GOLDEN2‐like1 is sufficient but not necessary for chloroplast biogenesis in mesophyll cells of C4 grasses

SUMMARY Chloroplasts are the site of photosynthesis. In land plants, chloroplast biogenesis is regulated by a family of transcription factors named GOLDEN2‐like (GLK). In C4 grasses, it has been hypothesized that genome duplication events led to the sub‐functionalization of GLK paralogs (GLK1 and GLK2) to control chloroplast biogenesis in two distinct cell types: mesophyll and bundle sheath cells. Although previous characterization of golden2 (g2) mutants in maize has demonstrated a role for GLK2 paralogs in regulating chloroplast biogenesis in bundle sheath cells, the function of GLK1 has remained elusive. Here we show that, contrary to expectations, GLK1 is not required for chloroplast biogenesis in mesophyll cells of maize. Comparisons between maize and Setaria viridis, which represent two independent C4 origins within the Poales, further show that the role of GLK paralogs in controlling chloroplast biogenesis in mesophyll and bundle sheath cells differs between species. Despite these differences, complementation analysis revealed that GLK1 and GLK2 genes from maize are both sufficient to restore functional chloroplast development in mesophyll and bundle sheath cells of S. viridis mutants. Collectively our results suggest an evolutionary trajectory in C4 grasses whereby both orthologs retained the ability to induce chloroplast biogenesis but GLK2 adopted a more prominent developmental role, particularly in relation to chloroplast activation in bundle sheath cells.


INTRODUCTION
Chloroplasts are endosymbiotic organelles and are the site of photosynthetic reactions in a range of organisms from algae to land plants (Sagan, 1967).Given the importance of chloroplast function for autotrophic growth, development of the organelle is a tightly controlled process that involves many levels of regulation (Waters & Langdale, 2009).First described in the mid-1990s, the golden2 (g2) mutant in Zea mays (maize) exhibits pale green leaves and impaired chloroplast development specifically in bundle sheath cells (Hall et al., 1998;Langdale & Kidner, 1994).Subsequent characterization revealed that ZmG2 (sometimes referred to as ZmGLK2) encodes a transcription factor (Rossini et al., 2001), a founding member of the GARP transcription factor family (Riechmann et al., 2000).In most plant species examined, two G2-like (GLK) genes are present in the genome (Wang et al., 2013).GLK genes are defined by a characteristic helix-turn-helix DNA-binding domain related to MYB or MYB-like transcriptional factors, a conserved AREAEAA motif located immediately after the last helix, and a GCT-box in the C-terminal region that is responsible for either homo-or heterodimerization (Fitter et al., 2002;Rossini et al., 2001;Safi et al., 2017).Functional characterization of GLK genes in maize and Arabidopsis thaliana (Arabidopsis) has shown that they are involved in the transcriptional activation of multiple photosynthesisrelated genes including those encoding proteins of the light-harvesting complexes and enzymes of the chlorophyll biosynthesis pathway (Cribb et al., 2001;Waters et al., 2009).This pivotal role in chloroplast biogenesis has been associated with a myriad of physiological processes including stomatal movement, hormone signaling, defense responses, and fruit ripening (Kobayashi et al., 2012;Li et al., 2020;Nguyen et al., 2014).The emerging picture is one where GLK proteins act as linchpins in any metabolic or developmental pathway that requires chloroplast function to be modulated.
During the evolution of land plants, the GLK gene family underwent multiple independent duplication events (Wang et al., 2013).In cases such as Arabidopsis where duplications likely occurred within species, the resulting paralogs are functionally redundant (Waters et al., 2009).However, duplication events in the monocots are more intriguing.The common ancestor of the Poales underwent a whole genome duplication that subdivided the GLK gene family into two sister clades: GLK1 and GLK2 (Wang et al., 2013).Notably, this event occurred before the many independent speciation events that led to the evolution of C 4 photosynthesis within the Poales group (Christin & Osborne, 2013).In monocot species such as rice that perform C 3 photosynthesis, the GLK paralogs are functionally redundant as in Arabidopsis (Wang et al., 2013;Waters et al., 2009).Unlike C 3 photosynthesis, where carbon reactions occur predominantly in mesophyll cell chloroplasts, in C 4 plants carbon assimilation and fixation reactions are compartmentalized in the chloroplasts of mesophyll and bundle sheath cells, respectively (Langdale, 2011).This biochemical compartmentalization is often accompanied by anatomical differences in chloroplast ultrastructure (Sage et al., 2011;Woo et al., 1970).In the C 4 monocot species maize, Setaria viridis (setaria) and Sorghum bicolor (sorghum), GLK1 is expressed preferentially in mesophyll cells whereas GLK2 transcripts accumulate preferentially in bundle sheath cells, with the level of specificity differing between species (Chang et al., 2012;John et al., 2014;Li et al., 2010;Tausta et al., 2014;Wang et al., 2013).Notwithstanding that protein levels dictate functional capacity, defects in bundle sheath but not mesophyll chloroplast development in g2 mutants of maize (Cribb et al., 2001;Langdale & Kidner, 1994;Rossini et al., 2001) demonstrated that the preferential localization of ZmG2 transcripts in bundle sheath cells is functionally relevant and that GLK paralogs in C 4 grasses can control chloroplast development in a cell-specific manner.
GLK paralogs have been extensively characterized in Arabidopsis, however, little is known about gene function in monocots, and in particular the function of GLK1 genes has not been investigated.Because GLK1 transcripts accumulate at much higher levels in mesophyll cells than bundle sheath cells of both maize (10-fold) and setaria (30fold) (Chang et al., 2012, John et al., 2014;Figure S1), and Zmg2 mutants of maize show bundle sheath specific chloroplast defects, it has been hypothesized that GLK1 controls chloroplast biogenesis specifically in mesophyll cells of C 4 grasses (Rossini et al., 2001).Here we test this hypothesis by generating and characterizing loss-offunction mutants in maize and setaria, two species that represent independent origins of C 4 .Our findings reveal that contrary to expectations, neither ZmGLK1 nor SvGLK1 is necessary for chloroplast development in mesophyll cells whereas loss of SvGLK2 function perturbs chloroplast development in both bundle sheath and mesophyll cells.

