Close this search box.

Engineered plants provide a photosynthetic platform for the production of diverse human milk oligosaccharides – Nature Food

Plant growth

N. benthamiana was grown from seed in 3.5 inch square pots in a controlled environment facility. Plants were grown with a 12 h/12 h day/night cycle at ~700 µmol of photons per m2 s−1. Daytime growth chamber temperatures were kept at 26 °C. Night-time growth chamber temperatures were kept at 25 °C. Relative humidity in the growth chamber was kept between 60% and 75%.

Plasmid construction and transient expression

For transient expression, the native sequences for candidate glycosyltransferases were PCR amplified and cloned into the binary vector, PMS057 (ref. 43), using Golden Gate assembly44, Gibson assembly45 or restriction-ligation (see Supplementary Table 5 for sequences). XL1-blue E. coli cells were transformed with the assembled plasmids via heat shock46. Transformed cells were selected by plating cells on Lysogeny broth (LB) agar plates containing 50 µg ml−1 of kanamycin. Plasmid assembly was confirmed by means of miniprep and Sanger sequencing (Azenta). A. tumefaciens str. GV3101 was transformed using sequence-verified plasmids by electroporation47. Transformed colonies were selected using LB agar plates containing 50 µg ml−1 of kanamycin, 50 µg ml−1 of rifampicin and 10 µg ml−1 of gentamicin. A. tumefaciens str. GV3101 harbouring individual candidate glycosyltransferases were grown in LB overnight to OD600 nm (VWR, V-1200) of 0.8–1.2. The cultures were centrifuged at 4,000g for 10 min and the supernatant was decanted. Cell pellets were resuspended in infiltration media (10 mM MES, 10 mM MgCl2, 500 µM acetosyringone, pH 5.6) and incubated at room temperature for 1 h with gentle rocking (Thermolyne, VariMix). A. tumefaciens strains harbouring each glycosyltransferase were mixed in equal amounts alongside a strain harbouring the p19 silencing suppressor48 to reach a final OD600 nm of 0.5. A. tumefaciens mixtures were injected into the abaxial side of a leaf on a 4-week-old N. benthamiana using a needleless syringe. Each experiment was performed with three biological replicates.

For the production of stable lines, HMO10 and HMO11 constructs were generated through a multipart Golden Gate assembly containing subcloned transcriptional units. Assembled plasmids were transformed and sequence verified as described above. N. benthamiana was transformed using A. tumefaciens str. EHA105 harbouring HMO10 or HMO11 by the UC Davis Plant Transformation Facility.

Quantitative PCR with reverse transcription

Total messenger RNA was extracted using E.Z.N.A. plant RNA kit (Omega Bio-tek) following manufacturer’s directions using the RB lysis buffer variation and on-column DNase digestion; complementary DNA synthesis was achieved with SSIV Vilo IV kit using random hexamers (Thermo Fisher Scientific). Quantitative PCR was performed using a CFX96 Real-Time thermocycler (Bio-Rad) programmed for detection of SYBR intercalating dye with the following temperature programming: 95 °C for 3 min, then 95 °C for 30 s, 60 °C for 45 s, repeated 34 times, then a gradual increase from 65 °C to 95 °C at 0.5 °C per minute to generate melt curves. Sso-Advanced Universal SYBR Green Supermix (Bio-Rad) was used for qPCR amplification. A previously validated primer set was used to amplify EF1α for internal normalization, primers for target genes were designed with Benchling’s qPCR primer design wizard and synthesized by IDT. One target gene from each transcriptional unit was chosen for both constructs (Supplementary Table 6). Melt curves for the product of all primer sets were unimodal and steep, suggesting only a single product was formed for each primer set. No reverse-transcriptase controls showed no amplification within the dynamic range of samples, confirming the efficacy of DNAse treatment and no template controls instituted at the beginning of RNA extraction with no plant matter and kept in parallel with real samples throughout all molecular steps did not amplify, confirming lack of contamination with extraneous DNA. Normalized relative expression was calculated using the ∆∆Cq method and normalized by setting the average level of amplification in the wild-type samples as 1.

