Introduction

Acute myeloid leukemia (AML) is a complex and dynamic disease. The malignant myeloid cells are composed of coexisting competing clones and the disease evolves over time [1]. People have discovered and used some molecular biomarkers, including genetic mutations, to help decipher this heterogenic and often deadly disease, to predict clinical outcome and guide treatment [2]. For example, DNMT3A and FLT3-ITD mutations are independent poor prognostic factors [3, 4], whereas the biallelic CEBPA mutation is associated with good prognosis [5, 6]. With the improvement molecular diagnostic technology, not only the mutations but also the aberrant expression levels of some genes could be integrated into the refined risk stratification of AML. The overexpressions of MN1, ERG, BAALC, EVI1, DOK4/5, PDK2/3, FHL2, and iASPP have been associated with poor prognosis, whereas high DOK7 expression is associated with good prognosis in AML [7,8,9,10].

The genes encoding the interferon (IFN)-induced transmembrane proteins (IFITMs) belong to the IFN-stimulated genes. These proteins are powerful suppressor of viral infections. Human IFITM genes are located on chromosome 11 and translates into four highly homologous membrane surface proteins IFITM1, IFITM2, IFITM3, and IFITM5, whereas IFITM4P is a fake gene [11]. At present, the functions and related mechanisms of IFITM1 and IFITM3 as tumor-promoting genes, if not oncogenes per se, have been reported in various solid tumors. For example, high expression of IFITM1 promotes the proliferation, invasion, and distant metastasis of squamous cell carcinoma of the head and neck [12], and also predicts adverse outcome of esophageal cancer [13]. In breast cancer tissue, the expression of IFITM3 is significantly higher than adjacent tissues and is closely related to the estrogen and progesterone receptors. Knocking down IFITM3 suppresses breast cancer cell growth and colony formation, and affects the cell cycle [14]. IFITM3 is also abnormally overexpressed in colon cancer, especially in patients with positive lymph node metastasis. It is an independent risk factor for disease-free survival (DFS) in colon cancer [15]. IFITM5 is only expressed in osteoblasts [16]. Overexpression of IFITM5 promotes osteosarcoma cell apoptosis, inhibits invasion, and promotes osteogenic differentiation [17]. Study on IFITM2 is lacking but there has been one study showing that it is significantly upregulated in intestinal cancer and has a p53-independent role in promoting apoptosis [18].

The prognostic significance of the IFITM family in AML has not been reported. In this study, we aimed to investigate the effects of IFITM on AML survival. Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is an effective treatment for AML, which can reduce recurrence and prolong the survival by significantly reducing the leukemia residual disease [19]. Herein, we also analyzed whether allo-HSCT could overcome the prognostic effects of the IFITM family.

Materials and methods

Patients

From The Cancer Genome Atlas database (https://cancergenome.nih.gov/), a total of 155 AML patients with IFITM family (IFITM1–5) expression data were included in this study [20]. Eighty-four patients received chemotherapy only and 71 also underwent allo-HSCT. Clinical characteristics at diagnosis, including peripheral white blood cell (WBC) counts, blast percentages in peripheral blood (PB) and bone marrow (BM), French–American–British (FAB) subtypes, cytogenetic risk group, and frequencies of common recurrent genetic mutations, were downloaded from the database. Event-free survival (EFS) and overall survival (OS) were the primary endpoints of the study. EFS was defined as the time from diagnosis to removal from the study due to relapse, death, or failure to achieve complete remission, or was censored at the last follow-up. OS was defined as the time from diagnosis to death from any cause or was censored at the last follow-up. Informed consents were obtained from all patients and the study protocol was approved by the Human Research Council of the University of Washington.

Statistical analysis

The clinical and molecular characteristics of patients were summarized using descriptive statistics. Data sets were described with median and/or range. Survival was estimated using the Kaplan–Meier method and the log-rank test. Numerical data were compared using the Mann–Whitney U-test and categorical data were compared using the χ2-test. Multivariate Cox proportional hazard models were constructed for EFS and OS using a limited backward elimination procedure. The confidence interval was 95%. All statistical analyses were performed by SPSS software 20.0 and GraphPad Prism software 7.0.

Results

Clinical and molecular characteristics of the patients

The clinical and molecular characteristics of all patients were shown in Table 2. Median age was 63 years (range 22–88), with 58 cases over 60 years old. Forty-five patients were men. The median WBC, BM blast, and PB blast count were 38.3 × 109/L, 67.5, and 36.3%, respectively. The major FAB subtypes were M1, M2, and M4 (72.6%). Forty-four patients had abnormal karyotypes. The proportion of good, intermediate, and poor-risk AML were 14.3%, 54.8%, and 28.6%, respectively. NPM1 had the highest mutation frequency (n = 27, 32.1%), followed by DNMT3A (n = 23, 27.4%), FLT3 (n = 22, 26.2%), IDH1/2 (n = 15, 17.9%), NRAS/KRAS (n = 12, 14.3%), TP53 (n = 12, 14.3%), TET2 (n = 11, 13.1%), and RUNX1 (n = 8, 9.5%).

