Diurnal Rhythms in the Red Seaweed Gracilariopsis chorda are Characterized by Unique Regulatory Networks of Carbon Metabolism

Abstract Cellular and physiological cycles are driven by endogenous pacemakers, the diurnal and circadian rhythms. Key functions such as cell cycle progression and cellular metabolism are under rhythmic regulation, thereby maintaining physiological homeostasis. The photoreceptors phytochrome and cryptochrome, in response to light cues, are central input pathways for physiological cycles in most photosynthetic organisms. However, among Archaeplastida, red algae are the only taxa that lack phytochromes. Current knowledge about oscillatory rhythms is primarily derived from model species such as Arabidopsis thaliana and Chlamydomonas reinhardtii in the Viridiplantae, whereas little is known about these processes in other clades of the Archaeplastida, such as the red algae (Rhodophyta). We used genome-wide expression profiling of the red seaweed Gracilariopsis chorda and identified 3,098 rhythmic genes. Here, we characterized possible cryptochrome-based regulation and photosynthetic/cytosolic carbon metabolism in this species. We found a large family of cryptochrome genes in G. chorda that display rhythmic expression over the diurnal cycle and may compensate for the lack of phytochromes in this species. The input pathway gates regulatory networks of carbon metabolism which results in a compact and efficient energy metabolism during daylight hours. The system in G. chorda is distinct from energy metabolism in most plants, which activates in the dark. The green lineage, in particular, land plants, balance water loss and CO2 capture in terrestrial environments. In contrast, red seaweeds maintain a reduced set of photoreceptors and a compact cytosolic carbon metabolism to thrive in the harsh abiotic conditions typical of intertidal zones.


Introduction
Physiological and behavioral cycles fluctuate over 24 h, even in the absence of external stimuli.Diverse organisms including cyanobacteria (Piechura et al. 2017;Taton et al. 2020), animals (Whitmore et al. 1998;Jones et al. 2013;Ashbrook et al. 2020;Krylov et al. 2021), insects (Nitabach and Taghert 2008), and land plants (Harmer 2009) display circadian rhythms.Day and night transitions and photoperiodic oscillations determine a large proportion of physiological regulation thus it is particularly important in photosynthetic organisms (Brown et al. 2012;Dodd et al. 2013;de Dios and Gessler 2018).Physiological cycles and their regulatory mechanisms in algae and plants have largely been studied in model species such as Arabidopsis thaliana and the green algae Chlamydomonas reinhardtii and Ostreococcus tauri (Matsuo et al. 2008;Corellou et al. 2009;De Caluwe ́ et al. 2016;Linde et al. 2017).In contrast, very little is known about diurnal rhythms and their impacts on physiology in red algae (Rhodophyta), which with green and glaucophyte algae, comprise the photosynthetic members of the Archaeplastida.
The photoreceptors, phytochrome (PHY) and cryptochrome (CRY) play central roles in the input pathways of circadian oscillation based on light cues (Hughes et al. 2012;Lopez et al. 2021;Liu et al. 2022), however, the PHY gene family is absent in red algae (Duanmu et al. 2014).Therefore, the oscillatory mechanisms in algae and plants differ and some shared circadian genes may have evolved through convergent evolution (Matsuo and Ishiura 2011;Linde et al. 2017;Serrano-Bueno et al. 2017).CRYs are blue light-sensitive proteins that are homologs of DNA photolyases (ultraviolet-damaged DNA repair enzymes; Fortunato et al. 2015).Despite this shared ancestry, CRYs have diverse functions in signaling and interaction, regulating physiological cycles such as circadian and diurnal rhythms (Fortunato et al. 2015;Lopez et al. 2021).CRYs are classified into three major groups: plant, animal, and DASH (Drosophila-Arabidopsis-Synechocystis-Homo) families (Kianianmomeni and Hallmann 2014).
Plant CRYs (pCRYs), such as those found in Arabidopsis thaliana, regulate the expression of nuclear-encoded genes involved in photoresponsive regulatory mechanisms in green lineages, including photosynthesis (e.g.Calvin cycle-related genes), growth, and development, and life cycle progression (Ohgishi et al. 2004;Fortunato et al. 2015;Wang and Lin 2020).In Arabidopsis, more than 30 pCRY-interacting proteins have been identified (Wang and Lin 2020), and light-activated pCRYs directly interact with the complex of Constitutive Photomorphogenic 1 (COP1) and Suppressor of PHY A (SPA).This interaction inhibits the E3 ubiquitin ligase activity of the COP1/SPA complex which degrades light-inducible transcription factors for photomorphogenesis in the dark (Fortunato et al. 2015;Wang and Lin 2020).
In plants, myeloblastosis-like transcription factors, CIRCADIAN CLOCK-ASSOCIATED 1 (CCA1) and LATE ELONGATED HYPOCOTYL (LHY), which have peak gene expression at dawn and the most reduced at dusk, repress the transcription of clock-associated genes like FLAVIN-BINDING, KELCH REPEAT AND F BOX 1 (FKF1), a key regulator in photoperiodic flowering (Schaffer et al. 1998, Wang andTobin 1998;Imaizumi et al 2005;Sawa et al. 2007;Nakamichi 2020;Lopez et al. 2021).REVEILLE8 (RVE8), homologous to CCA1/LHY, also shows a gene expression peak at dawn and acts as an activator of flowering along with the NIGHT LIGHT-INDUCIBLE AND CLOCK-REGULATED (LNK) genes (Farinas and Mas 2011;Nakamichi 2020).Homologs of the transcription factor CCA1/LHY are present in green algae, although little is known about the associated genes (e.g.FKF1 and LNKs) in these species (Linde et al. 2017;Serrano-Bueno et al. 2017).
Central carbon metabolism (e.g.glycolysis/gluconeogenesis) is an intermediate step in both photosynthetic carbon assimilation and energy metabolism (Walker et al. 2021).Pyruvate-related metabolism is important not only for glycolysis/gluconeogenesis, but also for cytosolic carbon (e.g.carbon concentration, and interconversion of C 3 to C 4 organic acids) and mitochondrial metabolism.Phosphoenolpyruvate (PEP) carboxylase (PEPC; PEP + HCO 3 − → OAA + Pi) plays crucial roles in carbon and nitrogen metabolism, including the intermediate step in the TCA cycle (Chollet et al. 1996;Shi et al. 2015).Carbon flow is closely linked to mitochondrial energy (i.e.adenosine triphosphate [ATP]) metabolism via oxidative phosphorylation, under both light and dark conditions (Raghavendra and Padmasree 2003).The cycling of mitochondrial energy production is directly related to the response to reactive oxygen species (ROS) (de Goede et al. 2018).ROS production in photosynthetic cells can be managed by "malate circulation" between plastids and mitochondria (Zhao et al. 2020), as well as through cell cycle controls (Miyagishima et al. 2014).
There are many different types of metabolite transporters involved in photosynthesis and energy metabolism in algae that have diversified via serial endosymbiosis (Facchinelli and Weber 2011).Some of these genes adapted to the limiting light levels and low CO 2 gas diffusion rate in aquatic environments (Raven 1970;Zeebe 2011).These differences have resulted in divergent physiological responses in algae with respect to genes, photoreceptors, and their modulators (Noordally and Millar 2014).However, red algal systems are poorly studied even though a large proportion of photosynthetic eukaryotes (i.e.cryptophytes, haptophytes, stramenopiles, and apicomplexans) contain a red-alga-derived plastid (Bhattacharya et al. 2004).Here, we studied gene expression over the diurnal cycle in the red seaweed Gracilariopsis chorda (class Florideophyceae) and described CRY-derived regulation and photosynthetic/ cytosolic carbon metabolism in this species.We find that G. chorda has a compact and efficient energy metabolism that is distinct from C 3 , C 4 , or Crassulacean acid metabolism (CAM) in plants.We postulate that rhythmic regulation of physiological cycles in eukaryotes could plausibly originate via multiple, independent horizontal gene transfers (HGTs).

