Preemptive interferon-α treatment could protect against relapse and improve long-term survival of ALL patients after allo-HSCT

Relapse was the major cause of treatment failure in patients with acute lymphoblastic leukemia (ALL) after allogeneic hematopoietic stem cell transplantation (allo-HSCT). We aimed to identify the efficacy and safety of preemptive interferon-α (IFN-α) treatment in ALL patients who had minimal residual disease (MRD) after allo-HSCT. Multiparameter flow cytometry and polymerase chain reaction assays were applied for MRD monitoring. Recombinant human IFN-α-2b injections were administered subcutaneously twice weekly in every 4 weeks cycle. Twenty-four (35.3%), 5 (7.4%), 6 (8.8%), and 13 (19.1%) patients achieved MRD negativity at 1, 2, 3, and > 3 months, respectively, after treatment. Seven patients showed grade ≥ 3 toxicities after IFN-α treatment. The 4-year cumulative incidence of total acute graft-versus-host disease (aGVHD), severe aGVHD, total chronic GVHD (cGVHD), and severe cGVHD after treatment was 14.7%, 2.9%, 40.0%, and 7.5%, respectively. The 4-year cumulative incidences of relapse and non-relapse mortality after treatment was 31.9% and 6.0%, respectively. The 4-year probabilities of disease-free survival and overall survival after IFN-α treatment were 62.1% and 71.1%, respectively. Thus, preemptive IFN-α treatment could protect against relapse and improve long-term survival for ALL patients who had MRD after allo-HSCT. The study was registered at https://clinicaltrials.gov as #NCT02185261 (09/07/2014).

www.nature.com/scientificreports/ and cytokine release syndrome) 14,15 , and remissions after CAR T-cell treatment was relatively brief because of poor CAR T cell persistence and/or leukemia cell resistance 16 .
Interferon-α (IFN-α) had shown activity in acute leukemia through immune activation 17 , which rekindled the interest of using IFN-α as an immunotherapy for patients receiving allo-HSCT 18 . Our pilot studies showed that IFN-α was a safe agent for allo-HSCT recipients 19 . We further confirmed that preemptive IFN-α treatment can clear the MRD effectively in patients with acute leukemia and high-risk myelodysplastic syndrome after allo-HSCT 4,[20][21][22] . IFN-α could also be used as a salvage treatment for patients with unsatisfactory response to preemptive Chemo-DLI 23 . However, the sample of ALL patients enrolled in these studies was relatively small, and no study had identified the efficacy of preemptive IFN-α treatment in a disease-specific population of patients with ALL after allo-HSCT. In addition, the follow-ups of these patients were short. Thus, the long-term efficacy of preemptive IFN-α treatment remains unknown in ALL patients following allo-HSCT.
Therefore, we aimed to identify the safety and long-term efficacy of preemptive IFN-α treatment in ALL patients following allo-HSCT.
Clinical outcomes of MRD-positive patients receiving preemptive Chemo-DLI. We also analyzed the data of patients who received preemptive Chemo-DLI during the same period (Supplementary Table 1 and Supplementary Fig. 1). The median cycle of Chemo-DLI was 1 (range 1-2 cycles), and 3 patients received Chemo-DLI for more than 1 cycle. The characteristics of patients in the Chemo-DLI group were summarized in Supplementary www.nature.com/scientificreports/ IFN-α treatment, a historical cohort between March 1, 2009 and May 31, 2013 including MRD-positive patients without any interventions was enrolled as controls (n = 18) 22 . T-ALL was more common in historical cohort and they had more cycles of chemotherapy before transplantation; however, the other patient characteristics were comparable between the present and the historical cohorts ( Table 1). The cumulative incidences of relapse and survival were worse in the historical cohort than those receiving preemptive IFN-α treatment in the present study ( Fig. 2A, C, D), and the cumulative incidence of NRM rates were comparable between the present and the historical cohorts (Fig. 2B). After adjusted by the MRD status, preemptive IFN-α treatment could also decrease the risk of relapse and improve survival (Supplementary table 6).

