Phytochrome Interacting Factors (PIFs) in Solanum lycopersicum: Diversity, Evolutionary History and Expression Profiling during Different Developmental Processes

Although the importance of light for tomato plant yield and edible fruit quality is well known, the PHYTOCHROME INTERACTING FACTORS (PIFs), main components of phytochrome-mediated light signal transduction, have been studied almost exclusively in Arabidopsis thaliana. Here, the diversity, evolution and expression profile of PIF gene subfamily in Solanum lycopersicum was characterized. Eight tomato PIF loci were identified, named SlPIF1a, SlPIF1b, SlPIF3, SlPIF4, SlPIF7a, SlPIF7b, SlPIF8a and SlPIF8b. The duplication of SlPIF1, SlPIF7 and SlPIF8 genes were dated and temporally coincided with the whole-genome triplication event that preceded tomato and potato divergence. Different patterns of mRNA accumulation in response to light treatments were observed during seedling deetiolation, dark-induced senescence, diel cycle and fruit ripening. SlPIF4 showed similar expression profile as that reported for A. thaliana homologs, indicating an evolutionary conserved function of PIF4 clade. A comprehensive analysis of the evolutionary and transcriptional data allowed proposing that duplicated SlPIFs have undergone sub- and neofunctionalization at mRNA level, pinpointing the importance of transcriptional regulation for the maintenance of duplicated genes. Altogether, the results indicate that genome polyploidization and functional divergence have played a major role in diversification of the Solanum PIF gene subfamily.


Introduction
Every aspect of plant physiology is influenced by light. Right after germination, etiolated growth (skotomorphogenesis) allows seedlings to seek for light at the soil surface and, upon light exposure, signal transduction initiates photomorphogenic development (deetiolation), characterized by chloroplast differentiation and initiation of photosynthetic activity. During autotrophic vegetative development, light provides the energy that fuels plant growth, designs architecture of mature plant and regulates flowering. Furthermore, light deprivation is an phosphorylation and subsequent proteasomal degradation of PIFs, which leads to physiological responses. A notable exception to this dynamic behavior is PIF7, which despite interacting with PHYB shows no detectable light-induced degradation [1]. Several target genes for A. thaliana PIF proteins have been identified. PIF3 mediates the initial phases of seedling lightinduced chloroplast development during deetiolation through the regulation of nuclear genes involved in photosynthesis and chloroplast biogenesis [15]. ChIP-PCR experiment confirmed that PIF4 binds to the E-box motifs of the promoters of both chloroplast activity maintainer genes GOLDEN 2-LIKE 1 (GLK1) and GLK2, repressing their expression [16]. Additionally, PIF4 and PIF5 act as transcriptional activators of the master senescence transcription factor ORESARA 1 (ORE1) and chlorophyll degrading enzyme encoding genes, such as STAY GREEN 1 (SGR1) and NON-YELLOW COLORING 1 (NYC1), during dark-induced senescence by direct interaction with the G-box motifs on the corresponding promoter regions [16][17][18]. Finally, PIF1 has been shown to directly bind the G-box motif of the promoter of the chlorophyll and carotenoid biosynthetic genes PROTOCHLOROPHYLLIDE OXIDOREDUCTASE and PHYTOENE SYNTHASE (PSY), inducing and inhibiting their transcription, respectively [19,20]. Only one tomato PIF gene has been characterized so far, PIF1a, and, in agreement with its Arabidopsis ortholog showed to modulate carotenoid biosynthesis during fruit ripening. During green stages of fruit development, as a consequence of self-shading, Chl reduces R/ FR ratio stabilizing PIF1a, which, in turn, represses the expression of the fruit-specific PSY1. After the onset of ripening, degreening allows the activation of Pfr and the consequent PIF1a degradation releases PSY1 transcription, enhancing carotenogenesis [12,21].
Considering the importance of light perception and signaling for plant development and fruit quality and, the poorly available knowledge about PIF genes in tomato; here we performed a comprehensive characterization of this gene subfamily in S. lycopersicum. By surveying the tomato genome, we identified eight PIF homolog sequences. The phylogenetic, divergence time estimation and selective pressure evaluation analyses allowed us to reconstruct the evolutionary history of PIF genes in S. lycopersicum and closely related Solanaceae species, the wild tomato S. pennellii and S. tuberosum. We further explored the transcriptional profile in four different developmental contexts, deetiolation, dark-induced senescence, daily cycle and fruit ripening, and identified expression patterns that suggest functional specificity. The data were discussed in the context of tomato genome evolution.

