The small cellulose-binding domain (sCBD) derived from Trichoderma reesei (AEP40512.1), SUMO domain (amino acids 21-97) of bdSUMO (XP_003564931.1), bdSENP1 (catalytic region; amino acid positions 242-481, XP_003567671.1), green fluorescent protein (GFP), hIL6 (mature region from amino acid positions 30-212, P05231), hLIF (active region from amino acid positions 23-202, P15018), and coat protein p38 (P06663) were codon-optimized based on the codon usage of N. benthamiana and were chemically synthesized (Bioneer corp., Daejeon, Korea).

The fusion constructs sCBD:bdSUMO : GFP and sCBD:bdSUMO:hIL6 were generated by overlap PCR using primers sCF/suR/GF/GR, and sCF/suILR/ILF/ILR ( Supplementary Table S1 ), respectively. The final PCR products were digested with BamHI and XhoI restriction endonucleases and introduced into the 326-plant expression vector (Kim et al., 2013) to give p326-sCBD:bdSUMO : GFP (p326-BsCSGH) ( Figure 1A ) and p326-sCBD:bdSUMO:hIL6 (p326-BsChIL6H) ( Supplementary Figure S1A ). His:bdSENP1 and bdSENP1:HA for expression in plants were generated by PCR amplification using the primer sets HisSF-1/SR-2 and SF-3/SHaR-4 ( Supplementary Table S1 ), respectively. The PCR products were digested with BamHI and XhoI restriction endonucleases and introduced into the plant expression vector, 326-GFP which had been digested with the same restriction endonucleases, to give p326-His:bdSENP1 ( Figure 1A ) and p326-bdSENP1:HA. The synthetic hLIF gene was incorporated to the C-terminus of M:CBM3:bdSUMO by overlapping PCR using primers csLF-1, csLR-2, csLF-3, and csLR-4 ( Supplementary Table S1 ) using 326-M:CBM3:bdSUMO:hIL6 as a template (Islam et al., 2019a). Six histidine residues (His) were added at the N-terminus of hLIF (His:hLIF) during PCR amplification. The final overlapping PCR product, M:CBM3:bdSUMO : His:hLIF (MCS:hLIF), and PCR amplified His:hLIF were digested with BamHI and XhoI restriction endonucleases and introduced at the same site in the plant expression vector, 326-M:CBM3:bdSUMO:hIL6 to give 326-MCS:hLIF and 326-His:hLIF, respectively. The synthetic p38 gene was PCR amplified using the primer set PF-1 and PR-2 ( Supplementary Table S1 ) and ligated into 326-MCS:hLIF vector after digestion with BamHI and XmaI restriction endonucleases to generate the 326-p38:bdSUMO:hLIF. An ER retention signal, His-Asp-Glu-Leu (HDEL) (Gomord et al., 1997), was also added to the C-terminus of the coding sequence of all expression constructs during PCR amplification.

To binary expression constructs, 326-MCS:hLIF, 326-His:hLIF, 326-p38:bdSUMO:hLIF, and 326-bdSENP1:HA constructs were digested with PstI and EcoRI restriction endonucleases and ligated to a binary construct p1300-C3bdSU-hIL6 (Islam et al., 2019a) digested with the same restriction endonucleases to give p1300-MCS:hLIF, p1300-His:hLIF, p1300-p38:bdSUMO:hLIF and p1300-bdSENP1:HA, respectively ( Figure 2A ). For the expression of genes in plants, cauliflower mosaic virus (CaMV) 35S promoter with the double enhancer (d35S), 5′ untranslated enhancer region (5′ UTR) (Kim et al., 2013), the BiP leader sequence (BAA13947; amino acid positions 1 to 34), and the HSP terminator from Arabidopsis thaliana (Nagaya et al., 2010) were used. The nucleotide sequences of all constructs were confirmed by DNA sequencing (Macrogen Inc., Seoul, Korea).

Plasmid DNA of p326-BsCSGH, p326-BsChIL6H, and p326-His:bdSENP1 was isolated by the Qiagen plasmid DNA isolation kit (Qiagen, CA, USA) and used for transformation into Arabidopsis protoplasts by polyethylene glycol (PEG)-mediated transformation (Yoo et al., 2007).

