Sphingolipid-Induced Programmed Cell Death is a Salicylic Acid and EDS1-Dependent Phenotype in Arabidopsis Fatty Acid Hydroxylase (Fah1, Fah2) and Ceramide Synthase (Loh2) Triple Mutants

Abstract Ceramides (Cers) and long-chain bases (LCBs) are plant sphingolipids involved in the induction of plant programmed cell death (PCD). The fatty acid hydroxylase mutant fah1 fah2 exhibits high Cer levels and moderately elevated LCB levels. Salicylic acid glucoside level is increased in this mutant, but no cell death can be detected by trypan blue staining. To determine the effect of Cers with different chain lengths, fah1 fah2 was crossed with ceramide synthase mutants longevity assurance gene one homologue1-3 (loh1, loh2 and loh3). Surprisingly, only triple mutants with loh2 show cell death detected by trypan blue staining under the selected conditions. Sphingolipid profiling revealed that the greatest differences between the triple mutant plants are in the LCB and LCB-phosphate (LCB-P) fraction. fah1 fah2 loh2 plants accumulate LCB d18:0, LCB t18:0 and LCB-P d18:0. Crossing fah1 fah2 loh2 with the salicylic acid (SA) synthesis mutant sid2-2 and with the SA signaling mutants enhanced disease susceptibility 1-2 (eds1-2) and phytoalexin deficient 4-1 (pad4-1) revealed that lesions are SA- and EDS1-dependent. These quadruple mutants also confirm that there may be a feedback loop between SA and sphingolipid metabolism as they accumulated less Cers and LCBs. In conclusion, PCD in fah1 fah2 loh2 is a SA- and EDS1-dependent phenotype, which is likely due to accumulation of LCBs.


Introduction
Sphingolipids are important molecules with diverse functions in eukaryotic cells. In addition to their structural role as membrane components, in plants they are critical for development and responses to biotic and abiotic stresses (Berkey et al. 2012, Saucedo-Garcia et al. 2015, Ali et al. 2018, Huby et al. 2020. Plant sphingolipids can be divided into four subgroups: free long-chain bases (LCBs), ceramides (Cers), glucosylceramides (GlucCers) and glycosyl inositol phosphoryl ceramides (GIPCs; Pata et al. 2010). Whereas GlucCers and GIPCs are important structural membrane components, Cers and LCBs are mainly described as biosynthetic intermediates and signaling molecules.
LCBs consist of a C18 core with an amino group at C2. Hydroxylation at C1 and C3 generates sphinganine (d18:0), the simplest form of LCB. It can be directly channeled into Cer synthesis or modified by additional hydroxylation at C4 (t18:0) or desaturation (d18:1; Luttgeharm et al. 2016b). Ceramide synthase (CerS) catalyzes the step from LCB to Cer by N-acylation of the LCB (Michaelson et al. 2016, Luttgeharm et al. 2016b). The fatty acid moiety can vary in chain length between C16 and C26. In Arabidopsis, there are three different CerS enzymes: Longevity Assurance Gene One (LAG1) Homologue1-3 (LOH1-3; Markham et al. 2011, Ternes et al. 2011. LOH1 and LOH3 preferentially catalyze the reaction with very long-chain fatty acids (VLCFAs) and t18:0 LCBs as substrates , Ternes et al. 2011, Luttgeharm et al. 2016a). LOH2 catalyzes the reaction preferentially with long-chain fatty acids (LCFAs) of C16 chain length and d18:0 LCB , Ternes et al. 2011, Luttgeharm et al. 2016a. Consequently, loh1 mutants accumulate Cers and GlucCers with C16 fatty acid and free trihydroxy sphingoid bases and show lesions late in development, whereas loh2 mutants show reduced LCFAcontaining Cer levels but exhibit no apparent growth phenotype , Ternes et al. 2011. These data suggest that spontaneous cell death in loh1 mutants is triggered by the accumulation of either free trihydroxy LCBs or C16 Cer species (Ternes et al. 2011).
