Sterol Side Chain Reductase 2 Is a Key Enzyme in the Biosynthesis of Cholesterol, the Common Precursor of Toxic Steroidal Glycoalkaloids in Potato

Potatoes ( Solanum tuberosum ) contain a -solanine and a -chaconine, two well-known toxic steroidal glycoalkaloids (SGAs). Sprouts and green tubers accumulate especially high levels of SGAs. Although SGAs were proposed to be biosynthesized from cholesterol, the biosynthetic pathway for plant cholesterol is poorly understood. Here, we identify sterol side chain reductase 2 (SSR2) from potato as a key enzyme in the biosynthesis of cholesterol and related SGAs. Using in vitro enzyme activity assays, we determined that potato SSR2 (St SSR2) reduces desmosterol and cycloartenol to cholesterol and cycloartanol, respectively. These reduction steps are branch points in the biosynthetic pathways between C-24 alkylsterols and cholesterol in potato. Similar enzymatic results were also obtained from tomato SSR2. St SSR2 -silenced potatoes or St SSR2 -disrupted potato generated by targeted genome editing had signi ﬁ cantly lower levels of cholesterol and SGAs without affecting plant growth. Our results suggest that St SSR2 is a promising target gene for breeding potatoes with low SGA levels.


INTRODUCTION
Steroidal glycoalkaloids (SGAs) are famous as toxic compounds in solanaceous plants that include the world's major food crops such as potato (Solanum tuberosum) and tomato (Solanum lycopersicum) (Friedman, 2002(Friedman, , 2006. a-Chaconine and a-solanine account for most of the SGAs in cultivated potato and are especially abundant in sprouts, the green peel of tubers, and other aerial parts. To avoid postharvest increases in SGAs, harvested potatoes must be stored in the dark at low temperature (Kozukue and Mizuno, 1990). In potato breeding, introducing wild potato species as a genetic resource could increase the SGA level (Jacobsen and Rousselle, 1993). Thus, reducing the SGA level is an important target for potato breeding.
Feeding studies with isotope-labeled substrates suggested that SGAs are biosynthesized via cholesterol as a common precursor ( Figure 1) Hulpke, 1966, 1967;Heftmann et al., 1967). Cholesterol is converted to SGAs by a series of modifications that are predicted to involve oxidation, transamination and glycosylation (Ohyama et al., 2013). The recent releases of the potato (Xu et al., 2011) and tomato (Tomato Genome Consortium, 2012) genome sequences have accelerated the identification of genes encoding these enzymes in the SGA biosynthetic pathway (Moehs et al., 1997;McCue et al., 2005McCue et al., , 2006McCue et al., , 2007Itkin et al., 2011Itkin et al., , 2013. Modifying the SGA biosynthetic pathway by targeting potato genes encoding the oxidases or glycosyltransferases results in decreased levels of SGAs; however, the decrease in SGAs is associated with an increase in levels of intermediates between cholesterol and SGAs or unexpected derivatives of the intermediates (McCue et al., 2005(McCue et al., , 2006(McCue et al., , 2007Itkin et al., 2011Itkin et al., , 2013. This unintended metabolic change may pose a risk to potato consumers. As a more ideal way of engineering potatoes, reducing the supply of cholesterol, a common precursor for SGA biosynthesis, should result in lower levels of SGAs without a corresponding increase in levels of undesired intermediates between cholesterol and SGAs or their derivatives. However, the biosynthetic mechanism for cholesterol in plants has not been fully elucidated. Generally, plants produce C-24 alkylsterols including campesterol, a precursor of the phytohormone brassinolide ( Figure 1). The primary structural difference between the C-24 alkylsterols and cholesterol is the presence of an alkyl group at the C-24 position of the side chain (Benveniste, 1986). The C-24 alkylsterols have a methyl or an ethyl group at C-24 resulting from the addition of one or two methyl groups to D 24(25) -sterols, such as cycloartenol and desmosterol catalyzed by sterol methyltransferases (SMTs) (Husselstein et al.,1996;Shi et al., 1996;Bouvier-Navé et al., 1997;Grebenok et al., 1997). Diminuto/Dwarf1 (DWF1) (Klahre et al., 1998;Choe et al., 1999) of Arabidopsis thaliana has been proposed to reduce alkylidene groups generated by SMT reactions to alkyl groups ( Figure 1). The dwf1 mutant has a dwarf phenotype due to the lack of brassinosteroids (Kauschmann, et al., 1996). In the cholesterol biosynthetic pathway of animals including humans, the C-24 position of D 24(25) -sterols, such as desmosterol, is reduced by 24-dehydrocholesterol reductase (DHCR24) (Waterham et al., 2001), a homolog of DWF1. Although cholesterol is generally present only in trace amounts in plants, solanaceous plants accumulate considerable amounts of SGAs derived from cholesterol. To meet this high demand for cholesterol, these plants may have a cholesterol biosynthetic mechanism distinct from that for ordinary plant sterol biosynthesis. In this study, a gene encoding sterol side chain reductase 2 (SSR2) committed to cholesterol biosynthesis in plants was identified and used to engineer potatoes with lower levels of SGAs.

