|Home | About | Journals | Submit | Contact Us | Français|
The DNA hypomethylating drug decitabine maintains normal hematopoietic stem cell (HSC) self-renewal but induces terminal differentiation in acute myeloid leukemia (AML) cells. The basis for these contrasting cell-fates, and for selective CpG hypomethylation by decitabine, is poorly understood. Promoter CpGs, with methylation measured by microarray, were classified by the direction of methylation change with normal myeloid maturation. In AML cells, the methylation pattern at maturation-responsive CpG suggested at least partial maturation. Consistent with partial maturation, in gene expression analyses, AML cells expressed high levels of the key lineage-specifying factor CEBPA, but relatively low levels of the key late-differentiation driver CEBPE. In methylation analysis by mass-spectrometry, CEBPE promoter CpG that are usually hypomethylated during granulocyte maturation were significantly hypermethylated in AML cells. Decitabine treatment induced cellular differentiation of AML cells, and the largest methylation decreases were at CpG that are hypomethylated with myeloid maturation, including CEBPE promoter CpG. In contrast, decitabine-treated normal HSC retained immature morphology, and methylation significantly decreased at CpG that are less methylated in immature cells. High expression of lineage-specifying factor and aberrant epigenetic repression of some key late-differentiation genes distinguishes AML cells from normal HSC and could explain the contrasting differentiation and methylation responses to decitabine.
An important goal in myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML) research is to develop treatment that is more selectively detrimental to MDS/AML cells and relatively sparing of normal hematopoietic stem cells (HSC). One such potential therapeutic is the deoxycytidine analogue decitabine, which at low, non-cytotoxic concentrations can deplete the chromatin modifying enzyme DNA methyl-transferase 1 (DNMT1) and hypomethylate DNA CpG (1–5): in a number of studies from different groups, decitabine and histone deacetylase inhibitors (HDACi, another class of drug that inhibit chromatin modifying enzymes) have been shown to maintain or increase normal HSC self-renewal (6–15); in contrast, decitabine and HDACi induce terminal differentiation of AML primary cells and cell lines encompassing the wide morphologic and cytogenetic spectrum of disease (5;16–24). The basis for these contrasting cell fate responses of AML cells and normal HSC to chromatin-relaxing drugs is poorly understood, impeding rational and optimal clinical translation of this mode of therapy. Since a major effect of decitabine is to hypomethylate DNA, pre- and post-decitabine patterns of DNA methylation could provide an insight into underlying mechanisms. However, changes in promoter CpG methylation are also an important and usual aspect of hematopoietic differentiation (25). Therefore, to better understand and interpret DNA methylation patterns before and after treatment with decitabine, promoter CpGs, with methylation measured by microarray and mass spectrometry, were categorized by the direction of methylation change with normal myeloid maturation. The methylation levels of maturation-responsive CpG were then compared in normal, MDS and AML cells. The methylation analyses were complemented by gene expression measurements of key lineage-specifying and late-differentiation transcription factors (TF), which together drive progressive myeloid maturation. These analyses exposed differences in baseline maturation and epigenetic context between AML cells and normal HSC that likely contribute to and explain contrasting cell fate and methylation responses to decitabine.
Informed consent for sample and data collection was obtained according to protocols approved by the Cleveland Clinic Institutional Review Board. Bone marrow aspirates were collected from MDS and AML patients between 2002 and 2007. High risk disease was defined as MDS or AML with ≥5% myeloblasts (n=130), and low risk disease as MDS with <5% myeloblasts (n=27). Clinical annotation of samples used for promoter CpG methylation analysis by microarray was described in detail in a previous publication (26). Normal total bone marrow (NBM) were aspirates from healthy individuals (n=42). Six AML cell lines were studied, of which K-562 and TF1 are described as erythro-leukemia cell lines, and KG1, TF1 and Kasumi-1 express CD34. A model of ‘first-hit’ abnormality (pre-leukemia) was also examined: cord blood CD34+ cells transduced with retrovirus to express the leukemia fusion protein RUNX1-ETO, as previously described and characterized (27).
