Introduction

T-cell acute lymphoblastic leukemia (T-ALL) is an aggressive hematological malignancy that is classified into different genetic subtypes based upon the aberrant expression of specific transcription factor oncogenes (TAL, TLX1, TLX3 or HOXA) or the arrest at a specific stage of T-cell differentiation (immature T-ALL)1,2,3,4. Notably, these molecular subgroups are characterized by unique mRNA and long non-coding RNA expression signatures, which partially reflect their putative cell of origin1, 2, 5.

MicroRNAs (miRNAs) are short non-coding RNAs that function as post-transcriptional repressors of specific target genes6, 7. Several studies have previously described a role for miRNAs in malignant T-cell transformation, including the identification of both an oncogenic (miR-19b, mir-20a, miR-26a, miR-92 and miR-223)8 as well as a tumor suppressor (miR-150, miR-155, miR-200 and miR-193b-3p)9, 10 miRNA network involved in T-ALL disease biology. However, studies that addressed the role of miRNAs in human T-ALL have largely been focused on previously recognized miRNA molecules as they consistently used RT-qPCR or microarrays as detection platforms.

More recently, small RNA-sequencing emerged as a more comprehensive technology that enables unbiased detection of the full spectrum of small RNA molecules. In addition, it also provides information on specific isoforms that differ from canonical miRNAs by the addition or deletion of one or more nucleotides at the 5′ or 3′ end of the miRNA11, 12. Notably, this heterogeneity in miRNA sequences, which is thought to result from RNA-editing, exonuclease activity or imprecise cleavage by DICER or DROSHA (ribonucleases involved in the miRNA processing)13, could be functionally relevant as shown for a number of miRNAs14,15,16,17.

Here, we used small RNA-sequencing to study the full spectrum of miRNAs that are expressed in human T-ALL samples. We demonstrate, for the first time, that molecular genetic subtypes of human T-ALL are characterized by unique miRNA expression signatures, delineate the pattern of miRNA isoforms that are expressed in malignant T-cells and use small RNA sequencing profiles of normal T-cell subsets to identify novel putative oncogenic or tumor suppressive miRNAs in the context of human T-ALL.

Results

Small RNA-sequencing of T-ALL patient samples, healthy donor thymocytes and T-ALL cell lines

To study the full spectrum of miRNAs involved in normal and malignant T-cell development, we performed small RNA-sequencing on 48 primary T-ALL patient samples of different T-ALL subgroups (13 immature, 14 TLX1 + or TLX3 +, 15 TAL-rearranged and 6 HOXA-overexpressing T-ALL samples), CD34+ and CD4+CD8+ normal thymocyte subsets from healthy donors and a panel of 7 T-ALL cell lines (Fig. 1a). While 1816 miRNAs were initially detected in the total panel of T-ALL patient samples, we further only considered the 574 miRNAs for which four reads were present in at least 60% of patients from one T-ALL subtype (Fig. 1b).

Figure 1
figure 1

Small RNA-sequencing of T-ALL samples, healthy thymocytes and cell lines detects novel T-ALL miRNAs and isomiRs. (a) Overview of samples profiled by small RNA-sequencing. (b) Dot plot representing the mean normalized expression levels of all 574 miRNAs detected by small RNA-sequencing of the 48 T-ALL patient samples. Each dot represents one miRNA and the miRNAs are ranked from highest to lowest mean expression. Black dots are miRNAs that were already detected by a qRT-PCR platform from previous studies. Red dots are the novel miRNAs detected in T-ALL samples. (c) Bar plot visualizing the distribution of the miRNAs by means of the amount of isomiRs they are represented by. (d) Correlation plot between the mean expression level of the miRNAs and their amount of isomiR forms. (e) Table representing the isomiR form that was represented by the highest expression level for each of the detected miRNAs. The first column denotes the 5′ overhangs or deletions, the second column the 3′ overhang or deletions of the isomiR in comparison to the canonical miRNA. The third column shows the amount of miRNAs of which the highest expressed miRNA had that isomiR form. (Graphics from www.somersault1824.com).

