Distinctive patterns of age-dependent hypomethylation in interspersed repetitive sequences

Pornrutsami Jintaridth, Apiwat Mutirangura

Abstract

Interspersed repetitive sequences (IRSs) are a major contributor to genome size and may contribute to cellular functions. IRSs are subdivided according to size and functionally related structures into short interspersed elements, long interspersed elements (LINEs), DNA transposons, and LTR-retrotransposons. Many IRSs may produce RNA and regulate genes by a variety of mechanisms. The majority of DNA methylation occurs in IRSs and is believed to suppress IRS activities. Global hypomethylation, or the loss of genome-wide methylation, is a common epigenetic event not only in senescent cells but also in cancer cells. Loss of LINE-1 methylation has been characterized in many cancers. Here, we evaluated the methylation levels of peripheral blood mononuclear cells of LINE-1, Alu, and human endogenous retrovirus K (HERV-K) in 177 samples obtained from volunteers between 20 and 88 yr of age. Age was negatively associated with methylation levels of Alu (r = −0.452, P < 10−3) and HERV-K (r = −0.326, P < 10−3) but not LINE-1 (r = 0.145, P = 0.055). Loss of methylation of Alu occurred during ages 34–68 yr, and loss of methylation of HERV-K occurred during ages 40–63 yr and again during ages 64–83 yr. Interestingly, methylation of Alu and LINE-1 are directly associated, particularly at ages 49 yr and older (r = 0.49, P < 10−3). Therefore, only some types of IRSs lose methylation at certain ages. Moreover, Alu and HERV-K become hypomethylated differently. Finally, there may be several mechanisms of global methylation. However, not all of these mechanisms are age-dependent. This finding may lead to a better understanding of not only the biological causes and consequences of genome-wide hypomethylation but also the role of IRSs in the aging process.

  • global hypomethylation
  • genome-wide methylation
  • aging
  • interspersed repetitive sequences

global hypomethylation, or the decreasing of genome-wide DNA methylation, is a common epigenetic alteration in senescent cells ( 23, 56, 57). Overall, total deoxymethylcytosine levels tend to decrease with aging in most vertebrate tissues, including the brain, liver, small intestine mucosa, heart, and spleens of salmon, mice, rats, cows and cattle, and aged human peripheral blood cells ( 7, 22, 23, 25, 36, 48, 53, 54, 57). Interestingly, although global hypomethylation is related to the replicative senescence associated with aging ( 56), it is also a common feature of cells with limitless replication potential, including stem cells ( 62) and cancer cells ( 8). Age-dependent global hypomethylation has been shown to be associated with many diseases ( 20, 27, 49, 55). Moreover, in cancer, global hypomethylation is associated with more advanced stages and poor prognosis ( 30, 31, 36, 39, 40, 43, 50, 51). Therefore, it is important to understand how the same epigenetic characteristic is associated with a variety of cellular phenotypes.

Interspersed repetitive sequences (IRSs) are a major contributor to genome size, accounting for ∼45% of human genomic DNA ( 35, 45). Under normal circumstances, IRSs are methylated to variable degrees depending on their location ( 46). Therefore, loss of IRS methylation was believed to be the major contributor of global hypomethylation ( 8). IRSs can be subdivided by size and the presence of transposable elements. Commonly known IRSs are short interspersed elements (SINEs) being <500 bp, long interspersed elements (LINEs) being 6–8 Kb, long terminal repeat (LTR)-retrotransposons, and DNA transposons ( 14). SINEs, LINEs, and LTR-retrotransposons possess transcriptional activity, and several reports have suggested they are cis regulatory elements ( 28, 34, 52, 61). Therefore, changes in SINE, LINE, or LTR-retrotransposon methylation should alter cellular functions. Because cellular biology of cancer and aging are different, we hypothesized that the age-dependent pattern of methylation losses in SINEs, LINEs, and LTRs should be different.