The binary plant expression vectors p1300-MCS:hLIF, p1300-His:hLIF, p1300-p38:bdSUMO:hLIF, and p1300-bdSENP1:HA were transformed into Agrobacterium tumefaciens (A. tumefaciens) EHA105 strain by electroporation. A. tumefaciens EHA 105 carrying p1300-MCS:hLIF, p1300-His:hLIF, p1300-p38:bdSUMO:hLIF, or p1300-bdSENP1:HA, or A. tumefaciens harboring p1300-p38 of the Turnip crinkle virus silencing suppressor ( Supplementary Figure S5 ) (Qu et al., 2003) were cultured in YEP liquid medium consisting of 1% yeast extract, 1% tryptone peptone, 0.5% NaCl (pH 7.0), supplemented with 50 µg/ml kanamycin and 50 µg/ml rifampicin at 27~28°C for 24–36 h by constant shaking at 200 rpm. A. tumefaciens cells were pelleted by centrifugation (3,500 × g for 10 min) and resuspended in the infiltration buffer (pH 5.6) containing 10 mM MgCl₂, 10 mM MES, and 200 µM acetosyringone at 0.8 of OD₆₀₀. For coexpression of multiple genes, two or three bacterial suspension cultures were mixed at a 1:1 ratio (v/v). N. benthamiana leaf tissues (5 to 7-week-old) were infiltrated with 1 ml syringe without a needle or either using the vacuum infiltration system with the mixture of A. tumefaciens cultures (Marillonnet et al., 2005). Subsequently, plants were further grown for 3 to 7-days post infiltration (DPI) in the greenhouse under the standard conditions.

To prepare total soluble protein (TSP), infiltrated N. benthamiana leaf tissues were ground to a fine powder in liquid nitrogen and homogenized in protein extraction buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM DTT, 0.1% [v/v] Triton X-100 and protease inhibitor cocktail). TSP was clarified by centrifugation (13,000 × g) for 20 min, and protein concentrations were determined by the Bradford method (Bradford, 1976) (Bio-Rad, Hercules, CA, USA).

TSP or purified proteins were separated using 10–12.5% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) under the reducing condition, and the proteins on the gels were either stained with 0.25% Coomassie blue R-250 (AMRESCO, cat. no: 6104-59-2, Zottegem, Belgium) solution or analyzed by Western blotting using mouse anti-His (Qiagen, Valenica, CA, USA), mouse anti-LIF (abcam, cat. no., ab34427, Cambridge, UK), mouse anti-GFP (Clontech, cat. No., 632381, CA, USA), rat anti-HA (Roche, cat. No., 1583 816, Basel, Switzerland), mouse anti-IL6 (abcam, cat. no., ab9324, Cambridge, UK), and anti-CBD (Bioapp, Pohang, Korea) antibodies. The horseradish peroxidase (HRP)-conjugated sheep anti-mouse IgG (Bethyl, cat. no., A90-146P, TX, USA), and goat anti-rat IgG (Bethyl, cat. no., A110-105P, TX, USA) were used as secondary antibody. Bands on the immunoblots were detected using chemiluminescence detection reagents (GE healthcare, Illinois, USA), and images were captured with a LAS 4000 image capture system (Fujifilm, Tokyo, Japan).

TSP containing His-tagged hLIF was clarified by centrifugation (15,000g × 30 min) at 4°C. His:hLIF was purified using an Ni²⁺–NTA agarose column (Qiagen, CA, USA) according to the manufacturer’s instructions. Finally, eluted fractions containing His:hLIF were concentrated using a 10 K centricon centrifugal filter (Millipore, cat. no. 4304, Darmstadt, Germany). During the concentration process, the elution solution was replaced with PBS buffer.

Purified His:hLIF (5 µg) was separated via 12% SDS-PAGE under the reducing condition and electroblotted onto polyvinylidene difluoride (PVDF) membrane (Millipore, Darmstadt, Germany). The PVDF membrane was stained with 0.02% Coomassie Blue R-250 until protein bands were visualized clearly. Proteins between 30 and 42 kDa bound to PVDF membrane were subjected to Edman degradation using an ABI 492 Procise Protein Sequencer System (Applied Biosystems, CA, USA).

To remove the N-glycan moiety of His:hLIF, purified His:hLIF was treated with endo-β-N-acetylglucosaminidase H (Roche, Cat. No. 11088726001, Basel, Switzerland). Briefly, 2 µg of purified hLIF was denatured using glycoprotein denaturing buffer (0.5% SDS and 1.0% β-ME) at 100°C for 10 min and then treated with 1 µl (0.005 U) of Endo-H at 37°C for 4 h.

