Development of spirulina for the manufacture and oral delivery of protein therapeutics

Abstract

The use of the edible photosynthetic cyanobacterium Arthrospira platensis (spirulina) as a biomanufacturing platform has been limited by a lack of genetic tools. Here we report genetic engineering methods for stable, high-level expression of bioactive proteins in spirulina, including large-scale, indoor cultivation and downstream processing methods. Following targeted integration of exogenous genes into the spirulina chromosome (chr), encoded protein biopharmaceuticals can represent as much as 15% of total biomass, require no purification before oral delivery and are stable without refrigeration and protected during gastric transit when encapsulated within dry spirulina. Oral delivery of a spirulina-expressed antibody targeting campylobacter-a major cause of infant mortality in the developing world-prevents disease in mice, and a phase 1 clinical trial demonstrated safety for human administration. Spirulina provides an advantageous system for the manufacture of orally delivered therapeutic proteins by combining the safety of a food-based production host with the accessible genetic manipulation and high productivity of microbial platforms.

Fig. 1. Homologous recombination into the spirulina chromosome.

a, Plasmid DNA containing an antibiotic-resistance (ABR) gene and a gene of interest (GOI) flanked by LHA and RHA. A double-crossover inserts ABR and GOI into the target locus, replacing genomic DNA. b, Diagram of primer pairs for PCR genotyping. Amplification of LHA and RHA includes one priming site (MP1 and MP4) present only in the spirulina genome at the target locus. Sanger sequencing of the PCR product of the central primer pair (MP5 + MP6) confirmed faithful integration. c, Segregation analysis of strain SP607. Spirulina strain SP3 was transformed on day 0 with donor DNA containing an antibiotic marker and a transgene and cultured under antibiotic selection. Spirulina was collected at the indicated time points, and full transgene products (MP5 + MP6) were amplified from genomic DNA by PCR. Genotyping was performed once. d, Long-term transgene stability. Spirulina strain SP79 was genotyped after continuous culture for >3 years. PCR from genomic DNA was performed with primers targeting the full transgene region (MP5 + MP6). Genotyping was performed once. e, Strategy of markerless transgene integration (see main text for details).

Fig. 2. VHH scaffolding strategies.

a, Cartoons of multimeric scaffolds with sample expression data for VHHs in spirulina. Monomeric (MBP and thioredoxin (TRX)), dimeric (5HVZ), trimeric (cTRP) and heptameric (4B0F) scaffolding proteins were used to multimerize VHHs expressed in spirulina. Intersubunit disulfides confer additional stability to the dimeric and heptameric scaffolds, and these forms were commonly expressed with an MBP tag to improve solubilization. Inset CEIA blots for each scaffold demonstrate spirulina expression of a SARS-CoV-2 RBD-binding VHH fused to the indicated scaffolding protein. CEIA analysis was performed performed once for each strain. All proteins were observed at the appropriate size. b, Increase in apparent binding activity by dimerization of VHHs, as measured by ELISA with purified VHH (top) and spirulina extract (bottom). E. coli-expressed and purified monomeric (PP622) and dimeric (PP661) forms of RBD-binding VHH were assayed with RBD and compared with the binding activity of identical proteins present in spirulina extracts (SP1464 and SP1477, respectively). The concentration of VHH in the spirulina extracts was determined by CEIA. ELISA was performed once with duplicate samples. Absorbance was normalized to the highest absorbance within each sample. EC₅₀ values of 18.3 and 52.3 nM were recorded for monomeric VHH in purified (PP622) and extract (SP1464) samples, respectively, while those for dimeric VHH were 0.32 and 1.29 nM for the purified (PP661) and extracted forms (SP1477), respectively.

Fig. 3. Characterization of spirulina-expressed, anti-campylobacter VHH.

