---
title: "Eight Things FDA Pushed Back On: Lessons From 16 Published Genome Editing Filings"
description: The NIH's SCGE consortium publishes its FDA correspondence. Eight places the agency did not agree with what a sponsor proposed on off-target work, with the quotes.
---

[![Logo\_horizontal\_colored logo with black text](https://www.countagen.com/hubfs/Logo_horizontal_colored%20logo%20with%20black%20text.png)](https://www.countagen.com/)

- [Technology](https://www.countagen.com/technology)
- [Bioinformatics](https://www.countagen.com/bioinformatics)
- [Sample Report](https://www.countagen.com/lockseq-sample-report)
- [Resources](https://www.countagen.com/resources)
- [FAQ](https://www.countagen.com/faq)
- [About Us](https://www.countagen.com/about-us)

[Request a quote](https://www.countagen.com/contact)

FDA Guidance

# Eight Things FDA Pushed Back On: Lessons From 16 Published Genome Editing Filings

The NIH's SCGE consortium publishes its FDA correspondence. Eight places the agency did not agree with what a sponsor proposed on off-target work, with the quotes.

[Countagen Team](https://www.countagen.com/countagen-insights/author/countagen-team)

 Nov 18, 2026

---

Most of what we know about FDA's expectations for off-target work comes from guidance documents — texts written in the abstract, for everyone. What almost nobody gets to see is the other half: what FDA actually tells a named sponsor about a named product, in writing, when the package is real.

That half exists in public, and hardly anyone reads it.

The NIH Common Fund's Somatic Cell Genome Editing (SCGE) consortium publishes its FDA interactions on the [SCGE Platform](https://scge.mcw.edu/platform/public/documents/regulatory): sponsor INTERACT and pre-IND briefing books paired with FDA's written responses and signed meeting summaries. Sixteen documents, seven rare disease programmes, from 2024 through to June 2026. Publication is a condition of consortium membership — which is also why you will not find a commercial sponsor's correspondence there, and never will.

We read all sixteen. What follows are the eight places where FDA did not agree with what the sponsor proposed, or asked for something the sponsor had not thought to include.

A note on how to read these. Pre-IND and INTERACT advice is specific to one product and one package. None of this is FDA stating a general rule, and we have been careful not to present it that way. What it does show is how the April 2026 NGS draft guidance is being applied in review, by the same reviewers, across modalities — and the consistency is striking.

## The seven programmes

| Programme | Sponsor | Editor | Delivery & target | Stage reached |
| --- | --- | --- | --- | --- |
| Phenylketonuria | CHOP / Penn | Adenine base editor | LNP, IV, liver | Pre-IND, Jul 2025 |
| Urea cycle disorders | CHOP / Penn | Adenine base editor | LNP, IV, liver | Pre-IND, Oct 2025 |
| Urea cycle disorders | CHOP / Penn | Prime editor | AAV8 template + LNP, liver | Pre-IND, Feb 2026 |
| Prion disease | Broad Institute | Epigenetic editor (zinc finger) | AAV9, IV, CNS | INTERACT 2025, pre-IND Jul 2026 |
| Leber congenital amaurosis 16 | UW-Madison | Adenine base editor | LNP, subretinal, RPE | INTERACT, 2025 |
| Best disease | UW-Madison | Nuclease, allele-specific | LNP, subretinal, RPE | INTERACT, Jun 2026 |
| Spinal muscular atrophy | Jackson Labs / UT Southwestern | Adenine base editor | Dual AAV9, intrathecal, CNS | INTERACT, 2024 |

Six of the seven are in vivo editing — the setting where the editing happens in cells you cannot easily obtain, and the setting where FDA's comments get most specific.