Gracilariopsis chorda
To study the rhythmic regulation of physiological cycles in G. chorda, algal samples were exposed to a day/night (DN) Lee et al. • https://doi.org/10.1093/molbev/msae012MBE cycle for 24 h (12 h-light:12 h-dark) and thereafter, continuous light (CL) for 24 h.We collected algal samples each 6 h from DN (DN4, DN10, DN16, DN22) to CL (CL4, CL10, CL16, and CL22) conditions for RNA-seq analysis (data available at SRR21594546-SRR21594568; details in Methods).The DN samples can be used to study diurnal gene expression that underlies diverse physiological traits in G. chorda.The CL samples comprise two periods as follows: (i) CL4 and CL10, which provides additional insights into light-based cues for physiological cycles after the DN period (Hughes et al. 2012;Liu et al. 2022), and (ii) CL16 and CL22, which abolishes the DN transition and can provide insights into the expression of genes that are independent of the initial dark period.Based on the G. chorda gene expression patterns, we identified 3,098 rhythmic genes in this red seaweed (supplementary table S1, Supplementary Material online).
Rhythmic gene expression patterns were divided into 12 major and two minor patterns that show diurnal fluctuation (Fig. 1).The up-and down-regulated genes from DN4 are shown with "+" and "−", respectively, which indicate relative gene expression patterns compared to the previous time point (see Materials and Methods).Based on these results, we defined morning-phased (a-b-c; peak at DN4), dusk-phased (d-e-f; DN10), evening-phased (g-h-i; DN16), and dawn-phased (j-k-l; DN22) rhythmic genes (Fig. 1).However, the trough time was different within each group; the trough time in pattern-a is at DN10, which is earlier than the others (i.e.pattern-b at DN16, and pattern-c at DN22) within morning-phased group (Fig. 1).Therefore, patterns a, d, g, and j show gradual increase and then rapid decrease in gene expression.In contrast, patterns c, f, i, and l show rapid increase and gradual decrease.The patterns b, e, h, and k show a moderate cycle of increase and decrease in gene expression.We postulate that gene expression patterns are driven by sequential regulation (e.g.feedback loop) of rhythmic genes in functional groups.It is possible that the rhythmic genes we have identified in G. chorda may include circadian oscillators, but our current data do not allow us robustly to test this idea.In other words, to characterize true circadian gene expression in G. chorda, additional sampling from CL conditions for at least 24 h, which are not part of the previous diurnal cycle, is required.The minor patterns are indicated as fluctuating rhythmic gene expression and the peaks are present before (pattern-m) or after (pattern-n) both DN and night/day transitions (Fig. 1).These genes are potentially involved in preparation for the transitional points in the G. chorda cycle.Additional experiments are needed to address this hypothesis.Therefore, we focus here on diurnal rhythms in G. chorda.
To study the role of these CRYs in G. chorda biology, we examined their expression under diurnal and CL conditions.We found that 5/8 CRY genes and key photosynthetic genes in G. chorda are under rhythmic regulation (Fig. 2, supplementary table S2, Supplementary Material online).In G. chorda, only Gc-pCRY2 among the two pCRY genes exhibits rhythmic gene expression (Fig. 2) belonging to a morning-phased gene group (pattern-B in Fig. 1).This pattern is similar to the rhythmic expression profile of plastidtargeted light-harvesting proteins (e.g.chlorophyll-a/b binding and phycobilisome linker genes; supplementary table S2, Supplementary Material online).In the diatom, Phaeodactylum tricornutum, 2 members of the CRY photolyase family (CFP1, and CRYP) participate in the lightdependent expression of photosynthetic light-harvesting genes and photoprotection (Coesel et al. 2009;Juhas et al. 2014;Taddei et al. 2016;Allorent and Petroutsos 2017).It is well-established that pCRYs stimulate transcription of nuclear-encoded plastid-targeted proteins in diatoms and land plants (Thum et al. 2001;Ichikawa et al. 2004;Noordally et al. 2013;Fortunato et al. 2015).Using a heterologous expression system (A.thaliana), the product of the Gc-pCRY2 gene in G. chorda was nuclear localized (Fig. 2).These results suggest that the expression pattern of Gc-pCRY2 gene is also potentially involved in regulating nuclear-encoded plastid-targeted proteins during the day (Fig. 2, supplementary table S2, Supplementary Material online).
The GcDASH-CRY1, GcDASH-CRY3, and GcDASH-CRY4 genes in G. chorda show rhythmic expression (Fig. 2).Interestingly, two of them (GcDASH-CRY1 and GcDASH-CRY3) exhibit an out-of-phase gene expression pattern (pattern-h) compared to Gc-pCRY2 (pattern-b) in G. chorda (Fig. 2).The product of one of these genes, GcDASH-CRY3, shows plastid localization in the heterologous system (Fig. 2).Therefore, we propose that a diurnal rhythm exists for this plastid-targeted GcDASH-CRY3 gene, with maximum expression in the evening phase.This pattern may be explained by the repression of photosynthetic activity in the dark period, as observed in Chlamydomonas reinhardtii CRY-DASH1 (Rredhi et al. 2021).This mechanism may allow photosystem repair, similar to cyanobacterial DASH-CRY (Syn-CRY) that is involved in PSII repair (Vass et al. 2014).

