Why Multigene Constructs Fail in Plants—and How to Design Them Better

Plant synthetic biology is rapidly outgrowing the single-gene era. Researchers now routinely try to rebuild entire metabolic pathways, produce pharmaceutical proteins, install programmable gene circuits, and engineer biosensors that respond to environmental signals—all of which require several foreign genes to work at once inside the same plant cell.

The review, authored by Nayoung Lee and Moonhyuk Kwon, lays out the layered biology that stands between a designed construct and a stably expressing plant line. Every multigene construct must survive DNA delivery and integration, adapt to an unfamiliar chromatin environment, sustain transcription and RNA processing, support translation and protein targeting, and avoid pushing the host’s physiology past its limits. In single-gene transformation, failure at any one layer could be compensated by simply screening many independent lines and picking a strong performer. In multigene constructs, the probability that every cassette transfers intact, integrates properly, expresses in balance, and inherits stably drops as complexity rises.

One of the most consequential failure modes is gene silencing, which the authors divide into transcriptional gene silencing (TGS) and post-transcriptional gene silencing (PTGS). TGS acts at the DNA level: methylation and repressive chromatin at the promoter or transgene locus block transcription factor and RNA polymerase II access. Repeated sequences are major risk factors—high copy number, tandem or inverted arrangements, and repeated promoters or terminators all raise the odds that the introduced locus is flagged as repetitive or foreign DNA. Yet the review stresses that repetition does not invariably trigger silencing, and promoter methylation is not explained by copy number alone. In gentian, CaMV 35S promoter sequences were methylated and transgene expression repressed, while the same regulatory sequence behaved differently in tobacco—evidence that host species recognize particular foreign regulatory sequences in species-specific ways.

PTGS operates after transcription, destroying transgene-derived RNA through small RNA pathways. The classic example is co-suppression of chalcone synthase genes in transgenic petunia, where introducing a CHS transgene unexpectedly reduced both transgenic and endogenous CHS expression. Mechanistically, strong transcription can cause incompletely processed, non-polyadenylated, truncated, antisense, or read-through RNAs to accumulate when termination and RNA quality control fall short. RNA-dependent RNA polymerases, particularly RDR6, convert these aberrant single-stranded RNAs into double-stranded RNA, which Dicer-like proteins process into small interfering RNAs that guide Argonaute complexes to cleave homologous transcripts. Inverted repeats, hairpins, and overlapping sense-antisense transcription can generate double-stranded RNA more directly, meaning cassette orientation and architecture directly influence silencing risk. The two silencing layers are mechanistically distinct but can occur independently, simultaneously, or sequentially, and small RNAs from PTGS can in some contexts reinforce chromatin-level repression.

Terminators emerge from the review as underappreciated players that deserve equal status with promoters. A terminator is not merely a stop signal but the entire 3′ regulatory region governing mRNA cleavage, polyadenylation, maturation, and transcription termination—events that shape transcript stability, nuclear export, translation, and silencing susceptibility. In dense multigene arrays, inefficient termination lets RNA polymerase II read through into vector sequences or neighboring cassettes, generating abnormal RNAs that feed silencing pathways. The Arabidopsis HSP18.2 terminator illustrates the point: an element within it reduced read-through transcription, decreased transgene-derived small RNAs, and improved expression. Terminator behavior is also context-dependent—strong Arabidopsis terminators in tobacco infiltration are U-rich, while strong maize terminators in maize protoplasts are GC-rich, and the rice glutelin B-1 terminator worked efficiently in seed but was dysregulated in leaves.

The review also dismantles the assumption that regulatory parts are modular. A promoter that performs well with one coding sequence may produce different output with another; a terminator that stabilizes one transcript may underperform elsewhere. The consequences for pathway engineering are striking. In a two-gene betalain production system in tobacco, stronger promoters tended to yield more product. But in a three-gene diterpene system, also in tobacco, constructs built with moderate promoters produced more of the target metabolite casbene than those with strong promoters—promoter strength does not always correlate with final yield. A construct can fail not only when expression is too low but also when expression ratios are wrong, allowing intermediates to pile up, toxic side products to form, or off-target physiological effects to emerge. Balanced expression, the authors argue, should replace maximal expression as the design goal.

