- © 2016 American Society of Plant Biologists. All rights reserved.
Abstract
Codon optimization based on psbA genes from 133 plant species eliminated 105 (human clotting factor VIII heavy chain [FVIII HC]) and 59 (polio VIRAL CAPSID PROTEIN1 [VP1]) rare codons; replacement with only the most highly preferred codons decreased transgene expression (77- to 111-fold) when compared with the codon usage hierarchy of the psbA genes. Targeted proteomic quantification by parallel reaction monitoring analysis showed 4.9- to 7.1-fold or 22.5- to 28.1-fold increase in FVIII or VP1 codon-optimized genes when normalized with stable isotope-labeled standard peptides (or housekeeping protein peptides), but quantitation using western blots showed 6.3- to 8-fold or 91- to 125-fold increase of transgene expression from the same batch of materials, due to limitations in quantitative protein transfer, denaturation, solubility, or stability. Parallel reaction monitoring, to our knowledge validated here for the first time for in planta quantitation of biopharmaceuticals, is especially useful for insoluble or multimeric proteins required for oral drug delivery. Northern blots confirmed that the increase of codon-optimized protein synthesis is at the translational level rather than any impact on transcript abundance. Ribosome footprints did not increase proportionately with VP1 translation or even decreased after FVIII codon optimization but is useful in diagnosing additional rate-limiting steps. A major ribosome pause at CTC leucine codons in the native gene of FVIII HC was eliminated upon codon optimization. Ribosome stalls observed at clusters of serine codons in the codon-optimized VP1 gene provide an opportunity for further optimization. In addition to increasing our understanding of chloroplast translation, these new tools should help to advance this concept toward human clinical studies.
Heterologous gene expression has facilitated our understanding of DNA replication, recombination, transcription, and translation and protein import in chloroplasts. The expression of precursor proteins via the chloroplast genome demonstrated that cleavage of transit peptides takes place in the stroma and not in the chloroplast envelope (Daniell et al., 1998). Most importantly, the role of nucleus-encoded cytosolic proteins that bind to regulatory sequences and their species specificity were demonstrated using transgenes expressed in chloroplasts (Ruhlman et al., 2010). When the lettuce (Lactuca sativa) psbA regulatory sequence was used to drive transgene expression in tobacco (Nicotiana tabacum) chloroplasts, there was greater than 90% reduction in the accumulation of foreign proteins. This underscores the importance of the species specificity of chloroplast regulatory sequences. Likewise, details of the homologous recombination process and the deletion of mismatched nucleotides were evident using heterologous flanking sequences (Ruhlman et al., 2010). The translation of native polycistrons without the need for processing to monocistrons has been demonstrated (Barkan, 1988; Zoschke and Barkan, 2015), but the similarity of this process using heterologous polycistrons engineered via the chloroplast genome offered even more direct evidence for this process (De Cosa et al., 2001; Quesada-Vargas et al., 2005). The insertion of replication origins into chloroplast vectors offered further insight into minimal sequences required to study this process (Daniell et al., 1990). Therefore, in this study, we use transgenes, chloroplast genome sequences, and cutting-edge tools to understand the process of translation in chloroplasts.
Each plant cell contains up to 10,000 copies of the chloroplast genome. Therefore, transgenes inserted into chloroplast genomes are expressed at high levels, up to 70% of total leaf protein (De Cosa et al., 2001; Ruhlman et al., 2010). A wide range of proteins, from very small antimicrobial peptides (Lee et al., 2011) or hormones (Boyhan and Daniell, 2011; Kwon et al., 2013) to very large proteins encoded by bacterial, viral, fungal, animal, and human genes, have been expressed successfully in plant chloroplasts (DeGray et al., 2001; Daniell et al., 2009; Verma et al., 2010; Shenoy et al., 2014; Sherman et al., 2014; Shil et al., 2014). Most importantly, expressed proteins are highly stable when lyophilized plant cells are stored at ambient temperature (Kwon et al., 2013; Lakshmi et al., 2013; Kohli et al., 2014; Jin and Daniell, 2015). Therefore, oral delivery of proinsulin or exendin-4 reduced blood sugar levels similar to injected proteins (Boyhan and Daniell, 2011; Kwon et al., 2013). Oral delivery of angiotensin and ANGIOTENSIN-CONVERTING ENZYME2 expressed in chloroplasts reversed or prevented pulmonary hypertension by shifting the renin-angiotensin system to its protective axis, resulting in a decrease in fibrosis, improvement in cardiopulmonary structure and function, and restoration of right heart function (Shenoy et al., 2014). Furthermore, ocular inflammation caused by decreased activity of the protective axis of the renin-angiotensin system was improved significantly (Shil et al., 2014). Likewise, oral delivery of myelin basic protein reduced Aβ plaques in advanced mouse and human Alzheimer’s brains (Kohli et al., 2014). Delivery of coagulation factors to hemophilic mice induced oral tolerance and suppressed inhibitor formation and anaphylaxis (Verma et al., 2010; Sherman et al., 2014; Wang et al., 2015a). The aforementioned examples illustrate the significance of this novel, cost-effective protein drug-delivery concept.
However, a major limitation in the clinical translation of human therapeutic proteins in chloroplasts is their low-level expression. Prokaryotic or shorter human genes are highly expressed in chloroplasts (De Cosa et al., 2001; Arlen et al., 2007; Daniell et al., 2009; Ruhlman et al., 2010). However, expression of larger human proteins is a major challenge. For example, cholera nontoxic B subunit (CNTB)-fused native human blood-clotting factor VIII heavy chain (FVIII HC; 86.4 kD) or ANGIOTENSIN CONVERTING ENZYME2 (92.5 kD) were expressed at very low levels (Shenoy et al., 2014; Sherman et al., 2014). Likewise, the expression of viral vaccine antigens is quite unpredictable, with high, moderate, or extremely low expression levels (Birch-Machin et al., 2004; Lenzi et al., 2008; Waheed et al., 2011a, 2011b; Inka Borchers et al., 2012; Hassan et al., 2014). Furthermore, viral antigens are highly unstable, with expression observed in youngest leaves but not in mature leaves (McCabe et al., 2008). It is well known that high doses of vaccine antigens stimulate high-level immunity and confer greater protection against pathogens; therefore, higher level expression in chloroplasts is a key requirement for vaccine development (Chan and Daniell, 2015; Chan et al., 2016).
Such challenges in transgene expression have been addressed by the use of optimal regulatory sequences (promoters and 5ʹ and 3ʹ untranslated regions [UTRs]), especially species-specific endogenous elements (Ruhlman et al., 2010). In vitro assays of inserted genes with several synonymous codons show that translation efficiency does not always correlate with codon usage in plastid mRNAs (Nakamura and Sugiura, 2007), but they have been used in several codon optimization studies (Lutz et al., 2001; Ye et al., 2001; Franklin et al., 2002; Lenzi et al., 2008; Jabeen at al., 2010; Madesis et al., 2010; Gisby et al., 2011; Wang et al., 2015b; Boehm et al., 2016; Nakamura et al., 2016). While some studies achieved significant increases in expression (75- to 80-fold) after codon optimization (Franklin et al., 2002; Gisby et al., 2011), other studies observed negligible enhancement (Ye et al., 2001; Lenzi et al., 2008; Daniell et al., 2009; Wang et al., 2015b; Nakamura et al., 2016). However, translation initiation and the elongation efficiency of codon-optimized sequences were enhanced when chloroplast gene N-terminal sequences were inserted downstream of 5ʹ UTRs (Ye et al., 2001; Lenzi et al., 2008). In a recent study (Nakamura et al., 2016), the importance of compatibility between the psbA 5ʹ UTR and its 5ʹ coding sequence was shown using codon-optimized heterologous genes. The aforementioned codon optimization studies used only smaller eukaryotic coding sequences (less than 30 kD), but there is a great need to express larger human genes (e.g. FVIII; greater than 200 kD) that would require not only the optimization of codons but also compatibility with regulatory sequences for optimal translation initiation, elongation, and greater understanding of tRNAs encoded by the chloroplast genome or imported from the cytosol. However, no systematic study has been done to utilize the extensive knowledge gathered by sequencing several hundred chloroplast genomes (Daniell et al., 2016a) to understand codon usage and the frequency of highly expressed chloroplast genes.