The level of endotoxin in the purified hLIF was determined by Pierce’s LAL Chromogenic Endotoxin quantitation kit (Thermo Fisher Scientific, Cat. no. 88282, MA USA). The three biological replicates were used for endotoxin determination. The standard curve was generated using a commercial E. coli endotoxin standard provided by the kit manufacturer.

Mouse embryonic stem cells (mESCs) (ATCC, SCRC-1002) were maintained on mitomycin C-treated MEF feeder cells. Mitotically inactivated MEFs were seeded at a density of 1.1 × 10⁵ cells per 12-well plate, and the next day, mESCs were seeded at a density of 1.4 × 10⁵ cells per 12-well plate. mESCs were incubated in DMEM (Thermo Fisher Scientific, MA, USA) containing 200 mM Ala-Gln (Sigma-Aldrich, Missouri, USA), 1× minimum essential media-nonessential amino acids (Thermo Fisher Scientific, MA USA), 100 µM 2-mercaptoethanol (Thermo Fisher Scientific, MA USA), 1% penicillin/streptomycin (Thermo Fisher Scientific, MA USA), 15% FBS (Millipore, Darmstadt, Germany), 10 ng/ml commercial LIF (Merck Millipore, MA, USA), or 10 ng/ml His:hLIF (phLIF) derived from plant. The medium was changed every day, and mouse mESCs were passaged every 2 days.

Dissociated mouse mESCs were plated on the gelatin-coated dish and incubated at 37°C for 15 min to deplete feeder cells. Total RNA was extracted from feeder-depleted mouse ESCs using RNeasy Mini Kit (Qiagen, CA, USA). Total RNA (2 μg) was used to prepare cDNA using high capacity cDNA Reverse Transcription Kit (Applied Biosystems, CA, USA). qRT-PCR was performed on a 7500 ABI Real-time PCR system (Applied Biosystems, CA, USA) using the Power SYBR Green PCR master mix (Applied Biosystems, CA, USA) to check the expression levels of pluripotent marker genes Oct4, Sox2, and Nanog. Gene expression was normalized to the housekeeping gene Hprt. The PCR mixture (20 μl) contained 200 ng of template, 0.5 μM forward and reverse primers, and 1× SYBR master mix. The PCR conditions were as follows: initial denaturation at 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. To confirm specific amplification, a melting curve was generated by heating at 95°C for 15 s and then at 60°C for 1 min and then increasing the temperature by 0.3°C every 15 s up to 95°C. All primers are listed in Supplementary Table S2 . The data were collected from six independent experiments, and the technical replicates were repeated three times. Statistical analysis was performed by one-way analysis of variance (ANOVA) with Tukey’s post-hoc multiple comparison analysis (GraphPad Prism 6).

The SUMO-specific protease (bdSENP1) from B. distachyon was selected to serve as the basis for a protocol for the removal of foreign domains from target proteins in vivo. This protease is highly active and leaves no extra residues attached to the target protein after cleavage. Although bdSENP1 functions in the cytoplasm (Bailey and O’Hare, 2004; Hickey et al., 2012), we intended target proteins to accumulate in the ER. The ER localization renders recombinant proteins to be subject to posttranslational modification such as N-glycosylation and disulfide bond formation, which are critical for protein folding, stability, and functionality (Gomord and Faye, 2004). In addition, targeting of recombinant proteins to the ER significantly improved production yields in plants (Schouten et al., 1996; Nausch et al., 2012; Kang et al., 2018). Therefore, we first determined whether bdSENP1 was capable of cleaving its substrate in the ER in vivo. We generated a bdSENP1 expression construct containing a His×6 tag, His:bdSENP1 (BiP : His:bdSENP1:HDEL) ( Figure 1A ) and a bdSENP1 substrate construct, BsCSGH (BiP:sCBD:bdSUMO : GFP:HDEL), that incorporated the leader sequence of an ER chaperone BiP, a small cellulose-binding domain (sCBD) derived from Trichoderma reesei, the bdSUMO domain from B. distachyon, and green fluorescent protein (GFP) tagged with HDEL, an ER retention sequence ( Figure 1A ).

