How We Built the Peppers

Starting with the Transcriptome

Before you can engineer a pepper to be hotter, you need to understand how it makes heat in the first place. Capsaicinoids — the molecules responsible for pepper heat — are synthesized in the placental tissue of the fruit, the white pith that holds the seeds. But not all at once. The biosynthesis genes activate in a precise developmental sequence, turning on and off across fruit maturation.

We profiled this sequence using RNA-seq — the technology that captures which genes are active and at what level. We extracted RNA from Capsicum chinense (habanero-type) placental tissue at five developmental time points: 4, 12, 24, 36, and 54 days post-anthesis (DPA). Salmon 2.8.0 quantified transcript abundance against the Capsicum annuum UCD10Xv1.1 reference genome.

PyDESeq2 0.5.4 performed differential expression analysis, comparing peak capsaicinoid production (24-36 DPA) against early development. The result: 4,365 genes showed significant differential expression (p_adj < 0.01, |log₂FC| > 2). Among them were every gene in the capsaicinoid biosynthesis pathway, each with its own expression trajectory and peak timing.

The key capsaicinoid pathway genes showed strong upregulation: Pun1/AT3 (capsaicin synthase) at +8.34 fold-change, PAL at +6.82, pAMT at +6.41, 4CL at +5.18, COMT at +4.73, KAS at +4.56, and BCAT at +3.91. But knowing which genes are active doesn’t tell you which one limits production. For that, we needed to model the entire pathway mathematically.

Modeling the Metabolism

Capsaicinoid biosynthesis is a convergent pathway. Two branches feed into a final condensation step:

• The vanillylamine branch (from phenylalanine): PAL → C4H → 4CL → HCT → COMT → pAMT → vanillylamine
• The acyl-CoA branch (from valine/leucine): BCAT → KAS → 8-methylnonenoyl-CoA
• Condensation: Pun1/AT3 joins vanillylamine + acyl-CoA → capsaicin

We built a constraint-based metabolic model of this entire system using COBRApy: 37 metabolites across 44 reactions. Flux balance analysis (FBA) optimizes the steady-state flux distribution through the network, subject to stoichiometric constraints and capacity bounds derived from our RNA-seq data.

The Vanillylamine Discovery

The FBA results overturned a fundamental assumption in the field.

For decades, researchers assumed that Pun1/AT3 — capsaicin synthase, the enzyme performing the final condensation — was the rate-limiting step. It seemed logical: it’s the last enzyme, and it’s the one that literally makes capsaicin. Multiple papers had focused on Pun1/AT3 as the key target for boosting heat.

The vanillylamine branch carries a 90% flux control coefficient. Pun1/AT3 carries 0%. The condensation enzyme has zero control over total pathway output. The real bottleneck is upstream — in the supply of vanillylamine, one of the two precursor molecules.

This single finding changed our entire engineering strategy. Instead of trying to make Pun1/AT3 work faster (which would accomplish nothing), we could boost vanillylamine supply by overexpressing PAL (the entry enzyme into the phenylpropanoid pathway) and COMT (a key enzyme in the vanillylamine branch). We also needed to eliminate the metabolic drain — the peroxidase enzymes that actively degrade capsaicinoids after they’re made.

Stopping the Drain

One of the most underappreciated factors in pepper heat is capsaicinoid degradation. Peroxidase enzymes in the placental tissue actively break down capsaicinoids, creating a metabolic drain that reduces steady-state SHU. Two peroxidase loci were identified: LOC107864929 (primary) and LOC107856092 (secondary).

CRISPR-SpCas9 knockout of these peroxidases is the foundation modification present in all 10 cultivar specifications. By eliminating capsaicinoid degradation, we raise the steady-state concentration before any supply-side engineering. This is the simplest modification — just two gene knockouts — but it provides the base upon which all higher heat levels are built.

AI-Guided Enzyme Design

Even though Pun1/AT3 doesn’t control flux, it still performs the condensation reaction. For the highest heat levels (Heats 06-10), we wanted enhanced enzyme variants that could handle the increased substrate flux from our vanillylamine-boosting modifications.