Clinical outcomes of patients receiving prolonged IFN-α treatment. Three patients received
IFN-α treatment for more than 6 cycles at the request of themselves, 2 received 7 cycles and 1 received 14 cycles Table 2. Characteristics of aGVHD after preemptive IFN-α treatment. Data was present as n (%) or median (range). aGVHD acute graft-versus-host disease, IFN-α interferon-α.

Discussion
In our study, the cumulative incidence of relapse and NRM at 4 years after preemptive IFN-α treatment were 31.9% and 6.0%, respectively; and the probability of DFS and OS at 4 years after preemptive IFN-α treatment were 62.1% and 71.1%, respectively. Our study is the first to study the efficacy of preemptive IFN-α treatment in a disease-specific population of patients with ALL. These results identify the undefined role of this intervention strategy in ALL patients following allo-HSCT. The graft-versus-leukemia (GVL) effect had been described in ALL since 1970s 24 , which was further supported by a large scale study recently 25 . IFN-α can exert an immunomodulatory effect, promote the GVL effect, and clear MRD after allo-HSCT 4,18,21 . Moreover, it also showed the growth-inhibitory or cytotoxic effects on human ALL cell in vitro [26][27][28] . Based on these results, IFN-α emerges as a useful agent which can clear MRD through different mechanisms. In fact, IFN-α had been used as adjuvant [29][30][31] or maintenance treatments in ALL patients 32,33 , which was reported to help to achieve CR again in ALL patients who experienced relapse after allo-HSCT 34,35 . Sumi et al. 33 also reported that IFN-α helped to achieve sustained molecular CR in an ALL patient with continuing detection of MRD following allo-HSCT. However, the evidences of IFN-α as a treatment option for ALL was generally derived from single case report or small sample studies, and its clinical utility in ALL has not been consistently established. In the present study, we observed that more than 70% of the patients achieved MRD negativity after preemptive IFN-α treatment. Up to now, this is the largest study confirming that IFN-α can indeed induce clinically relevant anti-leukemic responses in ALL patients.
MFC relying on the identification of cells with LAIPs and is widely believed to be sensitive for relapse prediction in ALL patients 3,36-38 . Thus, a patient was considered as MRD sin+ status when a single BM sample was tested www.nature.com/scientificreports/ positive by MFC for LAIPs in our study. In addition, the relapse rate of ALL patients who had one positive LAIPs result was 80.0% after allo-HSCT 39 . Approximately 60% of the patients used WT1 as an MRD marker in the present study. WT1 is still an important genetic marker for ALL patients [40][41][42][43] . In addition, Zhao et al. 39 reported that the relapse rate of ALL patients who had one positive WT1 result after allo-HSCT was 63.9%, and the sensitivity and specificity of WT1 was 62.2% and 90.6%, respectively, for indicating ALL relapse in allo-HSCT recipients. Thus, it is reasonable to use WT1 as the triggering marker for preemptive IFN-α treatment in ALL patients. However, some authors suggested that the sensitivity and specificity of WT1 monitoring might be relatively low 44,45 . Thus, MFC was used in the detection of MRD simultaneously, compensating for the relatively low sensitivity of WT1 expression. On the other hand, WT1 was not a specific molecular marker of leukemia. It is inevitable that some patients may receive IFN-α treatment because of high WT1 expressions which were actually not relevant to leukemia (i.e., receiving prophylactic IFN-α treatment), but Klingemann et al. 46 demonstrated that even prophylactic IFN-α treatment could also decrease the risk of relapse after allo-HSCT. In addition, only few severe toxicities were observed during IFN-α treatment, which might minimize the impact of the relatively low specificity of WT1 expression.