Potato and Tomato Have Two Types of Genes Encoding DHCR24 Homologs
Using the human DHCR24 amino acid sequence as a query, BLAST (W. Gish, http://blast.wustl.edu and http://www.advbiocomp.com/ blast/obsolete; Altschul et al., 1997) analysis against potato unigene databases from the Dana-Farber Cancer Institute Potato Gene Index (Quackenbush et al., 2001), and the Sol Genomics Network (Bombarely et al., 2011) allowed us to select two types of cDNA sequences designated as St SSR1 and St SSR2. In contrast, DWF1 is a single copy gene in Arabidopsis. We also found two fulllength cDNA sequences (Sl SSR1 and Sl SSR2) similar to DHCR24 from a tomato unigene database from the Sol Genomics Network. The nucleotide sequence identities between St SSR1 and St SSR2 and between Sl SSR1 and Sl SSR2 are 81 and 79%, respectively. Phylogenetic analysis of the coding sequences of DHCR24 homologous genes from plants ( Figure 2; Supplemental Data Set 1) indicates that SSR1 and SSR2 are paralogs in the Solanum genus. Although genes associated with SGA biosynthesis in potato and tomato were reported to be clustered on chromosomes 7 and 12 (Itkin et al., 2013), SSR genes in potato and tomato are positioned on chromosome 2 and are more than 10 Mbp away from each other (St SSR1, PGSC0003DMG400011801, and St SSR2, PGSC0003DMG400021142, in Xu et al., 2011;Sl SSR1, Solyc02g030170.2, and Sl SSR2, Solyc02g069490.2, in Tomato Genome Consortium, 2012). All four deduced proteins were predicted to have a flavin adenine dinucleotide binding domain and a transmembrane region by the PROSITE (Sigrist et al., 2002) and SOSUI (Hirokawa et al., 1998) programs, respectively (Supplemental Data Set 2).

In Vivo Enzymatic Activity Assays of SSRs
We presumed that one SSR would catalyze the reduction of D 24(25) -sterol involved in cholesterol biosynthesis and the other Plant C-24 alkylsterols and cholesterol, the common intermediates in steroidal glycoalkaloid biosynthesis, are biosynthesized from cycloartenol. Red arrows indicate reduction steps mediated by SSR2. Blue arrows indicate reduction steps mediated by SSR1. Open black arrows indicate transmethylation steps mediated by SMTs.

of 12
The Plant Cell SSR would mediate reduction of the D 24(28) -double bond of C-24 alkylsterol biosynthetic intermediates. To investigate the reduction activities of the four putative SSRs from potato and tomato, the genes were expressed in yeast strains T31 and T21 that have modified ergosterol biosynthesis ( Figure 3A; Supplemental Figure  1). The T31 strain accumulates a D 24(25) -sterol, desmosterol, and the T21 strain accumulates a precursor of campesterol, 24-methylenecholesterol. Ethyl acetate extracts of the yeast cells were analyzed by gas chromatography-mass spectrometry (GC-MS) and compared with standard samples (Supplemental Figure  2). The T31 cells expressing St SSR2 and Sl SSR2 accumulated remarkable amounts of cholesterol compared with cells expressing St SSR1 and Sl SSR1 ( Figure 3B). When T21 served as the host strain, the cells expressing St SSR1 and Sl SSR1 accumulated marked amounts of campesterol compared with the cells expressing St SSR2 and Sl SSR2. The T21 cells expressing St SSR2 and Sl SSR2 also accumulated cholesterol ( Figure 3C). In the T21 cells, 24-methylenecholesterol is thought to be biosynthesized via D 24 (25

