Rational domestication of a plant-based recombinant expression system expands its biosynthetic range

Plant molecular farming employs plants as living bioreactors to produce recombinant products. We demonstrate that rationally chosen genome edits can result in plants capable of accumulating previously recalcitrant recombinant peptides.

Abstract

Plant molecular farming aims to provide a green, flexible, and rapid alternative to conventional recombinant expression systems, capable of producing complex biologics such as enzymes, vaccines, and antibodies. Historically, the recombinant expression of therapeutic peptides in plants has proven difficult, largely due to their small size and instability. However, some plant species harbour the capacity for peptide backbone cyclization, a feature inherent in stable therapeutic peptides. One obstacle to realizing the potential of plant-based therapeutic peptide production is the proteolysis of the precursor before it is matured into its final stabilized form. Here we demonstrate the rational domestication of Nicotiana benthamiana within two generations to endow this plant molecular farming host with an expanded repertoire of peptide sequence space. The in planta production of molecules including an insecticidal peptide, a prostate cancer therapeutic lead, and an orally active analgesic is demonstrated.

Introduction

Plant-based production of therapeutics offers opportunities for production at scale, with a reduced environmental footprint (Fischer and Buyel, 2020). Although plant-based expression systems are relatively underexplored compared with bacterial or mammalian cell recombinant technologies, they offer capacity for post-translational modification that could enhance or expand the utility of recombinant products. Their green advantage arises from their serum-free production and culture at scale, with only inexpensive inputs required such as fertilizer, water, and light without costly infrastructure investment (Walwyn et al., 2015; Nandi et al., 2016). Furthermore, animal cell-free systems negate the threat of zoonotic contaminants, which have derailed product rollouts in the past (Barone et al., 2020). Unlike mammalian or bacterial cell production lines which have the benefit of 50+ years of strain selection (Wuest et al., 2012), plant molecular farming (PMF) is a frontier technology in pharmaceutical production, ripe for genetic and process improvements.

PMF typically employs transient gene expression, which is both rapid and flexible for the production of therapeutic antibodies, enzymes, and vaccines. Indeed, transient expression in Nicotiana benthamiana enabled the rapid scale-up and deployment of the first antibody therapy against Ebola (Olinger et al., 2012; Gomez et al., 2021). Plant-based production of a ­seasonal adjustable quadrivalent influenza vaccine is now in phase III clinical trials (Ward et al., 2021b), and the development of SARS-CoV-2 vaccines has matured to the point of being approved by national health authorities (https://covid-vaccine.canada.ca/covifenz/product-details) (Ward et al., 2021a). Although some vaccines, enzymes, and antibodies produced in plant-based systems are on the market or in late-stage trials, there remains a gap in terms of peptide production. There is an opportunity to develop plant systems for producing stabilized therapeutic peptides, by capitalizing on recent advances in the application of plant peptide ligation machinery (Jackson et al., 2020; Rehm et al., 2021).

Some plants natively produce stable head-to-tail cyclic disulfide-rich peptides (cycDRPs), with the best characterized being the three-disulfide-containing cyclotide peptide family (Craik et al., 1999) and the single-disulfide-containing sunflower trypsin inhibitor (SFTI) peptide (Luckett et al., 1999). These peptides are gene encoded and processed from larger precursor proteins as they transit through the plant endomembrane system. The final maturation step is predicted to occur in the vacuole where backbone cyclization is performed by a class of ligase-competent asparaginyl endopeptidases (AEPs), found only in five angiosperm families (Jackson et al., 2018, 2020). Once extracted, cyclotides are highly stable, and tolerant of a range of thermal, proteolytic, or chemical insults (Colgrave and Craik, 2004), making them valuable scaffolds for peptide engineering applications (Wang and Craik, 2018). One proven application is for bioactive epitope grafting, which helps stabilize and configure an epitope for improved efficacy and therapeutic half-life (Chan et al., 2011, 2016). Thus, cycDRPs are envisaged as customizable vehicles to carry therapeutic sequences.

Although grafted cycDRPs can be produced synthetically at laboratory scale, their production at commercial yields is highly suited to a plant-based system, where both precursor and ligase-capable AEP can be cooperatively stacked. However, in plants generally most suited as biofactory hosts, such as N. benthamiana, the endogenous AEPs have not evolved for peptide ligation, but rather retain their ancestral function as hydrolases. Thus, these endogenous AEPs may have a negative influence on the resulting cyclic peptide yield, either from outcompeting and hydrolysing the precursor, or by linearizing any cyclic peptide product formed. Once misprocessed, peptides lack the bond energy required for transpeptidation, thus endogenous AEP activity can have a significant negative impact on the sequence space available for PMF of cyclic peptides.

The development of gene editing tools such as CRISPR (clustered regularly interspaced short palindromic repeats) has enabled targeted genetic improvements of crop species, previously thought impossible (Doudna and Charpentier, 2014). Here we demonstrate the genetic customization of the industrialized PMF crop N. benthamiana through rapid and rational ‘domestication’, enabling the accumulation of heretofore unattainable peptides with therapeutic potential in a plant-based system. We describe our simple and effective genomic edits that enable the production of therapeutics to treat prostate cancer, Netherton syndrome, and neuropathic pain. We further show the production of a potent insecticidal peptide naturally produced in garden pea.

