A challenging question sometimes asked of biologists by children and physicists is: if leaves are solar panels, why are they green and not black [10-12]? This query in turn implies further questions about the nature of the earliest life-forms, how they used light as an energy-source, how the photosynthetic apparatus became organised and reorganised during evolution and (of particular relevance to the present paper) the evolutionary origin of pigment metabolism and its control in plant development. The light absorption profile of chlorophyll-based green membranes is characterised by a broad gap in the middle of the visible wavelength range. It has been proposed that the earliest phototrophs had light receptors that occupied the centre of the spectrum between the red and blue peaks of chlorophyll-based green membranes, and that chlorophyll evolved as a complementary pigment, filling in the missing wavelengths at the edges of the spectrum. According to this hypothesis the earliest photoautotrophs may have been similar to purple bacteria in using bacteriorhodopsin as the light-intercepting pigment-protein and bacteriochlorophyll at the reaction centre, driving a proton pump coupled to ATPase [10,13,14]. Phylogenetic analyses based on genes for photosynthesis as well as other markers such as rRNA and heat shock proteins are suggesting a number of alternative trajectories for the early evolution of pigment metabolism and organisation, but the widespread prevalence of horizontal gene transfer makes it difficult to agree the definitive scenario [13,15]. We can be certain, however, that isoprenoid-derived pigments continued to carry out the function of intercepting light in the middle of the visible spectrum while chlorophyll adopted an increasingly prominent role in light harvesting as well as charge separation, leading to modern streptophytes including land plants.

The significant early events in the evolution of the photosynthetic apparatus comprised diversification of reaction centre and antenna structure, based on chlorophyll and isoprenoids, and organisation and delineation of oxygenic and anoxygenic photosynthesis. According to Strother [21] 'the endosymbiotic origin of plastids...indicates that perhaps all of the original host cell lineages for the algae were phagotrophic...' Based on nutrient conditions and the evolution of food webs in the Palaeo- and Meso-proterozoic ocean [21,22], it is reasonable to suppose that facultative autotrophy would have been the rule in early photosynthetic organisms. The formation of endosymbiotic relationships resulting in the evolution of plastid-containing eukaryotic cells probably occurred just once in the lineage, although there is evidence of several secondary events, including horizontal gene transfer, too [23-25]. It seems likely that the foundation endosymbiont was mixotrophic in character and that the earliest green eukaryotes were facultatively autotrophic [21,26]. This is significant for the origin of plant senescence because aspects of the biochemistry and cell biology of plastids in terrestrial plants suggest that the transition from chloroplast to the plastids of senescing leaves (gerontoplasts) or fruit and floral parts (chromoplasts) is a change in status from auto- to hetero-trophy [27]. A strikingly senescence-like response during the change of trophic condition on removal of light and nitrogen source is seen in cultures of Auxenochlorella (formerly Chlorella) protothecoides. Cells turn yellow, accompanied by changes in protein and enzyme activities that resemble those of senescing mesophyll cells [28], and a red pigment is secreted into the medium [29-31]. Chemical analysis shows the pigment to be a bilin derived from chlorophyll by removal of phytol and Mg and by oxygenolysis of the methine bridge between pyrrole groups A and B [32]. This structure and its biochemical origin relate directly to the metabolic sequence for chlorophyll catabolism in angiosperms [33]. Similar production of chlorophyll catabolites has been reported in a Chlamydomonas mutant [34]. The luciferin of the bioluminescent dinoflagellate Pyrocystis lunula is also a bilin degradation product of chlorophyll, though in this case the chlorin macrocycle is opened between rings A and D [35]. A luciferin of similar structure is found in the krill organism Euphausia pacifica [35,36], presumably originating in ingested phytoplankton. We conclude that at least part of the enzymic pathway of chlorophyll catabolism was present during the aquatic phase of plant evolution and this has led us to seek phylogenetic continuity between genes for trophic responses and pigment catabolism from unicellular algae to angiosperms.