Another major challenge is the lack of reliable methods to quantify insoluble proteins; the only reliable method (ELISA) cannot be used due to the aggregation or formation of multimeric structures that are required for oral drug delivery. Although the FDA accepts ELISA for the quantitation of purified protein drugs, it is not suitable for quantifying protein drugs from impure extracts due to cross-reacting proteins, autoantibodies (Kim and You, 2013), or for the quantitation of insoluble, multimeric, or membrane proteins. Similarly, immunoblots used for quantitation also have several limitations (i.e. aggregation of proteins at high protein concentrations trapped in wells, alteration of mobility by incomplete solubilization or secondary structures, saturation of antibody-binding sites, and inefficient transfer of large proteins to membranes and variable quantitation due to short or long exposure to films). However, peptide-centric quantitation strategies (e.g. targeted mass spectrometry quantitation by parallel reaction monitoring [PRM]) can overcome most of the limitations mentioned above. In the preparation of protein samples for PRM, strong denaturing and reducing conditions are used (e.g. higher concentrations of SDS and DTT) in combination with optimal enzymatic proteolysis conditions (e.g. sodium-deoxycholate; León et al., 2013), especially suitable for insoluble, multimeric, or membrane proteins (Savas et al., 2011). Moreover, PRM can be used for relative and absolute protein quantitation of target proteins present in highly complex protein backgrounds based on its high specificity and sensitivity (Domon and Aebersold, 2010; Gallien et al., 2012; Picotti and Aebersold, 2012). In addition, PRM offers high specificity and multiplexing characteristics, which allow for specific monitoring of up to several hundred peptides in a single analysis (Gallien et al., 2012). Determination of protein drug dose in planta, especially of insoluble proteins without purification, is an unexplored area of research, and to our knowledge, we investigate this concept for the first time to quantify recombinant protein drugs made in chloroplasts.
This study explores heterologous gene expression utilizing chloroplast genome sequences, ribosome profiling, and targeted mass spectrometry (PRM) to enhance our understanding of the translation of foreign genes in chloroplasts. We developed a codon optimizer program based on the analysis of psbA genes from 133 plant species to compare the translational efficiencies of native and codon-optimized genes driven by identical regulatory sequences. PRM using peptides selected from the N or C terminus were used to study the complete or incomplete synthesis of proteins and to validate this approach to quantify the dosage of protein drugs made in plant cells when compared with current methods. The codon optimizer program was evaluated in chloroplasts from two different species to identify any species specificity. Ribosome profile was evaluated for its suitability to diagnose limiting steps in transgene expression. These observations provide new insight into limitations in the translation of heterologous genes and approaches to address this in future studies.
RESULTS
Codon Optimization of Human/Viral Transgenes
Differences in codon usage by chloroplasts frequently decrease translation. We observed that plants expressing native sequences of FVIII HC or VIRAL CAPSID PROTEIN1 (VP1) from polio virus showed very low levels of expression, less than 0.05% for FVIII and approximately 0.1% for VP1 (see below). The psbA gene is among the most highly expressed genes in chloroplasts, and the translation efficiency of the psbA gene is greater than 200 times higher than that of the rbcL gene (Eibl et al., 1999). The 5ʹ UTR of psbA also showed the highest translation activity in vitro among 11 5ʹ UTRs investigated (Yukawa et al., 2007). Therefore, among 140 transgenes expressed in chloroplasts, more than 75% use the psbA regulatory sequences (Jin and Daniell, 2015; Daniell et al., 2016a, 2016b). Most importantly, compatibility between the 5ʹ UTR of psbA and its coding region is important for efficient translation initiation (Nakamura et al., 2016). For these reasons, a new codon optimization program was developed using codon usage of the psbA genes from 133 sequenced chloroplast genomes (Fig. 1A). We first investigated the expression of synthetic genes using only the most highly preferred codon for each amino acid, which is referred to as the old algorithm in this study. When this resulted in even lower levels of expression than the native gene (see below), a new codon optimizer algorithm was developed using the codon usage hierarchy observed among sequenced psbA genes. Therefore, most of the rare codons in heterologous genes were modified based on codons with greater than 5% frequency of use in the psbA genes. Synonymous codons for each amino acid were ranked according to their frequency of use (Fig. 1B).
Development of a codon optimization algorithm for the expression of heterologous genes in plant chloroplasts. A, Process to develop the codon optimization algorithm. Sequence data of psbA genes from 133 plant species collected from the National Center for Biotechnology Information, and their codon preferences, were analyzed. Finally, the codon optimizer was developed using Java. B, Codon preference table. Codon preference is indicated by the percentage of use for each amino acid. Black and underlined codons indicate codons that were not used when optimizing sequences due to their low usage frequency among synonymous codons (less than 5% use or, for amino acids with six synonymous codons [Leu, Ser, and Arg], the two codons used least frequently).
In this study, native sequences for FVIII HC (2,262 bp) and VP1 (906 bp) were codon optimized using the old or new algorithm and synthesized. After codon optimization, the AT content of FVIII HC increased slightly, from 56% to 62%, and 406 codons out of 754 amino acids were optimized. For the VP1 sequence from the Sabin 1 polio virus strain, the 906-bp-long native sequence was codon optimized, which slightly increased the AT content from 52% to 59%, and 187 codons out of 302 amino acids were optimized. However, the CNTB coding sequence was not codon optimized because of its prokaryotic origin and high AT content (65.4%). Most importantly, the expression level of CNTB (native sequence) fused with proinsulin reached up to 72% of total leaf protein in tobacco chloroplasts (Ruhlman et al., 2010) and 53% of total leaf protein in lettuce chloroplasts (Boyhan and Daniell, 2011), indicating that there is no limitation on translation of the CNTB coding sequence in chloroplasts. All sequences, including native and codon-optimized synthetic genes (new and old algorithms), are shown in Supplemental Figure S1; rare codons in native genes are shown in red and modified codons are highlighted in yellow in Supplemental Figure S2.
