Prions, characterized by self-propagating protease-resistant prion protein (PrP) conformations, are agents causing prion disease. Recent studies generated several such self-propagating protease-resistant recombinant PrP (rPrP-res) conformers. While some cause prion disease, others fail to induce any pathology. Here we showed that although distinctly different, the pathogenic and non-pathogenic rPrP-res conformers were similarly recognized by a group of conformational antibodies against prions and shared a similar guanidine hydrochloride denaturation profile, suggesting a similar overall architecture. Interestingly, two independently generated non-pathogenic rPrP-res were almost identical, indicating that the particular rPrP-res resulted from cofactor-guided PrP misfolding, rather than stochastic PrP aggregation. Consistent with the notion that cofactors influence rPrP-res conformation, the propagation of all rPrP-res formed with phosphatidylglycerol/RNA was cofactor-dependent, which is different from rPrP-res generated with a single cofactor, phosphatidylethanolamine. Unexpectedly, despite the dramatic difference in disease-causing capability, RT-QuIC assays detected large increases in seeding activity in both pathogenic and non-pathogenic rPrP-res inoculated mice, indicating that the non-pathogenic rPrP-res is not completely inert in vivo. Together, our study supported a role of cofactors in guiding PrP misfolding, indicated that relatively small structural features determine rPrP-res' pathogenicity, and revealed that the in vivo seeding ability of rPrP-res does not necessarily result in pathogenicity.