Chloroplast transformation for bioencapsulation and oral delivery using the immunoglobulin G fragment crystallizable (Fc) domain

Proinsulin Like Growth Factor (prolGF1) and myostatin (Mstn) regulate muscle regeneration when intravenously delivered. We set out to test if chloroplast bioencapsulated forms of these proteins may serve as a non-invasive means of drug delivery through the digestive system. We created tobacco (Nicotiana tabacum) plants carrying GFP-Fc1, proIGF-I-Fc1, and Mstn-Fc1 fusion genes, in which fusion with the immunoglobulin G Fc domain improved both protein stability and absorption in the small intestine. No transplastomic plants were obtained with the Mstn-Fc1 gene, suggesting that the protein is toxic to plant cells. proIGF-I-Fc1 protein levels were too law to enable in vivo testing. However, GFP-Fc1 accumulated at a high level, enabling evaluation of chloroplast-made Fc fusion proteins for oral delivery. Tobacco leaves were lyophilized for testing in a mouse system. We report that the orally administered GFP-Fc fusion protein (5.45 μg/g GFP-Fc) has been taken up by the intestinal epithelium cells, evidenced by confocal microscopy. GFP-Fc subsequently entered the circulation where it was detected by ELISA. Data reported here confirm that chloroplast expression and oral administration of lyophilized leaves is a potential delivery system of therapeutic proteins fused with Fc, with the advantage that the proteins may be stored at room temperature.

2 . Myostatin, a member of the transforming growth factor-β (TGF-β) superfamily, is a powerful negative regulator of skeletal muscle mass in mammalian species 3,4 . Nearly all clinical studies or FDA approved treatments for muscular diseases regarding these agents are dependent on needle-based application, the efficacy of which is questionable, particularly as the administration of low levels of growth factors and the limitations of the circulatory system to distribute proteins to muscle tissues are recognized constrains to the benefits of systemic delivery of recombinant proteins like IGF-I [5][6][7][8] . In addition, high cost of fermentation facilities, prohibitively expensive protein purification technologies, cold storage/ transportation and sterile delivery make protein drugs unaffordable for a large global population earning < $2/day 9,10 . Expression and bioencapsulation of recombinant proteins for muscle diseases using chloroplast transformation biotechnology can procure orally available therapies to replace injections. Furthermore, this efficient system has economic potential by obviating costly purification procedures. This may, in turn, increase accessibility of muscle atrophy treatments to a wider population.
The highly conserved Igf1 gene encodes a prepropeptide consisting of a signal peptide to direct secretion, the mature IGF-I peptide, and a C-terminal extension called the E-peptide 11,12 . The predominant Igf1 isoform, Igf1a, constitutes 90-95% of the mammalian Igf1 mRNA transcripts 13,14 . Retention of the E-peptide (proIGF-I) has been shown to enhance receptor binding in vitro compared to the mature IGF-I alone 15 . Further, removal of the initial 3 residues of the mature protein (GPE) increases IGF-IR activity, in part through a reduction of affinity for the family of IGF binding proteins 16 . Thus, we set out to express the pro form of IGF-I with or without GPE, and without the signal peptide.
Myostatin (Mstn) encodes a prepropeptide consisting of a signal peptide, latencyassociated peptide (LAP) and mature growth factor (GF) protein. The LAP, or propeptide, is important for myostatin maturation and proper folding [17][18][19] , but also remains associated with the mature protein after secretion into the extracellular matrix via strong noncovalent bonds that impedes GF activity. Additional processing of the propeptide by a member of the BMP1/Tolloid (TLD) family of metalloproteases destabilizes the propeptide, which results in the full activation of the myostatin GF. A single amino acid substitution in this processing site abolishes recognition by BMP1/TLD-like proteinases thereby preventing myostatin activity, which results in increased muscle mass when injected into mice 20,21 . We sought to express recombinant myostatin propeptide containing a D76A mutation for BMP1 cleavage prevention, thus generating a dominant negative gene product which, upon uptake into the blood stream, is capable of binding mature GF and inhibiting the Smad2/3 signaling cascade that leads to muscle atrophy.
For efficient uptake by intestinal epithelial cells into the bloodstream it is necessary to fuse the chloroplast bioencapsulated proteins to a specific compound. Two systems are currently available and utilize N-terminal fusions to Cholera non-toxic B subunit and the protein transduction domain (PTD) for the expression of recombinant pro-IGF-I which has been codon optimized for expression in lettuce chloroplasts 10,22 . Murine oral gavage trials with these fusions have demonstrated successful uptake into the bloodstream, accumulation in skeletal muscles of treated mice, and improvements if bone fracture healing in diabetic mice. While promising, the N-terminal fusions can impair effective receptor activation if the fusion peptide or remnants of non-IGF-I sequences remain following cleavage. This is because the N-terminus of IGF-I is important for receptor engagement 23 . To avoid this issue, we look to generate an additional system for oral delivery of therapeutic proteins using the Immunoglobulin G (IgG) fragment crystallizable (Fc) domain for fusion of recombinant protein at the C-terminus, and a chloroplast expression system that consistently delivers high expression levels [24][25][26][27] . We generated chloroplast transformation vectors carrying recombinant proIGF-I and Mstn propeptide fused to the IgG fragment crystallizable (Fc) domain to promote absorption through the epithelial lining of the small intestine 28-30 . Transplastomic tobacco lines were generated, and heterologous protein expression was quantified and characterized in each of the generated transplastomic lines. We lyophilized the leaf tissues to concentrate the recombinant protein, then used the lyophilized tissue in murine oral-gavage trials and in vitro activity assays. Only GFP-Fc was expressed at sufficiently high levels to enable testing the orally delivered fusion protein. Our findings suggest chloroplastbioencapsulated therapeutic drug delivery after fusion with the immunoglobulin G Fc domain is a compelling alternative to current strategies.

