iiinova — where your digital self keeps evolving

Michael Lin, MD PhD @michaelzlin.bsky.social · 13h

Karl opined that there may be an evolutionary reason for this feeling, because we are a social species and survival depended on teaching. Now we just need to find the pathway from the cortex to the VTA for that.

040

Michael Lin, MD PhD @michaelzlin.bsky.social · 13h

Celebration at Stanford Bioengineering for Karl Deisseroth’s Nobel Prize. His commitment to teaching came up, and Karl said something very interesting: that one of the best feelings is finishing a classroom session and thinking it went well, that the students learned something. Completely agree!

1181

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

And would not have been possible without early support from Stanford Bio-X, the Focused Ultrasound Foundation, and the SMA Foundation, and sharing the same roof with the Hong lab at the Stanford Wu Tsai Neuroscience Institute

110

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

Dongin Lee and Hanlee Ji for NGS; Nicole Repina and David Schaffer for engineering a programmable multi-well illuminator; and of course Guosong Hong for co-directing the project!

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

Many people to thank on this fun and intense collaborative project: Michael Westberg for ideas on designs, Yichi Su for the reporter construct; Marigold Malinow, Xiang Wu, Shan Jiang, and Sa Cai for MLNT synthesis and ultrasound experiments, Dongyun Jiang for histology...

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

There is more to explore, e.g. cell-type-specific expression and generalizing ps-Cas12f to non-editing outputs. More broadly, we hope that remote control of engineered proteins in vivo inspires others to think of other ways in which engineering can lead to smarter medicines.

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

Actually we found Cas12f variants to have less off-target activity than Cas9, replicating earlier results. So we are taking a low risk and making it even lower. IMO all future CRISPR will be Cas12f CRISPR: smaller, more specific, and just as active — and now regulatable in 4D.

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

So why is local gene editing useful? For example to treat Huntington's Disease, we'd prefer to disrupt HTT in just the striatum if possible, in case it has essential functions elsewhere. Also caging Cas12f by default reduces off-target toxicity that would otherwise accumulate.

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

It's interesting to think about what's happening here: The MLNTs circulate in the bloodstream. The light they produce cross the BBB to activate ps-CasMINI whose genetic instructions had crossed the BBB earlier in AAV!

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

So we delivered ps2-CasMINI across the BBB using AAV serotype CAP.B10 (from the Gradinaru lab), then perform local ultrasound induction of reporter gene editing at one location in the brain. (We show the pseudocolor overlays because that's standard but we like to show raw luminescence images too)

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

The brain is one place where local gene editing would be useful. As neurologists know, many conditions are localized, so it would be nice if CRISPR-based genetic treatments can be local too. But location-specific promoters often don't exist, and it would be nice to avoid surgery.

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

And so in a close collaboration with the Hong lab, we injected AAV to express ps-Cas12f in muscle, then injected mechanoluminescent nanotransducers (MLNTs), then performed ultrasound to trigger light production at the focal point and induce Cas12f-mediated gene editing.

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

Okay but external illumination isn't really local or deep. That's where the 2nd engineering effort comes in, in the development of mechanoluminescent nanoparticles by Guosong Hong's lab. These are particles that generate cyan light upon ultrasound stimulation, perfect for pdDronpa!

110

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

And it works in vivo! We delivered each system in a single AAV to mouse legs, then applied external light Light induced luciferase transgene activation (due to gRNA-targeted editing of a stop signal in front of the luciferase) by ~19x vs no light, reaching similar activity levels as unfused CRISPR

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

We fixed this. We engineered a faster-switching pdDronpa2, then made "ps2" variants with newer higher-activity Cas12f variants, eCas12f and AsCas12f-HKRA The result was ~20x more gene editing in response to lower amounts of light, reaching the same activity as unfused Cas9.

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

But pdDronpa's photoresponse was fairly inefficient, occurring in <1% of photon absorbance events (the other 99% producing fluorescence). Thus psCasMINI needed more photons to switch vs. natural photosensory domains, e.g. LOV or cryptochrome. And CasMINI was not as active as the standard Cas9.

