The specific DNA fragment selected for suppressing the expression of both copies of CLCNt2 (CLCNt2-S and CLCNt2-T) spans the exon–intron borders between coding exons 4 and 5: 5′-gtcatcatcaggtgtgtgtgctgctttccgttctccagtaggtggtgtcctatttgctttagaggaagtggcaacatggtggagaagtgcactcctctggagaactttcttcagcacggcagttgtggtggtgatactgagggccttcattgaatactgcaaatctggcaactgtggactttttggaagaggagggcttatcatgtttgatgtgagtggtgtcagtgttagctaccatgttgtggacatcatccctgttgtagtgattggaatcataggcggacttttgggaagcctctacaatcatgtcctccacaaaattctgaggctctacaatctgatcaacgagaagggaaaactacataaggttcttctcgctctgagtgtctcccttttcacctccattt-3′.

This insert was cloned in both sense and antisense orientations (separated by an intron) between the strong constitutive Mirabilis Mosaic Virus (MMV) promoter and the 3′ nos terminator sequence of the nopaline synthase gene of Agrobacterium tumefaciens (Cheng et al., 1997) within a PMI-proprietary binary vector (WO2012098111).

For the suppressed expression, the BU tobacco variety TN90e4e5 was transformed using standard Agrobacterium-mediated transformation protocols (Horsch et al., 1985). TN90e4e5 represents a selection from an ethylmethane sulfonate-mutagenized BU population that contains KO mutations in CYP82E4 and CYP82E5v2 (Lewis et al., 2010) to prevent nornicotine production and, therefore, NNN accumulation. Minimizing the nornicotine content in tobacco is highly desirable, because this compound serves as the direct precursor in the synthesis of NNN. Nornicotine is likely produced almost entirely via the N-demethylation of nicotine, in a process called nicotine conversion that is catalyzed by the enzyme nicotine N-demethylase (NND). Previous studies have identified CYP82E4 as the specific NND gene responsible for the unstable conversion phenomenon, and CYP82E5v2 as a putative minor NND gene (Lewis et al., 2010). Therefore, we used TN90e4e5 as background to silence the expression of CLCNt2 to confirm specifically the impact of potential reduced nitrate on the accumulation of all TSNAs, particularly NNN which can be also generated in part by the accumulation of nornicotine. Ten independent T0 plants from each construct were assayed by quantitative reverse transcription PCR (qRT-PCR) to select transgenic lines. T1 seeds were harvested from three independent T0 plants that exhibited the strongest CLCNt2 silencing. T1 plants were grown in the greenhouse and selected for the presence of construct insertion in the genomic DNA by PCR with the following primers (5′–3′): MMV-F (gacgtctaatcccaacttcgtc) and IPMS2-R (gacgtctaatcccaacttcgtc).

A multiple sequence alignment was generated using Clustal Omega (version 1.2.4, with the options --full--full-iter parameters), and a phylogenetic tree was constructed from these multiple sequence alignments by using FastTree (version 2.1.10, with the options-gamma-pseudo-spr4-mlacc2-slownni). The resulting phylogenetic tree was re-rooted on the longest branch and ordered on the leaf labels, while preserving the tree topology by using the nw_re-root and nw_order programs from newick_utils (version 1.6). The phylogenetic tree was visualized by using ete3 (version 3.1.1). Labels on the nodes are local support values calculated with the Shimodaira–Hasegawa test. The sequences mentioned are available at https://www.solgenomics.net/organism/Nicotiana_tabacum/genome.

Specific primers were designed to amplify DNA fragments by using genomic DNA: CLCNt2-S (forward primer: aaaacatggctggttacaaacttc; reverse primer: tctaaagcaaataggacaccaccta) and CLCNt2-T (forward primer: agtgacagcctcccttatcaatct; reverse primer: cacttcctctaaagcaaataggaca). As internal control (house-keeping gene), tubulin (TUB) transcripts were amplified using forward primer atttgttgactggtgcccaac and reverse primer tcttcatcgtcaacttcagca. PCR amplification of genomic TUB, CLCNt2-S, and CLCNt2-T fragments was performed by using leaf DNA as matrix template isolated from leaves of the three main N. tabacum types BU (TN90), FC (K326), and OR (Basma Xanthi) and from the tobacco ancestors N. sylvestris and N. tomentosiformis cultivated in the greenhouse using DNeasy Plant Mini Kit (Qiagen, Hilden, Germany).

