Complex Changes in Membrane Lipids Associated with the Modification of Autophagy in Arabidopsis

Autophagy is a conserved mechanism among eukaryotes that degrades and recycles cytoplasmic components. Autophagy is known to influence the plant metabolome, including lipid content; however, its impact on the plant lipidome is not fully understood, and most studies have analyzed a single or few mutants defective in autophagy. To gain more insight into the effect of autophagy on lipid concentrations and composition, we quantitatively profiled glycerolipids from multiple Arabidopsis thaliana mutants altered in autophagy and compared them with wild-type seedlings under nitrogen replete (+N; normal growth) and nitrogen starvation (−N; autophagy inducing) conditions. Mutants include those in genes of the core autophagy pathway, together with other genes that have been reported to affect autophagy. Using Matrix-Assisted Laser Desorption/Ionization—Mass Spectrometry (MALDI-MS), we imaged the cellular distribution of specific lipids in situ and demonstrated that autophagy and nitrogen treatment did not affect their spatial distribution within Arabidopsis seedling leaves. We observed changes, both increases and decreases, in the relative amounts of different lipid species in the mutants compared to WT both in +N and −N conditions, although more changes were seen in −N conditions. The relative amounts of polyunsaturated and very long chain lipids were significantly reduced in autophagy-disrupted mutants compared to WT plants. Collectively, our results provide additional evidence that autophagy affects plant lipid content and that autophagy likely affects lipid properties such as chain length and unsaturation.


Introduction
Autophagy is a pathway for the vacuolar degradation and recycling of cellular organelles and macromolecules. It is activated as a developmental process, as a housekeeping mechanism, or in response to stress. It is conserved among eukaryotic organisms, and macroautophagy, referred to here as autophagy, is the best-studied form of autophagy [1][2][3]. In plants, a basal level of autophagy during growth and development ensures homeostasis by degrading obsolete cytoplasmic materials and recycling the breakdown products. Autophagy is upregulated under environmental stresses to aid plant survival through the catabolism and recycling of damaged organelles and the elimination of toxic protein aggregates. It is also upregulated by developmental cellular reprogramming processes such as senescence [4][5][6], xylem development [7], and by temporary cellular reprogramming increased basal autophagy (Table 1). We observed an impact of autophagy level on all lipid classes analyzed, with some lipids showing a decrease while others showed an increase in atg mutants compared with the WT control. We also observed changes in unsaturation and acyl chain length in autophagy mutants compared to WT, thus strengthening the hypothesis that autophagy impacts lipid metabolism. Additionally, we imaged lipids in situ using Matrix-Assisted Laser Desorption/Ionization-Mass Spectrometry (MALDI-MS), which revealed that there are no detectable changes in the spatial distribution of lipids across leaves of all the genotypes evaluated. Autophagy blocked under all conditions tested [26,32] atg7-2 Required for formation of ATG8-PE conjugate via a ubiquitin like conjugation system Autophagy blocked under all conditions tested [33][34][35] atg9-4 Lipid scramblase required for autophagosome formation Autophagy blocked under all conditions tested (phenotype is less severe than atg5 and atg7) [12] kin10 Catalytic subunit of the protein kinase SnRK1, activated by low energy Autophagy activation by stress blocked [11] raptor1b Subunit of the TOR complex, inactivated by low nutrients Constitutive autophagy [36] rns2-2 Vacuolar ribonuclease Constitutive autophagy [12] rns2-2; atg5-1

