Aberrant expression of microRNA in polycythemia vera

Abstract

Background Polycythemia vera is a clonal hematopoietic stem cell disorder in which the JAK2 V617F mutation is observed in >95% of patients, but an as yet unidentified process appears to initiate the clonal expansion of hematopoiesis. Because microRNA regulate hematopoietic differentiation, we hypothesized that dysregulated expression of microRNA may contribute to the pathophysiology of polycythemia vera.Design and Methods We performed gene expression profiling in five patients with polycythemia vera and in five controls using CombiMatrix MicroRNA CustomArray. ANOVA identified deregulated microRNA in polycythemia vera, and their expression was studied in a larger set of samples by quantitative reverse transcriptase polymerase chain reaction. The expression of these microRNA was also analyzed in other myeloproliferative disorders.Results We observed down-regulation of let-7a and up-regulation of miR-182 in polycythemia vera granulocytes, up-regulation of miR-143, miR-145 and miR-223 in polycythemia vera mononuclear cells, up-regulation of miR-26b in polycythemia vera platelets, and down-regulation of miR-30b, miR-30c and miR-150 in polycythemia vera reticulocytes. JAK2 V617F frequency was positively correlated with miR-143 expression and inversely correlated with let-7a, miR-30c, miR-342 and miR-150. Transcript level of predicted target genes was determined, and overexpression of IRAK2 was detected in all granulocytes from patients with myeloproliferative disorders and in polycythemia vera reticulocytes. Abnormally high HMGA2 microRNA was found in myelofibrosis granulocytes.Conclusions Our study demonstrates that peripheral blood cells from patients with polycythemia vera have microRNA signatures distinct from those of controls. Our findings of aberrant microRNA expression underline the complexity of the molecular basis of polycythemia vera.

Introduction

Polycythemia vera (PV) is a clonal myeloproliferative disorder (MPD) caused by somatic mutation(s) arising in a hematopoietic multipotent cell. Hematopoiesis in PV is characterized by the accumulation of phenotypically normal erythrocytes with variable overproduction of platelets and myeloid leukocytes. A single nucleotide substitution (G1849T) generates the JAK2 V617F tyrosine kinase that is present in the bone marrow and peripheral blood myeloid cells of more than 95% of PV patients and in approximately 50% of patients with essential thrombocythemia (ET) and primary myelofibrosis (PMF). JAK2 V617F has also been identified in a small proportion of patients with chronic neutrophilic leukemia, acute megakaryocytic leukemia, and in most patients with refractory anemia with ringed sideroblasts and thrombocytosis.1,2 The non-specificity of the mutation, the absence of JAK2 V617F in some PV patients, the existence of familial PV in which affected members can be either JAK2 V617F-negative or positive, only a proportion of clonal PV cells are JAK2 V617F-positive, and the existence of erythropoietin-independent burst-forming units-erythroid (BFU-E) that are either JAK2 V617F-negative or positive demonstrate that somatic mutation of JAK2 is not the sole pathogenic process in PV.3–7 Moreover, several mutations in exon 12 of JAK2 have been described in JAK2 V617F-negative PV patients.8,9

MicroRNA (miRNA) are non-coding 18-22nt RNA that regulate gene expression either by destabilizing target mRNA or by inhibiting protein translation.10 Both in vitro and in vivo data show that miRNA are important regulators of hematopoiesis and that they play a role in the pathogenesis of some acquired hematologic disorders,11 functioning either as tumor suppressors (miR-15/16) or as oncogenes (miR-17-92 cluster). For example, in chronic lymphocytic leukemia, somatic deletion of 13q14 is observed in more than 50% of clonal lymphocytes, resulting in loss of expression of miR-15 and miR-16.12 As these miRNA upregulate expression of the anti-apoptotic gene BCL2, mutant cells have a survival advantage.13 Abnormally high expression of miR-155 has been reported in both Hodgkin’s lymphoma and diffuse large B-cell lymphoma.14,15 The precursor sequence of miR-155 is located in the non-coding region of the BIC locus. BIC activation accelerates lymphopoiesis and is associated with overexpression of MYC (V-myc myelocytomatosis viral oncogene homolog).16

Microarray studies have defined miRNA signatures in some hematopoietic cell lineages and hematologic diseases,17–22 and comparison of samples from patients and controls samples revealed aberrantly expressed miRNA that correlated with disease phenotype.23–27 Disease-specific miRNA expression may have diagnostic significance, prognostic significance, or both and may provide new insights into the pathogenesis of disorders such as PV in which the etiology is incompletely understood. To define the miRNA profiles in PV, miRNA expression was analyzed in peripheral blood cells from healthy volunteers and from patients with PV.

