There is conflicting information as to the existence, prevalence and phenotype of cellular immune responses to insulin in pre-diabetic rodents and humans [1±6]. Furthermore, little is known about the immunological impact of insulin administration, other than it can boost insulin autoantibody levels ([7] and references therein). The subcutaneous administration of soluble antigen has been traditionally thought to be only weakly immunogenic in naive animals, necessitating the use of adjuvants to prime significant immune responses. However, it has recently been shown that the subcutaneous administration of prototypic soluble foreign antigens can prime naive T cells toward a Th2 phenotype [8]. This finding raises the question whether the subcutaneous administration of self-antigens, such as insulin to pre-diabetic and diabetic individuals, also induces Th2 responses. If insulin administration does have an immunological impact, the nature of this response (Th1/ Th2) may depend on whether an autoimmune response to insulin is already established and the phenotype of this response. If a memory Th1 response to insulin is present in pre-diabetic and diabetic animals, insulin injection may boost this pro-inflammatory response. On the other hand, if insulin adminis- Diabetologia (1998) Summary Little is known about the immunological impact of insulin administration other than it can boost insulin autoantibody levels. In particular, while the subcutaneous administration of a soluble foreign antigen (without adjuvant) is generally only weakly immunogenic in a naive animal, it is unknown what effect the subcutaneous administration of a soluble self-antigen has in animals with established autoimmune responses to the antigen. Addressing these questions in pre-diabetic nonobese diabetic (NOD) mice, we examined the effects of administering insulin, as well as the metabolically inactive B-chain of insulin, on insulin-specific cellular and humoral immune responses. We show that pre-diabetic NOD mice have a spontaneous Th1-biased response against insulin. Administering insulin, or the insulin B-chain, rather than boosting the established Th1 response, primed Th2 cellular and humoral immunity to insulin, shifting the predominant insulin response toward a Th2 phenotype. Despite the presence of a Th1 response against insulin, insulin treated mice failed to mount proliferative T-cell responses following immunization and challenge with insulin, demonstrating that the treatment induced an active form of tolerance to this autoantigen. Thus, the subcutaneous administration of a soluble antigen can engage Th2 responses and induce self-tolerance, even after the establishment of autoreactive Th-1 responses. Such immune deviation and induced regulatory tolerance may contribute to the protective effects of prophylactic insulin therapy, as well as the establishment of a ªhoneymoonº phase in new-onset insulin-dependent diabetic patients. [Diabetologia (1998) 41: 237±240]