In Vitro Enzymatic Activity Assays of SSRs
The in vivo cholesterol production by the T31 and T21 yeasts expressing SSR2 suggests that St SSR2 may be able to reduce a variety of the D 24(25) -sterols, including lanosterol, the first D 24 (25) -sterol in yeast, and desmosterol (Supplemental Figure 1). In plants, cycloartenol is the first D 24(25) -sterol. If cycloartenol is reduced to cycloartanol by SSR2, the biosynthetic pathway would be efficiently redirected toward cholesterol ( Figure 1) (Diener et al., 2000;Schaeffer et al., 2000;Sitbon and Jonsson 2001;Holmberg et al., 2002;Arnqvist et al., 2003). We conducted in vitro assays to examine whether SSR2 has cycloartenol reductase activity and to compare the substrate specificities of SSR1 and SSR2 ( Figure 4). St SSR1, St SSR2, Sl SSR1, and Sl SSR2 were heterologously expressed in an erg4 [deficient in sterol D 24(28) reductase gene] yeast strain to avoid contamination of the endogenous sterol D 24(28) reductase activity into the in vitro SSR enzymatic assays. Homogenates of the erg4 yeast cells expressing St SSR1, St SSR2, Sl SSR1, or Sl SSR2 were incubated with cycloartenol, desmosterol, or 24-methylenecholesterol, and the n-hexane extracts of the incubated samples were analyzed by GC-MS. St SSR2 and Sl SSR2 efficiently converted cycloartenol to the D 24(25) reduction product, cycloartanol, and the reaction was NADPH dependent ( Figure 4B). St SSR2 and Sl SSR2 also converted desmosterol to cholesterol in an NADPHdependent manner ( Figure 4C), a result consistent with the in vivo assay that indicated the production of cholesterol in yeast cells expressing SSR2 ( Figure 3B). On the other hand, St SSR1 and Sl SSR1 efficiently reduced 24-methylenecholesterol to campesterol. The weak activity was also shown by St SSR2 and   Sl SSR2 ( Figure 4D). St SSR1 and Sl SSR1 also have very weak reductase activity toward cycloartenol and desmosterol (Supplemental Figure 3). These results indicate again that SSR1 and SSR2 have different substrate specificities. In addition, the D 24(25) reduction activity against cycloartenol of SSR2 suggests that SSR2 effectively directs the sterol biosynthetic flow toward the cholesterol-SGA pathway in potato and tomato.

Analyses of SSR2-Silenced Potato Plants
To verify the contribution of St SSR2 in the biosyntheses of cholesterol and SGAs in potato plants, we constructed St SSR2silenced potato plants by transformation with an RNA interference (RNAi) vector (Supplemental Figure 4A). The RNAi vector was constructed from a 269-bp fragment of St SSR2 cDNA that did not match over 21 consecutive nucleotides with St SSR1. Four independent St SSR2-RNAi potato plant lines (#2, #10, #11, and #12 in Figures 5A to 5D) were identified that had remarkably reduced St SSR2 transcript levels (Supplemental Figure 4B). In these selected RNAi lines, the levels of predominant SGAs (a-solanine and a-chaconine) in the stems of in vitro-grown plants ( Figure 5A) and tuber peels ( Figure 5B) were significantly lower by 10% or more compared with the nontransformant. The cholesterol content was also significantly lower in the St SSR2-RNAi lines ( Figure 5C). The levels of campesterol and stigmasterol were unchanged despite the decrease in b-sitosterol content ( Figure 5C). The dwf1 mutant is well known for a dwarfed phenotype caused by a deficiency in brassinosteroids (Kauschmann, et al., 1996); however, the four St SSR2-RNAi lines showed normal growth and harvest yield ( Figure 5D; Supplemental Figure  4C). We also analyzed levels of the major biologically active brassinosteroids, brassinolide and castasterone (Kim et al., 2005;Nomura et al., 2005), in two St SSR2-RNAi lines and the nontransformant. The levels of castasterone, which were under one thousandth of the cholesterol level in the nontransformant, were similar in the two transformants, and brassinolide levels were too low to quantify (Supplemental Table 1). These results indicate that St SSR2 is devoted to the production of cholesterol and SGAs in potato plants. The data also suggest that St SSR1 is responsible for C-24 alkylsterols and brassinolide biosyntheses. Levels of C-24 alkylsterol intermediates, such as 14a-methylcholesta-8,24-dien-3b-ol, 4a,14a-dimethylcholesta-8,24-dien-3b-ol, and isofucosterol, in the St SSR2-silenced lines were higher relative to those of the non-transformant (Supplemental Figures 4D and 4E). Similar results were also obtained from Sl SSR2-silenced tomatoes (Supplemental Figure 5).