Design and Methods

Blood samples

Peripheral blood from patients and controls was obtained following written informed consent using a protocol approved by the Institutional Review Board of the University of Utah. The diagnosis of PV, ET and PMF was made using the criteria of the Polycythemia Vera Study Group, and these clinical criteria were further supported by the following laboratory studies: low serum erythropoietin, identification of erythropoietin-independent colonies, and clonal hematopoiesis in informative females.28 The allele frequency of the JAK2 V617F mutation was quantified in patients’ cells as described elsewhere.29 The PV group consisted of 17 patients, in whom the JAK2 V617F mutation frequency ranged from 0.8% to 100%. The patients were categorized according to JAK2 V617F allele frequency into two groups: those with a low allele burden (patients with a mutation frequency <50%; n=6) and those with a high allele burden (patients with a mutation frequency >50%; n=11). We tested nine patients with PMF (6/9 JAK2 V617F-positive) and nine with ET (4/9 JAK2 V617F-positive). All patients were negative for MPLW515L and MPLW515K mutations.30 The control group was composed of ten volunteer donors.

Platelets were separated from plasma by centrifugation at 500g for 10 min. Mononuclear cells and granulocytes were isolated by Ficoll-Paque (Sigma, St. Louis, MO, USA) density gradient centrifugation.

Gene expression profiling

Total RNA was extracted using Tri-Reagent (MRC, Cincinnati, OH, USA) according to the manufacture’s instructions. Reticulocyte RNA was isolated as described previously.22

CombiMatrix MicroRNA CustomArrays (#3725) were used for gene expression profiling. These arrays contained 326 probes for human mature miRNA including corresponding mismatch oligonucleotides. We used labeling and hybridization protocols as described elsewhere.31 Briefly, 6 μg of total RNA were 3′-end labeled with Cy3 dinucleotides (Dharmacon, Lafayette, CO, USA) using T4 RNA ligase (NEB, Ipswich, MA, USA) by incubating on ice. After 2 hours, the labeled RNA was precipitated and resuspended in 12% formamide, 5% sodium dodecyl sulfate (SDS), 0.8% bovine serum albumin (BSA) and 400 mM Na₂HPO₄ (pH 7.0). The sample was hybridized on array slides overnight at 37°C. The slides were washed by incubation (3 minutes) once in 2xSSC, 0.029% SDS at room temperature, three times in 1.6xSSC at 23°C and twice in 0.8xSSC at 4°C. Processed slides were scanned using a GenePix 4000B (Axon, Sunnyvale, CA, USA) scanner at a resolution of 10 μm.

Microarray data analysis

The raw data were extracted by CombiMatrixImager software (www.combimatrixcorp.com). The median signal from all mismatch probes was set to background. This value was subtracted from all miRNA probe signals. Probe signals greater than 1.5 times background were considered present. By using Genesis 1.6.0Beta1 software (http://genome.tugraz.at/), signal intensity values were normalized to per-chip mean values, and hierarchical clustering analysis was performed using average linkage and Pearson’s correlation. Additionally, the data were processed by one-way ANOVA (Welch analysis of variance) to determine differentially expressed miRNA (a p value of <0.05 was considered statistically significant).

Expression analysis by quantitative real time polymerase chain reaction (qRT-PCR)

Expression of selected miRNA was analyzed by qRT-PCR using TaqMan MicroRNA Expression Assays (Applied Biosystems, Foster City, CA, USA) as previously described.22 Briefly, 10 ng of total RNA were reverse transcribed under the following conditions: 16°C for 30 min, 42°C for 30 min, 85°C for 5 min. The conditions for the PCR reaction were as follows: 95°C for 10 min followed by 40 cycles of 95°C for 15 s and 60°C for 1 min using an ABI PRISM 7000 thermal cycler. The miRNA expression levels were normalized to RNU6B (Applied Biosystems).