When the psbA-based codon table is compared with total chloroplast codon usage tables, which are generated based on all chloroplast genes of lettuce (57,528 codons from 189 coding sequences) or tobacco (34,756 codons from 137 coding sequences; Nakamura et al., 2000), there was no significant difference in AT content of coding sequences: it varied between 59.59% and 61.76%. However, there are striking differences between psbA-based and total chloroplast gene-based codon tables when individual codons are compared. Native FVIII HC used CTC Leu codon 11 times, but codon-optimized (new algorithm) HC eliminated all CTC codons. However, if the total chloroplast codon table is used, codon-optimized HC would still use five CTC codons. As seen in ribosome profiles, discussed below, tandem repeat of CTC-CTC in the native FVIII HC sequence resulted in major stalling sites that were completely eliminated by psbA-based codon optimization (new algorithm). Likewise, another rare codon, TCA (Ser), is used 16 times in the FVIII HC and seven times in VP1 coding sequences. However, the TCA rare codon was eliminated completely in both genes after codon optimization using the new algorithm. However, if the total codon table is used for codon optimization, FVIII HC and VP1 would still contain 12 and five TCA codons. Collectively, the new codon optimization algorithm eliminated 105 and 59 rare codons from FVIII HC and VP1, respectively, resulting in enhanced expression of both genes. However, if the total codon table is used, there will be 75 and 35 rare codons in codon-optimized FVIII HC and VP1 coding sequences, respectively. All 13 codons (GCG [Ala], GGG [Gly], CTG [Leu], CTC [Leu], CCG [Pro], CCC [Pro], AGG [Arg], CGG [Arg], TCA [Ser], TCG [Ser], ACG [Tyr], GTC [Val], and CTG [Val]) rarely used in the psbA gene were eliminated using our codon-optimized table (new algorithm). More detailed information on the codon distribution between different codon tables is included in Supplemental Figure S3.
Synthetic gene cassettes were inserted into the chloroplast transformation vector, pLSLF for lettuce or pLD-utr for tobacco (Fig. 2A). Native and synthetic genes were fused to the native CTNB sequence, which is used for efficient transmucosal delivery of fused proteins via monosialotetrahexosylganglioside receptors present on intestinal epithelial cells. To eliminate possible steric hindrance caused by the fusion of two proteins and facilitate the release of tethered proteins into the circulation after internalization, nucleotide sequences for a hinge (Gly-Pro-Gly-Pro) and a furin cleavage site (Arg-Arg-Lys-Arg) were engineered between CNTB and fused proteins. Fusion genes were placed under the control of identical psbA promoters, 5′ UTR and 3′ UTR regulatory sequences, for specific evaluation of codon optimization (Fig. 2A). To select transformants, the aminoglycoside-3″-adenylyl-transferase gene was driven by the rRNA promoter to confer resistance to spectinomycin in transformed cells. Expression cassettes were flanked by sequences for isoleucyl-tRNA synthetase and alanyl-tRNA synthetase, which are identical to endogenous chloroplast genome sequences, leading to efficient double homologous recombination and optimal processing of introns with flanking sequences (Fig. 2A).
Construction of chloroplast vectors using native or codon-optimized genes, and evaluation of homoplasmy and transgene expression. A, Lettuce or tobacco chloroplast vector maps. aadA, Aminoglycoside 3′-adenylytransferase gene; CNTB, coding sequence of cholera nontoxic B subunit; FVIII HC, factor 8 heavy chain native (N) or codon optimized (CN) using the new algorithm; PpsbA, promoter and 5′ UTR of the psbA gene; Prrn, rRNA operon promoter; SB-P, BamHI fragment; TpsbA, 3′ UTR of the psbA gene; trnA, alanyl-tRNA; trnI, isoleucyl-tRNA. B and C, Southern-blot analysis of homoplasmic lines. Total genomic DNA (3 µg) from untransformed (UT), native (N), or codon-optimized CNTB-FVIII HC (new algorithm; CN) was digested with HindIII and separated on a 0.8% agarose gel, blotted onto a Nytran membrane, and probed with a BamHI fragment. Lanes 1 to 4 show four independent transplastomic lines. L.s., Lactuca sativa. D, Comparison of the expression level of CNTB-VP1 between transplastomic lines expressing the native (N) or codon-optimized genes using the old (CO) or new (CN) algorithm. Total extracted proteins were loaded as indicated protein concentrations and were probed with anti-CNTB antibody. CNTB, Standard protein of cholera nontoxic B subunit; IDV, integrated density values; N.t., Nicotiana tabacum.
Transformation vectors containing the native or synthetic sequences for FVIII HC and VP1 sequences were used to create transplastomic lettuce or tobacco plants. To confirm homoplasmy, Southern-blot analysis was performed on four independent lettuce and tobacco lines expressing native or codon-optimized FVIII HC and VP1. For lettuce plants expressing either native or codon-optimized CNTB-FVIII HC, chloroplast genomic DNA was digested with HindIII and probed with digoxigenin (DIG)-labeled probe spanning the flanking region. All selected lines showed the expected distinct hybridizing fragments and no untransformed fragment (Fig. 2, B and C). The homoplasmic tobacco lines expressing native or codon-optimized CNTB-VP1 sequences were confirmed already in a previous study (Chan et al., 2016). Therefore, these data confirm the homoplasmy of all transplastomic lines; therefore, transgene expression levels should be attributed to translation efficiency and not transgene copy number.
Translation Efficiency of Native and Codon-Optimized Genes in Lettuce and Tobacco Chloroplasts
Expression levels between native and codon-optimized genes in chloroplasts were compared using immunoblot and densitometry assays. Early studies in this project compared the translation efficiency of the old algorithm (using only the most preferred codons) with that of the new algorithm (using the psbA codon hierarchy) quantified by integrated density values of western blots (Fig. 2D). The CNTB-VP1 expression level in transplastomic plants using the old algorithm for codon optimization was 2.7- to 3.1-fold lower than that of the native VP1 viral gene sequence, and the increase in VP1 expression was 77- to 111-fold higher using the new algorithm (Fig. 2D). Therefore, the new algorithm of the codon optimizer program was used in all subsequent studies. In order to correct for overexposure or underexposure of western blots to x-ray film, data on variable exposures were collected. In order to account for extreme variation in the expression levels of native and codon-optimized genes, serial dilutions of extracted proteins were loaded on each blot (Fig. 3, A and B; Supplemental Fig. S4). In a densitometry assay of lettuce expressing native and codon-optimized CNTB-FVIII HC, which also was used for PRM, the concentration of FVIII HC from the codon-optimized gene (108.8–137.5 µg g−1 dry weight) was 6.3- to 8-fold higher than that of the native FVIII HC gene (16.9–17.4 µg g−1 dry weight; Supplemental Fig. S4). For tobacco plants expressing CNTB-VP1, the batch used for PRM mass spectrometry showed a 91- to 125-fold difference between codon-optimized (11.3–18.1 µg mg−1) and native sequence (0.12–0.15 µg mg−1; Fig. 3C; Supplemental Fig. S4). Based on these data, codon-optimized sequences obtained from our newly developed codon optimizer program improved the translation of transgenes to different levels, based on the coding sequence.
Quantitation of native or codon-optimized CNTB-FVIII HC or CNTB-VP1 gene expression using western blots. Extracted leaf proteins were resolved on gradient (4%–20%) SDS-PAGE and probed with anti-CNTB antibody (1:10,000). For a loading control, the same membranes were stripped and reprobed with anti-RbcL antibody (1:5,000). A, Lettuce leaf protein extracts (5 or 10 μg) expressing CNTB-FVIII HC or untransformed. For loading controls, Ponceau S staining of membrane prior to western blot or reprobed blot with the large subunit of Rubisco (RbcL) is provided. B, Serial dilution of the native (5–20 μg) or codon-optimized (1–4 μg) CNTB-FVIII HC lettuce leaf extracts. C, Serial dilution of the native (2–8 μg) or codon-optimized (0.1–0.4 μg) CNTB-VP1 tobacco leaf extracts. CO or CN, Codon optimized with old algorithm (CO) or new algorithm (CN); L.s., Lactuca sativa; N, native sequence; N.t., Nicotiana tabacum; UT, untransformed wild type.
