The TkPEL-like coding sequence was amplified by PCR from T. koksaghyz cDNA using primers TkPEL-like_fw and TkPEL-like_rev ( Supplementary Table S1 ). It was purified using the PCR clean-up gel extraction kit (Macherey-Nagel, Germany) and inserted into the cloning vector pJET1.2/blunt using the CloneJET PCR Cloning Kit (Thermo Fisher Scientific, USA). For the TkPEL-like overexpression construct, the strong quadruple cauliflower mosaic virus promoter (pQ35S) was amplified from pFGC5941-GW (GenBank: DQ231581.1) using primers Q35S-P_XmaI_fw and Q35S-P_SOE_rev_XhoI_mut, and the TMV Ω 5′-leader sequence was amplified from the same plasmid using primers TMV_Omega_TL_SOE_fw_XhoI_mut and TMV_Omega_TL_XhoI_rev. The PCR products were diluted 1:100 and used as templates for overlap extension PCR with primers Q35S-P_XmaI_fw and TMV_Omega_TL_XhoI_rev. This disrupted the XhoI site between pQ35S and the TMV Ω 5′-leader, enabling the subsequent cloning step. The PCR product was purified and digested (XmaI + XhoI) and inserted into pLab12.1 (Post et al., 2012), which had been digested with the same enzymes. The TkPEL-like sequence was amplified from pJET-TkPEL-like using primers TkPEL-like_NcoI_fw and TkPEL-like_XhoI_rev. After purification and digestion (XhoI + XbaI), the TkPEL-like sequence was inserted into the XhoI/XbaI-digested custom vector for constitutive expression. For transient expression in Nicotiana benthamiana, the TkPEL-like sequence was amplified from pJET-TkPEL-like using forward primer TkPEL-like_NcoI_fw and either TkPEL-like_XhoI_rev or TkPEL-like_XhoI_rev_nsc, yielding fragments with and without a stop codon. These fragments were purified, digested (NcoI + XhoI) and inserted into the NcoI/XhoI-digested Gateway vector pENTR4 (Invitrogen, USA), resulting in the cloning vectors pENTR4-TkPEL-like and pENTR4-TkPEL-like_NSC (NSC = no stop codon). Site-directed mutagenesis was carried out with primer pairs TkPEL-like_muta.t55a and TkPEL-like_muta.t55a_antis or TkPEL-like_muta.t85a and TkPEL-like_muta.t85a_antis using the QuikChange Lightning Site-Directed Mutagenesis Kit and the publicly available QuikChange Primer Design Program (Agilent Technologies, USA), leading to the nucleotide substitutions c.55T>A and/or c.85T>A. This yielded the following cloning vectors: pENTR4-TkPEL-like_t55a, pENTR4-TkPEL-like_t85a, pENTR4-TkPEL-like_t55a/t85a, pENTR4-TkPEL-like_NSC_t55a, pENTR4-TkPEL-like_NSC_t85a and pENTR4-TkPEL-like_NSC_t55a/t85a. Finally, the mutated and non-mutated sequences were transferred to the Gateway-compatible vectors pBatTL-ccdB, pBatTL-Cerulean-ccdB and pBatTL-ccdB-Cerulean (Epping et al., 2015) by LR recombination using LR Clonase (Thermo Fisher Scientific), resulting in the final plant transformation vectors pBatTL-TkPEL-like_t55a, pBatTL-TkPEL-like_t85a, pBatTL-TkPEL-like_t55a/t85a, pBatTL-Cerulean-TkPEL-like_t55a, pBatTL-Cerulean-TkPEL-like_t85a, pBatTL-Cerulean-TkPEL-like_t55a/t85a, pBatTL-TkPEL-like_t55a-Cerulean, pBatTL-TkPEL-like_t85a-Cerulean and pBatTL-TkPEL-like_t55a/t85a-Cerulean. All constructs were sequenced to confirm their integrity.

All T. koksaghyz plants were cultivated under controlled conditions in an indoor greenhouse (18°C, 16‐h photoperiod, 260 PPFD high-pressure sodium lamp with enhanced yellow and red spectrum) as previously described (Unland et al., 2018). Plant tissues were separated during harvesting, then immediately flash-frozen in liquid nitrogen and lyophilized. The root tissue was pulverized using a ZM 200 Ultra Centrifugal Mill (Retsch, Germany), whereas petiole, leaf and flower tissues were pulverized under liquid nitrogen with mortar and pestle.

The TkPEL-like overexpression construct was introduced into T. koksaghyz plants (accession number 203; kindly provided by the Botanical Garden of the University of Münster, Germany) as previously described (Stolze et al., 2017). The resulting transgenic plants were selected by cultivation on phosphinothricin-containing medium. The presence of the transgenes was verified in crude leaf extracts by PCR with gene-specific primers ( Supplementary Table S1 ) using the KAPA3G Plant PCR Kit (Kapa Biosystems, USA).

Total RNA was extracted from T. koksaghyz root, latex, leaf, petiole and flower tissues using the innuPREP RNA Mini Kit (Analytik Jena, Germany) according to the manufacturer’s instructions. To synthesize full-length cDNA, 500 ng of total RNA was reverse transcribed using PrimeScript RT Master Mix (TAKARA Clontech, USA) according to the manufacturer’s instructions.

