Comprehensive Analysis of Autophagy-Related Genes in Sweet Orange (Citrus sinensis) Highlights Their Roles in Response to Abiotic Stresses.

Autophagy is a highly conserved intracellular degradation pathway that breaks down damaged macromolecules and/or organelles. It is involved in plant development and senescence, as well as in biotic and abiotic stresses. However, the autophagy process and related genes are largely unknown in citrus. In this study, we identified 35 autophagy-related genes (CsATGs—autophagy-related genes (ATGs) of Citrus sinensis, Cs) in a genome-wide manner from sweet orange (Citrus sinensis). Bioinformatic analysis showed that these CsATGs were highly similar to Arabidopsis ATGs in both sequence and phylogeny. All the CsATGs were randomly distributed on nine known (28 genes) and one unknown (7 genes) chromosomes. Ten CsATGs were predicted to be segmental duplications. Expression patterns suggested that most of the CsATG were significantly up- or down-regulated in response to drought; cold; heat; salt; mannitol; and excess manganese, copper, and cadmium stresses. In addition, two ATG18 members, CsATG18a and CsATG18b, were cloned from sweet orange and ectopically expressed in Arabidopsis. The CsATG18a and CsATG18b transgenic plants showed enhanced tolerance to osmotic stress, salt, as well as drought (CsATG18a) or cold (CsATG18b), compared to wild-type plants. These results highlight the essential roles of CsATG genes in abiotic stresses.

The autophagy process maintains a basal level under normal conditions, but it can be quickly induced by external stimuli such as nutrition starvation, drought, salt, heat, and oxidative and osmotic stresses [3,12,[14][15][16]. Many studies have demonstrated the functions of ATGs in alleviating abiotic stresses in plants. For example, an apple MdATG8i (Malus domestica, Md) was significantly up-regulated in response to nitrogen depletion and oxidative stress, and overexpression of MdATG8i in apple callus conferred enhanced tolerance to nutrient-limited conditions [17]. MdATG3a, MdATG3b, MdATG7b, and MdATG18a of apple were shown to confer tolerance to nitrogen/carbon starvation, drought, salt, or osmotic stresses by ectopic-expressing in Arabidopsis and apple [18][19][20][21]. Li et al. reported that a banana MaATG8f improved drought stress tolerance in transgenic Arabidopsis through modulating reactive oxygen species (ROS) metabolism, abscisic acid biosynthesis, and autophagy activity [22]. Chen et al. overexpressed an ATG8 in Arabidopsis that increased autophagic activity and improved nitrogen remobilization efficiency [23]. Moreover, overexpression of a SiATG8a of foxtail millet conferred tolerance to both nitrogen starvation and drought in transgenic Arabidopsis and rice [11,24]. In contrast, silencing of AtATG18a in Arabidopsis reduced drought and salt tolerance [25]. Knockdown of the ATG6 of wheat inhibited autophagy and resulted in accelerated programmed cell death (PCD) under drought stress [26]. Loss of ATG10 and ATG18f significantly reduced HsfA1a-induced drought tolerance and autophagosome formation in tomato [27]. Under salt stress, autophagosome formation was rapidly induced, and the atg2 and atg7 mutants of Arabidopsis showed hypersensitive phenotype [28]. Under heat stress, atg mutants of Arabidopsis displayed visibly impaired pollen development (including atg2-1, atg5-1, atg7-2, and atg10-1) [29] or enhanced growth inhibition (including atg2-1, atg5-1, atg12ab, and atg18a-2) [30]. Recently, Shinozaki et al. proved that autophagy played important role in maintaining zinc bioavailability to avoid ROS accumulation under zinc deficiency in Arabidopsis [31]. Moreover, the potential roles of autophagy in waterlogging and excess aluminum and copper stresses were also reported [13,32,33].
As the largest fruit crop, citrus is widely grown around the world. In 2017, the global citrus growing area and yield were 14.1 million hectares and 146.6 million tons, respectively (FAO statistics, http://faostat.fao.org/default.aspx). However, citrus production is limited by abiotic stresses including drought, heat, cold, and nutrient disorders. Although the underlying physiological and molecular mechanisms triggered in response to these stresses have been well studied, the autophagy process and ATG genes in citrus are largely unknown. In this study, we systematically identified 35 ATGs from citrus by using the genome of sweet orange [34]. Comprehensive analysis including bioinformatic characterizations and expression profiles of these ATGs were conducted and discussed. Moreover, two ATGs of sweet orange, CsATG18a and CsATG18b (Citrus sinensis, Cs), were cloned and overexpressed in Arabidopsis. This further supported their functions in improving tolerance to drought, salt, cold, and osmotic stresses.

