A total of 17 independent primary transgenic lines were selected via the transformation of pOIL076 construct into potato (Solanum tuberosum cv Atlantic) on the kanamycin-containing media. The transgenic status of these plants was verified by polymerase chain reaction (PCR) of each of the three transgenes being overexpressed, including AtWRI1, AtDGAT1 and SiOLEOSIN1 from genomic DNA (data not shown). The analysis of the total fatty acid (TFA) contents of senescent leaves of the 17 independent T₀ transgenic potatoes showed a significant variation between 2.6 and 4.5% of leaf DW, compared to 2.4% in the wild type (WT) (Figure 1A). Among these transgenic lines, L3 and L5, which contained relatively high levels of TFA, 4.11 and 4.46% respectively, were selected for further analysis. Potato plants are typically propagated vegetatively by tuber-cutting. Consequently the transgenes in transgenic potato plants remain in their heterozygous status without segregation. In order to obtain synchronized growth and physiological status of potato plants for analysis, WT, L3 and L5 transgenic lines were propagated by tuber-cutting and grown under the controlled glasshouse conditions and sampled at three developmental stages including flowering, mature and senescent stages.

Transgene expression assessment was carried out through real-time quantitative reverse transcription polymerase chain reaction (qRT-PCR). The expressions of AtWRI1, AtDGAT1 and SiOLEOSIN were not detected in WT, but showed variable expression patterns in the two selected transgenic lines during plant development, relative to the reference gene S. tuberosum cyclophilin (stCYP) (Figures 1B,C). AtWRI1 displayed low yet consistent expression in both transgenic lines at the flowering stage, and was significantly increased afterward, particularly at the senescent stage in L3 (Figure 1B). AtDGAT1 constantly exhibited the highest expression among the three transgenes, which peaked at the senescent stage in both L3 and L5. The expression of SiOLEOSIN was relatively consistent over the three developmental stages in L3 (Figure 1B), in contrast to L5 in which the highest expression was observed at the flowering stage (Figure 1C).

Total fatty acid contents of both L3 and L5 showed consistently significant increases compared with WT over the three developmental stages, and reached the maximum level at the mature stage as 6.19 and 7.05% of leaf DW respectively (Figure 2B), but reduced thereafter. There were significant variations in the contents of starch and the total soluble sugars between transgenic lines and WT. Specifically, the contents of the total soluble sugars in L3 and L5 showed about 1.4-fold reduction at the flowering stage (Figure 2A), but were significantly increased at the mature and senescent stages relative to WT (Figure 2C). The starch contents of L3 and L5 were consistently lower than WT over the entire growth period. For example, the starch content in L3 was reduced drastically to as low as 3.12% (DW) at the senescent stage, which is a 2.6-fold drop compared to 7.96% in WT (Figure 2C). At the mature stage, starch content in WT reached as high as 14.91% of leaf DW, in contrast to transgenic lines with rather low starch accumulations (8.77% in L3 and 4% in L5, respectively) (Figure 2B).

In comparison to the rather moderate increases in TFA, TAG accumulation in both L3 and L5 was much more evident over the three developmental stages (Figure 3A). The highest TAG contents in L3 and L5 were recorded at the senescent stage as 0.84 and 0.82% of leaf DW respectively, which was nearly 30-fold increase compared to 0.03% in WT. TAG was clearly the predominant neutral lipid across the three developmental stages in L3 and L5, peaking at the senescent stage (Figure 3B). The accumulation of the two galactolipids, monogalactosyldiacylglycerol (MGDG) and digalactosyldiaclglycerol (DGDG), in transgenic leaves rose significantly at the flowering stage, with L5 displaying the highest MGDG content as 1.38% (DW), which was increased nearly 2-fold compared to WT (Figure 3C), but subsequently dropped to a lower level in the senescent stage (Figure 3E). DGDG accumulation showed a similar trend of change with MGDG during the development. Phospholipids including phosphotidylcholine (PC), phosphatidylethanolamine (PE), and phosphatidylglycerol (PG) were also correspondingly varied in L3 and L5, particularly at the flowering and senescent stages. Compared to WT, contents of PC were significantly increased in the two transgenic lines at the flowering stage, and PE and PG contents in L5 were nearly doubled (Figure 3C). At the senescent stage, only PC remained the higher level than WT, particularly in L3 (0.36% of DW), whereas PG was barely detectable (Figure 3E). Despite the highest production of TFA, no significant variation was observed in polar membrane lipids between WT and transgenic plants at mature stage (Figure 3D).