MBE
By contrast with the plastid localization of the GcDASH-CRY3 gene product, the GcDASH-CRY1 and GcDASH-CRY4 gene products are localized to the cytoplasm (Fig. 2).Based on their expression patterns, these genes may also function as photoreceptors at dawn.This suggests that darkness may be required to sustain the "dark phase" of expression.Cytoplasmic DASH-CRYs in G. chorda could therefore be involved in rhythmic regulation through diverse protein interactions in the cytoplasm, like other known CRY families that play important roles in signal transduction of cellular metabolism (Wu and Spalding 2007), suppression of nuclear-encoded genes (Chaves et al. 2011;Lee et al. 2015), and protein-protein interactions (Yoo et al. 2013;Damulewicz and Mazzotta 2020) in the cytoplasm.Owing to the similar gene expression pattern for plastid-targeted GcDASH-CRY3 and cytoplasmic GcDASH-CRY1, we propose that these proteins function at dawn in G. chorda.By contrast, gene expression of GcDASH-CRY2 is nonrhythmic, although it appears to increase when the cells are exposed to light (i.e. the dusk-phase; Fig. 2).Unfortunately, the subcellular localization of this protein is difficult to determine using heterologous expression due to protein aggregation in the cytoplasm (Fig. 2).Hence, the regulatory activity of GcDASH-CRY2 remains unclear in G. chorda.Our data suggest that the DASH-CRY family may function in both the plastid and cytoplasm which are likely to be involved in coordinating photosynthetic function in the plastid with nuclear gene transcription and regulation of protein functions in the cytoplasm at daybreak and mostly throughout the light phase.In comparison, the nonrhythmic Gc-aCRY and the rhythmic HGT-derived Gc-hCRY gene products show nuclear localization in the heterologous system (Fig. 2), thus, they too likely regulate nuclear-encoded genes.
A Role for COP1 in pCRY Signaling in G. chorda Although the origin of PHYs and CRYs remains controversial in eukaryotes, genes for both photoreceptor families were likely present in the common ancestor of Archaeplastida (Asimgil and Kavakli 2012;Duanmu et al. 2014;Rockwell andLagarias 2017, 2020).Given the lack of PHYs in red algae and in their heterotrophic ancestor Rhodelphidia (Rockwell and Lagarias 2020), it is not surprising that their signaling components are reduced in red algal genomes (Han et al. 2019).Surprisingly, most of the CRY-interactive components and typical circadian core oscillator genes found in the land plants are absent in Rhodophyta except for the COP1 gene and CCA1/LHY-RVE8 homologous genes (supplementary table S3  Diurnal Rhythms in the Red Seaweed Gracilariopsis chorda • https://doi.org/10.1093/molbev/msae012

MBE
COP1 is an E3 ubiquitin ligase involved in degradation of transcription factors (e.g.BBX21, HY5, and HYH) in darkness (Xu et al. 2016).In G. chorda, COP1 (PXF44200.1),whose expression increases rapidly at dawn and then gradually falls (pattern-l; supplementary table S3, Supplementary Material online).By comparison, COP1 gene expression in plants is not light regulated.Plant COP1 function is inhibited by photoactivated CRY and is active in darkness to degrade factors that promote photomorphogenesis (Wang et al. 2001;Holtkotte et al. 2017;Kim et al. 2017;Wang and Lin 2020).Assuming a conserved function for G. chorda plant-type CRYs, the rhythmic patterns of COP1 (PXF44200.1)and pCRY (PXF48820.1)expression would lead to an analogous light-dependent suppression of COP1 function in G. chorda, while also accounting for the increase in G. chorda COP1 gene expression following inactivation of pCRY in darkness (Fig. 2; supplementary table S3, Supplementary Material online).In plants, this process is mediated by COP1-interacting "suppressor of PHYA-105" (SPA) proteins (Holtkotte et al. 2017;Wang and Lin 2020;Ponnu and Hoecker 2021), which lacks supporting evidence in G. chorda (supplementary table S3, Supplementary Material online).These results suggest that the interplay between CRY and COP1 by light signaling and the physiological cycle differs in plants and red algae, but their signaling output may be similar.
Two genes encoding transcription factors (PXF45473.1 and PXF40765.1;BLASTp top hits, e-value cutoff = 1.e-05) in G. chorda which have partial sequence similarity to CCA1/LHY (AT2G46830/AT1G01060) and RVE8 (AT3G09600) show gene expression peaks at dawn (supplementary fig.S2, Supplementary Material online).CCA1/LHY and RVE8 repress and activate, respectively, the transcription of FKF1, a crucial regulator of photoperiodic flowering in Arabidopsis (Schaffer et al. 1998, Wang andTobin 1998;Imaizumi et al 2005;Sawa et al. 2007;Nakamichi 2020;Lopez et al. 2021).The FKF1 (AT1G68050) homolog (PXF45752.1;BLASTp top hit, e-value cutoff = 1.e-05) shows a reversed gene expression pattern to the dawn-peaked transcription factors in G. chorda (supplementary fig.S2, Supplementary Material online).Consequently, we propose that these transcription factors suppress the transcription of the FKF1 homologs in G. chorda.In addition, there were no genes homologous to LNKs (AT5G64170 and AT3G54500; BLASTp search, e-value cutoff = 1.e-05) in G. chorda.LNKs serve as interacting partners of RVE8 to activate FKF1 in Arabidopsis (Rawat et al. 2011;Fogelmark and Troein 2014;Gray et al. 2017;Nakamichi 2020).Therefore, the dawn-peaked transcription factors in G. chorda appear to function similarly to CCA1/LHY.However, the roles of these transcription factors, their interacting partners, and the associated transcription-translation feedback loops remain unknown in the developmental processes of red seaweeds.A more in-depth study, employing diverse molecular and genetic approaches targeting homologs of CCA1/LHY and RVE8 in G. chorda, will be necessary to address these shortfalls.

