MIR4435-2HG as a Novel Predictive Biomarker of Chemotherapy Response and Death in Pediatric B-Cell All

doi:10.21203/rs.3.rs-2857252/v1

Download PDF

Article

MIR4435-2HG as a Novel Predictive Biomarker of Chemotherapy Response and Death in Pediatric B-Cell All

https://doi.org/10.21203/rs.3.rs-2857252/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Although B-cell acute lymphoblastic leukemia (B-cell ALL) survival rates have improved in recent years, Hispanic children continue to have poorer survival rates. Our aim was to identify biomarkers of treatment response, which may also predict relapse and death, through identifying differentially expressed and methylated genes between patients who responded or did not respond to induction treatment. DNA methylation and mRNA sequencing assays were performed on 27 bone marrows from Hispanic children with B-cell ALL. Gene expression and differential methylation were compared between responders and non-responders at day 15 and at the end of induction chemotherapy. DAPK1, CNKSR3, MIR4435-HG2, CTHRC1, NPDC1, SLC45A3, ITGA6, and ASCL2 were overexpressed and hypomethylated in non-responders. The overexpression of DAPK1, ASCL2, SCL45A3, NPDC1 and ITGA6 can predict non-response at day 15 and refractoriness. Additionally, higher expression of MIR4435-2HG increases the probability of non-response, death, and the risk of death. MIR4435-2HG is also overexpressed in relapse samples. Finally, MIR4435-2HG overexpression, together with positive minimal residual disease, are associated with poorer survival, and together with overexpression of DAPK1 and ASCL2, it could improve the risk classification of patients with normal karyotype. MIR4435-2HG is a potential predictive biomarker in children with B-cell ALL.

Biological sciences/Cancer/Haematological cancer/Leukaemia/Acute lymphocytic leukaemia

Biological sciences/Cancer/Paediatric cancer

Biological sciences/Molecular biology/Epigenetics/Dna methylation

Biological sciences/Molecular biology/Transcriptomics

Health sciences/Biomarkers/Predictive markers

B-cell acute lymphoid leukemias (B-cell ALL) are the most frequent neoplasms in children [1]. Cure rates for ALL have improved remarkably in the last four decades; nevertheless, while developed countries achieve 80%, cure rates in developing countries are around 50% [2]. The mechanisms underlying these differences in survival rates are still unknown; however, several studies by the Children's Oncology Group (COG) show that Hispanic patients have worse survival and treatment response compared to White and Asian children, even using the same treatment protocols [3].

Currently, clinical parameters such as leukocyte count, age, extramedullary infiltration, chromosomal translocations, and minimal/measurable residual disease (MRD) classify patients into risk groups. MRD is the most used variable to define treatment response [4]. However, due to low survival in our patients, it is possible to propose that those variables do not fully define risk groups, which leads to incorrect selection of chemotherapy protocol, affecting patient survival.

In ALL, gene expression alterations not only results from mutations; alterations at the epigenetic level also play a relevant role in this pathology [5]. Thus, epigenetic alterations, including aberrant DNA methylation, could act as important molecular mechanisms in developing resistance to treatment of ALL [5]. In bone marrow (BM), DNA methylation patterns change during normal hematopoiesis and are believed to play an essential role in lineage differentiation [6]. As in normal cells, tumor cells may also depend on specific DNA methylation patterns to acquire their phenotype and maturation patterns [7]. Therefore, the characterization of aberrant patterns in DNA methylation in tumors can provide important clues about how gene expression is regulated in these pathologies [7]. Hogan et al [8], found that patients with relapses presented promoter hypermethylation and identified a clear signature of differentially expressed genes at the time of diagnosis and relapse; moreover, this signature differs between early-relapse patients compared to late-relapse patients. Although differential methylation and gene expression patterns have been observed between samples at diagnosis and in relapse, it should be determined whether these variables could be tools to predict treatment response, including relapse or death [9]. Also, a CpG island methylation analysis identified candidate genes as biomarkers of pediatric ALL subgroups and their correlation with disease prognosis [10]. Identifying genomic markers, derived from methylation and gene expression analysis, could improve risk classification and define patient prognosis.

Here we use RNA-seq technology and DNA methylation to identify potential predictive biomarkers of response to chemotherapy and prognosis of relapse and death in pediatric patients with B-ALL. To do this, we sequenced BM samples obtained at baseline and compared gene expression and DNA methylation profiles between responder and non-responder patients at 2-time points during induction chemotherapy, searching for overexpressed genes with hypomethylation status, verifying their expression using RT-qPCR, and investigating whether the selected genes could predict response to induction chemotherapy, relapse or death.

Table 1 describes the clinicopathological characteristics of patients. In the descriptive cohort, the average age was 10 years, and the disease was more frequent in males (57%). Sixty percent of patients had normal karyotype, and t(1;19) was the most frequent genetic alteration (14.3%). The most frequent risk group was intermediate (71.4%); extramedullary infiltration was 11%. More than 90% of patients responded to corticosteroid therapy. MRD was detected by FCM at day 15 and the end of induction; 20 patients did not respond at day 15 (71%), while only 9 patients did not respond (32%) at the end of induction. MRD-/- patients corresponded to 25%, while MRD+/+ (refractory) patients represented 29%. Partial response refers to patients who did not respond at day 15 but achieved complete remission at the end of induction (46%). Relapse occurred in 4 patients (14%), and 9 patients (32%) died. The data obtained in the validation cohort are like those of the descriptive cohort.