The fatty acid moiety of a Cer molecule can be modified by hydroxylation in the alpha position. In Arabidopsis, two fatty acid hydroxylases (FAHs) are described (FAH1 and FAH2; Nagano et al. 2009, König et al. 2012. FAH1 prefers Cers with VLCFAs and FAH2 with C16 fatty acids (Nagano et al. 2012). Hydroxylated Cers (hCers) are found in GlucCers and GIPCs in high amounts (Pata et al. 2010) and might be suppressors of cell death (Townley et al. 2005). In fah1 fah2 double-mutant plants, which represent a knockdown in the FAH1 gene and a complete knock out in the FAH2 gene, however, high amounts of non-hCers in relation to hCers were found but no cell death was observed (König et al. 2012). Additionally, free salicylic acid (SA) and its glucoside salicylic acid glucoside (SAG) accumulate in those plants.
Apart from loh and fah mutants, deregulation of sphingolipid synthesis in other steps of the pathway often leads to cell death with involvement of SA (Mortimer et al. 2013, Bi et al. 2014, Li et al. 2016, Yanagawa et al. 2017, Zienkiewicz et al. 2020. Different Arabidopsis mutants with defects in sphingolipid metabolic pathways spontaneously accumulate SA and undergo programmed cell death (PCD) (Greenberg et al. 2000, Brodersen et al. 2002, Wang et al. 2008, Mortimer et al. 2013, Bi et al. 2014, Zienkiewicz et al. 2020. So far, the regulation of SA accumulation by sphingolipids on the molecular level is mostly unknown, but there are many hints that Cers and/or LCBs are inducers of PCD signaling and SA induction.
Recently it was shown that ENHANCED DISEASE SUSCEPTI-BILITY 1 (EDS1) and PHYTOALEXIN DEFICIENT 4 (PAD4), two key players in regulation of SA synthesis in response to biological and abiotic stresses (Wiermer et al. 2005, Rietz et al. 2011, Cui et al. 2017, are required for Cer-induced cell death in ceramide kinase accelerated cell death 5 (acd5) mutants and LOH2 overexpression lines (Zeng et al. 2021). EDS1 and PAD4 are necessary for promotion of ISOCHORISMATE SYN-THASE 1 (ICS1) gene expression and SA induction in Arabidopsis immune responses (Cui et al. 2017). EDS1 is the key signaling protein in the interaction with Toll/interleukin 1 receptornucleotide-binding/leucine-rich repeat (TNL) receptors, during effector triggered immunity (ETI) for defense responses and cell death (Wiermer et al. 2005). Together with PAD4 or SENESCENCE-ASSOCIATED GENE101 (SAG101), it acts as a heterodimer to transcriptionally induce defense responses and ICS1-dependent SA synthesis (Rietz et al. 2011, Cui et al. 2017.
In this study, we crossed fah1 fah2 mutants with loh mutants to induce cell death and analyze which components of the sphingolipid pool might be responsible for PCD. We further looked into the involvement of SA in sphingolipid-induced cell death by crossing sphingolipid-deficient mutants with SA biosynthesis or signaling mutants.

Results
Crossing loh2 into fah1 fah2 mutants leads to lesions in distinct leaf areas The fah1 fah2 double mutant has elevated Cer levels and slightly elevated LCB levels, but no severe lesion phenotype (König et al. 2012). To further analyze if Cers with distinct chain lengths may lead to lesions in these mutants, as previous work suggested (Ternes et al. 2011), the double mutant was crossed with the three CerS mutants loh1, loh2 and loh3.