HGT-derived Rhythmic Genes in G. chorda
HGTs lead to increased genetic and functional diversity in many eukaryotic genomes (Husnik and McCutcheon 2017;Van Etten and Bhattacharya 2020;Pereira et al. 2022).We analyzed rhythmic genes in G. chorda to identify candidates that originated via HGT (supplementary table S4, Supplementary Material online).One such example is the HGT-derived Gc-hCRY gene (Fig. 2).Other examples include two 2-isopropylmalate synthase (K01649) genes (PXF49417.1 and PXF49429.1),a tyrosyl-tRNA synthetase (K01866) gene (PXF50084.1),and a chlamydia-derived ABC transporter gene (PXF50084.1)that encodes a NitT/ TauT family transport system permease (K02050).The latter may have been acquired during primary endosymbiosis because this chlamydia-derived gene is found in red algae as well as in the green lineage (supplementary table S4, Supplementary Material online).Whereas species-or lineage-specific gene transfers can provide novel functions related to nutrition, protection, and adaptation to extreme environments, most transferred genes are nonfunctional (Huang et al. 2004;Nowack et al. 2008;Schönknecht et al. 2013;Wybouw et al. 2014;Husnik and McCutcheon 2017;Rossoni et al. 2019;Lhee et al. 2020).We postulate that such HGT-derived genes could be independently and gradually established through diverse evolutionary processes, likely at the population/species-level (e.g.pangenome concept; Fan et al. 2020; Lee 2021) to the lineage-level (e.g.endosymbiotic gene transfer).Most of the prokaryotic HGT-derived genes in G. chorda have unknown functions (supplementary table S4, Supplementary Material online), therefore their potential role in host rhythmic gene expression remains a goal of future studies.
C 3 Cycle, and Glycolysis/Gluconeogenesis in G. chorda The C 3 (Calvin-Benson-Bassham) cycle corresponds to the light-independent chemical reaction of photosynthesis that fixes carbon dioxide via reductive generation of glyceraldehyde-3-phosphate (G3P).The C 3 cycle consists of three phases: (i) carbon fixation, (ii) reduction, and (iii) regeneration (Gurrieri et al. 2021).During carbon fixation, ribulose-1,5-bisphosphate carboxylase-oxygenase (RuBisCo) catalyzes carbon dioxide-dependent carboxylation of ribulose-1,5-bisphosphate (RuBP) to generate two molecules of glycerate-3-phosphate (3-PGA).Transcription of RuBisCo subunits in G. chorda increased during the light phase, and decreased during the dark phase (supplementary fig.S3, encoding RuBisCo subunits could be generally induced by light availability, but these patterns are not consistent (i.e.fluctuating or nonrhythmic) with species-specific patterns throughout the diel cycle (Recuenco-Mun õz et al. 2015; Perdomo et al. 2021).The distinct pattern of RuBisCo expression, i.e. gradual increase in the light period and decrease in darkness, likely reflects the fact that both subunits of this enzyme are plastid-encoded in red algae (supplementary fig.S3, Supplementary Material online).This pattern is unlike the case in the green lineage where the small subunit is nuclear-encoded (Lee et al. 2016).In addition, the discordance between transcript and protein abundance of RuBisCo subunits has been reported and reflects post-transcriptional regulation (Recuenco-Mun õz et al. 2015; Perdomo et al. 2021).Therefore, in this study, we focused on carbon metabolism downstream of RuBisCo over the diurnal cycle.
Genes for cytosolic carbon storage pathways in G. chorda, from hexose-phosphates to floridean starch and floridoside, primarily exhibit morning-phased expression, whereas genes for their breakdown for energy metabolism exhibit dusk-to dark phase maxima (supplementary fig.S3, Supplementary Material online).Such patterns of C 3 cycle and starch synthesis/breakdown-related gene expression mirror those in the green lineage.The cytoplasmic localization of starch synthesis and breakdown-involved genes in red algae is a striking difference from the plastid localization of these processes in the green lineage (Viola et al. 2001;Patron and Keeling 2005;Ball et al. 2011).Presumably, this allows for more rapid mobilization of stored energy reserves, which may be more critical for red algae, and provides a point of divergence for the regulation of these pathways between these two Archaeplastida lineages.
Genes involved in the glycolysis/gluconeogenesis pathways in G. chorda typically show dusk-to dawn-phased maxima with out-of-phase of those of the C 3 cycle and light-harvesting genes (supplementary fig.S3 and table S6, Supplementary Material online).Plastidial glycolysis genes exhibit a transcriptional peak at the evening phase (supplementary fig.S3 and table S6, Supplementary Material online).Gluconeogenesis is a key interface between organic acid and sugar metabolism-a process that generates glucose from oxaloacetate (OAA) in plants (Leegood and Walker 2003;Walker et al. 2021) Regulation of Cytosolic and Mitochondrial Carbon Metabolism in G. chorda PEPC is highly expressed at the dusk-phase in G. chorda (pattern-e; Fig. 3), and likely plays an important role in the induction of cytosolic OAA (Shi et al. 2015;Walker et al. 2021;Shu et al. 2022).This enzyme is supported by carbonic anhydrase (CA) which catalyzes the interconversion of carbon dioxide to bicarbonate (CO 2 + H 2 O ←→ HCO 3 − + H + )-a major CO 2 -concentrating mechanism (CCM) in G. chorda (Fig. 3, supplementary table S8, Supplementary Material online; Badger and Price 1994;Lindskog 1997).The pattern of gene expression in G. chorda is consistent with primary carbon fixation by C 3 cycle morning-phased genes in plastids followed by secondary carbon fixation in the cytosol with PEPC at dusk and in the early evening.For sequential carbon fixation to function, proper regulation of the cytosolic CO 2 and/ or bicarbonate ions (HCO 3 − ) concentrations is important and both cytosolic nicotinamide adenine dinucleotide phosphate-malic enzyme (cytoNADP-ME) and mitochondrial nicotinamide adenine dinucleotide-malic enzyme (mitoNAD-ME) are critical for both types of carbon fixation.Whereas cytoNADP-ME performs nonphotosynthetic functions (e.g.stress responses, cytosolic pH regulation, and TCA cycle metabolism) in C 3 and C 4 plants, cytoNADP-ME is utilized by CAM plants for CO 2 release in a process analogous to that performed by the plastid NADP-ME in C 4 plants.Both reactions produce CO 2 which is fixed by the C 3 cycle enzyme RuBisCo (Drincovich et al. 2001;Lai et al. 2002;Wheeler et al. 2005;Maier et al. 2011;Tronconi et al. 2018;Chen et al. 2019).We postulate that cytosolic CO 2 produced by cytoNADP-ME supports carbon fixation in both the C 3 cycle (morning) and the CCM (dusk) because cytosolic CO 2 during daytime can easily diffuse into plastids.In addition, mitoNAD-ME can elevate cellular CO 2 concentration (Fig. 3).Even though the major function of mitoNAD-ME is in the TCA cycle, CO 2 diffusion likely contributes to plastid CO 2 fixation in both C 4 and CAM plants (Hatch and Kagawa 1974;Artus and Edwards 1985;Tronconi et al. 2018) and, by analogy in G. chorda.In the C 3 A. thaliana, however, mitoNAD-ME genes are highly active during night (Tronconi et al. 2008(Tronconi et al. , 2018)).Therefore, the daytime reaction of mitoNAD-ME in G. chorda could induce CO 2 diffusion into the cytosol, and this contributes to both types (RuBisCo and PEPC) of light-activated carbon fixation like cytoNAD-ME (Fig. 3, supplementary fig.S3, Supplementary Material online).The TCA cycle (active at dusk; supplementary fig.S4, Supplementary Material online) also generates CO 2 thus it could be simultaneously recycled by PEPC in the cytosol.
The loss of PEPCK in red algae may indicate that it is unnecessary to supply additional CO 2 for photosynthetic carbon fixation as in C 4 plants, because of the bicarbonate-rich condition (about 90% of inorganic carbon sources) in most marine environments (Poschenrieder et al. 2018), which could be captured by PEPC.Moreover, the addition of sodium bicarbonate promotes growth in red seaweeds and Diurnal Rhythms in the Red Seaweed Gracilariopsis chorda • https://doi.org/10.1093/molbev/msae012MBE photosynthetic activity (Zhou et al. 2016).In contrast, the availability of water and HCO 3 − in land plants is frequently limiting in terrestrial environments.In addition, CO 2 availability could be limited by closed stomata in land plants under high temperatures and drought (Poschenrieder et al. 2018), thus land plants may have evolved enhanced photosynthetic carbon fixation in chloroplasts to inhibit photorespiration: e.g.C 4 -type photosynthesis (Sage et al. 2012).Photorespiration in aquatic algae could also occur under CO 2 deficiency but HCO 3 − availability by CA reduces photorespiration (Li et al. 2021).Although cytosolic PEPC is present both in red algae and land plants, the differences in availability of CO 2 and HCO 3 − could lead to divergent strategies for carbon metabolism.Algal species use inorganic carbon transport (not stomata), thus inorganic carbon is more accessible to these taxa (Raven and Beardall 2016;DiMario et al. 2017;Razzak et al. 2018).Aquatic green algae are similar, but they have chloroplast-centralized carbon metabolism, and energy metabolism is active under dark conditions, as in land plants (Zones et al. 2015).Therefore, we suggest that metabolic processes and cellular structures underpinning carbon metabolism in red algae and the green lineages have diverged significantly (Fig. 4).
Based on these results, we postulate that ancient carbon metabolism in the major red algal lineage Florideophyceae evolved independently as a compact (e.g.CRYs) and efficient system.This is in contrast to the spatial (pyruvate and malate synthesis in mesophylls and bundle sheath in C 4 plants) or temporal (carbon fixation at the daytime and dark energy metabolism in most C 3 /C 4 /CAM plants) gap to enhance photosynthesis in land plants.The evolutionary trends in carbon metabolism in red algae are quite different from the systems in C 3 , C 4 , or CAM plants.

