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

Human milk oligosaccharides (HMOs) are key bioactive components of human milk that shape infant gut microbiota and health. Despite widespread use of infant formula, current products lack the diversity of ~200 HMOs found in human milk, and microbial fermentation produces only a small subset at scale. Diverse HMOs with varying linkages and polymerization degrees support specific beneficial microbes, underscoring the need for platforms that can produce a broader HMO repertoire. This study investigates engineered plants as a photosynthetic, sustainable platform to produce diverse HMOs—including complex, higher-degree-of-polymerization species—addressing limitations of microbial production and enabling potential nutritional and therapeutic applications for infants and adults.

Prior work shows HMOs confer multiple health benefits in infants and potentially adults, including shaping the gut microbiota and improving gut barrier function. Commercial production has focused on microbial hosts, which to date can scale only a few simple HMOs (for example, 2′-fucosyllactose, lacto-N-neotetraose), leaving most of the >100 additional HMOs understudied. Beneficial gut microbes can have preferences for specific HMO structures, highlighting the importance of structural diversity. Plant metabolism offers robust nucleotide sugar biosynthesis and could be leveraged for complex glycan assembly, but plant-based HMO production had not demonstrated the breadth of human milk-like diversity. This work builds on knowledge of glycosyltransferases and nucleotide sugar pathways to reconstruct HMO biosynthesis in planta and compares plant versus microbial production economics.

- Expression system and host: Bacterial HMO biosynthetic enzymes were localized to the cytosol of Nicotiana benthamiana. Transient expression used Agrobacterium tumefaciens leaf infiltration (strains mixed with p19 silencing suppressor) to introduce genes for neutral, fucosylated, and acidic HMOs. Leaves were extracted and oligosaccharides purified by liquid–liquid extraction, C18 SPE, and PGC SPE, then analyzed by LC-MS/MS. - Neutral HMOs: Pathway comprised two β-1,4-galactosyltransferases (GalTPM1141, Hp0826), one β-1,3-galactosyltransferase (Cuβ3GalT), and one β-1,3-N-acetylglucosaminyltransferase (NmLgtA). Production of lactose and neutral HMOs with DP 3–7 was detected, including LNT and LNnT (m/z 708.2559), plus larger neutral oligosaccharides identified via MS/MS. - Fucosylated HMOs: An α-1,2-fucosyltransferase (Te2FT) coexpressed with the neutral pathway yielded 2′FL (m/z 489.1819) and LNFPI (m/z 854.3136), and several fucosylated hexasaccharide isomers (4 hexose, 1 HexNAc, 1 deoxyhexose). - Acidic HMOs: Because plants lack CMP-Neu5Ac, a mammalian CMP-Neu5Ac biosynthetic pathway was coexpressed with sialyltransferases. With α-2,6-sialyltransferase (St6), production of 6′SL (m/z 634.2191) and LSTc (m/z 999.3505) was detected; with α-2,3-sialyltransferase (PmST3), 3′SL (m/z 634.2187), LSTd (m/z 999.3510), and multiple acidic hexasaccharide isomers were detected. Together, all three HMO classes and both type I and type II structures were produced. - Pathway optimization for LNFPI: The LNFPI pathway was coexpressed with nucleotide sugar biosynthetic pathways (UDP-Gal, UDP-GlcNAc, GDP-Fuc) and quantified via Agilent 6530 Q-TOF MS using internal calibration. Overexpressing GDP-fucose increased LNFPI by 32.9% to 1,075.03 µg g−1 dry weight versus 808.91 µg g−1. Overexpression also led to lactodifucotetraose (LDFT, m/z 635.2394) and LNDFHI (m/z 1,000.3720) without exogenous α-1,3/α-1,4-fucosyltransferases, suggesting native plant fucosyltransferase activity. Combinatorial nucleotide sugar overexpression shifted overall HMO composition (more fucosylated with GDP-Fuc; increased hexose+HexNAc with UDP-GlcNAc). Statistics used heteroscedastic two-tailed Student’s t-test; n=3 leaves. - Stable transgenic lines: Two constructs were built—HMO10 (four enzymes to produce lactose, 2′FL, LNTII, LNT, LNFPI using 2A peptides under a constitutive promoter) and HMO11 (HMO10 plus GDP-D-mannose-4,6-dehydratase Gmd). Transgenic N. benthamiana lines were generated (Agrobacterium EHA105), with RT-qPCR confirming expression. Quantification by Thermo Q Exactive MS on three leaves per plant. LNFPI was detected in multiple lines with highest average 6.88 µg g−1 dry weight; 2′FL highest average 130.35 µg g−1 dry weight (HMO11 line 5). - Purification of HMOs for functional assays: From pooled transiently expressing leaves (LNFPI + GDP-Fuc pathways), water extraction, yeast fermentation (S. cerevisiae) to remove simple sugars, PVPP treatment to remove phenolics, and repeated C18 SPE yielded an HMO-rich extract with negligible simple sugars/phenolics. LC-MS confirmed presence of LNFPI, 2′FL, LNDFHI and additional oligosaccharides (hexose/deoxyhexose/HexNAc compositions). - Bifidogenic activity assays: Growth of Bifidobacterium longum subsp. infantis ATCC 15697 (HMO consumer) and B. animalis subsp. lactis ATCC 27536 (non-HMO consumer) was tested in anaerobic mMRSC media with plant-derived HMOs versus HMOs isolated from human milk. OD600 trajectories showed B. infantis growth on plant HMO similar to human milk HMO; B. lactis showed no growth on either, indicating selectivity and minimal simple sugars. - Technoeconomic analysis (TEA): SuperPro Designer v12 models compared LNFPI production in a plant-based integrated cellulosic biorefinery (biomass sorghum) co-producing ethanol vs. microbial E. coli system. Assumptions: sorghum accumulates 0.31% DW LNFPI; extraction efficiency 90%; >95% HMO purity plant route; microbial route uses reported highest LNFPI bioconversion yield 0.48% with 62% recovery; plant facility processes 2,000 bone-dry tonnes/day at US$95/BDT; ethanol price scenarios of US$1.44 and US$1.00 per L gasoline equivalent. MSP computed via discounted cash flow analysis. - Instrumentation and analytics: LC-MS on Thermo Vanquish + Q Exactive (PGC column) for identification; Agilent 6530 Q-TOF for profiling/quantification; Agilent 6470 QqQ for targeted quant; MS-DIAL and Agilent MassHunter for data analysis. Replicates generally n=3; lab-scale purifications in duplicate with technical replicates.