Crossing fah1 fah2 with loh1 and loh3 produced triple mutants with no additional severe phenotypes. However, crossing loh2 into fah1 fah2 produced triple mutants with cell death-derived lesions at distinct areas at the leaves ( Supplementary Fig. S1B), which was verified by trypan blue staining (Fig. 1). Lesion formation in these mutants starts between 14 and 21 d after sowing (Supplementary Fig. S2). Additionally, these plants are significantly smaller than the double-mutant plants. Whereas 35-day-old fah1 fah2 plants are >50% reduced in rosette area compared to Col-0 mutant plants, fah1 fah2 loh2 plants show <30% of the rosette leaf area of wildtype plants (Supplementary Fig. S1A). fah1 fah2 loh1 and fah1 fah2 loh3 mutants do not show visible lesions or changes in leaf area compared to fah1 fah2 grown for up to 35 d under long day conditions. fah1 fah2 loh2 triple mutants accumulate d18:0 and d18:0-P To reveal the reason for lesion development in fah1 fah2 loh2 mutant plants, sphingolipid profiles in all mutant lines were compared. Before analyzing large data sets, Cer and SA levels at early (14 days old) and late time points (35 days old) were compared to establish the best plant age for measurements of these metabolites (Supplementary Fig. S2). SA levels between Col-0, fah1 fah2, loh2 and fah1 fah2 loh2 are similar at both time points, but highest at 35 d. However, changes in Cer and hCer profiles are more distinct in 35-day-old plants. Thus, plants at this age were used for further analysis.
Analysis of all mutant lines revealed that the most obvious changes between the triple mutants are in their LCB profiles. As shown in earlier studies, changes in LCB levels in the fah1 fah2 double mutant are only minor (1.5-2-fold increase) compared to Col-0. However, fah1 fah2 loh2 triple mutants accumulate d18:0 (20-fold compared to Col-0) and d18:0-P (not detectable in Col-0) as well as t18:0 (7-fold compared to Col-0; Fig. 2A). In contrast, in fah1 fah2 loh1 triple mutants, t18:0 (7-fold compared to Col-0) and t18:1 (4.7-fold) are the dominant enriched species, but the amount is comparable to the loh1 single mutant. fah1 fah2 loh3 triple mutants do not show significant changes compared to fah1 fah2 mutant LCB levels. Although elevated levels of t18:0 and t18:1 LCBs could be detected in the different mutants, no phosphorylated versions of these species could be detected. In addition, d18:1 levels were not detected as well.
In addition to LCBs, Cers are suspected to be involved in cell death induction in leaves. Changes in Cer profiles are especially visible in the fah1 fah2 loh1 triple mutant (Fig.  2B). Compared to the fah1 fah2 mutant, LCFA (C16-C18)containing Cers increased up to three times, whereas in the fah1 fah2 loh2 triple mutant, these species are strongly reduced. VLCFA-containing Cers are comparable in the double and triple mutants.

SAG levels are highly elevated in mutants with fah1 fah2 background
As SA is described to be involved in induction of sphingolipidderived cell death, SA and SAG were measured in rosette leaves. Among all mutant lines tested, the highest SA levels are detectable in the fah1 fah2 loh2 mutant (5.3 nmol/g fresh weight [FW]) followed by the loh1 single mutant with 3.8 nmol/g FW (Fig. 3A). Compared to 1.7 nmol/g FW in Col-0, the induction is minor. However, SAG strongly accumulates in the fah1 fah2 double mutant (105 nmol/g FW) and in all three triple mutant lines (fah1 fah2 loh1: 124 nmol/g FW, fah1 fah2 loh2: 188 nmol/g FW, fah1 fah2 loh3: 205 nmol/g FW) compared to Col-0 (17 nmol/g FW; Fig. 3B).

Reduction of SA levels by disruptions in SA synthesis or signaling represses the lesion phenotype of the fah1 fah2 loh2 mutant
To connect SA accumulation with the lesion phenotypes, mutants were crossed with the SA biosynthesis mutant sid2-2 (defective in ISOCHORISMATE SYNTHASE 1) and additionally with ENHANCED DISEASE SUSCEPTIBILITY1-2 (eds1-2) and PHY-TOALEXIN DEFICIENT4 (pad4-1). eds1-2 and pad4-1 are signaling mutants defective in the induction of SA synthesis by ETI (Cui et al. 2017).
