01 · contextWhy bacteria broke the rule
Bacteria and the phages that infect them are locked in a chemical arms race that turns over far faster than mammalian immunity. To survive, microbes evolved an arsenal of immune systems — restriction enzymes, CRISPR, retrons, and a growing zoo of defense-associated reverse transcriptases (DRTs). Most immune systems destroy invader nucleic acid. DRTs do the opposite: they build DNA when they sense a phage, and the DNA they build is the weapon.
"Defense-associated reverse transcriptases are widespread bacterial anti-phage systems that use unconventional mechanisms of polynucleotide synthesis."
Deng et al., Science 2026 · DRT3 cryo-EM paper
As of mid-2026 the cataloged DRT families run from DRT1 through DRT10+ and a separate set called UG (UnknownGroup) RTs that includes UG10 — the protein the new Figiel preprint renames DRT7. Four of these families now have published mechanisms, and each one breaks a different assumption.
02 · the headline paperDRT7 / UG10 — DNA synthesized from a blank page
Figiel et al. (bioRxiv, 16 Feb 2026) solved cryo-EM structures of two UG10 enzymes — the family they reclassify as DRT7 — caught mid-synthesis as covalent protein–DNA conjugates. The architecture is two tightly-coupled domains: an RT-like polymerase plus a primase, fused to a poorly-characterized accessory element.
Step 1 · ssDNA
RT-like domain writes poly(dT)
No template, no primer. A tyrosine in the active site donates its hydroxyl to start the chain. The growing strand stays covalently bound to the protein.
Step 2 · dsDNA
Primase domain pairs the second strand
The poly(dT) product is then handed off to the primase half, which now uses it as a template to lay down complementary poly(A).
Trigger
Phage λ Gam mimic
Activation requires a phage-encoded RecBCD mimic (Gam). DRT7 senses the mimic and switches on.
Scope
Broad-spectrum antiphage
The system restricts multiple unrelated phages, not just λ — consistent with sensing a host nuclease threat rather than a specific phage.
What makes DRT7 unprecedented is the template-independent, protein-primed mode of the first step. Every previously characterized DNA polymerase needs a nucleic-acid template; even tdT, the closest mammalian analog, still requires a 3' OH on a pre-existing primer. DRT7 supplies its own primer (a tyrosine side chain) and its own sequence preference, then uses the resulting homopolymer as a stencil for the primase. Information flows from protein structure to DNA sequence.
03 · the other exceptionDRT3 — amino acids as a template
One month before DRT7, Deng et al. (Stanford, Science 2026) solved DRT3. The architecture is completely different — a 6:6:6 hexameric ring of two RTs (Drt3a, Drt3b) plus a noncoding RNA — but the punchline rhymes: a polymerase that does not need a nucleic-acid template.
Architecture
D3-symmetric 6:6:6 complex
Six Drt3a, six Drt3b, six noncoding RNA copies arranged in three-fold dihedral symmetry. Resolved at 2.6 Å.
Drt3a
RNA-templated poly(GT)
Uses a conserved ACACAC stretch in the ncRNA as a normal template, producing one strand of the duplex.
Drt3b
Protein-templated poly(AC)
Synthesizes the complement with no nucleic acid template. Conserved active-site residues act as “dA gate” and “dC gate”, enforcing strict base alternation.
Product
Alternating poly(GT/AC) dsDNA
The two strands anneal into a precise dinucleotide-repeat duplex — the antiphage payload.
The radical claim is that protein side chains alone can dictate DNA sequence with single-nucleotide precision. Deng et al. show that mutating the gate residues collapses fidelity — confirming the protein, not RNA, is the “template” for the AC strand. This is the first clean example of amino-acid-templated DNA polymerization.
04 · the lineageTwo more shocks from the same family
DRT7 and DRT3 didn't appear in a vacuum. The Sternberg lab at Columbia spent 2024–2025 cracking two other DRT families, and each one is its own dogma violation.
DRT2 from Klebsiella pneumoniae performs rolling-circle reverse transcription of a short ncRNA. The result is a concatenated cDNA that, upon phage infection, becomes double-stranded — and the resulting locus encodes a brand-new, stop-codon-less ORF the authors named Neo. Translation of Neo arrests cell growth, killing the infected cell before the phage can replicate. A gene that exists only after the cell is attacked.
DRT9 builds a long stretch of poly-dA — covalently anchored to a conserved tyrosine, exactly like DRT7 will later do for poly-dT. Phage infection releases the brake. The cell drowns in single-stranded polydeoxyadenylate and dies, taking the phage with it (“abortive infection”). DRT9 is the closest mechanistic cousin to DRT7 — same protein-priming trick, different nucleotide, no second-strand handoff.
05 · what nobody has answeredThe open questions
These are the gaps an undergrad can actually attack. Most don't need a multi-million-dollar cryo-EM scope.
How wide is the DRT7 family really?
The Figiel preprint characterizes two UG10 homologs. There are likely hundreds more in metagenomes that nobody has flagged. A simple HMM scan against IMG/M would expand the family map.
What activates DRT3 in vivo?
Deng et al. show the cryo-EM structure of the active hexamer, but the phage-encoded trigger — analogous to DRT7's Gam — is unidentified. Co-infection experiments and proteomics could nail it down.
Is the protein-templated DNA toxic by sequence, or by mass?
Does the cell die because alternating poly(GT/AC) does something specific (e.g. binds an essential protein), or because there is just too much DNA suddenly present? Synthetic mimics could decouple the two.
Can DRT7 be engineered for biotech?
