A total of 30 newly diagnosed B cell ALL and 10 normal donors were enrolled for gene expression microarray. Patient characteristics are shown in Supplementary Table 1. There were 19 (63%) males, and the median age was 33 years old. Most of them were common-B ALL (N = 22, 73%), and the most common cytogenetics abnormality is t (9;22)(q34;q11). Correspondingly, BCR::ABL1 fusion accounted for nearly half of the patients (N = 14, 47%). IKZF1 deletion (N = 14, 47%) often concurred with BCR::ABL1, and 8 ones both harbored BCR::ABL1 fusion and IKZF1 deletion (double-positive). There were 10 double-negative cases, while IKZF1-positive and BCR::ABL1-positive were 6 cases each. Five cases had KMT2A-r fusion, one had TCF3::PBX1 fusion and none of them had ETV6::RUNX1 fusion. As for donors, there were 6 males and the median age was 43 years old.

We firstly subdivided B cell ALL patients into 4 subgroups according to the BCR::ABL1 and IKZF1 status to unveiled the dysregulated genes in different background. As shown in Supplementary Fig. 1 and Supplementary Table 2, the most significant difference was seen in double-positive and double-negative group. In the double-negative group, the up-regulated genes were 538 more than that in the double-positive group, but the down-regulated genes were significantly reduced, reaching 1744.

We found that the two transcripts (NM_004425, NM_022664) of ECM1 were both highly transcribed in the four subgroups (Supplementary Table 3), indicating that the abnormal expression of ECM1 may have universal significance in B cell ALL. To elucidate the possibility of ECM1 as a potential molecular marker in B cell ALL, we further verified the transcription level of ECM1 in samples from newly diagnosed (N = 267), complete remission (CR, N = 41) and relapsed (N = 42) B cell ALL, as well as 22 healthy donors. Basic clinical information of 267 newly diagnosed patients was shown in Supplementary Table 4. Particularly, most of the patients (97%, 259/267) were adults (age ≥ 14 years), and only 8 patients aged < 14 years. Our data showed that the highest level of ECM1 transcription was found in newly diagnosed B cell ALL patients, with a median of 124.57% (0.79–3519.50%), significantly higher than in CR patients (median, 3.95% [1.50–11.22%], P < 0.001) and donors (median, 7.14% [1.54–10.27%], P < 0.001). Transcription level elevated again in relapsed patients, with a median of 107.62% (1.19–1142.81%), and was significantly higher than in CR patients (P < 0.001) and donors (P < 0.001) (Fig. 1A). We also detected ECM1 transcription level in AML, including 15 newly diagnosed and 13 CR samples. The median transcription level of newly diagnosed AML was 6.88% (1.50–35.76%), and the median in CR was 3.21% (1.43–5.36%). There was no significant difference in ECM1 transcription levels between newly diagnosed AML and normal donors (P = 0.551). ROC analysis of ECM1 transcription level in newly diagnosed B cell ALL patients showed that the area under the curve was 0.89 (95% confidence interval, 0.85–0.93; P < 0.001) (Fig. 1B). The maximum Youden index is 0.846, corresponding to an ECM1 transcription level of 10.28%. Taking 10.28% as the diagnostic threshold, the sensitivity and specificity for diagnosis was 84.6 and 100%, respectively. We further compared the diagnostic performance of ECM1 with WT1 in B cell ALL. In 265 patients who were tested for both ECM1 and WT1, the overall false negative rate of ECM1 in diagnosing B cell ALL was 15.1%, while the false negative rate of WT1 was 34.3%. The overall false negative rate of the two genes was 4.2% (Supplementary Table 5). Therefore, we suggested that ECM1 is more sensitive than WT1 in diagnosing B cell ALL, and these two genes could complement each other.

Up-regulated transcription level of ECM1 in newly diagnosed B cell ALL. A Transcription level of ECM1 in B cell ALL (D newly diagnosed, CR complete remission, R relapse), normal donors, and AML (D newly diagnosed, CR complete remission). The original transcription level (ECM1/ABL1) was multiplied by 1000 times and then logarithmically conversed. B The ROC curve of ECM1 transcription levels for B cell ALL diagnosis