Identification of 35 ATGs in the Sweet Orange Genome
After a BLASTP search of the sweet orange genome with known AtATG proteins as queries and confirmation of the existence of the ATG domains in the PFAM database, a total of 35 putative ATG proteins were identified from sweet orange (Table 1). These ATGs showed 47.6% to 90.9% sequence identity with AtATGs, and they were named and numbered as CsATGs according to the highest sequence identity and the closest phylogenic relationship of AtATG homologues (Table 1 and Figure 1). On the basis of known functional classification of AtATGs, all CsATGs were also classified in the categories of the ATG1/13 kinase complex, PI3K complex, ATG9/2/18 complex, ubiquitin-like ATG8 and PE conjugation pathway, ubiquitin-like ATG12 and ATG5 conjugation pathway, and SNARE (Table 1). Among the 35 CsATGs, 15 contained only one member, whereas 4 CsATGs were encoded by gene families (three members for CsATG1, six members for CsATG8, seven members for CsATG18, and four members for CsVTI12). The ORF length of the CsATGs ranged from 315 bp (CsATG12) to 7095 bp (CsTOR), encoding 104 to 2364 amino acids (Table 1). This suggested the existence of significant variations and potential function differentiation. Predicted subcellular localization suggested that most CsATGs express in the cytoplasm, but there are also members located in the mitochondria, plasma membrane, Golgi apparatus, vacuoles, and nucleus (Table 1).

Bioinformatic Characterizations of CsATGs
Neighbor-joining phylogenetic trees were constructed with 35 CsATGs, 40 AtATGs, 30 NtATGs, 31 OsATGs, and 33 VvATGs, and the results showed that most of the CsATG proteins closely clustered with their homologues in Arabidopsis and grapevine ( Figure 1). For the ATGs containing multiple members (such as CsATG1, CsATG8, CsATG18, CsVTI12, and CsVPS in sweet orange), the different members of same subfamily were clustered in two or three different branches. For example, three CsATG1 members were clustered in two branches (CsATG1b and CsATG1d in one branch, and CsATG1a in another branch), and seven CsATG18 members were clustered in three branches (CsATG18a, CsATG18c, and CsATG18e in one branch; CsATG18f, CsATG18g, and CsATG18h in one branch; and CsATG18b in another branch). This result suggests potential functional differentiations in these subfamilies.

Expression Patterns of CsATGs Under Abiotic Stresses
As shown in Figure 4, most of CsATG genes were up-regulated under different stresses. For example, 27, 28, 23, 22, and 29 of CsATGs were significantly up-regulated (log2FC > 1) during at least one time-point of drought, heat, cold, mannitol, or salt treatments, respectively. In particular,