Burley variety TN90 plants were grown in a Swiss field. After flowering, the petal, sepal, immature flower, stem, root, and upper, middle, and lower leaves were collected in triplicate and immediately frozen in liquid nitrogen. Total RNA was isolated from each tissue sample (n = 3) by using the RNeasy mini kit (Qiagen, Hilden, Germany). Before cDNA library construction, RNA purity and integrity were checked using the Nanodrop device (Qiagen, Hilden, Germany). mRNA was enriched from total RNA by using a polyA column to generate cDNA libraries for sequencing. All libraries were sequenced on an Illumina HiSeq-2500 sequencer (Illumina, San Diego, CA, United States) by using version 3 chemistry and flow cells with runs of 2 × 100 bases. The reads were cleaned using Trimmomatic (version .32; Bolger et al., 2014) and aligned to the genome using Hisat2 (version 2.0.1 beta; Kim et al., 2015). Cuffdiff2 (version 2.2.1; Trapnell et al., 2013) was used for calculating the FPKM values based on the aligned reads. The transcriptomes were derived by mapping RNA-Seq reads from each tissue to the following tobacco reference genome.¹ For each plant, three biological replicates were obtained for each tissue. Previously published gene models were used as the basis for differential gene expression analysis (Sierro et al., 2014).

The effectiveness of the RNAi constructs in silencing CLCNt2 expression was assessed by qRT-PCR in a greenhouse environment. RNA was isolated from transgenic plants from each independent transformation event and their corresponding control plants, and qRT-PCR was performed using primers (5′ to 3′) F1 (tcccaacatgtacggagcaa) and R1 (aactgctgcaatgcttccaa) and an internal probe (acactttttgtcaagatc). To measure isoform-specific expression of the two CLCNt2 isoforms in different tissues of the greenhouse-grown TN90 plants, the following primer pairs were used: Nt2S-F (cttttgggaagcctctacaatca) and Nt2S-R (agtaccaggacaagacccgg) for CLCNt2-S; and Nt2T-F (cttttgggaagcctctacaattg) and Nt2T-R (agtgccaggacaagaccctt) for CLCNt2-T. The qRT-PCR reactions which were normalized against a tobacco actin gene by using the following primers: NtACT-F (cctcacagaagctcctcttaatc) and NtACT-R (acagcctgaatggcgatatac). Each PCR well contained .01 μM of forward and reverse primers, 2 μl of cDNA, and 5 μl of iTaq Universal SYBR Green Supermix (Bio-rad) per a 10-μl reaction mix. The reaction mix was heated to 95°C for 3 min and then amplified over 40 cycles of 95°C for 5 s and 62°C for 15 s.

All field evaluations were conducted at Clayton (NC, United States) using a randomized complete block design with individual plants as the experimental units (39–50 plants per genotype). Agronomic practices typical for BU tobacco production, including topping, were employed. The plants grew for 95 ± 9 days from transplant to harvest time, after which they were cured for 56 ± 10 days. For field-grown samples, leaf tissues were analyzed for nitrate content at the North Carolina State University Tobacco Analytical Service Laboratory as previously described (Lu et al., 2021).

Nicotine, nornicotine, anatabine, and anabasine levels in cured upper-stalk leaf samples were quantitated using a Perkin–Elmer AutosystemXL Gas Chromatograph in accordance with previously established protocols (Jack et al., 2007). Total alkaloid content was calculated as the sum of nicotine, nornicotine, anatabine, and anabasine levels. NNN, NNK, NAT, and NAB were quantified in accordance with “Method 1” of Morgan et al. (2004). Total TSNA levels represent the sum of NNN, NNK, NAT, and NAB levels.

Test cigarettes were handmade and filled with the electric cigarette injector “Powermatic 2”² by using standard cigarette tubes. The weight of the cut filler was varied (average, 800 mg) to obtain a resistance-to-draw corresponding to 120–140 mm water gauge. The cigarettes were subjected to machine smoking as outlined by Health Canada Method T-115 (1999). Mainstream smoke was collected on Cambridge filter pads and extracted with 100 mM ammonium acetate containing D4-NNN and D4-NNK standards. The extracts were analyzed by LC–MS/MS, liquid chromatography (LC) tandem mass spectrometry (MS), coupled with ESI and quantitated using D4-NNN and D4-NNK internal standards.