![Where FDA pushed back, by stage of the off-target workflow: eight findings placed across nomination, confirmation and genome integrity, with two that apply across the whole programme](https://www.countagen.com/hubfs/blog-figures/SCGEFig1_Where_FDA_Pushed_Back.png)

## 1. Your nomination assay has to match your editor's mechanism

The SMA programme (dualAAV9/SMABE, PTS PS009261) proposed CHANGE-seq on HEK293T genomic DNA to nominate candidate off-target sites for an adenine base editor.

FDA's response was blunt: *"We do not agree with your proposal to implement CHANGE-seq using HEK293T genomic DNA."*

Two separate objections sit inside that one sentence. The material should be human cells or genomic DNA relevant to the product, not an immortalised line of convenience. And more fundamentally: CHANGE-seq is built around Cas9, which makes double-strand breaks, while an adenine base editor nicks one strand. Sites nominated by a double-strand-break assay would not tell you what happens in cells edited by a nickase.

The sponsor resolved it in the meeting by citing a publication supporting CIRCLE-seq, which FDA accepted — and also committed to adding in silico nomination as a further orthogonal method.

**The point:** an unbiased genome-wide assay is not automatically the right one. The April 2026 draft says this at §VI.B, noting that assays developed for double-strand breaks may not detect single-strand nicks, and that an endonuclease step can be added to convert them. If your editor nicks, your nomination assay needs to account for it.

## 2. A supersaturating dose is not a free pass

Running confirmation at a deliberately high editing dose is a common and reasonable move: rare events are easier to detect when you push the system. Two SCGE programmes proposed it. They got different answers eleven weeks apart.

In the PKU pre-IND (PS008879), the sponsor proposed confirming off-targets at 20× EC90. FDA accepted the rationale for surfacing rare events, but required the findings to be contextualised with testing at clinically relevant doses in primary human hepatocytes.

In the urea cycle disorders base editing pre-IND (PS010343), the same group proposed the same design. This time FDA drew a line through the dose while leaving the method standing: *"While your proposal to use targeted sequencing for confirmatory testing is acceptable, we do not agree with your proposal to use 20 times the effective concentration EC90 for this analysis. Based on the data published in a recent publication \[PMID: 39631713\], we are concerned that the use of high concentration of LNPs can be detrimental to adequate uptake and trafficking of your drug product to the nucleus of cells. This may potentially impede the genome editor activity and off-target editing rate outcome."*

Worth being precise about what was refused. The same response opens by telling the sponsor that their *"on-target edit analysis, bystander edit analysis, off-target nomination, and confirmatory testing methods are acceptable."* The assay was fine. The concentration it was to be run at was not.

The concern is subtle and worth sitting with. It is not that a high dose over-reports off-targets. It is that a high LNP concentration may impair delivery badly enough to *under*-report them — the editor never reaches the nucleus efficiently, and the off-target readout is distorted downward.

It came up a third time, in a different form. In the prime editing programme the sponsor proposed a saturating dose of AAV for its integration site analysis. FDA's reply: *"It is not clear what a saturating dose of AAV means. Additionally, it is also not clear why a saturating dose will be used in this study. Please provide justification with data to support your strategy."*

**The point:** the April 2026 draft asks at §VI.D that on-target editing rates in the confirmation study be comparable to the intended rates in the final drug product. A dose chosen for assay convenience has to be defended against that — and "saturating" is not a specification.

## 3. Immortalised cell lines are a liability, and more so for an epigenetic editor

The prion disease programme (PRNP-CHARM-001, PS009720) is the only epigenetic editor in the set — a zinc finger DNA-binding domain fused to a DNMT3A-based silencer and a KRAB repressor. The sponsor's position was that dose-responsive RNA-seq across five human cell lines characterised specificity, on the argument that any phenotypic off-target silencing shows up as a change in gene expression.