DISCUSSION
The data presented here show that St SSR2 is a key enzyme in the biosynthesis of toxic SGAs derived from cholesterol in potato. The methyl addition at C-24 of cycloartenol is the initial step in C-24 alkylsterol biosynthesis in plants. The presence of the alkyl group at C-24 that originates from this reaction is the structural characteristic differentiating C-24 alkylsterol from cholesterol. Therefore, cholesterol levels in plants are hypothesized to be modulated by SMT; this hypothesis is supported by the lower cholesterol level in transgenic plants that overexpress SMT and the higher cholesterol level in SMT knockout plants (Diener et al., 2000;Schaeffer et al., 2000;Sitbon and Jonsson 2001;Holmberg et al., 2002;Arnqvist et al., 2003). However, the considerable amounts of cholesterol and its derivatives, SGAs, in solanaceous plants have suggested that these plants have a mechanism to produce cholesterol that is distinct from other plants. The in vitro and in vivo analyses of both potato and tomato enzymes showed that solanaceous plants have two types of DWF1 homologs, SSR1 and SSR2, with different substrate preferences. SSR1 has D 24(28) reduction activity required for C-24 alkylsterol biosynthesis. On the other hand, SSR2 has D 24(25) reduction activity, which is essential for biosynthesis of cholesterol, an SGA precursor, as well as weak D 24(28) reduction activity. These observations suggest that solanaceous plants have obtained an additional DWF1 homolog and developed it into SSR2 to meet the high demand for cholesterol. The much reduced levels of cholesterol and SGAs in two distinct loss-of-function potato plants, i.e., silencing by RNAi and disruption by TALEN, strongly support the identification of SSR2 as a key enzyme in their biosyntheses. The increases of some C-24 alkylsterols and their intermediates in the St SSR2-silenced and St SSR2-disrupted potatoes were probably due to switching of the biosynthetic flow from the cholesterol-SGA pathway to the C-24 alkylsterol pathway caused by lowered or abolished SSR2 activity. The different effects of St SSR2 silencing and St SSR2 disruption were observed in the levels of campesterol, b-sitosterol, and isofucosterol produced. These inconsistencies can be attributed to an off-target effect of St SSR2-RNAi on St SSR1 transcription in the silenced lines (Supplemental Figure 4B). The TALEN-disrupted line #271-19 is considered to have only null or strong alleles of the St SSR2 locus because there was no or little intact genomic DNA for this locus (Supplemental Figure 6C). The residual SGA accumulated by this line ( Figure 5E) might result from the weak D 24(25) reduction activity of St SSR1 toward cycloartenol or desmosterol (Supplemental Figure 3), rather than leaky activity of St SSR2. Although roles for cholesterol and SGAs in plant growth and development have not been fully elucidated, the normal growth and development of St SSR2-RNAi and St SSR2-TALEN potatoes also suggests that cholesterol and SGAs may not be essential for the normal growth of potato (at least) under unstressed growth conditions.
In addition to SGAs in solanaceous plants, spirostan and/or furostan saponins such as dioscin in yam (Dioscorea spp; order Liliales) and trigoneosides in fenugreek (Trigonella spp; order Fabales) could also be biosynthesized via cholesterol as an intermediate (Bennett and Heftmann, 1965;Hardman and Fazli, 1972). Whether these plants have any genes homologous to DHCR24 is unknown. The paralogous relationship between SSR1 and SSR2 in the Solanum genus suggests that cholesterol biosynthetic mechanisms in plants were developed by convergent evolution in each order, family, or genus. That is, the cholesterol biosynthetic pathway in other plants may differ from that in Solanum.
In this study, we identified SSR2 as an enzyme that is responsible for the initial reaction from C-24 alkylsterol biosynthesis to cholesterol biosynthesis on the pathway to the production of toxic SGAs in solanaceous plants. In addition, the lower level of SGAs in the St SSR2-disrupted potato by targeted genome editing with TALEN is evidence that the St SSR2 gene identified in this study is an excellent target for breeding potatoes with decreased SGA levels. Breeding an St SSR2-knockout potato may also be possible by targeting induced local lesions in genomes (TILLING) (Muth et al., 2008); however, such methods that involve conventional mutagenesis would not rapidly generate new cultivars of important polyploid food crops, such as tetraploid potato. Our results from the St SSR2-TALEN transformants demonstrated that targeted genome editing with TALEN is able to generate potatoes with no remaining intact allele (Supplemental Figures 8A to 8C). An St SSR2-knockout potato without the transgene will be obtained by self-crossing the transformant (Supplemental Figure 8D). In the engineered potatoes, the absence of any increase in levels of intermediates between cholesterol and SGAs or their unexpected derivatives is expected because of the reduced supply of cholesterol as a common precursor of SGAs. In our analyses, there was no marked increase in the levels of intermediates (Supplemental Figure 7A). Further study of the SSR2-knockout potato will increase our understanding of the roles of cholesterol and SGAs in potato. Furthermore, as one of the possible gain-of-function applications of SSR2, overexpression of SSR2 in plants producing cholesterolderived saponins could increase the production of such saponins, i.e., dioscin, as a raw material for steroidal drugs.

