Here we present Sm-Nd, Re-Os, and Pb isotopic data of carefully screened, least altered samples of boninite-like metabasalts from the Isua Supracrustal Belt (ISB, W Greenland)that characterize their mantle source at the time of their formation. The principal observations of this study are that by 3.7-3.8 Ga melt source regions existed in the upper mantle with complicated enrichment/depletion histories. Sm-Nd isotopic data define a correlation line with a slope corresponding to an age of 3.69+/-0.18 Gy and an initial epsilonNd value of +2.0+/-4.7. This Sm-Nd age is consistent with indirect (but more precise) U-Pb geochronological estimates for their formation between 3.69-3.71 Ga. Relying on the maximum formation age of 3.71 Gy defined by the external age constraints, we calculate an average epsilonNd [T = 3.71 Ga] value of +2.2+/-0.9 (n = 18, 1sigma) for these samples, which is indicative of a strongly depleted mantle source. This is consistent with the high Os concentrations, falling in the range between 1.9-3.4 ppb, which is similar to the estimated Os concentration for the primitive upper mantle. Re-Os isotopic data (excluding three outliers) yield an isochron defining an age of 3.76+/-0.09 Gy (with an initial gammaOs value of 3.9+/-1.2), within error consistent with the Sm-Nd age and the indirect U-Pb age estimates. An average initial gammaOs [T = 3.71 Ga] value of +4.4+/-1.2 (n = 8; 2sigma) is indicative of enrichment of their source region during, or prior to, its melting. Thus, this study provides the first observation of an early Archean upper mantle domain with a distinctly radiogenic Os isotopic signature. This requires a mixing component characterized by time-integrated suprachondritic Re/Os evolution and a Os concentration high enough to strongly affect the Os budget of the mantle source; modern sediments, recycled basaltic crust, or the outer core do not constitute suitable candidates. At this point, the nature of the mantle or crustal component responsible for the radiogenic Os isotopic signature is not known.