Results
Vectors for expression of the Fc fusion proteins in chloroplasts. The Tobacco Vaccine Vectors (TVV) used in the study are depicted in Fig. 1. The basic vector TVV1 (Fig. 1a) carries a spectinomycin resistance (aadA) marker gene in a rbcL plastid gene promoter/ terminator cassette (P1/T1). The aadA coding region is translationally fused with cMyc tag to facilitate quantification of aminoglicosyde-3''-adenylyl transferase (AAD). The marker gene is flanked by minimal attB (34 bp) /attP (39 bp) sequences, which are target sites for the PhiC31 phage-site-specific integrase for post-transformation excision of the marker gene 31 . Between the multiple cloning site and attB site we have included a tRNA (trnP), known to be efficiently processed in polycistronic mRNAs 32,33 .
In vector pTVV2 we expressed a green fluorescent protein (GFP) fused with Fc at the C-terminus via 3x glycine linker (GFP-Fc1). Expression of proIGF-I fusion proteins contained either a noncleavable 3xglycine linker combined with deletion of GPE in TVV3 (proIGF-I-Fc1NC), or a cleavable proIGF-I-Fc1 fusion protein with a 2x furin linker and retention of GPE (proIGF-I-Fc1C) in TVV4 (Fig. 1a). Finally, the sequence for murine myostatin propeptide harboring a D76A mutation with Fc peptide at its Cterminus was expressed in TVV5. Each sequence was codon optimized for expression in chloroplasts using the OPTIMIZER program 34 . The synthetic coding region was inserted in the PrrnLatpH/ This/thr cassette. PrrnLatpH is a promoter-leader cassette, obtained by fusing the tobacco plastid rRNA operon promoter with the maize atpH gene leader. The maize atpH leader, in a dicistronic operon, yielded very high levels of GFP (23.5% of Total Soluble Protein or TSP) in potato leaves (pAI3) 35 . We expected therefore, that the ribosomal RNA operon promoter (Prrn) fused with the 5'-UTR of the Zea maize atpH gene leader harboring a PPR10 RNA binding site should elicit a strong translation initiation signal. The cassette also contains a histidine-threonine attenuator (This/thr) to stabilize the 3' UTR of the mRNA 36 . Each Fc1 fusion transgene was inserted into the multiple cloning site of pTVV1, placing it upstream of the selectable spectinomycin resistance gene (aadA). Sterile tobacco leaves were  bombarded with TVV1, TVV2, TVV3, and TVV4 vector DNA. After bombardment the leaves were dissected into 1 cm 2 sections and transferred to RMOP shoot regeneration medium containing 500 mg/L spectinomcyin, where wild type leaf sections bleach out and proliferate very slowly. Tentative transplastomic clones were identified as regenerating green shoots. The green shoots may be spontaneous spectinomycin resistant mutants, or plastid transgenic (transplastomic) events. Spontaneous mutants are resistant to spectinomycin only, while transplastomic events are resistant to both, spectinomycin and streptomycin, the property conferred by the aadA gene 37 . Double resistance was confirmed in over 70% of the evaluated spectinomycin resistant clones for TVV1, TVV2, TVV3, and TVV4. The clones were obtained in 3 bombarded samples, a relatively high plastid transformation efficiency (Table 1).