Phylogenetic and Evolutionary Analysis of PIF loci
By performing a BLAST search against fully sequenced genome databases using A. thaliana canonical PIF sequences as queries, 119 sequences from 16 species were retrieved including sequences of the bHLH superfamily that do not belong to the PIF subfamily [1] (see Material and Methods, S1 Table). In agreement with previous report, no PIF homologs were found in chlorophytes [22]. In the basal land plants Marchantia polymorpha (liverworth), Physcomitrella patens (moss) and Selaginella moellendorffii (lycophyte), one, four and three PIF homologs were identified, respectively. Spermatophyte species harbor several gene copies that, based on the phylogenetic reconstruction, are mainly divided in two super clades named according to the corresponding A. thaliana homolog representative. The first contains PIF1 and PIF4 sequences and, the second encompasses PIF3, PIF3-like 1 and 2 (PIL1/2) [23], PIF8, PIF7, ALCATRAZ (ALC) and SPATULA (SPT) [24] sequences. In the second clade, PIF3 and PIL1/2, PIF7 and PIF8 and, ALC and SPT clustered together, respectively (Fig 1, S1 Fig, S1 Text).
Whereas Arabidopsis has six PIF encoding genes, henceforth named AtPIFs, eight loci were identified in S. lycopersicum genome, corresponding to the following accessions in Sol  Table. Compacted clades encompassing more than one sequence are indicated by black triangles. Arabidopsis thaliana and Solanum lycopersicum sequences are indicated with green and red circles, respectively. PIF clades are highlighted with colored squares. Numbers at nodes represent bootstrap/ approximate likelihood-ratio test (aLRT) values. Genomics Network database [25]: SlPIF1a: Solyc09g063010, SlPIF1b: Solyc06g008030, SlPIF3: Solyc01g102300, SlPIF4: Solyc07g043580, SlPIF7a: Solyc03g115540, SlPIF7b: Solyc06g069600, SlPIF8a: Solyc01g090790, SlPIF8b: Solyc10g018510 (Fig 1). Aminoacid pairwise sequence alignments indicated that Arabidopsis and tomato homologs share 27-51% identity (S2 Table). Despite this low identity score, the APB-binding and bHLH domains were found in all tomato protein sequences, reinforcing their identity as PIF proteins. However, it is worth mentioning that tomato SlPIF1b, SlPIF4 and SlPIF8b display an amino acid substitution in the APB-binding domain that alters the conserved Q residue to G, E and E, respectively [26]. On the contrary, APA-binding domain was exclusively identified in SlPIF1s and SlPIF3 (S2 Fig). Interestingly, the tree topology clearly showed that Arabidopsis AtPIF4 and AtPIF5 genes were originated by a Brassicaceae exclusive duplication, explaining the existence of a single gene in tomato genome within the clade PIF4. No differences in gene copy number were observed between S. lycopersicum and the most distantly related species within Lycopersicon section (i.e. tomatoes), S. pennellii. For PIF1, PIF7 and PIF8 clades, the analyzed tomato species harbor two gene copies, while for PIF3 and PIF4 a single copy was identified. S. tuberosum has a similar PIF gene copy number, excepting for a single PIF8 locus (Fig 1).