Endogenous spatiotemporal expression patterns were determined by quantitative real-time PCR (qRT-PCR) as previously described (Laibach et al., 2015) using RNA from the root, latex, petiole, leaf without midrib and flower tissues of 12-week-old T. koksaghyz wild-type plants grown outdoors from May to August 2017 (University of Münster, Germany) or leaf RNA derived from wild-type plants aged 3, 4, 5, 6, 8, 10 or 12 weeks grown under controlled greenhouse conditions as above. For the spatial expression pattern, three pools consisting of four individual plants as biological replicates were tested for each tissue. For the temporal expression pattern, three independent wild-type plants as biological replicates were tested for each time point. TkPEL-like overexpression in transgenic plants was assessed for eight individual plants from each of the lines #3.4, #4.1 and #5.7 in comparison to nine individual plants of near-isogenic lines (NILs). For the validation of the transcriptomic data, leaf RNA from three pools each consisting of four plants was analyzed for the transgenic lines #3.4, #4.1 and #5.7. In comparison, three pools consisting of leaf RNA from two to three plants each was analyzed for the corresponding NILs (NIL#3.4, NIL #4.1, NIL #5.7). All samples were additionally analyzed in three technical replicates. We used elongation factor 1a (TkEf1a) and ribosomal protein L27 (TkRP) to calculate the relative normalized expression levels (Pütter et al., 2017) using Bio-Rad CFX Manager v3.1 (Bio-Rad Laboratories, USA). All oligonucleotides are listed in Supplementary Table S1 .

Transient expression in 4-week-old N. benthamiana leaves was achieved by agroinfiltration (Müller et al., 2010). N. benthamiana seeds (kindly provided by the Sainsbury Laboratory, John Innes Centre, Norwich, UK) were germinated and the plants were cultivated under controlled greenhouse conditions (Unland et al., 2018). After infiltration, the plants were grown for 4–5 days under steady light in a growth chamber (CLF Plant Climatics, Germany) at 19°C. The subcellular localization of recombinant proteins in epidermal cells was determined by screening tissue explants from infiltrated leaves for Cerulean fluorescence using the Leica TCS SP5 X confocal system (Leica Microsystems, Germany) with excitation at 405/458 nm and emission at 469–488 nm.

Leaf material was harvested, immediately flash-frozen in liquid nitrogen, lyophilized in a freeze dryer Alpha 1-4 LSCplus (Christ Gefriertrocknungsanlagen, Germany) and cryogenically ground in a Retsch mixer mill MM 400. We then extracted 10 mg of the pulverized leaf powder twice with 500 µl acetone (30 min, room temperature, on a shaking platform in the dark) and removed the cell debris by centrifugation (1000 g, 5 min, room temperature). The final 1 ml acetone leaf extract was diluted 1:10 with acetone and analyzed by absorption spectrophotometry at 470, 645, 662 and 750 nm using a quartz cuvette and a BioSpectrometer (Eppendorf, Germany). The carotenoid and chlorophyll contents were calculated as previously described (Lichtenthaler, 1987).

Poly(cis-1,4-isoprene), squalene/2,3-oxidosqualene and pentacyclic triterpenes/triterpenoids were measured in pulverized root extracts by ¹H-NMR and GC-MS (Stolze et al., 2017).

The maximum potential quantum efficiency of photosystem II (PSII) and non-photochemical quenching (NPQ) were measured in a single leaf from 12-week-old transgenic T. koksaghyz plants and controls using a Maxi-Imaging Pulse-Amplitude-Modulation (PAM) chlorophyll fluorimeter (Walz, Germany). Before measurement, the leaves were dark adapted for 20 min, exposed to photosynthetically active radiation with an intensity of 1076 μmol m^-2 s^-1 for 5 min, followed by relaxation in the dark for ~15 min. Fluorescence measurements were captured from 14 independent areas per leaf. The photosynthetic parameters, described in detail by Baker (2008), were calculated using ImagingWin v2.41a (Walz). NPQ was calculated using Equation 1, the maximum quantum yield of PSII after dark adaption (F_v/F_m) using Equation 2, and the operating efficiency of PSII under illumination (F_q′/F_m′ or Φ_PSII) using Equation 3.

The detection of distant TkPEL-like homologs was achieved by screening the non-redundant NCBI peptide database (NR, http://www.ncbi.nlm.nih.gov). Multiple sequence alignment was then carried out using Clustal Omega (https://www.ebi.ac.uk/Tools/msa/clustalo/) and MegAlign Pro 17 (DNASTAR, USA). MOTIF (https://www.genome.jp/tools/motif/) was used to screen the NCBI-CDD library (Marchler-Bauer et al., 2013), as well as the Pfam (Bateman et al., 2002) and PROSITE (Falquet et al., 2002) databases for matching protein domains. Evolutionarily conserved amino acids were visualized using the ConSurf server (http://consurf.tau.ac.il; Ashkenazy et al., 2016), and secondary structures were predicted using the JPred 4 server (Drozdetskiy et al., 2015). The subcellular localization of proteins was predicted using DeepLoc (Armenteros et al., 2017; Armenteros et al., 2019). The TkPEL-like promoter was analyzed using NSITE-PL (Shahmuradov and Solovyev, 2015).