Nitrogen Starvation Induces Autophagy and Causes Changes in the Lipidome
To explore the impact of autophagy on lipids in Arabidopsis, we carried out a lipidomic analysis on seedlings from nine different genotypes, with growth occurring in nitrogen replete (+N; control) and in nitrogen deplete (−N; autophagy-inducing) conditions [33] ( Table 1). This experimental design allowed us to distinguish between true effects of changes in autophagy from changes associated with N-starvation, as well as any potential additional functions of ATG proteins that are unrelated to autophagy. The genotypes studied included wild-type (accession Columbia-0) and the autophagy machinery mutants atg5-1, atg7-2, and atg9-4, hereafter labelled as atg5, atg7, and atg9; mutants in non-ATG genes that affect autophagy activity included kin10 (unable to induce autophagy), raptor1b, and rns2-2, hereafter labelled as rns2, (constitutive autophagy); and double mutants rns2; atg5 and rns2; atg9.
A principal component analysis (PCA) based on the lipid profile data showed that most of the variance was included in the first two principal components, explaining about 70% of the overall data variance (56% for PC1 explaining the N-treatment effect and 11% for PC2 explaining the genotype effect) ( Figure 1A). This shows that both N-treatment and genotype had an effect on the lipid content of the seedlings, with the N-treatment effect explaining most of the variation observed. In +N conditions, all genotypes clustered closely together, except for raptor1b. In −N conditions, which activates autophagy, separation was seen between genotypes, with autophagy-blocked mutants (atg7, atg5, atg9, kin10, rns2; atg5, and rns2; atg9) clustering together, and mutants with constitutive autophagy (rns2 and raptor1b) clustering with WT ( Figure 1A). The clustering pattern is therefore consistent with the extent of autophagy under different conditions and reinforces the observation that autophagy activity affects the lipid profile of seedlings. Figure 1. Nitrogen treatment and genotype both influence the plant lipidome. (A) Principal component analysis (PCA) for autophagy-blocked mutants (atg5, atg7, atg9, rns2; atg5, rns2; atg9, and kin10), constitutive autophagy mutants (rns2 and raptor1b), and WT. The values were determined from log10-transformed lipidomics data (n = 3 biologically independent samples; for WT and atg5, n = 2 biologically independent samples). The amounts of variation explained by the first two components are indicated on the axes. The dashed circles indicate biological replicates for each genotype/condition. (B) Total membrane lipid amount in Arabidopsis seedlings. Total lipid amount represents the sum of the amount of all lipid species measured from each of the genotypes under + and -N. Values (lipid/dry weight (nmol/mg)) are means ± standard deviation (SD), n = 3 biologically independent samples and for WT and atg5, n = 2 biologically independent samples. Different letters denote statistically significant differences between means (q < 0.1). (C) Nitrogen starvation causes a reduction in most lipid species. Log 2 Fold Change (FC) heat map of lipid species for each genotype under −N versus +N comparison. Blue indicates a reduction and red indicates an increase in lipids. Log 2 FC values were calculated from data presented in Supplementary Materials Table S2. The total amount of the measured lipids was significantly reduced in all genotypes in the −N condition. In the +N condition, the atg5, atg7, kin10, and raptor1b mutants showed a significant reduction in total lipid amount as compared to the WT control. Given that raptor1b shows constitutive autophagy, this observation reveals a complex relationship between autophagy activity and lipid content, and could reflect the enhanced catabolism of lipids by autophagy or disruption of active lipid assembly. Under −N conditions, atg5, atg7, atg9, rns2; atg5, and rns2; atg9 mutants showed a significant reduction in total lipid amounts as compared to WT plants, consistent with all of these mutants being deficient in autophagy, while kin10, rns2, and raptor1b mutants showed no significant difference compared to WT ( Figure 1B,C). Consistent with a previous study [37], the proportion of MGDG was reduced in −N conditions, with a concomitant increase in other lipids (Supplementary Materials Figure S1).
A heat map representation of the log 2 fold change in the amount of individual lipid species in response to N starvation shows that most lipid species are significantly reduced across all genotypes ( Figure 1C). The raptor1b and kin10 mutants have the fewest number of lipid species significantly changed by N starvation ( Figure 1C). Unlike other lipids, which generally decrease in −N, lysophospholipids increased in accumulation in a number of genotypes. Because lysophospholipids are primarily products of phospholipases, this finding is consistent with previous reports that showed an increased accumulation of these lipid breakdown products during carbon starvation [31] and freezing stress [38] in Arabidopsis, and under N starvation in maize [28]. The amounts of most PA species were significantly increased upon nitrogen starvation ( Figure 1C). This is consistent with previous studies in Arabidopsis, which showed that in response to N starvation, phospholipids are hydrolyzed by phospholipase D ε to generate PA [39]. Together, these results show that N treatment affects both the quantity and composition of lipids.

