PIF3 is a repressor of chloroplast development

To further understand the role of PIF3 in early seedling development, we constructed a pif1pif3 double mutant using an independently isolated pif3 T-DNA insertion allele that is identical to pif3-1 (14) and a previously undescribed pif1 allele designated pif1-101 (see SI Text and Fig. S1). The PIF1 protein has been shown to repress chloroplast development and protochlorophyllide (Pchlide) synthesis in the dark (20). To test whether PIF3 might be acting similarly, we followed accumulation of Pchlide in pif1, pif3, and the pif1pif3 double mutant (Fig. 1A). All lines showed an increase in Pchlide. This increase was clearly detectable 2.5 d after germination, and at all time points the response of the pif1pif3 double mutant appeared additive to that of pif1 and pif3. Analysis of Pchlide levels in the pif1-2 and pif3-3 alleles confirmed our results (Fig. S2A). Because the pif3-3 allele contains no detectable transcript (15) or protein (17), this result is consistent with the phenotype of pif3 mutants being due to loss of PIF3 function. In our experiments, seeds were routinely germinated after 2 h of white light (WL), a treatment reported to have no longer-term effects on seedling growth in the dark (21). Nevertheless, we checked whether this short pretreatment contributed to the pif mutant response. As shown in Fig. S3A, Pchlide was also elevated in the pif1pif3 double mutant even after germination in complete darkness. The increase in Pchlide was not simply due to an increased rate of germination, because all genotypes showed at least 95% germination by day 1 in these experiments.

In addition to the effects on Pchlide accumulation, we also observed that dark-grown pif mutant seedlings had open cotyledons and had lost their apical hook (Fig. 1B). This response was observed in the majority, but not all, of the single mutant seedlings but was more consistent and stronger in the pif1pif3 double mutant. Moreover, it was not due to the WL pretreatment because seedlings germinated completely in the dark showed the same response (Fig. S3B). To test whether pif mutant seedlings showed the full constitutive photomorphogenic phenotype, we also measured hypocotyl lengths of dark-grown seedlings. In all cases, pif mutants were shorter in the dark compared with WT, with the pif1pif3 double mutant again showing an additive phenotype (Fig. 1C). This was also true in the absence of the WL pretreatment (Fig. S3C). Finally, one distinctive feature of constitutively photomorphogenic seedlings such as cop1 is that they show a partially developed chloroplast in the dark that is characterized by a reduced prolamellar body (PLB) and increased prothylakoid membranes (22). We therefore examined etioplasts in dark-grown pif mutant seedlings (Fig. 2 A–D). After 4 d in the dark, WT etioplasts showed a characteristic, highly regular PLB with little prothylakoid development (Fig. 2A). In contrast, both pif1 and pif3 etioplasts showed increased membrane development and PLBs that were generally reduced in size (Fig. 2 B and C). The most significant differences to WT were seen with the pif1pif3 double mutant, where PLB size was severely reduced and prothylakoid membranes were extensive, although no membrane stacking was observed (Fig. 2D). In some cases, no PLB was observed in pif1pif3 double mutant seedlings, although full-size PLBs were detected occasionally.

When grown in the dark for 4 d before transfer to WL, pif1pif3 double mutants failed to green over the next 24 h, with pif1 and pif3 showing only moderate greening during this period (Fig. 3A and Fig. S4). An identical result was observed with the pif1-2 and pif3-3 alleles (Fig. S2 B and C) and has been observed for pif1 (20). Interestingly, detailed examination of the time course after transfer to WL showed that at 2 h pif1, pif3, and the pif1pif3 double mutant had more chlorophyll than WT but that this was already reversed after 4 h of WL (Fig. 3B). In contrast to the situation after 4 d in the dark, seedlings transferred to WL after 2 d in the dark were able to green, with the pif1pif3 double mutant accumulating the most chlorophyll (Fig. 3C). The pif1-2 and pif3-3 alleles were also able to green more efficiently under these conditions (Fig. S2 B and C), and this ability was gradually lost as seedlings aged at transfer to WL (Fig. 3C and Fig. S2C). The time course of chlorophyll loss after transfer to WL and the effect of increasing the dark period on subsequent greening ability are consistent with the loss of chlorophyll in the pif mutants being due to photooxidative destruction rather than reduced synthesis. To test this, we examined the effect of different WL fluences on chlorophyll levels in WT and pif mutant seedlings. As shown in Fig. 3D, as the fluence rate increased, the relative loss of chlorophyll in the pif mutants also increased compared with WT, consistent with photooxidation being the primary cause of chlorophyll deficiency in the pif mutants.