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

Amazing postdoc Daesun Song took up the project. He quickly made a psCasMINI that showed ~13-fold induction of gene editing with light, and induced activity similar to unfused CasMINI. All within 2.5kB, small enough to fit, along with a gRNA gene, into a single AAV particle

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

But the cool thing is: in the Cas12f:gRNA complex, the two N-termini of Cas12f already flank the gRNA. Thus if we fuse just *1* pdDronpa domain to the Cas12f NT, the pdDronpa dimer should form across the gRNA to cage it Here's my manual model from 2021. No AF, no ML. Just PyMol.

110

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

Previously, we had made single-chain photoswitchable proteins by attaching pdDronpa at 2 locations flanking an active site. The pdDronpa dimer cages the site in the dark, while light induces dissociation and uncaging Cartoon below for psMEK (red molecule is ATP in active site)

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

The smallest CRISPR systems are the Cas12f family. Two Cas12f chains wrap around one sgRNA, allowing each to be less than 1/2 the size of the original Cas9 proteins where 1 chain wraps around 1 sgRNA. The first Cas12f variant used in mammals was CasMINI: www.cell.com/molecular-ce...

cell.com

Engineered miniature CRISPR-Cas system for mammalian genome regulation and editing

Xu et. al developed a miniature CRISPR system for genome engineering via protein and guide RNA engineering. Whereas the natural Cas12f does not function in mammalian cells, engineered Cas12f mutants, ...

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

pdDronpa is dimeric in the dark. Cyan light dissociates it into monomers. Redimerization occurs spontaneously in ~1h, or immediately with a little violet light. pdDronpa is the first (and only?) synthetic photosensory domain, and nicely doesn't require a cofactor as it's a GFP.

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

There were two key protein engineering steps: using our pdDronpa photodissociable GFP to control the compact CRISPR protein Cas12f with light, and improving photosensitivity of pdDronpa to make photoregulation more efficient in vivo. The 2017 paper on pdDronpa: www.science.org/doi/10.1126/...

science.org

Optical control of cell signaling by single-chain photoswitchable kinases

Single-chain photoswitchable kinases based on engineered photodissociable proteins enable fine control of kinase activity.

110

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

The solution: Make a photoswitchable CRISPR system small enough to fit in a single AAV, and produce light at specific locations in vivo by ultrasound-triggered light production from mechanoluminescent particles. Hmm, some engineering was needed...

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

The challenge: Can we select not only which gene to disrupt with a CRISPR system, but also where in the body to disrupt it? If we could do this, then we can eventually control gene function in particular parts of the body (especially the nervous system) even without a specific promoter.

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 29/09/2026

Excited to share a new study from our lab in Cell: “Non-invasive control of gene editing in vivo by photoswitchable Cas12f and focused ultrasound” We engineered a compact CRISPR system that can be activated at specific locations in the body using focused ultrasound (🧵) www.cell.com/cell/fulltex...

cell.com

Non-invasive control of gene editing in vivo by photoswitchable Cas12f and focused ultrasound

Remote activation of a single-AAV-deliverable photoswitchable Cas12f system enables precise spatial and temporal control of genome editing in deep tissues without surgery or implanted devices.

13710

Reposted by Michael Lin, MD PhD

Sarafan ChEM-H @stanford-chemh.bsky.social · 25/09/2026

#ICYMI Faculty Fellow @michaelzlin.bsky.social and team have developed RASER, a novel therapeutic approach that sharpens the precision of certain cancer treatments while sparing healthy tissue: med.stanford.edu/cancer/about...

med.stanford.edu

Cutting through cancer with RASER: A new approach to metastatic disease

By prioritizing treatment precision and minimizing toxicity, the RASER system offers hope for patients living with metastatic cancers.

022

Michael Lin, MD PhD @michaelzlin.bsky.social · 26/06/2026

Human forward thinking is irreplaceable so easy solution is to have multiple types of recognition.

000

Michael Lin, MD PhD @michaelzlin.bsky.social · 26/06/2026

Competition is inherent to all types of recognition; it’s a good motivator as long as it’s not taken to the extreme where people stop working together. And I’d prefer the consistency of an automated system to the biases and clubbiness of human systems if we only had one choice. But

100

Michael Lin, MD PhD @michaelzlin.bsky.social · 21/06/2026

Love this blog post on our ErbB-OSV: "A virus with a very specific grudge" Nice explanations of the advantages and the historical context oncobriefs.com/2026/06/the-...