Matrix-Assisted Laser Desorption/Ionization-Mass Spectrometry (MALDI-MS) Imaging of Lipids in Arabidopsis under + and −N Conditions
Mass spectrometry imaging (MSI) has become a valuable tool for analyzing the spatial distributions of a wide range of metabolites in situ [40][41][42][43][44][45][46][47], including in plants [44,[46][47][48]. We therefore used MALDI-MS to image lipids in situ and to determine the effect of autophagy and −N or +N conditions on their spatial localization across a leaf in Arabidopsis.
We used a leaf fracture method, which has been demonstrated to maintain the original leaf anatomy [49], and imaged the distribution of lipids within the leaves. A leaf was "split" into two halves parallel to the plane of the leaf, and the two exposed leaf interior surfaces were imaged visually and by MALDI-MS. The resulting two images were then aligned and compared to identify the distribution of molecular features with respect to visually identified features. Lipid distributions were imaged in fractured leaves from several genotypes (WT, atg5, atg9, atg7, rns2, rns2; atg5, and rns2; atg9) that had been grown either in +N or −N conditions. Based on the ESI-MS lipidomics data, we selected specific lipid molecules to image, which showed significant changes in abundance in the −N vs. +N comparisons (i.e., PE34:2, PE34:3, PE36:4, PG32:1, PG34:2, PG34:3, PG34:2, PG34:4, PI34:2, and PI34:3); additionally, we imaged PA36:5, which showed no significant change in abundance as measured by ESI-MS. These lipid species were found to be uniformly localized throughout the exposed interior leaf surfaces examined ( Figure 2A). Moreover, this uniform distribution was unaffected by any of the mutations that we studied, nor was it affected by the N-treatment that was applied to these plants, both of which affect autophagy. Note that due to analytical and biological variation between samples, direct comparisons should be made only in reference to each corresponding wild-type sample within a single experimental replicate and cannot be made between experiments. These findings, therefore, indicate that the lipid metabolism response to changes in autophagic status are triggered simultaneously among all the cells visible throughout the leaf (Figure 2A), which may indicate that autophagy is active in all of the cells observed in the leaves. While different signal intensities are observed between replicates, the spatial distribution of the signal within an individual leaf is reproducible, as illustrated in Figure 2B, which shows replicate images of independent biological replicates of each half of a leaf for PC34:2.