The variation in TAG and polar membrane lipids in transgenic lines was accompanied by the alteration of fatty acid composition relative to WT (Figures 4, 5). The significant reduction in the level of α-linolenic acid (ALA, C18:3^Δ9,12,15) in TAG and PC represented the major fatty acid change in transgenic potatoes compared to WT. Correspondingly, palmitoleic acid (C16:1^Δ9), oleic acid (C18:1^Δ9) and linoleic acid (LA, C18:2^Δ9,12), as well as the long chain fatty acids (LCFAs) including arachidic acid (C20:0) and the others, were all increased, particularly in TAG at the flowering and mature stages (Figures 4A,B). However, a significant reduction in palmitic acid (C16:0) in TAG was observed in L3 and L5 at the senescent stage, and a 3-fold increase in stearic acid (C18:0) was particularly reflected in L3 (Figure 4C). By comparison, such distinct fluctuations in the levels of saturated fatty acids and monounsaturated fatty acids (MUFA) were not reflected in PC throughout the leaf development (Figures 4D–F). In galactolipids, the fatty acid composition of transgenic lines was mainly featured by the significantly increased hexadecatrienoic acid (C16:3) in MGDG at the flowering stage (Figure 5A), and enhanced LA in both galactolipids at all stages, relative to WT (Figure 5).

In parallel with the biochemical analysis, microscopic analysis of leaves sampled at the three developmental stages was undertaken. Both LD and plastoglobuli were observed in the mesophyll cells of potato leaves with the transmission electron microscopy (TEM) (Figure 6). LDs were found in the cytosol in all three samples: WT, L3 and L5, with proximity to chloroplast and mitochondria, and plastoglobuli were found inside the chloroplasts. Under the two-dimensional (2D) horizon, both LD and plastoglobuli spheres were visualized as irregular round shapes. The average diameters were therefore compared in order to reflect the possible variation in the morphology. In WT, LDs did not appear to vary significantly with the aging of the leaf, but plastoglobuli enlarged as leaves developed, the average diameter increasing more than ten times from less than 0.1 μm at the flowering stage to 1 μm in the senescent stage (Figures 6A,D,G). A similar observation in terms of plastoglobuli size increase was also made in L3 and L5. But, in addition to the expanding plastoglobuli, the transgenic mesophyll cells were featured with dramatically enlarged LDs often with irregular shapes. Particularly, at the flowering stage, the diameter of LDs in L3 and L5 were approximately 3 and 5 μm respectively, which was in sharp contrast to merely 1.5 μm in WT. At the senescent stage, LDs of the two transgenic lines had enlarged dramatically to about 10 μm in diameter (Figures 6H,I). However, the number of LDs did not show significant variation in L3 and L5 from WT when the scale bars were normalized to 1 μm in all the photographs. Starch granules were observed in abundance in all chloroplasts imaged. But compared to WT, L3 and L5 seemed to exhibit decreased numbers of starch granules and potential alteration in granule shapes (Figures 6E,F,H).

In order to explore the potential factors impacting TAG production in transgenic potato leaves, characterization of the positional distribution of fatty acids in lipids specifically distributed into chloroplast and cytosol was carried out, together with the fundamental analysis of the intracellular TFA allocation and identification of one of the major components, fatty acid phytyl ester, in plastoglobuli of plant chloroplasts at the senescent stage (Rottet et al., 2015). These experiments aimed at initially exploring the potential variation in lipid compartmentalization and plastoglobuli biogenesis in potato leaves.