To confirm the reduction of SA and SAG in the mutants, these phytohormones were measured in all mutant lines. SA levels in the fah1 fah2 loh2 plants crossed with eds1-2 and pad4-1 are lower than wild-type levels. SAG is still induced in those plants, but at a lower level than in fah1 fah2 loh2 plants ( fah1 fah2 loh2 pad4-1: 43% of fah1 fah2 loh2 levels, fah1 fah2 loh2 eds1-2: 33% of fah1 fah2 loh2 levels; Fig. 4C). fah1 fah2 loh2 with C16-C18 FA moiety; hLCFA, Cers with C16-C18 α-hydroxylated FA moiety; VLCFA, Cers with C20-C28 FA moiety; hVLCFA, Cers with C20-C28 α-hydroxylated FA moiety. Relative values were calculated according to 100% of wild-type peak area of the total LCB or Cer signals. Values represent the mean ± SD of four biological replicates (n = 4). The experiment was repeated once with similar tendencies. Statistical analysis was performed with log-transformed data by one-way analysis of variance (ANOVA) with Tukey's post hoc test (P < 0.05). Different letters indicate significant differences with P < 0.05. plants crossed with sid2-2 plants show no SA or SAG accumulation. In addition to the fah1 fah2 loh2 triple mutant, the fah1 fah2 double mutant was also crossed with sid2-2 to analyze which phenotypic changes are due to SA accumulation in those plants. Also in this mutant line, no SA or SAG is detectable (Supplementary Fig. S4).
Sphingolipid analyses of the mutants show that defects in SA synthesis and/or signaling also have an influence on the sphingolipidome of the plants. All SA mutant crosses showed a reduction of LCB accumulation relative to the fah1 fah2 loh2 triple mutant. d18:0 and d18:0-P are especially strongly reduced in the quadruple mutants (Fig. 5A). Additionally, LCFA-containing Cers are depleted in the quadruple mutants, and reductions of VLCFA-containing Cer are only minor (Fig. 5B). In contrast to this, fah1 fah2 mutants crossed with sid2-2 show a strong reduction in VLCFA-containing Cers and only a slight reduction in LCB levels compared to fah1 fah2 ( Supplementary  Fig. S4). In contrast to LCBs and Cers, levels of GlucCer and GIPC are not changed due to disruption of SA synthesis (Supplementary Fig. S3).

Discussion
Accumulation of LCBs and/or Cers is suspected to be responsible for sphingolipid-induced cell death in several mutants with disruptions in the sphingolipid pathway (Berkey et al. 2012, Bi et al. 2014, Yanagawa et al. 2017, Zienkiewicz et al. 2020). However, fah1 fah2 mutant plants exhibit strong accumulations of non-hCers and hydroxylated very long-chain fatty acid (hVLCFA)-containing Cers, moderate increases in LCB levels and enhanced SA levels, but no lesion phenotype (König et al. 2012). Therefore, the intention behind crossing fah1 fah2 plants with loh plants was to determine if specific changes in the sphingolipidome of those mutants could be correlated with the induction of PCD. fah1 fah2 mutants were crossed with loh1, loh2 and loh3 mutants to shift the Cer accumulation to VLCFA Cer or C16 Cer. C16 Cers are discussed as PCD-inducing sphingolipids in loh1 plants under certain conditions (Ternes et al. 2011) and in LOH2-overexpressing plants (Luttgeharm et al. 2015). Unexpectedly, fah1 fah2 crosses with loh2, but not those with loh1, produced triple mutants with PCD ( Figs.  1 and 2). In fah1 fah2 loh2 plants, LCFA-containing Cers are strongly reduced compared to fah1 fah2, but this is also more pronounced in the loh2 single mutant. Therefore, it seems that the other main difference in the sphingolipid profile of these plants, the elevated levels of d18:0 LCB and d18:0-P LCB, is most likely the underlying reason for the lesion phenotype. LCBs are also suspected to be involved in PCD induction upon Fumonisin B 1 treatment, a fungal toxin inhibiting LOH1 (Luttgeharm et al. 2016a). LOH1 inhibition induces a strong accumulation of d18:0 and t18:0 as well as LCFA-containing Cer species. This is supported by the analysis of mutants defective in different steps of the sphingolipid pathway or pathogen-infected plants, which collectively suggest that accumulation of LCBs induces cell death in plants (Shi et al. 2007, Peer et al. 2010, Saucedo-Garcia et al. 2015, Luttgeharm et al. 2016a, Yanagawa et