As well as promoting diversification of structure, and adaptation to terrestrial ecological niches, achieving landfall eventually exposed plants to new biotic and abiotic interactions. The ancient role of carotenoids as quenchers of excess light energy intercepted by photosystems [45] acquired new significance in environments where mismatch between photon flux, temperature and the availability of water and nutrients is the rule and a strongly oxidising atmosphere supports the uncontrolled propagation of damage by reactive chemical species. Coevolution between plant pigments and the visual systems of insects and other animals introduced a new role for carotenoids as signalling molecules, advertising to pollinators, to dispersers or to herbivores and other predators [46,47]. Investigations of the phylogenetic origin of carotenoids in relation to colour changes in senescence and ripening are frustrated by the ubiquity and antiquity of the corresponding genes and the inordinate degree of horizontal gene transfer apparent in the evolutionary record [48,49]. Nevertheless, we have identified four genes associated with the subcellular organisation and developmental regulation of carotenoid metabolism in senescing and ripening tissues that provide insights into how this aspect of senescence may have evolved. Lipid bodies (plastoglobules) accumulate during the transition from chloroplast to gerontoplast or chromoplast [50]. They have been observed in the plastids of Auxenochlorella during the switch from auto- to hetero-trophy [51]. Plastoglobule-like lipid bodies are very widely distributed amongst prokaryotes [52]. Existing and newly-synthesised carotenoids concentrated in plastoglobules are responsible for the colours of fruits and senescing leaves in some species [8,53,54]. We examined the distribution of sequences similar to PAP-Fibrillin, a plastoglobule protein with functions in plastid development and stress responses [55].

The observation of bilin products of chlorophyll catabolism in unicellular algae focuses attention on the enzymes responsible for opening the chlorin macrocycle, namely PaO and RCCR. Figure ​Figure22 is a Genevestigator heat map representation of the relative expression of all the Arabidopsis genes whose protein trees are reported in the present paper, including RCCR and PaO (AGI accession numbers respectively At4g37000 and At3g44880). Expression of RCCR is not particularly strong in senescing leaves. It gives a noticeably weak signal in roots and other non-green structures like pollen. PaO is also largely undetectable in non-green tissues but exhibits clear differential expression during development, being most prominent in senescing leaf, cauline leaf and sepal.

In a study of genes expressed during Arabidopsis leaf senescence, Hinderhofer and Zentgraf [44] identified Wrky53 as a transcription factor with a major coordinating role in senescence regulation. Phylogenetic analysis of the Wrky family by Zhang and Wang [74] established that Wrky53 is a member of Group 3 that appeared before divergence of monocots and dicots but later than diversification of the bryophytes 160 – 330 million years ago (mya). Wrky53 is expressed exclusively in mature and senescing leaves (Figure ​(Figure2),2), suggesting a role beginning in the earliest phase of senescence. Sequence similarities to the protein corresponding to Wrky53 gene At4g23810 (UniProt accession number NP_194112) were analysed and the resulting tree topology was in agreement with that of Zhang and Wang [74], with the exception of the appearance of a single hit in Chlamydomonas (Figure ​(Figure9).9). The E score for this protein is 2e^-11 and analysis of conserved domains reveals two Wrky superfamily sequences with E-values of 9.7e^-33 and 1.26e^-35 respectively (Figure ​(Figure9).9). We conclude that this represents a true algal Wrky53 homologue and is evidence that the appearance of Group 3 Wrkys may have been significantly earlier than the 330 mya originally inferred.

Plastids differentiate, dedifferentiate and morph from one type to another in a network of developmental transitions (Figure ​(Figure12;12; [75-78]). Orange and red carotenoids are unmasked, concentrated or synthesised de novo in gerontoplasts and chromoplasts during senescence and ripening [8] and are sequestered in plastoglobules that become abundant as chloroplasts transdifferentiate [55,79]. We examined the expression and sequence relationships of Arabidopsis PAP-Fibrillin (UniProt NP_192311, AGI At4g04020), a plastoglobule-associated protein [55]. Genevestigator profiling shows expression to be absent from roots and to be prominent in cotyledons, floral parts and leaves (Figure ​(Figure2).2). The protein tree for Fibrillin (Figure ​(Figure13)13) reveals dicot and monocot clusters and a conifer branch with a comparatively low bootstrap value. Eukaryotic unicellular green algae and cyanobacteria make a distinct separate group. The Ostreococcus sequence with a high E score appears to be a chimeric protein comprising two additional FHA (forkhead-associated, putative nuclear-signalling) domains, separated from each other and the Fibrillin domain by regions of low complexity (Figure ​(Figure1313).