RNA was extracted from the leaf material of four plants from each pQ35S::TkPEL-like line (#3.4, #4.1 and #5.7) and four plants of the corresponding NILs. The RNA obtained from four individual plants belonging to one line was pooled in equal amounts. The three RNA pools per genotype were then pooled again in equal amounts resulting in two RNA pools consisting of 12 plants each that were used for Illumina 150-bp paired-end sequencing (Novogene, USA). Reads were mapped to the T. koksaghyz reference genome (Lin et al., 2022) using Subread v2.0.3 and raw counts were determined using Feature Counts. Transcripts were annotated by conducting a BLASTX search against the NCBI NR and UniProt databases with a minimum e-value of 1 × 10^-3. Differential gene expression was analyzed using DEGseq v1.50.0. Datasets were filtered for minimum counts per million (CPM) of 0.2 and differential expression was assessed by comparing CPMs in NILs and pQ35S::TkPEL-like plants. Differentially expressed genes (DEGs) were accepted if they met the threshold for log₂ fold changes – log₂(FC) – in normalized expression ≥ 1 or ≤ –1 and a q-value (Storey) < 0.05 ( Supplementary Data S1 ). Gene Ontology (GO) term enrichment among modulated genes (–2.32 ≤ log₂(FC) ≥ 2.32) was calculated using WEGO (https://wego.genomics.cn/) with all transcripts detected in NILs as the background. Genes were also annotated against the Kyoto Encyclopedia of Genes and Genomes (KEGG) database using eggNOG-mapper v2 (http://eggnog-mapper.embl.de/). KEGG pathway enrichment analysis was carried out using the enricher tool of the clusterProfiler package in R Studio (Posit PBC, USA).

A homology search against T. koksaghyz sequence databases identified Contig19977 (length 751 bp, e-value 2 × 10^-29) with a translated sequence similar to DCAR_032551. Amplification with flanking primers yielded a shorter, 327-bp coding sequence, which we named TkPEL-like. A subsequent, less stringent BLASTP search with the obtained 108 aa sequence identified homologous protein sequences to be most prominent in clade Streptophyta ( Supplementary Figure S1 ). They included only a few annotated proteins: a DNA replication complex GINS protein, a putative angiotensin-converting enzyme 2 and a polyribonucleotide nucleotidyltransferase. Nucleotide-based homology searching revealed an additional match to an argininosuccinate lyase ( Supplementary Table S2 ). Multiple sequence alignment of the TkPEL-like protein, DCAR_032551 and AtRPGE2 showed that the 57 N-terminal amino acids are strongly conserved ( Figure 1A ), with three overlapping domains: ‘A_thal_3526’ (CDD: 401595) and ‘TIGR01589: A_thal_3526’ (CDD: 130650) are only found in plant proteins but their functions are unknown, whereas the PIN_Fcf1-like domain (CDD: 350212) is present in a large nuclease superfamily involved in RNA processing ( Figure 1D ). The N-terminal sequence also contains two highly conserved cysteines that are predicted to act as structural residues (arrows in Figure 1B ). Secondary structure prediction revealed five potential α-helices (H, marked in green) distributed throughout the peptide sequence ( Figure 1D ). Additionally, the protein contains two putative N-glycan accepter sites (AA90-93, 105-108) and two phosphorylation sites (AA25-28,75-78). Like its carrot and Arabidopsis counterparts, TkPEL-like was predicted to be a cytosolic protein ( Supplementary Figure S1 ). Given that the AtRPGE2 promoter contains sequences recognized by PIFs (Leivar et al., 2009; Zhang et al., 2013; Kim K. et al., 2016) and the bZIP transcription factor HY5, a key regulator of light-mediated transcriptional programming (Lee et al., 2007), we also analyzed the TkPEL-like promoter for the corresponding binding sites. First, we searched the T. koksaghyz genome assembly (Lin et al., 2022) for homology to the 327-bp amplified gene sequence to identify the promoter region, revealing two matching loci with 95–99% identity. We then screened 1 kb upstream of each coding sequence to identify cis-acting regulatory motifs. We found 20 motifs for GWHGBCHF020456 on pseudochromosome four, and 21 for GWHGBCHF024124 on pseudochromosome five, including a G-box motif with the core sequence CACGTG. This is known to mediate light-responsive transcriptional regulation by binding to HY5 and PIFs (Toledo-Ortiz et al., 2003; Lee et al., 2007; Kim J. et al., 2016) ( Figure 1C ; Supplementary Table S3 ). This motif also shows similarity to the light-dependent regulatory element ACE^CHS, which is present in the chalcone synthase (CHS) promoter and bound by the bZIP transcription factors CPRF1 and CPRF4 to regulate flavonoid biosynthesis (Kircher et al., 1998; Sprenger-Haussels and Weisshaar, 2000).

To examine the spatial expression profile of TkPEL-like, we used cDNA obtained from latex, root, leaf, petiole and flower tissues of 12-week-old T. koksaghyz plants for qRT-PCR analysis. This revealed predominant expression in leaves ( Figure 2A ). We then evaluated the temporal expression profile of TkPEL-like in leaves, which indicated low levels of gene expression during the growth stages (weeks 3–6) followed by a ~5-fold increase in weeks 8–12 ( Figure 2B ).

We generated expression constructs in which the fluorophore Cerulean was fused in-frame with the TkPEL-like coding sequence at the N-terminus or C-terminus in order to investigate its intracellular localization in vivo. Transient expression in N. benthamiana leaf epidermal cells was monitored by confocal laser scanning microscopy, showing Cerulean fluorescence signals in the cytosol and nucleus for TkPEL-like fusions as well as Cerulean itself ( Figure 3B ). This suggests TkPEL-like is localized in these compartments, although we cannot exclude the possibility that the spatial distribution was due to the small size of the fusion protein. Leaves transiently expressing native TkPEL-like or the Cerulean fusions showed visible signs of chlorosis after 4 days ( Figure 3A ). We therefore measured the pigment content of the leaves and compared them with non-infiltrated leaves and those infiltrated with a Cerulean construct alone. We found that chlorophyll and carotenoid levels were significantly lower in leaves expressing TkPEL-like constructs compared to controls ( Figures 3C, D ; Supplementary Table S4 ).