Table 1

CLINICAL CHARACTERISTICS OF THE PATIENTS INCLUDED IN THE STUDY
CLINICAL CHARACTERISTICS	DESCRIPTIVE COHORT (n = 28)		VALIDATION COHORT (n = 18)
	n (%)	MEAN (RANGE)	n (%)	MEAN (RANGE)	p - value
AGE (years)		10 (3–17)		8.26 (1–15)	0.17
SEX
Female	12 (43)		9 (50)		0.22
Male	16 (57)		9 (50)		0.22
WBC (cel/µ)		61.85 (1.43–487)		40.39 (0.8–269)	0.47
RISK
Low	1 (3.6)		3 (16.7)		0.06
Intermediate	20 (71.4)		11 (61.1)
High	7 (25)		4 (22.2)
EXTRAMEDULLAR INFILTRATION
Yes	3 (11)		3 (17)		0.55
No	25 (89)		15 (83)		0.55
CORTICOID RESPONSE
Yes	26 (93)		17 (94)		0.83
No	2 (7)		1 (6)		0.83
CARIOTYPE/MOLECULAR ALTERATIONS
Normal	17 (60.7)		11 (61.1)		0.08
Hipodiploid	0 (0)		0 (0)
Hyperdiploid	1 (3,50)		3 (16.7)
t(1;19)	4 (14,3)		1 (5.6)
t(4;11)	0 (0)		0 (0)
t(9;22)	1 (3.5)		0 (0)
t(12;21)	0 (0)		1 (5.6)
Amp Cr 21	1 (3.5)		0 (0)
Other	4 (14.3)		2 (11.1)
MRD DAY 15
Positive	20 (71)		11 (61)		0.46
Negative	8 (29)		7 (39)		0.46
MRD END OF INDUCTION
Positive	9 (32)		3 (17)		0.24
Negative	19 (68)		15 (83)		0.24
TRUE RESPONSE
MRDday15+ & MRDday33+	8 (29)		3 (17)		0.53
MRDday15- & MRDday33-	7 (25)		7 (39)
Parcial	13 (46)		8 (44)
RELAPSE
Yes	4 (14)		3 (17)		0.82
No	24 (86)		15 (83)		0.82
DEATH
Yes	9 (32)		3 (17)		0.17
No	19 (68)		15 (83)		0.17
MRD indicates minimal residual disease; RNAseq, RNA sequencing, RT-qPCR, quantitative reverse transcription PCR

To identify DEGs between responders and non-responders, we balanced the groups and randomly compared 8 responders vs. 8 non-responders. A supervised hierarchical cluster analysis compared gene expression profiles between BM samples at diagnosis from responders vs. non-responders at day 15 of induction treatment. We detected 159 DEGs (Fig. 1a). Similarly, a second supervised analysis compared BM samples at diagnosis from responders vs. non-responders at the end of induction treatment and found 200 DEGs (Fig. 1b). The third supervised approach was done between BM samples at the time of diagnosis from MRD+/+ vs. MRD-/- patients and found 153 DEGs (Fig. 1c). Finally, a Venn diagram was used to identify common DEGs between the three analyses, finding 50 of them (Fig. 1d). In non-responders, 46 of 50 common genes were up-regulated, and 4 of 50 were down-regulated (see Supplementary Table S1 online). Uniprot analysis showed that some genes were associated with different functions, including DNA binding (HOXB2, RFX8), transmembrane (SLC18A2, SLC45A3, BOC, ITGA3), lipidation (TNFRSF10C, PTGIR), among others. In addition, enrichment analysis showed that these genes are involved in processes like inflammatory response, transmembrane transport of sugars, cellular response to growth factor stimulus, etc.

To identify GDMCpGs, we used the same groups as in Fig. 1. A total of 1 834 GDMCpGs were obtained when comparing diagnosis BM of patients with MRD + and MRD- at day 15 (Fig. 2a); 2 382 at day 33 (Fig. 2b); and 2 077 between MRD+/+ and MRD-/- patients (Fig. 2c). We identified 687 common GDMCpGs across all comparison groups (Fig. 2d).

To identify genes with both differential expression and methylation, we matched the previously obtained DEGs and GDMCpGs. We found 38 genes at day 15 of induction treatment, 52 genes at the end of induction, and 40 genes between patients with MRD+/+ and MRD-/-. Finally, a Venn diagram analysis identified 10 common genes (HOXB2, TMEM51, F2RL1, PDZRN3, BOC, TANC1, ASCL2, FBN2, MIR4435-2HG, and SCL18A2) (Fig. 2e). We found an inverse correlation between gene overexpression and hypomethylation in only four of them, which were present in all three comparison groups (BOC, MIR4435-2HG, SCL18A2, and ASCL2)(Table 2). We selected six additional DEGs from the analyses between the MRD+/+ vs. MRD-/- comparison; these genes also had GDMCpGs, and there was an inverse correlation between them. The resulting genes were DAPK1, NPDC1, CNKSR3, ITGA6, SCL45A3, and CTHRC1(see Supplementary Table S2 online).

Table 2

COMMON GENES BETWEEN THREE ANALYSIS GROUPS AND THEIR CORRESPONDS CpGDM
GENE/CpGs			MRD DAY 15		MRD DAY 33		TRUE RESPONSE
	FC	METHYLATION STATUS	p-value	Rho	p-value	Rho	p-value	Rho
HOXB2	5,1
cg22777724		Hypermethylation	0.45	0.2	0.08	0.47	0.08	0.49
cg09313705		Hypermethylation	0.32	0.26	0.18	0.37	0.45	0.22
cg22807449		Hypermethylation			0.18	0.53
TMEM51	5,4
cg24835051		Hypomethylation			0.004	-0.73
cg15319451		Hypomethylation			0.009	-0.67
cg13210325		Hypermethylation	0.09	0.43			0.54	0.18
cg07180646		Hypomethylation	0.08	-0.45
cg11618537		Hypomethylation					0.2	-0.37
F2RL1	3,5
cg12518146		Hypermethylation	0.001	0.72	0.003	0.74	0.005	0.74
PDZRN3	5,4
cg04340156		Hypomethylation			0.27	-0.31
cg11461794		Hypomethylation			0.86	0.05	0.87	0.049
cg24353443		Hypomethylation	0.66	-0.11			0.62	-0.14
cg05502701		Hypomethylation	0.6	-0.14			0.45	-0.22
BOC	4,6
cg11205552		Hypomethylation			0.68	-0.11
cg22413784		Hypomethylation	0.05	-0.48	0.02	-0.58	0.01	-0.68
TANC1	3,5
cg21390755		Hypomethylation	0.11	-0.4	0.03	-0.57	0.12	-0.45
ASCL2	5,4
cg22346124		Hypomethylation			0.13	-0.42
cg10290276		Hypomethylation			0.03	-0.55
cg13930892		Hypomethylation	0.004	-0.66	0.01	-0.65	0.02	-0.6
cg20912770		Hypomethylation			0.05	-0.52
cg10526374		Hypomethylation			0.05	-0.51
cg06263495		Hypomethylation	0.01	-0.6	0.02	-0.59
cg26051413		Hypomethylation					0.52	-0.19
cg11644479		Hypomethylation					0.02	-0.62
cg13762320		Hypomethylation					0.03	-0.58
cg19284039		Hypomethylation					0.09	-0.48
FBN2	-6
cg23696863		Hypomethylation	0.05	0.49	0.09	0.46	0.12	0.45
MIR4435-2HG	4,4
cg24783876		Hypomethylation	0.02	-0.55	0.05	-0.51	0.06	-0.53
SLC18A2	8,7
cg03570973		Hypomethylation	0.001	-0.72	0.001	-0.77	0.0009	-0.8
cg19721867		Hypomethylation			0.003	-0.72
cg10245915		Hypomethylation			0.01	-0.65
cg01043119		Hypomethylation			0.0005	-0.8
cg19617377		Hypomethylation			0.003	-0.72
cg00512279		Hypomethylation			0.001	-0.76	0.03	-0.58
cg08521987		Hypomethylation			0.05	-0.51	0.007	-0.7
cg27186877		Hypomethylation					0.03	-0.59