TAG purified from the senescent leaves of WT and two transgenic lines were digested with the Rhizopus arrhizus lipase, which preferentially cleaves fatty acids bonded to the outer positions (sn-1 and sn-3) of the glycerol backbone of a TAG molecule, resulting in the production of hboxtextitsn-2 MAG and sn-1/3 FFA. Thus, the fatty acids in sn-2 MAG represent the fatty acids of sn-2 position of TAG and FFA molecules reflecting the acyl components of sn-1/3 positions. As displayed in Table 1, the relatively higher preference to outer positions by saturated fatty acids was observed in both WT and the transgenic lines, while unsaturated fatty acids, mainly the C18 polyunsaturated fatty acids (PUFA), showed a clear preference to the sn-2 position. The increase in LA largely at the expense of ALA in transgenic plants was also reflected on the sn-2 position. MGDG and DGDG were digested with the same lipase to yield lyso-MGDG and lyso-DGDG retaining the sn-2 acyl chain and releasing FFAs from the sn-1 position (Tables 2, 3). Similar to TAG, the unsaturated fatty acids, represented by C16:3, LA and ALA, showed preference to sn-2 position while saturated fatty acids, mostly palmitic acid and stearic acid, were enriched at sn-1 position.

Confocal scanning microscopy was applied to visualize the distribution of neutral lipids in potato leaves at the senescent stage. The presence of neutral lipids mainly in the form of LD and plastoglobuli was visualized following Bodipy staining which is specific to neutral lipids (Figure 7). Compared with WT, significantly more abundant LDs were observed in L3 (Figures 7C,D) and L5 (Figures 7E,F). Plastoglobuli were associated with chloroplasts and appeared to be visually smaller than LD in cytosol. Such an observation was consistent with the TEM analysis. As being particularly exemplified in L5 with a further magnification, plastoglobuli, as a type of neutral lipids storage structure, were found to be overlapping with chloroplast, whereas LDs which incorporate TAG as the predominant component were found in cytosol (Figures 7G,H).

The acyl fatty acids derived from chloroplast and cytosol were then compared (Figure 8). After normalizing based on the amount of chlorophyll, the cytosolic TFA contents in potato leaves showed higher accumulation compared to the chloroplasts. Relative to WT, L3 and L5 showed a significant increase in the amount of TFA in chloroplasts but significant reduction in cytosols (Figure 8A). The fatty acid compositions of chloroplast and leaf were both featured by significantly increased LA at the expense of ALA in transgenic plants relative to WT, but the production of LCFAs was only identified in the leaf fatty acids (Figure 8C). Fatty acid phytyl esters, as one of the major components of plastoglobuli, were obtained by TLC fractionation, but slightly co-migrated with the wax ester components. At the flowering stage, significant deposition of the fatty acid phytyl esters was identified in both L3 and L5, in sharp contrast to WT in which the bands representing fatty acid phytyl esters were barely detectable on the TLC plate (Figure 8D), but became visible at the leaf senescent stage (Figure 8E).

Mature potato tubers were sampled at the leaf senescent stage in parallel with leaves for analysis. Real-time PCR indicated that the transgenes of AtDGAT1 and SiOLEOSIN also showed considerable levels of expression in the tubers of L3 and L5, whereas the expression of AtWRI1 was barely identified (Figure 9A). Generally, the L5 tuber displayed the highest expression levels of the two transgenes relative to L3, with SiOLEOSIN being the most highly expressed. As a result, significant alteration in the contents of lipids and carbohydrates was observed. In particular, TFA content of L5 was doubled and TAG increased 5-fold compared to WT (Figure 9B). Similarly in L3, the TAG content increased by 2-fold. The total polar lipids which accounted for the major part of tuber TFA also increased significantly in both L3 and L5 relative to WT. Interestingly, in terms of the total carbohydrate variation, L3 showed significant increase in starch content and reduction in the total soluble sugars compared to WT, but not in L5 (Figure 9B). The fatty acid compositions of TAG (Figure 9C) and total polar lipids (Figure 9D) in the transgenic tubers were generally consistent with the transgenic leaves, except that L3 showed enhanced ALA in polar lipids relative to WT.