Expression Patterns of CsATGs Under Abiotic Stresses
As shown in Figure 4, most of CsATG genes were up-regulated under different stresses. For example, 27, 28, 23, 22, and 29 of CsATGs were significantly up-regulated (log 2 FC > 1) during at least one time-point of drought, heat, cold, mannitol, or salt treatments, respectively. In particular, CsATG2, CsATG8g, CsATG13, and CsATG18h under drought; CsATG2, CsATG18e, CsATG18g, and CsVTI12b under heat; CsATG4 under cold; and CsATG8a, CsATG18e, and CsVTI12c under mannitol and salt showed high expression levels, suggesting their potential key roles in response to these stresses. Apart from CsATG1d, CsATG5, CsATG7, CsATG8d, CsATG8g, CsATG16, CsATG18e, CsTOR, and CsVTI12d, the other 26 tested genes were significantly up-regulated during both drought and heat stresses. Twenty CsATGs showed significant up-regulation under both mannitol and salt stress. Remarkably, CsATG2, CsATG4, CsATG8i, CsATG9, CsATG10, CsATG18a, CsATG18b, and CsVTI12c were significantly up-regulated under all tested stresses simultaneously, suggesting their important roles in response to these stresses. In addition, some genes were significantly up-regulated at all tested time points of treatment, such as CsATG9 and CsATG13 under drought. Apart from these up-regulated CsATGs, we also noticed that CsATG1a was down-regulated under cold, mannitol, and salt treatments, which differed from the expression of the other two members of the ATG1 subfamily, CsATG1b and CsATG1d. The expression of CsATG5 was not obvious (absolute value of log 2 FC < 1) under most of the stresses, except for a significant down-regulation under 12 h of heat treatment. Similarly, the expression of CsATG8d was also steady, except for a significant up-regulation under mannitol and salt treatments.  Under excess Cu (+Cu), Mn (+Mn), and Cd (+Cd) treatments, the expression of most CsATGs were down-regulated or not obvious in roots and leaves, and only a few CsATGs were significantly up-regulated at one or two time-points of treatment ( Figure 5). Generally, CsATG2, CsATG4, CsATG5, CsTOR, and CsVPS15 were significantly down-regulated (log2FC < -1) in both roots and leaves, CsATG18g, CsATG18h, and CsATG20 were significantly down-regulated in leaves, and CsATG13 was significantly down-regulated in roots under +Cu, +Mn, and +Cd stresses simultaneously. In addition, Under excess Cu (+Cu), Mn (+Mn), and Cd (+Cd) treatments, the expression of most CsATGs were down-regulated or not obvious in roots and leaves, and only a few CsATGs were significantly up-regulated at one or two time-points of treatment ( Figure 5). Generally, CsATG2, CsATG4, CsATG5, CsTOR, and CsVPS15 were significantly down-regulated (log 2 FC < -1) in both roots and leaves, CsATG18g, CsATG18h, and CsATG20 were significantly down-regulated in leaves, and CsATG13 was significantly down-regulated in roots under +Cu, +Mn, and +Cd stresses simultaneously. In addition, the expression of CsATG1a in +Cd roots and +Cu leaves, as well as CsATG8g and CsATG18c in +Cu leaves showed extreme down-regulation. Expression of CsATG8g and CsATG9 in +Cd roots, CsATG8c in +Cu leaves, and CsATG12 in both +Cu and +Mn leaves showed obvious up-regulation ( Figure 5). These genes may play roles in response to heavy metal stress. leaves showed extreme down-regulation. Expression of CsATG8g and CsATG9 in +Cd roots, CsATG8c in +Cu leaves, and CsATG12 in both +Cu and +Mn leaves showed obvious up-regulation ( Figure 5). These genes may play roles in response to heavy metal stress.

Overexpression of CsATG18a/b Conferred Osmotic and Salt Tolerance in Arabidopsis
To confirm the functions of CsATGs in abiotic stresses, CsATG18a and CsATG18b were cloned and overexpressed in Arabidopsis. Three homozygous overexpression (OE) lines (T4 generation) and wild-type (WT, Col-0) of Arabidopsis were selected to perform germination experiments. The results showed that the OE lines of CsATG18a (OE1, OE3, and OE5) and CsATG18b (OE1, OE3, and OE4) had significantly higher germination rates than WT on 200 mM NaCl and 400 mM mannitol ( Figure 6 and Figure 7). These data suggest that overexpression of both CsATG18a and CsATG18b confer osmotic and salt tolerance.