Ten upregulated DEGs with hypomethylated CpGs (BOC, MIR4435-2HG, SCL18A2, ASCL2, DAPK1, NPDC1, CNKSR3, ITGA6, SCL45A3, and CTHRC1) were selected for verification by RT-qPCR. The genes showing correlation between the two techniques were DAPK1, CNKSR3, MIR4435-HG2, CTHRC1, NPDC1, SLC45A3, ITGA6, and ASCL2(see Supplementary Fig. S1 online). BOC and SLC18A2 did not demonstrate correlation and were excluded from further analysis.

All samples (descriptive and validation cohort) were used to verify whether gene expression was different between responders and non-responders at each time point of the evaluation. Interestingly, according to the RNA-seq results, MIR4435-2HG, CNKSR3, ITGA6, NPDC1, and DAPK1 were overexpressed in MRD + in all comparison groups and had differential FC values between conditions (see Supplementary Fig. S2 online).

We used logistic regression to test whether genes could predict response to induction chemotherapy. MIR4435-2HG was found to be the best predictor of whether a patient will be a true responder, refractory (Figs. 3a and b), or non-responder at day 15 of induction (Figs. 3c and d). Regarding DAPK1, ASCL2, SCL45A3, NPDC1 and ITGA6, it was observed that they could also predict response at day 15 or whether the patient will be a true responder or refractory (see Supplementary Fig. S3 and S4 online); however, both MIR4435-2HG and the other genes do not predict whether the patient will respond at the end of induction (see Supplementary Fig. S5 online).

To test whether MIR4435-2HG could also predict the risk of death, we performed logistic regression using our RNA-seq counts. We observed that MIR4435-2HG can predict death with good sensitivity and specificity (Fig. 3e). Finally, patient overall survival was compared based on MIR4435-2HG gene expression. We observed that patients with MIR4435-2HG overexpression have worse survival than those without it (Fig. 3f).

When analyzing the relationship between normalized RNA-seq counts of MIR4435-2HG and the probability of non-response to treatment at day 15 or being refractory, we found that patients with counts > 5.1 had a 66% probability of being refractory to treatment. This probability increased proportionally to gene expression. In addition, for patients with counts > 4.97, the probability of therapeutic failure at day 15 was > 60% and increased progressively as MIR4435-2HG overexpression increased. Similarly, the probability of death increased when counts were > 7.0 (see Supplementary Table S3 online).

To test whether the identified genes could also be useful as prognostic biomarkers of relapse or death, first, we compared FC values between patients with MRD-/- and patients with relapsed using RT-qPCR analysis. We found that DAPK1 and CNKSR3 had higher FC in relapse samples than in MRD-/- samples, while MIR4435-2HG showed a tendency (Fig. 4a). No difference was observed with ITGA6, NPDC1, ASCL2, CTHRC1 and SLC45A3(see Supplementary Fig. S6 online).

To validate the role of MIR4435-2HG, we analyzed its expression in a larger patient cohort with longer follow-up, using normalized RNA-seq counts obtained from the BM of 58 children with recurrent ALL from TARGET Pan-Cancer database. The 5-year survival analysis showed that MIR4435-2HG overexpression is a poor prognostic variable for patient survival (Fig. 4a). Interestingly, average counts in this cohort (7.82) correlate with our cutoff points, which would give a 100% probability of being refractory and an 80% probability of dying. Nevertheless, in this database, DAPK1 and CNKSR3 were not correlated with survival.

Four Cox regression models were performed to determine which variable might affect patient survival. The first model employed the clinical variables used currently to define the risk of death; in this model, no variable conferred risk of death. The second model used the same clinical variables and gene overexpression; in this model, no variable conferred risk either. However, in the third model, when using only MRD and the selected genes, MIR4435-2HG overexpression was the only variable that increased the risk of death 29 times. Similarly, in the fourth model, when evaluating the entire gene profile, MIR4435-2HG overexpression affected the risk of death (Table 3). Strikingly, survival analysis showed that patients overexpressing MIR4435-2HG had worse survival compared to those who did not overexpress MIR4435-2HG(Fig. 4b).

