Elemental Analysis
The leaves in the tillering period were collected and used for elemental analysis, among which the els1 mutant showed a few cell death dots. The tissues were completely dried in an oven, and the samples were treated as described by Hosmani et al (2013)(Hosmani et al., 2013; Z. Wang et al., 2019). Elemental analysis was performed on a CIC-260 ion chromatograph (K, Na, Mg, and Ca) and ICP-MS (Cd, As, Mo, Fe, and Mn), and nine elements were monitored.

RNA-seq

The second and third leaves from the top at the tillering stage were collected separately. There was an obvious cell death phenotype in the third leaf and no visible cell death phenotype in the second leaf at this stage in the els1 mutant (Fig. S19A). Each treatment included two biological repeats. Total RNA was extracted using the method of Bilgin et al.(Bilgin, DeLucia, & Clough, 2009) and then sent to Novogene company (service@novogene.com) for RNA-seq analysis. The resulting sequence data including 33.8 million ~ 42.6 million paired-reads to each sample was provided for preliminary analysis. The differentially expressed genes (DEGs) were revealed with the DeSeq programme(Andersen, Barberon, & Geldner, 2015), and the significant DEGs were compared by applying a Padj-value cut-off less than 0.05 (FDR BH corrected). The enrichment of DEGs in the gene ontology was analysed with g:Profiler (https://biit.cs.ut.ee/gprofiler/gost), and the enrichment in the KEGG pathway was revealed with KOBAS (2.0)(Xie et al., 2011). The Padj-value denotes the significance of a GO term enrichment in DEG clusters and/or pathway correlations (Padj-value <= 0.05 was considered significant).

Results

Characterization of the rice els1mutant and cloning of the ELS1gene

The els1 mutant was discovered in a paddy field and came from a high generation progeny of Jinhui2629 and TR-2. The mutant and wild-type (WT) used in the study showed stable inheritance. The phenotype of theels1 mutant appears at the beginning of the tillering stage and is not obvious before this stage. Cell death usually begins at the blade tip or upper part of the leaf and then expands through the whole leaf along with growth. The els1 mutant exhibits fewer tillers and a reduced plant height at the adult stage (Fig. 1A~C and Fig. S4A~D).
To reveal the genetic basis of els1 mutant, we performed a cross between the els1 mutant and Nipponbare and investigated 426 individual plants from the F2 population. Genetic analysis showed that the els1 mutant was controlled by a single recessive gene. We mapped the els1 locus on chromosome 4 and delimited the locus to approximately a 265-kb region between the insertion-deletion (Indel) molecular markers ID4-3 and ID4-10 using the F2 population. To further fine-map the els1 gene, we selected several F2 individual plants with a heterozygous genotype in the candidate region to establish the F3 population. The els1 locus was narrowed down to an 11.3-kb region between ID4-371 and ID4-3-8 using Indel markers, which included two genes - Os04g0684200 and Os04g0684300 (Fig. S4E). We sequenced the two genes and found a 15-bp deletion in the coding region of the Os04g0684300 gene in the els1 mutant and no mutation in the Os04g0684200 gene (Fig. 1D). Os04g0684300 encodes a Casparian strip membrane domain protein (CASP1), and the mutation of Os04g0684300 could induce early leaf senescence.
To further validate CASP1, we performed genetic complementation by introducing the OsCASP1 gene with its promoter into the callus from F3 seeds, which arose from a few individuals from the F2 population that exhibited mutant and Nipponbare-like phenotypes to reduce transformation difficulty. The results showed that all transgenic lines carrying the OsCASP1gene had a restored WT phenotype (Fig. 1E and 1F). We also generated transgenic plants to knock out the OsCASP1 gene with CRISPR-cas9 technology, and all the transgenic plants with the homozygous genotype exhibited obvious early leaf senescence, sterility or eventual death (Fig. S5). We found that some transgenic plants with the heterozygous genotype also showed mutant phenotypes, and further analysis revealed that both alleles at the OsCASP1 locus contained mutant sites (Fig. S5).

