greg Armstrong and Rob Larkin for the POR and GUN4 antibodies. We thank Dr Stefan Hortensteiner for the kind gift of the PAO gene.

Ancillary