We also compared the transcription levels of ECM1 among different subgroups (Fig. 2A–B). Among them, patients with KMT2A-r, TCF3::PBX1, and ETV6::RUNX1 were further grouped independently. Multiple comparison showed that there was no significant difference in transcription levels among the double-negative (median, 93.50% [0.83 – 3519.50%]), IKZF1-positive (median, 97.77% [0.79 – 1397.76%]), and the BCR::ABL1-positive subgroups (median, 141.22% [11.33 – 754.35%]) (both P > 0.05). Double-positive group showed the highest transcription level, with a median level of 210.78% (45.30 – 714.05%). Patients with KMT2A-r and TCF3::PBX1 had the lowest transcription levels, with a median of 39.48% (6.28 – 165.32%) and 30.02% (1.78 – 82.54%), respectively. However, the transcription level of patients with ETV6::RUNX1 was not low (median, 104.87% [50.84 – 424.05%]). The transcription level in the double-positive ones was significantly higher than that in the double-negative (P = 0.001), the IKZF1-positive (P = 0.005), KMT2A-r (P < 0.001), and TCF3::PBX1 ones (P < 0.001). As for immunophenotypes, Common-B ALL had the highest transcription level (median, 151.53% [0.79–3519.50%]), significantly higher than that in the Pre-B ALL (median, 12.45% [1.54–1373.54%]; P < 0.001). Transcription level in Pro-B ALL was also not high (median, 66.48% [0.85–1210.48%]; P = 0.053). We further investigated the diagnostic performance of ECM1 in patients except for KMT2A-r, TCF3::PBX1, and ETV6::RUNX1 rearrangements (Fig. 2C–F). In the double-negative group, the area under the curve was 0.82 (0.74–0.89; P < 0.001), slightly lower than that in the IKZF1-positive group (0.84 [0.76–0.93]; P < 0.001). However, in BCR::ABL1-positive and double-positive groups, the area under the curve was 1 (P < 0.001).

Transcription levels of ECM1 in different B cell ALL subgroups. A Transcription levels of ECM1 in different genetic backgrounds. B Transcription levels of ECM1 in different immunophenotypes. The original transcription level (ECM1/ABL1) was multiplied by 1000 times and then logarithmically conversed. C–F The ROC curve of ECM1 transcription level in the double-negative, IKZF1-positive, BCR:ABL1-positive, and double-positive groups, respectively. **: P < 0.01, ***: P < 0.001

We further evaluate the value of ECM1 transcription level for MRD monitoring. In 20 consecutive follow-up cases, it was found that ECM1 transcription level down-regulated after achieving CR, but up-regulated at the time of relapse (Fig. 3A). We further monitored MRD of 5 long-term follow-up cases and compared ECM1 detection with other MRD monitoring techniques (morphology, flow cytometry, other genes in qRT-PCR) (Fig. 3B–F). All patients showed high ECM1 transcription level at diagnosis. Patients with long-term CR also maintained low transcription level of ECM1 (patients 1, 4, and 5), while refractory or relapse patients maintained high transcription level (patient 3). Relapsed patients after transplantation also showed elevated ECM1 transcription level (patients 1, 5).

MRD monitored by different indicators. A The correlation between ECM1 transcription level and clinical course in consecutive 20 patients. The original transcription level (ECM1/ABL1) was multiplied by 1000 times and then logarithmically conversed. B–F Transcription level of ECM1 and other clinical MRD monitoring indicators in 5 long-term follow-up patients. Samples at diagnosis, relapse, molecular relapse, and extramedullary relapse were marked, while the remaining unlabeled points indicated complete remission. The actual MRD levels of each indicator were shown in patients 1–4. Due to the significant difference in transcription levels between ECM1 and WT1 in patient 5, ECM1 was multiplied by 1000 times and then logarithmically conversed, while WT1 was multiplied by 10,000 times and then logarithmically conversed, corresponding to the left axis. The blue dashed line represented to the WT1 threshold (0.60%). Bone marrow morphology and flow cytometry were showed as the actual MRD levels, corresponding to the right axis. D diagnosis, CR complete remission, R relapse, T transplantation, CNS central nerve system, BM bone marrow morphology, FC flow cytometry

As for morphology, the ECM1 transcription level was basically consistent with the blast at most bone marrow aspirations. In some patients, ECM1 transcription level detection was superior to morphology. For patient 2, after induction (the 2nd bone marrow aspiration), the morphology showed CR, but MFC and genes (BCR::ABL1, IKZF1) did not achieve, as well as ECM1 transcription level. Patient 4 relapsed during the 4th and 5th aspirations, and the sensitivity of ECM1 was superior to morphology. Molecular relapse occurred during the 12th and 13th aspirations, with an increase in ECM1 transcription level, but morphological relapse was at the 14th aspiration. Patients 1, 2, 3, and 5 provided follow-up MFC information. In patients with MFC-MRD, the transcription level of ECM1 was highly consistent with the MFC, especially in the relapsed samples. In terms of other genes, the transcription level of BCR::ABL1 (patients 2 and 4) and KMT2A-r (patient 3) showed good parallelism with ECM1 in disease course monitoring. Patient 2 also had IKZF1 deletion. The dynamics of IKZF1 deletion level and ECM1 transcription level was basically consistent. Patients 1, 3, and 5 provided WT1 follow-up information. The dynamics of WT1 transcription level was basically consistent with ECM1. Like other intramedullary MRD indicators, due to the fact that two patients with extramedullary recurrence (patients 1 and 5) did not suffer intramedullary relapse simultaneously, the transcription level of ECM1 did not indicate the status of extramedullary relapse well.