OsCASP1 was mainly localized on the nuclear membrane

To determine the subcellular localization of OsCASP1, we generated anOsCASP1pro:OsCASP1-GFP construct and transformed it into Zhonghua 11. Unexpectedly, we observed that the green fluorescence was concentrated in the nucleus in the lateral rootlet tips (Fig. 2A). The localization of OsCASP1 differed from that of AtCASPs, which was present at the CSD in the endodermis. We also noticed some green fluorescence on the plasma membrane in the main and lateral root elongation zones, but it was difficult to distinguish the fluorescence from GFP or autofluorescence (Fig. S6A). Thus, we transformed35S::DsRed-OsCASP1 and OsCASP1pro:OsCASP1-GFP into tobacco, onion and Arabidopsis protoplast, respectively. All of these results indicated that OsCASP1 protein was mainly localized to the nucleus (Fig. 2B and Fig. S6). Because the protein contains transmembrane domains, we speculated that OsCASP1 should be localized to the nuclear membrane. Moreover, the fluorescence was also detected on the plasma membrane in tobacco and onion, albeit with a much weaker signal (Fig. 2B, Fig. S6B~D). This result implied that the protein could be located on the plasma membrane. To reveal whether OsCASP1 was present on the plasma membrane in rice, we transformed35S::DsRed-OsCASP1 into Zhonghua 11 and detected red fluorescence on the plasma membrane in the small lateral root (Fig. S7A). We observed strong red fluorescence in the nuclei and some other intracellular compartments in roots (Fig. S7), which could be endoplasmic reticulum attached to the nucleus. Additionally, OsCASP1 protein aggregated in this region due to overexpression of this gene (Fig. S6E and S7D). The strong red fluorescence was concentrated in the stele of roots (Fig. S7C and S7E) and did not restrict the localization within the CS domain.
Because the cellular position of OsCASP1 differed from that of AtCASPs, and OsCASP1could contain the fifth transmembrane according to the TMHMM prediction (Sonnhammer, von Heijne, & Krogh, 1998) (Fig. S2), we generated two constructs containing the N-terminal region (1~50) and C-terminal region (62-224) of CASP1, respectively, and transformed them with the nuclear marker and the endoplasmic reticulum marker into rice protoplasts. We compared the protoplast cells of 35S::DsRed-OsCASP1-N with that of35S::DsRed and discovered no obvious difference. In addition, the red fluorescence from DsRed-OsCASP1-N did not completely overlap with the green fluorescence from the endoplasmic reticulum marker (Fig. S6F). These results indicated that OsCASP1 did not possess the fifth transmembrane domain. Further experiment revealed that the C-terminal region with four transmembrane domains was also localized in the nuclear membrane and endoplasmic reticulum similar to whole OsCASP1 protein. This result suggested that the localization signal was in the C-terminal region of OsCASP1 (Fig. 2 and Fig. S6).

Pattern and induction of OsCASP1 gene expression.

To detect the expression pattern of OsCASP1 , we transformedOsCASP1pro:OsCASP1-GUS into Zhonghua 11 and observed very strongOsCASP1 expression at the tips of small lateral roots (SLRs) and weak expression in other regions of the roots but not all tips of SLRs. Salt stress strongly induced OsCASP1 expression in roots and leaves. We examined the cross and longitudinal sections of primary root after staining and found that GUS activity was mainly concentrated in the stele of roots. We also discovered OsCASP1 gene expression in other root tissues treated with NaCl, especially sclerenchyma cells (Fig. 3 and Fig. S8). We could not detect OsCASP1 expression in leaves at the adult stage using GUS staining method and RT-PCR.

No significant difference of the primary root structure between els1 and WT

The upper results suggested that the function of OsCASP1 should differ from that of AtCASPs. To reveal the function of OsCASP1 in the root, we first checked the structures of the primary roots (embryonic crown roots, postembryonic crown roots) and large lateral roots of theels1 mutant by separate staining with phloroglucinol, berberine-aniline blue, and periodic acid. Though we observed the cross-sections from different zones of many primary roots and/or large lateral roots, we found no observe obvious differences in the root structure between the WT and els1 mutant, including the CS structure (Fig. S9 and S10). These experiments were repeated many times, and we found no significant difference in the endodermis and exodermis between the WT and els1 mutant. To reveal the fine structures of CSs, we used electron microscope to observe the sections of primary roots and found no difference between the CSs of els1 and WT. We also checked the CS after treatment with salt stress and still found no CS difference between the mutant and WT (Fig. 4A).
We evaluated the suberin deposition at different zones of primary roots stained with Fluorol Yellow 088 in 7-day-old, 28-day-old, and 35-day-old seedling, respectively. Because there were considerable differences among roots in WT or the els1 mutant, we could not obtain a consistent result that indicated a significant difference in suberin formation between our mutant and WT.