A polymerase that primes itself and copies without a template is a tool. Imagine a programmable cell-state sensor that lays down a DNA “tape recording” on demand. Mutate the dT gate to write other homopolymers?
Are there protein-templated polymerases in eukaryotes?
Every DRT discovered so far is bacterial. The closest eukaryotic activity is terminal deoxynucleotidyl transferase (tdT) — still primer-dependent. A targeted search of eukaryotic RT-fold proteins might surface a hidden analog.
What is the kinetic limit of protein-priming?
How long can the covalent Tyr–DNA adduct grow before it falls off? A simple in vitro time-course with radiolabeled dNTPs would answer it, no structural biology needed.
06 · the planHow an undergrad can actually contribute
The honest path forward — concrete, low-cost, accountable. I'm working through it in this order.
Read every primary paper end-to-end, including supplementaries
The four DRT papers below total ~60 pages of main text and ~200 pages of supplements. The mechanistic details — exact catalytic residues, mutant phenotypes, conditions used — only live in the supplements. Annotate each one. The goal is to be the person in any future conversation who has actually read the data.
PubMedbioRxivZotero
Pull every known DRT7/UG10 sequence and align them
Use the NCBI accession numbers in the Figiel supplement, plus a BLASTp + HMMER scan of UniRef and IMG/M. Build a multiple sequence alignment, mark the active-site tyrosine and primase motifs, and identify outliers — homologs that are missing the catalytic residue are candidates for “DRT7-like-but-different” activity worth a deeper look.
HMMERMAFFTIQ-TREEIMG/M
Run AlphaFold3 on the outliers
For every unusual homolog, predict the structure with cofactors (ssDNA, dNTPs). Compare to the published cryo-EM model. Mismatches in the active-site geometry are testable hypotheses — and AlphaFold is free, GPU-cheap, and undergraduate-tractable.
AlphaFold3PyMOLChimeraX
Build an interactive mechanism viewer for the public
This is where my portfolio kicks in. Most people will never read a Science paper. A clean, browser-based animation of the DRT7 protein-priming mechanism — built in Three.js or D3, embedded next to the cards above — is both a teaching artifact and a way to make the field legible. Ship it on this page.
Three.jsWebGLMol* viewer
Email the labs, with a real question
The Figiel, Deng, Sternberg, and Wiedenheft groups are all reachable. Don't ask for an internship — show up with a specific question their paper doesn't answer (e.g. “does mutating the second tyrosine in your alignment kill activity?”). At UCSD, look at the Pogliano, Wirbel, and Pogliano-adjacent microbial-defense labs. The Doudna network at UC Berkeley is also actively working in this space.
Cold emailUCSD MicroBioDoudna lab
Replicate a single biochemical assay
The DRT7 paper measures protein-DNA adduct formation on a gel. If a UCSD lab will host a few weeks of bench time, that one experiment — purified protein, radiolabeled dNTPs, denaturing PAGE — is reproducible, instructive, and a real data point. Start small. Confirm the published result before claiming a new one.
Wet labPAGEProtein purification
Write a public synthesis · journal-club or blog
Once the literature notes, the alignment, and the mechanism viewer are done, publish a written synthesis on this site — and submit a journal-club presentation at UCSD. The act of explaining it forces the gaps in your understanding to surface, and is the fastest way to be visible to a PI.
SubstackJournal clubThis page
Propose one falsifiable, novel experiment
The endpoint of contribution is not a summary — it's a question nobody has asked yet, framed crisply enough that an answer would be publishable. Mine, today: do DRT7 homologs that have lost the primase domain still confer antiphage activity through ssDNA toxicity alone, the way DRT9 does?
HypothesisExperimental design
07 · primary sourcesReading list (verified)
Every paper here is one I've pulled directly from PubMed or bioRxiv. DOIs are live links. PubMed-indexed papers are cited per PubMed's attribution requirement.
[1]
Figiel et al. — Structures and enzymatic mechanisms of DRT7/UG10 antiphage reverse transcriptases.
bioRxiv · 2026 · preprint 2026.02.16.706125
bioRxiv ↗
[2]
Deng, Lee, Armijo, Wang, Gao — Protein-templated synthesis of dinucleotide repeat DNA by an antiphage reverse transcriptase.
Science · 2026 · DOI 10.1126/science.aed1656 · PMID 41990131
DOI ↗
[3]
Tang, Conte et al. — De novo gene synthesis by an antiviral reverse transcriptase (DRT2 / Neo).
Science · 2024 · DOI 10.1126/science.adq0876 · PMID 39116258
DOI ↗
[4]
Tang, Žedaveinytė et al. — Protein-primed DNA homopolymer synthesis by an antiviral reverse transcriptase (DRT9).
bioRxiv · 2025 · DOI 10.1101/2025.03.24.645077 · PMID 40196691
DOI ↗
[5]
Liu et al. — Molecular mechanism of the type 2 defense-associated reverse transcriptase (DRT2 cryo-EM).
Nucleic Acids Research · 2025 · DOI 10.1093/nar/gkaf1135 · PMID 41206047
DOI ↗
[6]
Tang, Conte et al. — De novo gene synthesis by an antiviral reverse transcriptase (DRT2 preprint).
bioRxiv · 2024 · DOI 10.1101/2024.05.08.593200 · PMID 38766058
DOI ↗
[7]
Mestre, Mestre, Sorek — Prokaryotic reverse transcriptases: from retroelements to specialized defense systems (review).
FEMS Microbiology Reviews · 2021 · fuab025
FEMS ↗
Attribution: items [2]–[6] retrieved via PubMed. Items [1] and [7] retrieved from bioRxiv and Oxford Academic respectively.