Table 3

MULTIPLE COX REGRESSION USING CURRENTLY CLINICAL VARIABLES AND GENE PROFILE
	MODEL 1. CURRENTLY USED CLINICAL VARIABLES		MODEL 2. CURRENTLY USED CLINICAL VARIABLES AND GENE PROFILE		MODEL 3. GENE PROFILE AND MRD		MODEL 4. GENE PROFILE
PARAMETER			p-value	OR (95%CI)	p-value	OR (95%CI)	p-value	OR (95%CI)
Age	0.76	NC	0.46	NC
WBC	0.62	NC	> 0.99	NC
Extramedullar infiltration	> 0.99	NC	> 0.99	NC
Corticoid response	0.15	NC	> 0.99	NC
MRD + day 15	0.36	NC	0.51	0.32 (0.002–11.91)	0.34	0.25 (0.006–3.80)
MRD + final induction	0.22	NC	0.36	3.71 (0.1786–121.0)	0.23	3.49 (0.357–37.17)
MIR4435-2HG			0.22	NC	0.01	29.42 (2.52-1 067)	0.005	16.47 (2.49–157.2)
DAPK1			> 0.99	NC	0.41	3.12 (0.16–96.95)	0.32	2.87 (0.26–24.31)
ITGA6			0.18	NC	0.11	0.03 (0.00002–1.59)	0.14	0.09 (0.002–2.27)
NPDC1			> 0.99	NC	0.10	13.22 (0.68–631)	0.09	9.40 (0.90–221.5)
CNKSR3			0.92	NC	0.87	1.17 (0.11–9.55)	0.97	0.97 (0.11–7.07)
ASCL2			> 0.99	NC	0.73	1.58 (0.08–29.20)	0.92	0.88 (0.05–11.05)
CTHRC1			> 0.99	NC	0.52	0.43 (0.02–5.49)	0.24	0.26 (0.01–2.56)
SCL45A3			> 0.99	NC	0.51	2.55 (0.08–45.41)	0.54	2.11 (0.11–24.78)
MRD indicates minimal residual disease; NC, No calculable, WBC, white blood counts

However, since MRD is the most used variable to define risk of death, for survival analysis, we considered MRD at the end of induction, effectively showing that MRD- patients have better survival than MRD+ (Fig. 4c). Interestingly, when we compared survival combining MRD + with MIR4435-2HG overexpression vs. MRD- with down-expression of MIR4435-2HG, we observed that these two variables separate the survival curves better, where patients with MRD + and MIR4435-2HG overexpression presented worse survival (Fig. 4d).

Finally, since more than half of our patients had normal karyotype, we wanted to test whether selected genes could improve risk classification in patients with normal karyotype. Surprisingly, we found that simultaneous overexpression of MIR4435-2HG, DAPK1, and ASLC2 was associated with poorer survival in these patients compared to those who did not overexpress them (Fig. 4e).

Since ALL are diseases with high heterogeneity at the genetic and epigenetic level, new biomarkers are needed to improve patient prognosis [11]. The present study conducted an integrative analysis of genome-wide DNA methylation and gene expression by RNA-seq in a cohort of 27 pediatric patients with B-cell ALL to investigate whether differential DNA methylation genes and gene expression patterns could be proposed as potential predictive biomarkers that differentiate responder from non-responder patients and confer risk of relapse and death in pediatric patients with B-cell ALL.

Aberrant DNA methylation has been considered a hallmark in different types of cancer, including acute lymphoblastic leukemia (ALL) [12]. As reported by Borssén et al [13], our study revealed a clear separation in both DNA methylation and gene expression profiles between responders and non-responders in all comparisons made; finding that overexpressed genes conferred a more aggressive phenotype [5]. Previously, Figueroa et al [14], reported that aberrant DNA methylation in childhood ALL could be an important determinant of gene expression in disease-specific alterations. We observed correlation between hypomethylation of CpGs and overexpression of DAPK1, CNKSR3, MIR4435-HG2, CTHRC1, NPDC1, SLC45A3, ITGA6, ASCL2, BOC and SLC18A2 genes; demonstrating that changes in DNA methylation can influence gene expression.

Aberrant patterns of DNA methylation have been associated with clinical outcome in patients with ALL; however, additional research is required to assess the clinical utility of some of these findings [5]. In this study, we selected MIR4435-2HG, DAPK1, ASCL2, SCL45A3, NPDC1 and ITGA6 as potential biomarkers of treatment response, whose overexpression predicts, with high sensitivity and specificity, treatment failure or refractoriness. Unfortunately, none of the genes predicted response at the end of induction, although this could be explained by the lack of a significant number of patients with MRD + for comparison. This analysis could be verified with a larger group of patients with MRD+. Nevertheless, we highlight MIR4435-2HG because its overexpression is the one that best predicts therapeutic failure.

Recently, selected genes had been described as possible diagnostic and prognostic biomarkers in different types of cancer and other non-neoplastic diseases (see Supplementary Table S4 online) [15–43]. In particular, MIR4435-2HG, which is a long non-coding RNA, also named LncRNA-AWPPH, LINC00978, or MORRBID [15]. Interestingly, MIR4435-2HG overexpression has previously been correlated with hypomethylation in gliomas [34]. In patients with T-cell ALL, MIR4435-2HG was overexpressed compared to healthy patients, and has been shown to promote proliferation and inhibit apoptosis of ALL cell lines [38]. Although the biological role of MIR4435-2HG is still under study, it is known to participate by deregulating different signaling pathways related to proliferation, invasion, migration, epithelial-mesenchymal transition, and apoptosis; specifically, participating in signaling pathways such as TGF-β, WNT-βcatenin, MDM2/p53, PI3K/AKT, Hippo, and MAPK/ERK [15].

Unfortunately, no clinical variable was observed to increase the risk of death in our population; however, MIR4435-2HG overexpression increases risk, predicts death, and is associated with poorer survival. Similar results had been reported in acute myeloid leukemia (AML) by Zhigang Cai et al [44], who observed MIR4435-2HG overexpression and its association with poor survival in other cancer models, MIR4435-2HG overexpression has also been associated with worse progression-free survival and overall survival [34].

There are no studies that show the relationship between aberrant DNA methylation, gene overexpression, and treatment response, specifically at day 15 and the end of induction chemotherapy. However, some authors have found differences between the methylation profiles of patients who relapsed and those who did not [13], where less methylated CpG island methylator phenotype at diagnosis had an inferior overall survival compared to patients who had more methylated CpG island phenotype. Previously, Sandoval et al[45] reported hypomethylation in different genome regions, including Polycomb target genes, and its association with poor survival and relapse. Similarly, Hogan et al. reported epigenetic dysregulation in the acquisition of chemoresistance in relapse in genes CDKN2A, COL6A2, PTPRO, and CSMD1 [8]. Here, we reported overexpression of MIR4435-2HG, DAPK1, and CNKRS3 in relapse samples, although the samples were not matched; in any case, it is still a surprising result because these genes are elevated in both non-responder and relapse diagnostic samples, indicating that they could be biomarkers of relapse. These results were also confirmed in the TARGET Pan-Cancer. Interestingly, in our study, 3 of 4 patients who died had MIR4435-2HG overexpression; these patients were MRD- at the end of induction and were intermediate risk but relapsed and died.