Characterization of the transplastomic plants.
To achieve uniformity of transformed plastid genomes, shoots were repeatedly regenerated from the leaves of transplastomic plants. The clones are designated by the construct name and serial number. For example, TVV1-3AD is a plant that was obtained by transformation with vector TVV1, and was designated event No. 3. Because the plants regenerated from the initial event often contain mixed wild-type and transformed plastid genome copies, it is routine to regenerate new shoots from the leaves of primary events expecting that new shoots develop from a small cluster of cells which contain only transformed ptDNA copies. Shoots regenerated from the same leaf section are identified by the letters of the alphabet. The addition of AD indicates that the plant TVV1-3AD went through two cycles of consecutive shoot regeneration (Fig.1b).
Chloroplast transformation was confirmed by DNA gel blot analyses in multiple, independent clones by probing EcoRI-digested total cellular DNA. Hybridization with the 1.9-kb rrn16 probe yielded a 3.1 kb hybridization band for the wild type ptDNA, a 2.7 kb band for TVV1, a 4.6 kb band for TVV2 and 4.2 kb bands for TVV3 and TVV4. The EcoRI cleavage site within the multiple cloning site yielded an additional 1.8 kb hybridization band in each transplastomic plant lines. The 3.1 kb band in TVV2-10A plants indicates that wild-type ptDNA is still present indicating that the plant is still heteroplastomic (Fig. 1b, upper panel). However, the wild-type fragments segregated away in the T1 seed generation. Individuals from a single line, indicated in the T1 DNA gel blot analysis, were used to carry out the remaining experiments (Fig. 1b, lower).
The predicted transplastomic fragment sizes for TVV5 were 1.8 kb and 4.1 kb. We obtained a relatively small number of spectinomycin resistant clones (Table 1), but never obtain events with the predicted fragment sizes. We concluded therefore, that the myostatin propeptide-Fc fusion protein is toxic to plant calls and high-level expression is incompatible with the cellular metabolism due to potential interference with the plant brassinosteroid metabolism 38 .
Northern blot analysis was performed to visualize TVV transcripts in the transplastomic lines (Fig. 1c). Total leaf RNA was extracted, separated in agarose gels, transferred to nitrocellulose membranes and then hybridized with the Fc probe and the aadA probe. The transcript size detected by the Fc probe is approximately 1.7 kb in TVV2 and 1.3 kb in TVV3 and TVV4 (Fig. 1c). The absence of hybridizing band larger than 1 kb indicates the absence of readthrough transcripts from the Fc fusion genes due to efficient processing of readthrough transcripts from the upstream genes when trnP is excised. The aadA marker gene probe recognized a 1 kb mRNA in TVV1 through TVV4 lines, as expected (Fig. 1c).
Expression of transgenes in the chloroplasts may interfere with plastid function. Therefore, it was of interest to determine if expression of the aadA marker gene adversely affects biomass accumulation. To address this problem, we grew the transplastomic plants for 45 days in a randomized experimental design, then measured plant height and biomass accumulation. A general phenotypic comparison did not reveal any readily identifiable differences between TVV transplastomic lines and wild type. (Fig. 1d). To further discern potential metabolic differences, height and dry weight measurements were taken from pre-bolting mature greenhouse grown plants equally represented and randomly positioned across 8 randomized blocks. Two transplastomic lines, Nt-pMRR15 and Nt-pMRR18, were used as controls and were previously characterized with high level GFP expression of 29.3% and 12.7%, respectively 26 . A comparison of mean height and dry biomass for each line to wild type shows that only the high levels of recombinant protein expressed in the Nt-pMRR15 plants significantly reduced plant biomass (Fig. 1d).
Inheritance of spectinomycin resistance has been tested in the seed progeny with the unexpected outcome, that expression of spectinomcyin resistance can be most efficiently tested by germinating seed in the absence of sucrose ( Supplementary Fig. S1). The aadA marker gene has also been excised by introducing the Int site-specific integrase via crossing ( Supplementary Fig. S2). Introduction of the site-specific recombinase by crossing was less efficient for the removal of the marker gene than direct transformation (Supplementary on line information).

The expression of Fc fusion proteins in leaves.
Accumulation of the GFP-Fc1 protein was assessed in the protein extracts of transplastomic plants (Fig 2a). For consistency, we prepared protein extracts from the second youngest leaf (#2) of the apical meristem (Fig. 3a). The GFP-Fc1 fusion protein was not visible in the Coomassie stained SDS-PAGE gel containing protein extracts of TVV2 plants (Fig. 2a), therefore a source of GFP protein was needed to generate a standard curve for quantification on Western blots. We chose the pMRR13 transplastomic tobacco line known to accumulate high levels of GFP in the leaf total cellular protein (TSP) 39 . In our hands, the level of GFP in pMRR13 plants was about 27% of TSP (Fig. 2a). Using immunoblots hybridized with the Living Colors primary antibody for the detection of GFP, and MRR13 leaf protein extract as a reference, an average of three biological replicates showed GFP-Fc1 accumulation at 4.24% ± 1.31 of TSP, or 382.47 µg/g fresh leaf material (Fig. 2b). GFP-Fc1 fusion protein displayed two bands, the smaller of which had an electrophoretic mobility consistent with the predicted monomer, and the larger consistent with that of a dimer. Of the two, the dimer was the dominant band in protein samples extracted from young leaves. GFP is monomeric, but Fc forms dimers 40 , explaining the presence of dimers GFP-Fc1.
Protein extracts of TVV2 plants were the source of reference GFP-Fc protein for quantification proIGF-I-Fc1NC and proIGF-I-Fc1C proteins in TVV3 and TVV4 leaf extracts using the Fc Ab (Fig. 2c). Averages of four biological replicates were calculated for each. Extracts of both types of plants displayed two bands, the smaller of which had an electrophoretic mobility consistent with the predicted protein weight of a monomer (~37 kDa) and the larger consistent with that of a dimer. The leaves accumulated 0.20% ± 0.09 and 0.05% ± 0.02 TSP of proIGF-I-FcNC and proIGF-I-Fc1C, the equivalent of 24.07 µg/g and 5.90 µg/g fresh leaf, respectively ( Table 2).
Protein synthesis is active in young, developing leaves and slows down in older leaves. Therefore, unstable proteins are expected to accumulate only in young leaves whereas stable proteins may be present in most leaves. To decide which leaves to harvest for lyophilization we tested protein accumulation in the leaves of plants at the 9-leaf stage before the flower buds appeared (Fig. 3a). TVV2 leaves contained increasing amounts of GFP-Fc1 from the youngest leaf (#1) through the oldest leaf (#6) tested in the fully grown plants (Fig. 3b). In this regards GFP-Fc behaves as FaeG, a highly stable proteinaceous polymer with a capacity to evoke mucosal immune responses 41 . Recombinant protein proIGF-I-Fc1NC accumulated only in leaves 1-4 in TVV3 plants (Fig. 3c), and proIGF-I-Fc1C in TVV4 plants only in leaves 1 and 2 ( Fig. 3d). Therefore, we collected #1 and #2 leaves for lyophilization.