Autophagy Alters the Amount of Specific Lipid Species
Given that N starvation is known to induce autophagy [11,33,36], we examined whether the observed changes in lipid amounts in response to N conditions are dependent on autophagy. Lipids with different polar head groups have been shown to be important for the process of autophagy [50,51] and also as cargo degraded/remobilized by autophagy [16]. Based on this, we profiled the impact of autophagy on levels of polar membrane lipids. We compared changes in lipid amounts between WT plants and each of the mutants under control (+N) and autophagy-inducing conditions (−N).
An analysis of lipid amounts revealed differences between WT and autophagy-related mutants, both under +N and −N conditions. As an example, in −N (autophagy inducing) conditions, total PS was significantly reduced in all autophagy-blocked mutants except for kin10, while the constitutive autophagy mutants raptor1b and rns2 showed no significant change when compared with the WT control. Under +N conditions, only the raptor1b mutant showed a significant reduction in levels of PS as compared to WT (Figures 3 and 4). Total PE was decreased in atg5, atg7, atg9, and raptor1b mutants under −N, while under +N, it was significantly reduced in atg7, kin10, and raptor 1b. Overall, in conditions in which autophagy is active in WT plants, a reduction in the amount of PS and PE is seen in autophagy-deficient mutants. raptor1b and kin10 seem to behave differently to other mutants, possibly reflecting their roles in many cellular pathways in addition to autophagy, or as regulators rather than core components of the pathway.  We assessed the impact of changes in autophagy on individual lipid species within the major lipid groups. Under −N conditions, most PS molecular species were significantly reduced in autophagy-blocked mutants (atg5, atg7, and rns2; atg5), while no significant change was observed for most of the PS molecular species in kin10, rns2, and raptor1b ( Figure 4). Very few PS species showed significant changes in their amount compared to WT under +N conditions (Supplementary Materials Figure S2). Changes in PE species were less consistent, with some species increased and some decreased.
An increase in lysophosphatidylcholine (LysoPC) concentration was detected in all atg mutants compared to WT in +N. Under −N conditions, LysoPC was increased compared to WT in all genotypes tested ( Figure 4). This is consistent with previous studies that showed an increased accumulation of lysophospholipids in plants under different stresses [28,31,38]. Because lysolipids are considered as breakdown products of glycerol lipids, these findings suggest an increase in lipolytic activity in mutants compared to WT plants in response to N starvation. Other lipid species measured (PA and PG) showed fewer differences between genotypes in response to nitrogen conditions ( Figure 4).
Together, these results show that some lipids decrease in the atg autophagy-blocked mutants, while others increase in these mutants when plants are grown under −N conditions. One possibility is that some of the lipids are incorporated into the autophagosome membrane, and thus the activation of autophagy triggers an increase in the biosynthesis of such lipids to meet the demand for autophagosome formation. An increase in some lipids in the atg mutants compared to WT could be due to the degradation of these lipids in WT plants by autophagy, which is lost in in atg mutants.

Lipids with Longer Chain Unsaturated Fatty Acids Are Decreased in Autophagy Mutants in Response to N Starvation
In the autophagy-blocked mutants grown under −N conditions, we found a significant reduction in lipids with longer chain fatty acids and in lipids with higher levels of acyl chain unsaturation (i.e., 36:6, 38:4, 38:5, 38:6, 42:2, and 42:3 lipids). This alteration in lipid accumulation is not apparent in the autophagy-constitutive mutants ( Figure 5A). Except for 36:6, no difference in these lipids was seen in the mutants under N-replete conditions. Given that these changes are observed in the −N condition that induces autophagy, but disappear in the +N condition, they are most likely dependent on autophagy. Additionally, we looked at changes in the relative amount for individual lipid species from the groups identified in Figure 5A; we observed a significant decrease in representative individual lipid species in autophagy-blocked mutants compared to WT plants under −N conditions, with minimal changes in constitutive autophagy mutants ( Figure 5B). Most of the significant changes in the individual lipid species are diminished under +N conditions (Supplementary Materials Figure S3). Shown are the means ± SD (n = 3 biologically independent samples; for WT and atg5, n = 2 biologically independent samples). * denotes statistically significant differences between means (q < 0.1). (B) Heat map depicting log 2 fold changes in individual lipids that make up the lipid groups in mutants compared with WT plants under −N conditions, with shades of red (increased) or blue (decreased) colors depicting changes according to the scale bar. * denotes statistically significant differences (q < 0.1). Clustering for heat maps was performed using the pheatmap package in R (https://cran.r-project.org/web/packages/ pheatmap/pheatmap.pdf (accessed on 14 February 2022)).