We examined chloroplast ultrastructure in WT and pif mutant seedlings after transfer to WL after 2 or 4 d in the dark (Fig. 2 E–L). Wild-type chloroplasts were already well developed 24 h after transfer from 2 d in the dark, with some thylakoid stacking evident at this stage, although in some cases a residual PLB was observed (Fig. 2E). Consistent with the chlorophyll data, there was no evidence of any repression of chloroplast development in pif1, pif3, and pif1pif3 mutants, and in all cases there appeared to be more thylakoid stacking than in WT (Fig. 2 F–H). No residual PLBs were observed in pif1, pif3, or pif1pif3 under these conditions. The situation after transfer to WL from 4 d in the dark was more complex. Development of WT chloroplasts was similar to that seen after transfer from 2 d in the dark (Fig. 2I). In contrast, pif1 and particularly pif1pif3 double mutants contained chloroplasts with poorly defined membrane structure and no evidence of granal stacking (Fig. 2 J and L). The appearance of the chloroplasts was reminiscent of chloroplasts damaged through pigment-induced photooxidative stress (e.g., ref. 23) and was not due to problems of fixation, because other structures in these sections were well defined (for example, the mitochondrion to the right of the chloroplast in Fig. 2L). Chloroplasts in the pif3 mutant were generally more similar to WT in appearance than those for pif1 or pif1pif3 (Fig. 2K), consistent with the higher levels of chlorophyll in pif3 at this time point.

To understand the basis of the increase in Pchlide, we followed the expression of HEMA1 encoding glutamyl tRNA reductase, the rate-limiting step in tetrapyrrole synthesis (24). Expression was measured at 6 h time points from 1¼ d after germination using real-time PCR. HEMA1 expression was strongly induced in all pif mutants relative to WT at 2 and 3 d after germination, with the response severely diminished or lost at days 4 and 5 (Fig. 4A). The response in the pif1pif3 double mutant was again equivalent to both single mutants combined. We also analyzed 2 additional genes shown to be key regulatory targets in the tetrapyrrole pathway, CHLH encoding the H subunit of Mg chelatase and the chelatase regulator GUN4 (25–27). Both genes showed a similar pattern, with the strongest peak 3 d after germination, high expression after 2 d, and little induction if any after 4 d (Fig. 4B). Analysis of GUN4 expression in the pif1-2 and pif3-3 alleles gave similar results when measured 3 and 4 d after germination (Fig. S2D). Examination of glutamyl tRNA reductase protein levels showed an increase in pif1, pif3, and the pif1pif3 double mutant by 2 d in the dark (Fig. 4C).

The profile of the relative induction of HEMA1 in the pif mutants is quite unusual, with sharp peaks 2 and 3 d after germination but no induction at 2½ d. To understand the basis for this, we plotted the normalized level of HEMA1 mRNA (relative to YLS8) for WT and the pif mutants independently (Fig. 4D). This analysis revealed 2 main observations. First, HEMA1 expression was generally higher at early time points in pif1, pif3, and pif1pif3 compared with WT. Second, and most strikingly, HEMA1 expression was out of circadian phase in the pif mutants compared with WT seedlings. Although expression of HEMA1 oscillated with a period close to 24 h in WT seedlings, pif1, pif3, and pif1pif3 showed a reduction in the amplitude of oscillation and a period of oscillation that was variable ranging from ≈12 to 22 h for pif1pif3 (Fig. 4D). To test whether the circadian clock is functioning normally in dark-grown pif mutant seedlings, we examined the expression of the central clock genes CCA1, TOC1, and LHY in the same samples (Fig. S5 A–C). No major changes in expression were observed for all 3 genes, indicating that the clock is still functional in etiolated pif mutants. We also examined the expression of another circadian-regulated gene, CAX1, that is not involved in chloroplast development. Circadian expression of this output gene was unaffected in the pif mutants (Fig. S5D), indicating that pif1 and pif3 might specifically affect circadian regulation of chloroplast-related genes. Finally, we investigated the circadian regulation of PIF1 and PIF3 at this developmental stage. PIF1 and PIF3 showed a robust circadian rhythm in dark-grown seedlings, with a similar phase to that of HEMA1 (Fig. 4E).

Because PIF3 has been proposed to function positively in the light induction of nuclear-encoded chloroplast genes, we followed gene expression after transfer to 24 h of WL, a time at which chlorophyll levels are severely reduced in the mutants. Although, as noted previously, expression was higher in the dark for HEMA1, CHLH, and GUN4 in all pif mutant lines, all 3 genes were light induced to a similar degree, and the final expression levels of these genes in the light were still higher in pif1, pif3, and pif1pif3 than in WT (Fig. 4F and Fig. S6). We also tested whether pif mutants could respond to monochromatic lights sources and over shorter time periods. As shown in Fig. S2, induction of GUN4 was still apparent in pif1-2 and pif3-3 after 4 h FR and 8 h R light treatments.

