MicroRNA expression profiles of human leukemias

MicroRNAs (miRNAs) are small noncoding RNAs of 20–24 nucleotides (nt) that negatively regulate the translation of target mRNAs through incomplete base-pairing with their 3′-untranslated regions.1 Evidence indicates that miRNAs play an important role in the development of human cancers including leukemias, with one of the most well-characterized examples being association of miR-15a and miR-16a with chronic lymphocytic leukemia. Almost half of chronic lymphocytic leukemia patients harbor a chromosome deletion that encompasses 13q14, a region that includes the genes for miR-15a and miR-16a, and the abundance of these miRNAs is reduced in chronic lymphocytic leukemia cells with the chromosome deletion.2 Several other miRNAs, such as miR-155 and miR-17-92, have also been implicated in the pathogenesis of lymphoma.3 It is therefore important that the entire miRNA repertoire of clinical specimens be characterized and compared among various hematologic malignancies.

Reliable assessment of the global expression profiles of miRNAs, especially for the small amounts of clinical specimens available, is not straightforward, however. Microarray-based detection of miRNAs is prone to the generation of false-positive data that may result from mishybridization of probes, although improvements have recently been developed for this technology.4 A large-scale cloning strategy would be an ideal approach to reliable estimation of the expression level of miRNAs, provided that a sufficient number of clones were to be analyzed. However, conventional methods for isolation of miRNAs require >10 μg of total RNA, which is not always obtainable from clinical specimens.

We recently developed a sensitive method, mRAP (micro RNA amplification profiling)5 that readily allows the isolation of miRNA clones from 1 × 104 cells. To examine the miRNA expression profiles for leukemias with mRAP, we first purified CD34+ cells from individuals (n=12) with de novo acute myeloid leukemia, acute myeloid leukemia secondary to myelodysplastic syndrome, acute lymphoid leukemia or biphenotypic acute leukemia (Table 1). Column affinity-chromatography to isolate CD34+ cells yielded 10–50% of the input cells with a purity of 90% as judged by flow cytometry (data not shown). As a normal control, we also purified a CD34+ cell fraction from bone marrow mononuclear cells of a healthy volunteer. Then mRAP procedure was applied to 1.1 × 106–1.0 × 108 of the purified CD34+ cells from each individual in order to obtain short RNA clones.

Table 1 Clinical characteristics of the study subjects

Sequencing and computer filtering5 of the mRAP amplicons identified a total of 38 858 qualified reads for the 13 study subjects. BLAST analysis then isolated 32 867 reads that match the human genome sequence (ncbi 36 assembly), among which 2054 reads were mapped to transfer RNA genes, 2720 to ribosomal RNA genes and 9474 to repetitive sequences. From the remaining sequences, we identified 7191 reads corresponding to 143 independent known miRNAs (Supplementary Table 1). We further searched for candidate sequences corresponding to novel miRNAs whose surrounding genome sequences (of 100 nt) potentially fold into a hairpin structure with a single notch. In this analysis, we did not exclude miRNA candidates that were not detected in the genomes of other species, given that some miRNAs are species-specific or have arisen recently during evolution.6

We isolated an unexpectedly large number (n=170) of independent candidates for novel miRNAs among 296 sequence reads (Supplementary Table 1, Supplementary Data). The proportion of reads for such novel candidate miRNAs among all miRNA reads ranged from 1.7 to 9.5% per sample (mean, 4.7%). Of the 170 candidates, 19 were identified in at least two samples, supporting the notion that they are bona fide miRNAs. The surrounding genome sequence for one such candidate (designated Hsj_376) is conserved among human, cow and hedgehog (Figure 1a). Hsj_376 was found in two acute myeloid leukemia samples (corresponding to a total of 52 reads) in our data set and folds into a single hairpin (Figure 1a). In contrast, we obtained only one read for a candidate miRNA (Hsj_41) whose surrounding genome sequence also folds into a single hairpin structure (Figure 1b). However, this read was independently identified in our experiments performed both in Japan and in the Netherlands. The nucleotide sequence of all the miRNA candidates and their flanking sequences are presented in Supplementary Data.