To investigate the impact of codon optimization on transcript stability, northern blotting was performed using a probe for the psbA 5ʹ sequence (Fig. 4). Although loading controls show equal amounts of total RNA in each lane based on ethidium bromide staining, higher or lower levels of the endogenous psbA transcript are observed among samples, suggesting subtle changes in RNA loading. The mRNA levels of codon-optimized or native sequences for CNTB-FVIII HC and CNTB-VP1 were normalized to endogenous psbA transcripts using densitometry, and the normalized ratios in each sample were compared. Northern blots indicated that the increase of codon-optimized CNTB-FVIII HC and CNTB-VP1 accumulation is at the translational level rather than RNA transcript accumulation. Several previous studies on the expression of foreign genes have shown a lack of variation or modest increases in transcript abundance but significant variation in translation efficiency (Franklin et al., 2002; Gisby et al., 2011; Nakamura et al., 2016). Franklin et al. (2002) reported a lack of variation in transcript abundance for GFP expression in Chlamydomonas reinhardtii chloroplasts despite an 80-fold increase in GFP protein accumulation of the codon-optimized sequence. Even though there was a 3-fold increase in mRNA levels of codon-optimized TGF-β3 when compared with the native sequence (Gisby et al., 2011), the greater part of the 75-fold increase in synthetic TGF-β sequence was attributed to enhanced translation. A recent study also showed that the compatibility of the 5ʹ UTR and its coding sequence increased the efficient translation of codon-optimized sequences rather than mRNA abundance (Nakamura et al., 2016).
Northern analysis of transplastomic lines. Transgene transcripts of CNTB-FVIII HC (A) or CNTB-VP1 (B) were probed with 200 bp of lettuce psbA 5ʹ UTR (for FVIII HC) or tobacco psbA 5ʹ UTR (for VP1) regulatory sequences. Bottom and top arrowheads represent the endogenous psbA gene and CNTB-FVIII or CNTB-VP1 transgene, respectively. Ethidium bromide (EtBr)-stained gels are included for the evaluation of equal loading. CN, Codon-optimized sequence using the new algorithm; N, native sequence; UT, untransformed wild type.
Absolute Quantitation by PRM Analysis
Expression levels of codon-optimized and native gene sequences also were quantified using PRM mass spectrometry (Fig. 5). To select the optimal proteotypic peptides for PRM analysis of the CNTB and FVIII HC sequences, we first performed a standard tandem mass spectrometry analysis (data not shown) of a tryptic digest of lettuce plants expressing CNTB-FVIII HC to choose specific peptides. The expression of codon-optimized FVIII HC was 5.4- or 5.8-fold higher than that of the native sequence when the fold changes were normalized based on the housekeeping protein peptides or stable isotope-labeled standard (SIS) peptides (Figs. 5 and 6A). Peptides chosen from CNTB showed minor variations in fold change based on the locations of peptides and normalized with SIS or housekeeping protein peptides from Rubisco (small or large subunits) or ATP synthase subunit β: 4.9 (or 4.5; IAYLTEAK), 5.2 (or 4.8; IFSYTESLAGK), or 6.6 (or 6.1; LCVWNNK). Peptides chosen from FVIII HC also showed minor variations: 5.4 (or 5; FDDDNSPSFIQIR), 5.7 (or 5.2; YYSSFVNMER), or 7.1 (or 6.6; WTVTVEDGPTK; Fig. 6A). The locations of these selected peptides within CNTB-FVIII HC are shown in Supplemental Figure S5. For more details, see the raw data included in Supplemental Data Set S1.
PRM mass spectrometry analysis of CNTB-FVIII and CNTB-VP1 proteins at N- to C-terminal protein sequences. The y axis shows molarity (fmol on column) of peptides from CNTB-FVIII HC or CNTB-VP1 in codon-optimized or native genes. CNTB: peptide 1, IFSYTESLAGK; peptide 2, IAYLTEAK; peptide 3, LCVWNNK. FVIII: peptide 4, FDDDNSPSFIQIR; peptide 5, WTVTVEDGPTK; peptide 6, YYSSFVNMER. The median of four technical replicates is presented for each sample. Circles represent native sequences, and squares represent codon-optimized (c.o.) sequences using the new algorithm. CV, Coefficient of variation.
Fold change (increase) of CNTB-FVIII HC or CNTB-VP1 proteins based on targeted mass spectrometry analysis of CNTB and HC peptides. The reported data represent medians of the results from six and three peptides from CNTB-FIII HC (A) and CNTB-VP1 (B), respectively. The y axis represents the fold change increase (based on measured fmol on column) of peptides from plant materials expressing genes codon optimized using the new algorithm (CO) with respect to plant materials expressing native sequence (N). CNTB: peptide 1, IFSYTESLAGK; peptide 2, IAYLTEAK; peptide 3, LCVWNNK. FVIII HC: peptide 4, FDDDNSPSFIQIR; peptide 5, WTVTVEDGPTK; peptide 6, YYSSFVNMER. SIS-normalized values represent fold change as a ratio to each spiked SIS peptide. Housekeeping (HK) protein normalization values represent fold change as a normalized ratio to Rubisco large or small subunit and ATP synthase CF1 β-subunit protein peptides. For peptide ratio results for CNTB-FVIII and CNTB-VP1, see Supplemental Data Set S1.
The expression of codon-optimized CNTB-VP1 was 25.9- or 26.1-fold higher than that of the native sequence when their fold changes were normalized based on the SIS peptides or housekeeping protein peptides (Figs. 5 and 6B). Peptides chosen from CNTB showed minimal variations in fold changes based on their locations: 22.5 (or 22.5; LCVWNNK) to 26.1 (or 26; IAYLTEAK) to 28.1 (or 28; IFSYTESLAGK; Fig. 6B). The linearity of the quantification range also was investigated by spiking stable SIS peptides in a constant amount of plant digest (1:1:1:1 mix of all four types of plant materials) in a dynamic range covering 220 amol to 170 fmol (values equivalent on column per injection). These results are reported in detail in Supplemental Figure S7. For all six peptides, we observed an r2 value over 0.98.
Absolute quantitation can be achieved by spiking a known amount of the counterpart SIS peptide into samples. For each counterpart, SIS peptide (34 fmol) was injected on column mixed with protein digest (equivalent to protein extracted from 33.3 µg of lyophilized leaf powder). By calculating ratios of area under the curve of SIS and endogenous peptides, we estimated the endogenous peptide molarity, expressed as femtomoles on column (Fig. 6). The mean of all calculated ratios of femtomoles on column (six and three peptides for CNTB-FVIII HC and CNTB-VP1, respectively) for codon-optimized and native sequences is reported as the fold increase of protein expression in codon-optimized constructs. The high reproducibility of the sample preparation and PRM analysis is shown in Figure 5. All peptide measurements were the result of four technical replicates, two sample preparation replicates (from leaf powder to extraction to protein digestion), and two mass spectrometry technical replicates. Coefficients of variation among the four measurements per peptide ranged from 0.5% to 10% in all but two cases, where they were 17% and 22%.
Ribosome Profiling Studies
Ribosome profiling uses deep sequencing to map ribosome footprints, mRNA fragments that are protected by ribosomes from exogenous nuclease attack. The method provides a genome-wide, high-resolution, and quantitative snapshot of mRNA segments occupied by ribosomes in vivo (Ingolia et al., 2009). Total ribosome footprint abundance within an open reading frame can provide an estimate of translational output, and positions at which ribosomes slow or stall are marked by regions of particularly high ribosome occupancy.