The small lateral roots of the els1mutant display ectopic suberin deposition

We then evaluated suberin deposition in SLRs with Fluorol Yellow 088. The results showed that the patterns of suberin deposition depended on the position of SLRs in the primary root. The newborn lateral roots near the primary root tip (2~3 cm behind the root apex) usually had no suberin deposition, and the SLRs without suberin deposition occurred more frequently in the WT than in the mutant (Fig. 4C). The SLRs that were far from the primary root tip in WT usually exhibited even colouring or ectopic deposition in a few cells. This result indicated that suberin deposition was evenly distributed along the SLRs in the WT. However, most SLRs in els1 showed ectopic deposition and an uneven distribution along the SLRs, and suberization increased along with root growth. Suberin deposition in SLRs of the WT was sensitive to environmental conditions, and SLRs among different seedlings showed greater differences in suberin deposition (Fig. 4B, 4C, and Fig. S11). To determine the position of ectopic suberin deposition, we observed the cross-sections of the SLRs and found that ectopic suberin deposition mainly occurred in endodermal cells in theels1 mutant (Fig. 4D and Fig. S12). We also observed strong fluorescence in the sclerenchyma, which could allow the SLRs of WT to exhibit continuous fluorescence (Fig. 4D and Fig. S13). We were not sure whether the strong fluorescence of sclerenchyma arose from autofluorescence and/or suberin stained with Fluorol Yellow 088, but the sclerenchyma could be coloured with Sudan Red 7B, which is also used to stain suberin(Brundrett, Kendrick, & Peterson, 1991; Schreiber, Franke, Hartmann, Ranathunge, & Steudle, 2005). We observed the SLRs stained with berberine-aniline blue and discovered patchy white-blue zones from endodermal cells in the els1 mutant, as observed with Fluorol Yellow 088. However, the SLRs of the WT showed continuous white-blue fluorescence, which was strong or weak. This fluorescence could arise from both lignin and suberin stained with berberine-aniline blue and mainly from sclerenchyma cells of WT (Fig. S13).

The apoplastic barrier function ofels1roots

Propidium iodide (PI) was used as an apoplastic tracer to reveal the functional apoplastic barrier in roots of Arabidopsis(Alassimone, Roppolo, Geldner, & Vermeer, 2012), thus we applied this substance to rice. The results showed that the CS of WT rice could not hinder PI entry into the stele. However, different zones of the primary root showed different permeabilities to PI, and the zone far from the root tip exhibited more retardation of PI. This retardation did not seem to come from CS in exodermis and endodermis (Fig. S14A). The PI concentration (10 µg/ml) used to stain the roots of Arabidopsis did not seem to be sufficient for the primary roots of rice, which often causes only partial staining of transverse sections (Fig. S14A). We attempted to use a higher concentration of PI (500 µg/ml) and found that whole root was uniformly coloured (Fig. S14C). These results indicated that the primary roots could not completely block the penetration of PI into stele. We also detected the SLRs with the lower concentration PI, and the result suggested that the stele of root was strongly coloured. In addition, there was no difference in permeability to PI between the WT andels1 mutant (Fig. S14B). These results suggested that the CS structure and function of rice were different from those ofArabidopsis .
Periodic acid and berberine were used as apoplastic tracers in rice, and they were blocked at the outside of the exodermis of the WT roots under stagnant deoxygenated conditions (Shiono et al., 2014). We detected the permeability of CS with berberine chloride and periodic acid and found that the different root regions showed different permeabilities to berberine and periodic acid, and the region close to the root tip exhibited stronger permeability. However, we did not discover significant differences between els1 and WT (Fig. S10 and S15). We detected the root permeability by staining with phloroglucinol for different lengths of time and found that the staining of sclerenchyma and stele cells was stronger in the WT than the mutant (Fig. S16). The results indicated that the permeability of the exodermis and endodermis to phloroglucinol might be lower in the mutant than in the WT.
The els1 mutant displays ion homeostasis defects and different sensitivities to different nutrient stresses
Mutation of OsCASP1 changed the deposition of suberin in SLRs and probably altered the ion permeability of the mutant. We measured the content of 9 metal elements in leaves in the tillering period, in which cell death partly appeared. We found a significant accumulation of iron, ­­­­­­­­­­manganese, and sodium and a reduction in potassium and arsenic in the els1 mutant (Fig. S17). The results indicated that theOsCASP1 mutation altered ion uptake in the root, and the report by Wang et al. (2019) also indicated that the shoots of theOscasp1 mutant display nutrient homeostasis defects (Z. Wang et al., 2019). We speculated that the disorder of ion homeostasis in plants resulted in the mutant phenotype, and then we tested the growth of els1 plants in nutrient-poor solution in a climate incubator and found that the mutant displayed distinct phenotypes in different stress conditions. There was no visible leaf cell death in the els1 mutant in complete medium. The mutant was insensitive to the deficiency of phosphorous, iron, or nitrogen and to a high concentration of phosphorous or iron (data not shown), and there were slight differences in the leaves betweenels1 and WT in medium without potassium, magnesium, or aluminium (Fig. 5A~C, and Fig.S18A~C). The WT showed early senescence of lower leaves in medium without magnesium or with aluminium (Fig. 5B, 5C, Fig. S18B, and S18C). WT plants were sensitive to the medium with a low pH value (pH = 4.0) and showed no obvious differences in tillers under this condition from the mutant (Fig. 5D and Fig. S18D). In addition, the mutant showed curled and dry leaves in the medium with cadmium (Fig. S18E) and was sensitive to high concentrations of NaCl (Fig. 5E). These results suggested that theels1 mutant exhibited different sensitivity to different nutrient stresses, which could be relevant to the composition of mineral ions in the mutant plant.