This could be a reason to implement a genetic panel, including MIR4435-2HG, especially when 25% of pediatric leukemia patients have no detectable genetic alterations and have a low mutation rate making risk classification difficult [46]. In fact, 60% of our patients had no genetic alterations but overexpressed MIR4435-2HG, DAPK1 and ASCL2 correlating with worse survival, suggesting that expression of these genes could improve risk classification in these patients.

While MIR4435-2HG overexpression appears to be a poor prognostic factor, the association of DAPK1 expression with poor prognosis is controversial [47]. In this study, we associated DAPK1 overexpression and hypomethylation with therapeutic failure, relapse, and poorer survival in patients with normal karyotype. However, in the Pan-Cancer analysis, patients with this overexpression had better survival. Therefore, there is no consensus on its prognostic role. DAPK1 has also been associated with resistance to imatinib in chronic myeloid leukemia [48], autophagia [16], alterations in the p53 signaling pathway in chronic lymphocytic leukemia [49] and methylated in myelodysplastic syndrome [39].

The limitations of our study are related to the sample size. Although it is a descriptive study, these findings must be validated in a larger patient cohort to demonstrate that MIR4435-2HG allows improving risk classification. Due to the coexistence of new and classical technologies today to analyze gene methylation and expression, it is not easy to compare results between studies, and reproducible results may not be found.

Here, we propose a gene profile that can predict treatment response. Particularly, we demonstrate for the first time that MIR4435-2HG is overexpressed and hypomethylated in non-responder patients with B-cell ALL; it can predict treatment response and confers a higher risk of death to those patients who overexpress it. Its detection could be combined with MRD to improve risk classification in our patients, specifically those with normal karyotype and provide a new biomarker to evaluate post-treatment follow-up with the potential to be a useful predictive and prognostic biomarker in Hispanic children with B-cell ALL.

Patient samples

The study included 45 patients with B cell-ALL treated at the Instituto Nacional de Cancerología, Hospital Militar Central, and Hospital Universitario San Ignacio (Bogotá, Colombia). Patients with a new diagnosis of B-ALL, no prior chemotherapy, no history of cancer, no diseases associated with genetic alterations, and < 18 years of age were included. The diagnosis was confirmed using flow cytometry (FCM)[50] and morphological analysis of BM. This study was conducted following the recommendations of the Colombian Regulation for Research in Humans (Resolution 8430 of 1993, Ministry of Health of Colombia) and in accordance with the Declaration of Helsinki and approved by each participating institution’s Institutional Review Boards (IRB). All methods for nucleic acid analysis were approved by the LSUHSC Translational Genomics Core’s Institutional Biosafety Committee (IBC) protocol number 17370. Informed consent was signed by the parents of all participants. Treatment for ALL is based on the Berlin-Frankfurt-Munich (BFM) protocol, according to risk [51]. Patients with treatment abandonment or non-adherence to it were excluded. The descriptive cohort consisted of 27 BM samples in which RNA-seq/DNA methylation protocols were performed; 18 additional patients were included in the validation cohort by RT-qPCR, as well as BM from 15 relapsed patients.

According to the BFM protocol, response to induction therapy was evaluated by FCM detecting MRD at day 15, where patients with < 0.1% residual blasts in BM were classified as responders (MRD- day15), and patients with > 0.1% residual blasts were considered non-responders (MRD + day15). At day 33, patients classified as responders had < 0.01% residual blasts in BM (MRD- day33), while non-responders had > 0.01% residual blasts (MRD + day33) [51]. Therefore, we considered patients with MRD- day15 and MRD- day33 as true responders (MRD-/-) and patients with MRD + day15 and MRD + day33 as refractory patients (MRD+/+), this comparison was called a "true response”.

Blasts isolation and purification

BM samples (diagnosis and relapse) were collected by a hemato-oncologist and processed within 24 hours after sample collection. First, mononuclear cells were separated from BM by density-gradient centrifugation (Lymphoprep, Lonza). The blasts were separated using magnetic microbeads coated with anti-CD19 or anti-CD34 antibodies, followed by MACS column enrichment (Miltenyi, Bergisch Gladbach, Germany). The purity of sorted blasts was assessed with CD34-PERCPCy5.5, CD45-V500, CD19-PECy7, and CD10 APC antibodies. Data was acquired in a FACSCanto II flow cytometer (Becton/Dickinson Biosciences, San Jose, CA), using the FACSDiva software program. Infinicyt software (Cytognos SL, Salamanca, Spain) was used for data analysis [52].

DNA and RNA extraction

DNA and RNA were extracted from MACS-sorted blasts using the Allprep mini kit and robotic workstation QIAcube (Qiagen, Hilden, Germany). RNA quality was evaluated using the Agilent RNA 6000 Nano and Pico kits, and data were analyzed with Agilent 2100 Bioanalyzer. RNA concentration was calculated using the Qubit™ RNA High Sensitivity and Broad Range kits, while DNA concentration was calculated using the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific).

Library preparation and RNA sequencing

Samples with RIN > 6 and purity by FCM with > 90% blasts were selected for RNA-seq. For RNA library preparation, 300 ng of total RNA was used. TruSeq Stranded mRNA RNA libraries were prepared following Illumina’s protocol. Resulting libraries were sequenced at 2 x 75 bp on a NextSeq550 sequencer system at the Stanley S. Scott Cancer Center’s Translational Genomics Core at LSUHSC-New Orleans. On average, more than 50 million reads were obtained per sample.

DNA methylation assay

Bisulfite conversion was performed in 500 ng of DNA for each sample following the recommendations of the EZ DNA Methylation-Startup Kit (Zymo Research, USA). Bisulfite-converted DNA was amplified and hybridized to the Infinium Methylation EPIC Kit chips and scanned on the Illumina’s iScan.