To gain insight on the evolutionary history of PIF gene family, we estimated the divergence time of PIFs using molecular clock [27]. The duplication of PIF1, PIF7 and PIF8 was estimated in a range of time from 59.2 to 91.2 MYA (millions of years ago). As expected [28,29], our data indicated that tomato and potato PIF genes diverged around the species splitting event (Fig 2) estimated about 5.1 to 7.3 MYA [30]; excepting PIF7b, for which an estimate of 22.5-23.8 MYA was retrieved. Similarly, the divergence of S. lycopersicum and S. pennellii PIF genes dates close to the estimated age of the most recent common ancestors within the species, 2.2-3.1 MYA [27], with the exception of PIF8b, for which a value of 6.2 MYA was obtained. The high divergence times The divergence times between the duplicated PIF genes in Solanaceae are shown in green. The divergence times between tomatoes (S. lycopersicum and S. pennellii) and S. tuberosum and, S. lycopersicum and S. pennellii homologs are indicated in blue and red, respectively. Species divergence times are shown in black (Arabidopsis thaliana-Solanaceae [34], Solanum tuberosum-Solanum lycopersicum [30], S. pennelli-S. lycopersicum [27]). Values are expressed in million years ago. observed for PIF7b and PIF8b are consequence of high synonymous substitution values (dS). Aiming to test whether the high dS values were consequence of positive selection or neutral evolution, we evaluated the selective constraints under which PIF gene are evolving (Table 1). Indeed, PIF7b showed signatures of positive selection, particularly in threonine 343 and serine 369 (BEB test P>95%). The rest of the PIF clades showed to be evolving under purifying selection. Unfortunately, we were unable to perform the test for PIF8b because it is absent in S. tuberosum.   A. thaliana PIF proteins are known regulators of seedling deetiolation and dark-induced senescence, and are modulated by the circadian clock [1,17]. Particularly in S. lycopersicum, recently, the role of SlPIF1a in ripening-associated carotenogenesis was also been demonstrated [12]. To evaluate the functional diversity of tomato PIF genes, a comprehensive mRNA accumulation profiling was performed during seedling deetiolation, dark-induced senescence, diel cycle and fruit ripening. SlPIF8a and SlPIF8b were not considered for functional analyses because there are no publications demonstrating A. thaliana PIF8 and PHYs interaction, therefore, this clade was not considered a canonical PIF. The expression profile of SlPIF genes during deetiolation was analyzed in 4 day-old darkgrown seedlings exposed to 24, 48 and 72 h of constant light or dark conditions. The darktreated seedlings exhibited typical skotomorphogenic phenotype presenting long hypocotyls as well as closed, small and chlorotic cotyledons. Seedlings exposed to constant light underwent photomorphogenesis and showed shorter hypocotyls, opened apical hooks, expanded and green cotyledons and anthocyanin accumulation (   Table).
The expression pattern of tomato PIFs during 24 h under 12/12 light/dark photoperiod was analyzed in 3-week-old plants (Fig 4). SlPIF1a, SlPIF3 and SlPIF7a showed similar oscillation patterns, characterized by lowest transcript abundance at the end of the light period followed by a progressive increase during the dark period and maximum levels 4 h after dawn. SlPIF4 mRNA abundance was significantly reduced during the afternoon achieving the lowest level at dusk and progressively increasing over the night to reach the maximum level 8 h after dawn. SlPIF7b mRNA levels were high and constant during the light period, progressively decreasing during the night. SlPIF1a and its paralog SlPIF1b exhibited distinct diel expression patterns since during the night period SlPIF1a and SlPIF1b mRNA levels progressively increased and decreased, respectively. Interestingly, SlPIF1 genes were the most copiously expressed PIFs at beginning of the light period displaying over 2-fold more transcripts in leaves than the other PIF genes (S3 Table).
Furthermore, we explored the transcriptional profile of S. lycopersicum PIF genes in leaves sampled from 3-week-old-plants maintained in darkness for 0, 1, 2, 3 and 7 days. The leaves showed clear signs of senescence as evidenced by the reduction in chlorophyll content at the seventh day (S4A Fig). The degreening was explained by the increment in PHEOPHYTINASE expression, the enzyme responsible for chlorophyll dephytylation in tomato leaves [32] and accompanied by a reduction in SlGLK1 transcripts. Additionally, the induction of senescence was confirmed by the mRNA accumulation of the senescence marker SENESCENCE-ASSO-CIATED GENE 12 (SlSAG12, [33]) and A. thaliana ORE1 homologs SlORE1S02, SlORE1S03 and SlORE1S06 (S4B Fig). These data allowed us to conclude that after 7 days of dark treatment, the plants underwent dark-induced senescence. Transcriptional profiling revealed that SlPIF1a, SlPIF3, SlPIF7a and SlPIF7b were downregulated whereas SlPIF1b and SlPIF4 were upregulated upon darkness exposure, suggesting that probably the formers are involved in dark-induced senescence signaling (Fig 5).