ZmGLK1 is not necessary for chloroplast biogenesis in mesophyll cells of maize
To determine whether ZmGLK1 regulates chloroplast biogenesis in maize, loss-of-function mutants were generated using CRISPR/Cas9 (Figure 1a).Four independent mutant lines (referred to collectively as Zmglk1) were obtained, all of which exhibited the same overall growth phenotype as segregating wild-type siblings (Figure 1b,c; Figure S2).Ultrastructure analysis revealed that chloroplasts in Zmglk1 lines were similar to wild-type, with well-developed granal thylakoids in mesophyll cells and agranal thylakoids in bundle sheath cells (Figure 1d-i; Figure S2).Total chlorophyll levels were also similar in wild-type and Zmglk1 mutants (Figure 1j).As such, we concluded that ZmGLK1 is not necessary for thylakoid assembly or chlorophyll biosynthesis in bundle sheath or mesophyll chloroplasts of maize.
An evaluation of chloroplast function in Zmglk1 mutants was made by measuring photosystem II (PSII) parameters.PSII is a large protein complex comprised of the antenna and chlorophyll molecules that harvest light and together initiate the electron transfer reactions of photosynthesis (Nickelsen & Rengstl, 2013).In maize, PSII accumulates in the granal thylakoids of mesophyll cells and is largely absent from the agranal bundle sheath cell chloroplasts (Woo et al., 1970).As such, PSII function is a proxy for mesophyll cell chloroplast function.Fluorescence measurements were therefore taken in light and dark-adapted plants to quantify PSII operating efficiency (PhiPSII) and maximum quantum efficiency of PSII photochemistry (Fv/ Fm), in both wild-type and Zmglk1 mutant plants.Figures 1k,l show that PSII is operating normally in Zmglk1 mutants, suggesting that ZmGLK1 is not necessary for chloroplast biogenesis or function in maize mesophyll cells.

SvGLK1 is not necessary for chloroplast biogenesis in mesophyll cells of setaria
Maize and setaria have different origins of C 4 photosynthesis within the PACMAD clade (Li & Brutnell, 2011) but like the orthologous ZmGLK1, SvGLK1 transcripts accumulate preferentially in mesophyll cells (John et al., 2014).Phylogenetic analysis and amino acid sequence comparisons revealed that SvGLK1 shares many structural features with ZmGLK1 (Figure S3).To assess whether SvGLK1 plays a role in chloroplast biogenesis, two independent loss-offunction mutants were generated in setaria using CRISPR/ Cas9 (Figure 2a).As in maize, Svglk1 mutants exhibited overall plant growth (Figure 2b-d) and chloroplast ultrastructure (Figure 2e-j) phenotypes that were similar to segregating wild-type siblings, with thylakoid membranes developing normally in both mesophyll and bundle sheath cells (Figure 2e-l).Furthermore, there was no significant difference in chlorophyll levels between Svglk1 mutants and wild-type (Figure 2m), and electron flow through PSII was unperturbed (Figure 2n,o).Collectively, these results suggest that the preferential expression of GLK1 genes in mesophyll cells of C 4 grass species does not reflect a requirement for gene function during chloroplast development and/or maintenance.(h-l) Measurements of thylakoid occupancy in mesophyll cells (h), thylakoid occupancy in bundle sheath cells (i), total chlorophyll (j), PSII operating efficiency (k), and maximum quantum efficiency of PSII photochemistry (l).The WT segregant is depicted in blue and the four independent Zmglk1 lines are depicted in green.Each dot represents a biological replicate, the square is the average value for that line and the bars are the associated standard deviations.A minimum of five biological replicates were evaluated for thylakoid occupancy and chlorophyll quantification, and three biological replicates were assessed for PSII fluorescence.Different letters indicate statistically different groups (P value ≤0.05, one-way ANOVA and Tukey's HSD; see SvGLK2 regulates chloroplast development in both mesophyll and bundle sheath cells of setaria. Given that dimorphic chloroplast development is observed in bundle sheath and mesophyll cells of both maize and setaria, we next sought to determine whether SvGLK2, like its ortholog ZmG2 (Figure S3), regulates bundle sheath chloroplast development.To this end, CRISPR/Cas9 was used to generate loss-of-function Svglk2 mutants  (Figure 3a).In contrast to maize where g2 mutants are weaker than wild-type but are viable and fertile, any Svglk2 mutant plants that regenerated after transformation did not progress beyond the seedling stage before dying.The only viable line recovered from tissue culture was heterozygous for SvGLK2 and this was used to isolate a null homozygous line via self-pollination.Svglk2 8G5 mutants, henceforth referred to as Svglk2, exhibited severe growth defects, had pale yellow leaves, and failed to complete the life cycle (Figure 3b).Leaf cross sections showed misshapen and smaller chloroplasts in the bundle sheath cells of Svglk2 mutants compared to wild-type (Figure 3c,d), and observations of chloroplast ultrastructure revealed impaired chloroplast anatomy with discontinuous thylakoids (Figure 3e,f), but thylakoid occupancy within the organelle was not significantly different from wild-type (Figure 3i).More striking, given the bundle sheath-specific perturbations observed in maize g2 mutants, was the aberrant phenotype of mesophyll chloroplasts in Svglk2 mutants.Compared to the wild-type there was a significant reduction in thylakoid membrane coverage (15% less than the wild-type) and a noticeable reduction in the size and frequency of grana (Figure 3g,h, j).As in bundle sheath cells, thylakoid membranes in mutant mesophyll chloroplasts appeared fragmented (Figure 3h).The observed perturbations in bundle sheath and mesophyll cells were accompanied by a reduction in leaf chlorophyll content of more than 80% (Figure 3k).Together, the Svglk2 and Svglk1 mutant phenotypes suggest that SvGLK1 cannot compensate for loss of SvGLK2 function whereas SvGLK2 can compensate for loss of SvGLK1 function.As such, despite SvGLK2 transcripts comprising only 43% of the total GLK transcript pool in mesophyll cells (as compared to 98% in bundle sheath cells) (John et al., 2014), SvGLK2 is the primary regulator of chloroplast biogenesis in both cell-types.
GLK2 transcript levels are not upregulated in glk1 mutants of maize or setaria Given that ZmGLK1 and SvGLK1 transcripts represent 95% and 57% of the total GLK transcript pool in mesophyll cells of the respective species (Chang et al., 2012, John et al., 2014; Figure S1), the absence of chloroplast defects in this cell-type in loss of function mutants is intriguing.It is particularly notable because all GLK genes characterized to date have a role in chloroplast biogenesis, albeit a redundant role alongside paralogs in many cases (Fitter et al., 2002;Wang et al., 2013).Even a redundant role is unlikely in setaria given the Svglk2 mutant phenotype.However, redundancy could explain the lack of a Zmglk1 mutant phenotype if ZmG2 transcripts are translated very efficiently in mesophyll cells of maize.Alternatively, loss of ZmGLK1 function could lead to altered regulation of ZmG2 expression and/or to the upregulation of genes encoding the GATA transcription factors GNC and CGA1 which also regulate chloroplast development (Chiang et al., 2012;Hudson et al., 2013) and have been shown to partially rescue loss of GLK function in Arabidopsis (Zubo et al., 2018).To test this hypothesis, transcript levels of ZmGLK1, ZmG2, ZmGNC, and ZmCGA1 were quantified in maize using the Zmglk1 gene-edited mutants generated above alongside the previously reported transposon-induced Zmg2-bsd1-s1 mutant (Cribb et al., 2001).In each case the mutant allele produces transcripts that encode non-functional protein(s).
Figure 4 reveals that loss of Zmglk1 function does not lead to consistently elevated levels of ZmGLK1 (Figure 4a), ZmG2 (Figure 4b), ZmGNC (Figure 4c) or ZmCGA1 (Figure 4d) transcripts.A small but significant increase in ZmG2, ZmGNC and ZmCGA1 transcript levels was seen in one of the four Zmglk1 mutant lines but that cannot explain how chloroplasts develop normally in all four lines.We thus conclude that transcription of ZmG2, ZmGNC and ZmCGA1 is not upregulated in mesophyll cells when Zmglk1 function is lost.Whereas loss of ZmGLK1 function in maize does not impact on ZmGLK1 or ZmG2 transcript levels, loss of ZmG2 function in bundle sheath cells leads to a statistically significant reduction in the accumulation of both ZmGLK1 and ZmG2 transcripts (Figure 4a,b), and a similar situation is seen in setaria where both SvGLK1 and SvGLK2 transcript levels are dramatically reduced in Svglk2 mutants (Figure 4e,f).This observation suggests that the ZmG2/ SvGLK2 orthologs, either directly or indirectly, promote the accumulation of both GLK1 and GLK2 transcripts in the respective species.Consistent with this suggestion, multiple G-box elements which are bound by Arabidopsis GLK proteins (Waters et al., 2009) are present in ZmGLK1, ZmG2 and SvGLK2 promoter sequences (Figure S4), and a GLK binding site (Tu et al., 2022) is present 1840 bp upstream of the ATG in the SvGLK1 gene.Reduced levels of SvGLK1 transcripts in one of the two Svglk1 mutant lines (Figure 4e) suggest that SvGLK1 might regulate its own transcript levels via this binding site but the absence of such an effect in the other mutant line makes it equally likely that the specific edits in line #2062 lead to transcript turnover, possibly via the nonsense-mediated decay pathway.Collectively these data suggest that chloroplast development in C 4 grasses is predominantly regulated by genes of the GLK2 clade.
ZmGLK1 and ZmG2 are both sufficient to restore dimorphic chloroplast development in mesophyll and bundle sheath cells of Svglk2 mutants Our findings thus far show that SvGLK2 and ZmG2 play a more central role in controlling chloroplast biogenesis in C 4 grasses than their respective GLK1 paralogs.To evaluate whether this difference in necessity also applies to sufficiency, ZmGLK1 and ZmG2 were expressed under the   expression is a direct effect of the maize transgene and not via SvGLK2).In this regard, we must deduce that the transgenic lines expressing ZmGLK1 accumulate transcripts encoding functional ZmGLK1 and SvGLK1 proteins in both mesophyll and bundle sheath cells, and those expressing ZmG2 accumulate transcripts encoding functional ZmG2 and SvGLK1 proteins, similarly in both cell types.
To determine whether ZmGLK1 and/or ZmG2 can rescue chloroplast ultrastructure defects in mesophyll and/or bundle sheath cells of Svglk2 mutants, confocal images and transmission electron micrographs of transverse leaf sections from all six transgenic lines were examined.In all cases, chloroplasts were restored to normal size (Figure 6a-c; Figure S5), with predominantly agranal   thylakoids in bundle sheath cell chloroplasts (Figure 6d-f; Figure S5) and granal thylakoids in mesophyll chloroplasts (Figure 6g-i; Figure S5).Quantification revealed that with the exception of the one line that had low levels of transgene expression (#17808-15 254), thylakoid occupancy levels in the mesophyll (Figure 6j) and bundle sheath (Figure 6k) chloroplasts of transgenic lines were statistically equivalent (or in one case higher) than in wild-type plants.Consistent with the ectopic expression of GLK genes, mature chloroplasts were also observed in the leaf vasculature of all complemented lines, more with ZmG2 than ZmGLK1 (Figure 6a-c; Figure S5).Irrespective of which maize ortholog was overexpressed, the total chlorophyll content in leaves of transgenic plants was restored to wild-type levels (Figure 6l).Notwithstanding the fact that SvGLK1 may be functional in bundle sheath and mesophyll cells of all transgenic plants, collectively these data demonstrate that ZmGLK1 and ZmG2 proteins are functionally equivalent in the context of the C 4 setaria leaf and that the dimorphic identity of bundle sheath and mesophyll cell chloroplasts develops regardless of which maize ortholog is overexpressed.