a, Epitope mapping of VHH interaction with FlaA. Peptides derived from the D2/D3/D4 region of FlaA (amino acids 177–482) were panned by phage display. Enriched clones were sequenced after two or three rounds of panning. Results represent average positional frequency observed in two independent panning experiments. b, CEIA quantification of aa682 expressed in SP1182. Clarified lysate from SP1182 was displayed on a Jess system, with an anti-His-tag antibody used for detection. A single peak was observed at the predicted MW of 54.8 kDa. Using a standard curve of purified protein (Extended Data Fig. 7), the amount of soluble aa682 was ~3% of total biomass. Result is representative of dozens of independent experiments. c, Binding kinetics of spirulina-expressed aa682 with recombinant FlaA measured by BLI. Streptavidin-coated biosensors were loaded with biotinylated FlaA, and association and dissociation were measured. Curve fitting was performed using a 1:1 binding model. d, Binding of VHH to intact C. jejuni. Soluble extracts from spray-dried spirulina biomass containing an irrelevant VHH (SP257) or FlagV6-MBP (SP526) were incubated with C. jejuni 81–176 and stained with a fluorescently labelled anti-His-tag antibody. Fluorescence was measured in the allophycocyanin channel (APC-A) by flow cytometry. e, Inhibition of C. jejuni motility by aa682. Two strains of C. jejuni (81–176 and CG8421) were grown on soft agar plates in the presence of aa682 or an irrelevant VHH control (PP496). Halo areas (mean ± s.d.) were measured for triplicate samples at either 40 h (81–176) or 66 h (CG8421) after plating.

Fig. 4. Prevention of C. jejuni infection in mice.

a, Fecal shedding of C. jejuni in a mouse model of infection. Mice received a single 200-μl dose of spirulina strain SP227 (no VHH, n = 4), strain SP526 (analog of aa682, n = 7) or vehicle (n = 8) on days –1, 0, 1, 2 and 3 relative to challenge. Bacterial shedding in stool (CFU 10 mg^–1 feces) was measured 7 days after challenge. b, Fecal biomarkers of inflammation (LCN-2 and MPO) measured 11 days after infection. Mice received a single 200-μl dose of spirulina strain SP257 (irrelevant VHH, n = 4), strain SP526 (n = 4) or vehicle (n = 8) on days –1, 0 or 1 relative to challenge. Uninfected mice were treated with vehicle (n = 4). c, Fecal shedding after treatment with a single dose of SP526. Mice received a single 400-μl dose of spirulina resuspension or vehicle 1.5 h before challenge with C. jejuni. Bacterial shedding in stool was measured 24 h (n = 4 per group) and 72 h (n = 5 per group) after challenge. d, Fecal biomarkers of inflammation (LCN-2 and PMNs) measured 72 h after challenge and treatment with a single dose of SP526 (n = 5 per group). All data represented as mean ± s.e.m.. A one-tailed Mann–Whitney test was applied to assess statistical significance between cohorts, with P values indicated. Each animal experiment was performed once.

Fig. 5. Stability testing of SP1182 binding activity.

a, Batches of spray-dried SP1182 were stored at room temperature for the indicated lengths of time, and binding activity was assessed by ELISA. Purified aa682 binding to recombinant FlaA was used to generate a standard curve by linear regression. VHH activity calculated for spirulina samples was normalized to 100% assuming an expression level of 3% aa682 per unit of biomass. Each point represents a different batch of biomass, presented as the mean of two technical replicates. b, VHH binding activity of SP526 stored at a range of temperatures over 6 months. Aliquots of biomass were prepared in duplicate and stored at the indicated temperature with desiccant. Extracts prepared from spray-dried biomass were serially diluted and assessed for binding activity to recombinant FlaA by ELISA. Each experiment was performed once.

Fig. 6. Protease sensitivity of aa682.

a, SDS–PAGE analysis of purified aa682 incubated with simulated gastric fluid supplemented with 2,000 U ml^–1 pepsin. Results are representative of two independent experiments. b, CEIA of pepsin-digested spirulina biomass resuspension. Dried spirulina biomass of SP1182 was resuspended in simulated gastric buffer and incubated with pepsin for 0–120 min or overnight (O/N). Whole biomass samples were denatured and analyzed on a Jess system. Recombinant aa682 was detected with an anti-His-tag antibody, and data are representative of four independent experiments. c, CEIA of spirulina biomass extracted under different conditions. Dried spirulina biomass of SP1182 was resuspended in simulated gastric buffer (pH 3.0, no pepsin) and the presence of aa682 in buffer was analyzed (A). The biomass was first incubated (A), brought to intestinal pH (pH >5.0) (B) then compared with aa682 extraction by direct biomass resuspension in bicarbonate buffer (pH >7.0) (C). Data are representative of two independent experiments. d, ELISA-based, antigen-binding analysis of spirulina lysates prepared as in c assayed for binding to recombinant FlaA. Samples B and C yielded approximate EC₅₀ binding values of 85.8 and 29.2 μg ml^–1 biomass, respectively. Data are average of two replicates. e, ELISA binding activity of aa682 after in vitro exposure to intestinal proteases. Lysates from SP1182 were incubated with trypsin or chymotrypsin for 1 h. After protease neutralization, aa682 binding activity to recombinant FlaA was measured by ELISA. Data are average of two replicates. Unless noted otherwise, experiments were performed once.