To address the transcriptional behavior of all six SlPIF genes during tomato fruit ripening and evaluate their possible involvement in the light-dependent regulation of this developmental process, fruits at mature-green (MG) stage were harvested and left ripen under constant light or dark conditions. Total chlorophyll and carotenoids levels were measured and, as expected a concomitant reduction in total chlorophylls temporally coincided with the accumulation of the main carotenoids typically found in tomato fruits, thereby demonstrating that the detached fruits were undergoing normal ripening (S5 Fig). The levels of transcripts for both SlPIF7 paralogs were undetectable in all fruit stages analyzed. Under dark conditions, SlPIF1a, SlPIF1b, SlPIF3 and SlPIF4 mRNA levels peaked 2 days after the start of the treatment, when the fruits were still in MG stage, followed by a reduction at the breaker (BR) stage (Fig 6). During the progression of ripening, SlPIF1a showed to be transcriptionally induced and significantly higher in the presence of light, whereas the mRNA levels of its paralog, SlPIF1b, were clearly lower in light-than in dark-treated fruits and did not respond to ripening. Finally, SlPIF3 and SlPIF4 mRNA levels were relatively constantly low along ripening and did not show clear patterns of regulation by light and dark treatments. In terms of relative expression, the most abundantly expressed PIF gene in fruits was SlPIF3, with 17-, 6-and 1.7-fold more mRNA amount than SlPIF4, SlPIF1a and SlPIF1b, respectively (S3 Table).

Discussion
The key role played by light signaling on tomato plant development and fruit nutritional value has been widely studied by the use of mutants and transgenic approaches [3,4]. However, PIF genes have been lagged behind and almost exclusively studied in A. thaliana [11][12][13][14]. Besides the well described PHY-mediated proteasomal degradation mentioned above, PIF genes are under tight transcriptional regulation as indicated by database and genome-wide binding site analyses for several A. thaliana transcription factors. Moreover, it has been suggested that AtPIFs regulate their own expression by a complex autoregulatory mechanism [34], pinpointing the importance of studies approaching the expression regulation at transcriptional level.
Aiming to gain insights on the function of these genes in tomato, here we explored their genetic diversity and expression profiling along different physiological contexts.
PIF is one of the 26 gene subfamilies of the monophyletic bHLH superfamily of plant transcription factors [22]. By using the five well described PIF proteins sequences from A. thaliana as baits (i.e. AtPIF1/3/4/5/7), we retrieved 119 homologs from 16 species and performed a Evolution, Diversity and Expression of Tomato PIFs phylogenetic analysis that allowed the identification of eight clades in Spermatophyte species (Fig 1). A comparative study between tomato and grape genomes proposed that a wholegenome triplication affecting Rosids, which includes Arabidopsis, and Euasterids, which includes Solanum, occurred in a common eudicot ancestor more than 120 MYA [5,35]. Interestingly, the monocot representatives of our phylogenetic reconstruction, Oryza sativa and Sorghum bicolor, did not show PIF genes in all the eight identified clades (Fig 1, S1 Table). Further, another triplication estimated at 71 (± 19.4) MYA occurred in the Solanum lineage followed by widespread gene loss that predates the 7.3 MYA tomato-potato divergence [5,30]. This second event was likely the origin of the duplications within PIF1, PIF7 and PIF8 clades since the estimated divergence time between the duplicated genes coincided with the date of the whole-genome triplication (Fig 2). To confirm this hypothesis, the gene collinearity was analyzed along the flanking genomic regions of the duplicated genes. As demonstrated for Solanaceae PSY genes [5], the SlPIF1, SlPIF7 and SlPIF8 paralog regions showed recognizable small scale synteny (S6 Fig). Thus, these polyploidization events may represent the foundation of the PIF subfamily diversification.