To examine how codon optimization influenced ribosome behavior, we profiled ribosomes from plants expressing the native and codon-optimized CNTB-FVIII HC and CNTB-VP1 transgenes. Figure 7 shows the abundance of ribosome footprints as a function of position in each transgene; footprint coverage on the endogenous chloroplast psbA and rbcL genes is shown as a means to normalize the transgene data between the optimized and native constructs. Ribosome footprint coverage was much higher in the codon-optimized VP1 sample than in the native VP1 sample (Fig. 7A). However, the magnitude of this increase varies depending upon how the data are normalized (Fig. 7C): the increase is 5-, 16-, or 1.5-fold when normalized to total chloroplast ribosome footprints, psbA ribosome footprints, or rbcL ribosome footprints, respectively. These numbers are considerably lower than the 22.5- to 28.1-fold increase in VP1 protein abundance inferred from the quantitative mass spectrometry data. The topography of ribosome profiles is generally highly reproducible among biological replicates (see rbcL and psbA in Fig. 7B), which are at the same developmental stage and grown under the same conditions. In that context, it is noteworthy that the peaks and valleys in the endogenous psbA and rbcL genes are quite different in the native and optimized tobacco VP1 lines. It could be envisaged that competition with the endogenous psbA 5ʹ UTR could, in principle, reduce the translation of the endogenous psbA open reading frame. However, no such competition was observed for the lettuce construct. In addition, the degree of competition would depend on the abundance of the transgene mRNA. The abundance of the transgene mRNA was similar in the native and codon-optimized constructs, so competition via the psbA 5ʹ UTR is unlikely to contribute to differences in psbA ribosome occupancy in these lines. Many of the large peaks (presumed ribosome pauses) observed in these endogenous genes, specifically in the native VP1 line, map to paired Ala codons (asterisks in Fig. 7A). This suggests a limitation of Ala tRNA specifically in the native VP1 line. Although the basis for this is unclear, it is conceivable that it has to do with minor differences in the age of the plants used for the analyses (2.5 versus 2 months). It is also conceivable that introduction of the transgene had an unanticipated effect on the expression of the nearby gene encoding Ala tRNA. In the same vein, ribosome pause sites in the CNTB region would be expected, but the sites of the native and optimized VP1 constructs were not similar. This global difference in ribosome behavior at Ala codons may well contribute to differential transgene expression in the native and codon-optimized lines.
Ribosome profiling data from transplastomic plants expressing native and codon-optimized VP1 or FVIII HC. Read coverage for native transgenes (N), codon-optimized transgenes with new algorithm (CN), and the endogenous psbA and rbcL genes are displayed with the Integrated Genome Viewer. A, Data from tobacco leaves expressing native and codon-optimized VP1 transgenes. Asterisks mark each pair of consecutive Ala codons in the data from the native line. The + symbol marks three consecutive Ala codons. Many strong ribosome pause sites in the plants expressing native VP1 map to paired Ala codons, whereas this is not observed in the codon-optimized line. Triangles mark each pair of consecutive Ser codons in the codon-optimized line. A major ribosome stall maps to a region harboring five closely spaced Ser codons in the codon-optimized VP1 gene. nt, Nucleotides. B, Data from lettuce plants expressing the native and codon-optimized FVIII HC transgenes. A major ribosome stall in the native FVIII HC gene maps to a pair of adjacent CTC Leu codons, a codon that is not used in the native psbA gene. Ribosome footprint coverage is much more uniform on the codon-optimized transgene. C, Absolute and relative ribosome footprints counts.
The total number of ribosome footprints in the FVIII gene decreased approximately 2-fold in the codon-optimized line, whereas protein accumulation increased 4.5- to 6.6-fold. However, a major ribosome pause can be observed near the 3ʹ end of the native transgene, followed by a region of very low ribosome occupancy (see bracketed region in Fig. 7B). This ribosome pause maps to a pair of CTC Leu codons, a codon that is almost not used in native psbA genes (Fig. 1). These results strongly suggest that the stalling of ribosomes at these Leu codons limits the translation of the downstream sequences and overall protein output while also causing a buildup of ribosomes on the upstream sequences. Thus, overall ribosome occupancy does not reflect translational output in this case. Modification of those Leu codons in the codon-optimized variant eliminated this ribosome stall and resulted in a much more even ribosome distribution over the transgene (Fig. 7B, right). Taken together, the ribosome-profiling data revealed dramatic differences in ribosome dynamics between codon-optimized and native transgenes. Although total ribosome occupancy did not reliably predict protein output from transgenes expressed in chloroplasts, the detection of strong ribosome pauses at specific sites can provide insight into rate-limiting steps that could be mitigated through sequence modifications.
DISCUSSION
Past studies on transgene expression in chloroplasts reported abundant transcripts but variable levels of translation based on the origin of the coding sequence. Prokaryotic genes were translated more efficiently than eukaryotic genes. Transcript abundance is attributed to the high copy number of transgenes and the strength of the psbA promoter. Among more than 150 transgenes expressed in chloroplasts, more than 75% utilized psbA regulatory sequences (Jin and Daniell, 2015; Daniell et al., 2016a, 2016b). In addition, three ribosome-binding regions in the 5ʹ UTR of psbA recruit ribosomes and efficiently form the translational initiation complex (Zou et al., 2003). Therefore, it is expected that improvement of translation elongation in heterologous genes should increase transgene expression. There is a drawback of using a codon table based on all chloroplast genes, which assumes that all tRNA species are equally abundant. However, such translational selection is not possible (Surzycki et al., 2009). Therefore, in this study, we developed a codon optimizer program based on the codon usage of psbA genes across 133 plant species to increase the expression of heterologous genes in chloroplasts.
Codon Optimization Significantly Enhances Translation in Chloroplasts
The psbA promoter and 5ʹ UTR are the most widely used regulatory sequences for transgene expression in chloroplasts. Among more than 115 transgenes expressed via the chloroplast genome, 84 use the psbA regulatory sequence (Jin and Daniell, 2015; Daniell et al., 2016a, 2016b). A recent study (Nakamura et al., 2016) shows the absence of any detectable translation when codons for the tat coding sequence of HIV-1 were optimized using all 79 tobacco chloroplast mRNAs and regulated by the psbA 5ʹ UTR (Nakamura et al., 2016), but the same sequence was expressed well using the phage T7 GENE10 5ʹ UTR. However, when the 5ʹ psbA coding sequence was inserted between the psbA 5ʹ UTR and the tat sequence, translation was initiated. Therefore, compatibility between the psbA regulatory element and codons is vital for initiation and elongation during the translation of heterologous genes (Nakamura et al., 2016). Therefore, when heterologous genes are regulated by psbA, codon optimization based on psbA codon usage should facilitate the movement of ribosomes more efficiently from the translational initiation complex than codon-optimized sequences based on any other chloroplast genes.
In this study, we developed and tested two new codon optimizer programs based on the codon preference of psbA genes to improve the expression of heterologous genes in chloroplasts in concert with the psbA regulatory elements. The first old algorithm of the codon optimizer was programmed to use only the most highly used codons, resulting in lower expression than the native gene. The increase in expression of VP1 in chloroplasts between the old and new algorithm is 77- to 111-fold. Therefore, removal of rare codons and replacement with only highly preferred codons did not help in enhancing translation when tRNA pools were limited. Thus, the new algorithm of the codon optimizer program was used in all subsequent studies. The new algorithm of the codon optimizer used the codon distribution hierarchy observed among psbA genes. As a result, 105 rare codons out of 754 codons in the FVIII HC gene and 59 rare codons out of 302 codons in the VP1 gene were replaced with psbA preferentially used codons. However, the replaced codons are not identified as rare codons in codon tables using all chloroplast genes. Therefore, the total chloroplast codon table would have retained 75 rare codons in FVIII HC and 35 rare codons in the VP1 coding sequence.