Autophagy Induction Alters Lipid Unsaturation
Lipid groups and individual lipid species containing polyunsaturated fatty acids were decreased in autophagy mutants compared to WT seedlings ( Figure 6). We hypothesized that the induction of autophagy may lead to decreased lipid saturation. To test this hypothesis, we measured the total unsaturation index of lipid species, defined as the product of the amount of that species and the average number of double bonds per acyl chain [52,53]. Under N-replete conditions, there was no difference in the unsaturation index for any lipid in any of the genotypes. However, under N starvation, PE and PS both showed a decrease in the unsaturation index in autophagy mutants compared to the WT control. Other lipids, such as MGDG, DGDG, PI, PG, PC, and PA, showed no change in the unsaturation index across all genotypes under −N conditions (not shown).

Discussion
During autophagy, lipids can be both cargoes, and constituents, of autophagosome membranes. Several studies have recently reported on the alterations of lipid content in autophagy mutants under different stress conditions in plants. However, lipidomic changes in distinct autophagy mutants do not show a complete overlap [28,30,31], potentially because these mutants affect different steps in the autophagy pathway, allelic differences give rise to strong and weak mutant phenotypes [28], or ATG genes may function in other processes in addition to autophagy [54]. We reasoned that a lipidomic analysis of a wider range of mutants would help to pinpoint the specific effects of autophagy on lipid metabolism in Arabidopsis.
We used ESI-MS to quantitatively profile lipids and MALDI-MS to image lipids in situ from mutants affecting the core autophagy pathway (atg5, atg7, and atg9), mutants that do not affect the core autophagy machinery but that affect autophagy activity (kin10, raptor1b, and rns2) and double mutants (rns2; atg5 and rns2; atg9) to characterize changes in membrane lipid amounts in response to autophagy.

Autophagy Affects Lipid Concentrations under Both +N and −N Conditions
We observed that the raptor1b mutant clustered away from other genotypes in a PCA and has a significant reduction in total lipids compared to the WT control under +N conditions, but not under −N conditions. This suggests that the raptor1b lipidome is significantly altered, even under non-stress conditions. RAPTOR1B, a conserved TOR complex subunit, is a scaffold protein which recruits substrates for the TOR kinase. raptor1b mutants disrupt TOR signaling and exhibit molecular and physiological phenotypes of TOR inhibition [36,55].
TOR activation is reported to upregulate the expression of lipid synthesis genes and downregulate lipid degradation genes in Arabidopsis [56], as well as galactolipid synthesis genes in rice [57]. Thus, the reduction in total lipids in raptor1b could be due to reduced lipid synthesis. TOR also negatively regulates autophagy [58,59], and autophagy has been implicated in TAG synthesis and lipid droplet accumulation, likely from the breakdown of membrane phospholipids [15,16]. Since raptor1b has constitutive autophagy [36], lipid breakdown by autophagy may also contribute to the decrease in the total lipid amount in raptor1b compared to WT under +N conditions. However, rns2, which also has constitutive autophagy [12], behaved differently from raptor1b, showing no significant change compared to WT with regard to total lipid amounts, and also clustered closer to WT in the PCA under both +N and −N conditions. This suggests that changes in lipid synthesis may be a more significant factor affecting lipids in raptor1b rather than autophagic degradation.
The autophagy mutants atg5 and atg7 had significantly less total lipid than the WT control under +N conditions, while the atg9 mutant showed no significant change. The discrepancy in the atg9 result is probably due to the known weaker autophagy phenotype of atg9 [30,60]. It is not clear whether the observed significant reduction in total lipid amounts in atg5 and atg7 compared to WT is attributable to reduced lipid biosynthesis or increased degradation in these mutants. In contrast to our observation, a study by Fan et al., 2019 showed that a disruption of autophagy results in a decrease in membrane lipid turnover under non-stress conditions [16]. The differences in the systems used in our study (Arabidopsis seedlings) and Fan et al., 2019 (mature and senescing leaves) may explain this discrepancy. One possibility is that a disruption of autophagy results in the activation of other lipid catabolic pathways, therefore increasing lipid breakdown in autophagy mutants even under non-stress conditions. Consistent with this hypothesis, a transcriptomics analysis of maize plants undergoing N starvation revealed the upregulation of phospholipases in atg mutants when compared to WT plants [17].
kin10 showed a significant reduction in total lipids compared to WT under +N, and no significant change under −N conditions, an unexpected observation as this mutant has similar levels of autophagy as WT plants under +N conditions [11]. KIN10 is the catalytic subunit of Arabidopsis SnRK1 and regulates autophagy upstream of TOR. As KIN10 activates lipid catabolism genes and inhibits lipid biosynthesis genes [61,62], the reason for the observed changes in lipid content remains obscure and requires further investigation.