HMO extraction for identification of HMOs from individual leaves

N. benthamiana leaves transiently expressing HMO biosynthetic enzymes were harvested 5 days after infiltration. Three leaves of N. benthamiana stable lines transformed with HMO10 and HMO11 were harvested at 4 weeks old. Following harvest, vasculature was removed and leaves were frozen in liquid nitrogen and lyophilized (Labconco, Freezone 4.5) for 2 days. Lyophilized leaves were homogenized via a bead mill (Retsch, MM400) at 20 Hz for 10 min. Oligosaccharides were extracted from 20 mg of lyophilized leaf tissue by ethanol precipitation. To each sample, 1 ml of 80% ethanol was added before homogenization on a bead mill at 10 Hz for 1 min. Samples were then precipitated overnight at −20 °C and centrifuged at 10,000g for 15 min. The supernatant was transferred to a 2 ml screw-cap tube. The pellet was washed twice by adding 500 μl of 80% ethanol, homogenizing via bead mill for 1 min and centrifuging at 10,000g for 15 min. The supernatant and washes were combined and dried in a vacuum centrifuge (Genevac EZ-2, SP Scientific). Dried supernatants were reconstituted in 200 μl of water and subjected to both C18 and PGC SPE (Thermo Fisher Scientific) in 96-well plate format. C18 cartridges containing 25 mg of stationary phase were first conditioned by two additions of 250 μl of acetonitrile (ACN) followed by four additions of 250 μl of water. Samples were then loaded and eluted with two volumes of 200 μl of water. PGC cartridges containing 40 mg of stationary phase were conditioned by addition of 400 μl of water, 400 μl of 80% (v/v) ACN and water, followed by two volumes of 400 μl of water. The sample eluate from C18 SPE was then loaded, washed thrice with 500 μl of water and eluted using two volumes of 200 μl of 40% (v/v) ACN and water. The purified extracts were dried in a vacuum centrifuge and reconstituted in 100 μl of water before injecting 5 μl for liquid chromatography mass spectrometry (LC–MS) analysis.

For quantification of LNFPI in Figs. 2 and 3, LNFPI at known concentrations was added to the extraction solution. The extraction solution was then used on wild-type N. benthamiana. This was done to ensure accuracy of HMO quantification by accounting for HMO losses in the extraction processes and ion suppression that could occur due to the plant metabolites present.

LC–MS analysis of HMOs from individual leaves

For initial screening, chromatographic separation was carried out using a Thermo Scientific Vanquish UHPLC system equipped with a Waters BEH C18 Amide column (HILIC) (1.7 µm, 100 mm × 2.1 mm). A 10 min binary gradient was used based on ref. 49: 0.0–4.0 min, 25–35% A; 4.0–8.50 min, 35–65% A; 8.50–8.70 min, 25% A. Mobile phase A consisted of 3% ACN (v/v) in water with 0.1% formic acid and mobile phase B consisted of 95% ACN (v/v) in water with 0.1% formic acid.

For identification of HMOs produced, we performed LC–MS analysis using a Thermo Scientific Vanquish 3000 UPLC system connected to Thermo Scientific Q Exactive mass spectrometer. Chromatographic separation was carried out using a Hypercarb PGC column (5 µm, 150 mm × 1 mm, Thermo Scientific). A 40 min binary gradient using 3% ACN in water containing 0.1% formic acid (Solvent A) and 90% (v/v) ACN in water containing 0.1% formic acid was performed as follows: 100% A, 0–2.5 min; 100–84% A, 2.5–15 min; 84–42% A, 15–20 min; 42–0% A, 20–22 min; 0% A, 22–28 min; 0–100% A, 28–30 min; 100% A, 30–40 min.