Lyophilization of bioencapsulated Fc fusion proteins.
Lyophilization of transplastomic leaf tissues protects bioencapsulated recombinant proteins from degradation and enables the storage of desired aliquots or "doses" for murine oral gavage trials 42,43 . To create a source of rapidly growing young leaves in a short period of time, flowering plants were cut back to the oldest leaf. The plants then were treated with 14:14:14 slow-release fertilizer pellets. Numerous small shoots quickly emerged from lateral buds. After 2 days all shoot buds were removed with the exception of a single one that was allowed to develop to the 5-leaf stage (Fig. 3e). In 14 days the 2 nd and 3 rd youngest leaves were harvested from 20-25 plants to obtain a total 150 g of fresh leaf material. Freeze-drying 150 g fresh leaf yielded 5 to 9 g dry leaf with an approximate 12% residual water content. Protein-content and quality in fresh and lyophilized leaf material was compared after separating protein extracts in SDS-PAGE followed by Coomassie blue staining. We estimate, based on the signal intensity of the rubisco large subunit, that lyophilization concentrated the protein content on average 15 times (Fig. 3f). Western blots probed with the Fc antibody indicate that the monomeric and dimeric forms of the Fc fusions were preserved during lyophilization and that there was no readily detectable protein degradation during storage (Fig. 3g).
Bioactivity of Pro-IGF-I-Fc1. Even though there was insufficient production of either IGF-I sequence to evaluate oral uptake in mice, it enabled evaluation of IGF-I receptor activity in cell-based assays. IGF-IR phosphorylation was significantly increased with exposure to 10nM IGF-I from TVV3 (Fig. 4). Linear regression analysis revealed that equivalent phosphorylation levels were achieved by 12nM TVV3 as 5 nM recombinant IGF-I, and at 10nM for both sources, TVV3 phosphorylation was 45% of recombinant IGF-I. This assay indicated that the IGF-I produced in chloroplasts was remarkably active, even though production was lower than expected.

Oral uptake and detection in circulation of GFP-Fc in mice.
Adult mice expressing the humanized Fc neonatal receptor (hFcRn) provide a relevant model to test oral bioavailability of Fc fusion proteins 44 . A single dose of GFP-Fc delivered by oral gavage was detectable in circulation as early as 1 hour after delivery, and GFP levels returned to baseline by 18 hours (Fig. 5a). We compared two methods of delivery for a period of 3 days and found that trough levels of GFP in the circulation the following day were similar between oral gavage or dough diet (Fig 5b). Many tissues from the same mice were subjected to GFP ELISA measurements. However, there was no detectable signal found in any of the tissues (data not shown). To determine if oral bioavailability was achieved through transcytosis in the gut, immunostaining for GFP was performed on fixed sections of the jejunum. A subset of cells within the villi were detected with GFP signal surrounding the nucleus (arrow, Fig. 5c), supporting that these cells afforded uptake. Taken together, oral delivery of Fc fusion proteins expressed in the chloroplasts is achievable through gut absorption and paves the way for this strategy to be implemented for therapeutic proteins.