MALDI-MS Imaging of Lipids
Modern advances in mass spectrometry have made it possible to quantitatively profile acyl lipids, and have thus improved our understanding of lipid metabolism in plants [40,[63][64][65][66][67][68]. MS lipid profiling has mostly been performed through the extraction of pulverized tissue, yet it has recently been shown that lipids undergo dynamic spatial changes in plant tissues during development [69]. Therefore, we used MALDI-MS to image lipids in situ and assess the effect of autophagy and N-treatment on spatial lipid localization across the leaf in Arabidopsis. Our results show that neither N starvation, nor autophagy, impacted the spatial distribution of lipids in the seedlings, with changes in lipid profiles occurring evenly throughout the leaf, at least in the mesophyll cells that are imaged by this technique.

Autophagy Mutants Have Increases or Decreases in Different Lipid Species
The relative amount of some lipid species was increased, while others decreased in atg mutants compared to the WT control under −N conditions, either due to changes in lipid synthesis or degradation. Autophagy is marked by the formation of an autophagosome, a double membrane vesicle that carries cargo to the vacuole. A number of lipids are important for autophagosome formation and as constituents of autophagosome membranes [23,70]; for example, PE, which is conjugated to the key autophagy protein ATG8. We have observed changes in the amounts of some of these key lipid species in response to N depletion or in autophagy mutants. Studies in yeast and human cells have demonstrated the de novo synthesis of lipids for autophagosome formation [20,71]. Therefore, it is likely that lipids are synthesized in WT seedlings to meet the demand for autophagosome formation, but whether this synthesis is impaired in autophagy mutants is not clear. The observation that LPC, a breakdown product of PC, is increased in all autophagy-blocked mutants compared to WT under −N conditions is consistent with the proposal by McLoughlin et al., 2018 that autophagy might be essential for recycling membrane lipid turnover products [28].

Unsaturated and Long Chain FAs Are Reduced in Autophagy Mutants
Acyl chain length and unsaturation are fatty acid properties that have been reported to be impacted by autophagy [20][21][22]29,[71][72][73][74]. We, therefore, reasoned that the induction of autophagy may affect the saturation and carbon chain length of lipid fatty acyl chains in Arabidopsis. To this end, we looked at changes in the amount of lipids based on acyl chain length and the total number of double bonds. We observed a decrease in the amounts of lipids containing unsaturated FAs and with long acyl chains in autophagy-blocked mutants compared with the WT control. This observation is consistent with a previous study in rice that showed a decrease in unsaturated lipids extracted from anthers of Osatg7 compared to WT [75]. Lipidomic analysis of purified autophagosomes and autophagic membranes revealed an enrichment of unsaturated lipids in yeast and Drosophila [21,22], and defects in lipid desaturation resulted in autophagic defects [76]. This suggests that lipid desaturation may be important in autophagy.
VLCFAs are increased in plants under various abiotic constraints [77]. However, our observation that they are reduced in atg mutants compared to WT plants shows that LCFAs and VLCFAs might be important for autophagy. In contrast to our results, an increase in long acyl chains in atg5 compared with WT was previously reported [29]; that study included sphingolipids, potentially explaining this difference.
It is likely that changes in lipid species saturation and chain length could modify the physical properties of membranes, such as thickness, fluidity, and membrane curvature, which could be important for autophagosome formation. For example, the introduction of double bonds and reduced acyl chain length increase the membrane fluidity and flexibility, while increased fatty acyl chain length leads to thicker bilayer structures [78]. The appropriate balance of various phospholipid species might ensure the ideal physical properties for the membrane dynamics required when autophagy is induced.
In conclusion, by using multiple mutants, we show that autophagy has an impact on lipid content and affects lipid unsaturation and acyl chain length properties. A decrease in the amount of total lipids was seen under −N conditions for all genotypes, but this decrease was greater for mutants blocked in autophagy. Significant changes in individual lipid species were seen between mutants and WT plants, both in +N and −N conditions, but again, a greater number and more substantial changes were seen under −N conditions; this was particularly evident for PS and PE. Mutants blocked in autophagy also had significantly reduced relative amounts of polyunsaturated and very long chain lipids when compared with WT. Our data suggests an effect of autophagy on glycerol lipid degradation pathways, but also a potential role for autophagy in promoting the biosynthesis of these lipids.

