|Home | About | Journals | Submit | Contact Us | Français|
The timing and progression of DNA methylation changes during carcinogenesis are not completely understood. To develop a timeline of aberrant DNA methylation events during malignant transformation, we analyzed genome-wide DNA methylation patterns in an isogenic human mammary epithelial cell (HMEC) culture model of transformation. To acquire immortality and malignancy, the cultured finite lifespan HMEC must overcome two distinct proliferation barriers. The first barrier, stasis, is mediated by the retinoblastoma protein and can be overcome by loss of p16 INK4A expression. HMEC that escape stasis and continue to proliferate become genomically unstable before encountering a second more stringent proliferation barrier, telomere dysfunction due to telomere attrition. Rare cells that acquire telomerase expression may escape this barrier, become immortal, and develop further malignant properties. Our analysis of HMEC transitioning from finite lifespan to malignantly transformed showed that aberrant DNA methylation changes occur in a stepwise fashion early in the transformation process. The first aberrant DNA methylation step coincides with overcoming stasis, and results in few to hundreds of changes, depending upon how stasis was overcome. A second step coincides with immortalization, and results in hundreds of additional DNA methylation changes, regardless of the immortalization pathway. A majority of these DNA methylation changes are also found in malignant breast cancer cells. These results show that large-scale epigenetic remodeling occurs in the earliest steps of mammary carcinogenesis, temporally links DNA methylation changes and overcoming cellular proliferation barriers, and provides a bank of potential epigenetic biomarkers that may prove useful in breast cancer risk assessment.
Epigenetic dysfunction is a common and perhaps universal feature of human cancer. One facet of the epigenetic system that shows clear and dramatic differences between normal and cancer cells is DNA methylation. DNA hypermethylation generally occurs in CpG island gene promoters and is linked to transcriptional gene silencing, while DNA hypomethylation often occurs in repetitive elements, pericentromeric regions, and within the body of genes and has been linked to gene activation and chromosome instability. Studies into the biological consequences of hypo- and hypermethylation indicate that each plays a participative role in carcinogenesis (1, 2). While significant progress has been made in identifying aberrant methylation events that occur during carcinogenesis, the timing, origins, and consequences of these events remain incompletely understood.
We have used an HMEC model system that allows for assessment of changes in DNA methylation that occur at the earliest stages of multi-step human breast carcinogenesis. In this model, the transformation of normal finite lifespan HMEC to malignancy requires overcoming two distinct senescence-associated barriers to immortality (3). The first barrier, termed stasis or stress-induced senescence, is characterized by elevated levels of the cyclin-dependent kinase inhibitor p16INK4A (gene CDKN2A), which maintains the retinoblastoma protein in an active state (4). This barrier has been overcome or bypassed in cultured HMEC by various means, such as exposure to the chemical carcinogen benzo(a)pyrene, with the resultant post-stasis cells commonly exhibiting inactivation of CDKN2A by promoter hypermethylation or by gene mutation (4, 5). Loss of p16 expression due to silencing or mutation is also a frequent event during in vivo human breast cell transformation (6, 7). When grown in a serum-free medium, rare HMEC will “spontaneously” silence p16, generating a type of post-stasis HMEC population that has been called post-selection (4, 8). HMECs that have escaped stasis undergo further proliferation before encountering a second more stringent proliferation barrier resulting from critically shortened telomeres (3, 9). When approaching the telomere dysfunction barrier, HMEC exhibit increased chromosomal instability and a DNA damage response. Rare cells that gain telomerase expression may escape this barrier, and acquire immortal potential; additional perturbations can confer malignant properties on the immortally transformed cells (5, 10-12).
This HMEC system has proven useful for identifying and reflecting the molecular events involved in early human breast tumorigenesis (3-5, 9, 13-15). For example, the hypermethylation of CDKN2A seen in post-stasis HMEC has also been documented in precancerous lesions and histologically normal breast tissue (15, 16). Currently it is not clear if focal DNA methylation changes, such as in p16, are rare events that accumulate independently and gradually over time or if these are examples of larger sets of DNA methylation changes that occur in groups - concurrently and at distinct points in the immortalization process. Genomic instability and telomere erosion, characteristic of pre-malignant, in situ breast lesions, is seen in the cultured HMEC at the telomere dysfunction barrier (3, 9, 13, 17). The potential contribution of epigenetic changes to telomerase reactivation and immortality is currently not known. To begin to address these questions we examined the DNA methylation profiles of cultured normal finite lifespan HMEC, isogenic derivatives induced to escape the stasis and/or telomere dysfunction barriers, established breast cancer cell lines, and human breast cancer specimens.