Materials and methods

Generation of ΔAEP N. benthamiana genotype

An AEP knockout line was produced by introducing an array of four gRNA–tRNA repeats, each targeting one of the selected AEPs, into pKIR1.1 which is a CRISPR/Cas9 (CRISPR-associated peptide 9) expression vector carrying the pFAST seed selection system conferring expression of monomeric red fluorescent protein (mRFP) in seeds (Tsutsui and Higashiyama, 2016). Selection of AEP loci to be targeted by CRISPR/Cas9 to introduce mutations was based on the top TBLASTN hits in the version 6.1 Benthgenome annotation set (www.benthgenome.qut.edu.au) when Arabidopsis β-VPE (β-vacuolar processing enzyme; AT1G62710.1) was used as a query (Nakasugi et al., 2014). Targets were further restricted to those showing expression in the Benthgenome Atlas, did not share >90% pairwise identity with any other protein hit to maximize top target diversity, and restricted the maximum number of chosen loci to four to limit pleiotropic effects of reduced AEP function. In pKIR1.1, the crRNA is cloned into the relevant site using AarI, a type IIS restriction enzyme that liberates four bases of unique sequence on both sides of the plasmid backbone. Products were amplified using Phusion polymerase (ThermoFisher Scientific) as per the manufacturer’s protocol in a two-step protocol, except the addition of five cycles at the start of the program where annealing was set to 55 °C for 10 s followed by 30 cycles in a two-step cycle. DMSO at a final concentration of 3% was added to the pGEMT-MOD PCR. The template for pGEMT-MOD was linearized pGEM-T Easy (Promega), and the template for all other reactions was the pTG-mid dsDNA (Supplementary Table S1) fragment synthesized by Integrated DNA Technologies. Products were assembled using gel-purified fragments with the NEBuilder HiFi kit (New England Biolabs) into pGEMT-MOD, and subsequently cloned into pKIR1.1 using AarI in conjunction with T4 ligase (New England Biolabs).

To enable transformation of N. benthamiana, the pKIR1.1_ΔAEP construct was transferred into Agrobacterium tumefaciens (strain LBA4404) by electroporation. Transformation of N. benthamiana leaf discs and regeneration of plants was performed as described by Clemente (2006) with 30 mg l^–1 hygromycin B used for selection (Pavli et al., 2011). Shoots were rooted in rooting medium as described by Clemente (2006) supplemented with 10 mg l^–1 hygromycin B. Primary transgenics were acclimatized in a controlled-environment room and grown to maturity where progeny seed was scored for segregation of red fluorescence using a Nikon SMZ18 stereo microscope equipped with a mercury lamp and mRFP filter set. Non-fluorescent seed, devoid of Cas9 expression, were picked up with a moistened 27 gauge needle, stored in a 1.5 ml tube, and, when ready for planting, grown to maturity. Progeny were screened for lesions at AEP loci using cleaved amplified polymorphic sequence (CAPS) markers and Sanger sequencing using gDNA purified with the Jena Science Plant DNA kit. CAPS marker analysis was performed using primers (Supplementary Table S2) followed by digestion of PCR products. Restriction enzymes used were: NbAEP1 (AlwI), NbAEP2 (PspGI), NbAEP3 (PvuII), and NbAEP4 (StyI). Amplicons hosting polymorphisms remained undigested compared with digested fragments of wild-type alleles. PCR products from CAPS-positive plants were then selected for Sanger sequencing. A single line (termed ΔAEP) was selected for use in all infiltration experiments. See additional genotyping details at the Dryad Digital Repository (https://doi.org/10.5061/dryad.k6djh9w88).

Plant cultivation and material

Nicotiana benthamiana and ΔAEP plants were cultivated using a hydroponic nutrient system in a controlled plant growth facility as part of the Clive and Vera Ramaciotti Facility for Producing Pharmaceuticals in Plants. The temperature in the growth room was set at 28 °C and plants were grown under 170 µmol m^–2 s^–1 of LED illumination in 16 h daylight. (AP67 LED spectra, Valoya Oy, Finland).

RNA-seq analysis

To determine the global gene expression changes between ΔAEP and wild-type plants, two replicates of each plant were vacuum infiltrated with A. tumefaciens carrying pEAQ-eGFP. After 4 d, the plant tissue was sampled, ground to powder in liquid nitrogen, and RNA was isolated using TRIzol reagent (ThermoFisher Scientific) following the standard manufacturer’s protocol. Samples were treated with RNase-free DNase (Ambion), quantified by spectrophotometry, visualized on an agarose gel to check integrity, and submitted for Illumina NextSeq 500 RNA-seq for mRNA input (Australian Genome Research Facility, AGRF). Data are available from the National Center for Biotechnological Information Sequence Read Archive in BioProject PRJNA784697. Reads were trimmed and filtered using Trimmomatic (Bolger et al., 2014), mapped to the Sol Genomics N. benthamiana transcript set using Bowtie2 (Langmead and Salzberg, 2012), quantified with kallisto (Bray et al., 2016), and differential gene analysis performed by EBSeq (Leng et al., 2013). Significantly up-regulated or down-regulated genes were identified by a cut off P-value of 0.05 and differential genes further refined by a ±2-fold difference between the wild type and ΔAEP. See additional RNA-seq expression details at Dryad.