DISCUSSION
Prior to this study, it was known that GLK genes are duplicated in many land plant species and that in species that carry out C 3 photosynthesis the duplicate genes act redundantly to regulate chloroplast biogenesis in all photosynthetic cells (Bravo-Garcia et al., 2009;Wang et al., 2013;Waters et al., 2008).It had also been hypothesized that GLK1 and GLK2 genes regulate dimorphic chloroplast development in the two photosynthetic cell types of C 4 species, with each gene functioning in one of the two cell types (Rossini et al., 2001).This hypothesis was based on the observation that GLK1 genes are preferentially expressed in mesophyll cells of maize, sorghum and setaria whereas GLK2/G2 genes are preferentially expressed in bundle sheath cells (Chang et al., 2012;John et al., 2014;Li et al., 2010;Wang et al., 2014), and loss of Zmg2 function in maize perturbs bundle sheath but not mesophyll chloroplast development (Cribb et al., 2001).Through the phenotypic characterization of glk1 loss-of-function mutants in maize and setaria (Figures 1, 2 and 4), glk2 loss-of-function mutants in setaria (Figures 3 and 4), and glk2 setaria mutants complemented with maize GLK1 and GLK2 genes (Figures 5 and 6), we have revealed a more complex situation both with respect to the developmental processes operating within individual C 4 species and to likely evolutionary trajectories.Specifically, we reveal that spatially restricted GLK gene expression patterns do not predict sites of gene function.

GLK proteins have equivalent functional capacity
GLK genes duplicated in a common ancestor of the Poales, before the speciation of maize, sorghum, setaria, and rice, and before the evolution of C 4 photosynthesis (Wang et al., 2013).Given that GLK genes regulate chloroplast biogenesis in the moss Physcomitrium patens (Bravo-Garcia et al., 2009;Yasumura et al., 2005), which last shared a common ancestor with flowering plants over 450 million years ago (Kenrick & Crane, 1997;Wickett et al., 2014), it is reasonable to assume that prior to this duplication the ancestral GLK gene played a role in chloroplast development.The duplication event, which must have occurred between 85 and 95 million years ago (Gallaher et al., 2022), provided an opportunity for neo-and/or subfunctionalization of gene activity and thus for divergent roles to evolve in different grass species.It has previously been suggested that the compartmentalized expression of GLK1 and GLK2 genes in mesophyll and bundle sheath cells of the C 4 species reflects distinct roles in the biogenesis of granal versus agranal chloroplasts.This hypothesis was supported by the observation that GLK homologs are not differentially expressed in Gynandropsis gynandra (formerly known as Cleome gynandra), a C 4 dicot that exhibits granal thylakoids in chloroplasts of both mesophyll and bundle sheath cells (Marshall et al., 2007;Wang et al., 2013).However, data presented here show that appropriate cell-type specific thylakoid stacking is restored in mesophyll and bundle sheath cells of Svglk2 mutants when either ZmGLK1 or ZmG2 is constitutively expressed (Fig- ures 5 and 6).These results demonstrate that the two maize proteins have equivalent functional capacity.Furthermore, the data are consistent with a recent report which showed that the identity of genes activated by GLK transcription factors is determined by cis regulatory factors, such that a heterologous GLK protein will activate downstream targets that are normally activated by endogenous GLK proteins in the host species as opposed to targets that are activated by the heterologous protein in its species of origin (Tu et al., 2022).Together these observations suggest that the evolution of dimorphic chloroplasts in C 4 species was unlikely to have been mediated by subfunctionalization of GLK protein activity.