Lipid Extraction
Lipid extraction was carried out as previously described [68]. Seedling samples were immediately dropped into 3 mL of isopropanol with 0.01% butylated hydroxytoluene (BHT) preheated to 75 • C in a 50 mL glass tube with a Teflon lined screwcap (Thermo Fisher Scientific, Inc., Waltham, MA, USA). The sample-containing tubes were maintained at 75 • C for 15 min to deactivate lipid-hydrolyzing enzymes such as phospholipase D. Next, 1.5 mL of chloroform and 0.6 mL of water were added, followed by shaking on an agitating shaker for 1 h. Extracted lipids (first extract) were transferred to a new tube using a Pasteur pipette, leaving the seedling debris at the bottom of the original tube. Following the first extraction, 4 mL of chloroform: methanol (2:1) with 0.01% BHT were added to the seedling debris, and the samples were shaken on an orbital shaker at room temperature for 30 min, with the solvent then being transferred to the first extract. This was repeated 3 times, and for the 4th time, the samples were shaken on an orbital shaker at room temperature overnight while the combined extract was kept at −80 • C. After overnight shaking, the solvent was transferred to the first extract, and, at this stage, all of the seedling debris appeared white. The solvent was evaporated from the extract in an N-EVAP 112 nitrogen evaporator (Organomation Associates, Inc., Berlin, MA, USA). The lipid extract was dissolved in 1 mL of chloroform, transferred to a 2 mL clear glass vial with a Teflon-lined screw cap (DWK Life Sciences L.L.C., Millville, NJ, USA), and stored at −80 • C. Prior to shipping, the solvent (chloroform) was again evaporated from the lipid extract in the N-EVAP 112 nitrogen evaporator. The vials containing the lipid extracts were shipped overnight to the Kansas Lipidomics Research Center (KLRC) on dry ice. At KLRC, the lipid extract was again dissolved in 1 mL of chloroform and was used for lipid profiling (see below). The extracted seedling debris was dried in an oven at 105 • C overnight, cooled, and the dry weight was determined using a balance (Mettler Toledo ME104E, Mettler Toledo).

Lipid Profiling by MS
Lipids were profiled using automated electrospray ionization-tandem mass spectrometry to identify digalactosyldiacylglycerol (DGDG), monogalactosyldiacylglycerol (MGDG), phosphatidylglycerol (PG), phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylinositol (PI), phosphatidylserine (PS), phosphatidic acid (PA), lysophosphatidylethanolamine (LPE), and lysophosphatidylcholine (LPC) head group classes. The lipids in each head group class were quantified in comparison with internal standards of that class. To ensure consistency for data for each molecular species throughout the mass spectral data acquisition periods, quality control (QC) samples were prepared by pooling an aliquot from each lipid sample and were analyzed recurrently among the experimental samples [68,84]. The intensity of each lipid species in the experimental samples was normalized using the QC analyte intensities [68]. To maintain data quality, lipid analytes for which either the amount (normalized mass spectral signal) per milligram of leaf dry weight was less than the limit of detection, or which had a coefficient of variation (standard deviation divided by mean of the amount of the analyte in the QC samples) greater than 0.3 were removed from the data set. The lipid values were reported as normalized intensity per mg leaf dry weight (Supplementary Materials Tables S1 and S2).

