• The treatment of SGF after allogeneic stem cell transplantation has always been a difficult problem, fundamentally because the basic mechanism of SGF is not clear. SGF has been mainly studied in animals, and there are few relevant clinical studies. • We preliminarily explored the basic mechanism and clinical solutions of SGF. We found that the mechanism of SGF is closely related to residual host cellular and humoral immunity, mixed chimerism (MC), Foxp3 and regulatory T cells (Tregs). • We found that DLI from the second donor can treat SGF; it can even correct SGF, depending on the condition of the first donor, as we report for the first time in a clinical study. • Patients with SGF receiving DLI from the second donor achieved complete response, and MC gradually converted to CC; simultaneously, there were significant increases in Foxp3mRNA and the Treg fraction. Therefore the mechanism may be closely related to the conversion of MC and increasing Foxp3mRNA and Tregs, which is worthy of further study in randomized clinical trials. • It is hoped that interest in treatment for SGF can be stimulated from this clinical study. Secondary graft failure (SGF) is a fatal complication of allogeneic hematopoietic stem cell transplantation without effective treatment methods, especially after haploidentical transplantation. This study aimed to analyze the efficacy of donor lymphocyte infusion (DLI) from a second donor in treating SGF and the underlying immune mechanisms. A second donor is a candidate donor who did not initially provide stem cells for HLA-matched sibling donor or HLA-haploidentical donor transplantation. We conducted a retrospective study of 237 patients with a median age of 38 years (range 9–56) for whom the degree of mixed chimerism (MC) and complete donor chimerism (CC), mRNA expression levels of Forkhead box P3 (Foxp3), and the proportion of regulatory T cells (Tregs) were regularly assessed. The median time to SGF was 62 days (range 41–117) after transplantation. Twenty-one patients with SGF received DLI, including 12 patients who initially received DLI from a second donor (i.e. , a donor other than the transplantation [first] donor) and 9 patients who initially received DLI from the first donor but showed no response. Three of those 9 patients subsequently received DLI from a second donor. The incidence of acute GVHD and chronic GVHD induced by DLI from the second donor was significantly higher than that of DLI from the first donor (P = 0.006). Twenty-one patients with SGF exhibited synchronous MC, and the overall MC rate after transplantation was 65% (range 42%–85%).The proportion of Tregs significantly decreased in SGF patients, from a median of 2.61% ± 0.88% to 0.92% ± 0.23% at the indicated time point after transplantation (P = 0.03). Second-donor DLI resulted in a complete response (CR) in 13 patients, and MC gradually converted into CC; simultaneously, there was a significant increase in the mRNA level of Foxp3 and the proportion of Tregs (baseline, 0.92% ± 0.23% versus CR, 3.61% ± 0.82%; P = 0.01). For the patients who did not respond to DLI from either donor type, there was no significant change in donor chimerism, Foxp3 expression level or Treg proportion. Overall survival and disease-free survival 2 years after DLI were 66.7% ± 3.08% and 59.8% ± 4.11%, respectively. DLI from a second donor may be an effective treatment for SGF, and the mechanism is related to MC-to-CC conversion and activation of Foxp3 and Tregs. [ABSTRACT FROM AUTHOR]