Comparative Models of Photoreceptors and Carbon Metabolism Between Land Plants and Red Seaweeds
The central input pathways of circadian oscillation are controlled by light cues primarily mediated by the photoreceptors PHY and CRY in green plants (Lopez et al. 2021;Liu et al. 2022).PHYs are absent in red algae (Duanmu et al. 2014), therefore, CRYs, that are responsive to blue light provide the primary input to the oscillatory mechanism.These light wavelengths penetrate more deeply in water than does red light.For this reason, the CRY-based input system appears to be sufficient for G. chorda to flourish in blue light-enriched subtidal marine environments that have (depth-dependent) lower incident light levels.By contrast, green algae and plants are better adapted to shallower aquatic and terrestrial environments, where red wavelengths are more abundant (Fig. 4a), accounting for their retention of PHY sensors.For these reasons, the role of light sensing in driving physiological cycles in red algae is likely to differ significantly from that found in the green lineage.
Given this hypothesis, we did RNA-seq experiments to identify rhythmic genes and the role of light in G. chorda biology.Quantitative real-time PCR (qRT-PCR) analysis of gene expression patterns for most of the target genes provides results that are consistent with the RNA-seq data (See Methods).Based on the RNA-seq results, we identified 3,098 rhythmic genes in this red seaweed (Fig. 1, supplementary table S1, Supplementary Material online).Among genes of particular interest, were those involved in light sensing and photosynthesis.We detected rhythmic gene expression patterns in 5/8 CRYs (i.e. one pCRY, three DASH-CRYs, and one HGT-derived CRY-like) and many genes involved in photosynthetic/cytosolic carbon metabolism (Figs. 2 and 3, supplementary fig.S3 and  S2, Supplementary Material online).PHY-and CRY-based modulation of the oscillatory mechanism in the Archaeplastida evolved independently of lightmediated regulation of the kai family-based clock found in the cyanobacterial ancestor of plastids (Kondo and Ishiura 2000).Upon the loss of PHYs in red algae, rhythmic gene expression of pCRY and DASH-CRYs became associated with photosynthesis.The organization of these pathways is however likely to fundamentally differ from that in green lineage because highly conserved PHY-and CRY-interacting factors in plant species are absent in G. chorda (supplementary table S3, Supplementary Material online).Therefore, the diurnal cycle and its role in red algal physiology also diverged in a lineage-specific manner through processes such as gene loss, gene family expansion, and HGT (supplementary table S4, Supplementary Material online).
The green algae and most land plants retain the ancestral form of carbon metabolism, i.e. the C 3 pathway, present in cyanobacteria (Facchinelli and Weber 2011).However, some land plant lineages have independently evolved specialized forms of photosynthetic carbon metabolism (e.g.C 4 , and CAM) as an adaptation to CO 2 limitation (Fig. 4a).Mesophyll and bundle sheath cells allow C 4 plants to overcome CO 2 gas exchange limitation and reduce the rate of photorespiration (Wang et al. 2014).In marine environments, HCO 3 − is more abundant than dissolved CO 2 , owing to the high pH (∼ pH 8.1) of seawater (Fig. 4a).For this reason, CAs perform essential roles in marine species, not only in the interconversion of carbon sources (CO 2 + H 2 O ←→ HCO 3 − + H + ) for photosynthesis, but also in diverse forms of carbon metabolism (e.g.balancing cellular pH levels; DiMario et al. 2017).
Diurnal Rhythms in the Red Seaweed Gracilariopsis chorda • https://doi.org/10.1093/molbev/msae012MBE of CO 2 in G. chorda relies on malic enzymes (MEs) that feed photosynthetic carbon fixation and reduce photorespiration.(ii) HCO 3 − (and CO 2 via carbonic anhydrases, including remnants from photosynthetic carbon fixation) is captured by PEPC (rhythmic gene expression) at dusk (supplementary fig.S3, Supplementary Material online).The synthesis of OAA by PEPC could directly feed the TCA cycle in red algae during the dusk-phase.(iii) Pyruvate kinases (PEP → pyruvate) and the TCA cycle are also active at dusk, therefore carbon flow from C 3 organic acids (i.e.PEP) to energy metabolism in G. chorda is directly associated during the daytime, whereas energy metabolism in the green lineages is active in the dark (Figs. 3 and 4b).In addition, cytosolic CO 2 diffusion from the TCA cycle (dusk) and NAD-ME (daytime) could be recycled by photosynthetic (C 3 cycle in the morning) and cytosolic (PEPC at dusk) carbon fixation.Dusk activation of genes involved in mitochondrial pyruvate metabolism (e.g.pyruvate dehydrogenases and mitochondrial pyruvate carriers) and oxidative phosphorylation, including the F-type (mitochondria/chloroplast) ATPases (Kühlbrandt 2019), also support increased energy metabolism at dusk in G. chorda (Fig. 3, supplementary fig.S5, Supplementary Material online).Therefore, G. chorda exhibits a compact cytosolic carbon metabolism, whereby photosynthetic carbon fixation and energy metabolism are tightly coupled in daylight (Fig. 4b).(iv) One-way gluconeogenesis (from OAA to glucose) in G. chorda is controlled by the day (PEPC/NADP-ME) and night (PPDK/ MDH) cycle, which is closely related to efficient carbon recycling in G. chorda (Fig. 3).