Overexpression of CsATG18a/b Conferred Osmotic and Salt Tolerance in Arabidopsis
To confirm the functions of CsATGs in abiotic stresses, CsATG18a and CsATG18b were cloned and overexpressed in Arabidopsis. Three homozygous overexpression (OE) lines (T4 generation) and wild-type (WT, Col-0) of Arabidopsis were selected to perform germination experiments. The results showed that the OE lines of CsATG18a (OE1, OE3, and OE5) and CsATG18b (OE1, OE3, and OE4) had significantly higher germination rates than WT on 200 mM NaCl and 400 mM mannitol ( Figure 6; Figure 7). These data suggest that overexpression of both CsATG18a and CsATG18b confer osmotic and salt tolerance.   The data are the means ± SE of three independent repeats (n = 60). * indicates a significant difference at p < 0.05 using the least significant difference (LSD) test. Figure 8 shows that the OE lines of CsATG18a showed similar wilting phenotype with WT after 15 days of drought treatment. However, after rewatering for 2 days, all of the OE lines recovered to near normal, whereas the WT still exhibited wilting. In addition, the relative electrolyte leakage of OE lines was significantly lower than that of WT after either drought for 15 days or recovery for 2 days. These results suggest that overexpression of CsATG18a conferred drought tolerance in Arabidopsis. For the OE lines of CsATG18b, we also evaluated their tolerance to drought stress but observed no obvious tolerant phenotype ( Figure S1). The data are the means ± SE of three independent repeats (n = 60). * indicates a significant difference at p < 0.05 using the least significant difference (LSD) test. Figure 8 shows that the OE lines of CsATG18a showed similar wilting phenotype with WT after 15 days of drought treatment. However, after rewatering for 2 days, all of the OE lines recovered to near normal, whereas the WT still exhibited wilting. In addition, the relative electrolyte leakage of OE lines was significantly lower than that of WT after either drought for 15 days or recovery for 2 days. These results suggest that overexpression of CsATG18a conferred drought tolerance in Arabidopsis. For the OE lines of CsATG18b, we also evaluated their tolerance to drought stress but observed no obvious tolerant phenotype ( Figure S1).

Overexpression of CsATG18b Conferred Cold Tolerance in Arabidopsis
As shown in Figure 9, the OE lines of CsATG18b had a similar growth phenotype with WT before and after cold treatment (-20 °C). However, after recovery for 2 days, the three OE lines had less leaf damage. In addition, the relative electrolyte leakage of OE lines was significantly lower than that of WT after cold treatment for 5 min (Figure 10). These results suggest that overexpression of CsATG18b conferred cold tolerance in Arabidopsis. For the OE lines of CsATG18a, we did not observe a cold tolerance phenotype ( Figure S2).

Overexpression of CsATG18b Conferred Cold Tolerance in Arabidopsis
As shown in Figure 9, the OE lines of CsATG18b had a similar growth phenotype with WT before and after cold treatment (-20 • C). However, after recovery for 2 days, the three OE lines had less leaf damage. In addition, the relative electrolyte leakage of OE lines was significantly lower than that of WT after cold treatment for 5 min (Figure 10). These results suggest that overexpression of CsATG18b conferred cold tolerance in Arabidopsis. For the OE lines of CsATG18a, we did not observe a cold tolerance phenotype ( Figure S2).  Arabidopsis under cold stress. Relative electrolyte leakages of WT and OE lines were measured after -20°C treatment for 5 min. The data are the means ± SE of three independent repeats. * indicates a significant difference at p < 0.05 using an LSD test.

Discussion
ATG genes are known to be involved in the autophagy process [1]. Most ATGs were first identified and functionally characterized by mutagenesis studies in yeast [2]. More than 30 ATGs have now been identified in yeast, and 23, including ATG1-10, 12-14, 16-18, 20, 27, 29, 31, TOR, VPS15, and VPS34, are considered to be the core ATGs participating in autophagy [3,4]. In Arabidopsis, around 40 homologues to these core ATGs have been identified, except for ATG14, 17,27,29, and 31 with no homologues [1,3]. Similarly, a total of 19 core CsATGs, including 35 members, were identified from the sweet orange genome in this study. The quantity of ATGs in sweet orange is similar to the quantity   Arabidopsis under cold stress. Relative electrolyte leakages of WT and OE lines were measured after -20°C treatment for 5 min. The data are the means ± SE of three independent repeats. * indicates a significant difference at p < 0.05 using an LSD test.