The evolutionary history of a genome is the result of the interdependent diversification of different genetic features like regulatory sequences, mobile elements and coding regions. Although, it is expectable that the gene divergence time approximately coincides with the corresponding species splitting date, heterogeneity in the nucleotide substitution rates among genetic features within the genome can defy the molecular clock approach [36]. In this sense, the estimated divergence time for tomato and potato PIF7b genes significantly predated the splitting date between species; while a similar situation was observed for S. lycopersicum and S. pennellii PIF8b. Interestingly, the evolutionary analysis for PIF7b demonstrated signatures of positive selection, which can be associated to functional divergence. The absence of PIF8b in S. tuberosum might be attributed to stochastic gene loss; consequently, no evolutionary analysis was performed.
To explore the functional diversification of tomato PIF genes, a comprehensive expression profile was carried out under various physiological processes induced or regulated by light, such as deetiolation, daily cycle, senescence and fruit ripening. Interestingly, SlPIF genes displayed differential mRNA accumulation pattern at least along one of the analyzed contexts, suggesting that these genes have undergone functional specification. Little and fragmented information is currently available about the transcriptional regulation of PIF genes and only punctual similarities with our experimental conditions were found in literature. The transcription of AtPIF4 and AtPIF5 has shown to be upregulated in Arabidopsis seedlings upon white light exposure [37]. This result is consistent with our observation that tomato SlPIF4 mRNA levels increase during deetiolation (Fig 3). Tomato SlPIF4 and SlPIF7a transcript accumulation patterns during diel cycle resemble those observed for Arabidopsis AtPIF4, AtPIF5 and AtPIF7, whose mRNA levels are regulated by the circadian clock [38,39]. However, while AtPIF1 and AtPIF3 mRNA levels in Arabidopsis remained relatively constant along the diel cycle [37], the tomato SlPIF1a, SlPIF1b and SlPIF3 oscillated under 12 h/12 h light/dark photoperiod. SlPIF7b transcript levels also fluctuated during the diel cycle, suggesting that all tomato PIFs are transcriptionally regulated by the circadian clock (Fig 4). Besides the diel cycling, SlPIF7s have shown to be exclusively expressed in true leaves (S3 Table), reinforcing their role in circadian response regulation as demonstrated for AtPIF7 ortholog [40]. Similarities were also found with Arabidopsis during dark-induced senescence. In Arabidopsis, AtPIF4 and AtPIF5 exhibited a peak of transcript accumulation in leaves one day after dark treatment [16] triggering senescence through the activation of the master transcription factor AtORE1 [17]. Accordingly, tomato SlPIF4 mRNA reached the highest levels one day after the start of the constant dark treatment (Fig 4). Interestingly, tomato ORE1 homologs, SlORE1S02, SlORE1S03 and SlORE1S06 were also induced by the dark treatment, suggesting that a similar functional link between PIF4 and ORE genes regulates dark-induced senescence in both Arabidopsis and tomato (S4B Fig). A recent publication functionally characterized tomato SlPIF1a demonstrating its involvement as a negative regulator of fruit carotenogenesis [12]. Llorente et al. (2016) reported that SlPIF1a expression is induced along ripening and SlPIF1b is not expressed in fruits. Our data also showed that SlPIF1a transcripts do accumulate during ripening, however, the amount of SlPIF1b mRNA at MG stage was 3-fold higher than SlPIF1a (S3 Table, Fig 6). These apparent contrasting data might be the results of differences in experimental design, since the transcriptional profile showed here was performed from fruits ripened off vine and under constant light/dark treatments. Moreover, SlPIF1a transcript accumulated at higher levels in fruits ripened under light, while SlPIF1b transcription was repressed by this treatment. This opposite pattern of light response between SlPIF1 duplicated genes was also observed in the other physiological contexts analyzed in this work and might be the result from differences in transcriptional promoter activities. Therefore, we surveyed a fragment of 2 kb upstream the translation initiation site of these genes by a de novo search for cis-regulatory elements. Motifs recognized by A. thaliana transcription factors involved in light signaling, such as PIFs and HY5, were found in both sequences. SlPIF1a promoter showed PIF and HY5 binding-motifs, PBE-box and CA-hybrid, respectively [41,42]. Additionally, CArG motifs, which are recognized by the ripening inducer transcription factor RIN [43], were also found in SlPIF1a promoter region. CA-hybrid and CArG motifs were also identified in SlPIF1b regulatory region together with the HY5-binding ACE-motif (S7A Fig). The presence of gene-specific motifs and, different number and distribution of shared motifs might, at least in part, explain the different transcriptional behavior of SlPIF1 duplicated genes. In particular, SlPIF1a might be target of HY5-mediated light-induction and of the above mentioned PIF autoinhibitory mechanism. The comparison of the mRNA profiles of SlPIF7a and SlPIF7b duplicated genes was only possible in leaves because in other organs PIF7 mRNA levels were near or below the detection threshold. SlPIF7a and SlPIF7b mRNA levels displayed opposite accumulation pattern during diel cycle, while both genes were downregulated by dark-induced senescence, SlPIF7a mRNA levels being 10-fold higher than those detected for SlPIF7b (Fig 4, Fig 5, S3  Table). In this case, the analysis of the promoter sequences showed also differential number and distribution of PIF and HY5 binding motifs: PBE-box, ACE-motif and CG-hybrid in SlPIF7a and; PBE-box and ACE-motif in SlPIF7b regulatory region (S7B Fig). To further evaluate differentially selected motifs between these pairs of duplicated genes, a promoter phylogenetic analysis was performed. Resembling the topology of the tree obtained from amino acid sequences, regulatory fragments also revealed that gene duplication predates species divergence (S8 Fig). Interestingly, none of the motifs identified in SlPIF1s and SlPIF7s is conserved between paralogs, being either S. lycopersicum exclusive or shared with S. tuberosum and S. pennellii orthologs (S9 Fig). These data reinforced that the duplicated genes have undergone functional divergence in the Solanaceae common ancestral.
According to Force et al. (1999) [44], the loss of regulatory subfunctions in the promoter region by mutation and genetic drift is the main process by which duplicated genes are preserved, as long as they retain the complete set of subfunctions from the ancestral gene. In this context, it is expected that the duplicated loci should complement each other and show differences at the regulatory region. This model postulates that duplicated loci can undergo three different fates: nonfunctionalization, neofunctionalization and subfunctionalization. The first occurs when one copy acquires disabling mutations at the promoter region, leading to gene expression loss, while the other copy remains intact. The second takes place when a copy acquires new regulatory motifs, which confers a new regulatory function to this gene. The last is caused by degenerative mutations at both loci leading to loss or reduction of subfunctions. Our observations suggest that SlPIF1 genes suffered qualitative subfunctionalization, as evidenced by their opposite responsiveness to light, and neofunctionalization, since SlPIF1a acquired a regulatory function during fruit ripening. Whilst, SlPIF7 duplicated loci appeared to have undergone quantitative subfunctionalization possibly caused by fixed reduction-ofexpression mutations, which resulted in lowered expression of both copies. Moreover, it has been proposed that quantitative subfunctionalization is a transitory state to eventual neofunctionalization [45]. This seems to be the case of SlPIF7b gene for which, besides the reduced expression levels described above, positive selection has been also verified. A very interesting mechanism of neofunctionalization of duplicated genes in Solanum lineage has been recently described in tomato [46]. While in photosynthetic tissue, a CHLOROPLAST-SPECIFIC LYCOPENE β-CYCLASE (LCYβ) mediates the conversion of lycopene to β-carotene, in chromoplast, this reaction is executed by the product of the CHROMOPLAST-SPECIFIC LYCO-PENE β-CYCLASE gene (CYCβ), a LCYβ paralog. Sequence analysis of CYCβ gene from a repository of tomato and wild relative accessions showed that CYCβ undergoes purifying selection in tomato clade. However, the abundant and diverse variations in the promoter region are likely related to regulatory neofunctionalization that played a key role in fruit color development in tomato.
The data presented here bring evidences that SlPIF duplicated genes (e.g. SlPIF1a and SlPIF1b), originated during Solanum lineage specific whole-genome triplication, have undergone sub-and neofunctionalization most likely due to variations in promoter region than in the coding region, disclosing the impact of polyploidization events during the evolution of PIF gene subfamily.