al. 2017. The different roles of d18:0 and t18:0 LCBs are still not clear. In case of fah1 fah2 loh2, both species are induced, but compared with fah1 fah2 loh1 accumulation of d18:0 LCB is specific to the former. d18:0 LCB may therefore be responsible for the differences in lesion development between the mutants. This scenario is supported by reports from other sphingolipid mutants like the serine palmitoyltransferase mutant (lcb2a; Saucedo-García et al. 2011) and the sphingoid base hydroxylase double mutant (sbh1-1 sbh2-1; Chen et al. 2008), which show increased d18:0 contents in combination with PCD. Nevertheless, it could also be that d18:0 LCB is acting together with t18:0 LCB and that the combined accumulation of both LCBs passes a critical level  4). The experiment was repeated once with similar tendencies. Statistical analysis was performed with log-transformed data by one-way analysis of variance with Tukey's post hoc test (P < 0.05). Different letters indicate significant differences with P < 0.05. that leads to induction of PCD. LCB-phosphates (LCB-Ps) are thought to be the inactivated form of LCB that do not induce PCD and do not influence LCB-induced cell death (Yanagawa et al. 2017, Glenz et al. 2019. The high accumulation of d18:0-P in fah1 fah2 loh2 plants might be due to the partial removal of d18:0 by phosphorylation by LCB kinase to avoid the harmful effect of the molecule (Imai and Nishiura 2005).
LOH2 seems not to be essential under normal growth conditions, as loh2 mutants do not show a phenotype compared to wild type , Ternes et al. 2011, Luttgeharm et al. 2015. However, overexpression of LOH2 leads to dwarfed plants with high amounts of LCFA-containing Cers and to higher resistance to FB 1 treatment (Luttgeharm et al. 2015). Therefore, the authors suspected that LOH2 might be a safety valve for plants to sequester excess LCBs. In the case of the fah1 fah2 double mutant, LOH2 might be essential to channel excess LCBs into the Cer pool to avoid toxic LCB levels in the plant. In triple mutants with loh1 or loh3, activity of LOH2 and the other remaining LOH are high enough not to exceed the critical level of LCBs. In the fah1 fah2 loh2 mutant, this safety valve is not functional anymore as Cer synthesis of C16 fatty acids with d18:0 LCB is blocked and d18:0 LCB accumulate. Excess d18:0 can then be partially removed by phosphorylation resulting in accumulation of d18:0-P. It might be interesting to determine if blocking phosphorylation, e.g. in LCB kinase mutants, would lead to a more severe phenotype.
All mutants with fah1 fah2 background showed high levels of total SA. To test the role of SA in the development of lesions and reduction in rosette size, mutants were crossed with the SA synthesis mutant sid2-2 (isochorismate synthase 1 deficient). Additionally, as it is known that sphingolipid metabolism in plants is connected to plant defense (Huby et al. 2020), plants were crossed with eds1-2 and pad4-1. The same SA-mutant set was recently used by Zeng and colleagues to analyze the effect of SA and SA signaling on cell death in acd5 mutants and LOH2 overexpression mutants (Zeng et al. 2021). They could show that EDS1 and PAD4 contribute to Cer-mediated accumulation of SA and cell death in both transgenic lines. In our study, the lesion phenotype of fah1 fah2 loh2 was eliminated in fah1 fah2 loh2 sid2-2 and fah1 fah2 loh2 eds1-2 mutants but still present in fah1 fah2 loh2 pad4-1 although at a less severe level (Fig. 3), showing that the phenotype is dependent on SA and EDS1 but less on PAD4. Lapin and colleagues showed that EDS1 heterodimerizes with SAG101 and N REQUIRED GENE1 (NRG1) to promote TNL-dependent cell death in Arabidopsis, and that this strong cell death activity does not occur when PAD4 substitutes for SAG101 (Lapin et al. 2019). EDS1 and PAD4 instead mediate transcriptional promotion of defense reactions important for limiting bacterial growth. The distinction between both pathways seems not to be very strict and a small proportion of cell death can account from EDS1-PAD4 signaling. This could explain why no cell death is visible in fah1 fah2 loh2 eds1-2 mutants but is still found in fah1 fah2 loh2 pad4-1 mutants.