Our results demonstrate that aberrant DNA methylation patterns emerge at the earliest stages of HMEC transformation in vitro, in finite lifespan HMEC. Aberrant changes proceeded in a stepwise fashion, with each step temporally linked to escaping one of the two proliferation barriers. The first step of DNA methylation change occurs when stasis is overcome or bypassed, and results in few to hundreds of aberrant hyper- and hypomethylation events, depending upon the manner by which stasis was overcome. A majority of these events are found in breast cancer cells. The second step occurs in cells that have overcome the telomere dysfunction barrier and become immortal, and is associated with hundreds of aberrant methylation events, regardless of the manner of immortalization. A majority of these events are also found in breast cancer cells. Further methylation changes occur during malignant progression. These results suggest that groups of DNA methylation changes can arise concurrently and in a stepwise fashion during early breast carcinogenesis.
Organoids were obtained from reduction mammoplasty tissues as described (18). Finite lifespan pre-stasis HMEC from specimens 184 (batches D, E, F), 48 (batches RT, LT), and 240L (batch B), and post-selection HMEC 184 (batch B, telomere dysfunction arrest at ~passage 15), and 48 (batch RS, telomere dysfunction arrest at ~passage 23) were derived from tissue of women aged 21,16, and 19 respectively. Cells were initiated as organoids in primary culture in either serum-free MCDB 170 medium (MEGM, Clonetics Division of Lonza, Walkersville, MD) plus supplements (8), or serum-containing media MM (19) or M85 (composed of 50% MM and 50% supplemented MCDB 170 media) (Garbe et al, submitted). Post-selection HMEC were cultured in MCDB 170 as described (8, 18). The post-stasis 184Aa, 184Be, 184Ce cultures and the non-malignant immortal lines 184A1 and 184B5 were obtained from primary cultures of specimen 184 grown in MM, that had been exposed to the chemical carcinogen benzo(a)pyrene as described (5). 184A1-RF was obtained by retroviral transduction of the 184A1 line with the Raf-1 oncogene, and has gained anchorage-independent growth (AIG) (10). The p53-/- 184AA2 immortal line with AIG was obtained from 184Aa following insertional mutagenesis as described (20). The 184B5ME line with AIG was obtained by transfection of 184B5 with erbB2, and selection for anchorage-independent colonies. The post-stasis 184F-p16sh and 184D-p16sh cultures were obtained by retroviral transduction of pre-stasis HMEC cultures that were grown respectively in M85 and M87A+X (composed of 50% MM4 (19) and 50% supplemented MCDB 170 media plus 0.1% AlbuMax (Invitrogen) and 0.1 nM oxytocin (Bachem)). The non-malignant immortal 184ZNMY3 line was derived from post-selection 184B following retroviral transduction of ZNF217 and c-myc; 184ZNMY3-N with AIG was obtained following retroviral transduction of 184ZNMY3 with the mutated neu oncogene. Breast cancer cell lines were cultured as previously described (21).
Flash frozen specimens derived from normal or cancerous breast tissue were obtained from patients who underwent surgery for breast cancer, either lumpectomy or mastectomy, at the University Medical Center in Tucson, AZ., from 2003-2005. All patients signed surgical and clinical research consents for tissue collection in accordance with the University of Arizona IRB and HIPAA regulations. At the time of surgery, a 1-3 cm section of the tumor was immediately snap frozen in liquid nitrogen and stored in our prospective breast tissue bank at -80 degree Celsius. From each tissue block, a series of 5 micron sections were cut and stained with hematoxylin and eosin (H&E) for pathological evaluation. All H&E slides were reviewed by two independent pathologists to determine the integrity of the tumor specimen. A partial molecular characterization of these samples have been reported on previously (21, 22). Supplementary Table 1 provides the pathological assessment of each specimen.
RNA and DNA were isolated as previously described (21).