010

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

Finally, IMO engineering holds the key to curing cancer. "Cancer biology" has become the study of disease at ever-increasing detail rather than trying new ideas for cures. Big pharma is also content with the status quo. But give an engineer a problem and they will try to solve it 🛠️🧪🧬

160

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

Yes there were 9 years when we had to no specific funding. But if you really know you have something important, then you find a way to keep going. Funding rates are so abysmal (5%) that reviewers don't favor new ideas; they just want to keep the existing lights on.

130

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

And a big thanks to the organizations who funded this work: • 2011-2013: @acgtfoundation.bsky.social • 2012-2015: @damonrunyon.org • 2024-2026: Bachrach Foundation • 2025-2026: Stanford SPARK • 2026-2027: Harrington Discovery Institute

140

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

Looking forward, we think this can work IV for other HER2+ cancers (Xinzhi has promising results already), and yes we are looking to get this into trials. If you're a clinician interested in helping, feel free to contact me.

120

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

Huge credit and congrats to first author Xinzhi Zou, who single-handedly carried out all aspect of this project. That included protein engineering, VSV construction (probably has made more anyone in the world), potency/specificity characterization, tumor imaging, and efficacy.

120

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

…realistic dosing amounts and intervals, comparisons to prior state-of-the-art, and proof of MOA. We believe bringing the quantitative rigor of small-molecule drug discovery and enzyme/sensor engineering can help virotherapeutics more rapidly and rationally improve.

120

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

So disease-specific computation could ultimately provide a cleaner route to specificity than target binding alone. Finally, a minor comment. This is the most rigorous first-report we know of for a new virotherapeutic, with systematic comparisons vs standard-of-care…

110

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

Maybe the next frontier is not finding more context-dependent targets, but in engineering therapeutics to be triggered by the causes we already know well. And specificity need not come from the 3d structure of the target. It can come from integrating the 4th dimension: Time!

120

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

For many cancers, we know the driving causes: • HER2 amplification • EGFR mutation • RAS mutation • other kinase pathways So then instead of ever-more detail characterization of consequences of oncogenic signaling, can we exploit the signaling abnormalities themselves?

120

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

The broader idea is what excites me most. For decades we've programmed therapies to recognize molecules statically. Can we instead program therapies to recognize pathological states by activity? Cancer, inflammation, neurodegeneration — all involve abnormal signaling.

220

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

And, in results that even surprised us, if the mice are treated a little earlier after ovarian cancer implantation (but still at a standard time in the field), all mice receiving ErbB-OSV were cured.

120

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

Toxicity was ~10000x lower than wild-type VSV while retaining potent anti-tumor activity. In immunocompetent mice with advanced metastatic HER2⊕ ovarian cancer, ErbB-OSV synergized with standard chemotherapy to nearly triple survival times compared to chemotherapy alone.

120

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

If the answer is very high, ErbB-OSV kills the cell while making more of itself — in its specific and autoregulatory nature, it is kind of like an immune cell but capable of querying cancer states inside cells. This produced a large improvement in safety and thereby efficacy

120

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

This is a different concept from "targeted" inhibitors. Targeted therapies try to suppress HER2 function or kill cells expressing HER2. Both are toxic to normal cells. ErbB-OSV is more intelligent. It asks "let me integrate HER2 activity for a little while".

110

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

ErbB-OSV killed 100% of HER2⊕ ovarian cancer cells at concentrations that were benign to HER2⊖ cells, or to HER2⊕ treated with a drug that inactivates (dephosphorylates) HER2. This proves the mechanism of action of ErbB-OSV is indeed rewiring HER2 hyperactivity to cell death.

130

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

That way VSV can amplify only in the presence of hyperactive ErbB, such as the overexpressed HER2 of ovarian, breast, and gastric cancer. We call this an ErbB oncoselective VSV, or ErbB-OSV.

110

Michael Lin, MD PhD @michaelzlin.bsky.social · 17/06/2026

In the new work, we use a VSV backbone to deliver RASER. VSV is already in clinical trials for cancer and is the basis of the FDA-approved Ebola vaccine. We asked if RASER could be used to inactivate VSV in normal cells by sequestering a VSV RNA replication factor as effector.

110