Figure 1
figure1

Nucleotide sequence and expression of novel miRNA candidates. Nucleotide sequences (red) of genes for the predicted novel miRNAs Hsj_376 (a) and Hsj_41 (b) are aligned with genomic sequences of human and other species. Asterisks indicate conserved nucleotides. Possible base-pairing schemes for the respective miRNA precursors are shown in the upper insets. (c) Small-RNA fractions (800 ng per lane) purified from the indicated cell lines with the use of a mirVana RNA isolation kit (Ambion, Austin, TX, USA) were subjected to northern blot analysis with ‘locked’ nucleic acid probes for the candidate miRNAs Hsj_3 or Hsj_117 or for U6 small nuclear RNA (internal control). Hjs_3 has been very recently deposited into the miRBase database as hsa-miR-301b. The positions of 24- and 19-nt size markers are indicated on the left. miRNA, microRNA.

The genomic sequences for some of the candidate miRNAs mapped in the vicinity (20 kbp) of those for other miRNAs in the human genome. For example, the gene for one candidate (Hsj_360) and hsa-miR-560 are present on the long arm of chromosome 2 separated by a distance of 1 kbp (Supplementary Figure 1). In this instance, the genome sequences for the two miRNAs are not conserved in other species, indicative of recent evolution.

Expression of some of the candidate miRNAs was confirmed by northern blot analysis with small RNA fractions isolated from a variety of human cancer cell lines, including KCL22 (chronic myeloid leukemia), HL60 (acute myeloid leukemia), MiaPaCa (pancreatic carcinoma), RKO (colorectal carcinoma), MEC (cholangiocarcinoma), TKKK (intrahepatic bile duct carcinoma), TGBC (gallbladder carcinoma), NOZ (gallbladder carcinoma), LK2 (lung squamous cell carcinoma) and Jurkat (T-cell leukemia) (Figure 1c).

The relative expression profile of miRNAs was then calculated for each sample as shown in Figure 2. Whereas some miRNAs, such as miR-124a, miR-142, miR-143 and miR-146a, were expressed in different types of leukemia, most miRNAs were expressed in a sample-specific manner. For instance, miR-29b was abundant in only two samples (ID nos. 46 and 47), with the reads for this miRNA accounting for <1% of all miRNA reads in each of the other specimens. Similarly, the novel miRNA candidate Hsj_376 was abundant in the same two samples but not in the others. Both hsa-miR-183 and hsa-miR-590 were detected in only single samples (ID nos. 4 10, respectively).

Figure 2
figure2

Expression profiles of miRNAs in CD34+ specimens. The percentage contribution of each miRNA to the total miRNA population was calculated for each study subject. Abundant miRNAs are represented as color-coded, with candidates for novel miRNAs shown in red. The disease type of each individual is also indicated on the right. ALL, acute myeloid leukemia; AML, acute lymphoid leukemia; MDS, myelodysplastic syndrome; miRNA, microRNA.

To examine further the similarities and differences in the miRNA profiles among the study subjects, we performed a hierarchical clustering analysis for the subjects based on the expression patterns of all known and novel miRNAs (Figure 3a). Leukemia specimens with a normal karyotype were clustered in the same branch, indicative of a relative homogeneity of these samples, at least with regard to miRNA expression. Nevertheless, the healthy volunteer was placed in a different branch, suggesting that leukemic blasts with a normal karyotype possess a miRNA profile distinct from that of nonleukemic CD34+ cells with a normal karyotype.

Figure 3
figure3

Hierarchical clustering of the study subjects based on miRNA expression profiles. (a) Subject tree generated by two-way clustering analysis with the expression profiles of all known and novel miRNAs. Each row corresponds to a separate sample, and each column to a miRNA whose expression is color-coded according to the indicated scale. The disease type and karyotype of each subject are shown at the left. (b) Six karyotype-associated miRNAs identified with Student's t-test and a false discovery rate of <0.05 were used for two-way clustering analysis as in (a). ALL, acute myeloid leukemia; AML, acute lymphoid leukemia; BAL, biphenotypic acute leukemia; MDS, myelodysplastic syndrome; miRNA, microRNA.

We further attempted to identify miRNAs whose expression level was significantly linked to blast karyotype. Application of Student's t-test to the miRNA expression data with a Benjamini and Hochberg false discovery rate 7 of <0.05 resulted in the isolation of six miRNAs (hsa-miR-29c, hsa-miR-124a, hsa-miR-150, hsa-miR-183, hsa-miR-382 and hsa-miR-590). Hierarchical clustering of the study subjects based on the expression profiles of these ‘karyotype-associated miRNAs’ revealed that the healthy volunteer was again placed apart from the leukemic patients with a normal karyotype.