Transient expression in Nicotiana benthamiana leaf

Peptide expression constructs were ordered as gene blocks (Supplementary Fig. S1) (Integrated DNA Technologies, IDT) with codon usage optimized to N. benthamiana according to IDT optimization software. Gateway cloning enzymes (BP and LR Clonase II, Invitrogen) were used to clone the precursor coding sequence initially into the intermediate pDS221 vector (Du et al., 2020), followed by the plant expression vector pEAQ-DEST1 (Sainsbury et al., 2009). For AEP ligase expression, pEAQ-OaAEP1b and pEAQ-CtAEP1 were used and have been previously described (Jackson et al., 2018). For the expression of NbAEPs 1–4, coding sequences were amplified from prepared N. benthamiana leaf cDNA and subsequently transferred to pDS221 and to the plant expression vector pEAQ-DEST1 (Sainsbury et al., 2009). All vectors were validated by Sanger sequencing before transfer to A. tumefaciens strain LBA4404 for leaf infiltration experiments.

To compare relative peptide yields between N. benthamiana and ΔAEP, plants of 5–6 weeks of age were vacuum infiltrated with agrobacterium suspensions containing equal amounts of agrobacterium carrying AEP and precursor peptide constructs as estimated by mixing cultures for a final OD₆₀₀ of 0.5 for each component. Before infiltration, cultures were resuspended in infiltration buffer (10 mM MES pH 5.6, 10 mM MgCl₂, 100 µM acetosyringone) and allowed to rest for up to 2 h to induce virulence. Plants were harvested 5 d post-infiltration and immediately freeze-dried for storage and peptide extraction.

Peptide analysis and relative quantification

Freeze-dried tissue was ground to a fine powder using a GenoGrinder (SPEX SamplePrep). Peptides were extracted using a 50% (v/v) acetonitrile, 1% (v/v) formic acid solution at a ratio of 20 µl mg^–1 of tissue DW. Peptides were extracted overnight with gentle mixing before centrifugation to pellet insoluble material. A 10 µl aliquot of the peptide-containing supernatant was then mixed with an equal volume of an unrelated control peptide (GCCSDPRCNYDHPEICGGAAGN) and a further 80 µl of an 80% (v/v) acetonitrile, 1% (v/v) formic acid solution. For matrix-assisted laser deionization-time of flight (MALDI-TOF-MS), this diluted and spiked peptide mix was mixed 1:1 with a α-cyano-4-hydroxycinnamic acid [5 mg ml^–1 in 50% acetonitrile/0.1% trifluoroacetic acid (TFA)/5 mM (NH₄)H₂PO₄] solution before being spotted and dried on a MALDI plate. MALDI-TOF-MS spectra data was acquired using a 5800 MALDI-TOF MS (AB SCIEX, Canada) operated in reflector positive ion mode. For relative yield determination, the isotope cluster area corresponding to the peptide of interest was normalized to that obtained for the internally spiked peptide control.

Scale-up production of Pa1b and purification

To produce and purify N. benthamiana derived Pa1b peptide for cytotoxicity assessment, 10 ΔAEP plants were infiltrated by submerging whole plants in an agrobacterium suspension and applying and releasing a vacuum. At 5 d post-infiltration, plant leaves were sampled, freeze-dried, and ground to fine powder using a GenoGrinder. Peptides were extracted overnight using 50% (v/v) acetonitrile, 1% (v/v) formic acid at a ratio of 50 µl mg^–1 of tissue DW. After centrifugation to pellet insoluble material, the peptide-containing supernatant was again freeze-dried and resuspended in 10% (v/v) acetonitrile, 1% (v/v) formic acid before solid-phase extraction (SPE) using a Phenomenex Strata C18-E SPE cartridge with 10 g resin capacity. The eluted 10–50% (v/v) acetonitrile, 1% (v/v) formic acid fraction was then pooled, lyophilized, and reconstituted in 10% (v/v) acetonitrile, 0.1% (v/v) TFA in preparation for HPLC on a semi-preparative Phenomenex Jupiter C18 RP-HPLC column (250 m×10 mm, 5 µm particle size) connected to a Shimadzu LC-20AT pump system (Shimadzu Prominence). The purity of peptides was checked by MALDI-TOF-MS and analytic HPLC using a C18 column (Phenomenex, Jupiter® 5 µm, 300 Å, 150 × 2.0 mm).

Cytotoxicity assessment of recombinantly produced Pa1b

Spodoptera frugiperda (Sf9) insect cells were cultured in ESF 921 medium at 27 °C, 5% CO₂ in 75 cm² flasks until 80% confluent. Cells were plated into 96-well tissue culture plates (100 μL, 10 000 cells per well) and grown for 24 h prior to treatment. Peptides were added in triplicate with final concentrations ranging from 0.002 μM to 50 μM, while control wells were incubated with 1% Triton X-100 and medium only. The cells were further incubated for 24 h at 27 °C, 5% CO₂. A 10 μL aliquot of MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide, 5 mg ml^–1 in phosphate-buffered saline (PBS)] was added to each well to a final concentration of 0.5 mg ml^–1 and were incubated for 3 h at 27 °C, 5% CO₂. Supernatants were removed, and the insoluble formazan crystals were resuspended in 100 μL of DMSO. The plate was shaken at room temperature for 10 min to fully dissolve the formazan crystals, and the absorbance of the solutions was measured at 600 nm on a Tecan Infinite M1000Pro plate reader.