To assess the role of the two conserved cysteine residues in the highly conserved N-terminus, we introduced either single mutations (C19S or C29S) or a double mutation (C19S, C29S) in the fusion constructs. Microscopy revealed comparable cytosolic and nuclear fluorescence for the mutants and the wild-type variant ( Supplementary Figure S2 ). This was supported by in silico secondary structure predictions, which showed that the five α-helices were preserved in the mutants, confirming no significant impact on the integrity of the protein’s structure ( Supplementary Figure S3 ). Pigment analysis of leaves expressing one of the three mutant TkPEL-like sequences alone or as Cerulean fusions again showed a reduction in chlorophyll and carotenoid levels ( Figures 3C, D ; Supplementary Table S4 ). However, the amounts were comparable to those detected in leaves expressing Cerulean alone. Accordingly, leaves infiltrated with the TkPEL-like wild-type constructs had significantly lower levels of photosynthetic pigments than leaves expressing the mutant variants or Cerulean, suggesting the conserved cysteine residues are important for the function of TkPEL-like.

The ubiquitous overexpression of TkPEL-like in T. koksaghyz was achieved by placing the gene under the control of the pQ35S promoter ( Figure 4A ). Several transgenic lines were generated and crossed with wild-type T. koksaghyz to obtain the T₁ generation. The resulting progeny showed two alternative genotypes: transgenic plants that carried the TkPEL-like overexpression construct, which were named pQ35S::TkPEL-like, and NILs lacking the transgene. In common with the T₀ generation, the T₁ generation of pQ35S::TkPEL-like plants showed a pale green leaf phenotype and a white midrib ( Figure 4B ). The normally green parts of the petioles and flowers also appeared whitish ( Figures 4B, C ).

For in-depth analysis, we used three independent transgenic lines (#3.4, #4.1 and #5.7) grown from seeds harvested from T₁ plants crossed with wild-type plants, thus representing the T₂ generation. The resulting pQ35S::TkPEL-like plants showed no abnormal morphology ( Supplementary Figure S4 ) or differences in growth rate, size of the aboveground tissues or flowering ( Supplementary Figure S5 ) compared to NIL controls, but still had a pale green/whitish appearance, which confirmed the stable inheritance of this phenotype. Next, we used leaf cDNA for qRT-PCR analysis, which confirmed a massive increase in TkPEL-like expression (137-fold, on average) in the pQ35S::TkPEL-like plants compared to NIL controls ( Figure 4D ). We also observed a significant reduction in the amount of chlorophylls and carotenoids in the transgenic leaves compared to NIL leaves ( Figures 4E, F and Supplementary Table S5 ), explaining the observed phenotype. After observing these effects on leaf isoprenoids, we tested the roots of the pQ35S::TkPEL-like plants because T. koksaghyz roots normally accumulate large quantities of isoprenoid end-products such as pentacyclic triterpenes and poly(cis-1,4-isoprene), the main component of natural rubber (Niephaus et al., 2019; Pütter et al., 2019). TkPEL-like overexpression was confirmed in the roots of all three transgenic lines, but the quantity of pentacyclic triterpenes and poly(cis-1,4-isoprene) remained similar to normal ( Supplementary Figure S6 ; Supplementary Table S6 ). However, the dry root weight of lines #3.4 and #5.7 was significantly lower than that of NIL controls ( Supplementary Figure S6 ).

To compare the photosynthetic performance of wild-type and pale green leaves, we measured the leaf fluorescence of 12-week-old NIL and pQ35S::TkPEL-like plants. We measured the maximum quantum yield (F_v/F_m) and the effective quantum yield (F_q′/F_m′) of PSII in three consecutive periods: dark-adaption, exposure to intense light, and during recovery in darkness ( Figure 5A ). During short-term illumination, both genotypes showed stable and comparably low F_q′/F_m′ values. In contrast, the F_v/F_m values of both genotypes in the absence of light appeared to match those of leaves that were typically non-stressed (Björkman and Demmig, 1987), but this value was consistently and significantly higher for the pale green leaves of pQ35S::TkPEL-like plants. We also calculated the NPQ, which was the same for both genotypes ( Figure 5B ). NPQ appeared quickly under short-term exposure to intense light and then relaxed in the same way in both genotypes.

Given the clear negative impact of TkPEL-like overexpression on photosynthetic isoprenoid compounds in the leaf, but not on root isoprenoids, we investigated the underlying mechanism by comparing the transcriptomes of pQ35S::TkPEL-like and NIL leaves. We extracted RNA from the leaves of four individual 12-week-old plants from pQ35S::TkPEL-like lines #3.4, #4.1 and #5.7 and NILs as a control. The samples of each genotype were pooled so that two samples each representing 12 different plants were compared. The sequencing reads were mapped against the T. koksaghyz reference genome (Lin et al., 2022) with 93% efficiency, and were filtered for a minimum CPM of 0.2. The number of reads mapped to a gene was similar for both datasets ( Supplementary Figure S7 ) and 30,093 transcripts could be detected for NIL plants and 29,830 for the pQ35S::TkPEL-like lines.

Differentially expressed genes (DEGs) were defined as those meeting the threshold log₂FC (normalized expression) ≥ 1 or ≤ –1 with a q-value (Storey) of < 0.05. We identified 2646 (upregulated) and 3580 (downregulated) genes that were modulated at least two-fold in the pQ35S::TkPEL-like plants vs NIL, of which 1215 (upregulated) and 837 (downregulated) genes were modulated at least five-fold ( Figures 6A, B ). For independent validation of the transcriptomic data, expression levels of single genes were analyzed in the three separate lines representing each genotype using qRT-PCR ( Supplementary Figure S8 ). This confirmed the overall direction of transcriptional changes in pQ35S::TkPEL-like lines. Here it became obvious that, for specific genes, the transcriptional change seemed to be dependent on the level of TkPEL-like (over)expression.