- Engineered N. benthamiana produced all three HMO classes (neutral, fucosylated, acidic), including principal type I and II cores (LNT and LNnT; m/z 708.2559), fucosylated HMOs 2′FL (m/z 489.1819) and LNFPI (m/z 854.3136), and acidic HMOs 3′SL (m/z 634.2187), 6′SL (m/z 634.2191), LSTc/LSTd (m/z ~999.35). - Transient expression generated neutral oligosaccharides up to DP7 and multiple isomers, expanding accessible structural diversity beyond current microbial platforms. - Overexpressing GDP-fucose increased LNFPI titer by 32.9% to 1,075.03 µg g−1 DW (vs 808.91 µg g−1 DW). GDP-fucose overexpression also yielded LDFT (m/z 635.2394) and LNDFHI (m/z 1,000.3720) without exogenous α-1,3/α-1,4-fucosyltransferases, indicating native plant fucosyltransferase activity. Nucleotide sugar pathway modulation shifted overall HMO composition toward targeted classes. - Stable transgenic lines produced fucosylated HMOs: LNFPI up to an average of 6.88 µg g−1 DW and 2′FL up to an average of 130.35 µg g−1 DW (HMO11 line 5), demonstrating feasibility of production from photosynthetically fixed CO2 in stable plants. - Purified plant-derived HMO mixtures showed selective bifidogenic activity: B. infantis growth on plant HMOs matched growth on human milk HMOs, while B. lactis did not grow, confirming specificity and low residual simple sugars. - TEA showed plant-based LNFPI has lower minimum selling price (MSP) than microbial: plant MSP US$4.9 kg−1 (ethanol at cellulosic price) or US$18.4 kg−1 (target fuel price) versus microbial MSP US$45.0 kg−1, with microbial costs dominated by downstream purification and glucose feed due to low yield (0.48%) and 62% recovery. Results suggest plants can be cost-effective for complex HMOs when microbes underperform.

This work addresses the need for scalable production of structurally diverse HMOs by demonstrating that plants can synthesize neutral, fucosylated, and acidic HMOs, including complex and higher-DP species not currently available from microbial hosts. By tuning nucleotide sugar biosynthesis (e.g., GDP-fucose, UDP-GlcNAc), the HMO profile can be steered toward desired classes, providing a means to tailor product mixtures for specific microbial or clinical outcomes. The successful creation of stable transgenic lines producing 2′FL and LNFPI establishes a path from transient proof-of-concept to agricultural production using photosynthetically fixed carbon. Functionally, plant-purified HMOs recapitulate the selective bifidogenic effect observed with human milk HMOs, supporting their relevance as prebiotic ingredients. TEA indicates that, under realistic assumptions, plant platforms can economically outperform microbial routes for complex HMOs, and offer additional sustainability benefits via CO2 fixation and coproduct ethanol. Collectively, these results position engineered plants as viable, tunable, and potentially lower-cost systems to expand access to a broad spectrum of HMOs for research and nutrition.

The study establishes engineered plants as a photosynthetic platform capable of producing all major HMO classes, including complex structures such as LNFPI, and demonstrates compositional control by modulating nucleotide sugar pathways. Stable transgenic lines confirm feasibility of continuous, in planta production, and purified plant HMOs show selective bifidogenic activity comparable to human milk-derived HMOs. Technoeconomic analysis projects favorable minimum selling prices for plant-based LNFPI relative to microbial production, suggesting a cost-effective and sustainable route for complex HMOs. Future work should focus on improving yields in stable lines (e.g., optimizing construct design and expression of key glycosyltransferases), expanding to branched and additional complex HMOs, selecting optimal crop species and agronomic conditions, scaling and validating industrial purification processes, and conducting in vivo efficacy and safety studies to support nutritional and therapeutic applications.

- Stable transgenic yields of complex HMOs (e.g., LNFPI) were substantially lower than transient expression, possibly due to 2A peptide effects or reduced expression of key enzymes. - Structural elucidation of many isomeric oligosaccharides was limited to composition-level identification by m/z and MS/MS; full linkage/branching confirmation was not obtained for all species. - Functional assays were in vitro and limited to two bacterial strains; broader microbiome and in vivo responses remain to be evaluated. - The study used a model plant (N. benthamiana); performance and accumulation levels in commercial crops remain to be demonstrated. - TEA relies on assumptions (e.g., 0.31% DW accumulation, 90% extraction efficiency, coproduct ethanol pricing) that require validation at pilot/commercial scale. - Statistical analyses used small sample sizes typical for preliminary biological studies; experiments were not randomized or blinded, which may introduce bias. - Industrial-scale purification feasibility and product quality control were not validated beyond laboratory scale.