We divided 177 adult (age ≥ 14 years) B cell ALL patients with prognostic information into a high transcription level group (N = 86) and a low transcription level group (N = 91) based on the median transcription level of ECM1 (124.57%) (Table 1). In the high transcription level group, patients presented high-risk clinical features, such as older age (high transcription level group vs. low transcription level group, 35 y vs. 29 y, P = 0.008), high positive rate of MFC-MRD (≥ 0.01%) after induction (7 patients without MFC-MRD information) (64% vs. 43%, P = 0.005), high positive rate of MFC-MRD after 1 course of consolidation chemotherapy (18 patients without MFC-MRD information) (46% vs. 27%, P = 0.010) and after 2 courses of consolidation chemotherapy (33 patients without MFC-MRD information) (43% vs. 22%, P = 0.008). In terms of immunophenotype, patients with high transcription level were more likely Common-B ALL (P = 0.001). This result was consistent with the molecular genetics classification, because KMT2A-r is more common in the Pro-B ALL, while TCF3::PBX1 rearrangement is more common in the Pre-B ALL. The distribution of BCR::ABL1 and IKZF1 between the high and low transcription level groups was consistent with the above results. Patients with high transcription level were more likely to have BCR::ABL1 and IKZF1 deletion mutations (P < 0.05). There was no significant difference between the two groups in terms of gender, white blood cell count, hemoglobin concentration, platelet count, ETV6::RUNX1 rearrangement, and treatment regimen (all P > 0.05).

The specific decreased ECM1 transcription level in KMT2A-r (N = 8) and TCF3::PBX1 (N = 9) patients suggested that the mechanism of ECM1 in these two subtypes may be different from that of non-KMT2A-r and non-TCF3::PBX1 patients. So, subsequent prognostic analysis did not include these patients. In the case of transplantation as censored event, the 5-year OS of patients with high ECM1 transcription level was significantly lower than patients with low transcription level (low vs. high, 72.9% vs. 18.7%, P < 0.001) (Fig. 4A). When transplantation was not treated as censored event, high transcription level of ECM1 still had a worse 5-year OS (low vs. high, 71.1% vs. 56.8%, P = 0.038) (Fig. 4B). Six factors with P < 0.1, including ECM1 transcription level (high vs. low), age (≥ vs. < 35 years old), white blood cell (WBC, ≥ vs. < 30 × 109/L), BCR::ABL1 (positive vs. negative), IKZF1 deletion (positive vs. negative), and MFC-MRD after induction (positive vs. negative), were enrolled in the multivariate analysis (transplantation as censored event). Only high ECM1 transcription level, high WBC count, and positive MFC-MRD after induction were independent risk factors for OS (Supplementary Table 6). When transplantation was not considered as censored event, a total of four factors, including ECM1 transcription level, age, BCR::ABL1, and transplantation (yes vs. no), were enrolled in the multivariate analysis. Only BCR::ABL1 was an independent risk factor for OS, while receiving transplantation was an independent protective factor (Supplementary Table7).

Prognostic analysis of ECM1 in B cell ALL patients. A OS analysis with transplantation as censored event. B OS analysis with transplantation not as censored event

We further validated the clinical significance of ECM1 in patients by a B cell ALL dataset GSE34861 from the GEO database, with a total of 191 patients. Samples in this database were obtained from bone marrow and peripheral blood. For greater comparability with our samples, we only enrolled 91 patients with samples obtained from bone marrow. The median transcription level of ECM1 in these patients was 14.58 (9.22–15.57). According to the median transcription level, there were 45 cases in the low transcription level group and 46 in the high. The comparison of clinical information between the two groups was basically consistent with our patients (Supplementary Table 8). Pro-B and Pre-B were still mainly seen in the low transcription level group. BCR::ABL1 mostly occurred in the high transcription level group, while KMT2A-r and TCF3::PBX1 were still only found in the low transcription level group. A total of 74 patients were enrolled in the OS analysis (1 patient without follow-up information and 16 patients with KMT2A-r or TCF3::PBX1 were excluded). Transplantation was not treated as censored event due to the lack of transplantation information in the dataset. Similar to Fig. 4B, patients with high ECM1 transcription level still had a poor prognosis (37.4% vs. 10.6%, P = 0.059) (Supplementary Fig. 2).

We further explored the potential mechanism by which high ECM1 transcription levels affected the prognosis of B cell ALL. KEGG enrichment analysis of differentially expressed genes was performed in B cell ALL patients with high ECM1 transcription level. We found 3 pathways with the highest enrichment score were involved in the regulation of cell migration, including leukocyte transendothelial migration, adherens junction and regulation of actin cytoskeleton (Fig. 5). These results suggested that ECM1 may affect the invasion and metastasis of tumor cells. In the leukocyte transendothelial migration pathway, the genes that contribute primarily to enrichment score were CD99, ACTN1, and CTNND1 (Supplementary Table 9).

KEGG analysis of B cell ALL patients with high ECM1 transcription level. A Leukocyte transendothelial migration, B Adherens junction, C Regulation of actin cytoskeleton