Despite equivalent functional capacity, GLK2 plays a dominant role in both maize and setaria
The difference in the proportion of GLK2:GLK1 transcripts in mesophyll cells of maize (5:95) versus setaria (43:57) can partially explain the different phenotypes observed in loss-of-function mutants.In maize, where relative levels of ZmG2 are only 5% in mesophyll cells, Zmg2 mutants exhibit perturbed chloroplast development only in bundle sheath cells (where ZmG2 transcripts comprise 65% of the total GLK pool).By contrast, relative GLK2 levels of 43% in mesophyll and 98% in bundle sheath cells of setaria are associated with defective chloroplast development in both cell types (Figure 3).This observation was initially counterintuitive because it would be expected that, as in C 3 plants, the high levels of SvGLK1 transcripts in mesophyll cells would compensate for loss of SvGLK2 function.However, we observed that SvGLK1 transcript accumulation was virtually abolished in Svglk2 mutants, and levels of the nonfunctional Svglk2 transcripts were also reduced (Figure 4).As such, we conclude that SvGLK2 activates expression (or transcript stabilization) both of itself and SvGLK1, and that the Svglk2 mutant phenotype reflects loss of function of both genes.Collectively, these data suggest that SvGLK2 is the primary regulator of chloroplast development in both bundle sheath and mesophyll cells of setaria.
Whereas normal mesophyll chloroplast development in the absence of GLK1 function can be explained in setaria because SvGLK2 likely compensates for loss of GLK1 function, the mutant phenotype in maize was unexpected.In every land plant species examined to date, a role for GLK genes in chloroplast development has been revealed, albeit often a redundant role with paralogs operating in the same cell-type.Because ZmGLK1 is highly expressed in maize mesophyll cells and no other GLK-like transcripts appear as possible candidates for a redundant paralog, a previously unknown mechanism must be operating in this cell-type.Although GLK gene function has previously been shown to be cell-autonomous in Arabidopsis, it is possible that a non-cell-autonomous mechanism could operate in maize.Indeed, there is some evidence of this because ZmGLK1 transcript levels are reduced when G2 function is lost, even though G2 transcript levels in wild-type mesophyll cells are only 5% of the GLK pool (Figure 4).In such a scenario, loss of GLK1 function in mesophyll cells would be compensated for by direct or indirect activity of G2 proteins in the bundle sheath cells.Any such effect would only be revealed in double Zmglk1;g2 mutants, which would have defective chloroplasts in both cell-types.Regardless of whether ZmG2 compensates for loss of ZmGLK1 function, the GLK2 paralog plays the dominant role in chloroplast biogenesis in maize, as in setaria.
Cis-regulatory sequences are more variable in GLK2 than GLK1 genes Given that GLK gene paralogs encode functionally equivalent proteins both in C 3 species where they act redundantly and in the C 4 species maize, the question arises of why the paralogs are expressed in a cell-type preferential manner in C 4 grass species (at least in the ones that have been examined).Estimates based on the topology of BEP (C 3 species only) and PACMAD (C 3 & C 4 species) clades in the Poales suggest that maize diverged from rice ~80 million years ago (Gallaher et al., 2022) and that within the PAC-MAD clade, maize and sorghum share a common history of C 4 evolution whereas independent events occurred in maize and setaria around ~45 million years apart (Gallaher et al., 2022).Both GLK gene paralogs are predominantly expressed in mesophyll cells of C 3 rice leaves, with GLK1 transcripts accumulating at higher levels than GLK2 (Hua et al., 2021), and gene function being redundant (Wang et al., 2013).Notably, both transcripts can also be detected in bundle sheath cells, at levels roughly equivalent to those of GLK2 in mesophyll cells, suggesting that the few chloroplasts that develop in rice bundle sheath cells are also regulated by the redundant action of both genes.By contrast, an element of sub-functionalization is apparent in the C 4 species maize and setaria, with the GLK2 paralog expressed preferentially in bundle sheath cells.However, data from a number of different C 3 , C 4, and intermediate species (Washburn et al., 2021) (Figure S1) invoke a more complex scenario because preferential expression of GLK2 paralogs in bundle sheath cells is evident in species where decarboxylation reactions occur in the bundle sheath chloroplasts (NADP-ME type) but not in species where decarboxylation occurs in the mitochondria (NAD-ME type) or in the cytoplasm (PEP-CK type).This suggests that preferential localization of GLK2 in bundle sheath cells is associated with specific aspects of chloroplast function as opposed to general activation of photosynthesis.
The evolution of C 4 required activation of photosynthesis and a significant increase in chloroplast volume in bundle sheath cells and as such, selection on regulatory regions of GLK2 genes to enable high levels of activity in that cell type may have been a critical step.Notably, there is more within-species sequence variation in GLK2 promoters of different rice and maize varieties than in the corresponding GLK1 promoters (Figure S6), providing more opportunity for change through drift or selection.GLK2 promoter sequences are also less conserved between species than corresponding GLK1 sequences (Figure S7).ATAC-seq data from leaves shows that the proximal promoter and 5'-UTR regions are the main sites of protein binding in GLK genes from maize, setaria, and rice (Figure S4), with noticeably more DNA binding motifs related to light-responsive elements in the 5'-UTR sequences of GLK2 genes than in equivalent regions of GLK1 genes.In maize, an enrichment of BPC motifs is seen in the ZmG2 5'-UTR, and the sequences are almost completely conserved across 1200 maize varieties.This level of sequence conservation in the 5'-UTR is in stark contrast to the rest of the promoter region which is highly variable (Figure S6).Intriguingly, very few DNA binding motifs are found in either the SvGLK1 or SvGLK2 promoters (Figure S4), and no variation relevant to the domestication process was found when compared to Setaria italica (He et al., 2023).As in ZmG2, however, BPC motifs are present in both the SvGLK2 and OsGLK2 5'-UTR sequences but they are not found in any of the GLK1 gene sequences.This observation suggests that the GLK1 and GLK2 5'-UTR sequences diverged in the last common ancestor of maize, setaria, and rice and thus prior to the evolution of C 4 .Because BPC binding proteins act to recruit polycomb complex proteins and thus mediate chromatin changes (Theune et al., 2019), this divergence may have enabled novel post-transcriptional regulation of GLK2 orthologs during the evolution of C 4 .Differences in the presence and position of upstream open reading frames (uORFs) within the 5'-UTR may have further facilitated derepression of GLK2 translation (Zhang et al., 2020) in that putative uORFs in ZmGLK1 and SvGLK1 are situated less than 10 base pairs after the transcription start site (TSS) and thus if functional are predicted to be highly repressive, whereas the uORF in ZmG2 is over 200 bp downstream of the TSS and no uORFs are predicted in SvGLK2 (Figure S7).In rice, predicted uORFs are positioned 100 bp downstream of the TSS in both genes and may thus have no differential effect.Although requiring experimental verification, collectively these observations support the hypothesis that altered GLK2 gene regulation was associated with the evolution of functional chloroplasts in bundle sheath cells of C 4 grasses.