We used PEG-mediated transformation (Yoo et al., 2007) to examine bdSENP1 activity in the ER in Arabidopsis thaliana (A. thaliana) protoplasts. Protoplasts were transformed with either BsCSGH alone or BsCSGH plus His:bdSENP1; in addition, protoplasts were transformed with BiP : GFP:HDEL, which expressed GFP alone in the ER lumen, as a positive control (PC). A fluorescence microscope was used to determine the expression and subcellular localization of BsCSGH. Protoplasts transformed with BsCSGH displayed strong GFP signals in a net-like pattern, an indication of ER localization ( Figure 1B ), strongly suggesting that BsCSGH was targeted to this organelle. To determine the expression at the biochemical level, total soluble proteins were extracted from protoplasts and analyzed by western blotting with an anti-GFP antibody ( Figure 1D ); extracts from untransformed protoplasts were included as a negative control. A 42 kDa band was detected by the anti-GFP antibody in extracts in which BsCSGH alone was expressed, indicating a BsCSGH-specific protein band; by contrast, when His:bdSENP1 was coexpressed with BsCSGH, a 29 kDa protein was detected that was the same size as the GFP used as the control. These results indicate that bdSENP1 could digest the BsCSGH proteolytically by recognizing the bdSUMO domain when both were expressed in the ER ( Figure 1D ). Moreover, western blot analysis of protein extracts using anti-CBD antibody detected a 13 kDa band, corresponding to the sCBD:bdSUMO fragment ( Figure 1E ), further confirming the proteolytic cleavage of BsCSGH by bdSENP1 in the ER. Expression of His:bdSENP1 in the protoplasts was confirmed by western blotting with anti-His antibody ( Figure 1C ); this detected a 28 kDa protein band at the predicted position for His:bdSENP1. These findings suggested that when bdSENP1 was localized to the ER, it recognized the bdSUMO domain and cleaved the protein at the C-terminus of bdSUMO cleavage site.

To further validate this finding, we used another chimeric construct to express human interleukin-6 (hIL6) as the target protein. hIL6 is a cytokine and plays an important function as both a proinflammatory and an anti-inflammatory (Kishimoto, 2010). We previously used a chimeric protein (BiP:M:CBM3:bdSUMO:hIL6:HDEL) and in vitro cleavage by His:bdSENP1 expressed in Escherichia coli to produce tag-less hIL6 (Islam et al., 2019a). For the current study, we generated a chimeric construct BiP:sCBD:bdSUMO:hIL6:HDEL (BsCShIL6H) ( Supplementary Figure S1A ). BsCShIL6H was transformed into A. thaliana protoplasts, either alone or together with His:bdSENP1, and protein extracts were analyzed by western blotting with anti-IL6 and anti-CBD antibodies. Both antibodies detected a protein band at 34 kDa, which corresponded with full-length BsCShIL6H expressed alone ( Supplementary Figures S1B, C ). By contrast, following the coexpression of His:bdSENP1 with BsCShIL6H, anti-IL6-antibody detected a 21 kDa band, which matched the known size of hIL6 ( Supplementary Figure S1B ), whereas anti-CBD antibody detected a 13 kDa band that corresponded with the sCBD:bdSUMO fragment ( Supplementary Figure S1C ). Thus, bdSENP1 could act in the ER to cleave proteins containing the bdSUMO domain precisely.

To test further the effectiveness of removing foreign domains from the target protein in vivo, we designed a recombinant gene for expression of hLIF in N. benthamiana plants. For high-level expression of hLIF in planta, we incorporated the M domain, a highly glycosylated domain derived from human CD45 (Kang et al., 2018), to act as a translation enhancer domain (TED) at the N-terminus of hLIF to induce high levels of protein accumulation. Fusion of the M domain to a target protein causes an up to 6–8-fold increase in the accumulation of target protein in the ER in planta (Kang et al., 2018). To increase the solubility of the recombinant protein (Murashima et al., 2003), we included CBM3, a CBD from Clostridium thermocellum, next to the M domain (Islam et al., 2019a). We also tested the effect on protein expression of p38, a viral coat protein from Turnip crinkle virus (Qu et al., 2003); p38 is a gene silencing suppressor used when genes are transiently expressed in plants via Agrobacterium tumefaciens-mediated transformation (Qu et al., 2003; Thomas et al., 2003). As p38 accumulates to high levels in plants when used as a gene silencing suppressor, we speculated it might also act as a TED when fused to a target protein. To enable removal of N-terminal TED and CBM3 via bdSENP1-mediated cleavage in the ER in vivo, we fused the bdSUMO domain to the C-terminus of CBM3 or TED, respectively. In addition, we added a His×6 tag downstream from bdSUMO so that it could be used to purify recombinant hLIF from extracts of N. benthamiana leaves following in vivo cleavage of the recombinant protein. hLIF is a glycoprotein that has seven potential N-glycosylation sites ( Supplementary Figure S2 ), thus causing the N-linked glycosylation of recombinant hLIF in the ER. The BiP leader sequence and an ER retention signal HDEL (Gomord et al., 1997) were added to the N and C-termini, respectively, of the recombinant constructs ( Figure 2A ), thus giving BiP:M:CBM3:bdSUMO : His:hLIF : HDEL (MCS:hLIF) and BiP:p38:bdSUMO : His:hLIF : HDEL (p38:bdSUMO:hLIF) ( Figure 2A ). A glycine and serine-rich (glycine-glycine-glycine-glycine-serine × 2) flexible linker (L) was inserted to prevent steric hindrance between the domains.

