The glucagon response is the first line of defense against hypoglycemia and is lost in insulin-dependent diabetes. The β-cell " switch-off" hypothesis proposes that a sudden cessation of insulin secretion from β-cells into the portal circulation of the islet during hypoglycemia is a necessary signal for the glucagon response from downstream α-cells. Although indirect evidence exists to support this hypothesis, it has not been directly tested in vivo by provision and then discontinuation of regional reinsulinization of α-cells at the time of a hypoglycemic challenge. We studied streptozotocin (STZ)-induced diabetic Wistar rats that had no glucagon response to a hypoglycemic challenge. We reestablished insulin regulation of the α-cell by regionally infusing insulin (0.025 μU/min) directly into the superior pancreaticoduodenal artery (SPDa) of STZ-administered rats at an infusion rate that did not alter systemic venous glucose levels. SPDa insulin infusion was switched off simultaneously when blood glucose fell to <60 mg/dl after a jugular venous insulin injection. This maneuver restored the glucagon response to hypoglycemia (peak change within 5-10 min = 326 ± 98 pg/ml, P < 0.05; and peak change within 15-20 min = 564 ± 148 pg/ml, P < 0.01). No response was observed when the SPDa insulin infusion was not turned off (peak change within 5-10 min = 44 ± 85 pg/ml, P = NS; and peak change within 15-20 min = 67 ± 97 pg/ml, P = NS) or when saline instead of insulin was infused and then switched off (peak change within 5-10 min = -44 ± 108 pg/ml, P = NS; and peak change within 15-20 min = -13 ± 43 pg/ml, P = NS). No responses were observed during euglycemia (peak change within 5-10 min = 48 ± 35 pg/ml, P = NS; and peak change within 15-20 min = 259 ± 129 pg/ml, P = NS) or hyperglycemia (peak change within 5-10 min = 49 ± 62 pg/ml, P = NS; and peak change within 15-20 min = 138 ± 87 pg/ml, P = NS). Thus, the glucagon response to hypoglycemia that was absent in rats made diabetic by STZ was restored by regional infusion and then discontinuation of insulin. These data provide direct in vivo support for the β-cell "switch-off" hypothesis and indicate that the α-cell is not intrinsically abnormal in insulin-dependent diabetes because of STZ-induced destruction of β-cells.