Full size image

Given that we previously performed miRNA expression profiling using the same T-ALL patient cohort using an RT-qPCR approach10, we were able to directly compare the results from the RT-qPCR study with our small RNA-sequencing effort. Using a similar background selection as mentioned above, 283 miRNAs were previously detected by RT-qPCR10, which converted into 248 unique miRNAs that are currently annotated in the most recent version of miRBase18. From these, 198 miRNAs were also detected using our current small RNA-sequencing approach. Therefore, the RT-qPCR approach still detected 50 miRNAs that were not identified in the small RNA-sequencing dataset. In contrast, small RNA sequencing detected 259 miRNAs, that were not included in a previous RT-qPCR analysis from the same T-ALL patient population10.

In Fig. 1b and Supplementary Fig. S1, the mean expression values for the 574 miRNAs, detected by small RNA-sequencing in primary T-ALL patients, are plotted, with the black dots representing the 280 miRNAs that were previously identified by RT-qPCR and the red dots representing the miRNAs detected for the first time in human T-ALL. Most of the formerly known miRNAs reside amongst the most highly expressed in human T-ALL (Fig. 1b). However, some of the newly identified miRNA transcripts in T-ALL also show a very high expression pattern (the top 10 of highest expressed novel miRNAs are annotated in Fig. 1b), suggesting that they might possess oncogenic potential in the context of T-ALL disease biology. Nevertheless, the average expression level of most novel T-ALL miRNAs that were exclusively detected by small RNA-sequencing is median to low.

Small RNA-sequencing also enables the detection of so-called isomiRs, i.e. miRNA isoforms that deviate from the canonical sequence by one or a few nucleotide(s)11, 12. In our small RNA-sequencing dataset, we identified 2139 isomiRs covering 481 different miRNAs, losing some very low expressed miRNAs from the analysis above. Although 154 miRNAs were only represented by one isomiR, some others showed expression of more than 10 different isomiR forms (Fig. 1c). We observed a positive correlation between the number of isomiRs detected for a specific miRNA and its expression level in human T-ALL (r = 0.89, Fig. 1d). For example, we detected 39 different isomiR forms for miR-181a-5p, the miRNA that shows the highest average expression in T-ALL. Remarkably, for 106 out of 481 miRNAs, the canonical miRNA was not expressed in our patient series. In addition, for only half of the miRNAs (246 out of 481), the canonical isoform showed the highest expression level, suggesting that a substantial amount of miRNAs are mainly represented by alternative isomiRs. The distribution of isomiRs that show the highest expression for each miRNA is shown in Fig. 1e.

Small RNA-sequencing reveals a subtype specific expression pattern of miRNAs in human T-ALL

Previous studies have convincingly shown that molecular genetic subtypes of human T-ALL display unique mRNA1, 2 and lncRNA5 expression signatures. Here, we used the 574 miRNAs detected by small RNA-sequencing to define subtype specific miRNA expression signatures in human T-ALL (adjusted p-value < 0.05; Fig. 2a; Supplementary Table S1). Principal Component Analysis confirmed that the most pronounced differences in miRNA expression are observed between immature and TAL-R T-ALL patient samples (Fig. 2b). Of note, although TLX1 and TLX 3 patients have previously been considered as one genetic entity, miRNA based clustering did reveal some differences between members of the TLX subtype (Fig. 2b). For example, the miRNA expression signature of a specific subset of TLX3 positive T-ALLs is clearly more related to immature T-ALL, which was not observed for any of the TLX1 positive leukemias (Fig. 2b). The 10 miRNAs that show the highest expression level in each of the genetic subtypes are shown in Fig. 2c. The miRNAs, which were not previously detected by the RT-qPCR platform, are depicted in red (Fig. 2c).

Figure 2
figure 2

Small RNA-sequencing of primary T-ALL samples reveals a subtype specific expression pattern of miRNAs. (a) Heatmap representing the top 50 most significantly up- or downregulated miRNAs per subgroup in comparison to the other subgroups (adjusted p-value < 0.05). (b) PCA-plot showing the distribution of the patient samples. The different colors denote patient samples from a different subgroup. (c) Table representing the selection of 10 miRNAs per subgroup. These were the highest expressed miRNAs that were significantly upregulated in that subgroup compared to the other subgroups. MiRNAs denoted in red were not detected by a previously used qRT-PCR platform. (d) Density plots representing the distribution of the miRNA expression in the LOUCY, HPB-ALL and JURKAT cell line. Vertical bars show the expression level of the top 10 miRNAs selected for the subgroup these cell lines represent. LOUCY represents the immature T-ALL subgroup, HPB-ALL the TLX1/3 subgroup and JURKAT the TAL-R subgroup.