The CRISPR–Cas9 vector pGREB31 (Addgene) was used for targeted mutagenesis of the CLCNt2 genes. The backbone of pGREB31 is the CAMBIA vector pCAMBIA1300, which has the hptII gene as a selectable marker (hygromycin resistance). After cloning the sequences specifying the desired target site (5′-AGCTGTTGTGAACTATATCG-3′) into the sgRNA cassette of pGREB31, the vector was introduced into Agrobacterium strain LBA4404, and T0 plants (TN90) were generated using standard transformation protocols (Horsch et al., 1985). For such KO mutations, we used a TN90 as background, instead of TN90e4e5, to measure the impact of the lack CLCNt2 copies in a tobacco type used in agronomy, the TN90e4e5 plants being not implemented in the field to date. Nuclease-induced mutations were screened by PCR amplification by using primers flanking the target site, followed by DNA sequence analysis of the amplification products. The primers used for amplifying the CLCNt2-S isoform in order to identify CRISPR-induced mutations were Fw 5′-tggagctaaggtctcccacatc-3′ and Rev 5′-ccctgcagcagtaggtgcaaag-3′. The primers used for amplifying the CLCNt2-T isoform were Fw 5′-tggagctaaggtctcccatatt-3′ and Rev 5′-ccctgcagcagtaggcgcaaaa-3′.

Descriptive statistics, such as mean, minimum, maximum, and variance values, were computed per location and per genotype for all compounds investigated in this study. Potential outliers were identified and removed upstream the statistical modeling. One-way fixed effects ANOVA was performed to assess the variation in the different compounds across the different genotypes. The restricted maximum likelihood (REML) objective function was fit with the Newton–Raphson optimizer to obtain the statistical results. For each genotype tested, the genotypic means and the differences between genotypic means (along with their respective CIs) were calculated for each compound. Then, the relative variations of individual compounds were obtained for all comparisons of interest. Relative variation was defined as the effect of a compound concentration in a plant of a given genotype compared with that in a wildtype (WT) plant. Homogeneity of variances across the different genotypes was checked using the Akaike Information Criterion (AIC), and, when necessary, an additional residual variance was added into the linear model with Satterthwaite approximation for degrees of freedom. SAS software 9.4 was used for the statistical treatments, including Procedure MIXED for the computation aspects. R software 4.0.2 and ggplot2 library were used together for producing the charts.

Considering the vacuole as the main storage compartment for nitrate (Martinoia et al., 1981; Granstedt and Huffaker, 1982; van der Leij et al., 1998), and the AtCLCa (De Angeli et al., 2006) and AtCLCb (von der Fecht-Bartenbach et al., 2010) proteins as the tonoplast nitrate/proton antiporters involved in nitrate accumulation, our goal was to measure the impact of silencing or knocking out the tobacco orthologs of AtCLCa-b genes on nitrate storage in tobacco leaves. Using tobacco genome resources (Sierro et al., 2014), we were able to confirm that 17 genes encoding CLC proteins are located on the genome, as recently reported by Zhang et al. (2018). Ten NtCLC-like gene products cluster within the subfamily of Cl⁻/H⁺ antiporters, and the remaining seven proteins belong to the subfamily of Cl-channels.

A phylogenetic tree additionally populated with the seven and nine CLC proteins found in Arabidopsis thaliana (Marmagne et al., 2007) and Solanum lycopersicum, respectively, is shown in Figure 2A. This allowed us to assign a putative function to the tobacco CLC gene products, principally by sequence analogy subclustering with A. thaliana proteins. Eight of the tomato CLCs have duplicated copies in tobacco, one originating from each of the respective ancestors of N. sylvestris and N. tomentosiformis. The exceptions are Solyc02g094060, which has no homolog in tobacco, and Solyc01g112300, which clusters with the tobacco proteins NtCLCeA-S, NtCLCeB-S, and NtCLCe-T (Figure 2A). Interestingly, the Arabidopsis nitrate/proton vacuolar antiporters AtCLCa and AtCLCb (De Angeli et al., 2006; von der Fecht-Bartenbach et al., 2010) also have two orthologous proteins in tomato, Solyc02g0206880 and Solyc02g094060. Only the former isoform, however, appears to have been duplicated in tobacco. CLCNt2-S (N. sylvestris-related copy) and CLCNt2-T (N. tomentosiformis-related copy) correspond to NtCLC6 and NtCLC1, respectively, in the report by Zhang et al. (2018).