Alu is the most abundant SINE ( 4, 14). Human endogenous retroviruses (HERVs) are a major subset of LTR-retrotransposons ( 5, 14, 47). LINE-1, or L1, is the most abundant and well-known LINE ( 14, 29). The methylation status of Alu, LINE-1, and HERV-K, a well-known member of HERV, has been reported for many cancers ( 5, 8, 16, 19, 50, 51, 55, 60). Recently, Bollati et al. ( 6) reported an age-related methylation of peripheral blood mononuclear cells (PBMC) of Alu and LINE-1. They found a significant association between age and Alu hypomethylation, but the significance of the association between age and LINE-1 hypomethylation was borderline ( 6). In the present study, we describe patterns of methylation alternations in Alu, LINE-1, and HERV-K. Our analysis suggests that IRSs lose methylation at a certain age. Moreover, different IRSs lose methylation differently. Finally, there may be several global methylation mechanisms and not all of them are dependent on age.

MATERIALS AND METHODS

Subjects.

Our study population consisted of 42 men and 135 women aged 20–90 yr. Healthy volunteer subjects in the urban area of Bangkok were recruited for sample collection according to the following criteria: each individual 1) agreed to participate in the study and gave informed consent, 2) had no history of chronic or genetic diseases, 3) had a body mass index <30 kg/m2, 4) had no history of smoking and alcohol drinking 10 yr prior to sample collection, 5) engaged in no supplementation with vitamins for 3 mo before study, and 6) had normal living and eating habits. This study was approved by the ethics committee of the Faculty of Tropical Medicine, Mahidol University.

Blood collection.

Heparinized blood was collected, and PBMC were prepared by Ficoll density gradient centrifugation according to published methods ( 3). Blood hematological parameters and blood chemistries of participants were analyzed. Volunteers with normal hematological parameters and blood chemistries were chosen to participate. DNA methylation in PBMC from healthy subjects of different ages was measured. Hematological data and the amount of PBMC in whole blood were not significantly different between individuals.

DNA extraction and modified combined bisulfite restriction analysis.

DNA was extracted from PBMC using QiAmp DNA blood kits (Qiagen, Hilden, Germany). A total of 200 ng DNA (concentration 4 ng/μl) was used for bisulfite treatment. The combined bisulfite restriction analysis (COBRA) consisted of a standard sodium bisulfite PCR treatment followed by restriction digestion and quantitation ( 58). Bisulfite modification of genomic DNA was performed using previously published methods ( 8). In brief, 200 ng of DNA was dissolved in 50 μl distilled water and then denatured in 5.5 μl of 2 M NaOH for 10–30 min at 37°C. Next, 30 μl of freshly prepared 10 mM hydroquinone (Sigma) and 520 μl of 3 M bisulfite (pH 5.0) were added and mixed. The samples were incubated at 50°C for 16 h. The bisulfite-treated DNA was isolated using the Wizard DNA Clean-up System. The DNA was eluted with 50 μl of warm water and desulfonated with 5.5 μl of 3 M NaOH for 5 min. The DNA was precipitated (NH4OAC-EtOH) using glycogen as a carrier and resuspended in 20 μl of water. Bisulfite-treated DNA was stored at −20°C until ready for use.

COBRA-IRS.