Although we used a psbA-based codon optimization program to improve translation in chloroplasts, many other factors, including the size and origin of heterologous genes and the compatibility of the 5ʹ UTR and its 5ʹ coding region, are important. The CNTB-fused native sequence of human proinsulin (approximately 22 kD) was expressed up to 72% of total leaf protein (Ruhlman et al., 2010), and the expression of ZZTEV-IGF-1 (Staphylococcus aureus Z domains and TEV cleavage site fused to native human insulin-like growth factor1 gene; approximately 26 kD) was up to 32.4% of TSP (Daniell et al., 2009). However, human TGF-β3 (13 kD, 56% GC) was expressed in up to 12% of leaf protein only after codon optimization (Gisby et al., 2011). Also, the expression of GFP (approximately 26 kD) increased approximately 80-fold after codon optimization (Franklin et al., 2002). Therefore, proteins with shorter coding sequences are not ideal to evaluate codon optimization concepts and other limitations in translation. Consequently, a better understanding of codon usage and other rate-limiting steps (compatibility with regulatory sequences, efficiency of translation initiation, elongation, and availability of tRNAs) in translation is essential for the successful expression of human or other eukaryotic coding sequences.
Codon usage in psbA (our program) is different for preferred Arg, Asn, Gly, His, Leu, and Phe codons than those reported for 79 tobacco chloroplast mRNAs based on in vitro studies (Nakamura and Sugiura, 2007). Preferred codons are decoded more rapidly than nonpreferred codons, presumably due to higher concentrations of corresponding tRNAs that recognize preferred codons, which speeds up the elongation rate of protein synthesis (Yu et al., 2015). Higher plant chloroplast genomes code for a conserved set of 30 tRNAs. This set is believed to be sufficient to support the translation machinery in chloroplasts (Lung et al., 2006). In the ribosome profiling data for codon-optimized VP1, two major ribosome stalling sites correlated with an unusually high concentration of Ser codons (Fig. 7A). Five Ser codons were clustered at codons 71, 73, 75, 76, and 79, and three other Ser codons were found at codons 178, 179, and 182. Two adjacent Ser residues in each cluster, codons 75 and 76 (UCU-AGU) and codons 178 and 179 (UCC-UCU; see triangles in Fig. 7A), show a high level of ribosome stalling. Thus, it may be possible to further increase the expression of the codon-optimized VP1 transgene by replacing these codons with codons for a different but similar amino acid.
As seen in this study, the AT content of codon-optimized VP1 was increased marginally, but the protein level of the optimized CNTB-VP1 increased significantly, up to 22.5- to 28.1-fold (by PRM) and 91- to 125-fold (by western blot), over the native sequence when expressed in chloroplasts. Therefore, several other factors play key roles in regulating the efficiency of translation. As observed in ribosome profiling studies of CNTB-VP1, the availability and density of specific codons could severely impact translation. Similarly, FVIII HC ribosome footprint results showed that ribosome pauses mapped to CTC Leu codons, which are almost not used in psbA genes. This codon also is rarely used in the lettuce rbcL gene (2.44%) and is never used in tobacco rbcL. Native FVIII HC uses the CTC codon as much as 15.28%, but the CTC codon was eliminated from the codon-optimized sequence based on psbA codon usage. More detailed analysis of the codon frequency of the native FVIII HC and the psbA gene reveals further insight into rare codons: GGG for Gly is used 2.3% in psbA but 11.63% in native HC; CTG for Leu is 3.7% in psbA but 26.39% in native HC; CCC for Pro is 1.9% versus 11.9%; CGG for Arg is 0.5% versus 10.81%; and CTG for Val is 1.7% versus 25.49%. So, similar to the CTC codon, several other rare codons in native human genes should have reduced translational efficiency in chloroplasts. In the process of developing the codon optimizer, the cutoff value used for the determination of codons was set at 5% to eliminate rare codons. So, there is room to further modify the codon optimizer program.
New Solution for the Quantitation of Insoluble Multimeric Proteins
A major challenge is the lack of reliable methods to quantify insoluble proteins, because the only reliable method (ELISA) cannot be used due to the aggregation or formation of multimeric structures. CNTB fusion proteins expressed in chloroplasts form pentameric structures that are highly resistant to detergents, and this hampers solubilization due to tight interactions between CNTB monomers, mediated by 30 hydrogen bonds, seven salt bridges, and hydrophobic interactions (Miyata et al., 2012). In our previous studies (Boyhan and Daniell, 2011; Kwon et al., 2013; Kohli et al., 2014; Shil et al., 2014), multimeric forms exist even after treatment with DTT, detergents (SDS), and boiling. Also, acid (pH 2) could not completely dissociate CNTB pentamers due to the reformation of multimeric structures. Although such stability of pentamers is ideal for the oral drug delivery of CNTB fusion proteins, quantitation of the dose continues to be a major challenge.
Delivering accurate doses of protein drugs is a fundamental requirement for their clinical use. Therefore, in this study, we carried out PRM analysis for the absolute quantitation of CNTB-FVIII HC and CNTB-VP1 in plants carrying codon-optimized and native sequences. Limitations in quantitation using western blots, including protein aggregation and inefficient transfer of large proteins to membranes, inadequate solubilization, and differential exposure to films, were quite evident, resulting in unreliable quantification of drug dosage in planta. Use of strong denaturing and reducing conditions in combination with optimal enzymatic proteolysis conditions maximized the solubilization of multimeric CNTB proteins. PRM analysis has been broadly adopted in quantitative proteomics studies (e.g. biomarker discovery in plasma), due to its high sensitivity, specificity, and precise quantitation of specific protein targets within complex protein matrices (Gallien et al., 2012). These qualities clearly show the advantage of using PRM in the quantification of specific protein targets, independent of the protein matrix source (e.g. plant extracts from tobacco or lettuce) or complexity. Moreover, the development of a PRM assay for a handful of proteins can be achieved in a relatively short time and at low cost (not considering the mass spectrometry instrumentation). As a peptide-centric quantitation methodology, it also offers robustness and versatility of protein extraction methods, and keeping the protein of interest in a native conformation is not required. However, it is intrinsically biased by the enzymatic cleavage site access of the enzymes used for digestion. In order to overcome this bias, we used strong denaturing conditions (i.e. 2% SDS) and buffers that favor the activity of the proteolytic enzymes (i.e. sodium deoxycholate-based buffers; León et al., 2013). For FVIII HC (Figs. 5 and 6), there were no significant variations in the values for fold increases of codon-optimized over native sequences, which were determined by the peptides chosen for quantification. In addition, the fold increases were very similar between two different normalization approaches. Three peptides selected from the CNTB region (N terminus of the fusion protein) showed that the range of the fold increase was from 4.5 to 6.6, while the range was 5 to 7.1 for the peptides chosen from FVIII regions (C terminus of the fusion protein). Therefore, quantification results obtained from PRM analysis are consistent, irrespective of the selected region of the fusion protein (N or C terminus) or the component protein (CNTB or FVIII HC).