The Methylation Cancer Panel I and GoldenGate Assay kit with UDG (Illumina, San Diego, CA) was used for microarray methylation analysis of 1505 CpG sites mostly in the 5′-regulatory regions of 807 genes (known oncogenes, tumor suppressor, X-linked and imprinted, DNA repair, cell cycle control, differentiation and apoptosis genes) (28). DNA was bisulfite converted using the EZ DNA methylation kit (Zymo Research, Orange, CA). For each CpG site, there were two pairs of probes corresponding to either the methylated or unmethylated state of the CpG site. Through allele-specific extension and ligation, PCR templates were generated and then amplified by PCR using fluorescently labeled common primers. The resulting PCR products were hybridized to a bead array at sites bearing complementary address sequences. These hybridized targets contained a fluorescent label that denotes a methylated or unmethylated state for a given locus. Methylation status of the interrogated CpG site was then calculated as the ratio of fluorescent signal from one allele relative to the sum of both methylated and unmethylated alleles (β-value).
Our verification of the precision of the array, by analyzing technical replicate samples on 2 separate arrays, and our validation of microarray results with another method for measuring methylation levels, has been published previously (26).
The β-value provides a continuous measure of levels of DNA methylation at a CpG site, ranging from 0 for completely unmethylated sites to 1 in completely methylated sites (28). Three categories of CpG sites were defined (raw methylation data uploaded to GEO Datasets, GSE29047):
The image extraction and statistical analysis were done with Beadstudio and Genomestudio Methylation Module (Illumina). All Illumina recommended control oligos were used. Statistical significance of the difference between mean and median values of methylation level for different samples/groups was assessed using Beadstudio Methylation Module software (Illumina) and SAS V8 (SAS, Cary, N. Carolina). Hierarchical clustering was performed by Euclidean distance metric and centroid linkage (ArrayStar 3, DNAStar).
CEBPE promoter methylation was measured with the Sequenom MassARRAY Compact System (www.sequenom.com/). Gene-specific amplification of bisulfite-treated DNA was followed by in vitro transcription and analysis by matrix-assisted laser desorption ionization time-of-flight (MALDITOF) mass spectrometry (29). DNA (1 μg) was bisulfite converted using EZ DNA Methylation kit (Zymo Research). PCR primers specific for bisulfite converted DNA were designed (Epidesigner Online Tool -http://www.epidesigner.com/). Each reverse primer contained a T7-promoter tag for in vitro transcription (5′-cagtaatacgactcactatagggagaaggct-3′), and the forward primer was tagged with a 10mer to balance Tm (5′-aggaagagag-3′). Primer sequences used for the CEBPε promoter are; CEBPε F (5 ′-aggaagagagTGGGTTTGGTTAAAGTATTTGGTTA) and CEBPε R (5′–cagtaatacgactcactatagggagaaggctAAACTTCTCAAACCTCACAAAAAAA). The genomic coordinates for the CEBPε promoter region analyzed are chr14:23,588,987-14:23,589,309. Mass spectra were acquired using a MassARRAY MALDI-TOF MS (Bruker-Sequenom) and peak detection, signal-to-noise calculations, and quantitation of CpG methylation performed using proprietary EpiTyper software v1.0 (Sequenom). Differentiation-responsive CpG within this region were identified by time-course analyses of normal CD34+ hematopoietic precursors treated with granulocyte-colony stimulating factor (G-CSF). Putative TF binding sites in the analyzed region were identified by TFSEARCHv1.3 (“TFSEARCH: Searching Transcription Factor Binding Sites”, http://www.rwcp.or.jp/papia/)(30).
DNA was purified from cells using the Wizard Genomic DNA Purification Kit (Promega) and DNA (1 μg) was bisulfite converted using EZ DNA Methylation kit (Zymo Research). The PCR primers were 10 pmol of 5- TTTTTTGAGTTAGGTGTGGG-3 and 10 pmol of biotinylated-5-TTGTTTGGTATTTTTTAGTGAGA-3. PCR cycling conditions were; cycle temperature of 94°C for 30 seconds, 60°C for 30 seconds, and 72°C for 30 seconds, for 45 cycles to consume all the biotinylated primers. The biotinylated strand was captured on streptavidin sepharose beads (Amersham Biosciences, Uppsala, Sweden) and annealed with the sequencing primer 5-GGGTGGGAGTGAT-3. The methylation degree of LINE-1 was computed as the average of 4 CpG sites pyrosequenced with the Pyromark Q24 Pyrosequencer (Qiagen).