Sodium bisulfite (NaBS) treated genomic DNA was prepared according to manufacturer's instructions (Zymo Research, Orange, CA). NaBS treated DNA (5 ng) was seeded into a region specific PCR reaction incorporating a T7 RNA polymerase sequence as described by the manufacturer (Sequenom, San Diego, CA). Resultant PCR product was then subjected to in vitro transcription and RNase A cleavage using the MassCLEAVE T-only kit, spotted onto a Spectro CHIP array, and analyzed using the MassARRAY Compact System MALDITOF mass spectrometer (Sequenom, San Diego, CA). Each NaBS treated DNA sample was processed in two independent experiments. Data were analyzed using EpiTyper software (Sequenom, San Diego, CA) as described (22, 23). Primer sequences were designed using EpiDesigner (www.epidesigner.com). Primer sequences are available upon request.
Methylated fraction of DNA was obtained by immunoprecipitation as described (21). Input and immunoprecipitated DNA were amplified, labeled and analyzed on Human Promoter arrays as described (24). All microarray data were processed in R programming environment using Limma package was used and differentially methylated elements were identified using statistical approaches as described previously (25). To control for false discovery rate, a multiple testing correction was performed according to the methods described (24). A list of differentially methylated elements is provided in Supplementary Table 2. A region was considered differentially methylated if the adjusted p value was lower than 0.05 and there was at least a 1.25 fold change in methylcytosine immunoprecipitated DNA versus input DNA ratio relative to pre-stasis HMEC. The magnitude of the microarray methylation ratio was directly correlated with the degree of methylation, as determined by MassARRAY (Supplementary Figure 1). All raw and normalized microarray data with detailed protocols are available in the ArrayExpress database; accession E-MEXP-1889 (www.ebi.ac.uk/arrayexpress). The number of and samples analyzed by DNA methylation microarray are provided in Supplementary Table 3. Gene ontology terms over-representation testing was performed using GOstats package (26). Overlapping probabilities of DMR sets were calculated using a hypergeometric test (27).
Figure 1 shows the cells analyzed in this study, their relative position with respect to proliferation barriers, the predicted corresponding in vivo correlates (3, 9, 13), the timing and accumulation of genetic abnormalities (9, 13, 28-30), and finally the timing and accumulation of DNA methylation changes determined in this study. The DNA methylation state of HMEC at different stages in the transformation from normal finite pre-stasis to immortal with AIG, as well as of malignant breast cancer cell lines, was analyzed using a 13,500 element human gene promoter microarray (24); arrows in the bottom panel indicate when cells were examined.
The first step of DNA methylation change occurs when stasis is overcome or bypassed and post-stasis HMEC emerge (Figure 2A). When pre-stasis HMEC are compared to post-selection HMEC (48RS and 184B) that overcame stasis associated with silencing of p16 following culture in a stress-inducing serum-free medium (8), 191 differentially methylated regions (DMRs) were identified, in addition to the previously described CDKN2A methylation (4). This number represents approximately 2% of all promoters analyzed on this microarray. In contrast, HMEC that become post-stasis following exposure to the mutagen and complete carcinogen benzo(a)pyrene (184Aa, 184Be, 184Ce) (5, 31) exhibited only 10 DMR when compared to pre-stasis HMEC. Although this is a very small number of elements that could represent false discoveries or stochastic events, it appears that at least a few are valid and relevant, such as the HOXA gene cluster, since this is a common target among all these post-stasis HMEC samples, as well as clinical disease. Supplementary Figure 2 shows high resolution methylation analysis that confirms aberrant methylation of the HOXA4 locus in the benzo(a)pyrene treated HMEC. In contrast, HMEC that became post-stasis following transduction with p16 shRNA (184F-p16sh, 184D-p16sh) accumulated only 5 DMRs, and did not show changes in the HOXA gene cluster. These results suggest that different levels of DNA methylation changes may be associated with HMEC that overcome the stasis proliferation barrier via different genetic or epigenetic mechanisms. Considering that the number of aberrantly methylated genes in cancer has been estimated to be between several hundreds to low thousands (1, 22, 32), and the large number of DMRs seen in the post-selection HMEC, the transition through the stasis proliferation barrier may represent a critical epigenetic event in some pathways of carcinogenesis.
The second step of DNA methylation change occurs when telomere dysfunction is overcome and cells become immortal (Figure 2A). This second stepwise increase in DNA methylation changes produced hundreds of new DMRs, regardless of how telomere dysfunction or stasis was overcome. Breast cancer cell lines derived from malignant tumors displayed the greatest number of aberrant DNA methylation changes. Taken together, these results are consistent with progressive, stepwise increases in aberrant DNA methylation during the transformation process.