Discussion
ATG genes are known to be involved in the autophagy process [1]. Most ATGs were first identified and functionally characterized by mutagenesis studies in yeast [2]. More than 30 ATGs have now been identified in yeast, and 23, including ATG1-10, 12-14, 16-18, 20, 27, 29, 31, TOR, VPS15, and VPS34, are considered to be the core ATGs participating in autophagy [3,4]. In Arabidopsis, around 40 homologues to these core ATGs have been identified, except for ATG14, 17,27,29, and 31 with no homologues [1,3]. Similarly, a total of 19 core CsATGs, including 35 members, were identified from the sweet orange genome in this study. The quantity of ATGs in sweet orange is similar to the quantity Figure 10. Overexpression of CsATG18b decreased relative electrolyte leakage of transgenic Arabidopsis under cold stress. Relative electrolyte leakages of WT and OE lines were measured after −20 • C treatment for 5 min. The data are the means ± SE of three independent repeats. * indicates a significant difference at p < 0.05 using an LSD test.

Discussion
ATG genes are known to be involved in the autophagy process [1]. Most ATGs were first identified and functionally characterized by mutagenesis studies in yeast [2]. More than 30 ATGs have now been identified in yeast, and 23, including ATG1-10, 12-14, 16-18, 20, 27, 29, 31, TOR, VPS15, and VPS34, are considered to be the core ATGs participating in autophagy [3,4]. In Arabidopsis, around 40 homologues to these core ATGs have been identified, except for ATG14, 17, 27, 29, and 31 with no homologues [1,3]. Similarly, a total of 19 core CsATGs, including 35 members, were identified from the sweet orange genome in this study. The quantity of ATGs in sweet orange is similar to the quantity in grape (35 VvATGs) [13], banana (32 MaATGs) [12], rice (33 OsATGs) [7], and foxtail millet (37 SiATGs) [11], suggesting very conservative evolution of ATGs in these plants. Among the 19 core CsATGs, we found that CsATG1, CsATG8, and CsATG18 contained multiple members, which is similar to that in Arabidopsis and other plants [1,8,13]. However, unlike having two ATG4, ATG12, and ATG13 genes and one VTI12 in Arabidopsis, sweet orange has one member of CsATG4, CsATG12, and CsATG13 and four members of CsVTI12. This indicates a citrus crop-specificity in the number of these ATGs [1,8]. On the basis of gene duplication analysis, a high proportion of segmental duplications was predicted for those multiple members of CsATGs, especially for CsATG8. This suggests that expansion of these CsATG subfamilies was possibly derived from gene duplication during evolution [34].
Before functions were identified via transgenic strategy, the potential functions of unknown genes were often predicted according to their orthologous genes, conserved domains, and expression patterns. In this study, 29 of the CsATG genes showed higher than 60% sequence identity to homologues of Arabidopsis. Most of these high identities of CsATG and AtATG were also closely clustered together in phylogenetic trees. This result indicates that some CsATG genes may have functions similar to their homologues in Arabidopsis. On the basis of analysis of conserved domains, three CsATG1 members have similar serine/threonine-protein kinase domains that function in forming a complex with ATG13 and regulating the initiation of autophagy [35]. The UBQ and v-SNARE domains are conserved in all CsATG8 and CsVTI12 members, respectively. In general, the similar domain containing in the family members indicates the possible functional redundancy of this family [8]. However, the phylogenetic analysis showed that the members of CsATG1, CsATG8, and CsVTI12 were obviously clustered in two different branches, indicating that functional divergences are also existing in these subfamilies. A functional divergence had been reported for ATG8b and ATG8i during the hydrotropic response in Arabidopsis roots [36], which agrees well with their different phylogenetic branches in this study. In the CsATG18 subfamily, all members contain the WD40 domain, except CsATG18f, CsATG18g, and CsATG18h with an extra BCAS3 domain. The phylogenetic analysis also showed that the CsATG18f, CsATG18g, and CsATG18h were clustered in an independent branch, indicating functional divergence of these three CsATGs with the other members in this family [10].