Extended Data Fig. 1. Table of competence genes.

The presence of competence genes in sequenced arthrospira/limnospira genomes was determined using reciprocal best hits against the A. platensis NIES-39 competence genes previously identified. Genomes were retrieved on GenBank and BLASTp was used to identify reciprocal hits with e-values <1^–5 and query coverage >50%. Cell labels indicate the percent identity relative to the respective NIES-39 gene; cells in grey indicate that no homolog was identified. Lyngbya aestuarii BL J was included as an outgroup.

Extended Data Fig. 2. Determination of transgene copy number by qPCR.

Specific primer pairs were used in qPCR reactions to detect for the presence of three different sequences. Genomic DNA from spirulina strains SP205 (parental), SP985 (Pcpc600-driven expression of anti-C. difficile VHH 5D), and SP1182 (Pcpc600-driven expression of aa682) were used as templates. The gene for cpcB was used as an endogenous reference gene. The cpc600 promoter was expected to be present as a single copy in SP205, while both SP985 and SP1182 contained a second copy for transgene expression. The gene for aa682 should only be present in SP1182. Gene copy number ratio was calculated using the ∆∆Ct method for a. cpc600 promoter and cpcB and b. VHH transgene for aa682 and cpcB. Bars in both figures represents the mean ± SD of four independent replicates. Statistical significance was assessed by a two-tailed Mann-Whitney t-test.

Extended Data Fig. 3. Expression of representative transgenic proteins in spirulina.

Uniform expression of GFP in a population of spirulina transformed with a transgene (SP699). Shown are a bright field image (top), and fluorescence images of chlorophyll (middle) and GFP (bottom). b. Western blot of spirulina expressing a 12 kDa orally active and antibacterial peptide, MAA (mammary associated amyloid) (SP13); peptide represents 4-5% of cell protein. c. Western blot of spirulina expressing a vaccine antigen, a repeat domain of malaria P. yoelii circumsporozoite (CSP) protein fused to with core protein of the woodchuck hepadnavirus (WHcAg) (SP82), which self-assembles into a multivalent virus-like particle; antigen represents 4% of cell protein. d. CEIA of spirulina expressing a single-domain antibody, an anti-SARS-CoV-2 VHH fused to a self-dimerizing scaffold (SP1741); VHH represents 9.3% of cell protein. e. CEIA of spirulina expressing a single-domain antibody, an anti-SARS-CoV-2 VHH fused to a self-dimerizing scaffold (SP1825); VHH represents 29.8% of cell protein. f. CEIA of spirulina expressing an enzyme, the catalytic domain from a C. difficile-specific, bacteriophage-derived endolysin (SP1287); enzyme represents 0.54% of cell protein. g. Spectroscopic analysis of spirulina lysates containing a transgenic pigment protein (SP84); pigment protein represents 10% of cell protein. Phycocyanin, a natural blue pigment protein present in wild type spirulina, was overexpressed with a transgenic copy of the native gene. Absorbance spectra for wild-type (black) and transgenic (red) spirulina are shown. Each experiment was performed once.

Extended Data Fig. 4. Model representations of heterologous proteins designed for expression in spirulina.

a. Ribbon representation of a monomeric VHH (orange; PDB ID:6WAQ) with the solubility enhancer, MBP (green; PDB ID: 5M13). The mature, folded protein results in a monomeric VHH as a fusion to MBP and a C-termini 6X-his affinity tag. b. Ribbon representation of a VHH (orange) with a dimerization motif (blue; PDB ID: 5HVZ) and the solubility enhancer, MBP (green). The mature, folded protein results in a dimeric VHH where dimerization is facilitated by the disulfide-linked dimerization motif. The single polypeptide also contains the solubility enhancer MBP and C-terminal 6X-his affinity tag. c. Ribbon representation of a trimeric VHH (orange). The mature, folded protein results in trimeric VHH (orange) where trimerization is facilitated by the self-assembling homotrimer t-cTRP9X₃ (blue). The single polypeptide also contains a C-terminal 6X-his affinity tag. d. Ribbon representation of heptameric VHH (orange) with the heptamerization motif (blue; PDB ID: 4B0F). The mature, folded protein results in a heptameric VHH where heptamerization is due to intrachain disulfide bond between individual protomers. The polypeptide also contains an N-terminal solubility enhancer MBP fusion and C-terminal 6X-his affinity tag. All structures generated using Pymol (Schrodinger).