The preliminary assessment of several agronomic traits of potato tubers including density, yield, water content, the number of tubers per plant, and tuber size were tabulated in Supplementary Table 1, with photographs recorded of the mature and healthy tubers harvested from individual plants (Supplementary Figure 1). Significantly, tuber density and tuber water content of L3 and L5 were both reduced compared to WT, accompanied by relatively increased tuber yields (recorded as total fresh tuber weight). However, the number of tubers produced per plant, as well as the tuber size, has been significantly reduced in L3. Tubers were divided into four size groups (Size A, B, C, D) from the large size to small size by the maximum tuber length per potato. Tubers clustered in Size C (3 cm < tuber length < 6 cm) represented the largest proportion in WT and L5 (51.26 and 40.65%, respectively), while the Size A (tuber length > 9 cm) represented the largest proportion in L3 (34.38%). Accordingly, L3 showed the highest average weight of a single tuber, compared to WT and L5.

The extraction of total lipids, lipid fractionation and quantification were carried out following the methods previously described by Liu et al. (2017a). Neutral lipids and free fatty acids (FFAs) were separated in the solvent system of hexane: diethyl ether: acetic acid (70: 30: 1, by volume) on TLC. Polar lipids including PC, MGDG, DGDG, PE, and PG were separated using the solvent system consisting of chloroform: methanol: acetic acid: distilled water (30: 5: 3: 1, by volume).

Positional distribution of fatty acids in TAG, MGDG and DGDG was investigated via lipase digestion. The Rhizopus arrhizus lipase (Fluka, Buchs, Switzerland) was utilized to digest TAG, MGDG and DGDG respectively. TAG, MGDG or DGDG were isolated from the total lipids of potato leaf at the senescent stage by TLC fractionation, and further purified using chloroform: methanol (2: 1, by volume). The treatment of each lipid digestion and subsequent fractionation were performed as described in Liu et al. (2017b). Fatty acid methyl esters (FAME) analysis was carried out using the samples prepared from the TLC purified sn-2 MAG from TAG, and the sn-1 FFA and sn-2 lyso-MGDG/DGDG from galactolipids by GC analysis. It should be noted that the fatty acid composition of the sn-1/3 FFA released from TAG was calculated as previously described in Christie et al. (1984). The digestion of galactolipids using the same methods as the TAG was previously reported by Yongmanitchai and Ward (1993).

Fatty acid phytyl esters were fractionated from the total leaf lipids on TLC using a solvent system hexane: diethyl ether: acetic acid (85: 15: 1, by volume) as described in Ischebeck et al. (2006). To indicate the probable position of fatty acid phytyl esters, the palmitoyl hexadecanoate was synthesized for use as the C16- phytyl standard in TLC analysis (Pereira et al., 2002).

Three phases were visibly separated in the chloroform/methanol/0.1M KCl based lipid extraction after centrifugation. The upper phase contains soluble sugars and proteins whereas the insoluble substances such as starch and fiber remain in the interphase and the lower phase containing the lipids dissolved in chloroform, which were used for analysis of total soluble sugars, starch and lipids, respectively. The total soluble sugars was analyzed according to the anthrone coloration method. Briefly, 5–10 μL of supernatant obtained from the lipid extraction was isolated and boiled for 10 min in 500 μL anthrone solution (0.2% anthrone in 70% H₂SO₄), and measured under 630 nm for absorbance. Total starch content was measured by boiling the sample for 30 min in 0.2 M NaOH and neutralized with 4 μL glacial acetic acid prior to be analyzed with the Megazyme Total Starch Kit (Megazyme International Ireland, Bray, Ireland) following the manufacturer’s instruction.

Total RNAs from potato leaf tissues at different developmental stages were extracted using the RNeasy Mini Kit (Qiagen, Hilden, Germany), followed by quality check on a Nanodrop spectrophotometer ND 1000 (Thermo Fisher Scientific) and in an ethidium bromide-stained 1% agarose gel electrophoresis. The total RNA from mature potato tubers were extracted by the cetyl trimethyl ammonium bromide (CTAB) method (Reynolds et al., 2019) and purified using a RNeasy MinElute Cleanup Kit (Qiagen) following manufacturer’s instructions.