For identification of HMOs, the Q Exactive mass spectrometer equipped with an electrospray ionization source was operated in positive ionization mode with the following parameters: scan range m/z 133.4–2,000; spray voltage 2.5 kV, capillary temperature 320 °C, aux gas heater temperature 325 °C, sheath gas flow rate 25, aux gas flow rate 8, sweep gas flow rate 3. MS/MS analysis was performed using stepped collision energies of 20, 30, 40 eV. MsDIAL was used for data analysis50.

For quantification of LNFPI and HMO profiling, mass spectral analysis was carried out on an Agilent 6530 Accurate-Mass Q-ToF MS operated in positive mode using data-dependent acquisition. The gas temperatures were held at 150 °C. The fragmentor, skimmer, octopole and capillary were operated at 70, 55, 750 and 1,800 V, respectively. The collision energy was based on the empirically derived linear formula (1.8 × (m/z/100) − 3.6). The reference mass used for calibration was m/z 922.009798. The Agilent MassHunter Qualitative software was used for data analysis. Oligosaccharides were identified using an inhouse library, their MS/MS spectra and comparison to either authenticated standards or a pool of HMOs of known composition.

Extraction and purification of HMOs from pooled leaves

Five grams of lyophilized and ground N. benthamiana leaves transiently expressing the LNFPI and GDP-fucose biosynthetic pathways was mixed with 150 ml of water and agitated for 15 min at room temperature in a stirring plate. The mixture was centrifuged at 4,000g for 5 min and the supernatant was separated. The extraction was repeated two more times, combining the supernatant each time. The final supernatant was filtered using a 0.22 µm Millipore Steritop vacuum filter. The extraction process was carried out in duplicate to ensure reproducibility.

Yeast fermentation was carried out to eliminate simple sugars (glucose, sucrose and fructose) from the extracts51. Briefly, autoclaved extracts were inoculated with 0.4 g l−1 of commercial active dry yeast Saccharomyces cerevisiae52 (UCD 522 Montrachet, Lallemand) at 30 °C, 150 rpm for 24 h (Excella E24 Incubator Shaker Series, New Brunswick Scientific). After 24 h, the samples were centrifuged at 4,000g for 5 min and filtered using a 0.22 µm Millipore Steritop vacuum filter to remove the yeast. Samples were concentrated using a vacuum concentrator (Genevac miVac Centrifugal Concentrator) at room temperature and frozen until their purification.

PVPP (Sigma-Aldrich) was used to bind phenolic compounds within the sample following a previous protocol52. Briefly, 3 g of PVPP was conditioned by mixing it with 100 ml of 12 M HCl at 100 °C for 30 min with constant stirring in a stirring plate. After cooling off, the slurry was centrifuged at 4,000g for 5 min and filtered using a 0.22 µm Millipore Steritop vacuum filter. Subsequently, the PVPP was washed with nanopure water until the flow-through reached pH 7. Activated PVPP was mixed with water to a final concentration of 20 mg of PVPP per ml.

PVPP suspension was added to the extracts at a concentration of 6 mg of PVPP per ml and agitated at room temperature for 15 min on a stirring plate. After the time had elapsed, the sample containing the plant extracts and the PVPP was centrifuged at 4,000g for 5 min and filtered using a 0.22 µm Millipore Steritop vacuum filter to separate the PVPP containing the bound phenolics. To eliminate residual phenolics from the extracts, more PVPP was added to the supernatant (6 mg of PVPP per ml) and the process was repeated. The filtrate containing the oligosaccharides was concentrated and frozen until further purification.

Two SPE columns of 60 ml were packed with 15 g of Bondesil-C18, 40 µm suspended in 20 ml of ACN. After the ACN was drained, a frit was added to the C18. Before loading the samples, the columns were conditioned with three volumes of ACN and three volumes of nanopore water. PVPP-treated extracts were loaded onto the conditioned C18 columns and the oligosaccharides were washed with 250 ml of nanopure water divided into four washes. To ensure the complete removal of interfering compounds, C18 SPE was repeated two more times. The purified HMO fractions were dried in a vacuum concentrator (Genevac miVac Centrifugal Concentrator) at room temperature and frozen.

