Optimal immune-based therapies for type 1 diabetes (T1D) should restore self-tolerance without inducing chronic immunosuppression. CD4+Foxp3+ regulatory T cells (Tregs) are a key cell population capable of facilitating durable immune tolerance. However, clinical trials with expanded Tregs in T1D and solid-organ transplant recipients are limited by poor Treg engraftment without host manipulation. We showed that Treg engraftment and therapeutic benefit in nonautoimmune models required ablative host conditioning. Here, we evaluated Treg engraftment and therapeutic efficacy in the nonobese diabetic (NOD) mouse model of autoimmune diabetes using nonablative, combinatorial regimens involving the anti-CD3 (αCD3), cyclophosphamide (CyP), and IAC (IL-2/JES6–1) antibody complex. We demonstrate that αCD3 alone induced substantial T-cell depletion, impacting both conventional T cells (Tconv) and Tregs, subsequently followed by more rapid rebound of Tregs. Despite robust depletion of host Tconv and host Tregs, donor Tregs failed to engraft even with interleukin-2 (IL-2) support. A single dose of CyP after αCD3 depleted rebounding host Tregs and resulted in a 43-fold increase in donor Treg engraftment, yet polyclonal donor Tregs failed to reverse diabetes. However, infusion of autoantigen-specific Tregs after αCD3 alone resulted in robust Treg engraftment within the islets and induced remission in all mice. This novel combinatorial therapy promotes engraftment of autoantigen-specific donor Tregs and controls islet autoimmunity without long-term immunosuppression.

Vector space embedding models like word2vec, GloVe, fastText, and ELMo are extremely popular representations in natural language processing (NLP) applications. We present Magnitude, a fast, lightweight tool for utilizing and processing embeddings.Magnitude is an open source Python package with a compact vector storage file format that allows for efficient manipulation of huge numbers of embeddings. Magnitude performs common operations up to 60 to 6,000 times faster than Gensim. Magnitude introduces several novel features for improved robustness like out-of-vocabulary lookups.

This paper discusses recent developments in sensors for the Harvard RoboBee. The RoboBee is a sub-100 mg flapping-wing micro-aerial vehicle that is able to lift its own weight under external power, but, like flying insects, is unstable in flight without active feedback. We discuss design and characterization of two low-latency insect-inspired sensors for flight control: an antenna to sense airspeed and light-sensing ocelli to estimate attitude angle relative to a luminous sky.W e demonstrate accurate wind velocity estimation in a wind tunnel despite the effect of nearby flapping wings. We also demonstrate pitch angle control using the ocelli on a wire-mounted RoboBee that is free to rotate about its pitch axis. These flight-weight sensors are essential first steps toward autonomous upright stability and controlled forward motions.

No abstract