Forty days after Roche dropped a CD19 TCE, it licensed in another CD19/CD79a trispecific

In Roche Group’s H1 2026 R&D pipeline document (status as of July 23), the “Removed from phase I” column has one line: RG6382, CD19 × CD3, SLE. Roche’s stated reason was that the molecule did not have the profile needed to advance further in this indication; SLE was its lead indication. Forty days later, Roche took global rights to SIM0660.

This is where one must rein in the claims. Roche has not yet disclosed the intended indications for SIM0660. RG6382 was a molecule aimed at SLE, while SIM0660’s public positioning is first B-cell lymphoma and only secondarily autoimmune potential. Public information cannot support calling the latter a one-for-one replacement for the former. BD from first contact to signing also usually takes far longer than forty days, and the signing is not a consequence of the termination — the two events more likely share the same internal assessment, or may even be unrelated.

What can be established is the picture the two events jointly sketch. Roche’s B-cell depletion assets on the autoimmune side number three: Gazyva (anti-CD20 mAb, the most mature leg); Lunsumio (CD20×CD3) advanced into Phase 2 in SLE, single-arm subcutaneous, 30 patients; and P-CD19×CD20-ALLO1, the allogeneic CAR-T from the Poseida acquisition, in Phase 1 in SLE, 162 patients.

All three assets include CD20. After RG6382’s termination, Roche still holds one CD19-containing molecule in autoimmunity — that CD19×CD20 dual-target CAR-T is still advancing, so “a CD19 gap” doesn’t hold up. The accurate statement is: what Roche lost was the antibody-format, pure CD19-directed hand; the cell therapy leg remains, but it is a different mode of administration, a different cost structure, a different accessibility radius.

The oncology picture is similar. Among Roche’s marketed B-cell drugs, Rituxan, Gazyva, Columvi and Lunsumio are all anchored on CD20, and Polivy on CD79b. Columvi’s supplemental filing in second-line DLBCL received a complete response letter last July, after ODAC voted 8 to 1 against the applicability of the STARGLO study population to US patients — that study enrolled only 25 US patients. The more complete the backbone, the more a handle outside the backbone is needed. SIM0660’s CD79a plus CD19 do not overlap in target with any of those five.

CD79a and CD79b are the two signaling components of the B-cell receptor. There is only one marketed CD79-class drug, Roche’s own Polivy, which targets CD79b. A 2007 study published in Blood offered a specific observation: in that comparison, anti-CD79b antibodies downregulated surface BCR and were trafficked to lysosome-like compartments, and anti-CD79b ADCs were therefore more effective than anti-CD79a ADCs.

Note the boundary of this evidence. What it proves is that in that particular set of antibody and ADC designs, the CD79a approach performed worse than CD79b; it cannot be generalized to “CD79a as a target has been eliminated by the ADC route”, nor does it directly prove that CD79a stays on the cell surface long-term.

But directionally it does suggest a question worth thinking about: an ADC must be internalized and degraded to release its payload, whereas a T-cell engager needs the antigen to stay on the surface to form an immunological synapse. The same internalization property means opposite things for the two modalities. This is a reasonable line of thought and the most plausible explanatory framework for this target combination — but it is a hypothesis. Neither the announcement nor the press release provides any data supporting it, and SIM0660 has no externally verifiable preclinical dataset in the public domain.

As of the search date, no clinical trial registration explicitly using CD79a as a therapeutic target was found. That is both the source of the “potential first-in-class” claim and means that CD79a as a TCE target is completely unvalidated in humans.

Another company claim is that SIM0660 “limits cytokine release” while inducing potent T-cell cytotoxicity. This must be read in the context of the field: CLN-978 uses step-up dosing starting at 10 micrograms to manage cytokine release syndrome, KT501 uses CD3 masking, CMG1A46 uses a low-affinity CD3 arm — almost every CD19-class TCE going into autoimmunity is working on the same problem. Simcere did use a low-affinity CD3 antibody design on its sibling SIM0500, showing this is not empty talk, but for SIM0660 it remains only a design claim.

First, lay out the comparable deals.

Upfronts follow each deal announcement’s disclosure basis; headline total is upfront plus milestone cap, excluding royalties, in the announcement’s original currency. The deals differ in stage, indication, number of programs covered, milestone structure and royalty rates, none fully disclosed; the amounts cannot be directly used to compare asset quality or to derive the price of any single right.

What can be said: SIM0660’s upfront sits at the upper end of the preclinical, global rights tier, closest to Antengene’s ATG-201; the upfront is 4.9% of the headline total, consistent with the general structure of this set of deals.

SIM0660 is the first time Simcere has licensed out Greater China as well; the previous five all retained China:

This time it is “global”, and manufacturing is given up too.

There is a clock in the background: Simcere Zaiming filed its A1 with HKEX on January 9 this year, with CICC and Morgan Stanley as joint sponsors; after the spin-off, Simcere Pharmaceutical will still hold more than 50%. The group’s H1 2026 revenue was RMB 4.58 billion, profit for the period RMB 827 million, and cash at bank RMB 3.072 billion.

The group has ample cash, but that does not imply “this deal has nothing to do with financing” — the subsidiary’s own funding needs, risk transfer and reallocation of R&D resources are all legitimate motives, and not mutually exclusive. What can be said with certainty is that the deal also helps the spin-off listing narrative: in Simcere Zaiming’s IPO story, the TCE multispecific antibody platform is one of its three in-house platforms, and until now this line had only one external anchor, SIM0500, an option deal with undisclosed upfront; now it has become two deals, two multinationals, a disclosable upfront, and a second deal covering global rights.

And one longer-term observation to treat with more caution: the “removed” column in big pharma’s quarterly pipelines may be closer to their real gaps than the “added” column. Roche striking RG6382 in July and buying SIM0660 in September is a timing sample worth recording, but one sample does not make a pattern — it is worth treating as a lead to track, not as a conclusion.