ThermoCas9: Methylation-Sensitive CRISPR-Cas9 for Epigenetic Precision Editing

ThermoCas9 is a Type II-C CRISPR-Cas9 nuclease (1082 amino acids) from the thermophilic bacterium Geobacillus thermodenitrificans T12. Unlike SpCas9 from Streptococcus pyogenes, ThermoCas9 is methylation-sensitive: a 5-methyl group on the cytosine at position 5 of its PAM (5mCpG or 5mCpC) blocks the enzyme from binding DNA. This sensitivity creates a fundamentally new lever for CRISPR genome editing — selective targeting of hypomethylated cancer alleles while sparing the methylated copies in normal cells. The simulator above models this gating behavior across cell-line pairs validated in the published literature.

How methylation-sensitive editing works

ThermoCas9 recognizes two PAM sequences: 5'-NNNNCGA-3' (CpG-containing) and 5'-NNNNCCA-3' (CpC-containing). The fifth nucleotide — the cytosine — is the discriminating base. Cryo-EM structures resolved at 2.8 Å (pre-cleavage, PDB 9AR4) and 2.2 Å (post-cleavage, PDB 9AR7) show that residue Asp1017 contacts the C5 base through the major groove via two hydrogen bonds. A 5-methyl group at this cytosine sterically blocks the contact, preventing DNA binding. Ser1019 contributes to methylation discrimination — the S1019A mutant retains cleavage activity but loses the ability to distinguish methylated from unmethylated PAMs.

The discrimination happens at the binding step, before any cleavage attempt. This contrasts with AceCas9 from Acidothermus cellulolyticus, which also detects PAM methylation but binds methylated DNA and is then arrested at the open-to-closed conformational transition that activates the HNH nuclease domain. Two methylation-sensitive Cas9 enzymes, two distinct mechanisms.

Selective editing of hypomethylated cancer alleles

Many cancers display locus-specific hypomethylation at oncogenic or differentiation genes. The Roth et al. 2026 paper validates ThermoCas9 in two cell-line pairs:

MCF-7 (luminal/ER+ breast adenocarcinoma) vs MCF-10A (non-tumorigenic mammary epithelium)

Luminal-signature transcription factors ESR1 (Chr 6:151,690,043, β=0.07 in MCF-7 vs β=0.94 in MCF-10A) and GATA3 (Chr 10:8,045,463, β=0.02 vs 0.31) are hypomethylated in cancer. The catalytically enhanced (CE) variant of ThermoCas9 delivered as ribonucleoprotein (RNP) achieves up to 78% editing at GATA3 in MCF-7 with minimal editing of the methylated MCF-10A counterpart. GATA3 truncation mutations contribute to ~50% of GATA3 mutations in luminal/ER+ tumors and cause dominant-negative impairment of differentiation.

HCT116 (colorectal adenocarcinoma) vs HEK293T (reference kidney)

EMX1 target site 4 is methylated in HCT116 (β=0.95) and unmethylated in HEK293T (β=0.00). ThermoCas9 edits EMX1 to ~16% indel frequency in HEK293T but produces null editing in HCT116 — a clean on/off demonstration of methylation gating. SpCas9, used as a control, edits EMX1 in both lines regardless of methylation status, confirming that the differential is intrinsic to ThermoCas9 and not a chromatin or accessibility effect.

The Catalytically Enhanced (CE) variant

Wild-type ThermoCas9 has lower baseline activity than SpCas9 — common to type II-C effectors with weaker DNA-unwinding capability. To improve editing efficiency without sacrificing methylation sensitivity, the authors performed directed evolution of the HNH–RuvC linker (residues between the HNH and RuvC-III domains) and isolated a double mutant: Glu655Gly + Asn696Ile, designated CE-ThermoCas9. CE-ThermoCas9 produces substantially higher indel frequencies across all tested loci while preserving the methylation-discrimination profile.

How ThermoCas9 compares to other CRISPR-Cas9 enzymes

SpCas9 (S. pyogenes, Type II-A)

PAM: 5'-NGG-3'. Methylation-insensitive. Edits all sites with sufficient accessibility. Used here as a control to confirm that methylation-driven differential editing is ThermoCas9-intrinsic.

AceCas9 (A. cellulolyticus, Type II-C)

PAM: 5'-NNNNCC-3'. Methylation-sensitive at the activation step (binds methylated DNA but cannot rotate HNH into the closed conformation). Different mechanism from ThermoCas9.