Gene Ontology (GO) terms were compared with the background of all genes expressed in NIL leaves, showing that several terms were significantly (p < 0.05) enriched among the DEGs ( Figure 6C ). We focused on DEGs meeting the threshold –2.32 ≤ log₂(FC) ≥ 2.32. Among the genes downregulated at least five-fold, we observed enriched GO terms in the superordinate category ‘Biological Process’ related to ‘signal transduction’, ‘signaling’ and ‘secondary metabolic process’ and responses to different stimuli such as ‘response to stress’, ‘response to abiotic/biotic stimuli’ and ‘response to endogenous/external stimuli’. Furthermore, the terms ‘response to other organism’, ‘multi-organism process’, ‘secondary metabolic process’, and ‘response to biotic stimulus’ were enriched for both the upregulated and downregulated genes, and the terms ‘cell growth’, ‘cell wall organization or biogenesis’ and terms related to ‘cell or biological adhesion’ were overrepresented only among the upregulated genes. Also among the downregulated genes, we observed enriched GO terms in the superordinate category ‘Molecular Function’ relating to ‘DNA-binding transcription factor activity’, ‘transcription regulator activity’ and ‘signaling receptor activity’, whereas the term ‘oxidoreductase activity’ was enriched for both the upregulated and downregulated genes. In the superordinate category ‘Cellular Component’, the upregulated genes were enriched for the terms ‘apoplast’, ‘cell junction’, ‘external encapsulating structure’ and ‘cell periphery’.

We also screened for KEGG pathways significantly enriched (p < 0.05) among the DEGs (–1 ≤ log₂(FC) ≥ 1) compared to all genes expressed in the NIL plants ( Figure 6D ). We identified 42 pathways that were overrepresented among the DEGs, many of which related to processes and diseases only found in animals. This probably reflects the conservation of fundamental enzymes among eukaryotes, and such pathways were ignored for subsequent analysis. Enriched pathways present in plants included plant hormone signal transduction and the biosynthesis of phenylpropanoids, sesquiterpenoids, triterpenoids, flavonoids and isoflavonoids.

Recent progress has provided insight into the complexity of light-dependent regulation and isoprenoid biosynthesis in plants (reviewed by Vranová et al., 2012; Tholl, 2015; Li et al., 2022; Wang et al., 2022). Previous studies have indicated that the carrot protein DCAR_032551 (Iorizzo et al., 2016) and the Arabidopsis protein RPGE2 (Ichikawa et al., 2006; Kim K. et al., 2016; Kim et al., 2023) are involved in light signal transduction and the regulation of isoprenoids and isoprenoid-containing compounds such as carotenoids and chlorophylls. We have now identified TkPEL-like, a putative homolog of DCAR_032551/AtRPGE2 in T. koksaghyz. The overexpression of TkPEL-like in N. benthamiana and T. koksaghyz led to a pale green leaf phenotype and a reduction in chlorophyll and carotenoid levels ( Figure 4 ). This indicates the cross-species functional conservation of this protein in photosynthetic organisms, which is also supported by the expression of rice (Oryza sativa) RPGE in Arabidopsis and the interaction between RPGE and GLK homologs of different species (Kim et al., 2023). However, the faster stem elongation and early flowering observed in Arabidopsis plants overexpressing AtRPGE2 (Ichikawa et al., 2006) was not observed in our pQ35S::TkPEL-like T. koksaghyz plants. Contrary to the normal flowering time, different genes involved in the circadian rhythm, which regulate flowering in other plants, were among the DEGs in pQ35S::TkPEL-like vs. NILs ( Supplementary Table S7 ). To our knowledge, the control of flowering time in T. koksaghyz is not yet understood and we do not know the exact functions of the corresponding genes because different functions have been described for different isoforms (Beinecke et al., 2018; Lin et al., 2019; Schmidt et al., 2020). The vernalization-dependent flowering of T. koksaghyz appears to play an important role, but the various protein isoforms that regulate flower development in T. koksaghyz require further detailed analysis.

DCAR_032551 is more strongly expressed in highly pigmented carrot roots compared to pale roots, but the gene carries frameshift mutations that probably cause functional disruption (Iorizzo et al., 2016). The absence of a functional DCAR_032551 gene product therefore appears to be associated with high carotenoid levels whereas a functional DCAR_032551 protein has the opposite effect. This matches the phenotype of TkPEL-like overexpression in T. koksaghyz. Furthermore, neither the carrot nor dandelion plants discussed above showed any developmental abnormalities, suggesting that the synthesis of photosynthetic isoprenoids is affected rather than photomorphogenesis as a whole. However, in contrast to carrots, the isoprenoid content of T. koksaghyz roots was unaffected by TkPEL-like overexpression ( Supplementary Figure S6 ). The transcriptional downregulation of the MVA pathway (as detected in the leaves) may therefore be restricted to green tissues and root isoprenoid synthesis might be regulated differently. MEP pathway downstream products such as carotenoids have not been identified as typical latex/root metabolites, although some MEP pathway proteins have been detected in T. koksaghyz latex (Niephaus et al., 2019; Xie et al., 2019).

Transcriptomic analysis of pQ35S::TkPEL-like lines revealed the downregulation of genes involved in chlorophyll biosynthesis and the MEP and MVA pathways in leaves, providing an explanation for the observed phenotype. In highly pigmented carrots, the DXS gene (representing the first step in the MEP pathway) was induced, whereas in the pale green pQ35S::TkPEL-like dandelion plants it was strongly suppressed (Iorizzo et al., 2016). This suggests TkPEL-like is a negative regulator of DXS, and that other transcriptional changes may be secondary effects. For example, inhibiting the MEP pathway in Arabidopsis revealed that the MEP and tetrapyrrole pathways are co-regulated to maintain a metabolic balance and to prevent photo-oxidative damage by intermediates (Kim et al., 2013).