Compositional analysis of plant material

Total carbohydrate content was assessed by the anthrone method with modifications53. In a 96-well microplate, 40 μl of purified and diluted extracts were combined with 100 μl of anthrone reagent (2 mg ml−1 (w/v) in cold 98% sulfuric acid) and mixed through pipette tip aspiration. The microplate was incubated for 3 min at 92 °C in a water bath followed by 5 min at a room temperature water bath and then 15 min in a 45 °C Thermolyne Benchtop muffle furnace (Thermo Fisher Scientific). The plate was cooled for 3 min at room temperature before measuring the absorbance with a SpectroMax M5 UV/Vis spectrophotometer (Molecular Devices) at 630 nm. Total carbohydrate quantification calculations were based on a glucose standard curve. Each plant extract was prepared in duplicate and each sample was further analysed in duplicate.

Total phenolic content of the extracts was determined according to the Folin–Ciocalteu spectrophotometric method as described by ref. 54.

Simple sugars (glucose, sucrose and fructose) were quantified by high-performance anion exchange chromatography with pulsed amperometric detection on a Thermo Fisher Dionex ICS-5000+ HPAE-PAD system based on a method described by ref. 55 with modifications. Diluted extracts (1:100, v/v or 1:1,000 in nanopure water) were filtered through a 0.2 mm syringe filter (Agilent Captiva Econo Filter, PES, 13 mm, 0.2 μm) into 2 ml vials with septa. The samples (25 μl) were injected into a CarboPac PA200 guard column (3 × 50 mm) and a CarboPac PA200 analytical column (3 × 250 mm) and chromatographic separation was carried out with a 12 min gradient elution (from 0.6% to 25% B in 12 min), 0.5 ml min−1 flow rate. The solvent system consisted of A: 100% water; and B: 200 mM sodium hydroxide. Calibration curves (correlation coefficient ≥0.999) were prepared using glucose, sucrose and fructose standards.

Quantification of HMOs by QqQ LC–MS

Detection and quantitation of HMOs were performed using an Agilent 6470 triple quadrupole LC–MS system (QqQ LC–MS) equipped with an Advance Bio Glycan Map column (2.1 mm × 150 mm, 2.7 μm, Agilent). The mobile phase consisted of 10 mM ammonium acetate in 3% ACN, 97% water (v/v, pH 4.5; A) and 10 mM ammonium acetate in 95% ACN, 5% water (v/v, pH 4.5; B). The chromatographic separation was carried out at 35 °C with gradient elution at a flow rate of 0.3 ml min−1. The MS analysis was conducted in positive ion mode with source parameters as follows: the gas temperature was 150 °C at a flow rate of 10 l min−1; the nebulizer was 45 psi; the sheath gas temperature was 250 °C at a flow rate of 7 l min−1; capillary voltage was 2,200 V. See Supplementary Table 2 for gradient and multiple reaction monitoring transitions.

Characterization of HMOs by LC-QToF-MS

Oligosaccharides were purified by a two-step SPE using C18 (HyperSep C18–96, 50 mg bed weight; Thermo Fisher Scientific) and PGC (HyperSep Hypercarb-96, 25 mg bed weight; Thermo Fisher Scientific)56. The samples were filtered (Captiva Premium Syringe Filter PES membrane, 4 mm diameter, 0.2 µm pore size, LC/MS certified) into 200 µl vials.