The data presented here are consistent with PIF3 functioning as a repressor of chloroplast development in the dark. Pchlide synthesis was higher in pif3 than WT seedlings (Fig. 1A), and initial rates of chlorophyll synthesis were also greater (Fig. 3 B and C). pif3 seedlings also showed more advanced development of etioplasts and chloroplasts (Fig. 2). In these respects, the pif3 mutant behaved identically to the pif1 mutant, which has been identified as a negative regulator of chloroplast development (20). Consistent with these observations, the pif1pif3 double mutant showed an additive phenotype. Previously, the pif3 mutant had been described as showing inhibition of chloroplast development (15), and the hypothesis that PIF3 acts positively early in signal transduction (and negatively in the longer term) is still current (16, 17). Our data suggest that for the earliest stages of chloroplast development and for the target genes that we have analyzed (Fig. 4) that this is not the case. As discussed later, there are possible explanations for the previously reported loss of induction of chloroplast genes in pif3 (15, 17), but the observations that the overexpression of PIF3 is not sufficient for the induction of phytochrome-regulated genes and that DNA binding of PIF3 in the dark is required (17) are certainly consistent with a role for PIF3 as a repressor. Moreover, the phytochrome-interacting PIF proteins have generally been shown to be acting as repressors not activators of photomorphogenic responses (3, 4, 28), and our results are therefore consistent with a common molecular mechanism for this class of signaling protein.

The reason for the previous misinterpretation of the pif3 mutant phenotype is that seedlings transferred to WL after 4 d in the dark showed a reduced level of chlorophyll compared with WT (Fig. 3). This response, which is identical for pif1 and exaggerated in a pif1pif3 double mutant, is most likely due to photooxidative destruction of chlorophyll. Our results are entirely consistent with this explanation, because the loss of chlorophyll is dependent on the length of the dark period before transfer (and therefore the degree of excess Pchlide production), the fluence rate of WL, and the time of WL exposure. Misregulation of the tetrapyrrole synthesis pathway commonly leads to a photobleaching phenotype (e.g., refs. 29 and 30), and overaccumulation of Pchlide is well established as leading to photooxidative damage (29), for example, in the FR block of greening response (31).

One interesting phenotype that we observed for the pif1pif3 double mutant was that it showed a moderate constitutive photomorphogenic response in dark-grown seedlings (Fig. 1). This response was seen even when seeds were kept in complete darkness postimbibition (Fig. S3). Further investigation demonstrated that both pif1 and pif3 single mutants showed a similar, but less pronounced, response. A shorter hypocotyl in the dark has been seen for pif3 (14, 18) and pif1 (32), and a similar phenotype with expanded cotyledons, hook opening, and hypocotyl inhibition was recently observed for pif1, pif3, and a pif1pif3 double mutant (28). In this case, the authors reported a synergistic interaction between PIF1 and PIF3 in contrast to the additive phenotype reported here. A constitutive photomorphogenic phenotype of the pif1pif3 double mutant is expected based on the stronger, dominant negative phenotype of overexpressed truncated PIF1 (12). Presumably, in this case, the PIF1 protein is interfering with the function of additional PIFs, including PIF4 and PIF5 (28). Interestingly, constitutive activation of phytochromes in the dark also results in this phenotype, which could result from Pfr-mediated degradation of multiple PIFs (33). However, whether the pif1pif3 double mutant still requires the presence of seed Pfr (produced during seed set) to reveal the response remains to be seen. In our assays, we saw all aspects of the phenotype in seedlings that had only seen light during seed plating, before the seeds had fully imbibed, and Leivar et al. (28) were unable to block the pif1 and pif1pif3 response, even with a FR light treatment immediately after plating.