Statistical Analysis
We performed differential analyses on the lipidomics dataset by conducting three types of tests: (1) comparison between genotypes with a given treatment; (2) treatment effect test when the genotype is fixed; and (3) interaction effect test between a genotype and a treatment. Raw lipidomics data were transformed by using where r ij is the lipidomics data for the ith lipid in the jth sample, δ is the half of the smallest non-zero measure, and R j is the upper quartile for sample j. The transformed data was analyzed with the voom function in the R package limma (https://bioconductor.org/ packages/release/bioc/html/limma.html (accessed on 17 February 2022)). A linear model was fit to each lipid and an empirical Bayes moderated F-statistic was calculated for testing differential expression. To account for multiple testing effects, all p-values were converted to q-values [85]. The resulting q-values were compared to a significance level of 0.05 to control the false discovery rate around 5%. Lipids with q-values no larger than 0.1 were declared to be statistically different. A principal component analysis (PCA) was performed using the pcomp function of the factoextra package in R. Clustering for heat maps was performed using the pheatmap package in R. Spearman's correlation scatter plot was generated using the ggplot2 package in R.

Unsaturation Index
The unsaturation index for a lipid molecular species was calculated as the average number of double bonds per acyl chain, which is the number of double bonds in the lipid molecular species divided by the number of acyl chains. The unsaturation index of a lipid headgroup class was calculated as previously described [52].

MALDI Mass Spectrometry Imaging
Leaves were placed on packing tape and dried under vacuum for about 1 h. Once dry, the tape was folded, enclosing the leaf between the tape, and passed through a rolling press. The tape pieces were pulled apart, resulting in the fracturing of the leaf to expose internal metabolites [52]. The tape pieces were aligned on glass slides for analysis. Matrices were applied as follows: 2,5-dihydroxybenzoic acid (DHB) was sublimated for 5 min at 140 • C and the binary matrix DHB: Fe 3 O 4 (75 mm:5 mm in IPA) was sprayed using an oscillating capillary nebulizer (OCN) [86]. Negative mode data were acquired using 1,5-diaminonaphthalene (DAN). Data were acquired using a MALDI-linear ion trap (LIT)-Orbitrap mass spectrometer (MALDI-LTQ-Orbitrap Discovery; Thermo Finnigan, San Jose, CA, USA) with an external frequency-tripled Nd: YAG laser (UVFQ; Elforlight Ltd., Daventry, UK). The laser energy was optimized for each matrix and ten laser shots were used for every raster step. A raster step of 100 µm was used for imaging the fractured leaves. All images shown were generated using ImageQuest software version 1.1.0 (Thermo Scientific; San Jose, CA, USA). A mass tolerance of ±0.01 Da was used to produce MS images at each pixel. The maximum and minimum values for a particular metabolite in all sample types were set to the same values to facilitate comparisons. Peak assignments were made using accurate mass.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/metabo12020190/s1. Figure S1: Nitrogen starvation alters lipid composition of Arabidopsis seedlings, Figure S2: Lipid changes in autophagy-altered mutants under +N conditions, Figure S3: Changes in polyunsaturated and very long chain fatty acids in autophagy-blocked mutants compared to WT under +N conditions, Table S1: Lipid amounts (nmol/mg dry weight), Table S2: Normalized average amount of each lipid species for each genotype under +N and −N conditions.