IspH expression was also strongly reduced in the pQ35S::TkPEL-like plants, in contrast to the outcome in carrots. In Arabidopsis and tomato (Lycopersicon esculentum), IspH is a key enzyme in MEP-derived plastidial isoprenoid biosynthesis, and the upregulation of IspH increased the incorporation of isoprenoid precursors into a downstream recombinant pathway by 100% compared to the upregulation of DXS (Estévez et al., 2001; Botella-Pavía et al., 2004; Kishimoto and Ohmiya, 2006; Banerjee and Sharkey, 2014). The simultaneous downregulation of IspH and DXS may have reduced the allocation of IPP and DMAPP from the MEP pathway, which could not be compensated by the upregulation of DXR due to the limited amount of substrate.

The downregulation of chlorophyll biosynthesis genes and other photosynthesis-related genes was similar to the transcriptional changes observed following AtRPGE2 overexpression in Arabidopsis, although not all the genes downregulated in Arabidopsis were among the DEGs in T. koksaghyz (Kim K. et al., 2016; Kim et al., 2023) ( Supplementary Table S7 ). In Arabidopsis, RPGE2 prevents GLK target gene activation by the formation of heterodimers, which probably contributed to the transcriptional changes observed following AtRPGE2 overexpression (Kim et al., 2023). The upregulation of GLK as described in Arabidopsis could not be detected in our RNA-seq data. To test for a conserved molecular mode of action for TkPEL-like, as suggested in rice (Zhang et al., 2021), GLK homologs in T. koksaghyz and their potential interaction with TkPEL-like should be investigated in the future.

The higher maximum quantum yield in pQ35S::TkPEL-like plants compared to NIL controls ( Figure 5 ) was likewise reported for plants overexpressing AtRPGE2 (Kim et al., 2023). However, the associated lower total seed yields reported for Arabidopsis were not evaluated in our experiments. The increase in F_v/F_m indicates an increase of the proportion of the absorbed light energy that is used for photochemical reactions, which could be a consequence of the lower chlorophyll content in order to maintain sufficient photosynthesis. In chlorophyll-deficient rice mutants, an increased photosynthetic rate per chlorophyll molecule compensated for the negative impact of chlorophyll reduction (Li et al., 2013). It is possible that such an increase also contributed to sufficient photosynthesis and normal biomass accumulation in pQ35S::TkPEL-like plants ( Supplementary Figure S5 ). The lower chlorophyll content could have also reduced the photochemical damage and heat stress in leaves absorbing more light than required for maximum photosynthesis, and energy otherwise used for repair could therefore be used elsewhere, contributing to the comparable biomass accumulation in pQ35S::TkPEL-like and NILs (Hamblin et al., 2014). The comparable NPQ in the pQ35S::TkPEL-like plants and NIL controls supports the hypothesis that TkPEL-like overexpression does not trigger stress. This is particularly notable given that many essential photosynthetic components, such as plastoquinone (an intramembrane electron acceptor downstream of PSII) are also derived from the MEP pathway that was transcriptionally downregulated in pQ35S::TkPEL-like plants (Nowicka and Kruk, 2010).

Despite the lower carotenoid levels in pQ35S::TkPEL-like plants, carotenoid pathway genes tended to be upregulated. This may also be a primary effect, but we hypothesized it might have been a secondary feedback mechanism in response to low carotenoid levels (Avendaño-Vázquez et al., 2014). However, the limited pool of isoprenoid precursors may not have allowed for compensation via transcriptional upregulation. The possible upregulation of CCD, encoding a carotenoid cleavage dioxygenase, suggests an exacerbation of low carotenoid synthesis, given that this enzyme is responsible for the degradation of carotenoids in chrysanthemum, resulting in white petals (Ohmiya et al., 2006).

Three DEGs were found in the pathway leading from IPP and DMAPP to the formation of the first carotenoids. IPP and DMAPP are condensed by prenyltransferases to form pools of C₁₀ (geranyl diphosphate), C₁₅ (FPP) and C₂₀ (GGPP) precursors (Hemmerlin et al., 2012). For carotenoid biosynthesis, two GGPP molecules are condensed by phytoene synthase (PSY) to form phytoene, which is desaturated and isomerized in several steps (Shumskaya and Wurtzel, 2013). Interestingly, the genes encoding FFP synthase (FDPS) and GGPP synthase (GGPS) were upregulated in leaves of pQ35S::TkPEL-like plants, possibly to increase the flux from IPP/DMAPP in this direction. Another GGPS isoform was downregulated in the transgenic plants. Given the widespread occurrence of GGPS gene families together with the specialization of the individual enzymes (Beck et al., 2013; Ruiz-Sola et al., 2016), the opposing transcriptomic responses of the dandelion GGPS genes may reflect their spatial expression profiles or protein localization. In tobacco (Nicotiana tabacum), NtGGPPS1 and a light-regulated small subunit of GGPS were shown to interact with PSY, allowing the channeling of GGPP between successive pathway enzymes. The differential expression of GGPPS may therefore affect the efficiency of PSY in pQ35S::TkPEL-like plants, although the PSY gene itself was not differentially expressed. The specific functions of the differentially regulated isoforms should be examined in detail to validate the hypotheses derived from our expression data.