RT-qPCR

Total RNA from sorted blasts was treated with DNase I Amplification Grade (Invitrogen, USA) prior to reverse transcription. cDNA was synthesized using the SuperScript III First-Strand Synthesis SuperMix Kit (Invitrogen, USA), following the manufacturer’s procedures. TaqMan probes were used to quantify mRNA expression levels of candidate genes obtained by RNA-seq analysis (Assay IDs: DAPK1 Hs00234489_m1; NPDC1 Hs00209870_m1; CNKSR3 Hs00295109_m1; SLC18A2 Hs00996835_m1; CTHRC1 Hs00298917_m1; BOC Hs00264408_m1; SLC45A3 Hs00263832_m1; GAPDH Hs99999905_m1, ASCL2 Hs00270888_s1; MIR4435-2HG Hs03680374_m1). The reaction was amplified in a QuantStudio 12 K plex Real-Time PCR machine (Applied Biosystems). The 2^−ΔΔCT method was used to estimate the fold induction of each gene using GAPDH and Ct values to determine the fold change (FC) for each sample. A pool of samples was used as internal calibrator, as well as water as negative control. Assays were done in duplicate.

Statistical analysis

RNA-seq data quality control and filtering

The FASTQ files were uploaded to Partek Flow software to remove contaminants and align the reads to STAR 2.7.3 (reference genome hg38). Reads were quantified using RefSeq Transcrips 96. A quality analysis removed genes with zero or less than five reads. One sample with a low correlation (< 0.4) with respect to the others was removed.

Methylation DNA data quality control

Low-confidence probes (p-value > 0.05) and probes mapped to X and Y chromosomes were excluded. Normalization was done using NOOB (normalization for Illumina Infinium methylation arrays).

Transcriptomic and methylation data analysis

Normalization and differential expression analysis were performed using the Deseq2 library in RStudio. Differentially expressed genes (DEGs) were selected if they had a p-value < 0.05 and a FC > 2. Ggplot library was used to generate heatmaps, and GenomeStudio to calculate beta values for each hybridized probe. The Partek Genomic Suite was used for differential methylation analysis to identify and extract genes with differentially methylated CpGs (GDMCpGs). GDMCpGs were chosen if they had FC > 2 and FDR < 0.05. Enrichment analysis and functional gene annotation were performed using DAVID and Uniprot, respectively. Spearman correlation was used to determine correlation between overexpressed genes and hypomethylated probes; those with an inverse correlation less than − 0.50 and a p-value < 0.05 were selected.

Experimental design

Supplementary Fig. S7 online describes the methodological design of the study. The gene and methylation profile of induction treatment were compared between responders vs. non-responders in each comparison group. Treatment response was the only variable used to define profiles in each comparison.

RT-qPCR analysis

Spearman correlation was used to determine any correlation between RNA-seq counts and FC and Ct values for RT-qPCR. Genes with p-value < 0.05 and r > 0.50 were selected. T-tests were used to compare FC between responders and non-responders in comparison groups and to compare FC MRD-/- and relapse samples. GraphPad software was used for statistical tests and graphic images.

Clinical data analysis

To compare clinical variables between patient cohorts, t-test and chi-square tests were used. The follow-up time for relapse and death was 2 years. Logistic regression analysis was performed to evaluate whether candidate genes could predict treatment response, refractoriness, or true response. Survival analyses were estimated according to gene expression using Kaplan-Meier curves. Cox regression was used to determine whether gene expression conferred a higher risk of death. The Youden index of normalized RNA-seq counts was used to define the cutoff threshold for overexpression for each gene. The Xena Browser (http://xena.ucsc.edu/) was used to observe the relationship between MIR4435-2HG, CNKSR3, and DAPK1 overexpression and survival [53], using the following filters recurrent ALL samples of TARGET-Pan-Cancer cohort, ALL diagnosis, < 18 age, expression average of 7.82 for MIR4435-2HG, 10.58 for DAPK1, and 5.08 for CNKSR3, and 5-year overall survival.

Acknowledgments

We are mainly grateful to the patients who agreed to participate in the study, as well as to their tutors. We also thank the institutions that recruited the patients (Instituto Nacional de Cancerología, Hospital Militar Central and Hospital Universitario San Ignacio), especially Dr. Amaranto Suárez who provided support in the diffusion of the project within the Instituto Nacional de Cancerología, as well as Dr. David Garay, Dr. Eddie Pabón, Dr. Camila Prada, Dr. Jorge Buitrago, Andrea Naranjo, Giovanna Bedon, Bibiana Martinez, Paula Toro and Cindy Arevalo. Finally, special thanks to the Fundación Colombiana de Leucemia y Linfoma for the financial support to Yulieth Torres-Llanos in her PhD studies.

Author contributions

YTL: designed research, performed research, contributed vital analytical tools, collected data, analyzed, and interpreted data, performed statistical analysis, and wrote the manuscript. JZ: contributed vital new reagents and analytical tools, analyzed and interpreted data, performed statistical analysis, and wrote the manuscript. ALC: designed research, contributed vital new reagents and analytical tools, analyzed and interpreted data, and wrote the manuscript. PG: collected data, analyzed, and interpreted data, and wrote the manuscript. IR: collected data, analyzed, and interpreted data, and wrote the manuscript. NP: collected data, analyzed, and interpreted data, and wrote the manuscript. LL: contributed vital analytical tools, analyzed, and interpreted data, performed statistical analysis, and wrote the manuscript. AI: collected data, contributed vital new reagents and analytical tools, and wrote the manuscript. NCR: designed research, analyzed and interpreted data, and wrote the manuscript. SQ: analyzed and interpreted data and wrote the manuscript.

Competing interests: The authors declare no competing interests.

Data availability: patient's clinical data may be requested to [email protected] under a justifiable reason. Methylation and gene expression data are available in the GEO repository under accession number GSE229056. Link: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE229056. Reviewers token: cdonecwstpcfvst.

