LIN28B is over-expressed in specific subtypes of pediatric leukemia and regulates lncRNA H19

LIN28B is an RNA-binding protein with an oncofetal expression pattern. High LIN28B expression is crucial during human embryogenesis and is down-regulated in most tissues after birth.1 However, reactivation during oncogenesis is common in a plethora of adult cancers, including leukemia, and was also detected in solid pediatric tumors such as neuroblastoma, Wilms tumor, pediatric hepatoblastoma and hepatocellular carcinoma, germ cell tumors and rhabdoid tumors.2 More recently, we identified high LIN28B expression in a large subgroup of juvenile myelomonocytic leukemia (JMML), an aggressive childhood hematopoietic neoplasm.3

To further evaluate putative oncogenic activity of LIN28B in a broad panel of pediatric hematologic malignancies, we conducted a meta-analysis of LIN28B expression in different types of childhood leukemia using publicly available Affymetrix Human Genome U133 Plus 2.0 microarrays. For this, we investigated gene expression profiles from 1361 childhood leukemia patients in 14 independent studies: GSE71454,3 GSE29326,4 GSE19475,5 GSE47051,6 GSE20910,7 GSE13351,8 GSE11877,9 GSE32962,10 GSE68720,11 GSE14062,12 GSE17855,13 GSE35784,14 GSE1957715 and GSE52891.16 All available Affymetrix data were renormalized using RMA in R Bioconductor and, given the absence of LIN28B expression after birth, LIN28B was considered over-expressed if it exceeded the average expression of the array (Figure 1A, visualized using SinaPlot17). In contrast to JMML, LIN28B is not over-expressed in the evaluated pediatric myelodysplastic syndrome (MDS) cohort and, in general, LIN28B is more frequently over-expressed in pediatric acute myeloid leukemia (AML, 10.7%) compared to acute lymphoblastic leukemia samples (ALL, 2%). In a similar meta-analysis for LIN28A, only one patient with LIN28A overexpression was detected in the whole cohort, suggesting distinct functions for both human LIN28 homologs (Online Supplementary Figure S1A).

Figure 1.(A) Log2 LIN28B expression values in 14 publicly available pediatric leukemia datasets,^3–16 visualized using SinaPlot. Each dot is a patient and red dots represent values above the microarray average. The numbers on top represent the percentage of patients with LIN28B expression higher than the average of the study. (B) Relative growth compared to the day of cell seeding of control (gray) and LIN28B knockdown K562 cells (green). Lum: luminescence. (C) Differential expressed genes between K562 LIN28B knockdown and control cells, Oci-AML3 LIN28B over-expressing and control cells, and juvenile myelomonocytic leukemia (JMML) patients with high and low LIN28B expression. Venn diagram was generated using http://bioin-formatics.psb.ugent.be/webtools/Venn/ (D) RT-qPCR expression of LIN28B and H19 in K562 control (gray) and LIN28B knockdown cells (green). (E) RT-qPCR expression of LIN28B and H19 in Oci-AML3 control (gray) and LIN28B over-expressing cells (orange).

Despite the fact that increasing evidence points to the importance of LIN28B in malignant hematopoiesis, including the observation that LIN28B is able to reprogram adult hematopoietic progenitors to mediate fetal-like lymphopoiesis,18 the role of LIN28B in leukemic development is far from understood. Therefore, we altered LIN28B expression in normal and malignant hematopoietic cells. First, we retrovirally modulated LIN28B expression in two human myeloid leukemia cell lines with different LIN28B levels (Online Supplementary Methods). Stable shRNA mediated knockdown was achieved in K562, a cell line with very high intrinsic LIN28B levels that originates from a chronic myeloid leukemia patient in blast crisis (Online Supplementary Figure S1B and D). In addition, we stably over-expressed LIN28B in the acute myeloid leukemia cell line Oci-AML3, which lacks endogenous LIN28B expression (Online Supplementary Figure S1C and D). In line with previous reports,19 loss of LIN28B decreased leukemic cell growth in K562 cells (Figure 1B and Online Supplementary Methods). However, growth rates were comparable between Oci-AML3 cells with and without LIN28B expression (Online Supplementary Figure S2A).

To study the transcriptional consequences of LIN28B modulation in the context of human leukemia, we profiled both stable cell line models and their respective controls using Agilent gene expression microarrays containing probes for 27958 mRNA transcripts and 23042 long non-coding RNAs (GSE79981).20 In addition, expression profiles from 44 JMML patients were previously generated using the same platform (GSE71449). Differential gene expression (GenePattern;21 raw P-values<0.05) revealed 5904 differentially expressed probes between the LIN28B knockdown K562 cells and control cells, 4885 probes between the LIN28B over-expressing Oci-AML3 cells and control cells, and 4271 probes between LIN28B-high and LIN28B-low JMML patients. Interestingly, the intersection of these three analyses contained 37 probes, all of which correlated with LIN28B expression (Figure 1C). However, and most notably, a large proportion of probes within this intersection hybridized to lncRNA H19 (Online Supplementary Table S1 and Figure S2B). H19 regulation by LIN28B was validated by RT-qPCR in both cell lines, with the exception of one outlier (Figure 1D and E and Online Supplementary Methods). In addition, we confirmed the previously established reciprocal relationship between LIN28B and miRNA let-7 expression in these in vitro models (Online Supplementary Figure S2C and D and Online Supplementary Methods). We did not observe a correlation between LIN28B and H19 expression in 82 publicly available leukemia and lymphoma cell line data (Spearman’s rho=0.0929, P=0.406) (Online Supplementary Figure S3A).22

Although four potential let-7 binding sites have been shown previously identified in the human H19 sequence23 and H19 has been shown to be a target of let-7 in muscle cells,24 there has been no report of a direct regulation of H19 by LIN28B. Remarkably, H19 was described as the uttermost differentially expressed gene between fetal and adult murine hematopoietic stem cells, followed by Lin28b.25 In addition, H19 is known as an oncofetal gene with high expression before birth and has a critical role in tumor growth and metastasis.26 Recently, Zhou et al. evaluated the transcriptional effects of Lin28b expression in adult mouse bone marrow as well as let-7 activation in murine fetal liver hematopoietic cells.27 Notably, reanalysis of the data revealed strong H19 activation upon reintroduction of Lin28b in the adult bone marrow (Online Supplementary Figure S3B). In contrast, let-7 induction in the fetal liver triggered a profound decrease in H19 and Lin28b expression (Online Supplementary Figure S3C).