Plant growth conditions
Maize and S. viridis seeds were germinated in hydrated vermiculite or on paper, respectively, and placed in an Aralab Fitoclima D1200 PLH incubator under 12 h light/12 h dark cycle at an intensity of 320 lmol photon m À2 sec À1 , with temperatures ranging from 31°C during the day to 22°C at night and humidity between 50-60%.About 7-10 days after sowing (DAS), seedlings were transferred to pots containing John Innes no. 2 soil mixture and irrigated with water to keep the soil moist.Plants were kept under these conditions until sampled for experimental analysis.
For genetic crosses, maize plants were transferred to larger pots with the same soil mixture supplemented with Osmocote slow-release granules.Plants were grown in a greenhouse at the Department of Biology at the University of Oxford, UK with 16 h light/8 h dark cycle, daytime temperature 28°C and night-time temperature 20°C, with supplemental light provided when natural light levels were below 120 lmol photon m À2 sec À1 .

Construct design and assembly
ZmG2 and ZmGLK1 orthologs were identified in S. viridis via phylogenetic reconstruction of the GARP gene family (see Figure S2).Guide RNAs (gRNA) to drive CRISPR/Cas9 site-directed mutagenesis were designed using the CRISPOR (Concordet & Haeussler, 2018) online tool (http://crispor.tefor.net/) to achieve high editing efficiency for the target genes ZmGLK1 (Zm00001d044785), SvGLK1 (Sevir.4G123200),and SvGLK2 (Sevir.5G003100),and to minimize off-targets.For maize, the designed gRNAs were sent to the Wisconsin Crop Innovation Centre (University of Wisconsin) to be cloned and transformed using their standard pipeline for maize transformation.For S. viridis, the gRNA sequences were synthesized as oligonucleotides with appropriate flanking sites for Golden Gate cloning (Table S1; JLF_Ox43 and JLF_Ox44 for SvGLK1; JLF_Ox51 and JLF_Ox52 for SvGLK2).Prior to cloning, the oligonucleotides were annealed and phosphorylated by incubating 10 lM of each with 1x Ligase Buffer (New England Biolabs, Hitchin, UK) and 10 U of T4 Polynucleotide Kinase (New England Biolabs, Hitchin, UK) in a final reaction volume of 15 ll at 37°C for 1.5 h.After that period, 2 ll of 0.5 M NaCl was added, and the reaction was placed in a water bath heated to 95°C for 2 min.The reaction was cooled to room temperature before being diluted 100-fold and cloned downstream of the TaU6 promoter (module pJG310) using the Golden Gate one-step one-pot protocol (Engler et al., 2009) with an appropriate restriction enzyme (Esp3I).This module was then combined with another level 1 module containing TaCas9 downstream of the maize ubiquitin promoter (pJG471) into the final acceptor and binary vector pTC278.Both modules and the vector were provided by Dr Dan Voytas (U. of Minnesota).
To overexpress ZmGLK1 and ZmG2, the coding sequences of the maize genes were domesticated to avoid recognition sites for the type II restriction enzymes used in Golden Gate assembly: BsaI, BpiI, Esp3I, and DraIII.The final sequence was synthesized as a level 0 module with the appropriate SC flanking sites (Engler & Marillonnet, 2013).ZmGLK1 and ZmG2 level 0 modules (EC17054 and EC17056) were cloned downstream of the maize ubiquitin promoter (ZmUBI pro and EC15455) and upstream of a nos terminator (tNOS and EC41421) in level 1 backbone position forward 2 using the Golden Gate one-step one-pot protocol (Engler et al., 2009).The previously described hygromycin phosphotransferase (HygR) coding sequence cloned downstream of the rice actin promoter (OsACT pro ) was used as the selectable marker (Vlad et al., 2019).Level 1 modules were assembled into the binary vector pAGM4723 to obtain the constructs EC17808 and EC17810 (hereafter referred to as 17 808 and 17 810, respectively) depicted in Figure S5b,c.25 mM phosphate buffer pH 7.2.After fixation, leaves were stored in 0.2 M Na-EDTA pH 9.0.Free-hand cross sections were imaged using a Leica SP9 confocal microscope with an Argon laser set to excite chlorophyll at 663 nm and to emit from 650-800 nm.

Ultrastructure analysis via transmission electron microscopy
Sample fixation and preparation were performed by the Electron Microscopy Facility team at the Sir Willian Dunn School of Pathology (University of Oxford).Leaf discs of 2 mm diameter were collected from leaf 5 of 40 DAS maize or 25 DAS S. viridis plants and placed into 3% glutaraldehyde in 0.06 M Sorenson's phosphate buffer pH 7.2.Samples were transferred to a Leica AMW microwave tissue processor for fixation, osmification (with 1% OsO 4 ), dehydration with an acetone series, and resin infiltration (LRW Hard).The samples were then removed from the microwave and transferred to fresh resin for 4 days, renewing the resin twice a day.The embedded samples were transferred into gelatin capsules and filled with resin for polymerization at 60°C for about 65 h.Sections of 150 nm were cut using a Leica UC7 ultramicrotome, collected onto formvar-coated copper slot grids and poststained with lead citrate.Images were acquired on a Thermo Fischer Tecnai T12 TEM with a Gatan OneView CMOS camera.
Thylakoid occupancy of chloroplasts was quantified using ImageJ.The thylakoid area was measured by adjusting the image threshold to create a binary mask on which only the thylakoid membrane was apparent.This area value was divided by the chloroplast planar area to give the thylakoid occupancy level.