DISCUSSION
Our goal was high-level expression of the Fc fusion proteins in chloroplasts for testing oral bioavailability. Protein accumulation from transgenes depends on mRNA abundance determined by promoter strength and mRNA stability; the efficiency of translation determined by the 5'-UTR. The plastid rRNA operon PEP promoter (Prrn) is the strongest known promoter in chloroplasts 45 , which was linked to the maize psbH 5'-UTR (LatpH Zm ) , containing a PPR10 RNA binding protein target site, the most efficiently translated reading frame in chloroplasts 46,47 . This promoter-leader combination in a dicistronic construct (pAI3) resulted in the accumulation of 23% GFP in potato leaf 35 . The GFP-Fc protein, expressed from the PrrnLatpH Zm 5' regulatory region accumulated only at 4.27% of TSP. Protein models of the GFP-Fc1 fusion show low modeling scores at the interconnecting region between the C-terminus of the GFP and the N-terminus of the CH2 domain of Fc 48 . This region consists of a GGG linker, with the remaining GCKPCICT portion of the Fc hinge, and may represent a region of structural mobility or low structural integrity, both of which may subject the fusion protein to degradation. The non-cleavable proIGF-I-Fc1NC protein (TVV3) and proIGF-I-Fc1 with the cleavable linker (TVV4) accumulated at even lower level, 0.2% and 0.05% TSP. Reduced proIGF-I-Fc1 in TVV4 transgenic lines can be attributed to faster protein degradation due to differences of amino acid sequences in the linker region.
After conversion of 4.27 % TSP to µg/g fresh weight, protein accumulation of GFP-Fc1 measured approximately 0.3825 µg/mg of fresh weight and, after lyophilization, concentrated to 5.45 µg/mg of dry weight. These results are comparable to those obtained in previous reports where accumulation of CTB-GFP, PTD-GFP and DCpep-GFP in lyophilized transplastomic tobacco tissues, measured against a standard curve of purified GFP, was 5.6 µg/mg, 24.1 µg/mg and 2.16 µg/mg of dry weight, respectively 49 . The use of a purified, commercially available GFP as reference protein was also our first choice for quantification of the GFP-Fc1 fusion, however this method inflated accumulation measurements almost 9-fold. Heterologous protein expression at these levels should produce a heavy band in the Coomassie gels and a metabolic burden on the transplastomic plants, as our previous experience shows 26 but neither of which occurred here. We found greater accuracy using the previously characterized MRR13 transplastomic tobacco line as a reference for measuring GFP-Fc1 accumulation.
Protein accumulation of proIGF-I-Fc1 in TVV3 and TVV4 was 24.07 µg/g and 5.90 µg/g of fresh weight, respectively. Lyophilization concentrated the protein to 279.9 µg/g of dry weight in TVV3 and 116.0 µg/g of dry weight in TVV4. These results are comparable to those obtained in previous reports where accumulation of CTB-ProIGF-I and PTD-ProIGF-I in lyophilized transplastomic lettuce tissues was 370 µg/g and 270 µg/g of dry weight, respectively 10 . The difference in protein accumulation in the fresh tissues of TVV3 and TVV4 lines is likely due to differences in protein stability due to structural variations in the linker region, as discussed above.
We report here new TVV vectors with significant improvement over our currently available chloroplast expression vectors. (i) The spectinomycin resistance marker aadA is low impact, because it does not reduce the biomass of transplastomic plants (Fig 1d). We previously reported that the expression of spectinomycin resistance (aadA) marker gene in line Nt-pMRR15 significantly reduces biomass accumulation 26 . We also reported that expression of aadA in a PrrnLatpB promoter / This/thr terminator cassette reduced biomass by about 10%, and that excision of marker gene restored wild-type biomass levels 50 . (ii) Readthrough transcription can generate multiple RNA species from the transgenes producing complex patterns on RNA gel blots, see for example 51 . Inclusion of a tRNA between the marker gene and the gene-of-interest facilitates processing of readthrough transcripts. Indeed, RNA gel blots of plants obtained with the TVV vectors reveal complete processing of mRNA transcripts, simplifying quantitative evaluation of mRNA accumulation (Fig. 1c). (iii) Inclusion of PhiC31 recombinase target sites enable convenient post-transformation removal of the marker genes, as documented here. Transformation of the nuclear genome of T0 transplastomic plants typically results in excision of the marker gene in 100% of plastid genomes by the time a shoot regenerates and DNA gel blot analyses can be performed on the leaves. That is why we were surprised to find partial excision in each of the F1 progeny of TVV plants obtained by pollination with three independent nuclear transgenic lines carrying the Int gene 50 . Partial excision of the marker gene in the F1 generation is probably due to inactivity of the Agrobacterium 2' promoter driving Int expression in the female gametocyte, a problem that can be avoided by using female gametocyte-specific promoters.
Administering biopharmaceuticals by injection has disadvantages, including the need for refrigerated storage and transportation, and the requirement for skilled personnel to administer the medication. Oral delivery of these drugs could eliminate the need for refrigerated storage and transportation and would revolutionize healthcare. A major challenge associated with oral delivery is degradation of biopharmaceuticals by acids and proteases that can be avoided by bioencapsulating proteins the within plant cells in chloroplasts. In the gut, commensal microbes break down the cell wall and release the protein drug into the intestinal lumen from where it should get into the immune or circulatory system. Translocating the released protein requires fusing it with a transmucosal carrier protein.
In the early experiments on plant-made pharmaceuticals, no fusion protein was available for testing. Testing of the non-toxic Tet-C fragment of tetanus toxin was accomplished by mixing tobacco leaf extracts with a small amount of cholera toxin (CT) and applied to the nasal surfaces of mice. CT acted as an adjuvant, and immunization with the leaf extract provided protective immunity to the treated mice 25 .
In the meantime, a number of mucosal carrier proteins have been tested as carriers of plant-made pharmaceuticals. Cholera toxin B (CTB) is the most used fusion partner in chloroplasts 9 . CTB binds GM1 receptors on cell membranes and facilitates delivery of fusion proteins through endocytosis. Protein delivery was first shown using a CTB-GFP fusion protein. In addition to CTB, fusion of the GFP C-terminus with the protein transduction domain (PTD), a small cationic human peptide that can penetrate cell membranes without specific receptors and a dendritic cell peptide (DCpep), were shown to deliver GFP orally 10,49 .
We report here the GFP-Fc fusion protein for oral delivery. Examples of Fcfusion proteins made in plants have been published, including human prostatic acid phosphatase 52 , cocaine hydrolase-Fc 53 , osteopontin-Fc fusion protein 54 and a plantexpressed Fc-fusion protein tetravalent dengue vaccine 55 . These proteins have been expressed from nuclear genes and targeted for secretion via the ER, where Fc is glycosylated. One consideration in using Fc-fusion proteins is the role of glycosylation in modulating Fc receptor binding and specificity. The Fc Asn 297 is glycosylated but is within the core of the Fc dimer. Non-glycosylated Fc remains stable and folds properly, but the dimer formed by Fc domains lacking Asn 297 glycosylation has a tighter conformation, and this can reduce FcRn binding by ~10 fold 56 . Plant and metazoan glycosylation differ 57 , leading to possibly diminished FcRn binding and absorption. Glycosylation is absent in chloroplasts. We expect that, when high level chloroplast expression is obtained, high level expression compensates for reduced stability caused by the lack of glycosylation. This is particularly relevant in case of proIGF-Fc, because proIGF is more active when not glycosylated. The answers to these open questions will be delivered by future experimentation.