Cloning of SSR cDNAs
The tomato cDNA template was prepared from mRNA isolated from a flowering-stage leaf of Solanum lycopersicum cv Micro-Tom using the SV Total RNA Isolation System (Promega) and SuperScript First-Strand Synthesis System for RT-PCR (Life Technologies

Yeast in Vivo Assays
Yeast transformation was performed using a Frozen-EZ Yeast Transformation II Kit (Zymo Research) or a Gene Pulser Xcell electroporation system (Bio-Rad) according to the manufacturers' protocols. Yeast strains T21 and T31 described above were transformed with St SSR1, St SSR2, Sl SSR1, and Sl SSR2 cDNAs cloned into pYES-DEST52 or pYES2. T21 and T31 transformed with intact pYES2 were used as empty vector controls. The transformants were cultured in yeast Synthetic Minimum Medium containing 2% glucose supplemented with -His/-Ura DO supplement (Takara Bio) for 1 d at 30°C. The cells were transferred to new medium of the same composition except for containing 2% (w/v) galactose instead of glucose to induce the recombinant SSRs and cultured for two more days. The harvested cells were lyophilized and extracted with ethyl acetate, and the extracts were analyzed by GC-MS after trimethylsilylation.
In Vitro Assay S. cerevisiae BY4742 erg4 (Open Biosystems) (Kelly et al., 2001) was transformed with St SSR1, St SSR2, Sl SSR1, and Sl SSR2 cDNAs cloned into pYES-DEST52 or pYES2. BY4742 erg4 transformed with intact pYES2 was used as an empty vector control. The transformants were cultured in yeast Synthetic Minimal Medium containing 2% glucose supplemented with -Ura DO supplement (Takara Bio) for 1 d at 30°C. The cells were transferred to new medium of the same composition except for containing 2% galactose instead of glucose to induce the recombinant SSRs and cultured for one more day. The in vitro enzyme activity assays using yeast homogenates were performed as previously reported (Waterham et al., 2001;Ciufo et al., 2011). Briefly, the harvested cells were suspended in 20 mM Tris-HCl, pH 7.5, containing 50 mM NaCl and disrupted with a Beadbeater (BioSpec Products) containing 0.5-mm glass beads. After removing the glass beads and cell debris by centrifugation at 3000g for 5 min, protein concentrations in the homogenates were measured with Quick Start Bradford Protein Assay Kit 2 (Bio-Rad). The assay mixture contained 100 mM Tris-HCl, pH 7.23, 1 mM DTT, 20 µM flavin adenine dinucleotide, 2 mM NADPH, 0.5 mg mL 21 BSA, 0.56 mg mL 21 homogenate protein, and 168 µM sterol substrate. The substrate was added from a 420 µM stock solution in 100 mM Tris-HCl, pH 7.23, containing 1.25% (w/v) methyl-b-cyclodextrin. After incubation for 24 h at 30°C, samples were saponified with KOH and ethanol for 1 h at 80°C and extracted with n-hexane. The extracts were washed with water and analyzed by GC-MS after trimethylsilylation.