ATG genes have played key roles in plant responses to different abiotic stresses. We found that the expression of most CsATG genes were induced under drought, heat, cold, mannitol, or salt stresses. Our results support the previous findings regarding the involvement of autophagy in abiotic stresses [3,25,28]. Although most CsATGs were induced by these stresses, some CsATGs showed different expression patterns to different stresses. For example, CsATG1a was significantly up-regulated under drought and heat but down-regulated under cold. In contrast, CsATG1d was significantly up-regulated under cold. In the CsATG8 subfamily, CsATG8a, CsATG8c, CsATG8g, and CsATG8i were obviously induced by drought and heat, but this was not true for CsATG8d and CsATG8f. Similar results have been reported for ATGs of other plant species, such as VvATGs of grape [13], NtATGs of tobacco [8], and CaATGs of pepper [10]. These results indicate that there may be different pathways to regulate the autophagy process under different abiotic stresses [10,15,25]. Compared to the above tested abiotic stresses, the roles of autophagy in heavy metals toxicity are less well known in plants.
To provide more evidence, expression patterns of CsATGs were analyzed under high levels of Cu, Mn, and Cd. Unlike the predominant up-regulation of CsATGs under drought, heat, cold, and salt stress, more down-regulated CsATGs were found upon exposure to excess Cu, Mn, and Cd. In addition, the expression patterns were different between heavy metal-treated roots and leaves. Although we could not find similar reports to support our result at present, these data provide a new reference for understanding potential roles of ATGs in heavy metal stress.
ATG18 is a WD40 domain containing protein family with multiple members. They can function in vesicle formation in autophagy by binding to the phosphatidylinositol 3-phosphate sites in plants [2]. Among the multiple ATG18 members, ATG18a is the most studied. RNAi-AtATG18a transgenic Arabidopsis plants were autophagy-defective and more sensitive to oxidative, salt, and drought stresses than WT [25,37]. MdATG18a of apple showed significant up-regulation in response to leaf senescence, drought, heat, oxidative stress, and nitrogen starvation, and overexpression of MdATG18a in tomato and apple enhanced drought tolerance [21,38]. In the present study, overexpression of a CsATG18a in Arabidopsis also improved tolerance to osmotic stress, salt, and drought, which is consistent with previous results. Another member, CsATG18b, was shown to confer enhanced tolerance to osmotic stress, salt, and cold. This result provides new evidence regarding the functions of ATG18b in plants. Under abiotic stresses, a common response in cells is ROS accumulation and oxidative damage, including membrane breakage and protein oxidation and aggregation [21,37]. Thus, we speculate that enhanced tolerance in CsATG18a and CsATG18b transgenic plants might be because the damaged proteins and components are more efficiently degraded and recycled by the autophagy pathway, thereby reducing ROS accumulation and maintaining membrane integrity under stress [37,39]. Interestingly, a functional difference between CsATG18a and CsATG18b was noticed under drought and cold. Overexpression of CsATG18a conferred drought but not cold tolerance, whereas overexpression of CsATG18b conferred cold but not drought tolerance. This result appears reasonable and explainable. First, CsATG18a and CsATG18b have very different gene structures and predicted subcellular localization, which indicates their functional divergence (Table 1 and Figure 3). Second, qRT-PCR analysis shows that CsATG18a has a higher expression level under drought but a lower expression level under cold than CsATG18b ( Figure 4). Third, CsATG18a and CsATG18b may be regulated by different transcription factors (TFs) under drought and cold. There is evidence that ATG18a interacts with the WRKY33 transcription factor to mediate autophagosome formation and necrotrophic pathogen resistance in Arabidopsis [40]. The HsfA1a transcription factor binds to the ATG18f promoter to regulate drought tolerance in tomato [27]. ATG18a and ATG18f are regulated by two different TFs, indicating that CsATG18a and CsATG18b may be also regulated by different TFs, thereby resulting in their functional differentiations.