Three categories of CpG sites were defined: (i) CpG that undergo a significant (p<0.001, t-test) increase in methylation from normal bone marrow CD34+ precursors (NCD34, 95–99% CD34+ immature cells, n=9) to normal total bone marrow (NBM, <2% CD34+ immature cells, n=42) (‘hypomethylated in NCD34’, 108 CpG)(table S1); (ii) CpG that undergo a significant (p<0.001, t-test) decrease in methylation from NCD34 to NBM (‘hypermethylated in NCD34’, 162 CpG)(table S1); (iii) CpG that do not undergo a statistically significant change in methylation status between NCD34 and NBM (‘no methylation change’, 1236 CpG). In gene ontology analyses, gene expression in blood, leukocyte, neutrophil, platelet, leukemia, liver and spleen was significantly associated with maturation-responsive CpG (hypo- and hypermethylated in normal CD34) compared to ‘no methylation change’ CpG (table S2). In a comparison of pathway associations, hematopoietic pathways (eg., ‘hematopoietic cell lineage’) were the pathways most frequently associated with maturation-responsive CpG; this was not the case with ‘no methylation change’ CpG (table S3). In unbiased hierarchical cluster analysis of methylation data from an independent study (31), clusters generated using maturation-responsive CpG discriminated best between NCD34 and NBM, despite 7 to 8-fold more CpG sites in the ‘no methylation change’ category (figure S1).
CpG sites that became more methylated with normal myeloid maturation (hypomethylated in NCD34) were even more methylated in bone marrow cells from patients with low-risk MDS (n=27) and high-risk MDS/AML (≥5% myeloblasts, n=130) (figure 1, red box plots). CpG sites that became less methylated with normal myeloid maturation (hypermethylated in NCD34) were less methylated in low-risk MDS and high-risk MDS/AML bone marrow (figure 1, blue box plots). CpG sites that do not undergo significant methylation changes with normal myeloid maturation (‘no methylation change’) were individually selected based on no significant difference in methylation between NCD34 (n=9) and NBM (n=42). However, when combined for analysis, these 1236 CpG sites were significantly more methylated in NBM compared to NCD34, although the increase was small in magnitude (figure 1, grey box plots). These CpG sites were also significantly more methylated in MDS and AML bone marrow compared to NCD34, however, the increase was substantially smaller than the methylation shifts in maturation-responsive CpG (figure 1, grey box plots).
To strengthen these observations, the analyses were repeated in a dataset of promoter CpG methylation generated by other investigators (31). Ninety-five CpG sites from ‘Hypomethylated in NCD34’, 157 CpG sites from ‘Hypermethylated in NCD34’ and 1157 CpG sites from ‘no methylation change’ were present in the independent dataset. Methylation of maturation-responsive CpG (hypo and hypermethylated in NCD34) changed in the expected directions from independent NCD34 to NBM (figure S2, red and blue box plots). Although ‘no methylation change’ were significantly more methylated in NBM compared to NCD34, the methylation increase was smaller in magnitude than the methylation changes at maturation-responsive CpG (figure S2, grey box plots). CpG sites that become more methylated with normal myeloid maturation were not significantly more methylated in independent dataset MDS/myeloproliferative disease (MDS/MPD) (n=13) or AML bone marrow (n=116) (figure S2). Nonetheless, of the 25 CpG sites that were >20% more methylated in AML cells than NCD34, 11 were from the category of 95 CpG sites that become more methylated with normal myeloid maturation and 14 were from the 1157 ‘no change in methylation’ CpG sites (figure S3). Therefore, maturation-responsive CpG sites were over-represented amongst the most hypermethylated CpG sites in AML bone marrow (p<0.0001, Chi-Square test). As expected, CpG sites that become less methylated with normal myeloid maturation were significantly less methylated in the MDS/MPD and AML cells (figure S2). All 6 CpG sites that were >20% less methylated in AML cells than NCD34 were from the category of 157 CpG sites that become less methylated with normal myeloid maturation (figure S3). ‘No methylation change’ CpG were significantly more methylated in MDS/MPD and AML (figure S2).