To confirm that the CLCNt2-S and CLCNt2-T genes are present in all three major commercial tobacco types—BU, FC, and OR—and in the respective N. sylvestris and N. tomentosiformis tobacco species, we performed PCR amplification of genomic DNA by using specific primers. The data confirm the presence of the N. sylvestris and N. tomentosiformis copies in the three major commercial types of tobacco (Figure 2B). RNA-seq data of the field-grown BU variety TN90 suggests that the CLCNt2-S gene is more highly expressed than CLCNt2-T in the main organs of the plant, with the highest expression in the petals and stem (Figure 2C). The results of qRT-PCR assays of the greenhouse-grown TN90 plants also support the conclusion that CLCNt2-S is the more highly expressed of the two genes (Figure 2C), despite the differential not appearing as clear as observed with RNA-seq data.

We used a genetic approach to determine the impact of CLCNt2-S and CLCNt2-T on leaf nitrate content. For this purpose, we generated an RNAi construct to suppress the expression of CLCNt2-S and CLCNt2-T by incorporating a 407-bp DNA fragment exhibiting 98% identity to both CLCNt2 sequences. The resulting RNAi constructs were expressed under the control of the constitutive MMV promoter in BU tobacco line TN90e4e5, which harbors KO mutations in CYP82E4 and CYP82E5v2 (Lewis and Dewey, 2015) to prevent excess nornicotine production and, therefore, variable NNN accumulation. To assess the effectiveness of the RNAi suppression, we performed qRT-PCR assays to measure the relative CLCNt2 transcript accumulation in several T0 individuals and selected the three plants that exhibited the strongest inhibition for further investigation. The cumulative levels of CLCNt2-S and CLCNt2-T transcripts were decreased by as much as 80% in the CLCNt2-RNAi lines (designated CLCNt2-1, CLCNt2-2, and CLCNt2-3) relative to WT controls (Table 1).

The CLCNt2-RNAi lines and WT control were grown in a field in the United States by using a randomized complete block design. No obvious phenotypic differences were observed between the WT and CLCNt2-RNAi lines grown in the field (Figure 3A). Plant biomass—determined by directly weighing plants freshly harvested by cutting at the base of the stalk—was not impacted by the suppressed expression of the CLCNt2 genes (Table 1). After each plant on the stalk was air-cured for 10 weeks in a dedicated barn in accordance with the standard practice for BU tobaccos,³ the lamina of the fourth leaf from the top of each plant was analyzed for nitrate content. In this experiment, the nitrate content in the three independent CLCNt2-RNAi lines was reduced significantly (up to 68% lower than the nitrate content in WT plants; Figure 3B).

In a randomized subset of the same tobacco samples, we analyzed the impact of decreasing CLCNt2 expression on alkaloid and TSNA content. Interestingly, the alkaloid content was minimally impacted in the CLCNt2-RNAi lines (Figure 4A). In contrast, the TSNA levels in the CLCNt2-RNAi lines, individually and cumulatively, exhibited reductions ranging from 33 to 45% relative to the WT. The impact of CLCNt2 silencing was significant for each TSNA species (NNN, NNK, NAT, and NAB) as well as across the three RNAi lines tested (Figure 4B), which suggested that the nitrosation process is similarly impacted for each of the different alkaloid substrates, nicotine, nornicotine, anatabine, and anabasine.

We determined the TSNAs released in the mainstream smoke of cigarettes made from leaf material from the CLCNt2-RNAi and WT plants. The presence of individual TSNAs in mainstream cigarette smoke is primarily due to the thermal release of preexisting TSNAs in the cut filler; although, in some circumstances, pyrosynthesis (which involves TSNA formation in the aerosol or in the hot zone of the cigarette by nitrosation of alkaloids precursors) can additionally contribute to the TSNA yield in cigarette smoke (Jaccard et al., 2018). To investigate the effect of CLCNt2 suppression on TSNA transfer and/or synthesis in smoke, the cigarettes were made from pure lamina of CLCNt2-1 plants and WT controls collected from the same location over 2 successive years. Cut filler for cigarette production was generated by finely shredding the cured leaf lamina. Differences in environmental growth and curing conditions can lead to substantial differences in TSNA formation, even in tobaccos that share the same genetic background. This phenomenon was evident when the TSNA levels in cigarette smoke were compared between the 2 years (Table 2). For example, twice as much NNN was observed in the smoke from cigarettes produced from materials grown in year 1 than those grown in year 2, even though the smoke nicotine levels were quite similar between the 2 years. The relative differences between the smoke TSNA levels in the CLCNt2-1-RNAi and WT control; however, were similar across both trial years, with the exception of NAB levels, where the differential was greater in year 2 than in year 1 (Table 2). These data are consistent with the observed reduction in leaf nitrate levels in the transgenic lines, which was also relatively comparable between the two growing seasons: 72% reduction in year 1 and 68% in year 2 for line CLCNt2-1 (data not shown).