The levels of putative target mRNA (MYB: Hs00920571_m1, HMGA2: Hs00171569_m1, CCND2: Hs00922418_g1, HIC2: Hs00740546_s1; IRAK2: Hs00176394_m1, KRAS: Hs00270666_m1) were quantified by TaqMan Expression Assays (Applied Biosystems) according to the manufacturer’s instruction. Briefly, 500 ng of total RNA were reverse transcribed using SuperScript III First-Strand Synthesis SuperMix for qRT-PCR kit (Invitrogen, Carlsbad, CA, USA), under the following incubation conditions: 25°C for 10 min, 50°C for 30 min and 85°C for 5 min. Two microliters of cDNA were used for TaqMan qRT-PCR. The PCR reaction was performed under the following conditions: 95°C for 10 min, followed by 40 cycles of 94°C for 15 s and 60°C for 1 min. Expression levels of the genes were normalized to 18S rRNA (Hs99999901_s1, Applied Biosystems).

Relative fold changes of gene expression were calculated by the ΔΔC_T method and the values are expressed as 2

Prediction of putative miRNA targets

Putative gene targets of miRNA were predicted using the following two algorithm tools: TargetScan 4.0 (www.targetscan.org) and Pictar (http://pictar.bio.nyu.edu/). The target genes predicted in both databases are reported.

Results

miRNA signatures of peripheral blood cell lineages

For initial gene expression profiling by microarrays, we used peripheral blood cells from five PV patients and five controls. Hierarchical clustering analysis determined miRNA expression clusters for particular peripheral blood cell lineages in both controls and PV patients (e. g., cluster –R: down-regulation of miR-23a, miR-23b, miR-342, miR-27a, and miR-150 in reticulocytes; cluster R: up-regulation of miR-185, miR-195, let7-c/b/i, miR-92, miR-363, and miR-182 in reticulocytes; cluster G: up-regulation of miR-338, miR-148a, miR-424, miR-143, and miR-199b in granulocytes; cluster P: up-regulation of miR-199a*, miR-221, and miR-130a in platelets, and down-regulation of miR-451 in mononuclear cells) (Online Supplementary Figure S1).

Validation of miRNA array data by qRT-PCR

The array data of 12 miRNA were validated in the tested cell lineages by qRT-PCR. The relative fold changes of miRNA expression, controls versus PV patients, determined by qRT-PCR were similar to those detected by microarrays (r=0.95, p<0.01) (Online Supplementary Figure S2). Relative gene expression was assessed using both quantitation derived from a standard curve and the ΔΔC_T method. A high concordance (r>0.96, p<0.01) was observed when the two methods were compared.

Differentially expressed miRNA in polycythemia vera peripheral blood cells

The ANOVA of array data showed significantly different expression of 40 miRNA in particular PV cells as compared to in control cells (p<0.05) (Figure 1). Of these, we selected 25 miRNA (p<0.03) for further testing in a larger set of PV patients (n=17) and controls (n=10) by qRT-PCR. We confirmed down-regulation of let-7a (p<0.05) and up-regulation of miR-182 (p<0.01) in PV granulocytes; up-regulation of miR-143 (p<0.01), miR-145 (p<0.01) and miR-223 (p<0.01) in PV mononu-clear cells; up-regulation of miR-26b (p<0.05) in PV platelets; and down-regulation of miR-30b (p<0.05), miR-30c (p<0.05) and miR-150 (p<0.05) in PV reticulocytes (Figure 2). miR-30b and miR-30c belong to the same gene family and their levels of expression were significantly correlated with each other (r=0.96, p<0.01). Array data showed miR-342 down-regulation in PV cell lineages except reticulocytes but we did not confirm this by qRT-PCR at a p value of <0.05.

miRNA expression in cells from patients with primary myelofibrosis and essential thrombocythemia

To investigate PV specificity of deregulated miRNA, we analyzed miRNA expression in peripheral blood cells of patients with PMF (n=9) and ET (n=9). In granulocytes, the down-regulation of let-7a was specific for PV, but abnormally high expression of miR-182 was also found in all other MPD patients (p<0.05). In addition, we detected significantly decreased levels of miR-30c in ET granulocytes (p<0.01). We did not confirm PV specificity of miR-143, miR-145 or miR-223 overexpression in mononuclear cells, as increased levels of those miRNA were also detected in mononulcear cells from patients with PMF and ET (p<0.01). Because of the limited amount of blood available from non-PV MPD patients, we did not attempt to isolate reticulocyte RNA from patients with PMF or ET.