Plant Materials and Treatments
"Hamlin" sweet orange (Citrus sinensis (L.) Osbeck) was used as the plant material in this study. The seeds of "Hamlin" sweet orange were collected from the Citrus Research Institute, Southwest University, Chongqing, China. The seeds were sterilized and germinated, and the seedlings were hydroponically cultured at 25 • C and 16 h photoperiod (50 µmol m −2 s −1 ), as described by Fu et al. [41]. After 30 days of growth in hydroponic solution, uniform seedlings were divided into different groups (3 biological repeats and 10 plants per repeat for each group) for the following sets of treatments. For the drought treatment, the seedlings were dehydrated under room temperature (25 • C) by removing the hydroponic solution. For the heat and cold treatments, two groups of seedlings were put in a growth chamber under 42 • C and a refrigerator under 4 • C, respectively. For the mannitol and NaCl treatments, two groups of seedlings were transferred to 400 mM mannitol and 200 mM NaCl solutions, respectively. For the excess Cu, Mn, and Cd treatments, the seedlings were transferred to new hydroponic solutions containing 0.01 mM CuSO4 (+Cu), 0.1 mM MnSO4 (+Mn), and 0.038 mM CdSO4 (+Cd), respectively. The seedlings grown under normal conditions were used as controls (CK). The leaves were sampled after 3 h and 12 h of drought, heat, and cold treatments or 2 h and 3 h of mannitol and NaCl treatments, and the leaves and roots were sampled upon 5 days and 15 days of excess Cu, Mn, and Cd treatments. Samples were immediately frozen in liquid nitrogen and stored at −80 • C until use.

Genome-Wide Identification of ATG Genes in Sweet Orange
To identify all putative ATG proteins of the sweet orange, a BLASTP search of the sweet orange genome database (http://citrus.hzau.edu.cn/orange/) [34] was performed using the known Arabidopsis ATG proteins (AtATGs) as query sequences [1,3]. The results were filtered using a score value of ≥100 and an e-value ≤ e −10 [42]. After removal of redundant sequences, all putative ATG genes were submitted to the PFAM database (http://pfam.xfam.org/) to confirm the existence of the ATG domains. All of the identified ATG genes of sweet orange were named as CsATGs.

Bioinformatic Analysis of CsATGs
The subcellular locations of the CsATG proteins were predicted online by using the ProtComp v. 9.0 (http://www.softberry.com/berry.phtml?topic=protcomppl&group=programs&subgroup=proloc). To construct phylogenetic trees, all putative CsATGs and the ATG proteins from Arabidopsis thaliana [1,3], tobacco (Nicotiana tabacum) [8], rice (Oryza sativa) [7], and grapevine (Vitis vinifera) [13] were first collected (as shown in Table S1) and then submitted to ClustalW for performing multiple sequence alignment [43]. The generated files were used to construct the phylogenetic tree with the neighbor-joining method and 1000 replicates of bootstrap analysis in MEGA 6 software [44]. The bootstrap values lower than 50% are not shown in the phylogenetic trees. The chromosomal positions of the CsATG genes were provided by the sweet orange genome database, and MapInspect software (http://mapinspect.software.informer.com) was used to draw the location images. For the gene structure analysis, the generic feature format (GFF) file of each CsATG was downloaded from the sweet orange genome database and then submitted to the Gene Structure Display Server (GSDS, http://gsds.cbi.pku.edu.cn/) to display intron-exon structures [45]. The GFF files were also inputted into MCScanX software to identify gene duplications and collinear correlations [46]. The conserved domains of CsATG proteins were predicted online in the NCBI Conserved Domain Database (https://www.ncbi.nlm.nih.gov/cdd/).