Conclusion
The green lineage, in particular land plants, evolved a system to balance water loss and CO 2 capture in terrestrial environments using spatial or temporal isolation of pyruvate and malate via the C 4 or CAM system, respectively.In contrast, red algae maintain a reduced set of photoreceptors and a compact cytosolic carbon metabolism to exploit limited light and carbon (i.e.∼10,000 times lower CO 2 gas diffusion rate than in air; Raven 1970; Zeebe 2011) conditions in aquatic environments (Fig. 4b).We note that our results and hypotheses regarding diverse enzymatic processes in red seaweeds are based solely on transcriptome analysis.Given that post-transcriptional regulation is common in eukaryotes and often unlinks gene expression from protein accumulation (Fang et al. 1998;Yu 2007;Yu et al. 2021), our data need to be interpreted with some caution.In addition, the putative functions of rhythmic genes and the encoded proteins we have identified are based on inferences using other model systems and are therefore provisional in nature.Nevertheless, our results provide novel insights into the regulation of physiological cycles that can be tested using other omics and genetic approaches.In addition, industrial applications using carbon metabolism in the red seaweeds (e.g.CO 2 hydration; Razzak et al. 2020) may be used to mitigate carbon emissions.More generally, our study suggests that gaining a deeper understanding of red algal carbon metabolism may contribute to the understanding of ecological carbon fluxes and storage in marine ecosystems (e.g.blue carbon; Krause-Jensen et al. 2018;Macreadie et al. 2019).

Preparation of Algal Samples and RNA Sequencing
The thallus of G. chorda was collected from an aquaculture farm at Jangheung (South Korea) and washed several times with sterile seawater to remove surface contaminants.The seaweed was cut into 10 cm fragments, pre-cultured in L1 medium (L1 media kit, https://ncma.bigelow.org/MKL150L) for two weeks.Algal cultures were maintained at 15 °C under a 12L/12D photoperiod (12:12 h light and dark cycle) with 1,500 lux light intensity.To examine rhythmic gene expression in G. chorda, algal samples were exposed to a DN cycle for 24 h (12 h-light:12 h-dark) and then CL for 24 h.We collected samples after 4 h-light exposure (DN4) initially to analyze morning-phased gene expression, and then each 6 h in DN (DN10, DN16, and DN22) and CL conditions (CL4, CL10, CL16, and CL22).For example, DN10 and DN16 designated the experimental condition of 10 h-light exposure and 12 h-light/ 4 h-dark exposure, respectively.CL4 indicated 4 h-CL exposure after the DN period for 24 h.Algal samples were prepared in triplicate (except for one condition; we used duplicate sequencing data of CL16 because of failed sequence library construction) and stored at −80 °C prior to RNA extraction.
Total RNA was extracted from 3 different independent samples in each condition using the RNeasy Plant Mini Kit (QIAGEN, Germany).The quality of purified RNAs was determined using a 2100 Bioanalyzer (Agilent Technologies, USA).RNA sequencing libraries were constructed using the TruSeq RNA Sample Prep Kit (Illumina, USA).Libraries were sequenced using 100 bp paired-end reagents on the Illumina HiSeq2500 platform (Illumina, USA).All experiments were done using the manufacturer's instructions.The quality scores of RNA-seq reads (average Q20 = 97% and Q30 = 92%) were analyzed using Fastp (v0.23.4;default options;Chen et al. 2018;Chen 2023).