FDA could not agree the strategy was sufficient: *"We are concerned that the methylation profile of immortalized human cell lines may not recapitulate the methylation profile of primary human cells (PMID: 24371281). Please conduct additional off-target analysis in relevant primary human cells to characterize the off-target profile of PRNP-CHARM-001. If off-target analysis in relevant primary human cells is not feasible, please provide justification."*

Note the shape of it. FDA did not refuse the category — §VI.A allows immortalised lines with scientific justification where primary cells are not feasible. It asked for the justification, and pointed out that the panel did not contain the primary therapeutic target cell. In the meeting the sponsor agreed to add iPSC-derived induced neurons, primary hepatocytes, and heart and adrenal lines chosen from biodistribution data.

**The point:** for an editor whose mechanism *is* a change to the epigenome, the epigenetic state of your test system is not a detail. It is the assay.

## 4. Nomination and confirmation can both be fine and the package still has a hole

The Best disease INTERACT (UW-LNP-SpCas9-BEST1-R218C, PS010799, signed 1 June 2026) is the most recent document in the corpus, and the most instructive about where the new guidance is heading.

FDA found the sponsor's nomination and confirmation strategies generally acceptable. Then it asked for something that had not been provided at all: *"Please use sensitive, quantitative next generation sequencing (NGS)-based method(s) or equivalent methods to assess the genomic integrity of cells post-genome editing. Additionally, please use appropriate sequencing-based and/or equivalent methods to assess the target edit site for the presence of large structural changes such as large insertions, large deletions, etc."*

This is also the first document in the set where FDA points a sponsor directly at the April 2026 draft as a review instrument — for PAM selection in in silico nomination, for what to include in biochemical off-target study reports, for what to include when reporting nominated and confirmed sites, and for how to describe the sequencing strategy, bioinformatics analysis and results of whole-genome and long-read studies.

**The point:** off-target site analysis and genome integrity are two different obligations. §VIII covers chromosomal integrity for modalities that create double-strand breaks or medium-to-large indels, including translocations between on-target sites and between on-target and confirmed off-target sites. A complete off-target package does not satisfy it.

## 5. If the editor stays expressed, one timepoint is not enough

Two programmes, both AAV-delivered, ended up with longitudinal obligations.

At the prion INTERACT, FDA recommended the sponsor *"evaluate the pharmacokinetic profile of CHARM to assess persistence of protein expression and discuss how this may impact the rate of off-target silencing over time."* Characterise how long the editor lasts, then say what that does to off-target risk. The sponsor's answer by the pre-IND was a dedicated long-term in vivo study comparing differentially expressed gene burden at day 30 against day 180 after a single dose — citing §VI-G of the April 2026 draft by name.

The SMA programme planned an 18-month longitudinal mouse study to understand whether off-target edits accumulate. FDA did not simply accept it. It asked the sponsor to *"provide your rationale for the study duration of 18 months"*, and recommended *"a stepwise approach to characterize the in vivo persistence and BD of ABE prior to conducting a long-term safety study"* — supported by the sponsor's own data or published kinetics for similar editors, because that information governs *"selection of the appropriate sacrifice time points and study durations."*

That is the more demanding version of the same idea. It is not enough to run a long study; the length has to come from what you know about how long the editor persists.

No LNP/mRNA programme in the corpus was asked for anything equivalent, which fits: §VI.G frames this around durability of editor expression, contrasting short-lived mRNA delivery with sustained AAV expression.

**The point:** delivery choice creates a downstream evidence obligation. If your editor is still being expressed at six months, a single early timepoint describes a system that no longer exists — and the timepoints you do choose are expected to follow from persistence data rather than from convenience.

## 6. Allele-specific editing means measuring the allele you are not targeting

In the Best disease programme the guide differs from the normal genic site by a single nucleotide. FDA asked the sponsor to report the editing rate at the mutation-harbouring site, at the normal allele site in heterozygous cells, and at the endogenous BEST1 gene — separately — and to describe the bioinformatics steps used to assign indel-harbouring sequences back to either the mutation-harbouring site or the normal genic site.