Statistical analysis

One-way ANOVA was performed using GraphPad Prism version 9.00 for Mac OS (GraphPad Software, La Jolla, CA, USA).

Results

Cyclic SFTI-1 therapeutic peptide candidates harbouring Asn residues are inefficiently produced in N. benthamiana

SFTI-1 is a 14 amino acid cyclic peptide naturally produced in sunflower seed, and is a favoured scaffold for peptide ­engineering applications (de Veer et al., 2021). Previously, using N. benthamiana as a host for transient gene expression, we demonstrated the successful in planta production of SFTI-1 as well as a variant displaying low picomolar inhibition of the human serine protease plasmin (Jackson et al., 2019; Swedberg et al., 2019). In the same study, we attempted to improve the cyclization yield by substituting the Asp residue of SFTI-1 with an Asn, which we predicted, for some AEPs, could improve the cyclization efficiency. However, unlike the native SFTI-1, SFTI-1_N could not be produced in planta, which led to the hypothesis that the Asp–Asn residue exchange is detrimental to the stability of the peptide in planta.

We first aimed to determine if this problem of instability is more broadly applicable to other SFTI-1 therapeutic candidates, including examples proposed as potential leads for the treatment of prostate cancer. To approach this question, we assembled three additional expression constructs, two encoding SFTI-1 variants designed to inhibit human kallikrein 4 (SFTI-1_KLK4_D and SFTI-1_KLK4_N) (Swedberg et al., 2011; Riley et al., 2019) and one encoding a kallikrein 5 inhibitor (SFTI-1_KLK5_N) (de Veer et al., 2016). The two kallikrein 4 inhibitors were identical, apart from an Asp–Asn exchange at the cyclization residue. This single residue change (Asp–Asn) was previously demonstrated to confer a 125-fold increase in potency (K_i=0.04 nM) and enhanced selectivity over off-target serine proteases (Swedberg et al., 2011). Thus, of the two peptides, SFTI-1_KLK4_N represents the preferred candidate for plant-based production. The kallikrein 5 inhibitor was likewise chosen as a good test peptide, where the best performing peptide harboured an Asn residue at the AEP processing site and was most potent (K_i=4.2 nM) (de Veer et al., 2016).

For expression in planta, we assembled SFTI-1 peptide expression constructs by incorporating back-translated SFTI- therapeutic candidates into the Oak1 precursor framework, replacing the sequence encoding the cyclotide kB1, and retaining the papain-like cysteine protease (PLCP) cleavage site (Rehm et al., 2019) (Fig. 1A). To ensure high-level transient expression in N. benthamiana, we assembled each designer precursor peptide gene and AEP ligase gene into the pEAQ vector system, which allowed for co-infiltration (Sainsbury et al., 2009). As previously demonstrated for SFTI-1 (Jackson et al., 2019), infiltration of N. benthamiana leaf with agrobacterium harbouring precursor gene constructs alone produced no detectable cyclic peptide masses as assessed using MALDI-TOF-MS. However, upon co-infiltration with agrobacterium harbouring the expression vector for ligase-competent AEP (OaAEP1b), backbone cyclic masses were observed for native SFTI-1 and SFTI-1_KLK4_D (Fig. 1B). For the remaining three peptides, each containing an Asn at the cyclization residue, no cyclic masses were evident, suggesting that these peptides are highly unstable in N. benthamiana leaf cells.

Fig. 1.

Therapeutic peptide expression in N. benthamiana. (A) SFTI-1, SFTI-1_N, and therapeutic peptide candidates were prepared for plant-based expression by insertion of the peptide-coding sequence into the Oak1 gene, replacing the sequence domain for the cyclotide kB1. The processing of the engineered SFTI-1 precursors is predicted to be controlled by a papain-like cysteine protease (PLCP) and an asparaginyl endopeptidase (AEP) at the N- and C-terminus, respectively. The Oak1 signal peptide (SP), N-terminal propeptide (NTPP), and N-terminal repeat (NTR) sequence ensure that the precursor enters the endomembrane system with targeting towards the vacuole. (B) Co-expression of engineered SFTI-1 precursors with the AEP ligase OaAEP1b in N. benthamiana resulted in MALDI-MS detection of either degraded peptides (shaded in peach) or cyclic peptides (shaded in purple).

A quadruple AEP knockout depletes competing AEP activity with minimal deleterious phenotypic effects

As the minimal change required to induce in planta instability of SFTI-1 and the tested variants was a single Asp–Asn substitution, we hypothesized that endogenous AEP activity at Asn residues is problematic. To test this hypothesis, we simultaneously produced lesions in four N. benthamiana AEP sequences using CRISPR/Cas9. The chosen loci Niben101Scf04675g08014.1, Niben101Scf04539g04014.1, Niben101Scf18356g00003.1, and Niben101Scf04779g01004.1 were given the names NbAEP1, NbAEP2, NbAEP3, and NbAEP4, respectively.