To extend the analyses to more homogenous populations of AML cells, including CD34+ AML cells, promoter CpG methylation, with CpG sites classified as above, was analyzed in a model of first-hit or pre-leukemia (CD34+ cells retrovirally transduced to express the leukemia fusion protein RUNX1-ETO [CD34 RUNX1-ETO]; the phenotype of these cells has been previously described (27)), and six AML cell lines, including the CD34+ cell lines KG1, TF1 and Kasumi-1. The control sample was normal CD34+ cells isolated from cord-blood (NCD34).
CpG sites that become more methylated with normal myeloid maturation were significantly more methylated in three of the AML cell lines compared to NCD34 (figure 2, red box plots). CpG sites that become less methylated with normal myeloid maturation were significantly less methylated in five of the cell lines compared to NCD34 (figure 2, blue box plots). ‘No methylation change’ CpG were significantly hypermethylated in six of the cell lines compared to NCD34 (figure 2, grey box plots), however, the increase in methylation was substantially smaller in magnitude than the changes in methylation in the categories of myeloid maturation-responsive CpG (figure 2). Therefore, the AML cell lines recapitulated the methylation pattern observed in primary MDS and AML cells.
CD34+ normal, pre-leukemia (CD34 RUNX1-ETO) and leukemia cells (Kasumi-1) were treated with identical concentrations of decitabine, to compare the cell fate and promoter CpG methylation response.
Normal CD34+ cells treated with decitabine 0.5μM continued to proliferate exponentially, although cell counts were lower than in vehicle treated control. In contrast, CD34+ RUNX1-ETO and CD34+ Kasumi-1 cells treated with decitabine 0.5μM decreased in cell numbers (figure 3A). A higher concentration of decitabine (1μM) decreased normal CD34+, RUNX1-ETO and Kasumi-1 cell numbers. Cytarabine (AraC) is another clinically utilized cytosine analogue which is the mainstay of AML chemotherapy. Unlike decitabine 0.5μM, AraC 0.5μM decreased cell numbers of normal CD34+ as well as RUNX1-ETO and Kasumi-1 (figure 3A).
Although decitabine 0.5μM decreased RUNX1-ETO and Kasumi-1 cell numbers, this concentration of decitabine did not cause early apoptosis (measured by annexin/7AAD-staining 24 hours after drug exposure)(figure 3B, figure S4, data not shown). Decitabine at 1μM caused early apoptosis (figure 3B, figure S4, data not shown), although not to the same extent as AraC 0.5μM (figure 3B, figure S4).
On day 9 of liquid culture, vehicle treated normal CD34+ cells showed changes of early myeloid differentiation (decreased nuclear-cytoplasmic ratio, increase in pale blue cytoplasm, vacuolization), however, in the decitabine-treated wells, immature morphology was retained (high nuclear-cytoplasmic ratio, fine chromatin) (figure 3C). In contrast, RUNX1-ETO cells treated with decitabine 0.5μM exhibited extensive myeloid differentiation (decreased nuclear cytoplasmic ratio, nuclear condensation and segmentation, cytoplasmic granulation and vacuolation), whereas vehicle treated cells demonstrated immature morphology (figure 3C). Similarly, decitabine 0.5μM induced morphologic changes of differentiation in Kasumi-1 (nuclear segmentation and lobulation, increase in cell-size, and cytoplasmic vacuolization), although these changes were not as prominent as those noted in decitabine treated RUNX1-ETO cells (figure 3C).
Colony formation in semi-solid media is an assay for stem cells and progenitors. Larger colony size and/or mixed character of the colony (more than one cell lineage) suggest greater immaturity of the colony forming cell. Normal CD34+ cells treated with decitabine 0.5μM produced fewer colonies than vehicle treated normal cells, however, the colonies formed were larger and mixed. Furthermore, normal CD34+ cells treated with decitabine 0.5μM retained colony-forming ability after extended liquid culture (14 and 21 days), while normal CD34+ cells cultured in the identical conditions without decitabine had rapidly diminished colony forming ability (figure 3D). RUNX1-ETO and Kasumi-1 cells treated with decitabine 0.5μM lost their otherwise vigorous colony-forming ability (figure 3D, figure S4).