We used ESM2 — Meta’s 650-million-parameter protein language model — to scan every possible single-amino-acid substitution in the Pun1/AT3 sequence. Each mutation was scored by log-likelihood ratio (LLR): a higher LLR means the model predicts that substitution is more compatible with the protein’s evolutionary context.

These mutations are predicted to alter substrate specificity and catalytic efficiency. Heat 06 uses the single S39L mutant. Heat 08 uses the S39L+L345G double mutant. Heat 10 uses a triple mutant: S39L+L345G+C175S — three simultaneously optimized positions creating an enzyme variant never seen in nature.

Designing the CRISPR Constructs

With engineering targets identified, we designed 24 CRISPR-SpCas9 guide RNAs across four genomic targets. Each guide was scored for GC content, PAM site (NGG for SpCas9), and off-target risk against the Capsicum genome.

The modification complexity scales with heat level. Heats 01-04 use only peroxidase knockout — one CRISPR construct each. Heat 05 adds PAL overexpression. Heat 06 adds the Pun1 S39L mutant. Heat 07 adds COMT overexpression. Heats 08-10 use increasingly aggressive Pun1 multi-mutant combinations while maintaining the full overexpression architecture.

Tuning the Heat Profile

SHU alone doesn’t define the experience. The five-capsaicinoid blend ratio determines the character of the heat: onset speed, burn location, duration, and intensity. Each capsaicinoid has a different TRPV1 receptor binding profile.

Capsaicin (SHU coefficient: 16,000,000) delivers sharp, immediate oral burn. Dihydrocapsaicin (15,000,000) provides deep, radiating body heat. Nordihydrocapsaicin (9,100,000) adds mild, fruity warmth. Homodihydrocapsaicin (8,600,000) extends duration. Nonivamide (9,200,000) multiplies intensity with a sharp sting.

Across the lineup, the capsaicin fraction climbs from 40% (Heat 01) to 75% (Heat 10), driving three distinct burn profile tiers: slow-building body burn (Heats 01-03), aggressive mid-mouth assault (Heats 04-06), and instantaneous sinus-stripping strike (Heats 07-10, peaking in under 5 seconds).

Assembling the DNA

The final stage of the pipeline generated 25 complete, synthesis-ready DNA constructs totaling 31,911 base pairs. Each construct includes all regulatory elements — promoters, terminators, selection markers, homology arms for genomic integration — formatted for direct ordering from commercial DNA synthesis providers.

The constructs range in complexity from single-modification peroxidase knockouts (one construct per cultivar for Heats 01-04) to four-component systems combining peroxidase knockout, PAL overexpression, COMT overexpression, and triple-mutant Pun1/AT3 (Heat 10). All 25 constructs have been submitted to NCBI GenBank (SUB16548149).

Beyond the Plant: The Yeast Chassis

As an additional proof of concept, we designed a complete Saccharomyces cerevisiae CEN.PK2-1C chassis for capsaicinoid fermentation. Nine pathway genes were codon-optimized for yeast expression and integrated across four genomic loci (chromosomes X, XI, XII plus an episomal vector). Host metabolism was rewired with feedback-resistant ARO4^K229L and ARO7^G141S mutants, plus ΔARO10 and ΔPAD1 knockouts to redirect precursor flux.

Expected yield: 100-500 mg/L total capsaicinoids. This opens the door to producing pharmaceutical-grade capsaicin through industrial fermentation — no pepper plants required.

The Result

Ten engineered Capsicum chinense cultivar specifications. Three million to thirteen million SHU. Twenty-five DNA constructs. Twenty-four CRISPR guide RNAs. Every sequence synthesis-ready. Every design documented in a scientific preprint (bioRxiv BIORXIV/2026/758036). Every construct deposited at GenBank.

It started with RNA-seq. It ended with synthesis-ready DNA. And the key that unlocked it all was a metabolic model proving that the bottleneck wasn’t where anyone thought it was.