AEP knockout lines were produced by introducing an array of four gRNA–tRNA repeats, each targeting one of the selected AEPs, into pKIR1.1, which is a CRISPR/Cas9 expression vector carrying the pFAST seed selection system conferring expression of mRFP in seeds (Tsutsui and Higashiyama, 2016). A hemizygous primary transformant was allowed to self, with the resulting seed negatively selected for mRFP expression, thus giving rise to a series (n=7) of genome-edited genotypes in the T₁ generation that were no longer carrying the pKIR1.1 CRISPR/Cas9 cassette. Sites targeted by the crRNA array have restriction enzyme sites present near the protospacer adjacent motif (PAM) end of the crRNA to enable the use of the site as a CAPS marker (Supplementary Fig. S2). In this way, transgene-free T₂ genotypes carrying some form of mutation at the target loci were identified.

Seven plant lines were selected from the T₂ population for further analysis. Amplicons from gDNA containing target sites were examined by Sanger sequencing to categorize mutations present. The lesions identified in the T₂ led to non-silent point mutations, frameshifts, and premature stop codons at the loci targeted. However, some combination of bi-allelic states were observed for all NbAEPs in the population of selected plants, highlighting the need to select for a quadruple homozygous individual. Plants were allowed to self, and quadruple mutants homozygous at the chosen loci were identified in the T₄ generation by CAPS screening and again by validating allelic states with Sanger sequencing of amplicons (Supplementary Fig. S2). In this way, a single line harbouring mutations in all four targeted AEPs was selected and named ΔAEP. The ΔAEP genotype is homozygous for an 18 bp deletion encompassing the splice acceptor site of exon 4 in NbAEP1, insertion of a single nucleotide at position 388 of the NbAEP2 coding sequence, a single nucleotide insertion at position 544 of the NbAEP3 coding sequence, and a missense mutation at position 538 of the NbAEP4 coding sequence.

RNA-seq analysis was performed on T₅ ΔAEP and wild-type plants to further validate the structure of the resulting NbAEP transcripts and observe global gene expression changes between genotypes in a transient expression experiment. Both ΔAEP and wild-type plants were infiltrated with pEAQ-eGFP and samples of infiltrated tissue taken for RNA-seq (n=2 per genotype). Reads of ΔAEP were mapped against the Sol Genomics N. benthamiana Niben101 transcript set and visualized to validate changes in NbAEP mRNA. The ΔAEP NbAEP1 allele produced transcripts with disrupted exon splicing at exon 4 as seen by a lack of read coverage at the crRNA target site (Supplementary Fig. S2), thus confirming the NbAEP1 allele present in ΔAEP as null. Expression levels were estimated, and this revealed that NbAEP1, NbAEP2, and NbAEP3 were significantly down-regulated in ΔAEP (>2.9-fold down-regulated, P-values <0.007) (SupplementaryTable S3). NbAEP4 exhibited very low expression in both genotypes such that it did not result in mapped reads above the threshold for abundance estimation. In addition to the down-regulated AEP transcripts in ΔAEP, we observed one other transcript that was significantly down-regulated <2-fold, and 13 transcripts significantly up-regulated >2-fold (Supplementary Table S3). In summary, our ΔAEP genotype resulted in minimally significant perturbations to the agrobacterium-infiltrated wild-type N. benthamiana transcriptome. Furthermore, ΔAEP plants were morphologically indistinguishable from the wild type. A simple biomass experiment using a randomized plot design under artificial light and growth conditions failed to reveal significant differences in biomass between genotypes [mean DW per 4-week-old plant (n=10): wild type 4.256 ± 0.371 g, ΔAEP 4.281 ± 0.665 g].

To determine the effectiveness of our AEP knockout strategy in transient expression of peptides, we first assessed and compared processing of a modified cyclotide precursor gene Oak1_HIIAA in wild-type and ΔAEP plants (Fig. 2). It was demonstrated previously that Oak1 expression in N. benthamiana without co-expression of a helper AEP ligase results in accumulation of linear, linear extended, linear truncated, and a small MS signal representing cyclic kB1 (Saska et al., 2007; Gillon et al., 2008). Of these, only the cyclic, full-length linear, and linear minus a Gly at the N-terminus can be attributed to AEP processing, with the remainder representing processing by carboxypeptidases that compete for the substrate (Fig. 2A). For our analysis, to ensure that MALDI-TOF-MS could be used to differentiate all predicted processed products, we modified the C-terminal propeptide (CTPP) residues of Oak1 to HIIAA from the natural GLPSLAA, as with the latter construct it is impossible to differentiate the linear kB1 mass from a linear form harbouring a truncated N-terminal Gly and a C-terminal extended Gly. Expression of Oak1_HIIAA in wild-type N. benthamiana resulted in MS signals (as a percentage of total kB1-related signals) of 7.6 ± 2.7% for cyclic, 20.1 ± 1.8% for linear, and 16.4 ± 6.4% for linear Gly, with the remaining ~58% of signal being for peptides carrying C-terminal extensions with or without the N-terminal truncation event. In contrast, expression within the ΔAEP line produced almost no detectable signal for cyclic, linear, or linear Gly peptide, with ~100% of the signal representing non-AEP processed forms (Fig. 2B, C). These results clearly demonstrate the effectiveness of our AEP knockout strategy to remove interfering endogenous AEP activity.

Fig. 2.