The 4-year CIR of patients who had MRD sin+ after preemptive IFN-α treatment was only 17.8% in the present study. Thirty-five patients with MRD sin+ were tested repeatedly 2 weeks after obtaining the first positive results. Among the 35 patients who showed MRD sin+ but did not receive IFN-α treatment, although immunosuppressions were tapered in 21 patients, only 1 patient achieved MRD negativity and the other 34 patients were tested positive for 2 consecutive BM samples (i.e., MRD co+ ). This indicated that MRD sin+ and MRD co+ might be different stages of the ALL progression. Reducing immunosuppression alone could not clear the MRD effectively and the preemptive IFN-α treatment for MRD sin+ patients could help to control the disease more timely. In addition, Zhao et al. 39 reported that patients with MRD co+ had higher relapse rate (WT1 + twice: 100%; MFC + twice: 87.5%; MFC + and WT1+: 100%) compared to that of MRD sin+ (WT1 + once: 63.9%; MFC + once: 80.0%), suggesting that MRD co+ may represent a higher risk of relapse compared with MRD sin+ . Our results also showed that the clinical outcomes seemed to be better in the MRD sin+ group than the MRD co+ group among patients receiving preemptive IFN-α treatment. Thus, preemptive IFN-α treatment may not completely overcome the poor prognostic significance of MRD co+ status of ALL, and patients with MRD sin+ may benefit more from preemptive IFN-α treatment after allo-HSCT.
We previously reported that preemptive Chemo-DLI could significantly decrease relapse and improve survival of patients with MRD 7 . In this study, the 4-year CIR, DFS, and OS rate of Chemo-DLI were 60.1%, 39.9%, and 67.4%, respectively. However, approximately 40% of our patients received preemptive IFN-α treatment for MRD sin+ and most of them could clear the MRD. These patients would not be classified as MRD co+ and they did not need to receive Chemo-DLI. Thus, it would be premature to derive conclusions regarding the superiority of IFN-α treatment over Chemo-DLI in patients with MRD.
A limitation to our current study was that it was not a randomized trial and the number of patients in historical control was relatively small. In addition, the ratio of PCR positive and MFC positive simultaneously seemed to be higher in non-IFN-α group although P value was 0.057, which meant that the risk of relapse may not be totally equivalent between IFN group and non-IFN-α group. In future, prospective, randomized trial may further confirm the efficacy of preemptive IFN-α in these patients. Secondly, the sensitivity of PCR for WT1 transcript and MFC for LAIPs was only 10 −3 -10 −4 . With a deep detection limit and high specificity, next-generation sequencing for MRD may represent a promising tool for the ALL patients 47 , and it may further improve the efficacy of preemptive IFN-α treatment. Thirdly, IFN-α can exert anti-leukemia effect through activating NK cells; however, we did not examine the number of NK cells in the present study, and we would identify the association between the number of NK cells and MRD negativity in our future study. Lastly, besides of the CAR-T therapy, several monoclonal antibodies (MoAbs) can also target certain surface antigens on ALL cells resulting in their destruction. However, the efficacy of these MoAbs in allo-HSCT recipients with MRD was unclear. Our future prospective studies can further compare the efficacy among MRD-directed preemptive Chemo-DLI, IFN-α treatment, MoAbs, and CAR-T therapy in ALL patients following allo-HSCT [48][49][50] .
In conclusion, preemptive IFN-α treatment could protect against relapse and improved long-term survival of ALL patients who had MRD after allo-HSCT. Because IFN-α may tend to be started in patients with relatively low leukemia burden 17 , it could not only unlock its therapeutic potential in ALL, but also spare the patients in remission from further therapy. Moreover, IFN-α is a simple treatment with increased accessibility as it could be performed on an outpatient basis. Based on our results, future randomized clinical trials are needed to further compare the efficacy of preemptive IFN-α treatment and cytotherapies in ALL patients who had MRD after allo-HSCT.