Full size image

As mentioned above, small RNA-sequencing was also performed on a panel of 7 human T-ALL cell lines. These in vitro model systems reflect most of the different genetic subtypes of human T-ALL and included TLX1/3 positive (ALL-SIL, DND41 and HPB-ALL), TAL-R positive (PF-382, JURKAT and KE-37) and immature/HOXA overexpressing (LOUCY) tumor lines. Notably, based on the miRNAs represented in Fig. 2a, most T-ALL cell lines clustered separate from the T-ALL patient samples, with the exception of LOUCY, which clustered together with immature T-ALL patient samples, and ALL-SIL (Supplementary Fig. S2). Nevertheless, subtype specific miRNAs, which were identified in primary T-ALL patient samples, also showed high expression in the T-ALL cell lines from their corresponding genetic subtype (Fig. 2d and Supplementary Fig. S3). Therefore, the subtype specific tumor lines can be used as valuable in vitro tools to evaluate the role of specific miRNAs in the pathogenesis of this disease.

MiRNA profiling of normal thymocyte subsets reveals oncogenic subtype specific miRNAs

Small RNA-sequencing was also performed on CD34+ and CD4+CD8+ normal thymocyte samples from two healthy donor controls. First, DESeq2 analysis revealed 190 miRNAs that show significant differential expression between these CD34+ and CD4+CD8+ normal T-cell subsets (126 high in CD34+ and 64 high in CD4+CD8+; adjusted p-value < 0.05; Fig. 3a and Supplementary Table S2). The top ten most significant miRNAs are listed in Fig. 3b, with the miRNAs depicted in red those that were not covered in previous RT-qPCR analyses. Of note, small RNA-sequencing data were highly concordant between both donors for each subset (Fig. 3c and Supplementary Fig. S4).

Figure 3
figure 3

MiRNA profiling of healthy thymocyte subsets reveals different miRNA expression profiles between CD34+ and CD4+CD8+ subsets. (a) Diagonal plot showing the expression of the miRNAs in the different thymocyte subsets. Red dots represent miRNAs that are significantly higher expressed in the CD34+ subset, blue dots are the miRNAs significantly higher expressed in the CD4+CD8+ subset. (b) Top 10 most significantly upregulated miRNAs for the CD34+ subset and for the CD4+CD8+ subset. MiRNAs denoted in red were not detected by a previously used qRT-PCR platform. (c) Dot plots of two representative miRNAs for the subsets. **Significant difference with an adjusted p-value < 0.001.

Full size image

Next, we integrated these miRNA expression data obtained from healthy donors with the subtype specific miRNAs that were identified in the primary T-ALL patient cohort. For example, 4 miRNAs (hsa-miR-222-3p, hsa-miR-146a-5p, hsa-mir-221-3p and hsa-miR-126-5p) from the top 10 immature T-ALL specific miRNAs (Fig. 2c) also showed significant higher expression in CD34+ vs. CD4+CD8+ T-cell subsets (Fig. 4a). Similarly, three miRNAs (hsa-miR-16-5p, hsa-miR-16-2-3p and hsa-miR-450b-5p) from the top 10 TAL-R specific miRNAs (Fig. 2c) are significantly upregulated in CD4+CD8+ T-cell subsets (Fig. 4b). Therefore, these miRNAs most probably reflect the specific T-cell maturation arrest associated with these molecular genetic subtypes of T-ALL and their respective cell of origin.

Figure 4
figure 4

Subtype specific miRNAs can be either oncogenic or representative for the differentiation arrest of lymphoblasts. (a + c) Box plots showing the expression of the immature T-ALL specific miRNAs hsa-miR-126-5p and hsa-miR-222-3p in the immature T-ALL patients and the thymocyte subsets. (b + d) Box plots showing the expression of the TAL-R T-ALL specific miRNAs hsa-miR-16-2-3p and hsa-miR-450b-5p in the TAL-R T-ALL patients and the thymocyte subsets. *Significant difference with an adjusted p-value < 0.05; **significant difference with an adjusted p-value < 0.001.