Chlorophyll extraction and quantification
Leaf discs were harvested from fully expanded leaf 5 at 30 DAS for maize and 21 DAS for S. viridis, and immediately frozen in liquid nitrogen.For total chlorophyll extraction, finely powdered tissue was resuspended in 80% acetone buffered in 100 mM Tris-HCl pH 8.0 (Chazaux et al., 2022).Following a centrifugation step, the absorbance of each sample was read in technical triplicates in a FLUOstar Omega spectrometer (BMG Labtech) at 646 and 663 nm.Background absorbance was measured at 750 nm.Absorbance values recorded at 750 nm were subtracted from values recorded at 646 and 663 nm for normalization purposes.Technical replicates were averaged for each sample.Concentration of chlorophylls a + b were calculated in micrograms (lg) per leaf area using the following the equation: (19.54 x Abs 646 + 8.29 x Abs 663 ) x dilution factor (Porra et al., 1989).

Fluorescence quantification of PSII
Photosystem II (PSII) fluorescence was measured using a LiCOR Li-6800 portable photosynthesis system.LiCOR chamber conditions were set to a flow intensity of 500 umol sec À1 , delta pressure of 0.2 kPa, relative air humidity of 50%, reference concentration of CO 2 at 400 umol mol À1 and temperature exchange at 25°C.For quantification of PSII operating efficiency (PhiPSII), plants were adapted to light under the conditions described above for at least 1.5 h and actinic light intensity on the LiCOR chamber was set to 1000 lmol photon m À2 sec À1 .After clamping, each leaf was adapted for a few minutes until stabilization was reached before logging the fluorescence value.Quantification of maximum quantum efficiency of PSII photochemistry (Fv/Fm) was performed on plants that were dark-adapted for at least 16 h.LiCOR actinic lights were off.In both cases, the fluorescence action log was set to FoFm (dark), FsFm' (light), rectangular flash type with red target at 8000 lmol photon m À2 sec À1 , duration of 1000 ms, output rate of 100 Hz and margin of 5 points.

Gene expression analysis via qPCR
The mid-portion of fully expanded leaf 5 from at least three biological replicates of 30 DAS maize or 21 DAS S. viridis were harvested and immediately frozen in liquid nitrogen.Tissue was ground prior to total RNA extraction using a RNeasy Plant Mini Kit (Qiagen, Manchester, UK) according to the manufacturer's instructions.To avoid any DNA contamination, samples were treated with TURBO DNA-free DNase (Thermo Fischer Scientific, Waltham, CA, USA) according to the manufacturer's guidelines.For first strand cDNA synthesis, 100 ng of total RNA was added to 1 ll of Maxima Enzyme Mix (Thermo Fischer Scientific) and 1x Reaction Mix to a final volume of 10 ll.The mixture was incubated at 25°C for 10 min, followed by 50°C for 15 min.Enzyme inactivation was carried out at 85°C for 5 min.
Amplification reactions were performed in a StepOnePlus Real-Time PCR System (Thermo Fischer Scientific) using SYBR-Green to monitor dsDNA synthesis.The reaction mixtures contained 5 ll of diluted cDNA (1:25), 0.2 lM of each primer and 10 ll of SYBR TM Green PCR Master Mix (Thermo Fischer Scientific) in a total volume of 10 ll.The reaction cycles began with a 5 min denaturation step at 94°C, followed by 40 amplification cycles of 15 sec at 94°C, 10 sec at 60°C, 15 sec at 72°C.After each cycle, the fluorescence was measured at 60°C for 35 sec.The melting curve was produced by a cycle of 95°C for 15 sec, 60°C for 1 min, 95°C for 30 sec and finally 60°C for 15 sec.Every qPCR reaction was repeated three times to make technical replicates.
Transcript levels for multiple target genes were measured with primers designed to specifically amplify the gene of interest (Table S1).Two reference genes were used to normalize gene expression for maize (Lin et al., 2014) (Table S1; ZmCYP_F and ZmCYP_R, ZmEF1a_F and ZmEF1a_R) and S. viridis (Lambret-Frott e et al., 2015) (Table S1; JLF_Ox26 and JLF_Ox27, JLF_Ox28 and JLF_Ox29).Exported Rn values were used to calculate amplification efficiency for each primer pair and the Cq values for each qPCR reaction using the 'qpcR' package from R (Ritz & Spiess, 2008).Normalized relative expression of the target genes was calculated using the R package 'EasyqpcR' (https://www.bioconductor.org/packages//2.12/bioc/html/EasyqpcR.html), according to the DDCq model (Pfaffl, 2001) based on the previously calculated Cq and amplification efficiencies.