Prognostic significance of IFITM family in AML

To evaluate the prognostic significance of the IFITM family in AML, all patients were divided into high- and low-expression subgroups by the median expression levels of each IFITM member (IFITM1/2/3/5). EFS and OS of the expression subgroups of each gene were analyzed with the Kaplan–Meier method and the log-rank test (Table 1). In the chemotherapy-only group, high IFITM3 expression had adverse effects on EFS and OS (P = 0.018 and P = 0.010, Fig. 1a, b). None of the IFITM members had impact on survival in the allo-HSCT group.

Table 1 Comparison of EFS and OS between different expression levels of IFITM1–5
Full size table
Fig. 1
figure 1

Kaplan–Meier curves of event-free survival (EFS) and overall survival (OS) in patients who received chemotherapy only. a, b High IFITM3 expressers had shorter EFS and OS than the low expressers

Full size image

Association of IFITM3 expression with other clinical and molecular characteristics in the chemotherapy-only group

The clinical and molecular characteristics of high and low IFITM3 expression subgroups were compared (Table 2). IFITM3high had more age ≥ 60 patients (P = 0.018), more FAB-M0 (P = 0.006), and fewer FAB-M5 (P = 0.002) patients, fewer normal karyotype patients, and more complex karyotype (all P < 0.001). No significant differences were found in gender distribution, peripheral WBC count, BM blasts, PB blasts, risk-group distribution, and frequency of common genetic mutations (FLT3, NPM1, DNMT3A, RUNX1, TET2, TP53, IDH1/IDH2, and NRAS/KRAS) between the two groups.

Table 2 Comparison of clinical and molecular characteristics in different groups
Full size table

Multivariate analysis of EFS and OS in the chemotherapy-only group

To further evaluate prognostic value of IFITM3, multivariate Cox proportional hazard models were constructed, selecting the expression levels of IFITM3 (high vs. low), age (≥ 60 vs. < 60 years), peripheral WBC count (≥ 15 × 109/L vs. < 15 × 109/L), BM blasts (≥ 70% vs. < 70%), PB blasts (≥ 70 vs. < 70%), FLT3-ITD (positive vs. negative), and common AML mutations (NPM1, DNMT3A, CEBPA, RUNX1, IDH1/IDH2, and NRAS/KRAS, mutated vs. wild type). Results were shown in Table 3.

Table 3 Multivariate analysis of EFS and OS
Full size table

Multivariate analysis showed that high IFITM3 expression and age ≥ 60 years were independent risk factors for both EFS and OS (all P < 0.05). Besides, BM blasts ≥ 70%, PB blasts ≥ 70%, and DNMT3A mutation were independent risk factors for EFS (all P < 0.05) and RUNX1 mutation was an independent risk factor for OS (P < 0.05)

Discussion

In this retrospective study, we found that high IFITM3 expression was an adverse prognostic factor for AML, but not in those who underwent allo-HSCT, implying that allo-HSCT might be able to overcome its prognostic impact.

Increasing number of studies have shown that IFITM3 participates in the development and progression of various tumors and is involved in myriads of cell biology processes, including cancer cell proliferation, invasion and metastasis, apoptosis, and the epithelial-to-mesenchymal transition (EMT). A study indicated that downregulating IFITM3 in U251 cells could inhibit cell proliferation and cloning, arrest the cell cycle in the G0/G1 phase, especially in the pre-G1 phase that could lead to apoptosis. In addition, the cell migration was also significantly suppressed after downregulation of IFITM3 [21]. In gastric cancer, high IFITM3 expression was found to promote tumor cell migration, invasion, and proliferation by activating Wnt/β-catenin signaling pathway. Another study revealed that IFITM3 silencing would effectively reverse the EMT phenotype and reduce MMP-2 and MMP-9 expression [22]. Overexpression of IFITM3 may also predict poor prognosis in stage IIA esophageal squamous cell carcinoma patients after Ivor Lewis esophagectomy [23]. Consistent with these findings, our study pointed out that IFITM3 might also be a tumor-promoting gene or oncogene in AML. Its overexpression coincided with other established poor prognostic factors, such as older age and complex karyotype, although its effect was independent.

Out results concurred with previous studies that age ≥ 60 years had unfavorable effects on AML survival, probably due to the higher mutation burden, poorer baseline performance status, and more co-morbidities in this age group [24]. We identified that BM blasts ≥ 70% and PB blasts ≥ 70% also were independent risk factors for EFS, consistent with a former finding that abnormal proliferation of BM blasts and PB blasts had significant negative effects on survival in AML [25]. In our study, DNMT3A mutation was an independent risk factor for EFS and RUNX1 mutation was an independent risk factor for OS, which was in line with other reports that DNMT3A mutation was associated with inferior DFS and a trend toward shorter OS in cytogenetically normal AML [26], and RUNX1 mutation being a strong independent predictor for inferior OS in complex karyotype AML [27].

In conclusion, high IFITM3 expression was associated with poor prognosis in AML, but its effects on survival could be overcome by allo-HSCT. Due to the small sample size, larger prospective researches are needed to further validate the role of IFITM3 as an independent poor prognostic factor for AML. In addition, precise experiments need to be designed to explain the mechanisms of IFITMs in tumorigenesis.