The function of the chloroplast is depressed and nutrient recycling is enhanced in the els1mutant

To reveal the mechanism that regulates cell death, we performed RNA-seq with the second and third leaves from the top, respectively. There was no obvious cell death phenotype in the second leaf and an obvious phenotype in the third leaf in the els1 mutant (Fig. S19A). We separately found 2752 and 1668 differentially expressed genes (DEGs) in second and third leaves between the WT and els1 mutant (Fig. S19B). To achieve a functional annotation of the genes related to leaf cell death, we performed GO enrichment analysis. The significantly enriched DEGs were mainly concentrated in the chloroplast and downregulated in the els1 mutant compared with the WT. Accordingly, the DEGs related to photosynthesis as a biological process and electron transfer activity as a molecular function were significantly enriched (Table S3). In addition, downregulation of genes involved in the response to light stimulus were enriched in the mutant, and most proteins encoded by these genes were located in the chloroplast. KEGG pathway enrichment analysis revealed that the DEGs encoding proteins involved in the light-harvesting chlorophyll protein complex, electron transport chain and ATP-synthesizing apparatus in the thylakoid membrane and carbon fixation of photosynthesis were downregulated (Fig. S19). We also noticed that the DEGs involved in ion homeostasis and transport biology process and ion binding molecular function were also significantly enriched and upregulated in the second leaf of the mutant (Table S3). All these results implied that the ion imbalance in the mutant could induce autophagy and nutrient recycling and mobilization from the lower leaf to sink tissues. We then checked the DEGs and found that some autophagy-related genes (ATGs) were upregulated in the mutant and significantly enriched in macroautophagy, which encode key components for autophagosome formation and participate in autophagic process(Q. Chen et al., 2019; Have, Marmagne, Chardon, & Masclaux-Daubresse, 2017) (Fig. S20A). The changes in expression of these genes and repression of chloroplast function indicated that autophagy was initiated in the second leaf of the mutant. Nutrient recycling also requires various proteases and transporters, and chloroplast material and unwanted cytoplasmic material are driven to the vacuole by autophagosome for degradation by proteases and hydrolases, and the released nutrients (including amino acids, organic acids, glyceride, and mineral elements) are exported to the cytosol and then transported to newer tissue via transporters(Q. Chen et al., 2019). Thus, we checked the DEGs of proteases, proteases inhibiters, transporters, and channels. The genes of proteases and proteases inhibiter were selected according to figure 2 and table S1 published by Have et al. (2019), which are orthologues of these Arabidopsisgenes(Q. Chen et al., 2019). The results showed a significant enrichment of the genes encoding aspartic-type peptidase, cysteine-type peptidase, serine-type peptidase, and metallopeptidase and the peptidases in the vacuole and chloroplast envelope (Fig. S20B~D), which take part in chloroplast dismantling (Q. Chen et al., 2019; Have et al., 2017). There was an enrichment of genes encoding anion transporters (including amino acids, phosphate, and sulphate transporters), cation transporters (including metal ion transporter), channels, and solute:cation symporters, which participate in nutrient remobilization (Q. Chen et al., 2019; Have et al., 2017) (Fig. S21). These results suggested that nutrient starvation in the mutant plants resulted in nutrient recycling and accelerated lower leaf senescence.