MATERIALS AND METHODS Plant material and growth conditions.
Experiments were carried out using Nicotiana tabacum L. cv. Petit Havana plants. The tobacco seed was a gift of Canada Department of Agriculture, Ottawa Research Station to the Institute of Genetics, Hungarian Academy of Sciences, Budapest. PM received the seed from the institutional stock in 1970 and maintains it since. Seed is also available from Lehle Seed, Round Rock, TX 78681, United States. Seed is available for free and without any restriction upon request to PI, after paying for postage.
Seedlings were germinated under aseptic conditions from surface-sterilized seeds under 16h illumination at 28°C using cool-white, fluorescent tubes (2,000 lx, CXL F025/741), followed by an 8h dark period at 21°C. Seeds germinated under greenhouse conditions were illuminated for 16 hrs.

Plastid transformation vector construction.
The vector backbone in which all subsequent vectors were created is pPRV1-II 58 , a PUC119/pZS192 derivative 59 containing N. tabacum plastid sequences to facilitate homologous recombination between the plastid trnV gene and the 3'rps12/7 promoter. TVV1 carries the spectinomycin resistance (aadA) 37 selection marker with a C-terminal c-Myc tag facilitating detection of AAD on protein gel blots. The selectable marker is expressed in a rbcL promotor 60 and psbA terminator 61 cassette in which the trnP from the Medicago trunculata plastid genome (AC093544 ) is incorporated between the multiple cloning site and the aadA-cMyc gene to facilitate efficient cleavage of potential readthrough transcripts 32,33 . The resistance marker is also flanked by minimal attB (34 bp) /attP (39 bp) sites 31 for posttransformation excision by the PhiC31 phage-site-specific integrase 62 . Vector TVV2 contains the eGFP-Fc1 fusion gene with a non-cleavable 3x glycine linker; TVV3 carries a proIGF-I-Fc1 fusion with a non-cleavable 3x glycine linker; TVV4 carries a proIGF-I-Fc1 fusion with a cleavable 2x furin linker; TVV5 carries the myostatin propeptide-Fc1 fusion. The genes of interest in TVV2, TVV3, TVV4, and TVV5 are expressed from the tobacco plastid ribosomal RNA PEP promoter (Prrn) 63 fused with the Zea maize atpH 5'-UTRs, referred to as leader sequences (L) 46 . Termination is controlled by a histidinethreonine attenuator. TVV1 was generated through a single insertion into the ScaI site of pPRV1-II and the TVV2, TVV3, TVV4, and TVV5 were cloned as a SacI-HindIII fragment into the multiple cloning site of TVV1. The schematic map of the vectors is shown in Fig. 1a. DNA sequences are listed in Supplemental Table S1.
Transformation of the tobacco plastid genome. Tobacco plastid transformation was carried out as described 58,64 . Briefly, a Bio-Rad PDS-1000 biolistic gun and Hepta Adaptor was used to bombarded sterile tobacco leaves with 0.6 µm gold particles coated in vector DNA. Leaves transferred to selective RMOP shoot regeneration medium containing 500 mg/L spectinomycin form pigment deficient callus. Transplastomic events express the selective marker gene and emerge as green shoots from calli. Primary events were subjected to one or two additional rounds of regeneration on spectinomycin medium to dilute out residual wild-type copies of the plastid genome Transplastomic events are distinguished from spontaneous spectinomycin resistant mutants by streptomycinspectinomycin resistance while spontaneous spectinomycin resistant mutants are sensitive to streptomycin. Transgene integration into the plastid genome is confirmed by Southern blot analyses. In genetically stable homoplastomic plants all wild type chloroplast genome copies are absent. The shoots were rooted on hormone-free medium and transferred to the greenhouse.

PCR analysis.
PCR was used for the initial identification of transplastomic events (T0). The forward primer, LTRF4 5'-CACCACGTCAAGGTGACACT-3,' is in the trnV gene of the left targeting region of the chloroplast genome target site. The reverse primer, aadAR2 5'-GTTGAGTCGATACTTCGGCG-3,' is in the aadA coding region. Expected band sizes for transplastomic events are as follows: 0.5 kb for TVV1, 2.3 kb for TVV2, 1.9 for TVV3, 2.0 kb for TVV4, and 2.3 kb for TVV5. A single EarI restriction digest site within the TVV4 amplicon allows for further discernment from TVV3. Table 3 contains primers used to verify transplastomic events, generate probes, and analyze marker excision.
Compliance with the digital image and integrity policies. Where gels/blots are used in figures , we have checked their compliance with the digital image and integrity policies of the journal (https://www.nature.com/srep/journal-policies/editorial-policies#digitalimage).

Southern blot analysis.
DNA gel blot analysis was carried out as described 37 . Briefly, the cetyltrimethylammonium bromide (CTAB) protocol was used to extract total cellular DNA from leaf tissue. Two micrograms of total leaf DNA was digested with the EcoRI restriction endonuclease followed by separation on 1% agarose gels. The DNA was then transferred to Hybond-N membranes (GE Healtcare RPNBL/02/10) by capillary blotting. Two probes, the plastid rrn16S and aadA, were prepared by random-primed 32 P labeling using the Takara Bio. Inc., kit (Cat. No. 6045). The rrn16S probe was the 1.9 kb ApaI/BamHI ptDNA fragment encoding part of the 16S rRNA gene, while the aadA probe was a 0.7-kb fragment amplified from the aadA coding region using primers aada Fw 5'-GTTGCTGGCCGTACATTTG-3' and aadA Rv 5'-TCGCCTTTCACGTAGTGGAC-3'.