Correlation of JAK2 V617F allele frequency with miRNA expression

To analyze the relationship between miRNA expression and JAK2 V617F allele frequency, PV patients were grouped into two categories according to their mutational status and Spearman’s correlation analysis was applied. The patients with an allelic burden of <50% were categorized into a low allele burden group (n=6) and those with an allele frequency of >50% were categorized into a high allele burden group (n=11). We determined a positive correlation of miR-143 (r=0.62) (p<0.05) in PV mononuclear cells; an inverse correlation of let-7a (r=−0.67)(p<0.01), miR-30c (r=−0.63)(p<0.01), miR-150 (r=−0.74)(p<0.001) and miR-342 (r=−0.63) (p<0.01) in PV granulocytes; and an inverse correlation of miR-150 (r=−0.75)(p<0.01) in PV platelets. In PV reticulocytes, only miR-150 showed a slight trend towards an inverse correlation (r=−0.42)(p>0.05). We also correlated expression levels of these miRNA in PMF and ET. Although we had a limited number of JAK2 V617F-positive PMF and ET patients, we found an inverse correlation of miR-199a (r=−0.77)(p<0.05) and miR-342 (r=−0.82)(p<0.05) in PMF platelets. Generally, we observed the same correlation trend (either positive or negative) in PV and PMF, but in ET, the correlation trend was in the opposite direction for some miRNA (e.g., miR-143, miR-342, miR-27b).

Figure 1.Differently expressed miRNA in PV peripheral blood cells detected by microarrays. The one-way ANOVA of array data revealed significantly deregulated miRNA (p<0.05) in PV mononuclear cells (A), granulocytes (B), platelets (C) and reticulocytes (D). The PV patients are listed according to their JAK2 V617F level, from the lowest level in PV1 to the highest level in PV5. The relative gene expressions are expressed by a gradient intensity of color, as shown in the color scale at the right upper corner of this Figure. The dark blue color indicates low expression and the lightest red color indicates maximal expression. NORM: control; PV: polycythemia vera patient; MNC: mononuclear cells; GRAN: granulocytes; PLAT:platelets; RET: reticulocytes.

Transcript levels of putative target genes

Putative miRNA targets were predicted by TargetScan 4.0 and PicTar software (Table 1), and transcript levels were measured by qRT-PCR (Figure 3). The highest scored target of miR-150 is MYB (V-myb myeloblastosis viral oncogene), but we did not detect a significant difference in MYB mRNA levels between control and MPD granulocytes or between control and PV reticulocytes. Another high scored target of miR-150 is IRAK2 (interleukin-1 receptor-associated kinase 2), and we detected its significant overexpression in all MPD granulocytes (p<0.01) and in PV reticulocytes (p<0.05). The highest scored target of let-7a is HMGA2 (high mobility group AT-hook 2), and we observed significantly increased expression of HMGA2 mRNA in PMF granulocytes (p<0.05) but not in PV and ET granulocytes. CCND2 (cyclin D2) may be regulated by let-7a, miR-182 or miR-145, and we found decreased expression of CCND2 in all MPD granulocytes; however, in PMF and ET, the p values were of borderline significance (0.06 and 0.05, respectively). HIC2 (hypermethylated in cancer 2 gene), a potential target of let-7a, 181a or miR-145, was not expressed aberrantly in MPD granulocytes. As let-7a, miR-143 or miR-150 may target KRAS (v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog), we found a significantly decreased level of KRAS in PMF granulocytes (p<0.05).

Figure 2.Gene expression of selected miRNA in myeloproliferative disorder peripheral blood cells detected by qRT-PCR. Gene expression of the miRNA was determined by qRT-PCR in peripheral blood mononuclear cells (A), granulocytes (B), platelets (C) and reticulocytes (D) from polycythemia vera, primary myelofibrosis and essential thrombocythemia patients. Relative fold changes of expression were calculated by the ΔΔC_T method and the values are expressed as 2^−ΔΔCt. Data are presented as the mean plus standard error. The statistical significance between miRNA expression in controls and patients was calculated by Student’s t-test. Reticulocyte RNA was available only from poly-cythemia vera patients. PV: polycythemia vera, PMF: primary myelofibrosis, ET: essential thrombocythemia, *p<0.05, **p<0.01.