Anthocyanin production and leaf yellowing are associated with senescence responses to nitrogen limitation, but are genetically independent [81]. We analysed the expression and protein sequence relationships of two genes with functions in anthocyanin biosynthesis. Bronze1 (Bz1) encodes the glycosyl transferase that synthesises the red vacuolar pigment cyanidin-3-glucoside. Zea Bz1 (UniProt P16167) was aligned with similar sequences; the homologous Arabidopsis gene is AGI At5g17050. The anthocyanin pathway is under the control of myb transcription factor C1 (UniProt accession number of Zea C1 P10290; AGI code of the homologous Arabidopsis gene At4g34990). Both genes are widely expressed in different organs and tissues (Figure ​(Figure2).2). Bz1 is most prominent in senescing leaf. C1 is also expressed in senescing leaf but is most strongly represented in stigma and hypocotyl tissues. The Bz1 tree reveals the occurrence of homologues in angiosperms, conifers and mosses (Figure ​(Figure16).16). At an E value of 1e^-08, only a single BLAST hit was recorded for a non-tracheophyte. The Chlamydomonas protein has a UDP glycosyl transferase region towards the C terminal end but low similarity to Bz1 elsewhere in the polypeptide. A BLAST search against the Chlamydomonas protein identifies other plant UDP glycosyl transferases, but E scores are high, and the algal protein aligns more closely with a range of nucleoside glycosyl transferases from metazoa. We conclude that this algal protein has a distant structural relationship to tracheophyte Bz1-like sequences but functions as a transferase in a different metabolic pathway from that leading to anthocyanin synthesis. It follows that the anthocyanin pathway appeared relatively late in the evolution of senescence processes, around the time of cryptogam diversification. By contrast, genes related to C1, the myb transcription factor responsible for regulating expression of anthocyanin synthesis, are well represented throughout plant taxa, from Zea to Ostreococcus (Figure ​(Figure17).17). BLAST also identifies similar genes in the actinomycete Frankia and other bacteria.

The Archaeozoic and Mesoproterozoic eons lasted from 3800-1000 mya. Although the detailed phylogeny of this period is difficult to infer from molecular analysis of contemporary prokaryotes [84], it is probable that the fundamental design of the photosynthetic apparatus was established early, in the form of pigment-protein photoreceptors driving an ATP-generating proton pump [13]. The ubiquity of genes for isoprenoid synthesis [49] is evidence of their ancient and fundamental role in light interception and photochemical quenching [45,85,86]. Genes for (bacterio)chlorophyll biosynthesis have been used to create well-rooted trees leading back to the putative earliest photosynthetic organisms [15]. Assembling complexes requires translational regulation [87] and efficient facilitated shape-matching [88], to coordinate the supply of structural elements. In modern autotrophs a major factor in achieving stoichiometry is post-transcriptional culling of excess unmatched components [89-91]. Thus in mutants deficient in pigment or protein components of photosystems, most of the genes responsible for building the complexes are transcribed and translated normally, but because they cannot make stable assemblages the proteins are degraded by fidelity-testing peptide hydrolases. For example, Fu et al. [92] showed that the cyclophilin AtCYP38 is required to stabilise thylakoid complexes in Arabidopsis: mutants deficient in CYP38 make normal amounts of photosystem structural components but these fail to assemble correctly and are turned over. Significantly, the CYP38 gene (At3g01480) is strongly down-regulated in senescing compared with mature and juvenile Arabidopsis leaves (data not shown). Xiong [15] described how a functional photosynthetic apparatus could have evolved in the earliest prokaryotes by multi-staged recruitment of reaction centre and antenna proteins with previously diverse evolutionary histories and functions. We consider that disassembly of complexes is correspondingly hierarchical in origin. Phylogenetic analyses of plant proteases [93] and of proteasome components and organisation [94] shows convincingly that the catabolic machinery for protein turnover is as ancient as the Chlorobacteria and Proteobacteria at the base of the evolutionary tree.

The origin of the chloroplast as a cyanobacterium-protozoan symbiosis occurred, according to most estimates, around 1500 mya or earlier [98], though there is an alternative view placing it as late as 600 mya [23]. The primary event soon gave rise to two lineages, the glaucophytes and the green plant/red algal group. Mixotrophy and opportunist organotrophy were probably well established in the earliest phototrophic prokaryotes and, in response to increasing competition for nutrients, vestiges of facultative trophic behaviour have persisted throughout the period from the enslavement of the endosymbiotic ancestors of eukaryote organelles to the plastid differentiation network of modern angiosperms (Figure ​(Figure12;12; [24]). The metabolism of green cells of angiosperms changes from auto- to hetero-trophic as the photosynthetic apparatus is remodelled during the transdifferentiation of chloroplasts into chromoplasts or gerontoplasts. Light is a declining source of energy in these cells and respiratory processes become increasingly significant [99]. Fruit ripening in many species, and leaf senescence in a few, is accompanied by a respiratory burst (climacteric; [100]). The amino acid products of protein recycling are subjected to transamination, deamination and amide synthesis to provide carbon skeletons for respiration while salvaging nitrogen in a transportable form [101]. A gene encoding an amino acid permease is upregulated when Auxenochlorella protothecoides is transferred from auto- to hetero-trophic culture conditions; the Arabidopsis homologue is also upregulated in senescing leaves (Figure ​(Figure2).2). The exact identity of the protease(s) responsible for releasing amino acids from photosynthetic proteins is not agreed [102]. A number of plastid-located proteases are known [90], amongst which at least four (ClpP3, ClpD, FtsH7 and FtsH8) are upregulated in senescence (Figure ​(Figure20).20). BLAST searches show that highly homologous sequences to all four proteins are widely distributed across taxa including phototrophic and non-phototrophic bacteria. There is evidence that Sgr is essential for making chlorophyll-binding proteins available to proteases [9,69,89]. Sgr is not detectable in photosynthetic prokaryotes, although Sgr-like sequences are present in some firmicutes, possibly as the result of horizontal transfer (Figures ​(Figures6,6, ​,7).7). We suggest that Sgr was recruited to the system for dismantling pigment-proteolipid complexes when subcellular topology became more complex with the appearance of the eukaryotic condition. This in turn suggests that protein mobilisation and colour change in tracheophyte senescence have their origins in the intraplastidic turnover of pigment-proteins related to regulated assembly of macromolecular complexes and trophic flexibility.