Statistical analysis
Datasets were tested for Normal distributions using the Shapiro-Wilk test (Shapiro & Wilk, 1965).For thylakoid coverage, chlorophyll concentration and PSII fluorescence, variance was calculated using a one-way ANOVA followed by a pairwise comparison analysis with Tukey's HSD test (Tukey, 1949).For relative gene expression, statistical significance was calculated based on a Student's t-test, comparing the mean variance of each mutant line with their respective controls.All statistical analyses were performed using scipy v.1.5.0 library and plots were prepared using matplotlib v3.5.1 library; both from Python v.3.7.6, using Visual Studio Code as a code editor.
experiments and analyzed the data.JL-F and JAL wrote the manuscript.published by Li et al. (2010) and Chang et al. (2012), and setaria data were published by John et al. (2014).PEPC and NADP-ME transcripts accumulate specifically in mesophyll and bundle sheath cells, respectively, and thus act as markers for cross-contamination of RNA samples between cell types.As such, mesophyll contamination of bundle sheath transcriptomes was 22% (Li et al), 4.5% (Chang et al), and 3.5% (John et al), whereas bundle sheath contamination of mesophyll transcriptomes was 10.7% (Li et al), 7% (Chang et al), and 6.6% (John et al).Disregarding the Li et al. (2010) data because of the high level of cross-contamination, GLK1 transcript levels were higher in mesophyll than bundle sheath cells by 8-fold in maize and 21-fold in setaria, whereas GLK2 transcript levels were higher in bundle sheath than mesophyll cells by 4.8-fold in maize and 3.2-fold in setaria.(b) Log2 fold change between bundle sheath and mesophyll transcripts for S. viridis (John et al., 2014), Z. mays (Chang et al., 2012;Li et al., 2010), and three different species of Paniceae grasses (Urochloa fusca, Panicum hallii, and Digitaria californica) (Washburn et al. 2022).All species carry out C 4 photosynthesis, but each uses a different decarboxylation pathway in the bundle sheath cells.PEPC was used as a mesophyll cell marker in each case and genes encoding the relevant decarboxylation enzyme (i.e., NADP-ME, PCK, and NAD-ME) were used as bundle sheath cell markers.Protein sequences of the extensively characterized GLK members ZmGLK1, ZmG2, OsGLK1, and OsGLK2 were used to create a HMMER profile that was used as a query to search for homologs in whole protein databases of 11 species (Zea mays, Setaria viridis, S. italica, Panicum virgatum, Sorghum bicolor, Oryza sativa, Brachypodium distachyon, Vitis vinifera, Arabidopsis thaliana, Marchantia polymorpha, and Physcomitrium patens).The presence of the MYB domain in those sequences was confirmed using InterproScan, and only the sequences containing the conserved AREAEAA domain characteristic of the GLK subfamily were selected.The resulting 25 amino acid sequences were aligned using Clustal Omega.Phylogenetic reconstruction was performed using Bayesian inference on Mr. Bayes using Poisson with fixed substitution rates as a model.Two independent MCMC runs with 1 000 000 generations were performed and no further improvement of log-likelihood of prior probabilities was observed.The tree was visualized on FigTree.Clade support on each node was assessed using Bayesian posterior probabilities.Members of the GLK1 subgroup are depicted in green and GLK2 in blue.(b) Amino acid sequence alignment showing conservation between ZmGLK1, SvGLK1, ZmG2, and SvGLK2.The diagonal lines show the HLH associated with MYB or MYB-related proteins; the AREAEAA conserved site is depicted in horizontal lines and the last exon that contains the GCT-box domain is depicted in vertical lines.showing the number of DNA binding motifs found in regulatory regions of each gene.Plant non-redundant motifs (JASPAR) were used to screen the above-mentioned regions of each gene, using the FIMO tool with a P value of <0.0001.The photosynthesis-related/light responsive motifs (BPC, ABI3-like, GT-1, GT-4, CNA, MYB-B, STZ, bZIP/bHLH(G-box), GATA, MYB-like/I-box, TCP, TGA, and WRKY) were as described by Sing et al. (2023) and rice and maize GLK-specific binding sites were as described in Tu et al. (2022).(c,d) Schematics of 800 bp upstream of the transcription start site plus the 5'-UTR regions for each gene, showing the position of all predicted DNA binding motifs (c) and of known light-responsive motifs (d).The 5'-UTR is shown in white and promoter sequence is in gray. .Sequence variability in regulatory regions of GLK genes.The SNPs from 3025 variants of rice (SNP seek, IRRI) and 1210 variants of maize (Maize SNPDB) were used to calculate the frequency of each nucleotide for each SNP position.The nucleotide with a higher prevalence in each position is plotted on the graphs for OsGLK1, OsGLK2, ZmGLK1, and ZmG2.The 2Kb upstream of the transcription start site is shown in gray, the 5'-UTR in white, and the first exon in black.Figure S7.Alignment of GLK gene promoter and 5'-UTR regions.(a, b) GLK1 (b) and GLK2(a) sequences (800 bp upstream of the transcription start site (TSS) plus the 5'-UTR) aligned using MAFFT (Katoh et al., 2019) and visualized using JALview (Waterhouse et al., 2009).Conserved bases are indicated by blue shading, the predicted TATA box by the red squares, actual (rice- Murray et al., 2022), or predicted TSS by orange squares and putative uORFs by yellow squares.(c) Amino acid sequences of the predicted uORFs in (a) & (b).Table S1.Oligonucleotides used for cloning, genotyping, and qPCR.Table S2.Data and statistical analyses supporting Figures 1-6.

Figure 1 .
Figure 1.Chloroplast development is not perturbed in Zmglk1 mutants of maize.(a) Gene model of ZmGLK1 showing predicted exons in black boxes and the target site of CRISPR/Cas9 mediated mutagenesis in yellow and red below.The HLH associated with MYB or MYB-related proteins is depicted by diagonal lines; the AREAEAA conserved site is highlighted with horizontal lines and the last exon that contains the GCT-box domain is depicted by vertical lines.The table shows the sequence of the wild-type (WT) and mutated alleles with the predicted protein length.Each line represents a single transformation event.In the case of line 3b3 the T-DNA could not be segregated away and thus mutations continued to be generated in each generation.The phenotyped plants all lacked a wild-type allele but contained different combinations of the four mutant alleles indicated.The gRNA sequence is highlighted in yellow and the PAM sequence in red.(b,c) Whole plant phenotype of wild-type (b) and a representative Zmglk1 mutant 30 days after sowing (c).Scale bars = 10 cm.(d-g) Electron micrographs of chloroplast ultrastructure in bundle sheath (BSC; d,e) and mesophyll (MC; f,g) cells of WT (d,f) and a representative Zmglk1 mutant (e,g).Scale bars = 2 lm.(h-l)Measurements of thylakoid occupancy in mesophyll cells (h), thylakoid occupancy in bundle sheath cells (i), total chlorophyll (j), PSII operating efficiency (k), and maximum quantum efficiency of PSII photochemistry (l).The WT segregant is depicted in blue and the four independent Zmglk1 lines are depicted in green.Each dot represents a biological replicate, the square is the average value for that line and the bars are the associated standard deviations.A minimum of five biological replicates were evaluated for thylakoid occupancy and chlorophyll quantification, and three biological replicates were assessed for PSII fluorescence.Different letters indicate statistically different groups (P value ≤0.05, one-way ANOVA and Tukey's HSD; see TableS2for raw data).

Figure 2 .
Figure 2. Chloroplast development is not perturbed in Svglk1 mutants of setaria.(a) Gene model of SvGLK1 showing predicted exons in black boxes and the target site of CRISPR/Cas9 mediated mutagenesis in yellow and red below.The HLH associated with MYB or MYB-related proteins is depicted by diagonal lines; the AREAEAA conserved site is highlighted with horizontal lines and the last exon that contains the GCT-box domain is depicted by vertical lines.The table shows the sequence of the wild-type (WT) and mutated alleles with the predicted protein length.The gRNA sequence is highlighted in yellow and the PAM sequence in red.(b-d) Whole plant phenotype of wild-type (b) and Svglk1 mutants 20 days after sowing (c,d).Scale bars = 5 cm.(e-j) Electron micrographs of chloroplast ultrastructure in bundle sheath (BSC; e-g) and mesophyll (MC; h-j) cells of WT (e,h) and Svglk1 mutants (f,g,I,j).Scale bars = 2 lm.(k-o) Measurements of thylakoid occupancy in mesophyll cells (k), thylakoid occupancy in bundle sheath cells (l), total chlorophyll (m), PSII operating efficiency (n) and maximum quantum efficiency of PSII photochemistry (o).The WT segregant is depicted in blue and the two independent Svglk1 lines are depicted in green.Each dot represents a biological replicate, the square is the average value for that line and the bars are the associated standard deviations.Twelve chloroplasts were assessed for thylakoid occupancy quantification, and five biological replicates were assessed for chlorophyll quantification and PSII fluorescence for each sample.Different letters indicate statistically different groups (P value ≤0.05, one-way ANOVA and Tukey's HSD; see TableS2for raw data).