MCS:hLIF and p38:bdSUMO:hLIF were expressed in N. benthamiana leaves using A. tumefaciens-mediated transient expression (Marillonnet et al., 2005). Leaf tissue from 6-week-old plants was syringe-infiltrated with A. tumefaciens harboring MCS:hLIF or p38:bdSUMO:hLIF, together with A. tumefaciens harboring the p1300-p38 construct ( Supplementary Figure S5 ) (Qu et al., 2003). To determine the levels of protein expression, agroinfiltrated leaves were collected 3–5 days post infiltration (DPI), and total soluble proteins (TSPs) were extracted and analyzed using western blotting with anti-LIF antibody ( Figure 2B ). MCS:hLIF and p38:bdSUMO:hLIF-specific protein bands were detected at a range of molecular weights between 66 and 97 kDa. The full-length monomers of MCS:hLIF and p38:bdSUMO:hLIF were predicted to be 55 and 67 kDa, respectively. The multiple bands may represent recombinant proteins with different degrees of N-glycosylation. hLIF contains multiple N-glycosylation sites that display varying levels of N-glycosylation in animal cells; moreover, recombinant hLIF expressed in rice cells showed a variable glycosylation pattern (Youngblood et al., 2014). Blotting with anti-LIF antibody detected multiple protein bands ranging from 30 to 42 kDa ( Figure 2B ). These protein species represented fragments of MCS:hLIF and p38:bdSUMO:hLIF. From the observed sizes of the fragments, recombinant hLIF appeared to have been cleaved downstream of the bdSUMO domain; however, GFP or hIL6 fusion proteins containing the bdSUMO domain did not show any nonspecific cleavage in plants (Islam et al., 2019a; Islam et al., 2019b). The cause of in vivo degradation of MCS:hLIF and p38:bdSUMO:hLIF immediately adjacent to the bdSUMO domain was unclear. The expression level of p38:bdSUMO:hLIF was slightly higher than that of MCS:hLIF in N. benthamiana leaves ( Figure 2B ), and we accordingly selected p38:bdSUMO:hLIF for further study and to produce hLIF in plants.

To determine the optimum conditions for high-level expression of p38:bdSUMO:hLIF in N. benthamiana leaves, we examined its expression in leaf tissues at different time points after infiltration. TSP was extracted from leaves harvested at 3, 5, and 7 DPI and separated by SDS-PAGE, followed by Coomassie brilliant blue (CBB) staining ( Figure 2C ) and western blot analysis with anti-LIF antibody ( Figure 2D ). As observed previously ( Figure 2B ), multiple bands representing p38:bdSUMO:hLIF were detected in two different regions of the blot, between 66 and 97 kDa and between 30 and 42 kDa ( Figure 2D ). The intensity of the signal from both the larger and smaller regions increased with DPI, and was strongest in samples collected 7 DPI. Coexpression of the silencing suppressor p38 greatly enhanced expression of p38:bdSUMO:hLIF (Qu et al., 2003; Thomas et al., 2003).