Comparison of recombinant peptide endogenous processing in wild-type versus ΔAEP N. benthamiana. (A) The Oak1 precursor is predicted to be processed initially at the N-terminus by a papain-like cysteine protease (PLCP) (blue triangle), followed by asparaginyl endopeptidase (AEP) processing (green triangle) at the C-terminus. Without the co-expression of an AEP ligase, endogenous AEPs that prefer hydrolysis over ligation will compete for the expressed substrate with amino- and carboxypeptidases (red triangles). (B) MALDI-TOF-MS analysis (representative) of Oak1_HIIAA expression in wild-type N. benthamiana (NB) alongside the ΔAEP gene-edited accession. Only the signals highlighted in green can be attributed to endogenous AEP processing, with the remainder representing carboxypeptidase C-terminal processing events. (C) Mean and SD (n=3) of the individual kB1 identified masses as a percentage of the total kB1 peptide MS signal detected by MALDI-TOF-MS.

AEP depletion expands accessible peptide sequence space enabling the expression of cyclic therapeutic peptide leads

Having established an N. benthamiana plant line with reduced endogenous AEP activity, we next tested our hypothesis that reduced endogenous AEP activity would have a positive influence on the yield of Asn-containing SFTI-1_KLK4_N and SFTI-1_KLK5_N. We repeated infiltration experiments and compared relative yields obtained from wild-type and ΔAEP plants. As previously demonstrated, close to no MS signal for cyclic SFTI-1_KLK4_N and SFTI-1_KLK5_N could be detected in infiltrated wild-type plants. In contrast, MS signals for cyclic SFTI-1_KLK4_N and SFTI-1_KLK5_N were readily detectable when constructs were infiltrated into the ΔAEP genotype (Fig. 3A, B), representing a considerable breakthrough. For the expression of cyclotides kB1 and kB2, relative yields were slightly reduced in ΔAEP plants. A ­subsequent test of another AEP ligase CtAEP1 in ΔAEP plants gave a similar result (Supplementary Fig. S3). Further clarification of the negative effect in the ΔAEP genotype for kB1 and kB2 substrates is planned, with a reduced transactivation of AEP ligase probably playing a role.

Fig. 3.

Comparative expression of cyclic peptides in wild-type versus ΔAEP N. benthamiana. Representative MALDI-TOF-MS of cyclic peptides accumulated in (A) wild-type N. benthamiana (NB) and (B) the ΔAEP accession in transient co-expression with OaAEP1b. Test cyclotides are encoded in the Oak1 precursor that natively contains kB1, or with kB1 swapped for kB2, a cyclotide that ends with an Asp residue. MS signals for cyclic peptides are highlighted to match the colours in (C) and (D). An arrow indicates the MS signal for the internally spiked peptide control that served to normalize MS signals for relative quantification. (C) Mean and SD (n=3) of relative kB1 and kB2 MS signals detected in crude peptide extracts of infiltrated N. benthamiana (NB) and the ΔAEP accession. (D) Mean and SD (n=3) of relative SFTI-1_KLK4_N, SFTI-1_KLK4_D, and SFTI-1_KLK5_N MS signals detected in crude peptide extracts of infiltrated N. benthamiana (NB) and the ΔAEP accession.

Cyclotides are more resistant than SFTI-1 molecules to endogenous AEP activity

In common with all plant species, cyclotide producers encode AEPs as a multigene family with isoforms differing in their propensity for peptide ligation (Serra et al., 2016; Harris et al., 2019). Thus cyclotides, which naturally co-locate with AEPs in vegetative cell vacuoles (Conlan et al., 2011; Slazak et al., 2016), may have evolved structures that are more resistant to hydrolytic AEPs than for the Asn-containing SFTI-1 peptides tested in this study. To gain further insight into this, we set up an infiltration experiment in our ΔAEP plant where we co-expressed AEP ligase (OaAEP1b), individual NbAEPs, and the precursor genes of Oak1 or SFTI-1_KLK4_N (Fig. 4). For SFTI-1_KLK4_N, co-expression of any of the four tested N. benthamiana AEPs resulted in a complete elimination of cyclic product formation, as is observed in wild-type plants (Fig. 4A). In contrast, Oak1 processing appears to be only moderately affected, with cyclic kB1 predominating in the MS profile, irrespective of NbAEP overexpression (Fig. 4B, C).

Fig. 4.

Rescue of AEP function in ΔAEP N. benthamiana. Representative MALDI-TOF-MS of (A) cyclic SFTI-1_KLK4_N and (B) cyclic kB1 accumulation in the ΔAEP accession upon co-expression of peptide precursor, AEP ligase, and NbAEP genes. (C) Mean and SD (n=3) of the percentage of MS signal representing cyclic kB1 upon co-expression.

Bioactive linear peptide accumulation

Having shown that we could improve the in planta yield of cyclic SFTI-1 therapeutic leads, we next aimed to determine if the same could be true for bioactive peptides harbouring internal Asn sites. We first chose to test expression of the pea albumin-1 gene (PA1) that encodes the 37 amino acid disulfide-rich insecticidal peptide Pa1b (Higgins et al., 1986). This peptide is of high interest for development as it represents the first ever peptide that specifically inhibits insect vacuolar proton pumps (Chouabe et al., 2011). Although produced naturally in many legumes, the development of Pa1b as a commercial insecticide would benefit from an expression platform allowing the rapid testing of variants for improved yield and potency, and with the capacity for scaled up production (Eyraud et al., 2013). Pa1b has two internal Asn sites that represent putative processing sites of endogenous AEPs, thus Pa1b represents a good peptide to test for yield improvements in our ΔAEP N. benthamiana line devoid of AEP activity (Fig. 5A). Similar to the PA1 expression results reported by Eyraud et al. (2013), we found that PA1 expression in N. benthamiana results in a number of processed forms, with the predominant signals representing Pa1b with the C-terminal Gly37 removed, with or without oxidation (plus 16 Da) of the Met residue. We found this similar pattern of processing in both wild-type and ΔAEP plants; however, relative yields were calculated to be ~3.7-fold and ~1.9-fold higher in ΔAEP plants for Pa1b-Gly and Pa1b-Gly+Met^ox, respectively (Fig. 5B, C). By scaling up plant infiltration, Pa1b yield was calculated at ~0.2 mg g^–1 tissue DW at 95% purity, and the plant-derived peptide was shown to be cytotoxic to Sf9 cells, with a CC₅₀ of 13.58 nM (Supplementary Fig. S4).

