Female NOD mice, the canonical mouse model for type 1 diabetes, develop disease in 77 percent of animals by 30 weeks if untreated. A five-day course of low-dose anti-CD3 antibody, the same class the FDA approved in 2022 to delay T1D onset in at-risk humans, cut that to 66 percent. A beta-cell-targeted GLP-1 conjugate carrying 17β-estradiol, given subcutaneously for 18 weeks, cut it to 61 percent. Together, in a four-arm randomised study published online May 18 in Diabetologia ↗, the two interventions dropped diabetes incidence to 38 percent (p≤0.001 vs untreated). The combination delayed disease onset by six weeks, and the protection held for five weeks after the last dose.
What anti-CD3 is, and what it cannot do alone
Teplizumab, sold as Tzield, became the first disease-modifying T1D therapy when the FDA approved it in November 2022. The label is narrow. It is for people in stage 2 of the disease, when the autoantibodies are positive and glucose handling is starting to fail but the patient is not yet insulin-dependent. In the pivotal TN-10 trial, a 14-day infusion course delayed progression to clinical (stage 3) diabetes by a median of about two years. Responders gained more, non-responders barely moved. The drug works by transiently exhausting the autoreactive T cells that destroy pancreatic beta cells.
Two years is something. It is not a cure. The mechanism is one-sided. The immune attack is suppressed for a window, but the beta cells themselves are left to fend off whatever damage was already in motion. Once the drug clears, the autoimmunity resumes and the surviving beta cells, already stressed and partially dedifferentiated, finish failing. Most of the past decade of T1D-prevention work has been searching for the missing second mechanism.
What GLP1-E2 actually is
The new piece is a fusion compound. The GLP-1 ligand binds the GLP-1 receptor, which beta cells express at high density and most other tissues at much lower levels. Conjugated to the GLP-1 backbone is 17β-estradiol, the same estrogen that has been known since the 1990s to protect beta cells in models of stress and apoptosis through estrogen receptor alpha. Free estradiol cannot be given systemically for diabetes prevention. The off-target effects on reproductive tissue and breast tissue close that door. GLP1-E2 uses the GLP-1 receptor ↗ as a delivery address. Beta cells get the estrogenic payload. The rest of the body sees mostly a GLP-1 ligand, the same chemistry semaglutide ↗ and the rest of that drug class use every day in millions of patients.
Dose in the trial: 100 nmol per kg per day, subcutaneously, for 18 weeks. Anti-CD3 was given intravenously at 2.5 micrograms per day for five days, the minimal-pulse design that has carried over from human Phase 2 work.
What the spatial transcriptomics showed
The authors ran spatial transcriptomics on pancreas tissue from all four arms. The progression signature in untreated NOD mice is well-characterised. Beta cell stress markers go up (Hspa5, Eif2ak3, Xbp1, Ddit3, the unfolded-protein-response cassette). Beta cell identity is lost as dedifferentiation markers like Cd81 climb. Antigen-presentation genes (H2-K1, H2-Q6, H2-Ab1, H2-Eb1) and inflammatory chemokines (Cxcl10, Cxcl9, Ccl5) light up.
In the treated arms, all three groups attenuated this. The asymmetry is the part worth quoting. Anti-CD3 mostly restored beta cell identity, knocking down dedifferentiation and antigen-presentation signatures. GLP1-E2 mostly reduced beta cell stress and immunogenicity, the upstream events that recruit T cells in the first place. The combination hit both axes. CD81 staining and TUNEL labelling of dying cells in 17-week-old treated mice looked like 12-week-old normoglycaemic NOD mice, rewinding five weeks of disease progression at the histological level. Beta cell mass at 26.5 weeks was preserved compared with newly diabetic untreated controls. GLP1-E2 also reduced islet immune-cell infiltration to roughly the same extent as anti-CD3, without depleting peripheral lymphocytes, which is the safety angle most T1D-prevention drugs trip over.
What this changes
T1D prevention has been searching for combination therapy for a long time. Verapamil to protect beta cells. GAD-alum vaccines for antigen-specific tolerance. Anti-thymocyte globulin or rituximab for deeper immune suppression. Each got partial credit in human trials. None compounded with teplizumab cleanly enough to justify a label expansion.
GLP1-E2 is the first asset that shows clean mechanistic complementarity with anti-CD3 in the same study, with the immune and beta-cell-protective arms each carrying roughly half the effect and the combination strictly better than either alone. Mouse data does not predict humans, and in T1D it almost never has. But the translational stack is unusually friendly. The GLP-1 chemistry is the best-tolerated drug platform in medicine right now. Anti-CD3 is already approved with a known safety profile. A combination Phase 2 in stage 2 T1D patients identified by autoantibody screening is the obvious next step, and the data here is the first preclinical readout in years that argues for running it.