Cabaletta Bio (CABA, $473M) is a Phase 1/2 CD19 autoimmune CAR-T company with three RMAT designations. The Q1 2026 10-Q (filed May 14) contains a buried regulatory disclosure that materially changes the BLA risk profile — and a peer 10-Q filed two days earlier shows the opposite FDA outcome on the same architecture.

What the filing says

The new risk factor, below the MD&A: "In March 2026, the FDA noted that details are needed in order for the FDA to assess suitability of the proposed registry as an external control, including information on potential baseline differences between the enrolled study population and registry population (e.g., diagnosis classification and refractory definition) and the risk of missing or unavailable data." This is procedural pushback on the 17-patient single-arm DM/ASyS registrational design — the first BLA path.

The May 2026 follow-on raised $141M net at $2.90/share (above the $2.50 warrant strike), extending runway to mid-2027 from a prior Q4 2026 going-concern horizon. Pro forma cash $257M. Shares outstanding 163M plus 50.3M warrants at $2.50 expiring Sept 12, 2026.

The 13-patient RESET-MG dataset (presented at AAN April 2026, data cutoff March 6) shows 10/13 (77%) clinically meaningful MG-ADL improvement off immunomodulators; 13/13 no ICANS; 11/13 no CRS. The cohort included 6 AChR-negative patients — a CD19 response that is mechanistically novel since seronegative MG lacks the canonical pathogenic autoantibody. The 10-Q does not break out response rate by serotype.

Manufacturing: April 2026 commercial supply agreement with Cellares (Cell Shuttle + Cell Q). First two autoimmune patients already manufactured on the Cellares platform; translational data due at ASGCT May 18-21, 2026.

What the market thinks

CABA is +221% over the trailing year, but recent moves have been beta-explained: the +24.9% 1W rally tracked XBI +9.1% × β 3.29. Empirical 12-month regression: α = -0.4%, σ_idio = 99.6%, idio variance = 82.3%. Institutional ownership is 82% but quant-heavy (Millennium, Two Sigma, Jane Street); biotech specialists (Perceptive, OrbiMed, RA Capital) are absent from the top 20. Big Pharma autoimmune CAR-T M&A: $7.35B+ into in vivo and allogeneic over nine months, zero into autologous. CABA was mentioned in zero of 6,186 Q4 2025 Big Pharma earnings transcripts.

Three observables: easy money has been extracted, the marginal buyer is statistical not fundamental, and pharma's revealed preference excludes autologous CAR-T from M&A optionality.

Why the gap exists

Two days before CABA filed, Kyverna (KYTX) filed a 10-Q on the same regulatory architecture — CD19 autologous CAR-T, RMAT, single-arm Phase 2 + retrospective natural history external comparator — and disclosed a "positive pre-BLA meeting with the FDA and gained alignment on our regulatory path for miv-cel in SPS, including a rolling BLA submission." Rolling BLA initiated May 2026; target Q4 2026.

Same week, two filings, opposite FDA outcomes. The discriminator is indication: SPS has no approved therapy (clean reference class); DM/ASyS is a heterogeneous basket where "diagnosis classification" is exactly the concern FDA cited. The market treats the cohort as homogeneous — the filings show it isn't. The pushback is buried in CABA's risk factors, not the MD&A; analyst notes have not picked it up.

The AChR-negative signal compounds this. CABA disclosed 7 AChR+ / 6 AChR- explicitly; KYTX (KYSA-6, same AAN session) did not break out AChR status, and its FcRn-failure inclusion criteria imply an AChR-positive skew (FcRn inhibitors are only approved for AChR+ gMG). Same audience, same week, different transparency.

Risks (ranked)

1. FDA rejects the DM/ASyS external control. Forces randomized control arm — 1-2 year delay, $50M+ additional capital. The 10-Q says FDA "noted details are needed," not that FDA accepted. Resolution likely Q3-Q4 2026.
2. H1 catalyst stack disappoints. Four independent readouts in approximately six weeks: SLE preconditioning Phase 1/2 complete, SLE no-precondition initial, SSc Phase 1/2 + registrational design, ASGCT manufacturing translational data.
3. Specialist absence persists. Zero biotech-specialist funds in top 20 holders. They've seen the file and not bought.
4. Cohort-wide autologous discount. Big Pharma M&A flow ($7.35B) routed entirely to in vivo and allogeneic. Strategic bid structurally diverted.

Catalysts

• ASGCT May 18-21, 2026: Cellares manufacturing translational data
• H1 2026 (≈6 weeks): SLE no-preconditioning data, SSc Phase 1/2 + registrational design
• Mid-2026: RESET-MG registrational cohort design announcement
• Q3 2026: FDA Type B/C response to CABA's external control submission (the binary)
• Sept 12, 2026: 50.3M warrants at $2.50 expire
• Year-end 2026: KYTX miv-cel SPS BLA acceptance (predicted 82%)

What would change our mind

• 8-K disclosing FDA acceptance of the registry as external control → compresses the bear case
• 8-K disclosing FDA requirement for randomized control arm → kills 2027 BLA path
• ASGCT data showing Cellares-manufactured product matches traditional process on cell expansion and potency → de-risks manufacturing factor
• A second autoimmune CAR-T sponsor disclosing similar FDA pushback in upcoming 10-Qs → sectoral, compresses the divergence
• AChR-negative response published at 5/6 or 6/6 → mechanistic differentiation is real
• AChR-negative response published at 2/6 or 3/6 → the headline 77% was AChR+ driven; no platform-extension argument

Evidence

The 1Y +221% has already discounted the cohort re-rating. The remaining mispricing sits in two specific places: a buried regulatory disclosure that contradicts the consensus assumption of a homogeneous autoimmune CAR-T cohort, and an AChR-negative MG signal that one peer chose not to address at the same conference. Both compress if the next six weeks of catalyst data resolve to the bull side; both expand if Q3 FDA feedback formalizes the procedural pushback.