By using absolute quantified SIS peptides at identical concentrations in all samples and by examining the entire length from N to C terminus, one could accurately quantify the absolute amount of the target protein (Streng et al., 2016). Furthermore, the accuracy of PRM assays in this study was further consolidated by using two different normalization methods: SIS peptides and peptides for housekeeping proteins (large or small subunit proteins of Rubisco and ATP synthase β-subunit). Incomplete/cleaved proteins can be detected using targeted peptide located closer to the C or N terminus or in the midregions. Quantification results obtained from PRM analysis of both CNTB fusion proteins in our study are consistent, irrespective of the selected region of the fusion protein (N or C terminus or elsewhere), and offer data for reliable quantitation. Also, the same three CNTB peptides for CNTB-VP1 showed consistent fold increases, ranging from 22.5 to 28.1. PRM analysis is better than western blotting because it eliminates variation introduced by mobility and the transfer of different-sized proteins and the saturation of antibody probes. Overall, the PRM workflow included selection of the proteotypic peptides from CNTB and FVIII HC sequences and synthesis of the counterpart SIS peptides (Supplemental Fig. S6). Six peptides were selected and scheduled for PRM analysis on the Q Exactive mass spectrometer, based on observed retention time on the chromatograph with a window of ±5 min and mass-to-charge ratio (m/z) of the double and/or triple charge state of these peptides. This double way of targeting the selection of precursor ions, in addition to the high resolution of the Q Exactive mass spectrometer, contributes to the high specificity of the assay. The PRM data analysis, postacquisition, also offers a high specificity to the assay. The five most intense fragment ions, with no clear contaminant contribution from the matrix, are then selected for the quantification of the peptide. The confidence of the fragment ion assignment by the bioinformatics tool used (i.e. Skyline; MacLean et al., 2010) is finally achieved by the comparison of the reference tandem mass spectrometry spectra and the retention time profiles, generated with each of the counterpart SIS peptides. The high sensitivity, specificity, versatility, and robustness of PRM offer a new opportunity for characterizing translational systems in plants.
CONCLUSION
This study explored heterologous gene expression utilizing chloroplast genome sequences, ribosome profiling, and targeted mass spectrometry to enhance our understanding of the synthesis of valuable biopharmaceuticals in chloroplasts. Targeted proteomic quantification by mass spectrometry showed that codon optimization increases translation efficiency 4.5- to 28.1-fold based on the coding sequence, validating this approach, to our knowledge, for the first time for the quantitation of protein drug dosage in plant cells. The lack of reliable methods to quantify insoluble proteins due to the aggregation or formation of multimeric structures is a major challenge. Both biopharmaceuticals used in this study are CNTB fusion proteins that form pentamers, which is a requirement for their binding to intestinal epithelial monosialotetrahexosylganglioside receptors. Such a multimeric structure excludes the commonly used ELISA for the quantitation of dosage. However, delivering accurate doses of protein drugs is a fundamental requirement for their clinical use, and this important goal was accomplished in this study. Indeed, plant biomass generated in this study has resulted in the development of a polio booster vaccine that has been validated by the Centers for Disease Control and Prevention, a timely invention to meet the World Health Organization requirement to withdraw the current oral polio vaccine, which causes severe polio in outbreak areas, in April 2016 (Chan et al., 2016).
Such an increase of codon-optimized protein accumulation is at the translational level rather than any impact on transcript abundance. The codon optimizer program increases transgene expression in chloroplasts in both tobacco and lettuce with no species specificity. In contrast to previous in vitro studies, these in-depth in vivo studies of heterologous gene expression using a wealth of newly sequenced chloroplast genomes helped us to understand the codon optimization process. While the removal of rare codons is very important, replacing those with the most highly used psbA codons indeed decreased translation efficiency. Therefore, the key factor in enhancing translation is the replacement of rare codons following the hierarchy of a highly expressed gene. Ribosome footprints obtained using profiling studies did not increase proportionately with VP1 translation or even decreased after FVIII codon optimization, but it is a valuable tool for diagnosing rate-limiting steps in translation. A major ribosome pause at CTC Leu codons, a rarely used codon in chloroplasts, was eliminated from the native gene after codon optimization. Ribosome stalls observed at clusters of other codons in codon-optimized genes provide opportunities for further optimization. These observations provide further insight into limitations in chloroplast translation and approaches to address these in future studies.
MATERIALS AND METHODS
Codon Optimization
To maximize the expression of heterologous genes in chloroplasts, a chloroplast codon optimizer program was developed based on the codon preference of psbA genes across 133 seed plant species. All sequences were downloaded from the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/genomes/GenomesGroup.cgi?taxid=2759&opt=plastid). The usage preference among synonymous codons for each amino acid was determined by analyzing a total of 46,500 codons from 133 psbA genes. The optimization algorithm (Chloroplast Optimizer version 2.1) was made to facilitate changes from rare codons to codons that are frequently used in chloroplasts using Java.
Creation of Transplastomic Lines
The native sequence of FVIII HC was amplified using the pAAV-TTR-hF8 mini plasmid (Sherman et al., 2014) as the PCR template. The codon-optimized HC sequence obtained using Codon Optimizer version 2.1 was synthesized by GenScript. The native VP1 gene (906 bp) of Sabin 1 (provided by Dr. Konstantin Chumakov, Food and Drug Administration) was used as the template for PCR amplification. The codon-optimized VP1 sequence also was synthesized by GenScript. Amplified and synthetic gene sequences were cloned into chloroplast transformation vectors pLSLF and pLD-utr for lettuce (Lactuca sativa) and tobacco (Nicotiana tabacum ‘Petite Havana’), respectively. Sequence-confirmed plasmids were used for bombardment to create transplastomic plants as described previously (Verma et al., 2008). Transplastomic lines were confirmed using Southern-blot analysis as described previously (Verma et al., 2008), except for probe labeling and detection, for which the DIG High Prime DNA Labeling and Detection Starter Kit II (Roche; catalog no. 11585624910) was used.
Evaluation of Translation
To compare the level of protein expression between native and codon-optimized sequences, immunoblot and densitometric assays were performed using anti-CNTB antibody. For total plant protein, powdered lyophilized plant cells were suspended in extraction buffer (100 mm NaCl, 10 mm EDTA, 200 mm Tris-Cl, pH 8, 0.05% [v/v] Tween 20, 0.1% SDS, 14 mm β-mercaptoethanol, 400 mm Suc, 2 mm phenylmethylsulfonyl fluoride, and proteinase inhibitor cocktail) in a ratio of 10 mg per 500 µL and incubated on ice for 1 h for rehydration. Suspended cells were sonicated (pulse on for 5 s and pulse off for 10 s; sonicator 3000; Misonix) after vortexing (approximately 30 s). After Bradford assay, equal amounts of homogenized protein were loaded and separated on SDS-polyacrylamide gels with known amounts of CNTB protein standard. To detect CNTB fusion proteins, anti-CNTB polyclonal antibody (GenWay Biotech) was diluted 1:10,000 in 1× phosphate-buffered saline + 0.1% Tween 20, and then membranes were probed with goat anti-rabbit IgG-horseradish peroxidase secondary antibody (Southern Biotechnology; 4030-05) diluted 1:4,000 in 1× phosphate-buffered saline + 0.1% Tween 20. For loading controls, protein-blotted membrane was stained with Ponceau S (Sigma; P-3504) prior to immunoprobing with anti-CNTB antibody, and anti-RbcL antibody (Agrisera; AS03 037; 1:5,000) was used on the same blots after stripping anti-CNTB antibody. Chemiluminescent signals were developed on x-ray films, which were used for quantitative analysis with ImageJ software (IJ 1.46r; National Institutes of Health).