Expression of transgenes was analyzed using the qRT-PCR in triplicate biological samples, each with two technical replicates. Reverse transcription of total RNAs into cDNA was performed with a SuperScript^TM IV First-Strand Synthesis System kit (Thermo Fisher Scientific, Waltham, MA, United States). The oligo nucleotide sequence used for each transgene was as same as described in the PCR verification of the T₀ transgenic potato plants. The reference gene and its corresponding primers selected to calibrate the expression analysis was the S. tuberosum CYCLOPHILIN (stCYP), as previously described in Liu et al. (2017a). The FastStart Universal SYBR Green Master (ROX) kit (Roche, Indianapolis, IN) was utilized to conduct the real-time qRT-PCR reaction, with the reaction program as 95°C for 3 min, 39 cycles of 95°C for 10 s, 58°C for 30 s and 72°C for 30 s using a Biorad 96 well PCR machine (Bio-Rad, Hercules, CA, United States). Calculation was made by following the 2^{–Δ Δ Ct} method (Livak and Schmittgen, 2001).

Chloroplasts were isolated from the senescent potato leaves following the method described in Hosaka and Hanneman (1987), the collected chloroplast pellets were gathered for the subsequent purification by using a sugar-gradient centrifuging method (Elias and Givan, 1978), through which intact potato chloroplasts were well separated and stored in 4 mL stock buffer (0.33 M sucrose in extraction buffer C) under 4°C. The integrity and quality of chloroplast was checked by light microscopy using Leica-DMR (Leica Microsystems, Wetzlar, Germany) (Supplementary Figure 3).

For the TFA extraction, 1.5 mL stock buffer containing purified chloroplasts on ice was first mixed with 0.5 mL 1M KCl solution, then 6 mL chloroform: methanol (2: 1, by volume) was added following 5 min vortex and 5 min centrifugation at 1,700 rpm. The lower phase was collected and evaporated under nitrogen, then dissolved in 200 μL chloroform as stock under −20°C. As chloroplast cannot be accurately quantified as leaf DW, the amount of chlorophyll was used as the reference standard for normalization. Chlorophyll from chloroplast and leaf lipid solutions was measured via the spectrometric method described in Warren (2008). FAME analysis of TFA were then proceeded using GC. The TFA in cytosol was calculated via subtracting the TFA of chloroplast from the TFA content in leaf.

Transmission Electron Microscopy was applied to visualize the cellular distribution and morphology of LDs in potato leaf tissues. Dissected tissues were submerged in fixative 2.5% glutaraldehyde and 4% paraformaldehyde in 0.1 M phosphate buffer, and post fixed with 1% osmium tetroxide for 2 h. Fixed samples were infiltrated and embedded in LR white resin after gradual ethanol dehydration. 70 nm ultrathin sections were prepared with a Leica EM UC6 Ultramicrotome (Leica Microsystems). Each section was stained with 2% UA for 15 and 5 min with lead citrate. The sections were examined with a Hitachi H7100 transmission electron microscope (Hitachi, Tokyo, Japan) at 75 kV accelerating voltage. Confocal scanning microscopy analysis was processed by using freshly sampled potato leaf at the senescent stage, as described in Vanhercke et al. (2018).

For the basic tuber physiological trait analysis, three healthy independent potato plants as biological replicates of WT, L3 and L5 respectively were measured. After harvesting all the mature tubers from an individual plant, the maximum tuber length was measured and divided into four groups according to the size (Size A, tuber length > 9 cm; Size B, 6 cm < tuber length < 9 cm; Size C, 3 cm < tuber length < 6 cm; Size D, tuber length < 3 cm). From each group, an intact tuber was selected then analyzed with the tuber water percentage (tuber water content/tuber fresh weight) and tuber density (fresh tuber weight/tuber volume). The tuber water content was calculated with the oven-drying method, and the tuber volume with the dewatering method.

GenStat 9.0 software was used to calculate the least significant difference (LSD) value of all data for multiple comparison.