Northern blot analysis.
Leaves were frozen in liquid nitrogen and total cellular RNA was isolated using the Qiagen RNeasy Plant Mini Kit (Cat. No. 74904). The RNA was separated in a 1.5% formaldehyde gel and transferred to Hybond-N membranes by capillary blotting (GE Healthcare RPNBL/02/10). The probes were: aadA, the same which was used for Southern blot analysis, and Fc1, a 0.6-kb fragment amplified from the Fc1 coding region using primers Fc1-F1 5'-ATGCAAACCCTCGTGCAGTA-3' and Fc1-R1 5'-GCAAGCCCTGTATTTGTACGG-3'. The probes were prepared by randomprimed 32 P-labeling using the Takara Bio. Inc., kit (Cat. No. 6045).

SDS-PAGE and immunoblot analysis.
Leaves were harvested from plants grown under greenhouse conditions and frozen in liquid nitrogen. To obtain total soluble protein (TSP), about 100 mg of leaf material was pulverization with stainless steel grinding balls (SPEX Sample Prep, Metuchen, NJ) and suspended in 100 µl of extraction buffer containing 50 mM HEPES-KOH (pH 7.5), 10 mM potassium acetate, 5 mM magnesium acetate, 1 mM EDTA, 1 mM dithiothreitol, 2 mM phenylmethylsulfonyl fluoride, and protease inhibitor cocktail (30 µl/mL) (Sigma-Aldrich, Cat. No. P9599). The suspension was centrifuged at 14,000g for 5 min at 4 o C and the supernatant was transferred to a new Eppendorf tube. This step was repeated once more. To analyze lyophilized material, 10 mg of lyophilized leaf power was rehydrated in 100 μL of extraction buffer. Protein concentration was determined with the Bio-Rad Bradford Protein Assay (Cat. No. 5000002). The proteins were separated in urea PAGE or Bis-Tris PAGE before transfer to an Immun-Blot PVDF membrane (Bio-Rad, Cat. No. 1620177).
Detection of EGFP required the Living Colors® A.v. Monoclonal Antibody (JL-8) (Takara, 632381) as the primary antibody and the AntiMouse IgG Peroxidase antibody (Sigma-Aldrich, A4416-1) as the secondary antibody. For the detection of Fc1, the Anti-Mouse IgG1 Rabbit Monoclonal Antibody (RevMAb Biosciences, 31-1002-00) was used as the primary antibody and the Goat Anti-Rabbit IgG Peroxidase antibody (Sigma-Aldrich, A0545-1) as the secondary antibody. Detection of the AAD-cMyc tag required the c-Myc Antibody (9E10) (Santa Cruz Biotechnology, sc-40) as the primary antibody and the Anti-Mouse IgG Peroxidase antibody (Sigma-Aldrich, A4416) as the secondary antibody.

Lyophilization.
Plants grown to the 9-leaf stage were cut down to the basal node and regenerated to the 5-leaf stage. The second and third leaves from the top were harvested from the regenerated plants, washed in soapy water, and rinsed 5 times before removal of the central vein and subsequent slicing into 1-2 cm 2 pieces. 150 g of leaf material was placed in a tray and frozen at -80 o C before being lyophilized in the VirTis Genesis 25EL Pilot Lyophilizer vacuum (10 mTorr