The mechanism of early leaf senescence in the els1mutant

We carefully checked the DEGs and their orthologs in Arabidopsisto reveal genes that directly regulate leaf senescence and in which mutations result in leaf cell death. Many of these DEGs were uncovered, even in the second leaf, which implied that cell death was initiated in the second leaf. Further analysis showed that these DEGs could be divided into two groups, and the changes in gene expression in group 1 could cause leaf cell death in the mutant. The effects of the genes in group 2 were opposite to those in group 1, which showed expression in the mutant that should prevent cell death (Fig. 6)(Bruggeman, Raynaud, Benhamed, & Delarue, 2015; Leng, Ye, & Zeng, 2017; Woo et al., 2019). It was difficult to understand the changes in gene expression in group 2, which could be induced or repressed by leaf cell death. These and the above results implied that the metabolism in the mutant maintained a fragile balance under the right conditions, and the balance was easily broken by some adverse environmental factors, such as nutritional stress, which should result in dominance of the genes in group 1 leading to early leaf senescence.
More than half of the genes in group 1 encoded transcription factors (TFs), and most of the transcription factors belonged to NAC TFs. Among them, overexpression of OsNAP genes caused early leaf senescence, and suppression of the gene expression delayed leaf senescence (Liang et al., 2014; Zhou et al., 2013). AtNAP binds to the promoter of AtSAG113 (protein phosphatase 2C) and regulates gene expression, and SAG113 in turn inhibits stomatal closure in leaves, which results in faster water loss and consequently triggers leaf senescence(Zhang & Gan, 2012).OsNAP and OsSAG113 genes were significantly upregulated in the els1 mutant (Fig. 6A and 8C), and OsNAP should use the same route to modulate leaf senescence and regulate chlorophyll breakdown. Chlorophyll breakdown is widely used as a tool for monitoring physiological senescence in plants. The expression of genes involved in chlorophyll breakdown, such as ACD1 (OsPAO ), ACD2(OsRCCR1 ), SGR1 (OsSGR ), AND PPH(OsNYC3 ), was directly regulated by OsNAP/AtNAP, AtORE1, and AtNAC019/055/072(Liang et al., 2014; Woo et al., 2019). All of these genes were upregulated in the els1 mutant, which accelerated the degradation of chlorophyll a (Fig. 6) and could result in chlorosis and cell death in the mutant leaves. In addition, these overexpressed NAC-TFs result in cell death, and the expression levels of these genes influence a variety of processes associated with the induction of senescence and modulate the expression of SAGs except upper genes(Woo et al., 2019).