Extended Data Fig. 5. Effect of MBP fusion on VHH expression in spirulina.

CEIA was used to evaluate the relative expression level of fusions of two norovirus-binding VHHs in spirulina: a) M6 (ref. ) and b) Nano26 (ref. ) The VHH M6 was expressed with no fusion (SP1785), MBP fused to the C-terminus (SP1786), or MBP fused to the N-terminus (SP1790). The VHH Nano26 was expressed with no fusion (SP1772) or MBP fused to the C-terminus (SP1773). All proteins contained a C-terminal His-tag. Clarified lysates from the indicated spirulina strains were normalized by total soluble protein, run on a Jess system, and the protein of interest (POI) detected with an anti-His-tag antibody. The POI was quantified with a purified His-tagged reference standard and the percentage in lysate calculated relative to total soluble protein. Results represents a single measurement for each strain.

Extended Data Fig. 6. CEIA analysis of the oxidation state of disulfides in dimeric proteins expressed in spirulina strain SP1313.

Strain SP1313 expresses the anti-tcdB VHH 7F fused to MBP through the dimerization domain 5HVZ. The dimeric form of the protein should have one disulfide within each VHH and two disulfides to maintain association between 5HVZ subunits. Relative percentage of monomer and dimer peak areas observed under reducing and non-reducing conditions is shown. Protein was detected with an anti-His primary antibody followed by an anti-mouse secondary NIR antibody. SP1313 lysate was run in reducing conditions with DTT for 6 independent experiments with 3 lysates assessed in each run, and in non-reducing conditions for 5 independent experiments with 2 lysates assessed in each run. Each data point represents a peak area as quantified by Protein Simple’s SW Compass Software. Bars indicate mean ± SD. Four representative spirulina strains expressing different VHHs on the 5HVZ scaffold have been analyzed in this manner, and the portion of scaffolded VHH in the dimeric state in non-reduced samples ranged from 50 to 100%.

Extended Data Fig. 7. Linear regression analysis of CEIA standards for aa682 quantification in SP1182.

A reference standard curve of purified aa682 protein measured on a Jess system by anti-His-tag detection is shown. Clarified lysate from spray dried SP1182 was loaded at a concentration of 0.2 mg biomass/mL. Using the standard curve, soluble recombinant aa682 was measured at ~3% of total dried biomass.

Extended Data Fig. 8. Cost optimization.

a. Cost components of cGMP biomass production. b. Spirulina productivity is a function of light intensity and is empirically determined in the described system with SP1182 as the production organism and with current operating parameters. Cost per unit biomass includes labor, capitalized cost of operating lighting system (varies by light intensity), and capitalized costs of other upstream components (independent of light intensity). Minimal cost per unit biomass was achieved at a light intensity of approximately 100 μmol/m²/sec.

Extended Data Fig. 9. VHH stability during spray drying.

FlaA binding activity of aa682 in biomass versus drying temperature. Biomass of strain SP1182 was dried across a range of temperatures, extracted at 10 mg/ml biomass, and the extracts were diluted to a constant 0.039 mg/ml assay concentration. Binding activity of the extracts to FlaA was measured by ELISA. Binding activity was unaffected by drying temperatures <73 °C.

Extended Data Fig. 10. Mass spectrometry analysis of purified aa682.

a. Intact mass spectrum of aa682 measured by LC-MS. The theoretical mass was 54913 Da, with a measured mass of 54783 Da. The difference in mass would be consistent with loss of the N-terminal methionine. b. LC-MS peptide mapping of aa682. Denatured aa682 was digested with trypsin and chymotrypsin, with the resulting peptide fragments characterized by LC-MS. Trypsin and chymotrypsin peptide fragments mapping to aa682 are indicated below the sequence of aa682 in blue and red respectively. Tryptic peptides could be mapped to 52% of the full-length, while chymotryptic peptides mapping to 94% of aa682. Together, the peptide mapping covered 98% of the expected sequence for aa682.