COBRA is a standard technique for measuring methylation levels of IRSs ( 2, 8, 9, 25, 32, 38, 4143, 50, 51, 59). A schematic representation and examples of COBRA of LINE-1, Alu, and HERV-K are shown in Fig. 1. The primer sequences that correspond to the nucleotides in the regulatory region of the LINE-1 sequence (GenBank: M80343) are forward (F), 5′-CGT AAG GGC TTA GGG AGT TTT T-3′ and reverse (R), 5′-(AG)TA AAA CCC TCC (AG)AA CCA AAT ATA AA-3′ ( 8). The PCR reactions consisted of 35 cycles of 95°C for 1 min, 53°C for 1 min, and 72°C for 1 min. The PCR products were subsequently digested with 2 U of TaqI (MBI Fermentas) and 2 U of TasI (MBI Fermentas) in TE buffer 3 (Biolab) at 65°C overnight and were then run on an 8% nondenaturing polyacrylamide gel. The gel was stained with SyBr Green, and band intensities were measured by PhosphoImager using Image Quant software (Molecular Dynamics). The Alu primer sequences, which correspond to the nucleotides of the Alu Sx subfamily sequence ( 4), are F, 5′-GG(T/C) G(C/T)G GTG GTT TA(C/T) GTT TGT AA-3′ and R, 5′-CAC CAT ATT AAC CAA ACT AAT CCC GA-3′. Alu PCR conditions and restriction digestion conditions were similar to those of LINE-1. The HERV-K primer sequences (GenBank accession number AF394944.1), which correspond to the nucleotides in the regulatory region of the HERV-K sequence, are F, 5′-TGG GAA GGG AAA GAT TTG AT-3′ and R, 5′-ACA AAA AAC AAA TAC CTT CCT CTT-3′. The PCR reactions consisted of 30 cycles of 95°C for 30 s, 60°C for 30 s, and 72°C for 30 s. The PCR products were then digested with 2 U of TaqI (MBI Fermentas) and in TE buffer 3 (Biolab) at 65°C overnight and were run on an 8% nondenaturing polyacrylamide gel. The gel was stained with SyBr Green, and band intensities were measured by PhosphoImager using Image Quant software (Molecular Dynamics). COBRA LINE-1, Alu, and HERV-K amplicons were 160, 99, and 126 bp, respectively. After digestion, the LINE-1, Alu, and HERV-K methylated bands were 80, 57, and 94 bp, respectively. Unmethylated bands of LINE-1 were 97 and 63 bp, and the unmethylated band of Alu was 78 bp. Methylation levels were calculated as the intensity of methylated bands divided by the sum of the methylated and unmethylated bands. Each COBRA-IRS was performed two to four times. To ensure that measurements of IRS methylation from band intensities of samples reduced interassay variation, we used Daudi, Jurkat, HeLa, and Molt4 cell lines from the same stocks as a control to validate the interassay variation. IRS methylation levels of the stock cell lines was standardized, and IRS methylation in each experiment was adjusted so that all experiments had the same control IRS methylation levels. The average differences in IRS methylation of the LINE-1, Alu, and HERV-K repetitive elements between adults of different ages were evaluated. IRS methylation of LINE-1, Alu, and HERV-K repetitive elements were also analyzed for associations with age intervals. Associations between IRS methylation of LINE-1, Alu, and HERV-K in each group of age intervals were also determined.

Fig. 1.

Combined bisulfite restriction analysis-interspersed repetitive sequence (COBRA-IRS) of sodium bisulfite-treated peripheral blood mononuclear cell (PBMC) DNA of healthy subjects. A and B: COBRA of long intersperse element (LINE)-1. Unmethylated CpGs within AACCG sequences and methylated CpGs within CCGA sequences create TasI-specific AATT(G) and TaqI-specific TCGA sites, respectively, after sodium bisulfite treatment. Methylation levels can be assessed by TasI-TaqI double digestion within the 160 bp amplicon. The presence of an 80 bp fragment was yielded from TaqI digestion, whereas TasI digestion yielded 63 and 97 bp products. C and D: COBRA of Alu. TasI-specific unmethylated AATTTT and TaqI-specific methylated TCGAG sites were created from the original AATCC and CCGAG sequences, respectively, of Alu after sodium bisulfite treatment. A 99 bp amplicon was cut and a 78 bp methylation-specific band was generated in the presence of TasI, and a 57 bp band was present with TaqI digestion when the CpG of the AATCC was methylated. E and F: COBRA of human endogenous retrovirus (HERV)-K. A 126 bp amplicon was cut, resulting in 94 and 32 bp fragments when the CCGA sequences of the original bisulfite-treated HERV-K sequences were methylated and converted to TCGA sequences. M, DNA marker; -ve, negative (water and no template) control. Percentage of methylation is listed above each test.

Statistical analysis.

Data were analyzed using SPSS statistical software. The Mann-Whitney U-test was used to make nonparametric comparisons of the median methylation levels of LINE-1, Alu, and HERV-K between males and females. To analyze different age intervals, ages were divided into 20 yr intervals, and the age interval of each group overlapped adjacent intervals by 15 yr. Some groups were not normally distributed. The averages of methylation levels of LINE-1, Alu, and HERV-K in each group were determined as the median value and the Kruskal-Wallis test was used to make nonparametric comparisons between the groups. To evaluate the correlations between ages and IRS methylation levels in the population and groups of age intervals, Pearson and Spearman correlations were used to determine associated 95% confidence intervals. The associations between LINE-1, Alu, and HERV-K relative methylation levels in the population and groups of age intervals were examined while controlling for confounding variable such as age. Partial correlation was used to determine 95% confidence intervals. All tests were two-tailed analyses.