Biomass assay.
T1 seed was germinated under greenhouse conditions and transferred two times: first to 1"x2" cells and then to 4"x4" pots. The larger pots were situated in a randomized block design of 8 blocks where a table represented a block containing 14 randomized plants: two individuals from each of the 4 successful vector lines, two wild type plants, and two plants from 2 control lines previously characterized 65 : pMRR15, and pMRR18. Height and biomass data was collected when the flower buds appeared. Height was measured from the base of the soil line to the base of leaf no. 1 (Fig. 3a, e). Then the plants were cut down at the soil line and the entire shoot was placed in a paper bag and dried in an oven at 80 o C for 5 days. Fresh weight and dry weight measurements were taken before and after drying.
Mouse studies. Animal studies were approved by the University of Florida Institutional Animal Care and Use Committee (IACUC), and the experiments were performed in accordance to relevant guidelines and regulations. In addition, animal studies were reported in accordance with ARRIVE guidelines. Male and female mice with ablation of murine FcRn and transgenic expression of the human FcRn regulated by the native human promoter (FcRn -/-hFcRn (32)Tg, Jax #014565) 44 were utilized. Oral bioavailability of GFP-Fc1 fusion proteins was evaluated in male and female hFcRn Tg mice (all age 10 -14 weeks). Initial testing used 20mg lyophilized plant TVV2 expressing GFP-Fc1 suspended in 200uL PBS as a single bolus delivered by oral gavage after a 4 hour fast (N=5). Blood was collected from the facial vein prior to gavage, and at 1, 2, 4, 8, 18, and 24 hours after oral gavage. Blood was centrifuged to collect serum for GFP quantification and stored at -80C until analysis.
Subsequent testing compared oral gavage to ingesting lyophilized plant mixed into a dough diet (Bioserv Transgenic Dough Diet (S3472)). The oral gavage procedure was performed as described above for 3 days using TVV1 (vehicle) (N=4) or TVV2 (GFP-Fc1) (N=8). For the dough diet, mice were familiarized to the diet provided in 300 mg pellets including 20mg TVV1 plant material for 5 days. The following week, mice were given TVV2 (15 mg TVV2/300 mg dough, 1 pellet/day) for 3 days (N=6), with N=2 mice receiving no treatment. For all groups, mice were euthanized 18-20 hours after the final dose using exposure to carbon dioxide followed by cervical dislocation in accordance with American Veterinary Medical Association's (AVMA) Guidelines for the Euthanasia of Animals. Blood, brain, heart, liver, kidney, small intestine, and skeletal muscles (diaphragm, soleus, tibialis anterior, quadriceps and extensor digitorum longus) were collected, rapidly frozen in liquid nitrogen, and stored in -80C for GFP quantification.
GFP Detection in small intestine. hFcRn Tg mice were subjected to a single dose of TVV2 by oral gavage as described above and were euthanized 4 hours after TVV2 delivery. Following euthanasia, the jejunum of the small intestine was removed, flushed with PBS, and then fixed overnight in 4% paraformaldehyde at 4C. Tissues were incubated serially in 10% and 20% sucrose, surrounded in optimal cutting temperature compound (Sakura, Torrance, CA), and frozen in liquid nitrogen-cooled isopentane for histological analysis. 10 µm cryosections were subjected to immunostaining for GFP. Sections were washed in PBS 3 times, with 10 minutes each wash, followed by permeabilization in 0.5% Triton-X in PBS. Following blocking in 5% bovine serum albumin (BSA) in PBS for 1 hour at room temperature, sections were incubated overnight at 4 o C with primary antibodies diluted in 5% BSA for GFP (1:3000 rabbit pAb anti-GFP, Cat#ab6556, Abcam). After washing slides in PBS 3 times, with 10 minutes each wash, tissue autofluorescence was quenched with 0.1% Sudan Black incubation with subsequent washes. Next, sections were incubated for 1 hour at room temperature in the dark in secondary antibodies diluted in 5% BSA (1:1000 Alexafluor 568 IgG anti-rabbit (#A11036) (Invitrogen)). Sections were washed again in the dark (PBS 3 times, with 10 minutes each wash.), then airdried, and covered with mountant (ProLongTM Diamond Antifade with Dapi, Cat#P36962, ThermoFisher) and cover slip. Samples were visualized with a Leica STELLARIS confocal microscope (Leica Microsystems, Buffalo Grove, IL, USA). Images were acquired and processed with the Leica Application Suite and Microscope Imaging software (Leica Microsystems).

IGF-IR activation Assay.
To evaluate the potency of IGF-IR activation by IGF-I produced by TVV3 plants, a kinase receptor assay (KIRA) was performed as previously described . Briefly, 2.5 × 10 4 P6 cells, which overexpress IGF-IR (kind gift from Dr Renato Baserga, Thomas Jefferson University, Philadelphia, Pennsylvania) were seeded into 96-well plates. They were maintained in growth media supplemented with 200 μg/mL G418. The cells were serum starved for 6 hours, and then treated for 15 min with protein extracts obtained from lyophylized plants. Controls included P6 cells treated with protein extracts from GFP-Fc1/TVV2, or with recombinant IGF-I (0.5-10 nM). The P6 cells were lysed and IGF-IR was captured onto an ELISA plate coated with an antibody to IGF-IR (MAB1120, Millipore Corp.). A horseradish peroxidase-conjugated antibody to phosphorylated tyrosines Millipore Corp.) and TMB substrate (N301, Thermo Scientific, Rockford, Illinois) were used for colorimetric quantification. Absorbance was read at 450 nm via the SpectraMax M5 plate reader (Molecular Devices, Sunnyvale, California). Measurements were done in technical triplicates, and performed in entirety twice.  Quantification of GFP-Fc in the TVV2 line on immunoblots, using GFP Ab and a dilution of pMRR13 extract as reference. GFP-Fc abundance was measured against GFP in the MRR13 reference line (four biological replicates; mean ± 1 SD) with the ImageJ software (version 2.0.0-rc-38/1.50b) (c) Quantification of proIGF-I-GGG-Fc1 in the TVV3 line and proIGF-I-2xFurin-Fc1 in TVV3 line on immunoblots, using Fc Ab and a dilution of TVV2 extract as reference. Protein concentrations were calculated with the ImageJ software (version 2.0.0-rc-38/1.50b; four biological replicates; mean ± 1 SD). LSU, rubisco large subunit; SSU, rubisco small subunit. (f) Comparing fresh and lyophilized (Lyo) protein extracts after separation in polyacrylamide gels. Proteins were separated in Bis-Tris MES SDS polyacrylamide gel (12.5%) and stained with Coomassie brilliant blue. Protein content in dried leaf relative to fresh protein extracts was estimated by comparing LSU signal intensity in Coomassie stained gels (three to five replicates; mean ± 1 SD). Fold differences in LSU signal intensity were quantified with ImageJ software (version 2.0.0-rc-38/1.50b). (g) Western blot probed with the Fc antibody was used to detect the physical state of lyophilized protein. Western blots of protein extracts from two different lyophilization batches (lyo1, lyo2) are shown.  . Table 1. Frequency of streptomycin resistant / spectinomycin resistant transplastomic lines among spectinomycin resistant clones   (b) DNA gel blot confirms transgene integration into the chloroplast genome and the absence of non-transformed ptDNA copies. Total cellular DNA was digested with the EcoRI restriction endonuclease and probed with the 1.9-kb ApaI-BamHI fragment (Fig 1a).