Individual oligosaccharide compositions were analysed with an Agilent 6520 NanoChip LC-QToF mass spectrometer. Oligosaccharides separation was achieved with a microfluidic high-performance liquid chromatography PGC chip containing an enrichment (4 mm, 40 nl) and an analytical (75 μl × 43 mm) column as well as a nanoelectrospray tip, using a binary solvent gradient of solvent A (5 mM ammonium acetate in 3% ACN, 97% water (v/v)) and solvent B (5 mM ammonium acetate in 90% ACN, 10% water (v/v)) based on a previously optimized method55. The gradient was 0–16% B at 0–20 min, 16–44% B at 20–30 min, 44–100% B at 30–35 min, 100% B at 35–45 min and 100–0% B from 45 to 45.1 min, followed by a 15 min re-equilibration of 100% A57. The mass spectrometer was operated in positive ionization mode with a range of m/z 320–2,500 and an electrospray capillary voltage of 1,800–1,900 V. Reference masses of m/z 922.009 and 1,221.991 provided continuous internal calibration. All samples were analysed using MS/MS with tandem fragmented peaks selected by the automated precursor selection of the six ions with highest signal intensity with a medium isolation width. The Q-ToF MS had a ramped collision energy slope of 0.02 based on m/z values with an offset of −3.5 V. The acquisition rate of 1 spectrum per s was used for both MS and MS/MS. Each spectrum was manually examined and molecular masses were confirmed with Agilent MassHunter Qualitative Analysis B.07.00 software using the molecular feature extraction and a maximum tolerance of 20 ppm.

Bacterial strains and growth conditions

B. longum subsp. infantis ATCC 15697 and B. animalis subsp. lactis ATCC 27536 were cultured at 37 °C in a Coy vinyl anaerobic bubble with an atmosphere of 2.5% H2, ~5% CO2 and balance N2. Routine culturing was done with Difco MRS + 0.05% l-cysteine HCl (MRSC) and carbohydrate-specific culturing was done with modified MRS (mMRSC), which was prepared per litre as follows: 10 g of Bacto proteose peptone no. 3, 10 g of Bacto casitone, 5 g of Bacto yeast extract, 2 g of triammonium citrate, 5 g of sodium acetate trihydrate, 200 mg of magnesium sulfate hexahydrate, 34 mg of manganese sulfate monohydrate, 0.5 g of l-cysteine HCl and 1.063 g of Tween-80. Normally, 2 g of anhydrous dipotassium phosphate would also be added but it was precipitated by the plant HMO preparation.

Growth curves

One colony was used to inoculate 1 ml mMRSC + dipotassium phosphate + 1% lactose monohydrate and incubated for 24 h. The growth curve inoculum was cultured by diluting the 24 h culture 1:100 in mMRSC + dipotassium phosphate + 1% lactose, then incubating for 12–15 h. The inoculum was prepared by washing the cells twice with one volume of 1× PBS. Growth curve cultures (160 μl) were contained in flat-bottomed, optically clear, 96-well, lidded plates and they had a final inoculum and sugar concentration of 1% in mMRSC. Lactose was the growth control substrate, water was the no-growth control substrate and pooled HMO58 was the HMO-growth control substrate. Cultures were done in triplicate. Wells were overlaid with 40 μl of sterile mineral oil and incubated in a BMG SpectroStar Nano. The plate reader was set to read the OD600 nm of each well 30 times in a spiral pattern every half-hour with medium orbital shaking before each read. All media had uninoculated controls whose OD600 nm was subtracted from that of the inoculated medium.

Technoeconomic analysis

In this study, we used SuperPro Designer v.12 to develop technoeconomic models of HMOs which can be produced from both plant and microbial systems. To do this, we first developed process models and then applied discount cash flow analysis of the theoretical production of LNFPI in plants and microbes. The simplified process flow diagram can be found in Supplementary Fig. 8. In the plant system, we adopted integrated cellulosic biorefinery design to coproduce HMO and ethanol to maximize the use of plant biomass. Biomass sorghum was used as the representative plant because its characteristics, such as high yields and drought tolerance, are ideal for biofuel production. Previous studies demonstrated that using biomass sorghum as the representative plant to accumulate value-added bioproducts could improve the economic performance of an integrated biorefinery39,40. Since ethanol is coproduced in the biorefinery, two ethanol selling prices were considered: (1) baseline cellulosic biofuel selling price of US$1.44 l−1 of gasoline equivalent and (2) target fuel price of US$1.00 l−1 of gasoline equivalent.