Conclusions
Solanum lycopersicum genome harbors eight PIF encoding loci, SlPIF1a, SlPIF1b, SlPIF3, SlPIF4, SlPIF7a, SlPIF7b, SlPIF8a and SlPIF8b. SlPIF1, SlPIF7 and SlPIF8 duplications occurred during the Solanum lineage polyploidization event 71 (± 19.4) MYA, prior to the divergence between tomato and potato species. Transcriptional profiling revealed coincident expression patterns between tomato SlPIF4 and Arabidopsis AtPIF4 and AtPIF5, highlighting the evolutionary conserved function of PIF4 clade. Combined evolutionary analysis and transcriptional profile data indicated that SlPIF7a and SlPIF7b may have suffered quantitative subfunctionalization that reduced their expression level, followed by neofuctionalization process, supported by the differential pattern of light responsive motifs and the positive selection signatures observed. Finally, SlPIF1a and SlPIF1b promoter regions showed differential pattern of light and fruit ripening transcriptional factor binding motifs, providing also evidences for regulatory sub-and neofunctionalization. In summary, our data underlined the importance of polyploidization events on PIF subfamily diversification.

Phylogenetic, Gene Divergence Time and Evolutionary Analyses
The amino acid sequences of the five A. thaliana canonical PIF proteins (S1 Table) were used as queries to perform a BLAST search against Viridiplantae in Phytozome [47], DNA Data Bank in Japan [48], Dendrome [49], SustainPine [50], Sol Genomics [25] databases. 119 sequences with complete bHLH domain from 16 species, representing liverworts, mosses, lycophytes, gymnosperms and flowering plants, were retrieved. T-Coffee Structural-Alignment algorithm [51] was used to perform an alignment of gymnosperms and flowering plants sequences. The phylogenetic reconstruction was performed by maximum-likelihood method using JTT substitution model and validated by approximate Likelihood Ratio Test (aLRT) with the Shimodaira-Hasegawa-like (SH-like) and 100 bootstrap replicates procedures available at PhyML Interface [52].
Gene divergence time was estimated using T = dS/2K equation, where T is the divergence time, dS is the pairwise synonymous distance calculated in the MEGA 6 software using the corrected Nei-Gojobori method (Jukes-Cantor) [53] and, K is the mean substitution rate estimated for 27 loci belonging to three different chromosomes of S. pennellii and S. lycopersicum ([28]; 4.38x10 -9 substitutions per site per year).
Evolutionary analysis was conducted individually for PIF1a, PIF1b, PIF3, PIF4, PIF7a, PIF7b and PIF8a genes using the sequences of S. lycopersicum, S. pennellii and S. tuberosum. Non-synonymous (dN) and synonymous (dS) distances and their SE values were estimated with MEGA 6. In order to preserve the reading frames, the alignment gaps were deleted prior to estimation of dS and dN. Codon bias was determined by the effective number of codons (Nc) value computed in the CodonW program [54]. Nc varies between 21 for maximum codon bias, when only one codon is used per amino acid, and 61 for minimum codon bias, when synonymous codons for each amino acid are used at similar frequencies. Three evolutionary models were evaluated using the Codeml program implemented in the PAML4.8a package and the graphical interface PAMLX 1.3.1 [55]. Phylogenetic trees were constructed using manually adjusted alignments of the coding sequences and neighbor-joining method with the optimal model of nucleotide substitution estimated by "Find Best DNA/Protein Model" using MEGA 6 software. To test for neutral evolution, the nearly neutral model (M1a) was compared with the null hypothesis, one ratio model (M0). To test positive selection, the model M2a was compared with M1a. The M0 model assumes that all codons across the sequences have the same level of dN/dS. The model M1a proposes that there two classes of codon, some with 0 dN/dS < 1 and the remainder with dN/dS = 1. Finally, model M2a divides codons into three classes: those with 0 dN/dS < 1, dN/dS = 1, and dN/dS > 1. The fit of model M1a versus M0 or M2a versus M1a is evaluated by a likelihood ratio test (LRT) comparing twice the difference in log likelihoods with a χ 2 distribution [56]. In M1a versus M0 and M2a versus M1a the degrees of freedom (df) are 1 and 2, respectively. Bayes empirical Bayes (BEB) analyses were performed to identify positively selected residues with a BEB posterior probability 95%.