Gracilariopsis chorda
The RNA-seq reads (NCBI SRA database; SRR21594546-SRR21594568; BioProject PRJNA872288) of G. chorda were trimmed using Trimmomatic (v0.39;default options;Bolger et al. 2014) and mapped to coding sequences (NCBI genome accession NBIV00000000.1;Lee et al. 2018) from this species using Salmon (v1.4.0; default options; Patro et al. 2017).Based on the transcripts per million (TPM) value of each gene, those with low TPM values (<0.1) were removed.Organelle genes were analyzed with ChloroSeq using the reads per kilobase million (RPKM) values (filtration cutoff: < 0.1), which is an optimized organelle Lee et al. • https://doi.org/10.1093/molbev/msae012MBE RNA-seq bioinformatic pipeline including the Tophat2 aligner (Kim et al. 2013;Smith and Lima 2017).After the filtration step, a total of 7,985 genes were selected for analysis of gene expression.To validate the gene expression results in triplicate (or duplicate) RNA-seq data for each condition, we conducted Pearson correlation analysis (pearsonr; scipy.statspython module v0.13.0b1; python v2.7.16).The average correlation coefficient values for each condition ranged from 0.93 to 0.99, with P-value < 0.05.These results indicate that the observed gene expression patterns under each experimental condition are consistent and the observed variation is not statistically significant.To identify rhythmic gene expression, we generated average TPM or RPKM values for each condition, and did 3 types of comparative analyses based on the z-scores (["TPM value"-"average TPM values of all conditions in each gene"]/"Standard deviation of all conditions in each gene") between the 24 h-DN period (12 h-light and 12 h-dark; DN4, DN10, DN16, and DN22) and the latter period (24 h-light; CL4, CL10, CL16, and CL22).
The first approach was a selection of shared gene expression patterns with respect to the time between the two periods (i.e."DN4-DN22" and "CL4-CL22").We used the relative gene expression patterns (i.e.z-scores) with "up (+)" and "down (−)" indicating "increase" and "decrease" of gene expression compared to the previous time point, respectively (supplementary fig.S6, Supplementary Material online).For example, z-scores of "PXF48091.1"were 1.71 (DN4), −0.94 (DN10), −0.49(DN16), 0.41 (DN22), 0.92 (CL4), −1.16 (CL10), −0.67 (CL16), and 0.23 (CL22) that indicated a "down-up-up/ up/down-up-up" pattern.The same patterns occurred between the 2 periods (i.e."down/up/up") but we did not consider the transition point between DN22 and CL4 because it is not directly related to the rhythmic pattern at this stage.Based on the up and down gene expression patterns, a total of 3,098 rhythmic gene expressions were selected (the green asterisk in supplementary fig.S6, Supplementary Material online).The second approach was analysis of periodicity in each gene expression pattern.To this end, we used MetaCycle (meta2d; Wu et al. 2016) and BioCycle (Agostinelli et al. 2016) based on the DN period for 24 h (12 h-light:12 h-dark; DN4, DN10, DN16, and DN22).MetaCycle (Wu et al. 2016) analyzes periodic genes using three methods: ARSER (ARS), JTK_CYCLE (JTK), and Lomb-Scargle (LS).To implement MetaCycle, timepoints with an equal number of biological replicates are required, thus we selected 2 RNA-seq libraries from the existing triplicate datasets because 1 condition (CL16) had only 2 libraries available.For the data collection, the most correlated (duplicate) RNA-seq data from each sample were selected based on Pearson correlation analysis (the highest correlation coefficient with P-value < 0.05; pearsonr; scipy.statspython module v0.13.0b1; python v2.7.16).From the 3 types of MetaCycle results (ARS, JTK, and LS), we identified 1,056 periodic genes that show a P-value < 0.05 under all algorithms.BioCycle (Agostinelli et al. 2016) is a deep learning method used to recognize periodic genes.We identified 2,370 periodic genes with P-value < 0.05 and combined these with the MetaCycle results into an initial candidate list of 2,518 rhythmic genes (the blue asterisk in supplementary fig.S6, Supplementary Material online).However, among the selected rhythmic genes, we excluded 971 genes because they show low sensitivities, which indicates different gene expression patterns between the two periods (e.g.up/up/down and up/down/down in PXF48668.1;supplementary fig.S6, Supplementary Material online), although we used several statistical tests supported by MetaCycle and BioCycle.Therefore, we selected 1,547 genes with high sensitivities as rhythmic genes, which show the same gene expression pattern between the 2 periods with statistical supports as described above (P-value < 0.05; MetaCycle and BioCycle).As an additional validation step, we identified 670 genes that have a P-value < 0.05 in the Pearson correlation analysis (pearsonr; scipy.statspython module v0.13.0b1; python v2.7.16) of gene expression patterns between the two periods among the 1,547 rhythmic genes in G. chorda.These genes are marked as "significant correlation" (supplementary table S1, Supplementary Material online).We defined the remaining 1,551 rhythmic gene candidates as "general-rhythm" genes due to the same expression patterns between "DN4-DN22" and "CL4-CL22" (supplementary fig.S6 and table S1, Supplementary Material online).To create graphs of gene expression patterns with the normalized values (z-score), the "matplotlib.pyplot"(v1.3.1),we used a module in Python (v2.7.16).
Rhythmic gene expression patterns were divided into 12 major and two minor patterns that show fluctuating gene expression (Fig. 1).The up-and down-regulated genes from DN4 are shown with "+" and "−", respectively, which indicate relative gene expression relative to the previous time point.Based on these results, we defined morning-phased (a-b-c; DN4), dusk-phased (d-e-f; DN10), evening-phased (g-h-i; DN16), and dawn-phased (j-k-l; DN22) rhythmic genes (Fig. 1).Within each group, however, up-regulation of gene expression starts at different times.For example, the up-regulation of gene expressions in pattern-a starts at DN10, which is earlier than the others (i.e.pattern-b at DN16, and pattern-c at DN22) within this group (Fig. 1).To infer the CRY family data in G. chorda, we relied on published data (Kianianmomeni and Hallmann 2014).