Full size image

In order to identify miRNAs with putative oncogenic potential in specific T-ALL subgroups, we also performed differential expression analysis between immature T-ALLs and CD34+ thymocytes, and between TAL-R T-ALLs and CD4+CD8+ thymocytes, as these thymocyte subsets represent the stage of differentiation arrest during T cell development leading to these specific T-ALLs (Supplementary Tables S3 and S4). From the top ten immature specific miRNAs, three miRNAs (hsa-miR-21-5p, hsa-miR-222-3p and hsa-miR-101-3p) were significantly upregulated in the immature samples compared to the healthy control CD34+ samples. Remarkably, hsa-miR-222-3p was also significantly upregulated in the CD34+ subset, but its expression is further increased in immature T-ALLs (Fig. 4c). Similarly, three miRNAs (hsa-miR-182-5p, hsa-miR-29c-3p and hsa-miR-450b-5p), from the TAL-R subtype specific signature, show significant higher expression in the TAL-R T-ALLs as compared to their CD4+CD8+ normal counterparts. Here, hsa-miR-450b-5p was already higher expressed in the CD4+CD8+ double positive subset, but showed a further increase in activity in TAL-R rearranged leukemias (Fig. 4d).

Small RNA-sequencing reveals novel putative oncogenic miRNAs in human T-ALL

Finally, we aimed to identify novel T-ALL miRNAs with potential oncogenic activity in human T-ALL irrespective of the genetic subtypes. Differential expression analysis between 48 T-ALL samples and four normal thymocyte samples (Supplementary Table S5) resulted in the identification of 87 significantly upregulated miRNAs and 69 downregulated miRNAs in human T-ALL (Fig. 5a). Several miRNAs with a known oncogenic role in T-ALL8, 19 were recovered from this analysis and are depicted in Fig. 5a. However, and most notably, this analysis also identified different miRNAs, which were not previously detected in the context of normal and malignant T cell development, and which could potentially act as novel oncomiRs or tumor suppressor miRNAs involved in the biology of this disease (Fig. 5b,c).

Figure 5
figure 5

Novel oncogenic miRNAs are detected by small RNA-sequencing. (a) Diagonal plot showing the expression of the miRNAs in T-ALL compared to healthy donor thymocytes. Red dots represent miRNAs that are significantly higher expressed in the T-ALL patient samples, blue dots are the miRNAs significantly higher expressed in thymocyte subsets. (b) Top 10 most significantly upregulated miRNAs in the T-ALL patients and in the healthy donor samples. MiRNAs denoted in red were not detected by a previously used RT-qPCR platform. (c) Dot plots of two representative miRNAs from the table. **Significant difference with an adjusted p-value < 0.001.

Full size image

Discussion

More than a decade ago, Ferrando et al. and Soulier et al. described different T-ALL subtypes according to specific transcriptional profiles1, 2. Last year, we were able to show that these molecular genetic subtypes of human T-ALL also display unique long non-coding RNA expression signatures5. Here, we performed small RNA-sequencing on 48 T-ALL patient samples to finalize the transcriptional characterization of human T-ALL by a comprehensive analysis of miRNA expression in this disease.

Small RNA-sequencing enabled the detection of miRNAs that were not included in a previous RT-qPCR analysis of the same patient population10. Although most of the newly detected miRNAs in T-ALL were expressed at very low levels, we also identified a selection of novel T-ALL miRNAs with very high expression levels in T-ALL patient samples, including hsa-miR-181b-5p, hsa-miR-423-3p, hsa-miR-486-5p and hsa-miR-92b-3p (Fig. 1b). Interestingly, some of these miRNAs have previously been associated with malignant transformation in different tumor entities. For example, hsa-miR-181b-5p is a known oncogene in several cancer types as reviewed by Liu et al.20. In the context of leukemia, overexpression of hsa-miR-181b-5p was shown to enhance proliferation in acute myeloid leukemia by targeting MLK2. Furthermore, hsa-miR-92b-3p was identified as an oncomiR in glioblastoma by targeting SMAD321, which is known to be lost in several cases of pediatric T-ALL22, and PTEN 23, a well-established T-ALL oncosuppressor24, 25. In addition, hsa-miR-92b-3p is also specifically higher expressed in TLX1/3 T-ALL compared to the other T-ALL subtypes.

Interestingly, small RNA-sequencing also detects deviations from canonical miRNA sequences. Indeed, our analysis revealed 2139 different isomiR forms, corresponding to 481 miRNAs. Remarkably, for only half of the miRNAs, the canonical form showed the highest expression (Fig. 1e). In addition, most highly expressed isomiRs displayed modifications at their 3′ end, suggesting that these alterations would not affect the miRNA seed sequence and, therefore, have no functional effect on target recognition. Nevertheless, a number of studies have shown that these 3′ modifications might impact target specificity and stability of the miRNA17, 26.