The Ordovician-Silurian period (450-410 mya) saw the rise of the charophytes, which are considered to link the unicellular chlorophytes with the first land-plants. The lack of sequence data for members of the Charophyta is a major gap in the phylogenetic record, but it is possible to identify features within charophyte algae likely to be relevant to senescence evolution, notably the elaboration of multicellular morphology, branching, 3-D cell division and interconnection of cells by plasmodesmata [43]. A key genus is Coleochaete, which includes branching filaments and thallus-like parenchymatous forms. Molecular phylogenies based on chloroplast and ribosomal sequences put this group close to the evolutionary line between unicellular chlorophyceans and bryophytes [114].

Green plants made landfall in the Silurian-Jurassic period, more than 410 mya. In this new and hostile environment, desiccation was avoided and ionic homeostasis maintained by the early development of an all-enveloping epidermal and cuticular 'space suit' [123] and a water-transport system [40,124]. The capacity for lysigeny and schizogeny was established in the first land plants, in connection with both the differentiation of vascular systems and the shedding of parts [124,125]. Along with mechanisms for smoothing out extreme variations in water-supply, it was necessary to establish physiological and developmental adaptations to deal with wide fluctuations in the inputs to photosynthesis. Many of the adjustments to the primary events of light capture and CO₂ assimilation within chloroplasts are regulated by systems present in Chlamydomonas and other unicellular chlorophyceans [45]. Atmospheric CO₂ concentration 410 mya is believed to have been around 3000 to 4000 ppmv [126]. This, combined with exposure to light fluxes no longer attenuated by the water-column, has led to the suggestion that the move to the terrestrial environment stimulated carbon fixation and utilisation to the point of incontinence – in the words of Harper [127] 'the green plant may indeed be a pathological overproducer of carbohydrates'. It allowed plants to develop profligate, throw-away lifestyles based on lysigeny and schizogeny on a vast scale. For example, primitive trees such as Archaeopteris (359-349 mya) probably controlled canopy morphology through abscission of lower megaphylls [128]. Abscission may have also been important in dehiscence of sporangia even before it became significant in vegetative shaping of plants [128], and a deciduous Glossopteris flora was already well established in Gondwanaland during the Carboniferous era [129]. Preserved senescent plastids have distinctive features that enable them to be identified in fossilised material [130]. Thomas and Sadras [131] discussed senescence as a strategy evolved by early land plants to deal with the promiscuous productivity of a photosynthetic machinery optimised for the relatively stable ancestral aquatic environment.

Angiosperms appeared in the Jurassic period, 150 mya. Carpels, flowers and tectate pollen occur in the fossil record at around this time and within 30 my angiosperms had become the dominant component of global floras. Before evolution of the angiosperms the colour world of vegetation would have consisted largely of greens, yellows and browns, much as it does in modern conifer-dominated forest ecosystems. A rapid increase in the profusion of natural pigmentation accompanied the Cretaceous explosion in angiosperm evolution and was likely to have been an adaptive response to a combination of abiotic and biotic factors [152,153]. Radiation of the first angiosperms into a wide range of new and often hostile environments was made possible by physiological and morphogenetic innovations or consolidations including, for example, the development of flexible photosynthetic mechanisms and the proliferation of architectures and life-forms [152,154]. Climate change may also have been a driving force at the onset of angiosperm diversification, since a transition between icehouse and greenhouse conditions occurred at this time [155].

BLAST, including protein BLAST (blastp), nucleotide BLAST (blastn, megablast), blastx, tblastn, tree viewer and BLINK precomputed BLAST

All protein trees reported here were generated using, as appropriate, BLASTP and FASTA for local alignments, ClustalW2 for global alignment, and TREE-PUZZLE, PUZZLEBOOT and the PHYLIP Phylogeny Inference Package [185,186] for phylogeny and bootstrap analysis.