The rate-limiting step for Pchlide (and chlorophyll) synthesis is the enzyme glutamyl tRNA reductase (24). Light regulation of this step is mediated through changes in expression of the HEMA1 gene (34), and HEMA1 is one of a small group of highly regulated tetrapyrrole genes including CHLH and GUN4 (25, 27). The substantial increase in HEMA1 expression and consequent increase in glutamyl tRNA reductase protein can fully account for the observed increase in Pchlide levels in the pif1 and pif3 mutants. The increase in tetrapyrrole synthesis in pif1 was previously suggested to be due to a subtle down-regulation of the ferrochelatase gene (FCII) and a concomitant up-regulation of the heme oxygenase HO3, resulting in less free heme and less inhibition of glutamyl tRNA reductase activity (35), the opposite of the phenotype of the phytochrome-chromophore-deficient mutants in which the heme branch of the pathway is almost completely blocked (36). We have not tested these genes directly, but because HO3 has exceptionally low expression in seedlings and its loss has no impact on chromophore synthesis in the presence of HO1 (37), these changes are unlikely to make more than a minor contribution compared with the substantial increase in levels of the rate-limiting enzyme of the pathway. One reason that previous studies did not observe the changes seen here is that microarrays using dark-grown pif1 (35) and pif3 (15) and their follow-up analyses were performed using seedlings that had been grown for 4 d in the dark. As is clear from our current studies (Fig. 4 and Fig. S2D) differences between WT and the pif mutants are minor at this time.

We observed that both the pif1 and pif3 mutations affected circadian regulation of HEMA1, CHLH, and GUN4. HEMA1 and CHLH have been shown to be circadian regulated in the light (25), but circadian regulation for GUN4 has not been reported previously. The altered clock regulation of HEMA1 was not due to a major defect in the circadian clock, because the pif mutants did not have a strong effect on the expression of the core clock components CCA1, LHY, and TOC1. The control output gene CAX1, a H⁺/Ca²⁺ antiporter (38) unrelated to chloroplast function, was also unaffected, suggesting that PIF1 and PIF3 function specifically in circadian control of genes involved in chloroplast development. A circadian clock has been shown to be functional in dark-grown Arabidopsis seedlings, with entrainment initiated through changes in temperature or imbibition (39), and can be observed just 2 d after imbibition (39) or even earlier (40). Moreover, the importance of this clock in controlling chloroplast development is supported by the observation that a range of clock mutants fail to green normally after transfer to WL (40). We therefore propose that PIF1 and PIF3 function in circadian control of chloroplast development as shown in the model in Fig. 5. Furthermore, we favor a role for the PIF proteins in the output from the clock. Although phytochrome has a major role in the entrainment of the circadian clock by light (41), PIF3 does not play a significant role in controlling light input or function of the clock (15, 42, 43). Although we cannot completely rule out a role in entrainment, the apparent specificity of the response for chloroplast development genes suggests otherwise.

In our experiments, PIF1 and PIF3 showed robust circadian regulation in dark-grown seedlings, suggesting that clock regulation of PIF function is via circadian control of expression. Analysis of multiple circadian microarray experiments suggests that PIF1, but not PIF3, expression is under circadian control (44). However, a low-amplitude circadian rhythm has also been observed for PIF3 using a PIF3:LUC⁺ reporter construct (43). Within the resolution of our experiments, the PIF genes appear to cycle in the same circadian phase as HEMA1, precluding a simple mechanism of circadian-regulated PIF repression of HEMA1 expression. However, a small difference in phase could still permit such a mechanism. Alternatively, because both PIF1 and PIF3 have been shown to interact directly with TOC1 (45), a model in which the clock controls PIF function through direct protein interaction is also plausible.

In summary, our results show that both PIF1 and PIF3 are negative regulators of chloroplast development that function to integrate light and circadian control of this critical process. Exactly how they achieve this will be the focus of future studies.

Arabidopsis (Arabidopsis thaliana L.) seeds were imbibed at 4 °C for 2 d in darkness, followed by 2 h of WL (110 μmol·m⁻²·s⁻¹), and returned to darkness at 23 °C, indicating the start of the respective experiment (unless otherwise stated).

For hypocotyl measurements, 15 Arabidopsis seedlings were measured, and the longest 10 were averaged for 1 biological repeat. For Pchlide measurements, 100 μg of seedling material was extracted twice in acetone/0.1 M NH₄OH, 90:10 (vol/vol) as described in ref. 36. Chlorophyll was measured as described in ref. 27. Cotyledon samples for transmission electron microscopy were prepared and examined as described in ref. 36. Numerous plastids in at least 2 independent samples were viewed for each genotype and experimental condition and photographs were taken of representative plastids.

RNA extraction and real-time PCR methods were exactly as described in ref. 27, with one exception (see SI Text). To assess the expression of genes between genotypes at different time points, the absolute C(t) value of the YLS8 control gene was subtracted from the absolute value of the experimental gene for each biological replicate, and the average C(t) value for all biological replicates was used for comparison between genotypes. For primers, see SI Text. Protein extraction and immunoblotting were conducted exactly as described in refs. 27 and 36, with 50 seedlings extracted in 100 μL of SDS extraction buffer.