At the part level, the review recommends treating promoters as graded control elements rather than on/off switches. Quantitative studies in Arabidopsis show that expression rises with copy number only below a gene-specific threshold, beyond which higher dosage can trigger post-transcriptional silencing—and that threshold differed among reporter genes carrying the same regulatory elements. Graded promoter collections such as the PCONS series allow different expression strengths to be assigned to different components, and benchmarking in tobacco and BY-2 cells found a 326-fold difference in reporter output between the strongest and weakest promoter-terminator combinations. Practical rules include avoiding the same strong constitutive promoter on every gene, diversifying promoters, terminators, leaders, and introns to reduce internal sequence homology, and using synonymous codon diversification to reduce homology among related coding sequences while preserving protein sequence.

Architecture-level choices carry their own trade-offs. Independent transcriptional units offer maximum flexibility but multiply repeated regulatory parts. Polycistronic strategies compress constructs: viral 2A peptide systems encode multiple proteins from one transcript, with ribosomal skipping separating the products, reducing repeated promoters and terminators. But coordinated transcription does not guarantee equal protein accumulation—2A processing efficiency varies among peptides, protein contexts, and gene positions, incomplete skipping produces fusion proteins, and residual 2A sequences can affect protein activity or localization. Plastid transformation offers true operon-based polycistronic expression with high copy numbers and homologous recombination, avoiding insertion-site position effects and conventional epigenetic silencing, yet protein accumulation can still vary among cistrons and the platform remains technically difficult in many species. Bidirectional promoters cut construct size but couple the regulation of both genes. Targeted integration via CRISPR/Cas, ZFNs, or TALENs can reduce insertion-site variation, but remains demanding and does not eliminate the need for sound internal construct architecture.

Perhaps the review’s sharpest warning concerns validation. Transient assays in protoplasts or agroinfiltrated tissues are fast and useful for spotting gross imbalances, but they do not capture insertion-site effects, chromatin-level regulation, inheritance, developmental exposure, or multigenerational stability that become critical after stable integration. A construct that shines in a transient assay may lose activity, balance, or physiological compatibility once integrated. The authors call for structural validation of the entire transgenic locus, individual RNA- and protein-level assessment of every introduced cargo gene rather than relying on a marker, reporter, or final metabolite, and reporting of expression stoichiometry, copy number, and inheritance. When instability appears, analyses of DNA methylation, transgene-derived small RNAs, read-through transcription, and aberrant polyadenylation can distinguish structural failure from TGS, PTGS, interference, or RNA processing defects.

Burden-aware validation completes the framework. Even fully transcribed and translated multigene programs impose costs: foreign proteins demand amino acids, ribosomes, energy, chaperones, cofactors, and organelle-import systems, while engineered pathways can drain precursors, redirect carbon and energy, disturb redox balance, and accumulate reactive intermediates. Constitutive expression of multiple high-load genes may be tolerated transiently but fail in stable plants where the burden persists across development and generations. The review recommends tracking growth rate, biomass, fertility, morphology, stress sensitivity, and metabolic-intermediate profiles, and evaluating expression across multiple independent events, several generations, and the tissues and environmental conditions relevant to the intended application. Quantitative frameworks for measuring expression burden in plants, the authors note, still lag behind those in microbial synthetic biology—a gap they identify, along with stable-system part benchmarking and standardized reporting, as priorities for making multigene plant engineering genuinely predictable.

Subject of Research: Design principles for stable and balanced multigene expression in plant synthetic biology

Article Title: Design principles for stable and balanced multigene expression in plant synthetic biology

Article References: Lee, N., & Kwon, M. (2026). Design principles for stable and balanced multigene expression in plant synthetic biology. Applied Biological Chemistry, 69(1), Article 66. https://doi.org/10.1186/s13765-026-01135-9

Image Credits: AI Generated

DOI: 10.1186/s13765-026-01135-9

Keywords: plant synthetic biology, multigene expression, gene silencing, transcriptional gene silencing, post-transcriptional gene silencing, terminators, promoter design, 2A peptides, plastid transformation, metabolic burden, transgene integration, CRISPR targeted integration