Normal CD34+ cells treated with decitabine 0.5μM demonstrated significant decreases in methylation in both categories of myeloid maturation-responsive CpG (hypo- and hypermethylated in NCD34) (figure 4, red and blue box plots). However, CpG that are not responsive to normal myeloid maturation (‘no methylation change’) were not significantly hypomethylated by decitabine (figure 4, grey box plots). In RUNX1-ETO and Kasumi-1 cells treated with decitabine 0.5μM, consistent with the morphologic response, the largest and statistically significant decreases in methylation were at CpG that become less methylated with normal myeloid maturation (‘hypermethylated in NCD34’, figure 4, blue box plots). Although decitabine also decreased methylation at CpG that become more methylated with normal myeloid maturation (‘hypomethylated in NCD34’), this decrease was smaller in magnitude and not statistically significant (figure 4, red box plots). CpG that are not responsive to normal myeloid maturation (‘no change in methylation’) were also significantly hypomethylated by decitabine (figure 4, grey box plots), however, the methylation shift was substantially smaller than the methylation change at ‘hypermethylated in NCD34’ (figure 4).
Primary AML samples (n=5) and the erythro-megakaryoblastic AML cell lines UT7 and K562 were also treated with decitabine (the classification and chromosome abnormalities in the primary cells are listed in table S4, the cell lines contain complex chromosome abnormalities). Compared to vehicle treatment, decitabine 0.5μM on day 1, 4 and 7 decreased cell numbers accompanied by cellular changes of differentiation (decreased nuclear-cytoplasmic ratio, nuclear condensation, segmentation or lobulation, cytoplasmic granulation or vacuolization) (figure 5).
Lineage commitment and maturation is driven by and absolutely requires key DNA-binding TF (32–34). Therefore, differentiation of AML cells by decitabine suggests high baseline and/or induced expression of key myeloid differentiation-driving TF.
Using QRT-PCR, the levels of the key myeloid lineage-specifying TF CEBPA (CEBPα), which is essential for producing granulocytes, and the lineage-specifying TF PU.1 (SPI1), which is essential for producing monocytes and B-cells, (32) were measured in bone marrow from healthy controls (n=8), low risk MDS (n=4), and high risk MDS/AML (average myeloblasts 40%, n=12) (MDS and AML sub-types listed in table S5). Compared to normal CD34+ and total bone marrow, both MDS and AML bone marrow cells expressed significantly higher levels of CEBPA, and a trend towards higher PU.1 (p=0.026 and p=0.06 respectively) (figure 6A). To restrict the comparison to cells with a similar precursor surface-phenotype, CD34+ cells were isolated from AML (n=3) and normal donor bone marrow (n=3) (AML sub-types listed in table S6). Compared to CD34+ normal cells, CD34+ AML cells expressed 10 to 100-fold higher CEBPA (figure 6B). CEBPA levels in the AML cells were 2 to 30-fold higher than HOXB4 levels in the same cells (figure 6B). These observations were extended and recapitulated in analyses of public gene expression data (figure S5)(35–40).
Since AML cells express high levels of CEBPA, levels of CEBPE (CEBPε), a key late-differentiation TF necessary for progressive maturation, and a gene target of CEBPA, were measured (normally, CEBPE levels peak relative to CEBPA in the transition from proliferating pro-myelocytes to non-proliferating myelocytes (33;34;41)). CD34+ AML cells expressed 2 to 10-fold lower CEBPE levels than CD34+ normal cells, despite expressing substantially higher levels of CEBPA (figure 6B). These observations were extended and recapitulated in analyses of public gene expression data (figure S5)(35–40).
The CEBPE locus on chromosome 14 is not cytogenetically deleted in AML, suggesting the repression may be by epigenetic means. Using mass spectrometry, CEBPE promoter CpG that become less methylated during G-CSF-induced differentiation of normal CD34+ precursors into granulocytes were identified (figure 7A). Notably, the three identified CEBPE promoter CpG were in proximity to putative CEBPA and RUNX1 binding sites (figure S6). The methylation levels of these CpG sites were compared in bone marrow samples from normal volunteers, AML in remission, AML, and promyelocytic leukemia (cytogenetic abnormalities in the AML samples are listed in table S7). Two of the three CpG sites were significantly hypermethylated in AML compared to normal and/or remission bone marrow (figure 7B), but were not hypermethylated in promyelocytic leukemia cells, which express higher levels of CEBPE than other AML sub-types (figure 7B, figure S7). In contrast to the CEBPE promoter CpG, methylation levels at LINE-1 repetitive DNA element CpGs were similar in normal, remission, and AML bone marrow (figure 7C). The effect of decitabine treatment on CEBPE promoter and LINE-1 CpG methylation was examined in an AML cell line (THP1). Decitabine 0.5μM decreased CEBPE promoter CpG methylation by a much greater extent (>40–60%) than the ~20% decrease at LINE-1 CpG (figure 7D, figure S8). Decitabine-induced CEBPE promoter hypomethylation was accompanied by a substantial increase in CEBPE levels (figure 7D).