The induced genes in the downstream parts of the carotenoid pathway included ZEP, encoding zeaxanthin epoxidase (responsible for violaxanthin biosynthesis) whereas the gene encoding 9-cis-epoxycarotenoid dioxygenase (which converts violaxanthin to xanthoxin) was suppressed, indicating that violaxanthin levels were specifically increased in the pQ35S::TkPEL-like lines. In the xanthophyll cycle, violaxanthin is reversibly de-epoxidized to zeaxanthin by violaxanthin de-epoxidase under intense light to efficiently reduce photo-oxidative stress and lipid peroxidation (Havaux et al., 2007). The induction of ZEP may therefore prepare the cell to deal with reactive oxygen species by providing enough substrate for the detoxifying reaction that converts violaxanthin to zeaxanthin, and the limited quantity of carotenoids available was optimally deployed.

Finally, genes involved in flavonoid/isoflavonoid metabolism were also significantly upregulated in pQ35S::TkPEL-like plants compared to NIL controls ( Figure 6B ). One example is the key enzyme CHS, which also plays a role in cell wall organization and biosynthesis (Zuk et al., 2016), a process significantly enriched among upregulated genes ( Figure 6C ). CHS is also regulated by light, via HY5 and COP1 (Ang et al., 1998; Thain et al., 2002). These data suggest that flavonoid pigments, such as anthocyanins, were enriched to compensate for the low carotenoid levels in the plastids thus ensuring photoprotection, given there was no difference in fitness between the pQ35S::TkPEL-like plants and NILs. Such compensatory functions have been proposed in response to UV radiation (Guidi et al., 2016).

TkPEL-like overexpression was negatively correlated with the expression of various modulators of the light response, in contrast to the highly pigmented carrots in which there was a positive correlation (Iorizzo et al., 2016). Transcription factors (HY5 and PIFs) and the E3 ubiquitin ligase COP1 are key regulators of the light response. PIFs negatively regulate photomorphogenesis and repress the MEP pathway (Chenge-Espinosa et al., 2018), as well as chlorophyll and carotenoid biosynthesis (Huq et al., 2004; Moon et al., 2008; Toledo-Ortiz et al., 2010; Tang et al., 2012; Job and Datta, 2021). PIF1 and PIF3 repress transcription, and their degradation in the presence of light thus enables the expression of MEP pathway genes such as IspH (Chenge-Espinosa et al., 2018). In pif1 and pif3 mutant seedlings, protochlorophyllide accumulates in the dark, leading to bleaching under illumination, in parallel to higher levels of light-induced chlorophyll and increased carotenoid accumulation in the dark (Huq et al., 2004; Toledo-Ortiz et al., 2010; Job and Datta, 2021).

PIFs are also potential COP1 cofactors that represses photomorphogenesis in the dark by accumulating in the nucleus, facilitating the degradation of positive regulators of light signaling such as HY5 (Deng et al., 1991; Osterlund et al., 1999; Saijo et al., 2003; Jang et al., 2005; Xu et al., 2014). Accordingly, cop1 mutants show photomorphogenesis in the dark (Deng et al., 1991). These reports contrast with the indicated downregulation of PIF and COP1 in pale green pQ35S::TkPEL-like plants. The identification of a G-box element in the TkPEL-like promoter hints that TkPEL-like may be regulated by HY5, PIFs, or both. HY5 expression was unaffected in pQ35S::TkPEL-like plants, suggesting that HY5 may act upstream of TkPEL-like in the regulatory hierarchy. In contrast, PIFs were differentially expressed in the pQ35S::TkPEL-like plants, but AtRPGE2 was identified as a direct target of PIFs (Kim K. et al., 2016). The exact relationship between PIFs and TkPEL-like should be analyzed in future studies to gain a more detailed understanding of this regulatory network in T. koksaghyz.

One explanation for these contradictory results is that COP1 and PIF1 were downregulated as a feedback response to low carotenoid and chlorophyll levels in order to de-repress photomorphogenesis and associated pigment synthesis, but the limited pool of isoprenoid precursors prevented the normalization of carotenoid and chlorophyll levels. Furthermore, although three transcripts annotated as PIF1 or PIF3 with sequence identities > 50% were indeed downregulated in pQ35S::TkPEL-like plants, one further transcript with high similarity to the ATP-dependent DNA helicase PIF1-like from Cynara cardunculus var. scolymus was strongly upregulated. The overexpression of a single PIF1 isoform together with TkPEL-like may have been sufficient to effectively repress photomorphogenesis, including the downregulation of MEP/MVA and chlorophyll biosynthesis pathway gene expression. However, we found no evidence for the transcriptional modulation of PIF target genes such as PSY, as described in other plants (Toledo-Ortiz et al., 2010). PSY, among other photomorphogenic targets, was shown to be antagonistically regulated by PIFs and HY5 (Toledo-Ortiz et al., 2014). Therefore, unaffected HY5 expression may have been sufficient to regulate HY5 targets in a normal manner, despite differential PIF expression. This would further support the idea that TkPEL-like acts downstream of HY5 in the signaling hierarchy.

We observed the downregulation of COP1, but this gene was co-expressed with nonfunctional DCAR_032551 in highly pigmented carrots (Iorizzo et al., 2016). COP1 is excluded from the nucleus under prolonged illumination (Osterlund and Deng, 1998; Subramanian et al., 2004), so the transcriptional modulation of COP1 in response to light may not be critical for photomorphogenesis as long as nuclear exclusion is not impaired. Notably, translational and post-translational regulation can also influence metabolic pathways in a way that is not fully evident from our data (Hemmerlin et al., 2012; Shumskaya and Wurtzel, 2013; Banerjee and Sharkey, 2014).