Gene collinearity was addressed by BLASTN search against tomato genome [25] using the CDS sequences within a window of 100 Kb upstream and downstream the SlPIF1, SlPIF7 and SlPIF8 duplicated genes as queries.

Plant Material
All the experiments were performed with Solanum lycopersicum (cv. Micro-Tom). For deetiolation, diel cycle and dark-induced leaf senescence experiments, tomato seeds were surface sterilized and directly sown in vitro as described by Lira et al. [57]. After 120 h pre-germination in absolute darkness, seedlings were transferred to specific treatment conditions as described below. For deetiolation experiment, seedlings were either transferred to continuous white light (~100 μmol m -2 s -1 ) or maintained in absolute darkness for 0, 24, 48 and 72 h, after which cotyledons and hypocotyls were separately harvested. For daily cycle experiment, plants were grown under 12 h/12 h light/dark (~300 μmol m -2 s -1 ) photoperiod for three weeks and the second fully expanded leaves were harvested every 4 hours, for 24 hours. For dark-induced senescence, plants were grown in the same conditions as the daily cycle experiment and subsequently the plants were transferred to darkness for 0, 1, 2, 3 and 7 days for inducing leaf senescence. The second fully expanded leaves were harvested 4 h after the beginning of the light period since this day point has been shown to exhibits the highest mRNA levels of most tomato PIF genes. All experiments were conducted at 25 ± 1°C.
For fruit ripening experiments, plants were grown in 1 L pots in a greenhouse under automatic irrigation, at an average mean temperature of 25 ± 2°C, 11.5 h/13 h (winter/summer) photoperiod and 250-350 μmol m -2 s -1 PAR irradiance. Fruits at MG stage were harvested about 30 days after anthesis (dpa) and were transferred to continuous white light (400 to 800nm,~50 μmol m -2 s -1 ) or maintained under absolute darkness until reaching distinct ripening stages in a temperature-controlled growth chamber maintained at 25 ± 2°C and air relative humidity at 80 ± 5%. Top and bottom illumination was applied in order to homogenize the light environment surrounding the fruits. Since the beginning of the treatments, fruits were placed into a 0.5 L sealed transparent vessel and continuously flushed with ethylene-free, humidified air (approximately 1 L min -1 ) in order to avoid accumulation of ethylene inside the containers. Pericarp samples (without placenta and locule walls) were harvested at MG (displaying jelly placenta, 2 days after harvesting), BR (breaker), one day after BR (BR1), three days after BR (BR3), six days after BR (BR6), twelve days after BR (BR12) stages.
Seedling, plant and fruit tissues were harvested either under the specific light conditions used for the treatments or under dim green light (~0.01 μmol m -2 s -1 ), as appropriate. All samples were harvested and immediately frozen in liquid nitrogen, powdered and stored at -80°C.

Chlorophyll and Carotenoid Measurement
Chlorophyll and carotenoid extraction and analysis were carried out as described by Lira et al. [58]. When a data set showed homoscedasticity, an ANOVA test followed by a Tukey test (P< 0.05) was used to compare genotypes and fruit developmental stages. In the absence of homoscedasticity, a non-parametric ANOVA test was performed by applying the Kruskal-Wallis test (P< 0.05).

RNA Extraction and Quantitative Polymerase Chain Reaction (qPCR) Analysis
RNA extraction, cDNA synthesis and qPCR reactions were performed as described by Quadrana et al. [59]. The primers used for qPCR are listed in S4 Table. All reactions were performed with two technical replicates and at least three biological replicates. mRNA levels were quantified using a 7500 Real-Time PCR system (Applied Biosystem) and SYBR Green Master Mix (Applied Biosystem). Absolute fluorescence data were analyzed with LinRegPCR software [60] to obtain Ct values and to calculate primer efficiency. Expression values were normalized to the mean of two constitutively expressed genes: GAGA and CAC for seedlings [61], TIP41 and EXPRESSED for leaves and CAC and EXPRESSED for fruits [59]. A permutation test lacking sample distribution assumptions [62] was applied to detect statistical differences (P< 0.05) in expression levels between mutants and the control using the algorithms in the fgStatistics software package [63]. For senescence analysis, the normalized expression pattern was presented by a heat map constructed with GENE-E program [64].