It also asked for total reads covering the mutation-harbouring allele and the wildtype allele both *before* and *after* editing, in heterozygous patient-derived cells.

**The point:** when your specificity argument rests on one base, the read-assignment logic is part of the evidence, not an implementation detail. Three numbers, not one, a documented method for telling the reads apart, and a baseline to compare against.

## 7. An engineered cell model creates an attribution problem

Where sponsors built cell models carrying the target variant on a lentiviral cassette or as a knock-in, FDA asked the same question across three programmes: are you measuring the cassette or the native locus? In the two urea cycle programmes it asked for appropriate primers enabling independent measurement at the two distinct genomic locations. In Best disease it asked instead for a description of the bioinformatics steps used to assign indel-harbouring sequences back to one site or the other. Different mechanism, same requirement: the number has to be attributable.

In the prime editing programme (PS010700), it went further — raising that DNA nicking at the target site could cause unintended integration of the AAV-delivered template, and asking for long-read sequencing to assess it, with read counts supporting integration events reported at both the cassette and the native site.

**The point:** a model built to make the edit measurable can make it ambiguous. Whatever you report has to say which copy it came from.

## 8. An INTERACT agreement is a snapshot, not a commitment

At the prion INTERACT in February 2025, FDA agreed that humanised TFRC knock-in mice were an appropriate single species for the pivotal toxicology study.

At the pre-IND in July 2026, it withdrew that position — stating explicitly that the advice differed from its INTERACT feedback because new developments in the field had changed its understanding of the risks of capsids engineered to cross the blood-brain barrier. A large animal safety study was now expected.

This is one instance, with a stated scientific reason, in a fast-moving area. It is not evidence that agency advice is unreliable.

**The point:** seventeen months is a long time in this field. If a programme is building on early agreement, it is worth knowing what has changed underneath it.

## The thing we did not expect to find

Running through the Bioinformatics sections of these responses is something more concrete than any individual comment.

FDA sent two different sponsors — different institutions, different meeting types, months apart — the same two template workbooks: `NGS_qc+metadata_v0.3.xlsx` and `Software_tools_and_CLI_v0.3.xlsx`. The second asks for tool versions, licensing, publication and repository information, operating system and hardware requirements, and command-line detail for every computational step.

Elsewhere, when the prion sponsor summarised its RNA-seq findings, FDA asked for the results in Excel including raw and normalised counts, the complete list of differentially expressed genes for each cell line, and both raw and adjusted p-values at gene level. A table of selected hits was not enough.

Read together with §VI.F — which specifies genomic coordinates, mismatches and bulges, PAM information, genic context, and a literature-based risk assessment for every confirmed site — a picture emerges that is easy to miss when reading the guidance alone. The reporting format is being standardised, in practice, ahead of the document that formalises it.

Which means the shape of what you hand over is not an afterthought at the end of the work. It is part of the work.

### Sources

All sixteen documents are published by the SCGE Consortium at [scge.mcw.edu](https://scge.mcw.edu/platform/public/documents/regulatory) under a [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) licence, and were read in full in November 2026.

Programmes cited: phenylketonuria (CHOP/Penn, PTS PS008879) · urea cycle disorders, base editing (CHOP/Penn, PS010343) · urea cycle disorders, prime editing (CHOP/Penn, PS010700) · prion disease (Broad Institute, PS009720) · Leber congenital amaurosis 16 (UW-Madison, PS009898) · Best disease (UW-Madison, PS010799) · spinal muscular atrophy (Jackson Labs / UT Southwestern, PS009261).

Guidance referenced: FDA, *Safety Assessment of Genome Editing in Human Gene Therapy Products Using Next-Generation Sequencing*, draft guidance, April 2026, docket FDA-2026-D-1255. Draft guidance represents the agency's current thinking and is not binding.

The SCGE Consortium, the sponsors named above, NIH and FDA have no involvement in this article and no association with Countagen.