Total RNA Isolation and Quantitative Real-Time PCR (qRT-PCR) Analysis
Total RNA was extracted from the CK and stress-treated leaves or roots of "Hamlin" sweet orange seedlings using a RNAprep pure plant kit (Tiangen Biotech Co., Ltd., Beijing, China), and the RNA concentration and quality were determined with a Nanodrop 2000 spectrophotometer (Thermo Scientific, Waltham, MA, USA). Then, 1 µg of high-quality RNA was used for cDNA synthesis with an iScript cDNA synthesis kit (Bio-Rad) according to the manufacturer's instructions. qRT-PCR was performed on the Bio-Rad CFX Connect RealTime system using ChamQ Universal SYBR qPCR Master Mix (Vazyme Biotech Co., Ltd., Nanjing, China). Each PCR reaction contained 5.0 µL SYBR mix, 0.2 µM primers, and 1.0 µL diluted cDNA in a final volume of 10 µL. Specific primers of CsATG genes were designed using the online primer-blast program in the NCBI website, whereas Actin (Cs1g05000.1) was used as a reference gene to normalize the relative expression levels of the tested genes. Three biological replicates and three technical replicates were performed for each treatment.

Construction of Transgenic Arabidopsis Expressing 35S-CsATG18a and 35S-CsATG18b
The CDS (coding sequence) fragments of CsATG18a and CsATG18b were first isolated from "Hamlin" sweet orange. After being confirmed by sequencing, the full CDS fragments of CsATG18a and CsATG18b were recombined into a pGBO vector to produce 35S-CsATG18a and 35S-CsATG18b constructs, respectively. The two constructs were then transformed into Agrobacterium tumefaciens strain GV3101 by electroporation. The Arabidopsis thaliana wild-type (WT; ecotype Col-0) plants were used to perform Agrobacterium-mediated transformation by the floral dipping method [47]. The transgenic plants were selected on 1/2 MS medium containing 50 mg L −1 kanamycin until a homozygous T4 generation was obtained. The expression levels of transgenes were measured by qRT-PCR.

Phenotypic and Physiological Analysis of Transgenic Arabidopsis Lines
To evaluate the tolerance of CsATG18a and CsATG18b transgenic Arabidopsis plants to different abiotic stresses, three independent T4 homozygous lines with relatively high expression levels of CsATG18a and CsATG18b were selected for NaCl, mannitol, drought, and cold treatments, and WT (Col-0) plants were used as controls. For the NaCl and mannitol treatments, the seeds of transgenic lines and WT were sown on 1/2 MS with 200 mM NaCl or 400 mM mannitol, and the germination rate was calculated at 3 to 7 days. For the drought treatment, 30 day old transgenic and WT plants were dewatered for 15 days and then watered for recovery. For the cold treatment, 30 day old transgenic and WT plants were transferred to a −20 • C freezer for 5 min and then back to room temperature for recovery. After drought and cold treatments, the relative electrolytic leakage was measured as described by Shi et al. [48].

Conclusions
A total of 35 CsATG genes were identified from the genome of sweet orange. These CsATGs showed high sequence and phylogeny similarities with the AtATGs of Arabidopsis. Most CsATG genes were significantly up-or down-regulated in response to drought, heat, cold, salt, and osmotic stresses, as well as to excess Cu, Mn, and Cd stresses. Overexpression of CsATG18a and CsATG18b enhanced tolerance to both salt and osmotic stresses in Arabidopsis. Overexpression of CsATG18a in Arabidopsis conferred drought but not cold tolerance, whereas overexpression of CsATG18b conferred cold but not drought tolerance. This study analyzed CsATG genes in citrus and preliminarily revealed their important roles in response to abiotic stresses. In future, their functions and the underlying mechanisms will be further uncovered by overexpressing in citrus plants.

Conflicts of Interest:
The authors declare no conflict of interest.