We investigated whether bdSENP1:HA recognized the bdSUMO domain and cleaved the recombinant hLIF fusion protein (p38:bdSUMO:hLIF) in the ER in planta. We also examined the effect of p38 fusion on the expression of recombinant proteins in plants. As a control, we generated an hLIF construct without the p38 fusion (His:hLIF) ( Figure 2A ). N. benthamiana leaf cells were infiltrated with A. tumefaciens harboring the constructs His:hLIF alone, p38:bdSUMO:hLIF alone, or p38:bdSUMO:hLIF together with bdSENP1:HA ( Figure 2A ); for all three transformations, A. tumefaciens harboring the construct p38 ( Supplementary Figure S5 ) was also coinfiltrated to ensure high-level expression of both proteins. TSPs extracted from the leaves sampled on 7 DPI were analyzed by western blotting with anti-LIF antibody ( Figure 3B ). In the absence of bdSENP1:HA coexpression, anti-LIF antibody detected a band at approximately 66 to 97 kDa that corresponded with full-length p38:bdSUMO:hLIF; by contrast, hLIF-specific bands approximately 30 to 42 kDa in size were observed when bdSENP1:HA was coexpressed ( Figure 3B ), His:hLIF expressed alone was detected in the same size (30~42 kDa) ( Figure 3B ). Blotting with anti-HA antibody produced a 33 kDa band, corresponding with bdSENP1:HA, when bdSENP1:HA was coexpressed with hLIF recombinant protein ( Figure 3A ). bdSENP1:HA, when coexpressed with p38:bdSUMO:hLIF in planta, is thus able to recognize the bdSUMO domain of the recombinant protein and to cut at its cleavage site, thereby releasing the C-terminal His:hLIF.

We examined the cleavage by bdSENP1:HA in vivo. We performed N-terminal amino acid sequencing of His:hLIF. The multiple bands between 30 and 42 kDa were excised from the PVDF membrane after SDS-PAGE electroblots and subjected to Edman degradation. We found that the N-terminal 5 aa residues are H-H-H-H-H ( Supplementary Figure S4 ), indicating that all these multiple bands to be His:hLIF with the same His×5. Moreover, this result suggests that bdSENP1 cleaves the exact recognition site at the C-terminus of bdSUMO moiety in the ER in planta.

Next, we examined whether p38 fusion enhances the protein level in plants. Densitometry analysis was performed to compare the intensity of the signals detected by anti-LIF antibody from TSPs of plant tissues expressing His:hLIF alone, p38:bdSUMO:hLIF alone, or p38:bdSUMO:hLIF together with bdSENP1:HA. Following the coexpression of bdSENP1:HA with p38:bdSUMO:hLIF, the level of released His:hLIF expression increased 1.3-fold to compare with His:hLIF alone ( Figure 3C ), confirming that p38 fusion enhances the expression level of a target protein.

Following bdSENP1-mediated proteolytic cleavage of the full-length hLIF recombinant protein (p38:bdSUMO:hLIF) in planta, we purified His:hLIF from plant extracts. Agroinfiltrated leaves were harvested at 7 DPI, and TSPs were prepared from 40 g samples of fresh leaf tissue. His:hLIF was purified by Ni²⁺–NTA affinity column chromatography using 250 mM imidazole as the eluent. The purity of His:hLIF was examined by SDS-PAGE followed by CBB-staining ( Figure 4A ). The CBB-stained gel exhibited multiple bands of His:hLIF between 30 and 42 kDa. No other bands were detected in the second elution fraction, indicating a high degree of purity. The purified His:hLIF protein was confirmed by western blotting with anti-LIF ( Figure 4B ) and anti-His antibodies ( Figure 4C ). These results confirmed that our in vivo system employing the bdSUMO domain and bdSENP1 to remove foreign domains was highly effective in plant cells at removing domains fused to the N-terminus of a target protein.

Purified His:hLIF was quantified by the Bradford method, using bovine serum albumin (BSA) as a standard (Bradford, 1976). To confirm this method of protein quantification, we used western blot analysis with the anti-LIF antibody ( Figure 4D ) and densitometry analysis ( Supplementary Figure S3 ) to compare the intensity of the signals detected from His:hLIF produced by plants and commercial hLIF (Cat. No. 14890-H08H) produced in a human cell line (HEK293 cells). The two proteins showed a similar band intensity that matched the estimated concentrations, confirming the accuracy of our estimate of His:hLIF protein concentration. The yield of His:hLIF from production in plants was 32.49 mg/kg fresh leaves weight (FW) with approximately > 95% purity. hLIF produced in human cells was, however, detected at a slightly higher position than that produced in plants (30 to 42 kDa). The underlying cause of this difference was unknown. Nevertheless, the slightly higher molecular weight of commercial hLIF can be explained by the difference in the type of N-glycans. Commercial hLIF was produced in the HEK293 cell suspension system through the secretory pathway. Thus, it is likely that hLIF has a complex type of N-glycans via modification in the Golgi apparatus before secretion in the media (Ren et al., 2019). However, the plant-produced His:hLIF should have high mannose type N-glycan because it was retained in the ER via ER retention signal (see Figure 5 ).