Finally, small RNA sequencing of human CD34+ and CD4+CD8+ thymocytes enabled the identification of subtype specific and oncogenic miRNAs in the context of human T-ALL. An interesting example is hsa-miR-486. Hsa-miR-486-5p is one of the highest expressed newly identified miRNAs in T-ALL (Fig. 1b) and, together with hsa-miR-486-3p, it also shows higher expression in T-ALL samples as compared to healthy donor thymocytes. Hsa-miR-486-5p is an oncomiR in Down syndrome myeloid leukemias, where it is regulated by GATA127. Hsa-miR-486-3p has also been linked to erythroid development downstream of MYB, a known oncogene in T-ALL, and targeting MAF 28. The differences in miRNA expression could however also be due to the underrepresentation of normal thymocyte differentiation stages in this study. Additional experiments are thus needed to prove the oncogenic nature of the differentially expressed miRNAs.

Altogether, this study provides the first comprehensive overview of miRNA expression in molecular-genetic subtypes of human T-ALL. Integration of these signatures with miRNA expression profiles in normal T-cell subsets provides a unique resource to study novel miRNAs that are implicated in T-ALL disease biology.

Materials and Methods

Study design

Small RNA-sequencing was performed on 48 primary T-ALL patient samples, 7 T-ALL cell lines and 2 CD34+ and 2 CD4+CD8+ healthy donor thymocyte subsets, to profile the full T-ALL miRNA transcriptome.

Primary human T-ALL patient samples and T-ALL cell lines

Blood samples and bone marrow lymphoblast from T-ALL patients were collected after informed consent according to the Declaration of Helsinki from Saint-Louis Hospital, Paris, France. This study was approved by both the Institut Universitair d’Hématologie Institutional Review Board and the Ethical Committee of Ghent University Hospital. Total RNA was isolated using the miRNeasy mini kit (Qiagen). These samples are part of a cohort previously investigated by mRNA29 and lncRNA profiling5, 30. The T-ALL cell lines LOUCY, DND-41, HPB-ALL, ALL-SIL, PF-382 and JURKAT were purchased from DSMZ. KE-37 was a kind gift from the Cools lab.

Thymocyte subset selection

Thymus tissue was derived from children undergoing cardiac surgery (UZ Gent) and was obtained and used with informed consent and according to the guidelines of the Medical Ethical Commission of Ghent University Hospital (Ghent, Belgium). Both thymocyte subsets were each purified from different healthy donors in order to obtain two independent replicates for each subset. Immature CD34+ thymocytes were purified based on MACS purification using CD34 microbeads (Miltenyi Biotec)31 without lineage depletion, while CD4 and CD8 labeling was used to sort the CD4+CD8+ double positive subset by a FACSAriaIII (BDBiosciences)32. The purity of each subset was at least 98%. Total RNA was isolated using the miRNeasy mini kit (Qiagen).

MicroRNA profiling by small RNA-sequencing

The libraries for small RNA-sequencing were prepared using the TruSeq small RNA library kit from Illumina with 50 ng of total RNA as input for T-ALL samples and 100 ng of total RNA as input for T-ALL cell lines and thymocyte subset samples. According to the manufacturer’s use, 3′ and 5′ RNA adapters were ligated to the RNA followed by reverse transcription and PCR amplification (with bar-coded primers). The PCR products were separated using a Pippin Prep System to recover the 147 nt and 157 nt fractions. Sequencing of the small RNA libraries was performed on a NextSeq500 (Illumina), with an average of 14.4 million reads per sample. After read quality control and adapter trimming, reads were mapped to the reference genome (GRCh38) using Bowtie33. Raw data files are submitted into the GEO database34 with accession number GSE89978.

Differential expression analysis

MicroRNA expression data was filtered with a background correction that only retained miRNAs detected by at least 4 reads in at least 60% of samples from one T-ALL subgroup or in at least all samples from one thymocyte subset. Differential expression analysis was performed using the DESeq2 algorithm in R35. The expression was normalized using de Variance Stabilizing Transformation from the DESeq2 algorithm.

Data availability

Reviewers can access the data submitted to the GEO database trough this private link: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=kzszmioallifrmv&acc=GSE89978.