To further address whether aberrant DNA methylation progresses in a gradual or stepwise fashion in the earliest stages of breast carcinogenesis, we examined multiple early and late passage pre-stasis and post-selection HMEC, including cell passages in telomere dysfunction, that are known to have genomic instability (9) (Figure 2B). A comparison of the DNA methylation patterns of early and late passage pre-stasis HMEC revealed no DMRs. Similarly, no additional DMRs were detected when early passage post-selection HMEC were compared to those at telomere dysfunction. These results showed that there is not a gradual progression or increase in the number of DMRs between proliferative barriers, but rather new DMRs emerge in cells that overcame stasis or telomere dysfunction. Such step-wise changes are demonstrated on Figure 2B which shows the methylation status of post-stasis specific DMRs and breast cancer cell line specific DMRs at various stages of transformation. The abrupt increase in DNA methylation changes in post-stasis and immortal cells show the breakpoints of epigenetic reprogramming.
If the DNA methylation changes found in post-stasis HMEC and immortal HMEC cell lines are relevant to malignant transformation, then the identified methylation changes should resemble those seen in malignant breast cancer cell lines. Figure 2C is a Venn diagram relating the DMR found in post-stasis and immortal HMEC, and genetically distinct breast cancer cell lines. Of the 203 DMRs found in post-stasis HMEC populations, 136 DMRs (67%) are also aberrantly methylated in breast cancer cell lines (overlapping probability p value < 2.2 ×10e-16). Furthermore, of the 484 DMRs identified in the immortalized HMEC lines, 327 (68%) are also found in breast cancer cell lines (p value < 2.2 ×10e-16). Supplementary Figure 3 shows the overlap and differences in DMRs between immortal cell lines that bypassed this proliferation barrier by different mechanisms. These data indicate that a significant overlap exists between the targets of aberrant DNA methylation in breast cancer cells and the targets of aberrant DNA methylation in the pre-malignant stages represented by the HMEC model.
To further examine the timing of specific methylation events, two gene clusters known to undergo aberrant methylation during breast carcinogenesis were examined in the gene promoter microarrays. Previous studies have shown frequent hypermethylation and silencing of the HOXA and PCDH gene family clusters in breast cancer (21, 22). Our data indicate that a prevalence of aberrantly methylated HOXA genes is also seen in both post-stasis and immortal HMEC. Detailed exploration of the HOXA genomic region showed that silencing of the whole cluster can originate early in breast carcinogenesis (Figure 3). Early DNA methylation changes occur in HOXA3, HOXA4, HOXA9, HOXA10 and HOXA13 genes in the post-selection HMEC. Interestingly, the HOXA cluster is also one of the few regions where DMRs emerged in post-stasis cells that bypassed stasis following exposure to benzo(a)pyrene; however, the HOXA cluster was not targeted in HMEC that bypassed stasis via silencing of p16 by shRNA. Further progression in aberrant DNA methylation occurs in the HOXA gene cluster in the immortal HMEC cell lines that overcame the telomere dysfunction barrier. Continued malignant progression is ultimately associated with aberrant DNA hypermethylation of the whole gene cluster as seen in the breast cancer lines.
A similar progression in aberrant DNA methylation occurs in another genomic region, the PCDH gene family cluster. However, detailed exploration of the PCDH genomic region showed that in this cluster, most changes are observed in the transition from post-stasis to immortal HMEC (Supplementary figure 4). Whereas almost no changes were seen in the post-stasis HMEC, the non-malignant immortal cell lines already displayed most of the changes seen in the breast tumor lines. These results clearly illustrate the step-wise nature of the DNA methylation changes in association with overcoming the proliferation barriers to immortality.