Table 1.Putative target genes of deregulated miRNA in poly-cythemia vera peripheral blood cells. Target genes of the miRNA were predicted by TargetScan 4.0 and PicTar software and the genes predicted in both databases are reported.

Discussion

We determined the miRNA expression profiles of peripheral blood cells from patients with PV and compared these data to those of control samples and samples from patients with PMF and ET (the other two Philadelphia chromosome negative MPD). We identified aberrant expression of eight miRNA (let-7a, miR-182, miR-143, miR-145, miR-223, miR-26b, miR-30b, miR-30c and miR-150) in either granulocytes, mononuclear cells, platelets, or reticulocytes of PV patients. We observed correlations between aberrant expression of miR-143, let-7a, miR-30c, miR-342 and miR-150 and JAK2 V617F mutation frequency. In some cases (miR-30c, miR-150 and miR-342), a statistically significant difference in granulocyte expression was observed only after samples were categorized into low allele burden (<50% JAK2 V617F allelic frequency) and high allele burden (>50% JAK2 V617F allelic frequency) groups. Granulocytes from patients with a high allele burden had significantly lower expression of these miRNA compared to controls, and patients with a low allele burden expressed the miRNA at significantly higher levels than the patients with a high allele burden. For example, a comparison of granulocytes from high allele burden PV patients and those from controls showed a significant difference in miR-150 expression (p=0.02), but the expression was not significantly different between low allele burden PV patients and controls. This demonstrates that JAK2 V617F allelic heterogeneity must be considered in comparative analyses.

Figure 3.Transcript levels of putative targets of deregulated miRNA in myeloproliferative disorder peripheral blood cells. Putative miRNA targets were predicted by TargetScan 4.0 and PicTar software and transcript levels of selected genes were tested by qRT-PCR in myeloproliferative disorder granulocytes and polycythemia vera reticulocytes. Relative fold changes of expression were calculated by the ΔΔC_T method and the values are expressed as 2^−ΔΔCt. Data are presented as the mean plus standard error. The statistical significance between miRNA expression of controls and patients was calculated by Student’s t-test. PV: polycythemia vera, PMF: primary myelofibrosis, ET: essential thrombocythemia, Δp=0.06, *p<0.05, **p<0.01.

In our previous study, we tested miRNA expression in in vitro expanded erythroid progenitors and found that miR-150 was reduced in PV erythroid cells.22 In concordance with these in vitro data, we also found significant down-regulation of miR-150 in native PV reticulocytes. Together, these results suggest that the down-regulation of miR-150 expression plays a role in the pathophysiology of PV. In contrast, up-regulation of miR-150 was reported in chronic lymphocytic leukemia.33

Transcript levels of some miRNA targets were analyzed to determine their potential dysregulation at the mRNA stability level. High scored targets of miR-150 are the MYB and IRAK2 genes. We did not detect aberrant expression of MYB transcript in MPD granulocytes or in PV reticulocytes; however, the level of mRNA IRAK2 was significantly up-regulated in these cells. IRAK2 plays a key role in innate immunity through Toll-like receptor signaling pathways that interface with nuclear factor κB.34 Its role in MPD is unclear and thus represents an interesting target for further studies.

In this study, let-7a down-regulation was specific for PV granulocytes and inversely correlated with the JAK2 V617 allele burden. Previously, the down-regulation of let-7a had been shown in etiologically heterogeneous disorders such as chronic lymphocytic leukemia35 and gastric carcinoma,36 demonstrating its relevance in clonal disorders. The target of let-7a is HMGA2,37 an architectural transcription factor, whose aberrant expression contributes to clonal hematopoiesis in some cases of paroxysmal nocturnal hemoglobinuria.38 Compared to levels in normal controls, we detected a significantly higher HMGA2 mRNA level in PMF granulocytes but not in granulocytes from patients with PV and ET. Similar findings have been reported by others.39 However, expression of HMGA2 mRNA in PMF granulocytes did not correlate with let-7a expression in either PV (r=−0.46) or PMF (r=0.28) granulocytes, suggesting an alternative mechanism for controlling HMGA2 expression. For example, in head and neck squamous cell carcinoma, HMGA2 expression was reportedly controlled by miR-98,40 whereas in Burkitt’s lymphoma, let-7a down-regulated MYC.41