St SSR2-Silenced Potatoes
A 269-bp fragment of St SSR2 cDNA that did not match over 21 consecutive nucleotides with St SSR1 was PCR amplified using the primer set 59-GAGCTCTAGACCCTAGGAGGAAGATCCAG-39/59-GGAT-CCATATGCGTTTCTCATTCCAACAACA-39. An RNAi binary vector pKT251 targeting the St SSR2 gene was constructed from the binary vector pKT11 (Umemoto et al., 2001) by locating two St SSR2 fragments in opposite orientations interposing the third intron of the Arabidopsis thaliana At4g14210 gene under the control of the cauliflower mosaic virus 35S (CaMV35S) promoter in the T-DNA region (Supplemental Figure 4A). The RNAi binary vector pKT251 was transformed into Agrobacterium tumefaciens GV3110. Potatoes (S. tuberosum cv Sassy) were transformed using Agrobacterium GV3110 cells with pKT251 as previously reported (Monnma, 1990). In vitro-grown plants were cultured at 20°C under a 16-hlight/8-h-dark cycle. Thirty-nine transformants were individually selected by genomic PCR of the shoots with the primer set 59-TAAAGCAC GAGGAAGCGGT-39/59-GCACAACAGACAATCGGCT-39 targeting the kanamycin resistance gene on the T-DNA region integrated into the potato genome. RT-PCR analyses of St SSR1 and St SSR2 were Cholesterol Biosynthesis in Potato 7 of 12 performed using the primer sets 59-TTAGGTTTTTCTTTGGATGGGC-39/ 59-TCACCCTGCCTCTTGTGCAG-39 and 59-CTCTGCTCAAAGCCACA-CAA-39/59-TCAATTCGCAGGTTCATCAG-39 and total RNA prepared from stems of five independent lines of in vitro-cultured plants, #2, #10, #11, #12 and #9, and nontransformants. As a control, we used the potato elongation factor 1a gene (EF1a) and the primer pair 59-ATTGGAAACG-GATATGCTCCA-39/59-TCCTTACCTGAACGCCTGTCA-39 (Nicot et al., 2005). For the quantitative comparison of the transcripts, the cycles of PCR reactions were optimized to make sure of both clear visualization of the amplified products on an agarose gel and amplification in the exponential phase. The optimal cycles were 20 and 25 for St SSR1, 25 for St SSR2, and 20 for EF1a (Supplemental Figure 4B).