To confirm the methylation microarray data and extend our model, we analyzed the methylation status of post-stasis specific DMRs in pre-stasis, post-stasis, and immortalized HMEC, a set of primary breast tumors, and breast epithelial organoids using MassARRAY (Figure 4, Supplementary Figure 3). Results obtained confirmed the microarray data and extended them by showing that the DNA methylation changes seen in the in vitro model of post-stasis HMEC can also be found in primary breast tumors, but are not present in the normal epithelial organoids used to establish the HMEC system (18). In addition to the inappropriate methylation of CDKN2A as reported in focal aggregates of histologically normal mammary epithelia(15, 16) and in atypical ductal hyperplasia (ADH) (33), we also found the promoter region of PGR to be inappropriately methylated in post-stasis cells, similar to findings in ADH (34). Overall, these results indicate that post-stasis HMEC are a relevant model of early epigenetic changes in breast carcinogenesis, and that the hundreds of DMRs discovered may serve as potential markers of pre-malignant breast lesions.
To evaluate the potential functional importance of the DNA methylation changes, we analyzed the ontology of the genes where the promoters were targeted for changes in DNA methylation, and we found several overrepresented groups of genes (Supplementary figure 5). A significant number of genes involved in the extracellular matrix and the extracellular region were hypermethylated in post-stasis cells, consistent with published gene expression data that showed differential expression of extracellular matrix and cell-cell communication genes between pre-stasis and post-stasis HMEC (14). Another noteworthy group of affected genes that are hypermethylated in both post-stasis and breast cancer cell lines is the gene ontology group related to adhesion and the plasma membrane. This gene ontology analysis revealed groups of genes targeted by aberrant DNA methylation in premalignant mammary epithelial that are also found to be dysregulated in breast cancer, suggesting that epigenetic changes important to the malignant phenotype may occur early in the multistep process of malignant transformation and prior to immortalization.
In this report, we have shown that the transition of cultured HMEC from normal finite lifespan to immortality is associated with a step-wise progression of DNA methylation changes, and that these steps are coincident with passage through the defined epithelial cell proliferation barriers of stasis and telomere dysfunction. HMEC that emerged from these proliferation barriers acquired hundreds of DNA methylation changes compared to cells examined just prior to the proliferation barrier. In contrast, no changes in DNA methylation were observed between early and late passage cell populations that preceded the proliferation barriers. These results suggest a direct mechanistic link between epigenetic dysfunction and escape from senescence barriers thought to function as tumor suppressor mechanisms. Importantly, a majority of the two-step DNA methylation changes identified using this HMEC system are also seen in both breast cancer cell lines and tumor specimens. Since these methylation changes occurred at the earliest stages of transformation, in still finite lifespan HMEC populations, our results suggest that numerous aberrant methylation changes may be present in pre-malignant lesions of breast cancer. Indeed, several genes previously shown to have hypermethylated promoters in histologically normal mammary epithelia and premalignant stages of breast carcinogenesis were also identified in this study, including the CDKN2A and PGR promoters (15, 16, 33-37).
Gene Ontology analysis of the promoters targeted by DNA methylation showed that genes involved in the biological process of multicellular organismal development were overrepresented in both post-stasis cells and cancer cell lines, suggesting these targets may participate in the initiation and maintenance of the malignant phenotype (Supplementary Figure 5). HOXA genes are prominent in this list and are represented by HOXA2, HOXA4, HOXA9, HOXA10, HOXA11, and HOXA13. The disruption of these transcriptional regulators, which are involved in control of cell identity, are logical genes to be targeted in the malignant transformation process, and the consequences of their dysregulation are likely manifest through changes in the expression of HOXA target genes. The fact that aberrant DNA methylation and transcriptional dysregulation of the HOXA gene cluster is seen in a variety of human tumor types suggests that these events play a critical role in human carcinogenesis (21, 38-40). Based on this information we speculate that HOXA genes are critical targets in the epigenetic initiation of malignant transformation.
The HOXA genes also represent a gene family cluster that frequently undergoes Long Range Epigenetic Silencing (LRES) in invasive breast cancer (21, 22, 41, 42). Interestingly, the PCDH gene family clusters (α, β, γ) on chromosome 5, another known target of LRES epigenetic silencing in invasive breast cancer, are also represented in the early steps of our in vitro model of transformation (22). The increased DNA methylation observed in these two known LRES targets provide evidence that 1) LRES is initiated early in mammary epithelial cell transformation, 2) LRES likely initiates in a step-wise fashion, similar to the focal events observed in this study, and 3) the in vitro HMEC system is an accurate reflection of the clinical disease. Taken together these results suggest that the post-stasis specific and the immortalization specific DMRs identified in this study may serve as potential markers of pre-malignant events in breast carcinogenesis. The specific DNA methylation changes identified in this study can potentially provide a bank of epigenetic biomarkers for assessing breast cancer risk, and allow for the analysis of multiple post-stasis and immortalization specific DNA methylation changes that when combined with additional types of genomic data (e.g. SNPs, gene expression) will help develop increasingly robust risk assessment models.