Quantitative Real-time PCR Experiments
To verify the reliability of gene expression patterns in the RNA-seq data, we conducted qRT-PCR experiments.All Diurnal Rhythms in the Red Seaweed Gracilariopsis chorda • https://doi.org/10.1093/molbev/msae012MBE total RNA samples were pretreated with RNase-free DNase I (Ambion, USA) to eliminate genomic DNA contamination before cDNA synthesis.Reverse transcription of RNA samples was done using the RevertAid First Strand cDNA Synthesis Kit (Fermentas, Lithuania) according to the manufacturer's instructions, and the resulting cDNA products were diluted 1:20 with nuclease-free water.Specific PCR primer pairs for the target genes were designed (supplementary table S9, Supplementary Material online).The qRT-PCR was conducted on a CFX connect Real-Time System (Bio-Rad, Germany) using SsoFast EvaGreen Supermix (Bio-Rad, Germany) in a 10 μL reaction.Preliminary quantitative RT-PCR assays and melting curve analyses were performed to ensure that the primer pairs could efficiently amplify a single product without genomic DNA contamination.Based on the preliminary test, we selected two housekeeping genes in G. chorda, eukaryotic translation initiation factor 4E (eIF4E) and eukaryotic peptide chain release factor subunit 1 (ERF1), which showed the least variation in gene expression in our experiments.The qRT-PCR experiments for the target genes were performed with biological triplicates.The relative quantification of each gene expression among samples was evaluated using the comparative ΔΔC T method provided by Bio-Rad program (Bio-Rad CFX Manager 3.1) with two internal controls.To assess the quality of the qRT-PCR experiments, we calculate relative gene expression with error bars (standard error of the mean; supplementary fig.S7, Supplementary Material online).

Homologous Gene Search and Phylogenetic Analysis
Homologs of target genes were identified using Protein Basic Local Alignment Search Tool (BLASTp) search (top 500-1,000 hits; e-value cutoff = 1.e-05), and aligned with MAFFT (default option: -auto;v7.487;Yamadat et al. 2016).Available red algal EST (expression sequencing tags) data were collected from the MMETSP database (Keeling et al. 2014;Johnson et al. 2019).Phylogenetic analysis using the alignments was done using maximum likelihood (ML), with IQ-tree v1.6.12 (Nguyen et al. 2015).The ML trees were constructed using model test (-m TEST), and ultrafast bootstrapping of 1,000 replications (-bb 1000).When we constructed a combined alignment (e.g.CRYs) including gene families, we collected the top 5 or 10 hits in each taxonomic group from each blast result, and these were combined.The combined protein dataset was aligned using MAFFT (v7.487;Yamada et al. 2016) under default settings and used for phylogenetic analysis.

Experimental Procedure for Subcellular Localizations
Full-length cDNA fragments lacking a stop codon that encoded the seven CRY isoforms in G. chorda were amplified by PCR with their specific primer pairs (supplementary table S10, Supplementary Material online).PCR products were digested with XbaI/SpeI and BamHI/BglII, and ligated in-frame upstream of the green fluorescence protein (GFP) gene in the vector 326-sGFP.The Gc-pCRY1 (PXF43553.1)gene was not used in this study because we were unable to construct the in-frame GFP fusion.To analyze subcellular localizations, A. thaliana leaf protoplasts were isolated according to Yoo et al. (2007).The in-frame GFP fusion constructs were introduced into A. thaliana protoplasts with the polyethylene glycol-mediated method (Abel and Theologis 1994), and then incubated for 12-16 h under dark conditions.Images of GFP fluorescence and red chlorophyll autofluorescence in the transformed protoplasts were taken by a cooled charge-coupled device camera and an Olympus BX53 Light/Fluorescence microscope at 40× magnification.We used the fluorescence filter sets of eGFP (Ex: BP470/40, 495DC, Em: BP525/50), and Cy5 (Ex: BP620/60, 660DC, Em: BP700/75) for GFP and chlorophyll autofluorescence, respectively.
Summary of Carbon Metabolisms in Green Algae and C 3 /C 4 /CAM Plants A simplified model of carbon metabolism in C 3 /C 4 /CAM plants was generated based on insights gained from a wide variety of studies (Drincovich et al. 2001;Lai et al. 2002;Wheeler et al. 2005;Aubry et al. 2011;Maier et al. 2011;Tronconi et al. 2008Tronconi et al. , 2018;;Zones et al. 2015;Rao and Dixon 2016;Shen et al. 2017;Chen et al. 2019;Khoshravesh et al. 2020;Tay et al. 2021;Winter and Smith 2022).data analysis and identification of protein subcellular localizations.J.M.L. conducted the bioinformatic analysis for transcriptome data including collection of rhythmic genes, analysis of gene expression patterns, metabolic functions, phylogeny, and gene transfer analysis.J.M.L., J.H.Y., W.Y.K., A.P.M.W., D.B., and H.S.Y. wrote the manuscript draft.All authors read and approved the final manuscript.

Fig. 1 .MBEFig. 2 .
Fig.1.Rhythmic gene expression patterns in the red seaweed Gracilariopsis chorda.The dark areas in the plots are nighttime hours.The major peak patterns are labeled as pattern-a to pattern-l, grouped into morning-phased (a-b-c), dusk-phased (d-e-f), evening-phased (g-h-i), and dawnphased (j-k-l) genes.The two fluctuating rhythmic gene expression patterns are considered minor peak patterns (m and n).The relative gene expression patterns are marked with the "+" and "-" symbols, indicating "increase" and "decrease" in gene expression when compared to the previous time point, respectively (see Materials and Methods).

Fig. 3 .
Fig. 3. Model for the DN cycle of gene expression involving metabolism of cytosolic C 3 and C 4 organic acids in G. chorda (asterisk: significant correlation in rhythmic pattern).