Ó
2023 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2024), 117, 416-431

Figure 3 .
Figure 3. Chloroplast development is perturbed in both mesophyll and bundle sheath cells of Svglk2 mutants of setaria.(a) Gene model of SvGLK2 showing predicted exons in black boxes and the target site of CRISPR/Cas9 mediated mutagenesis in yellow and red below.The HLH associated with MYB or MYB-related proteins is depicted by diagonal lines; the AREAEAA conserved site is highlighted with horizontal lines and the last exon that contains the GCT-box domain is depicted by vertical lines.The table shows the sequence of the wild-type (WT) and mutated alleles with the predicted protein length.The gRNA sequence is highlighted in yellow and the PAM sequence in red.(b) Whole plant phenotype of wild-type, a heterozygous and a homozygous Svglk2 mutant 20 days after sowing.Scale bar = 5 cm.(c,d) Confocal images of leaf cross sections showing overall chloroplast structure in WT (c) and the Svglk2 mutant (d).Bundle sheath cells are outlined in white.Scale bars = 50 lm.(e-h) Electron micrographs of chloroplast ultrastructure in bundle sheath (BSC; e,f) and mesophyll (MC; g,h) cells of WT (e-g) and the Svglk2 mutant (f,h).Yellow arrows point to fragmented membranes.Scale bars = 2 lm.(i-k) Measurements of thylakoid occupancy in bundle sheath cells (i), thylakoid occupancy in mesophyll cells (j) and total chlorophyll (k).The WT segregant is depicted in purple and the Svglk2 line is depicted in pink.Each dot represents a biological replicate, the square is the average value for that line and the bars are the associated standard deviations.Twelve chloroplasts were assessed for thylakoid occupancy quantification, and five biological replicates were assessed for chlorophyll quantification and PSII fluorescence for each sample.Different letters indicate statistically different groups (P value ≤0.05, one-way ANOVA and Tukey's HSD; seeTable S2 for raw data).

Figure 4 .
Figure 4. GLK1 and GLK2 transcript levels are reduced in Svglk2 and Zmg2 mutants of setaria and maize.(a-f) Quantitative analysis of gene expression via qPCR in segregating wild-type (WT) siblings and glk mutants of maize (a-d) and setaria (e,f).Relative normal- ized expression of ZmGLK1 (a), ZmG2 (b), ZmGNC (c), ZmCGA1 (d), SvGLK1 (e) and SvGLK2 (f) was calculated based on the expression of two reference genes for each sample.Each dot represents a biological replicate, the square is the mean value and the bars are the associated standard deviations.A minimum of three and five biological replicates were evaluated for maize and setaria samples, respectively.Statistical significance was calculated using a Student's t-test and pairwise comparison with the respective WT sample.For each species comparison, different letters indicate statistically different groups (P value ≤0.05; see TableS2for raw data).

Figure 5 .
Figure 5. ZmGLK1 and ZmG2 complement the Svglk2 phenotype and activate transcription of the setaria homologs.(a,b) Whole plant phenotype of Svglk2 mutant overexpressing ZmGLK1 (a) or ZmG2 (b) 30 days after sowing.Scale bars = 5 cm.(c-f) Quantitative analysis of gene expression via qPCR of ZmGLK1 (c), ZmG2 (d), SvGLK1 (e) and SvGLK2 (f).Relative gene expression was calculated based on the expression of two reference genes for each sample.Each dot represents one of the five biological replicates evaluated, the square is the mean value and the bars are the associated standard deviations.Statistical significance was calculated using a Student's t-test and pairwise comparison with the respective wildtype (WT) sample.Different letters indicate statistically different groups (P value ≤0.05; see TableS2for raw data).

Figure 6 .
Figure 6.ZmGLK1 and ZmG2 are sufficient to induce dimorphic chloroplast development in both mesophyll and bundle sheath cells of setaria.(a-c) Confocal images of leaf cross sections showing overall chloroplast morphology in the Svglk2 mutant (a, same as Figure 3d), and representative lines expressing ZmGLK1 (b) or ZmG2 (c).Bundle sheath cells are outlined in white.Yellow arrows indicate ectopic chloroplasts developing in vascular cells.Scale bars = 50 lm.(d-i) Electron micrographs of chloroplast ultrastructure in bundle sheath (BSC; d-f) and mesophyll (MC; g-i) cells of the Svglk2 mutant (d-g; same as Figure 3g,h) and one representative line overexpressing ZmGLK1 (e-h) or ZmG2 (f-i).Scale bars = 2 lm.(j-l)Measurements of thylakoid occupancy in mesophyll cells (j), thylakoid occupancy in bundle sheath cells (k) and total chlorophyll (l).The wild-type (WT) segregant is depicted in purple, the Svglk2 line in pink, the three independent lines overexpressing ZmGLK1 in orange and those overexpressing ZmG2 in yellow.Each dot represents a biological replicate, the square is the average value for that line and the bars are the associated standard deviations.A minimum of 10 chloroplasts were evaluated for thylakoid occupancy quantification and five to ten biological replicates were used for chlorophyll quantification.Different letters indicate statistically different groups in (P value ≤0.05, one-way ANOVA and Tukey's HSD; see TableS2for raw data).

Ó
2023 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2024), 117, 416-431

Figure S3 .
Comparison of GLK sequences.(a) Phylogenetic reconstruction of the GLK gene family.

Figure S4 .
DNA binding motifs in regulatory regions of GLK genes.(a) Schematics of maize, setaria, and rice GLK genes showing binding sites revealed from ATAC assays with leaf RNA.Data was retrieved from https://epigenome.genetics.uga.edu/PlantEpigenome/index.html.The yellow shading highlights the 800 bp upstream of the transcription start site plus the 5'-UTR regions.(b) Table

Figure S5 .
Phenotypic characterization of Svglk2 mutant lines complemented with ZmGLK1 or ZmG2.(a) Schematic of the gene-edited Svglk2 mutant allele.(b,c) Schematic of the constructs used to express ZmGLK1 (b) or ZmG2 (c) in the Svglk2 background.HygR depicts the hygromycin phosphotransferase gene and OsACT pro and ZmUBI pro represent the constitutive rice actin and maize ubiquitin promoters, respectively.LB and RB refer to left and right borders.(d-m) Transmission electron microscopy images showing mesophyll (MC) and bundle sheath (BSC) cell chloroplast Ó 2023 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2024), 117, 416-431 ultrastructure in the mutant line (d,e) and in lines overexpressing ZmGLK1 (f-i) or ZmG2 (j-m) in the mutant background.Scale bar sizes are indicated on each image.Panels d & e are the same images as Figure 3g,h).(n-r) Confocal images of leaf cross sections from the mutant (n) and from lines overexpressing ZmGLK1 (o,p) or ZmG2 (q,r) showing the morphology of chloroplasts.Bundle sheath cells are outlined in white.Scale bars = 50 lm.Yellow arrows indicate ectopic chloroplast formation in vascular cells Figure S6 Table S2 for raw data).
Table S2 for raw data).
Ó 2023 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2024), 117, 416-431 Table S2 for raw data).
Ó 2023 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2024), 117, 416-431