We examined cells that escaped stasis by three distinct means, exposure to the chemical carcinogen benzo(a)pyrene, growth in a stress-inducing serum-free medium, and direct inactivation of p16 by p16shRNA; the first two methods generated clonal post-stasis populations. The extent of DNA methylation changes at this step varied greatly, depending upon the manner by which the HMEC had become post-stasis. Post-selection HMEC displayed DNA methylation changes at hundreds of gene promoters, including CDKN2A, while HMEC that bypassed stasis following benzo(a)pyrene exposure or direct genetic inactivation of p16 showed few changes. These results suggest that different pathways to post-stasis produce different epigenetic signatures. Notably, however, the DNA methylation state of the HOXA gene cluster is perturbed in all post-selection and benzo(a)pyrene exposed post-stasis cultures, regardless of pathway taken. These results suggest that within the emerging transformation-associated epigenetic signatures, there may be common critical epigenetic targets affected by the distinct paths and etiologies of breast cell transformation. The epigenetic aberrations that are common to the different pathways may play a critical role in driving the transformation process forward - the early deregulation of the HOX gene family clusters, which are decisively linked to human carcinogenesis, are one clear example that is consistent with this possibility (21, 39, 40). In contrast, the epigenetic aberrations that are specific to particular pathways to post-stasis and therefore occur at this earliest stage of transformation may allow for the emergence of distinct phenotypes if these cells pass to immortality and malignancy.
Although the stasis proliferation barrier could be bypassed with minimal epigenetic changes, all the immortalized lines examined acquired hundreds of additional DNA methylation changes, regardless of how they became post-stasis or immortal. These results suggest that DNA methylation changes may be necessary to achieve immortality. The role of epigenetic control in telomerase reactivation remains controversial and incompletely understood (43, 44). To address the potential role of DNA methylation in the control of hTERT in this HMEC model system, future studies will perform high resolution epigenetic analyses of the hTERT region. It is also possible that critical epigenetic changes within the multitude of genes affected during the transition to immortality are important and participate in telomerase reactivation.
Taken together, these results indicate that epigenetic alterations in DNA methylation found in breast cancer cells in vivo may arise during the earliest stages of HMEC transformation, prior to and coincident with attainment of immortality. Thus full understanding of the timing and nature of epigenetic alterations in early breast carcinogenesis will require examination of cells prior to attaining immortality. While non-malignant immortalized lines such as 184A1, 184B5, and MCF10A can be useful starting points for studying epigenetic events involved in the progression to invasive malignancy, they already possess many of the epigenetic aberrations found in breast cancers. Similarly, the post-selection HMEC, although finite, also possess many epigenetic changes found in breast cancer cells. This finding is of particular note since post-selection HMEC are sold commercially as normal primary cells.
Overall, these results support an epigenetic progenitor model whereby genome-wide DNA methylation changes occur early, in a stepwise fashion, may precede genetic mutations, and allow for an inappropriate proliferation and expansion of epigenetically comprised progenitor cells (45). The large number of genes affected by aberrant DNA methylation provides a foundation for phenotypic variability in cells that transform to immortality and malignancy, and may be an important force that drives the significant biological heterogeneity seen in the clinical disease. The DNA methylation changes identified in this HMEC model system can potentially provide a bank of epigenetic biomarkers for assessing breast cancer risk in pre-malignant lesions, and provide targets for therapeutic interventions.
We thank Jose Munoz-Rodriguez (UA) and Batul Merchant (LBNL) for outstanding technical support. Grants R01CA65662 and R33CA091351 to B.W.F. supported this work. Center Grants P30ES06694 and P30CA023074, and the BIO5 interdisciplinary biotechnology center at the UA supported the Genomics Shared Service. Training grants ES007091 and CA09213 helped support T.J. J.C.G. and M.R.S. were supported by NIH U54 CA112970, DOD BCRP BC060444, and the Office of Energy Research, Office of Health and Biological Research, U.S. Department of Energy under Contract No. DE-AC03-76SF00098.