Fig. 5.

Expression of two recombinant peptides negatively affected by endogenous AEP activity. (A) The insecticidal peptide Pa1b is encoded by the pea albumin 1 gene (PA1 gene) that additionally codes for the larger PA1a domain of unknown function. The Pa1b N-terminus is predicted to be released via a signal peptide (SP) cleavage event during co-translation into the endoplasmic reticulum. The protease responsible for the release of the Pa1b C-terminus is unknown. Two internal Asn sites are present (red triangles) that may represent putative AEP processing sites. (B) Representative MALDI-TOF-MS of crude peptide extracts prepared from infiltrated leaves of N. benthamiana (NB) and ΔAEP. Indicated by an arrow is the MS signal for the internally spiked peptide control that served to normalize MS signals for relative quantification. (C) Mean and SD (n=3) of the relative MS signal representing Pa1b-G and Pa1b-G(Mox) peptides. (D) The conotoxin Vc1.1 and its analogue Vc1.1[N9W] were prepared for plant-based expression by insertion of peptide-coding sequence into the Oak1 gene, replacing both the cyclotide kB1 domain and the C-terminal tail. Processing was predicted to be controlled by a papain-like cysteine protease (PLCP) (brown arrow). One internal Asn site (indicated by a red arrow) is present in Vc1.1, substituted for a Trp in Vc1.1[N9W]. (E) Representative MALDI-TOF-MS of crude peptide extracts prepared from infiltrated leaves of N. benthamiana (NB) and ΔAEP. (F) Mean and SD (n=3) of the relative MS signal representing Vc1.1 and Vc1.1[N9W].

As an example of a non-plant peptide with therapeutic potential, we prepared constructs (Fig. 5D) for the expression of the 16 amino acid two-disulfide-containing conotoxin Vc1.1, which is a lead compound for the treatment of neuropathic pain (Clark et al., 2010). Native Vc1.1 harbours one internal Asn site, but structure–activity studies have identified that this residue can be substituted with a Trp with minimal changes to activity (Yu et al., 2013). This allowed us to test and compare the expression of the Asn-containing Vc1.1 and Vc1.1[N9W] in wild-type plants and in our ΔAEP line. Of the two variants, only Vc1.1[N9W] could be detected in wild-type plants, indicating that this Asn processing site is liable to processing (Fig. 5E, F). In our ΔAEP line, both Vc1.1 and Vc1.1[N9W] could be detected, suggesting that reducing endogenous AEP levels is beneficial for the accumulation of linear Vc1.1, in a similar fashion to Pa1b.

Discussion

The uptake of N. benthamiana for industrial PMF is gaining momentum, with commercial entities employing N. benthamiana as the preferred biofactory host in Africa, North America, and Europe. In particular, the ‘Lab’ strain, derived from a wild N. benthamiana accession, which naturally carries a genetic lesion that reduces transgene silencing, has been the dominant accession for transient expression-based production for PMF (Bally et al., 2018). Genetic customization reported to date of the ‘Lab’ accession has been limited to the alteration of the glycosylation pathway, to humanize recombinant glycoproteins (Jansing et al., 2019). Altering N. benthamiana to create new genotypes exhibiting expanded expression capability represents great untapped potential. The work presented here demonstrates that potential, made possible by multiplex CRISPR gene editing, whereby knocking out four endogenous AEPs enabled the production of previously non-producible peptide products.

The N. benthamiana ΔAEP genotype is similar to the quadruple AEP mutant described from Arabidopsis in that it does not exhibit a deleterious phenotype in controlled growth environment settings (Gruis et al., 2004). In Arabidopsis, the quadruple AEP knockout genotype is however impaired in programmed cell death and in the hypersensitive response when challenged with Pseudomonas syringae, indicating a role for AEPs in pathogen response (Rojo et al., 2004). Further roles for AEPs include storage protein processing (Shimada et al., 2003), and separation of the endosperm from the testa during seed development, demonstrated in Arabidopsis (Nakaune et al., 2005). In our N. benthamiana ΔAEP genotype, no deleterious phenotypes were observed in controlled growth conditions, and biomass was unaffected. In field conditions or in controlled environments where pathogen infection may occur, the ΔAEP genotype might, however, exhibit greater ­susceptibility, and further exploration of pathogen susceptibility due to reduced AEP function is warranted.