Sl SSR2-Silenced Tomatoes
A fragment of Sl SSR2 cDNA that did not match over consecutive 21 nucleotides with Sl SSR1 was PCR amplified using the primer set 59-GAGCTCTAGATGTCGGATGCTAAGGCC-39/59-GGATCCATATGCA-TTCCTCTGGCCAAG-39. An RNAi binary vector pKT262 targeting Sl SSR2 was constructed with the Sl SSR2 fragment in a way similar to the construction of pKT251. Tomatoes (S. lycopersicum cv Micro-Tom) were transformed using Agrobacterium GV3110 cells with pKT262 as previously reported (Sun et al., 2006). Integration of the T-DNA region into the tomato genomic DNA was investigated by genomic PCR of the tomato leaves targeting the kanamycin resistance gene on the T-DNA region integrated into the tomato genome.
The TALEN vector pKT271 was transformed into Agrobacterium GV3110. Potatoes (S. tuberosum cv Sassy) were transformed using Agrobacterium GV3110 cells with pKT271 as previously reported (Monnma, 1990). Twenty-nine transformants were individually selected by genomic PCR of the shoots with the primer set 59-TAAAGCACGAGGAAGCGGT-39/ 59-GCACAACAGACAATCGGCT-39 targeting the kanamycin resistance gene on the T-DNA region integrated into the potato genome. The selected transformants were cultured in plant boxes (Murashige and Skoog medium [Murashige and Skoog, 1962] containing 3% sucrose, 100 mg L 21 carbenicillin, and 0.8% agar) at 20°C under a 16-h-light/8-hdark cycle.
To induce TALEN expression, 6-to 8-cm-tall subcultured shoots grown in plant boxes were dipped in a solution containing 10 mM 17b-estradiol and 0.02% Silwet L-77 and then transferred to soil. They were grown at 23°C under a 16-h-light/8-h-dark cycle.
Two weeks after the transgenic potatoes were transferred to soil, genomic DNA was extracted from TALEN-induced transgenic potato plants to detect somatic mutants in the target site. Regions surrounding the St SSR2 target sites were amplified by PCR using the primer set 59-TGTTCTCTGACACTGTTGTAGCACT-39/59-TCGAAGCATACATACCG-GTCATCAT-39 and PrimeSTAR Max DNA polymerase (Takara Bio). Amplified products were separated on 5% polyacrylamide gels (Ota et al., 2013). Two samples (lines #271-5 and #271-19) had extra bands probably due to the formation of heteroduplexes within the target sites (Supplemental Figure 6C). The PCR products were cloned into the pCR4-Blunt-TOPO (Life Technologies) vector, and 8 and 15 clones were randomly selected for sequencing from lines #271-5 and #271-19, respectively. The presence of targeted deletions or insertions was confirmed in two of eight sequenced clones from line #271-5 and in all 15 sequenced clones from line #271-19 (Supplemental Figures 6D and 6E). The corresponding region of the St SSR1 gene was also PCR amplified from line #271-19 (with the primer set 59-TGTTCTCAGACACTG-TTGTGTCATA-39/59-TTGAAGCATATCTACCAGTCATGCA-39), as the most homologous potential off-target site (Supplemental Figure 6B); however, no mutations were detected in any of the eight sequenced clones.
The levels of sterols and SGAs in TALEN-induced transgenic potato lines were analyzed 6 to 7 weeks after the transgenic potatoes were transferred to soil as described below.

Construction of SSA Reporter Plasmids and the SSA Assay Using HEK293T Cells
Construction of SSA reporter plasmids and the SSA assay were performed as previously described (Sakuma et al., 2013). SSA reporter plasmids for SSR2_A, SSR2_B, and SSR2_C TALENs and pGL4-SSA-SSR2_A, pGL4-SSA-SSR2_B, and pGL4-SSA-SSR2_C were constructed as follows: sense and antisense oligonucleotides (Supplemental Table 4) were annealed and inserted between the BsaI sites of the pGL4-SSA vector. The SSA reporter plasmid for the negative control samples, pGL4-SSA-ZFA36, was constructed as previously described (Ochiai et al., 2010). pGL4-SSA- HPRT1_B-15 (Sakuma et al., 2013) was used as the SSA reporter plasmid for the positive control TALEN.
For the TALEN-positive control, TALEN expression vectors, pcDNA-SSA-HPRT1_B_L and pcDNA-SSA-HPRT1_B_R, were constructed as 8 of 12 The Plant Cell previously described (Sakuma et al., 2013) and cotransfected with reporter plasmids as described herein.