Although RNAi suppression is a powerful tool for assessing the consequences of knocking down transcript accumulation of a targeted gene(s), this approach cannot help to establish the effect of complete gene inactivation or assess the relative contribution of multiple genes sharing high sequence identity. To test the effects of complete disruption of CLCNt2-S and CLCNt2-T gene function, both individually and collectively, we created KO mutations in CLCNt2 gene copies in the BU background TN90 by using the CRISPR–Cas9 system. In the T0 generation, an individual was identified (designated CLCNt2#8) that was monoallelic for the same 2-bp deletion both at the CLCNt2-S and CLCNt2-T loci (Figure 5). Given that these small deletions cause a frameshift that would result in the production of highly truncated CLCNt2 proteins that lack over 80% of the normally present amino acids, it can reasonably be assumed that the mutant gene products are incapable of functioning as nitrate transporters.

Plant CLCNt2#8 was self-pollinated, and over 200 T1 plants were screened by PCR using primers specific for the nptII selectable marker gene. Those plants that had lost the CRISPR transgene through segregation were genotyped by DNA sequence analysis in order to select plants homozygous for mutations in the individual CLCNt2-S (nt2s) and CLCNt2-T (nt2t) genes (nt2s/Nt2T and Nt2S/nt2t) as well as the double-homozygous mutant genotype (nt2s/nt2t). A single line of plants carrying each individual mutation and two double-homozygous mutant lines were grown and evaluated in the field together with the WT as control. After harvest, plant biomass was determined by weighing the plants directly in the field before air-curing, in the same manner as that described above for the CLCNt2-RNAi plants. Plant biomass was not affected by the presence of the clcnt2-KO mutations in the Nt2S/nt2t or double-homozygous mutant lines. The homozygous mutant clcnt2_s line (nt2s/Nt2T), however, unexpectedly displayed a statistically significant 29.5% reduction in biomass relative to the rest of the lines (Figure 6A). After curing, the leaf lamina from the same upper stalk position of each plant was analyzed for nitrate content. Despite the observation that CLCNt2-S transcripts were more prevalent than those corresponding to CLCNt2-T across every plant tissue examined (Figure 2C), neither single-mutant line displayed a nitrate phenotype that differed significantly from that of the WT control or each other (Figure 6B). In contrast, a significant 61.8% reduction in nitrate was observed in the two double-mutant lines (nt2s/nt2t), consistent with the data obtained with the CLCNt2-RNAi lines (Figure 3B).

Alkaloids and TSNAs were measured in the two independent double-KO lines (nt2s/nt2t). Surprisingly, the inactivation of both CLCNt2 copies in tobacco also significantly impacted the alkaloid content in leaves (Figure 7A), which was not observed when CLCNt2 transcript accumulation was knocked down in the CLCNt2-RNAi lines (see Figure 4A). The decrease was particularly marked for anatabine, which showed a decrease of 45.2% relative to the WT levels. Lower nicotine, nornicotine, and anabasine levels were also found to be statistically relevant in the double-homozygous clcnt2-KO plants, with reductions ranging from 23.9 to 34.9% relative to the WT (Figure 7A). As expected, because of the low-nitrate phenotype, the TSNA content in leaves was also substantially affected in the clcnt2-KO plants. The levels of all four TSNAs (NNK, NNN, NAT, and NAB) were clearly decreased in the two clcnt2 double-mutant lines (nt2s/nt2t), with a reduction of 44% in NNK, 35.7% in NNN, 62.8% in NAT, and 68.6% in NAB levels relative to the WT (Figure 7B). Despite the mean NNN values being lower, this measurement was not deemed significant owing to the exceptionally large variation measured among the plants, as attested by the minimum and maximum percentage values (−21.8 and −93.2%, respectively). Such variations were likely due to the use of TN90 lines as background for developing the clcnt2-KO plants, instead of TN90e4e5 in case of the CLCNt2-RNAi plant materials. Indeed, the TN90e4e5 cultivar (Lewis and Dewey, 2015) is similar to the Zyvert™ varieties, having lower and more stable nornicotine levels than TN90. Cumulatively, the data show that suppressed expression or inactivation of CLCNt2 genes resulted in a nitrate-underaccumulation leaf phenotype that impacted the TSNA formation.