RESULTS

Association between age and methylation of LINE-1, Alu, and HERV-K.

This study consisted of PBMC of 177 study subjects aged between 20 and 90 yr old (mean age 52.92 yr, SE = 1.19). Methylation levels of LINE-1, Alu, and HERV-K were measured by COBRA ( 2, 8, 9, 25, 32, 38, 4143, 50, 51, 59). Examples of COBRA-IRS were demonstrated ( Fig. 1, A–F). Our tests produced limited result deviations (Supporting Fig. S1). 1

First, we evaluated the relationship between IRS methylation levels and age. LINE-1 methylation was not associated with age (r = 0.145, P = 0.055) ( Fig. 2A). However, DNA methylation of both Alu and HERV-K was inversely correlated with age (r = −0.452, P < 10−3 for Alu; r = −0.326, P < 10−3 for HERV-K) ( Fig. 2, B and C). Therefore, age-related genomic hypomethylation occurred at only specific types of IRSs, i.e., Alu and HERV-K, but not LINE-1. We then evaluated whether sex influences IRS methylation. Similar to a previous report ( 8), there were no differences in the methylation of LINE-1, Alu, or HERV-K between males and females ( Fig. 3, A–C).

Fig. 2.

Association between age and IRS methylation of LINE-1 (A), Alu (B), and HERV-K (C). Significance of correlation coefficients (r) between age and IRS methylation was set at P < 0.05. %5 mC, %5-methylcytosine.

Fig. 3.

Methylation differences of LINE-1 (A), Alu (B), and HERV-K (C) in males and females. Comparisons of methylation levels were made between males and females. Error bars represent 95% confidence intervals. %5 mC, %5-methylcytosine.

Association between LINE-1, Alu, and HERV-K methylation.

We evaluated the correlations of methylation levels between IRSs. Because only Alu and HERV-K are demethylated with age, we expected a positive correlation exclusively between Alu and HERV-K. In contrast, we found a significant association between LINE-1 and Alu (r = 0.290, P < 10−4) or (r = 0.272, P < 10−3 when excluding outlier) ( Fig. 4A). There was no association between LINE-1 and HERV-K methylation (r = −0.058, P = 0.45) ( Fig. 4B). Unexpectedly, Alu and HERV-K methylation were also not correlated (r = 0.018, P = 0.81) ( Fig. 4C).

Fig. 4.

Association between methylation of repetitive elements. Each plot represents methylation levels of individuals, and error bars represent 95% confidence intervals. Pearson's correlation coefficients (r) with P values are indicated. Alu and LINE-1 methylation levels (A) are significantly correlated (P < 0.0001). In contrast, no correlation was observed either for Alu and HERV-K methylation levels (B) or for LINE-1 and HERV-K methylation levels (C). %5 mC, %5-methylcytosine.

Changes in LINE-1, Alu, and HERV-K methylation levels in different age intervals.

To determine the certain age at which hypomethylation occurs, we grouped samples according to age into 20 yr intervals, with the intervals of group overlapping each adjacent interval by 15 yr ( Figs. 57). Figure 5, A, B, and C, shows the median, minimum, and maximum of LINE-1, Alu, and HERV-K methylation, respectively, of each interval. Figure 6 shows the r values of the correlation between IRS methylation and each interval of age. No significant LINE-1 losses of methylation were observed in any age, but levels of LINE-1 methylation might increase during ages 40–59 yr (r = 0.32, P = 0.02) ( Fig. 6A). An increase of LINE-1 methylation has been seen previously in placentas ( 44) and in some specific loci in cancer ( 46). More importantly, we discovered that age-related hypomethylation was found at certain ages. Age-dependent Alu loss of methylation was significant from 34 to 68 yr of age (r = −0.477, P < 10−3) ( Figs. 5B, 6B, and Supporting Table S1). The shape of the association between Alu methylation and the age interval 34–68 yr had significant linear and curvilinear trends (Supporting Fig. S2, Supporting Table S2). HERV-K lost methylation twice, during the 40–63 and 64–83 age intervals ( Figs. 5C, 6C, and Supporting Table S1). The shape of association between HERV-K methylation and the age interval 40–83 yr did not fit with both linear and nonlinear models due to errors from curve fit having nonnormal distribution (Supporting Fig. S3, Supporting Table S3). These data support the previous correlation results that the age-related hypomethylations of Alu and HERV-K did not occur via the same process.