Ethics declarations: this study was approved by the Comité de Ética en Investigaciones del Instituto Nacional de Cancerología (C-19010300404), the Comité de Investigaciones y ética del Hospital Universitario San Ignacio (2017/114) and the Comité de Ética en Investigación del Hospital Militar Central (2019-006). Informed consent was signed by the legal guardians of all participants.

Funding: This study was supported by the Instituto Nacional de Cancerología and Hospital Universitario San Ignacio.

Belson, M., Kingsley, B. & Holmes, A. Risk Factors for Acute Leukemia in Children: A Review. Environ Health Perspect 115, 138–145 (2006).
Vera, A. M., Pardo, C., Duarte, M. C. & Suárez, A. Experiencia en el análisis de la mortalidad por leucemia aguda pediátrica en el Instituto Nacional de Cancerología. Biomédica 32, 355–364 (2012).
Pagna Sok et al. Disparities in relapse among a large multi-ethnic population of children diagnosed with acute lymphoblastic leukemia (ALL): A report from the Reducing Ethnic Disparities in Acute Leukemia (REDIAL) Consortium. (2022).
Alaggio, R. et al. The 5th edition of the World Health Organization Classification of Haematolymphoid Tumours: Lymphoid Neoplasms. Leukemia vol. 36 1720–1748 Preprint at https://doi.org/10.1038/s41375-022-01620-2 (2022).
Nordlund, J. & Syvänen, A. C. Epigenetics in pediatric acute lymphoblastic leukemia. Semin Cancer Biol 51, 129–138 (2018).
Ji, H. et al. Comprehensive methylome map of lineage commitment from haematopoietic progenitors. Nature 467, 338–342 (2010).
Scala, G., Federico, A., Palumbo, D., Cocozza, S. & Greco, D. DNA sequence context as a marker of CpG methylation instability in normal and cancer tissues. Sci Rep 10, 1–11 (2020).
Hogan, L. E. et al. Integrated genomic analysis of relapsed childhood acute lymphoblastic leukemia reveals therapeutic strategies. Blood 118, 5218–5227 (2015).
Pui, C., Mullighan, C. G., Evans, W. E. & Relling, M. v. Pediatric acute lymphoblastic leukemia: where are we going and how do we get there ? Blood 120, 1–3 (2017).
Dominique J. P. M. Stumpel,1 Pauline Schneider, Eddy H. J. van Roon, Judith M. Boer, Paola de Lorenzo, Maria G. Valsecchi, Renee X. de Menezes, Rob Pieters, and R. W. S. Specific promoter methylation identifies different subgroups of MLL-rearranged infant acute lymphoblastic leukemia, influences clinical outcome, and provides therapeutic options. Blood 114, 5490–5498 (2009).
Lejman, M., Chałupnik, A., Chilimoniuk, Z. & Dobosz, M. Genetic Biomarkers and Their Clinical Implications in B-Cell Acute Lymphoblastic Leukemia in Children. International Journal of Molecular Sciences vol. 23 Preprint at https://doi.org/10.3390/ijms23052755 (2022).
Hanahan, D. Hallmarks of Cancer: New Dimensions. Cancer Discovery vol. 12 31–46 Preprint at https://doi.org/10.1158/2159-8290.CD-21-1059 (2022).
Borssén, M. et al. DNA methylation holds prognostic information in relapsed precursor B-cell acute lymphoblastic leukemia. Clin Epigenetics 10, 1–7 (2018).
The Journal of … 123, 3099–3111 (2013).
Ghasemian, M. et al. Long non-coding RNA MIR4435-2HG: a key molecule in progression of cancer and non-cancerous disorders. Cancer Cell International vol. 22 Preprint at https://doi.org/10.1186/s12935-022-02633-8 (2022).
Movahhed, P. et al. The impact of DAPK1 and mTORC1 signaling association on autophagy in cancer. Mol Biol Rep 49, 4959–4964 (2022).
Qin, Y. et al. Effect of DAPK1 gene on proliferation, migration, and invasion of carcinoma of pancreas BxPC-3 cell line. Int J Clin Exp Pathol vol. 7 www.ijcep.com/ (2014).
Gasimli, K., Raab, M., Becker, S., Sanhaji, M. & Strebhardt, K. The Role of DAPK1 in the Cell Cycle Regulation of Cervical Cancer Cells and in Response to Topotecan. J Cancer 13, 728–743 (2022).
Lake, S. L. et al. Single nucleotide polymorphism array analysis of uveal melanomas reveals that amplification of CNKSR3 is correlated with improved patient survival. American Journal of Pathology 182, 678–687 (2013).
Sørensen, K. D. et al. Genetic and epigenetic SLC18A2 silencing in prostate cancer is an independent adverse predictor of biochemical recurrence after radical prostatectomy. Clinical Cancer Research 15, 1400–1410 (2009).
Lebedev, T. D. et al. Two receptors, two isoforms, two cancers: Comprehensive analysis of kit and trka expression in neuroblastoma and acute myeloid leukemia. Front Oncol 9, (2019).
Meng, C., Zhang, Y., Jiang, D. & Wang, J. CTHRC1 is a prognosis-related biomarker correlated with immune infiltrates in colon adenocarcinoma. World J Surg Oncol 20, (2022).
Sial, N. et al. CTHRC1 expression is a novel shared diagnostic and prognostic biomarker of survival in six different human cancer subtypes. Sci Rep 11, (2021).
Nguyen, C. H. et al. SOCS2 is part of a highly prognostic 4-gene signature in AML and promotes disease aggressiveness. Sci Rep 9, (2019).
Fattahi, S., Nikbakhsh, N., Ranaei, M., Sabour, D. & Akhavan-Niaki, H. Association of sonic hedgehog signaling pathway genes IHH, BOC, RAB23a and MIR195-5p, MIR509-3-5p, MIR6738-3p with gastric cancer stage. Sci Rep 11, (2021).
Mathew, E. et al. Dosage-dependent regulation of pancreatic cancer growth and angiogenesis by Hedgehog signaling. Cell Rep 9, 484–494 (2014).
Hu, X. G. et al. Elevated expression of ASCL2 is an independent prognostic indicator in lung squamous cell carcinoma. J Clin Pathol 69, 313–318 (2016).