Given these observations, we wondered whether the regulation of H19 by LIN28B in hematopoietic cells could be driven, at least in part, by the inhibitory effect of LIN28B protein on the let-7 family of microRNAs (Online Supplementary Figure S4A). To this end, we introduced a let-7 mimic cocktail in fluorescence activated cell sorted (FACS) stable retrovirally transduced LIN28B-over-expressing Oci-AML3 cells and evaluated LIN28B and H19 expression after 48 h (Online Supplementary Methods). Given that the ectopically expressed LIN28B has no 3′UTR, we hypothesized that every effect of let-7 on H19 should be LIN28B-independent and not the result of diminished direct binding of LIN28B to H19. As expected, let-7 mimics induced a significant reduction of H19 expression (Online Supplementary Figure S4B), whereas LIN28B levels were not significantly affected. Nevertheless, there was a trend towards lower LIN28B expression in let-7 electroporated cells, suggesting the presence of putative let-7 binding sites in the protein-coding region of LIN28B (Online Supplementary Figure S4C). In contrast, introducing a let-7 mimic cocktail in the empty control K562 cell line significantly down-regulated LIN28B (with 3’UTR), but not H19 (Online Supplementary Figure S4D and E). All experiments were replicated with the puromycin-selected cells and yielded the same results (data not shown). These results indicate that investigating the role of let-7 is hampered by the double negative feedback interaction between LIN28B and H19 and that other regulatory mechanisms, such as direct binding of LIN28B to H19, could be involved in this complex regulatory network, a notion that warrants further investigation. In the light of our previous study, we analyzed the expression of 768 miRNAs, including the H19-encoded miR-675, in 21 JMML patients using an RT-qPCR-based platform.3 Although half of the patients showed enhanced LIN28B and H19 levels, miR-675 was not expressed in any of the samples.

To further evaluate the association between LIN28B and H19 in non-malignant cells, we investigated two replicates of G4 ROSALUC mouse embryonic stem cells (G4 ESCs) and generated two independent ESC clones with ectopic ROSA26 driven Lin28b expression (R26-Lin28b ESCs) (Online Supplementary Methods).28 After G418 selection, RT-qPCR showed significant upregulation of murine H19 in Lin28b ESCs as compared to G4 ESC controls (Figure 2A). These observations extend the regulation of H19 by Lin28b to a murine, non-malignant model system. Subsequently, the four ESC clones described above were differentiated into embryoid bodies for ten days and seeded on methylcellulose plates for fourteen days. All colonies were harvested and subjected to a 12-color FACS-analysis. We measured 13 different cell surface markers associated with distinct hematopoietic populations (CD45, CD3, CD19, NK1.1, CD115, Ly6C, CD11b, Ly6G, MHCII, CD170, CD11c, Ter119 and FceR1a) (Online Supplementary Methods). Lin28b overexpression clearly affected FceR1a receptor levels, with declining FceR1a upon Lin28b activation (Figure 2B). Interestingly, these observations are in line with a recent report from Wang et al. in which bone marrow-derived mast cell cultures and common myeloid progenitors from induced iLIN28B mice were characterized by reduced levels of Fcer1a.29

Figure 2.(A) Lin28b and H19 expression in G4 embryonic stem cells (ESCs) (gray) and Lin28b ESCs (blue). (B) FACS analysis of G4 and Lin28b DIV10 embryoid bodies that were grown on methylcellulose plates for 14 days. (Top) CD11c is plotted versus FceR1a in CD45⁺ cells. The bottom graph represents the distribution of the FceR1a receptor. (C) RT-qPCR results showing Lin28b and H19 expression in Vav-Lin28b mice. Mice without Lin28b expression (Vav^+/+Lin28b^+/+ and Vav^+/+Lin28b^tg/+) are indicated in gray; mice with Lin28b expression (Vav^tg/+ Lin28b^tg/+) in brown. Tg: transgene. *P-value for Lin28b could not be calculated in qBasePlus due to complete absence of gene expression in the left mouse.

Finally, we used a previously established conditional ROSA26 Lin28b knock-in mouse model30 to provide independent confirmation of the putative association between Lin28b and H19 in murine hematopoietic cells. Enforced Lin28b expression in the hematopoietic compartment of adult mice using Vav-iCre resulted in increased levels of murine H19 when compared to littermate controls, thereby confirming the regulation of H19 by Lin28b (Figure 2C).

In conclusion, we show that lncRNA H19, one of the first discovered long non-coding RNAs, is regulated by LIN28B. H19 has been shown to alter DNA methylation genome wide and associates with EZH2, a member of the Polycomb Respressive Complex 2 (PRC2), which was recently identified to be mutated in JMML. LIN28B-driven deregulated H19 expression might provide yet another mechanism driving aberrant methylation in JMML and other leukemias. The discovery of LIN28B as a regulator of H19, known to be involved in hematopoiesis, embryonic development and oncogenesis, can help to clarify the function of this intriguing lncRNA.