Evaluation of Transcripts
Total RNA was extracted from leaves of plants grown in agar medium in a tissue culture room using the easy-BLUE Total RNA Extraction Kit (iNtRON; catalog no. 17061). For the RNA gel blot, equal amounts of total RNA were separated on a 0.8% agarose gel (containing 1.85% formaldehyde and 1× MOPS) and blotted onto a nylon membrane (Nytran SPC; Whatman). For northern blot, the PCR-amplified product from the psbA 5ʹ UTR of the chloroplast transformation plasmid was used as the probe. Hybridization signals on membranes were detected using a DIG labeling and detection kit as described above.
Lyophilization
Confirmed homoplasmic lines were transferred to a temperature- and light-controlled greenhouse. Mature leaves from fully grown transplastomic plants were harvested and stored at −80°C before lyophilization. To freeze dry plant leaf materials, frozen, crumbled small leaf pieces were sublimated under 400-mTorr vacuum while increasing the chamber temperature from −40°C to 25°C for 3 d (Genesis 35XL; VirTis SP Scientific). Dehydrated leaves were powdered using a coffee grinder (Hamilton Beach) at maximum speed; tobacco was ground three times for 10 s each, and lettuce was ground three times for 5 s. Powdered leaves were stored in containers under air-tight and moisture-free conditions at room temperature with silica gel.
Protein Extraction and Sample Preparation for Mass Spectrometry Analysis
Total protein was extracted from 10 mg of lyophilized leaf powder by adding 1 mL of extraction buffer (2% SDS, 100 mm DTT, and 20 mm TEAB). Lyophilized leaf powder was incubated for 30 min at room temperature with sporadic vortexing to allow rehydration of plant cells. Homogenates were then incubated for 1 h at 70°C, followed by overnight incubation at room temperature under constant rotation. Cell wall/membrane debris was pelleted by centrifugation at 14,000 rpm (approximately 20,800 rcf). The procedure was performed in duplicate.
All protein extracts (100 µL) were enzymatically digested with 10 µg of trypsin/Lys-C (Promega) on a centrifugal device with a filter cutoff of 10 kD (Vivacon) in the presence of 0.5% sodium deoxycholate, as described previously (León et al., 2013). After digestion, sodium deoxycholate was removed by acid precipitation with 1% (final concentration) trifluoroacetic acid. SIS peptides (greater than 97% purity, C-terminal Lys and Arg as Lys U-13C6;U-15N2 and Arg U-13C6;U-15N4; JPT Peptide Technologies) were spiked into the samples prior to desalting. Samples were desalted prior to mass spectrometry analysis with OligoR3 stage tips (Applied Biosystems). The initial protein extract (10 µL) was desalted on an OligoR3 stage tip column. Desalted material was then dried on a speed vacuum device and suspended in 6 µL of 0.1% formic acid in water. Mass spectrometry analysis was performed in duplicate by injecting 2 µL of desalted material into the column.
PRM Mass Spectrometry Analysis and Data Analysis
Liquid chromatography-coupled targeted mass spectrometry analysis was performed by injecting the column with 2 µL of peptide, corresponding to the amount of total protein extracted, and digested from 33.3 µg of lyophilized leaf powder, with 34 fmol of each SIS peptide spiked in. Peptides were separated using the Easy-nLC 1000 (Thermo Scientific) on a home-made 30-cm × 75-µm i.d. C18 column (1.9 µm particle size; ReproSil; Dr. Maisch HPLC). Mobile phases consisted of an aqueous solution of 0.1% formic acid (A) and 90% acetonitrile and 0.1% formic acid (B), both HPLC grade (Fluka). Peptides were loaded on the column at 250 nL min−1 with an aqueous solution of 4% solvent B. Peptides were eluted by applying a nonlinear gradient for 4%-7%-27%-36%-65%-80% B in 2-50-10-10-5 min, respectively.
Mass spectrometry analysis was performed using the PRM mode on a Q Exactive mass spectrometer (Thermo Scientific) equipped with a nanospray Flex ion source (Gallien et al., 2012). Isolation of targets from the inclusion list involved a 2-m/z window, a resolution of 35,000 (at m/z 200), a target AGC value of 1 × 106, and a maximum filling time of 120 ms. Normalized collision energy was set at 29. Retention time schedules were determined by the analysis of SIS peptides under equal nano-liquid chromatography. A list of target precursor ions and a retention time schedule are reported in Supplemental Data Set S1. PRM data analysis was performed using Skyline software (MacLean et al., 2010).
Ribosome Profiling
Second and third leaves from the top of the plant were harvested for ribosome profiling. Lettuce plants were approximately 2 months old. Tobacco plants were 2.5 or 2 months old, for native and codon-optimized VP1 constructs, respectively. Leaves were harvested at noon and flash frozen in liquid nitrogen. Ribosome footprints were prepared as described by Zoschke et al. (2013), except that RNase I was substituted for micrococcal nuclease. Ribosome footprints were converted to a sequencing library with the NEXTflex Illumina Small RNA Sequencing Kit version 2 (BIOO Scientific; 5132-03). rRNA contaminants were depleted by subtractive hybridization after first-strand cDNA synthesis using biotinylated oligonucleotides corresponding to abundant rRNA contaminants observed in pilot experiments. Samples were sequenced at the University of Oregon Genomics Core Facility. Sequence reads were processed with cutadapt to remove adapter sequences and bowtie2 with default parameters to align reads to the engineered chloroplast genome sequence.
Accession Numbers
Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers NM_000132.3 for FVIII HC and AY184219 for VP1.
Supplemental Data
The following supplemental materials are available.
Supplemental Figure S1. Sequences of native and codon-optimized FVIII HC and VP1 genes.
Supplemental Figure S2. Comparison of native and codon-optimized (new and old) sequences.
Supplemental Figure S3. Three different codon tables for the expression of heterologous genes in chloroplasts.
Supplemental Figure S4. Plot of integrated density values for the quantification of CNTB-FVIII HC and CNTB-VP1 based on standard curves.
Supplemental Figure S5. Peptide sequences used for targeted mass spectrometry.
Supplemental Figure S6. Comparison of CNTB-FVIII HC and VP1 by PRM analysis.
Supplemental Figure S7. Evaluation of PRM assay linearity.
Supplemental Data Set S1. Codon usage table and mass spectrometry data.
Acknowledgments
We thank Mark Yarmarkovich for help with developing the codon optimization algorithms, Nick Stiffler for help with the bioinformatic analysis of ribosome profiling data, and Non Chotewutmontri for helpful discussions.
Footnotes
The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (www.plantphysiol.org) is: Henry Daniell (hdaniell{at}upenn.edu).
K.-C.K. organized codon tables, created and characterized transplastomic plants, and interpreted and wrote sections of the article; H.-T.C. created and characterized transplastomic plants and contributed data; I.R.L. performed MS and PRM analyses, interpreted data, and wrote this section of the article; R.W.-C. contributed ribosome profiling data analyses; A.B. interpreted ribosome profiling data and wrote this section of the article; H.D. conceived and designed the project, analyzed and interpreted data, and wrote and revised several sections and versions of the article.
↵1 This work was supported by the National Institutes of Health (grant nos. R01 HL107904, R01 HL109442, and R01 EY 024564), the Bill and Melinda Gates Foundation (grant no. OPP1031406 to H.D.), and the National Science Foundation (grant no. IOS–1339130 to A.B.).
↵[OPEN] Articles can be viewed without a subscription.
Glossary
- CNTB
- cholera nontoxic B subunit
- UTR
- untranslated region
- PRM
- parallel reaction monitoring
- DIG
- digoxigenin
- SIS
- stable isotope-labeled standard
- Received June 20, 2016.
- Accepted July 25, 2016.
- Published August 29, 2016.
REFERENCES