Promoter CpG methylation patterns reflect differentiation context and stage (25). In primary MDS and AML cells, the pattern of methylation at maturation-responsive promoter CpGs mimicked the pattern in mature myeloid elements, and was the opposite of that in normal CD34+ precursors. In high risk MDS/AML samples, which contain more myeloblasts, and in uniformly morphologically immature CD34+ and CD34− AML cell lines, this pattern was emphasized. These observations suggest that MDS and AML cells are in at least some aspects maturation progressed from normal CD34+ precursors. Accordingly, MDS and AML cells express relatively high levels of key lineage-specifying TF CEBPA and PU.1 (SPI1), compared to normal CD34+ precursors or total bone marrow, and AML leukemia initiating cells frequently have surface-phenotype features of lineage-commitment (CD34+38+, CLL-1+, CD71+, CD90−, c-Kit-)(42–52). Despite high CEBPA expression, expression of CEBPE, a key downstream gene target of CEBPA, was relatively repressed in AML cells, accompanied by hypermethylation of differentiation-responsive CpG in the CEBPE promoter. Hypermethylation has also been reported for CpG in the promoter of CEBPD (CEBPδ), another late-differentiation gene target of CEBPA (53).
Why are key late-differentiation genes such as CEBPE and CEBPD epigenetically repressed in AML cells, despite high expression of lineage-specifying TF that should activate these genes? Genetic abnormalities in genes for lineage-specifying TF (eg., CEBPA) or their cofactors (eg., RUNX1) that compromise transactivation by these differentiation-drivers could have a role: in mice engineered to express mutated Cebpa (p30) with reduced transactivating capacity, the leukemia initiating cells that arose were lineage-committed with impaired expression of key late-differentiation genes including Cebpd (43). Similarly, Runx1 haploinsufficiency, a common abnormality in MDS and AML, impairs the usual cooperative gene activation by Runx1 and the differentiation driver Pu.1/Spi1, producing increased corepressor recruitment to Pu.1 and epigenetic repression of late-differentiation target genes (eg., Gm-csfr and Mcsfr) (54). Mutations in some chromatin modifying enzyme genes (e.g., ASXL1, EZH2) increase in frequency from MDS to AML (55–57). Possibly these genetic abnormalities additionally favor corepressor over coactivator recruitment at late-differentiation genes. DNMT1 depletion by decitabine, which not only hypomethylates DNA but disrupts DNMT1 containing histone modifying protein complexes involved in transcription repression, can potentially redress such corepressor/coactivator imbalance, and thereby trigger late-differentiation gene expression in these cells already expressing high levels of lineage-specifying TF. In contrast, in normal HSC which do not express high levels of lineage-specifying TF, DNMT1 depletion prevents stem cell gene repression by differentiation stimuli, thereby maintaining stem cell phenotype (6–15).
An important goal in MDS and AML research is to identify differences between malignant cells and normal HSC that can be exploited for therapy. In MDS/AML cells, the pattern of expression of key myeloid differentiation-driving TF, and methylation patterns at myeloid maturation-responsive CpG, suggests differentiation is impaired after lineage-commitment, mediated by aberrant epigenetic repression of some key late-differentiation driver genes. This maturation and epigenetic profile, distinct from that of normal HSC, likely plays a major role in the contrasting differentiation response of AML cells and normal HSC to decitabine and other chromatin-relaxing drugs.
We gratefully acknowledge the assistance of Dr Ying Jiang for bio-informatics analysis and sample preparation, Dr Pieter Faber and the Cleveland Clinic Genomics Core for processing of samples, and the following gifts: Mary Laughlin and Nick Greco at the Abraham J and Phyllis Katz Cord Blood Foundation and Cleveland Cord Blood Center for cord blood samples. YS is supported by NIH (1R01CA138858, U54HL090513), Dept. of Defense (PR081404). KPN is supported by Scott Hamilton CARES Foundation.