TkPEL-like is related to a PNPase and the presence of a PIN_Fcf1-like domain prompted us to analyze genes involved in mRNA processing. This indeed revealed several DEGs that are homologs of genes involved in RNA degradation and surveillance. PNPases play various roles in RNA metabolism. In Arabidopsis, PNPase was shown to facilitate plastid mRNA, tRNA and 23S rRNA metabolism, including poly(A) tail formation and degradation, as well as RNA degradation in general (Walter et al., 2002; Holec et al., 2006; Germain et al., 2011). Knockout mutants revealed a phenotype similar to pQ35S::TkPEL-like plants, with pale young leaves but near normal mature tissues (Sauret-Güeto et al., 2006). The PIN_Fcf1-like domain (~120 residues) may facilitate pre-rRNA cleavage, nonsense mediated mRNA decay and RNA interference (Clissold and Ponting, 2000; Fatica et al., 2004). However, the level of similarity between TkPEL-like, and PNPase and the PIN_Fcf1-like domain was only low, and PNPases are much larger (e.g., 922 amino acids for AtPNPase) than the 108-amino-acid TkPEL-like protein (Walter et al., 2002). Furthermore, plant PNPases have only been found in chloroplasts and mitochondria (Li et al., 1998; Yehudai-Resheff et al., 2001; Holec et al., 2006; Germain et al., 2012), whereas our data suggest that TkPEL-like is localized in the nucleus and cytosol. The secondary structure of the PIN_Fcf1-like domain also differs from that predicted for TkPEL-like ( Figure 1 ) (Senissar et al., 2017). However, a number of conserved acidic residues are found in both sequences. These data suggests that the function of TkPEL-like differs from that described in other species.

Our results support the hypothesis that the protein family containing AtRPGE2 and TkPEL-like acts as a negative regulator of photomorphogenesis (specifically chlorophyll and carotenoid accumulation) and represents an early step in the light-dependent signal transduction system (Iorizzo et al., 2016). The involvement of TkPEL-like proteins in light signal transduction and photomorphogenesis would also explain why this protein family is almost exclusive to photosynthetic organisms thus far, with highly conserved functions (Kim et al., 2023). However, RNAi-induced knockdown in Arabidopsis (Ichikawa et al., 2006) and our attempt to generate T. koksaghyz knockdown lines consistently resulted in a lethal phenotype, whereas an Atrpge1/2/3 triple mutant was viable (Kim et al., 2023). The phenotype of the triple mutant contrasted with the results of overexpressing AtRPGE1, AtRPGE2 or TkPEL-like, which suggested partial redundancy among the different isoforms in Arabidopsis (Kim et al., 2023). Additionally, highly pigmented carrots expressing a presumably nonfunctional version of DCAR_032551 were also viable (Iorizzo et al., 2016). This might reflect a functional specialization of DCAR_032551 so that only root isoprenoids derived from the MEP pathway are affected. The authors did not report low chlorophyll levels in the aboveground organs, and TkPEL-like overexpression did not influence MVA pathway isoprenoids in T. koksaghyz roots.

Predominant expression of TkPEL-like in leaves from the ninth week onwards ( Figure 5B ) coincided with the development of a visible leaf rosette under our greenhouse conditions ( Supplementary Figure S5 ). This expression profile correlated with that of AtRPGE2 (Klepikova et al., 2016). Given the hypothesis that the corresponding protein is a repressor of photosynthetic gene expression, de-etiolation and photomorphogenesis, low expression levels in young plants may promote leaf development and coloration to cope with the initial exposure to light, and higher expression levels in mature leaves are necessary to restrict the aforementioned processes to an appropriate level.

Our data also support recent findings on the molecular function of AtRPGE2 (Kim et al., 2023). The interaction between AtRPGE2 and GLK in the cytosol and nucleus of N. benthamiana cells agrees with our results showing the localization of Cerulean–TkPEL-like fusion proteins in the cytosol and the nucleus in the same species. However, TkPEL-like lacks a nuclear localization signal, and may be transported into the nucleus passively due to its small size or actively transported in response to light, as reported for photoreceptors (Genoud et al., 2008; Galvão and Fankhauser, 2015).

TkPEL-like mutants with either one or two conserved cysteines replaced did not suppress chlorophyll or carotenoid biosynthesis to the same degree as the wild-type protein, suggesting the cysteine residues are functionally important ( Figure 3 ). The cysteine residues found in the N-terminal segment of TkPEL-like are highly conserved between Arabidopsis, carrot and T. koksaghyz ( Figure 1A, B ), and may therefore be required for their molecular function. The conserved cysteine residues could also serve as points of attack for redox signaling (Couturier et al., 2013), which is known to regulate transcription, posttranslational modifications and retrograde signaling (Surpin et al, 2002; Balmer et al., 2003; Lemaire et al., 2004).

TkPEL-like is a promising candidate for the regulation of isoprenoid biosynthesis in leaves because it affects leaf isoprenoid but not root isoprenoid levels when constitutively overexpressed. It may act within the light-dependent signal transduction pathway and react to the redox status of the cell, thereby enabling responses to environmental cues. Our transcriptomic data provide a broad overview of the pathways affected by TkPEL-like overexpression and can be used as a basis for future functional studies to validate our hypotheses and to fully understand the complex regulatory network controlling isoprenoid biosynthesis and photomorphogenesis. A better understanding of the metabolic network will facilitate future breeding approaches aiming to modulate plant isoprenoid levels such as chlorophylls and carotenoids, or possibly other metabolites that we have not studied yet. It could also lead to the development of a suitable plant-based production platform for valuable metabolites.