ThermoCas9 wild-type (G. thermodenitrificans T12, Type II-C)

PAM: 5'-NNNNCGA-3' or 5'-NNNNCCA-3'. Methylation-sensitive at the DNA binding step. Lower baseline activity than SpCas9.

CE-ThermoCas9 (engineered variant, Glu655Gly + Asn696Ile)

Same PAM and methylation profile as wild-type, with substantially improved cleavage efficiency. The variant most relevant to therapeutic applications.

Delivery formats validated in the paper

Ribonucleoprotein (RNP) nucleofection

Pre-assembled ThermoCas9-sgRNA complex delivered via Lonza nucleofector. Highest editing efficiency observed (~78% at GATA3 in MCF-7 with the CE variant). Transient — no exogenous DNA or RNA persistence.

ThermoCas9-mRNA + chemically modified sgRNA

In-vitro-transcribed mRNA with codon-optimized ThermoCas9-NLS, delivered via Lipofectamine. Compatible with lipid-nanoparticle (LNP) delivery for in vivo work.

Plasmid co-transfection

ThermoCas9 + sgRNA + EGFP reporter plasmids. Lowest efficiency, but useful for FACS-sortable screening workflows.

Why methylation-sensitive editing matters for cancer research

Standard CRISPR-Cas9 editing is sequence-specific but agnostic to cell state — it cleaves the target wherever it is present. In a cancer therapy context this means off-tumor editing is governed purely by delivery selectivity and gRNA specificity. Methylation-sensitive enzymes like ThermoCas9 add an orthogonal layer: even if the enzyme reaches a normal cell, editing only proceeds where the target locus exhibits the cancer-associated hypomethylation. This restricts editing to genomic sites that have lost 5mCpG methylation, which can be a useful filter when the intended targets are tumor-associated alleles. The trade-off is reduced editing flexibility and a narrower set of accessible loci, since the PAM must include a CpG or CpC dinucleotide.

Status and limitations

ThermoCas9 is a research-stage molecular tool. The Roth et al. 2026 Nature paper demonstrates methylation-sensitive editing in cell-line models. The simulator on this site is an educational illustration of the published mechanism. Numerical efficiencies in the simulator are illustrative ranges drawn from the paper's figures and supplementary data; exact indel frequencies depend on chromatin context, delivery, sgRNA design, and replication timing. ThermoCas9 is not an approved therapy. A patent application covering methylation-sensitive editing with ThermoCas9 (PCT/US2025/014770, US 2025/0250641) was filed by Florida State University Research Foundation and Wageningen University.

More on this site

Translational hypothesis stack

Seven testable hypotheses, three concrete preclinical programs, and a Phase 0 / Phase 1 / platform-trial framework for moving methylation-sensitive Cas9 editing toward clinical use.

Summer research project guide

Bounded undergraduate and graduate project scopes — computational target discovery, in vitro cleavage assays, biomarker prediction — with a concrete 10-week example and publication targets.

Glossary

5mC (5-methylcytosine)

Cytosine with a methyl group at the 5-carbon position. The principal mammalian DNA methylation mark; concentrated at CpG dinucleotides in healthy cells, often disrupted in cancer.

β-value

A quantitative methylation level for a given CpG site between 0 (fully unmethylated) and 1 (fully methylated). Measured here using Illumina Infinium MethylationEPIC arrays and reduced- representation bisulfite sequencing (RRBS).

PAM (protospacer-adjacent motif)

A short DNA sequence next to the gRNA target site that the Cas9 enzyme requires for binding. Different Cas9 orthologs have different PAMs.

sgRNA / gRNA

Single-guide RNA. The 20-nt programmable spacer that directs Cas9 to a complementary genomic locus.

RNP (ribonucleoprotein)

Pre-assembled Cas9 protein + sgRNA complex. Delivered transiently into cells without persistent DNA or RNA.

HNH and RuvC domains

The two nuclease domains of Cas9. HNH cleaves the target strand; RuvC cleaves the non-target strand. A double-strand break is produced when both domains are active.

CRISPRi / CRISPRa

Transcriptional interference / activation using catalytically dead Cas9 (dCas9) fused to KRAB (CRISPRi) or VP64 (CRISPRa). Not the focus of the Roth et al. 2026 paper, which addresses cleavage by active ThermoCas9.