GC-MS Analysis
GC-MS analyses were conducted with the same conditions as described before (Choi et al., 2014). MS spectra of trimethylsilylated authentic standards are shown in Supplemental Figure 2. 14a- 4a, are tentatively identified compounds that were deduced by comparison of their mass spectra (Supplemental Figure 9) with published spectra and the relative retention times (Kornfeldt and Croon, 1981;Berman et al., 1986;Akihisa et al., 1988;Narumi et al., 2001;Wretensjö and Karlberg, 2002). The endogenous levels of sterols in potato and tomato plants were calculated by comparing the peak area values of molecular ions (cholesterol, 14a-methylcholesta-8,24-dien-3b-ol, 24-methylenecholesterol, 4a,14a-dimethylcholesta-8,24-dien-3b-ol, campesterol, stigmasterol, and b-sitosterol)  Extractions and LC-MS analyses of the plant materials were performed with the same method as previously described (Ohyama et al., 2013). Quantifications of a-solanine and a-chaconine were calculated from the ratio of peak area at m/z 868 and 852, respectively, from positive ion scans using a calibration curve of authentic a-solanine (with a coefficient-of determination: r 2 > 0.992).

LC-MS Analysis of a-Tomatine
The sample preparation procedures and LC-MS conditions were identical to those for LC-MS analysis of a-chaconine and a-solanine in St SSR2disrupted potatoes described above. Quantification of a-tomatine was calculated from the ratio of peak areas at m/z 1034 from positive ion scans using a calibration curve of authentic a-tomatine (with a coefficient of determination: r 2 > 0.998).

Quantification of Brassinosteroids
Frozen potato leaves (;1 g fresh weight) were lyophilized and crushed to a fine powder with 10-mm ceramic beads by vortex and then soaked in 10 volume of extraction solvent (80% acetonitrile, 1% acetic acid) with stable isotope-labeled castasterone and brassinolide. After centrifugation, the supernatant was concentrated and reconstituted with 1% acetic acid. To remove interfering compounds, the extract was passed through an Oasis HLB column (Waters), Oasis MCX column, and Oasis WAX column equilibrated with 1% acetic acid, and brassinosteroids-containing fraction was eluted with 80% acetonitrile and 1% acetic acid, 80% acetonitrile and 1% acetic acid, and 80% acetonitrile, respectively. In each column step, the eluate was evaporated and reconstituted with 50% methanol. The castasterone and brassinolide were measured with UHPLC-Q-Exactive (Thermo Fisher Scientific) with an ODS column (AQUITY UPLC BEH C 18 , 1.7 mm, 2.1 3 100 mm; Waters). Brassinosteroids were separated at a flow rate of 0.25 mL min 21 with linear gradients of solvent A (0.05% formic acid) and solvent B (0.05% formic acid in acetonitrile) set according to the following profile: 0 min, 90% A + 10% B; 20 min, 5% A + 95% B. Data were processed by Xcalibur software (version 2.2; Thermo Fisher Scientific).

Statistical Analysis
Dunnett's tests (Dunnett, 1955) were performed using the R Project for Statistical Computing software (R Core Team, 2013) with the multcomp package (Hothorn et al., 2008).

Phylogenetic Analysis
Sequence alignments (Supplemental Data Set 1) were constructed in MEGA5 (Tamura et al., 2011) using ClustalW (Thompson et al., 1994). The maximum likelihood tree was inferred in MEGA5 (Tamura et al., 2011) using Kimura's two-parameter model (Kimura, 1980) with a discrete gamma distribution (five categories) and the partial deletion option applied for gaps/missing-data treatment. Bootstrap values were performed with 1000 replications (Felsenstein, 1985). GenBank/EMBL/DDBJ accession numbers for the nucleotide sequences of the DHCR24 homologs are shown in Figure 2.

Multiple Alignments of Deduced SSR Amino Acid Sequences
Multiple alignments of deduced SSR amino acid sequences were constructed in GENETYX-MAC version 16 (Genetyx) with ClustalW2 (Larkin et al., 2007).

Accession Numbers
The sequences reported in this article have been deposited in the GenBank/EMBL/DDBJ database under the following accession numbers: AB839749 (St SSR1), AB839750 (St SSR2), AB839356 (Sl SSR1), AB839357 (Sl SSR2), and AB839751 (St DWF5). The accession numbers of the pER8 and potato EF1a in the GenBank/EMBL/DDBJ database are AF309825 and AB061263, respectively.

Supplemental Data
The following materials are available in the online version of this article.