Briefly, biomass sorghum with 0.31% dry weight HMOs accumulated in the plant biomass are harvested and transported to the biorefinery gate for preprocessing and short-term onsite storage. HMO extraction, separation and recovery is then conducted on the basis of our laboratory processes and as described in previous texts. Water is used to extract HMO from the biomass (room temperature for 6 h) since the industrial extraction process will last longer than the laboratory-scale process given the large quantity of biomass being processed. The extraction efficiency is assumed at 90% based on previous studies39. After extraction, the slurry is first cooled down to room temperature and is sent to centrifugation to remove water. This water is sent to the wastewater treatment unit located in the biorefinery for recycling and reusing. Afterwards, multistage ultrafiltration is used to recover HMO from the extraction stream. Before transporting to HMO onsite storage, another centrifuge is applied to ensure the recovery and purity of the final product. In this plant system, the extracted HMOs are considered as the main product with the purity of >95%. The remaining biomass from biomass sorghum is routed to ionic liquid pretreatment for biomass deconstruction. After ionic liquid pretreatment, enzymatic hydrolysis and ethanol fermentation are conducted to produce ethanol, followed by distillation and molecular sieve to remove excess water. Wastewater from the overall process is routed to the wastewater treatment sector to produce reusable process water and biogas, which can be combusted in the onsite turbogenerator, along with other solids from biomass, to generate heat and electricity that can satisfy the facility’s need. In the microbial system, HMOs were produced as the single product in the biorefinery. Unlike the plant system, pure sugar is used in the microbial system as the sole feedstock and no wastewater treatment sector and onsite combustion are designed for the microbial system. The downstream processing of microbial production is adopted from peer-reviewed publication41. After LNFPI is produced in the bioreactor under the bioconversion condition of 30 °C for 52 h, ultrafiltration is first applied to remove biomass and ion exchange chromatography to remove ions and other charged impurities. Multistage nanofiltration is further used to reduce the total volume and remove excess impurities. Gel filtration is used for the final purification of HMO from the microbial production process. The purity of the final product is >98% through this process.

After developing technoeconomic models in SuperPro Designer, we performed mass and energy balance and then applied discounted cash flow analysis to quantify the MSP of HMOs (US$ kg−1). For both systems, we assume the biorefinery can operate 24 h per day and 330 days per year for 30 years. The unit price of biomass sorghum is assumed at US$95 per bone-dry tonne and we assume in the plant system the cellulosic biorefinery can intake 2,000 bone-dry tonnes of biomass sorghum per day. The unit price of ionic liquid is US$2 kg−1 with a range of US$1–5 kg−1. Total capital investment includes installed equipment cost, piping costs, engineering costs, warehouse, site development, construction fees, contingency costs, land costs, startup and working capital. Annual operating costs include raw materials costs, utility costs, labour costs and facility-dependent costs such as insurance. These parameters are kept constant as per the 2011 National Renewable Energy Laboratory report59.

Statistics and reproducibility

No statistical method was used to predetermine sample size as performing experiments in biological triplicate is standard of the field. No data were excluded from the analyses. The experiments were not randomized. The Investigators were not blinded to allocation during experiments and outcome assessment. All experiments looking at individual leaves were conducted in biological triplicate. For the laboratory-scale purifications, extractions were completed in duplicate and each duplicate was measured with two technical replicates. Microbial growth assays were conducted in biological triplicate. For statistical analysis of LNFPI optimization, a heteroscedastic two-tailed Student’s t-test with the LNFPI pathway expressed alone was used as the reference group in RStudio v.1.2.5033. GraphPad v.10 was used to plot the microbial growth assays.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.