We next examined the nature of the multiple bands observed in plant-produced His:hLIF. hLIF has seven N-glycosylation sites: Asp-9, Asp-34, Asp-63, Asp-73, Asp-96, Asp-105, and Asp-116 ( Supplementary Figure S2 ). It was possible that not all these sites were fully N-glycosylated in planta, thereby resulting in multiple bands. To test this hypothesis, His:hLIF was treated with endoglycosidase H (Endo-H) as we reasoned that, since hLIF fusion protein contained the ER retention signal HDEL at its C-terminus, the N-glycans were likely to be the high mannose type and thus sensitive to Endo-H (Islam et al., 2019a). Endo-H-treated His:hLIF was analyzed by western blotting with anti-LIF antibody ( Figure 5 ). Following Endo-H treatment, the intensity of the multiple His:hLIF bands observed between 30 and 42 kDa was greatly diminished and a new protein band appeared at 20 kDa, the predicted size of His:hLIF. This result and the N-terminal amino acid sequencing analysis ( Supplementary Figure S4 ) indicate that the original multiple bands were forms of His:hLIF with varying degrees of N-glycosylation. It thus appears that His:hLIF is N-glycosylated to different levels in plants, as has been observed with hLIF produced in animal cells. Furthermore, the N-glycosylation pattern of His:hLIF confirms that it is localized to the ER.

Endotoxin contamination is a critical concern for in vitro culture of ESCs whenever growth factors or cytokines produced in E. coli are used. Studies show that stem cells exposed to >1 EU/ml endotoxin exhibit morphological alterations and significantly reduced rates of proliferation (Nomura et al., 2017). As we had used A. tumefaciens to induce transient expression of His:hLIF in plants, we examined endotoxin contamination in His:hLIF (phLIF) produced in plants using the chromogenic Limulus amebocyte lysate (LAL) assay. This indicated that phLIF contained <0.61 EU/µg (< 0.06 ng/µg) endotoxin, which was beneath the accepted safety limit [1 EU/µg (< 0.1 ng/µg)] (Magnusdottir et al., 2013). Despite the use of A. tumefaciens to deliver the recombinant gene construct to plant cells, endotoxin levels in the final product were thus very low (Islam et al., 2019a).

We next determined whether phLIF was biologically active. LIF is a crucial regulator that maintains pluripotency in human and mouse embryonic stem cells (mESCs) by suppressing stem cell differentiation (Shabbir et al., 2010; Hirai et al., 2011). We examined growth and pluripotency of mESCs in the presence of phLIF, using commercial LIF produced in E. coli as a PC. As elimination of LIF from the culture medium results in the rapid differentiation of mESCs (Shabbir et al., 2010), we also cultured mESCs in the absence of hLIF as a negative control. mESCs were cultured with 10 ng/ml of phLIF or commercial LIF for four passages, and cellular morphology was observed using phase-contrast microscopy. There were no apparent differences between the two LIF treatments; mESCs treated with either phLIF or commercial LIF appeared similar, and both showed spherical cellular morphology, a characteristic feature of undifferentiated status and pluripotency ( Figure 6A ). By contrast, mESCs cultured without LIF supplement showed spontaneous cellular differentiation ( Figure 6A ). phLIF thus appeared to be active, with a similar effect to commercial LIF. To further confirm phLIF activity, we analyzed mRNA levels of the mouse transcription factors, Oct4, Sox2, and Nanog, which are pluripotency markers (Theunissen & Jaenisch, 2017). mESCs cultured in the presence of phLIF or commercial LIF showed significantly elevated expression of all three genes compared to those cultured without LIF ( Figure 6B ). Although mRNA levels of Oct and Nanog were slightly higher in cells cultured with phLIF than with commercial hLIF, the differences were statistically insignificant ( Figure 6B ). phLIF was thus able to maintain the undifferentiated state of mESCs to a similar degree as the commercial product.