Discussion

Recently, OsCASP1 has been characterized and was thought to be required for CS formation at endodermal cells (Z. Wang et al., 2019). To reveal the function of OsCASP1, three aspects of research were needed. The first concerns the CS structure of theOscasp1 mutant, which was thought to be different from that of the WT based on the photographs of confocal laser-scanning microscope (Z. Wang et al., 2019). However, our transmission electron micrographs displayed more clear structure of CS and showed no difference between the mutant and WT (Fig. 5A). The second concerns the localization of OsCASP1, which was considered to accumulate at the CSD of rice roots and on the cell membrane in protoplast. This conclusion could be from the result of autofluorescence of CS and the oversight to the strong green fluorescing area at plasma membrane, which could be the nucleus (Z. Wang et al., 2019). Our results indicated that OsCASP1 was mainly located on the nuclear membrane. We also detected the protein on the cell membrane, especially in overexpressed transgenic plants, but the signal was weak. Moreover, its expression was concentrated at the SLR tip and in the root stele (Fig. 2, Fig. S6 and S7). The third concerns PI penetration, which was used to detect the permeability of the CS in Arabidopsis(Barberon, 2017; Hosmani et al., 2013; Naseer et al., 2012). However, our results indicated that PI was not suitable for detecting the CS permeability of rice. In short, OsCASP1 could not accumulate at the CSD and regulate CS formation.
Until now, there have been no reports showing a CASP or CASPL protein mainly localized on the nuclear membrane, and only one CASPL has been detected in the endoplasmic reticulum(Roppolo et al., 2014). In addition, CASPL genes can be expressed in tissues other than the endodermis(Roppolo et al., 2014), andAtCASPL4C1 is widely expressed in a variety of organs and is cold-inducible; its mutant shows elevated tolerance to cold stress(Yang et al., 2015).AtCASPL1D2 is exclusively expressed in suberized endodermal cell and could regulate the deposition of suberin induced by NaCl stress (Champeyroux et al., 2019). The function of OsCASP1 differs from that of AtCASPs and can appear similar to some AtCASPLs. OsCASP1 is strongly expressed in sclerenchyma cell and could regulate the suberin deposition of sclerenchyma (Figure S13), andOscasp1 mutant is more sensitive to upland condition(Z. Wang et al., 2019) and shows different tolerances to different nutrient stresses (Fig. 5 and S18). These result implied that OsCASP1 could play a unique role in the adaption to the growth condition. How OsCASP1 on the nuclear membrane functions is a new and interesting topic. We found that OsCASP1 could interact with some membrane transcription factors (data not shown), which suggested that OsCASP1 could regulate suberin deposition by modulating the activity of the transcription factors.
We found no obvious difference between Oscasp1 and WT in primary roots, but the recent report indicated that there was a clear distinction of suberin deposition between Oscasp1 and the WT by analyses of cross-sections. Unfortunately, the conclusion lacked statistical support (Z. Wang et al., 2019). We observed many cross-sections from different zones of many primary and large lateral roots and could not obtain consistent results to support the difference. In Arabidopsis , the number of endodermal cells from the first fully expanded cell can be counted to quantitatively describe the formation of the CS, the deposition of suberin, and permeability to PI (Hosmani et al., 2013). We attempted to detect autofluorescence of CS with whole SLRs like Arabidopsis , but the fluorescence emitted by CS was masked by the lignified sclerenchyma cells, which were outside the endodermis in rice and emitted strong fluorescence. Until now, there has been no appropriate method to observe CS in SLRs using the whole mount method. However, we detected ectopic deposition of suberin in the mutant SLRs after staining with Fluorol Yellow 088 or berberine with whole SLRs. SLRs contain more root hairs than primary and large lateral roots and should be the main tissue for nutrient uptake. OsCASP1 mutation results in a nutrient imbalance in the plant (Fig. S17)(Z. Wang et al., 2019), which should be due to the ectopic suberin deposition in the SLRs.
Leaf senescence is a developmental process, in which cellular structures and biomolecules are progressively broken down to reallocate nutrients to juvenile and reproductive organs. This process enhances the chances of plant survival to adapt to biotic/abiotic stresses (Q. Chen et al., 2019). Nitrogen is the element that is predominantly remobilized during leaf senescence, along with potassium. Many other elements are also remobilized, although less efficiently than nitrogen (Have et al., 2017; Himelblau & Amasino, 2001). OsCASP1 mutation leads to mineral element imbalance in the mutant shoot(Z. Wang et al., 2019), which enhances autophagy activity and accelerates nutrient recycling(Q. Chen et al., 2019). Our results indicated that the DEGs involved in nutrient recycling were significantly enriched, and the main functions of the chloroplast were depressed in the mutant, even in the second leaf, which showed no visible cell death phenotype. As state above, we rendered a model to address the mechanism of early leaf senescence in the Oscasp1mutant (Fig. 7).

ACKNOWLEDGMENTS

We thank Songbiao Chen of Institute of Oceanography, Minjiang University, Fuzhou, China for providing vectors and protein-tagged organelle markers, and Fangyu Chen of College of Agriculture, Fujian Agriculture and Forestry University for helping in electron microscopy, and Zhu Li of NERCS of Fujian Agriculture and Forestry University for helping in transient expression.
This work was supported by grant from the National Natural Science Foundation of China (31571574)