Patients and methods
Patients. From June 1, 2014 to December 31, 2017, consecutive Ph-negative ALL patients receiving allo-HSCT at the Peking University Institute of Hematology (PUIH) and showed MRD positivity were enrolled if they met the following criteria: (1) ALL defined as first or second complete remission (CR) without t(9;22) mutations (Supplementary Table 1 58 . When a single BM sample was tested positive by PCR or MFC, we considered this patient to have an MRD-positive status because the use of multiple methods could ensure sensitivity and specificity in the detection of the MRD 39 . Cases in which a single BM sample was tested positive by PCR or MFC were defined as the MRD sin+ group. Cases in which 2 consecutive BM samples within a 2-week interval were tested positive by PCR or MFC or those in which a single BM sample was tested positive by both PCR and MFC were defined as the MRD co+ group (Supplementary Fig. 1).
Patients in the MRD sin+ group were recommended to receive preemptive IFN-α treatment. For the patients with MRD co+ , the efficacy of Chemo-DLI had been confirmed 7 but the role of IFN-α treatment was undefined when this study started. Thus, preemptive Chemo-DLI was the first choice for patients with MRD co+ , and those who were unable to receive Chemo-DLI (e.g., patient or provider refusal) could receive IFN-α treatment ( Fig. 3 Preemptive IFN-α treatment and Chemo-DLI protocol. The detailed protocols for preemptive IFN-α treatment and Chemo-DLI was according to the routine protocols of PUIH which had been described in detailed (Supplementary methods) 4,7,21,22 . In brief, recombinant human IFN-α-2b injections (Anferon; Tianjin Hualida Biotechnology Co., Ltd., Tianjin, China) were administered subcutaneously for 6 cycles (twice weekly in every 4 weeks cycle), at dosages of 3 million units for patients older than 16 years and at 3 million units per square meter for those younger than 16 years (capped by 3 million units). Prolonged treatment with IFN-α was permitted at the request of patients. MRD status was monitored 1, 2, 3, 4.5, 6, 9, and 12 months after preemptive IFN-α treatment and at 6-month intervals thereafter. The patients who had persistent and increasing levels of MRD or those regained MRD positivity after achieving MRD negativity after IFN-α treatment could receive salvage Chemo-DLI (Supplementary methods) 7 .
Treatment of GVHD after preemptive immunotherapy. The treatments of acute GVHD (aGVHD) and chronic GVHD (cGVHD) were according to accepted international criteria (Supplementary methods) 59-61 . Definitions and assessments. Disease risk index (DRI) before allo-HSCT was described according to the criteria of Armand et al. (i.e., ALL patients in CR1 were categorized into intermediate risk, and ALL patients in CR2 were categorized into high risk group) 62 . GVHD was diagnosed according to accepted international criteria 63,64 . Definition of relapse, non-relapse mortality (NRM), early-onset MRD (EMRD), late-onset MRD (LMRD), high-level MRD, and low-level MRD were described in supplementary method 4 . www.nature.com/scientificreports/ Statistical analysis. The primary endpoint was relapse, and secondary endpoints were NRM, disease-free survival (DFS), and overall survival (OS). This study was planned to detect a relapse rate of 55% in patients with MRD receiving preemptive IFN-α treatment, from the reference rate of 75% in patients with PCR or MFC positivity but did not receive interventions in our previous study, controlling for type I and II error rates at 5% and 10%, respectively. Considering an expulsion rate of 15%, a total of 68 patients was planned to be enrolled. Comparisons of patient characteristics between the groups were performed using the Mann-Whitney U-test for continuous variable and χ 2 and Fisher's exact tests for categorical data. The probability of survival was calculated using the Kaplan-Meier estimator. The incidences of GVHD were calculated using the cumulative incidence function, with death and relapse as competing risks 65 . Cumulative incidences were estimated for NRM and relapse, to account for competing risks. Relapse was the competing event for NRM and vice versa. Hazard ratios (HRs) for clinical outcomes were estimated from Cox regression analyses. P values were 2-sided. The SPSS 24 (SPSS Inc./IBM, Armonk, NY, USA) and the R software package (version 2.6.1; https ://www.r-proje ct.org) were used for data analyses.