Interestingly, not only the phenotype of the plants changed due to SA mutations but also the sphingolipid profile (Fig. 5).
In fah1 fah2 loh2 eds1-2, fah1 fah2 loh2 pad4-1 and especially in fah1 fah2 loh2 sid2-2 plants, the accumulation of Cers and LCBs is significantly reduced, showing that SA level has an influence on the sphingolipidome of Arabidopsis. Indeed, Shi et al. (2015) showed that SA influences sphingolipid fluxes in Arabidopsis seedlings by in silico flux balance analysis and suggested SA could also act upstream of LCB or Cer signals (Shi et al. 2015). Interestingly, Zeng and colleagues could show that primarily EDS1 and only to a much lesser extent PAD4 are involved in regulation of Cer and LCB metabolism (Zeng et al. 2021). Cer and LCB accumulation in acd5 mutants and in LOH2 overexpression lines was strongly reduced only in crossings with eds1-2. Our data therefore only partly confirm these results that SA levels in the plant influence LCB and Cer levels and that there might be a feedback loop between SA signaling and sphingolipid biosynthetic pathways.
Together, our data strongly suggest that lesion formation as a result of changes in sphingolipid metabolism in fah1 fah2 loh2 plants is SA-dependent and transduced via EDS1 signaling. SA double and fah1 fah2 loh2 triple mutants with SA synthesis (sid2-2) and signaling mutants (eds1-2, pad4-1). Rosette leaves of 35-day-old plants grown under long day conditions were extracted and analyzed. LCFA, Cers with C16-C18 FA moiety; hLCFA, Cers with C16-C18 α-hydroxylated FA moiety; VLCFA, Cers with C20-C28 FA moiety; hVL-CFA, Cers with C20-C28 α hydroxylated FA moiety. Relative values were calculated according to 100% of wild-type peak area of the total LCB or Cer signal. Values represent the mean ± SD of four biological replicates (n = 4). The experiment was repeated once with similar tendencies. Statistical analysis was performed with log-transformed data by one-way analysis of variance with Tukey's post hoc test (P < 0.05). Different letters indicate significant differences with P < 0.05. signaling itself influences the accumulation of Cers and LCBs in Arabidopsis, suggesting a positive feedback loop between both pathways.

Plant material
The Arabidopsis thaliana plants used in this study were of the Columbia (Col-0) ecotype background. The seeds of the T-DNA insertion mutants-fah1 (SALK_140660), fah2 (SAIL_862_H01), loh1 (SALK_069253), loh2 (SALK_018608) and loh3 (SALK_150849)-were obtained from the Nottingham Arabidopsis Stock Centre. All mutants are knockout mutants except fah1 that is a knock-down mutant (Ternes et al. 2011, König et al. 2012). The sid2-2, eds1-2 and pad4-1 mutants were kindly provided by Prof. Christiane Gatz (University of Goettingen, Germany). Double, triple and quadruple mutants were generated by crossing the above mutants. The homozygous insertion of T-DNA was verified by polymerase chain reaction analysis performed on genomic DNA. Primer sequences used for genotyping are provided in Supplementary  Table S1. For soil-grown plants, Arabidopsis seeds were sown on soil and then stratified at 4 • C for 2 d. Plants were grown at 22 • C under long day conditions (16 h light : 8 h dark), with a light intensity of 120-150 µmol m −2 s −1 .

Trypan blue staining
To visualize the cell death phenotype, trypan blue staining was performed as described previously (Fernández-Bautista et al. 2016) and analyzed using a stereomicroscope Olympus SZX12 (Olympus, Hamburg, Germany).

Supplementary Data
Supplementary data are available at PCP online.