The proteolytic stability of recombinant proteins produced in plants remains a significant challenge. It is clear that there is not a one size fits all approach to improving stability, with larger recombinant proteins possibly degraded by numerous proteases in planta or during the extraction phase. Strategies developed to counter proteolytic interference have included the co-expression of protease inhibitors (Grosse-Holz et al., 2018; Ma et al., 2021), the re-engineering of protease-susceptible residues, and in directing subcellular compartmentation to protein-friendly organelles (Streatfield, 2007). In the case of therapeutic peptide production in plants, our approach for enhancing recombinant peptide stability involves co-expression of a ligase-capable AEP with a peptide substrate amenable to AEP-mediated backbone cyclization (Poon et al., 2018). With their lack of termini, backbone cyclic peptides have been demonstrated to have improved stability, both in human serum stability assays (Wong et al., 2012; Ganesan et al., 2021; Muratspahić et al., 2021) and during accumulation in plant cells (Jackson et al., 2019). Ligase-type AEPs, essential for backbone cyclization, however, have only been identified from cyclotide-producing plant species (Jackson et al., 2020), thus they must be exogenously co-expressed in biofactory hosts such as N. benthamiana.

Through CRISPR/Cas9 gene editing, we have shown here that the reduction of endogenous AEP activity in N. benthamiana positively influences the accumulation of the Asn-containing cyclic SFTI-1 therapeutic peptide candidates [SFTI-1_KLK4_N (Swedberg et al., 2011; Riley et al., 2019) and SFTI-1_KLK5_N (de Veer et al., 2016)] (Fig. 3). This interference by endogenous AEPs appears particularly problematic for Asn-containing SFTI-1 molecules, and not for cyclotides, which have probably evolved to remain resistant to AEP-mediated hydrolysis (Fig. 4). Mechanistically, breakdown of an Asn-containing SFTI-1 molecule may occur through either endogenous AEPs outcompeting the transgene-derived AEP ligase, or by endogenous AEP activity re-cleaving any cyclic peptide generated. We further demonstrate improved production levels upon transient expression in the ΔAEP genotype of the linear peptides Vc1.1 and Pa1b, that each carry internal Asn residues (Fig. 5). Of particular interest was the 1.9 to 3.7-fold increase in the relative yield of Pa1b-Gly+Met^ox and Pa1b-Gly, respectively, when compared with expression in wild-type N. benthamiana. Interestingly, using MALDI-TOF-MS, no intermediate Asn-specific cleavage products were detectable, suggesting rapid hydrolysis of peptide fragments after the primary AEP cleavage event. Thus, this inefficiency in Pa1b production in wild-type N. benthamiana could not be forecast, suggesting that many other small Asn-containing peptides could unknowingly benefit from production in the ΔAEP genotype. Similarly, when the Vc1.1 peptide was expressed in wild-type N. benthamiana, no peptide at all could be detected, which could be attributed to either protease-mediated instability or inefficient peptide folding. Only by expressions in the ΔAEP genotype could we identify that AEP hydrolysis was limiting the accumulation of the Vc1.1 peptide. Likewise, other non-accumulating PMF products containing internal protease sites may benefit from select knockout of offending proteases.

Here we have demonstrated genome editing as a valuable strategy to enable the accumulation of cyclic bioactive peptides in the framework of PMF. In reducing endogenous AEP activity of N. benthamiana, we have demonstrated an expansion of the repertoire of peptides that this favoured biofactory plant can achieve. The targeting of interfering proteases is an approach that may be applied to other recombinant products that may have failed previously. The work thus potentially has broad implications for the efficient and economical recombinant production of therapeutic and insecticidal peptides.

Supplementary data

The following supplementary data are available at JXB online.

Fig. S1. Gene sequences ordered as dsDNA gene blocks used in this study.

Fig. S2. crRNA target sites for NbAEP1, 2, 3, and 4, CAPS marker sites, and resulting mutations for the ΔAEP genotype.

Fig. S3. CtAEP1 (butelase-1) activity assessment in the ΔAEP accession.

Fig. S4. Purity and cytotoxicity assessment of Pa1b.

Table S1. Primers used in construction of the crRNA array.

Table S2. Primers for genotyping genome-edited plants.

Table S3. Genes significantly up-regulated and down-regulated >2-fold.

Author contributions

MJ: design and assembly of plant expression constructs, and performing and analysing plant infiltration experiments; EKG: preparation of the CRISPR/Cas9 AEP knockout construct, transforming, genotyping, and performing global differential gene expression analysis of the ΔAEP line; LYC and MDH: purification of the Pa1b peptide and conducting purity checks and cytotoxicity assays; DJC, MJ, and EKG: planning the experiments.

Conflict of interest

The authors declare no conflict of interest related to this work.

Funding

This work was supported by grants from the Australian Research Council (ARC) Discovery Program [DP150100443 and DP200101299] awarded to DJC, Thomas Durek, and EKG. DJC was an ARC Australian Laureate Fellow [FL150100146] during the period of this work. This work was supported by access to the facilities of the ARC Centre of Excellence for Innovations in Peptide and Protein Science [CE200100012] and the Clive and Vera Ramaciotti Facility for Producing Pharmaceuticals in Plants.

Data availability

The RNA-seq data comparing gene expression between wild-type and ΔAEP N. benthamiana are available through the National Center for Biological Information-Sequence Read Archive BioProject PRJNA784697. Data used to create figures and perform analyses are available online in the Dryad repository: https://doi.org/10.5061/dryad.k6djh9w88; Jackson et al. (2022).