In conclusion, application of the mRAP procedure to CD34+ leukemic blasts yielded 7487 reads for potential miRNA clones. We previously showed that mRAP readily allows the isolation of >1 × 106 miRNA concatamers from 1 × 104 cells and is thus suitable for miRNA profiling of clinical specimens.5 Indeed, mRAP functioned well with the small number of purified specimens in the present study, with the result that sequencing capacity, rather than specimen quantity, is likely to be the limiting factor for the size of the final data set in most studies.

Although, in the present study, the total number of sequence reads per sample (average=2989 reads) was not high, we were able to discover a relatively large number (n=170) of novel miRNA candidates from our sequence reads. Candidates for novel miRNAs continue to be identified, making it likely that the total number of human miRNAs has not yet reached saturation.8 Our results show that CD34+ leukemic blasts express a wider range of miRNAs than previously appreciated and that overall miRNA expression profiles generally reflect blast karyotype. Such karyotype-specific miRNAs may play a role in the malignant transformation of blasts of the corresponding karyotype, a possibility that needs to be confirmed by analysis of a large number of samples.

It is possible that some of the miRNA candidates identified in our study are not genuine miRNAs but rather degradation products of RNA or DNA. We believe, however, that a substantial proportion of the candidate miRNAs are indeed novel miRNAs because (i) many of them were identified in different samples in different laboratories (in Japan and in the Netherlands), (ii) many of them (together with the surrounding sequences in the genome) are conserved across various species and (iii) the expression of some of them was confirmed by northern blot analysis.

We have identified 170 novel miRNA candidates in, and demonstrated a high level of diversity in miRNA profiles among, leukemic blasts. Our data thus suggest that the miRNA repertoire of human leukemias has not yet been exhausted, and they should provide a framework for future studies in this regard.

Note added in proof

Hsj_117 and Hsj_360 have the miRBase accession numbers hsa-miR-590 and hsa-miR-663b, respectively.

Accession codes

Accessions

GenBank/EMBL/DDBJ

References

  1. 1

    Bartel DP . MicroRNAs: genomics, biogenesis, mechanism, and function. Cell 2004; 116: 281–297.

  2. 2

    Calin GA, Dumitru CD, Shimizu M, Bichi R, Zupo S, Noch E et al. Frequent deletions and down-regulation of micro-RNA genes miR15 and miR16 at 13q14 in chronic lymphocytic leukemia. Proc Natl Acad Sci USA 2002; 99: 15524–15529.

  3. 3

    He L, Thomson JM, Hemann MT, Hernando-Monge E, Mu D, Goodson S et al. A microRNA polycistron as a potential human oncogene. Nature 2005; 435: 828–833.

  4. 4

    Nelson PT, Baldwin DA, Scearce LM, Oberholtzer JC, Tobias JW, Mourelatos Z . Microarray-based, high-throughput gene expression profiling of microRNAs. Nat Methods 2004; 1: 155–161.

  5. 5

    Takada S, Berezikov E, Yamashita Y, Lagos-Quintana M, Kloosterman WP, Enomoto M et al. Mouse microRNA profiles determined with a new and sensitive cloning method. Nucleic Acids Res 2006; 34: e115.

  6. 6

    Berezikov E, Thuemmler F, van Laake LW, Kondova I, Bontrop R, Cuppen E et al. Diversity of microRNAs in human and chimpanzee brain. Nat Genet 2006; 38: 1375–1377.

  7. 7

    Reiner A, Yekutieli D, Benjamini Y . Identifying differentially expressed genes using false discovery rate controlling procedures. Bioinformatics 2003; 19: 368–375.

  8. 8

    Berezikov E, Guryev V, van de Belt J, Wienholds E, Plasterk RH, Cuppen E . Phylogenetic shadowing and computational identification of human microRNA genes. Cell 2005; 120: 21–24.

Download references

Acknowledgements

This study was supported in part by a grant for Third-Term Comprehensive Control Research for Cancer from the Ministry of Health, Labor, and Welfare of Japan as well as by a grant for Scientific Research on Priority Areas ‘Applied Genomics’ from the Ministry of Education, Culture, Sports, Science and Technology of Japan. The authors declare no competing financial interests.

Author information

Correspondence to H Mano.

Additional information

Supplementary Information accompanies the paper on the Leukemia website (http://www.nature.com/leu)

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Takada, S., Yamashita, Y., Berezikov, E. et al. MicroRNA expression profiles of human leukemias. Leukemia 22, 1274–1278 (2008) doi:10.1038/sj.leu.2405031

Download citation

Further reading