[FDA Guidance](https://www.countagen.com/countagen-insights/tag/fda-guidance) [Regulatory](https://www.countagen.com/countagen-insights/tag/regulatory) [Off-Target Confirmation](https://www.countagen.com/countagen-insights/tag/off-target-confirmation)

## Similar posts

<https://www.countagen.com/countagen-insights/fda-april-2026-guidance-off-target-confirmation>

FDA Guidance

### [What FDA's April 2026 Draft Guidance Means for Off-Target Confirmation](https://www.countagen.com/countagen-insights/fda-april-2026-guidance-off-target-confirmation)

 Countagen Team  Sep 15, 2026

<https://www.countagen.com/countagen-insights/fda-vs-china-cde-off-target-studies>

FDA Guidance

### [Off-Target Studies for Gene-Edited Therapies: FDA, the EU and China](https://www.countagen.com/countagen-insights/fda-vs-china-cde-off-target-studies)

 Countagen Team  Oct 15, 2026

<https://www.countagen.com/countagen-insights/fda-2026-genome-editing-drafts-off-target-methods>

FDA Guidance

### [How FDA's 2026 Genome Editing Drafts Connect on Off-Target Methods](https://www.countagen.com/countagen-insights/fda-2026-genome-editing-drafts-off-target-methods)

 Countagen Team  Nov 14, 2026

### Get notified on blog articles and insights

Be the first to know about new LockSeq insights and related articles.

[![Countagen\_Logo\_White](https://www.countagen.com/hubfs/Logo_horizontal_colored%20logo%20with%20white%20text.png)](https://www.countagen.com)

| Countagen AB Nobels väg 16 171 65 Solna Sweden | Registered supplier on [Scientist.com](https://app.scientist.com/providers/countagen-ab) [![Scientist.com](https://www.countagen.com/hs-fs/hubfs/Scientist-com.png?width=1244&height=259&name=Scientist-com.png)](https://app.scientist.com/providers/countagen-ab) |
| --- | --- |

<https://www.linkedin.com/company/countagen/>

### LockSeq

- [Overview](https://www.countagen.com)
- [Technology](https://www.countagen.com/technology)
- [Bioinformatics](https://www.countagen.com/bioinformatics)
- [Sample report](https://www.countagen.com/lockseq-sample-report)

### Resources

- [Resources](https://www.countagen.com/resources)
- [FAQ](https://www.countagen.com/faq)

### Company

- [About us](https://www.countagen.com/about-us)
- [Contact](https://www.countagen.com/contact)

© 2026 Countagen AB All rights reserved [Terms of Use](https://www.countagen.com/terms-of-use) [Privacy Policy](https://www.countagen.com/privacy-policy) [Cookie Policy](https://www.countagen.com/cookie-policy) [Service Terms & DPA](https://www.countagen.com/legal)

[Powered by Atlas - a B2B SaaS HubSpot theme](https://www.kalungi.com/atlas-hubspot-theme-for-b2b-saas-software)

```json
{
  "@context" : "https://schema.org",
  "@type" : "BlogPosting",
  "author" : {
    "@type" : "Person",
    "name" : "Countagen Team",
    "url" : "https://www.countagen.com/countagen-insights/author/countagen-team"
  },
  "dateModified" : "2026-11-18T14:11:07.709Z",
  "datePublished" : "2026-11-18T14:08:42.000Z",
  "headline" : "Eight Things FDA Pushed Back On: Lessons From 16 Published Genome Editing Filings",
  "mainEntityOfPage" : {
    "@id" : "https://www.countagen.com/countagen-insights/fda-pushback-published-genome-editing-filings",
    "@type" : "WebPage"
  },
  "publisher" : {
    "@type" : "Organization",
    "logo" : {
      "@type" : "ImageObject",
      "url" : "https://www.countagen.com/hubfs/Countagen%20logo.png"
    },
    "name" : "Countagen AB"
  }
}
```