Wu, L. et al. ASCL2 Affects the Efficacy of Immunotherapy in Colon Adenocarcinoma Based on Single-Cell RNA Sequencing Analysis. Front Immunol 13, (2022).
Esgueva, R. et al. Prevalence of TMPRSS2-ERG and SLC45A3-ERG gene fusions in a large prostatectomy cohort. Modern Pathology 23, 539–546 (2010).
Pin, E. et al. Identification of a Novel Autoimmune Peptide Epitope of Prostein in Prostate Cancer. J Proteome Res 16, 204–216 (2017).
Song, S. et al. Downregulation of ITGA6 confers to the invasion of multiple myeloma and promotes progression to plasma cell leukaemia. Br J Cancer 124, 1843–1853 (2021).
Yamakawa, N., Kaneda, K., Saito, Y., Ichihara, E. & Morishita, K. The increased expression of integrin α6 (itga6) enhances drug resistance in evi1 high leukemia. PLoS One 7, (2012).
Ouyang, W., Ren, L., Liu, G., Chi, X. & Wei, H. Lncrna mir4435-2hg predicts poor prognosis in patients with colorectal cancer. PeerJ 2019, (2019).
Zhong, C., Xie, Z., Zeng, L. H., Yuan, C. & Duan, S. MIR4435-2HG Is a Potential Pan-Cancer Biomarker for Diagnosis and Prognosis. Frontiers in Immunology vol. 13 Preprint at https://doi.org/10.3389/fimmu.2022.855078 (2022).
Zhu, L., Wang, A., Gao, M., Duan, X. & Li, Z. LncRNA MIR4435-2HG triggers ovarian cancer progression by regulating miR-128-3p/CKD14 axis. Cancer Cell Int 20, (2020).
Shen, H. et al. MIR4435-2HG regulates cancer cell behaviors in oral squamous cell carcinoma cell growth by upregulating TGF-β1. Odontology 108, 553–559 (2020).
Zhou, Z. H. et al. Chromatin accessibility changes are associated with enhanced growth and liver metastasis capacity of acid-adapted colorectal cancer cells. Cell Cycle 18, 511–522 (2019).
Li, X., Song, F. & Sun, H. Long non-coding RNA AWPPH interacts with ROCK2 and regulates the proliferation and apoptosis of cancer cells in pediatric T-cell acute lymphoblastic leukemia. Oncol Lett 20, (2020).
Greco, M. et al. Promoter methylation of DAPK1, E-cadherin and thrombospondin-1 in de novo and therapy-related myeloid neoplasms. Blood Cells Mol Dis 45, 181–185 (2010).
Dong, R. et al. Models for Predicting Response to Immunotherapy and Prognosis in Patients with Gastric Cancer: DNA Damage Response Genes. Biomed Res Int 2022, 4909544 (2022).
Tong, C.-W. et al. Novel genes that mediate nuclear respiratory factor 1-regualted neurite outgrowth in neuroblastoma IMR-32 cells. Gene 515, 62–70 (2013).
Bloomston, M. et al. Identification of Molecular Markers Specific for Pancreatic Neuroendocrine Tumors by Genetic Profiling of Core Biopsies. Ann Surg Oncol 11, 413–419 (2004).
Potter, C. et al. Maternal Red Blood Cell Folate and Infant Vitamin B12 Status Influence Methylation of Genes Associated with Childhood Acute Lymphoblastic Leukemia. Mol Nutr Food Res 62, 1800411 (2018).
Cai, Z. et al. Targeting Bim via a lncRNA Morrbid Regulates the Survival of Preleukemic and Leukemic Cells. Cell Rep 31, (2020).
Sandoval, J. et al. Genome-wide DNA methylation profiling predicts relapse in childhood B-cell acute lymphoblastic leukaemia. Br J Haematol 160, 404–418 (2012).
Iacobucci, I. & Mullighan, C. G. Genetic basis of acute lymphoblastic leukemia. Journal of Clinical Oncology 35, 975–983 (2017).
San Jose-Eneriz, E., Agirre, X., Rodriguez-Otero, P. & Prosper, F. Epigenetic regulation of cell signaling pathways in acute lymphoblastic leukemia. Epigenomics 5, 525–538 (2013).
Guru, S. A. et al. Aberrant hydroxymethylation in promoter CpG regions of genes related to the cell cycle and apoptosis characterizes advanced chronic myeloid leukemia disease, poor imatinib respondents and poor survival. BMC Cancer 22, (2022).
Wang, L. Q. et al. Epigenetic inactivation of mir-34b/c in addition to mir-34a and DAPK1 in chronic lymphocytic leukemia. J Transl Med 12, (2014).
van Dongen, J. J. M. et al. EuroFlow antibody panels for standardized n-dimensional flow cytometric immunophenotyping of normal, reactive and malignant leukocytes. Leukemia 26, 1908–1975 (2012).
Stary, J. et al. Intensive chemotherapy for childhood acute lymphoblastic leukemia: Results of the randomized intercontinental trial ALL IC-BFM 2002. Journal of Clinical Oncology 32, 174–184 (2014).
Cruz-Rodriguez, N. et al. High expression of ID family and IGJ genes signature as predictor of low induction treatment response and worst survival in adult Hispanic patients with B-Acute lymphoblastic leukemia. Journal of Experimental and Clinical Cancer Research 35, 1–14 (2016).
Goldman, M. J. et al. Visualizing and interpreting cancer genomics data via the Xena platform. Nat Biotechnol 38, 675–678 (2020).

No competing interests reported.

SupplementaryInformationSR.pdf

Download PDF

Version 1

posted

You are reading this latest preprint version

MIR4435-2HG as a Novel Predictive Biomarker of Chemotherapy Response and Death in Pediatric B-Cell All

Status:

Version 1

Abstract

Figures

Introduction

Results

Discussion

Methods

Patient samples

Blasts isolation and purification

DNA and RNA extraction

Library preparation and RNA sequencing

DNA methylation assay

RT-qPCR

Statistical analysis

RNA-seq data quality control and filtering

Methylation DNA data quality control

Transcriptomic and methylation data analysis

Experimental design

RT-qPCR analysis

Clinical data analysis

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1