RAS oncogenes can be regulated by let-7.42 We observed decreased expression of KRAS in all MPD granulocytes; however, only in PMF granulocytes did this decrease reach statistical significance. In the PV group, the p value was 0.10, but when the patient with the lowest JAK2 V617F allele frequency (0.8%) was excluded from the analysis, the p value was <0.01, suggesting apparent down-regulation of KRAS in patients with a high allele burden. Deregulation of Ras signaling has been reported in myeloid malignancies through alternative genetic mechanisms that include somatic mutations in NRAS and KRAS.43 Mutations of RAS genes are found occasionally in MPD with NRAS being affected most commonly and KRAS rarely.44 Interestingly, somatic activation of KRAS blocks erythroid differentiation and causes anemia in a mouse model.45 We may speculate that decreased expression of KRAS in PV patients is associated with exaggerated erythropoiesis.

We tested two other genes (HIC2 and CCND2) that are putative targets of deregulated miRNA. CCND2, a positive regulator of G₁ phase promotion of the cell cycle, has been shown to be involved in JAK2 V617F-mediated signaling.46 A borderline low CCND2 transcript level was observed in PMF and ET (p=0.06 and 0.05, respectively).

Although most of the mRNA targets did not show differential expression in the tested MPD cells, this does not rule out their suppression at the translational level, which represents a major mechanism of miRNA-mediated regulation in animals. The functional relevance of the dysregulated targets in the pathogenesis of PV will be studied in detail. These data will need to be compared to those of other heterogeneous acquired and congenital, as well as primary, secondary and Chuvash polycythemias; these studies are in progress.

We analyzed expression of miRNA in ET and PMF cells to determine whether aberrant miRNA expression was specific to PV. This comparison showed that some miRNA are aberrantly expressed in other MPD (e.g., we demonstrated overexpression of miR-143, miR-145 and miR-223 in all MPD mononuclear cells and abnormally high expression of miR-182 in granulocytes of all MPD). In contrast, decreased levels of miR-143 and miR-145 were shown in B-cell malignancies.47 Up-regulated expression of miR-182 has already been reported in PMF granulocytes.48 Overlapping dysregulation of some miRNA is not surprising because miRNA may regulate multiple targets or may co-operate to regulate a particular gene. Our study underscores the complexity of aberrant miRNA expression in clonal premalignant and malignant processes.

To our knowledge, this study is the first comparison of the miRNA signatures of PV peripheral blood cells and normal peripheral blood cells. We show that the expression of some dysregulated miRNA correlates with JAK2 V617F mutation frequency, suggesting a possible effect of JAK2 V617F on miRNA expression. However, we also identified aberrantly expressed miRNA without a correlation with JAK2 V617F mutation frequency. As expected from mRNA profiles,39,49 some miRNA (miR-143, miR-145, and miR-150) are deregulated with expression-specific patterns also in other clonal hematopoietic disorders. Nonetheless, our findings of aberrant miRNA expression support the concept that factors other than constitutive activation of JAK2 may contribute to the pathogenesis of PV. Additional studies will be needed to determine the molecular consequences of aberrant expression of miRNA in the pathophysiology of PV.

Acknowledgments

the authors thank Sabina Swierczek, PhD (Hematology Division, University of Utah) for analyzing JAK2 V617F, MPLW515L and MPLW515K mutations. We wish to thank Charles Parker, MD. (Hematology Division, University of Utah) for helpful critical comments of the manuscript. We also thank Joshua Mendell, MD, PhD (Johns Hopkins University School of Medicine, Baltimore, MD, USA) who made the chip available to us for the initial screening.

Footnotes

• Funding: this study was supported by grants 1P01CA108671-O1A2 (MPD) Consortium project#1; R01HL50077-11 (PI JTP) and NR/9236-3 IGA MZD CR (PI HB).
• The online version of this article contains a supplementary appendix.
• Authorship and Disclosures HB performed the research and wrote the manuscript; MM analyzed microarray data; JTP designed the research and critically revised the manuscript. The authors reported no potential conflicts of interest.

• Received January 3, 2008.
• Revision received February 18, 2008.
• Accepted February 19, 2008.