Fig. 5.

Changes in LINE-1 (A), Alu (B), and HERV-K (C) methylation levels in different age intervals. Error bars indicate ± SD, which was calculated using the averaged methylation level of all individuals in each group. %5 mC, %5-methylcytosine.

Fig. 6.

Association between age and IRS methylation [LINE-1 (A), Alu (B), and HERV-K (C)] in different age intervals. Different levels of significance are denoted by asterisks: *P < 0.05, **P < 0.01, and ***P < 0.001.

Associations between LINE-1, Alu, and HERV-K methylation levels in different age intervals.

We analyzed the association between Alu and LINE-1 methylation to study how LINE-1 was not demethylated by age despite its methylation levels being directly correlated with Alu. Alu and LINE-1 methylation were not correlated at younger ages but were significantly correlated at ages 49 yr and older (r = 0.49, P < 10−3) ( Fig. 7A and Supporting Table S4). These data support the previous conclusion that Alu methylation may be classified as both age-dependent and -independent. Moreover, the age-independent process may be the same mechanism as that for LINE-1. Alu and HERV-K methylation were directly and significantly correlated during the 40–59 yr age interval ( Fig. 7B and Supporting Table S4). Interestingly, LINE-1 and HERV-K were inversely correlated during the 44–68 yr age interval ( Fig. 7C and Supporting Table S4). Therefore, there may be a link between LINE-1 hypermethylation and HERV-K hypomethylation during these ages. Finally, there was direct association between LINE-1 and HERV-K during the 69–88 yr age interval ( Fig. 7C and Supporting Table S4). This may be similar to the association between Alu and LINE-1 at advanced ages.

Fig. 7.

Pair-wise association of methylation of repetitive elements in different age intervals. LINE-1 and Alu (A), Alu and HERV-K (B), LINE-1 and HERV-K (C). Asterisks are used to denote different levels of significance: P < 0.05 (*), 0.01 (**), and 0.001 (***).

DISCUSSION

COBRA-IRS.

There is a premise that COBRA-IRS should have been more subject to error. COBRA detects methylation of only one CpG dinucleotide. If there is a mutation, for example C→T, the methylation status will be misinterpreted. COBRA-IRS, however, detects thousands of loci. Consequently, if a mutation does occur, it will not interfere with the interpretation. Moreover, methylation levels of each LINE-1 CpG dinucleotide is in positive linear correlation with the others on the same locus ( 46). Moreover, pyrosequencing demonstrated frequent positive linear correlations between methylation levels of means and each IRS CpG dinucleotide, including TaqI CpG sites of COBRA-Alu, LINE-1, and HERV-K (Supporting Fig. S4). Therefore, COBRA-IRS is an accurate and reliable technique. Up to now, there are several IRS methylation studies. Nonetheless, there has been no conflict result yielded because of different techniques ( 6, 8, 10, 11, 13, 15, 24, 25, 31, 39, 40, 43, 46, 4951, 55). It is important to note that IRS methylation studies will not display absolute levels. Therefore, methylation levels between different IRS types or PCR protocols cannot be compared. IRS sequences vary. Therefore, each PCR protocol will target different number of IRS templates. Consequently, IRS methylation levels are relative numbers and the levels between samples can be compared only by the same detection protocol. As a result, we do not compare Alu ( Fig. 3B) or LINE-1 methylation levels in this study with Bollati et al. ( 6).

Age- and cancer-related global hypomethylation.

In the past, IRSs were wrongly called “junk DNA” because they were thought not to possess any physiological role ( 33). Consequently, it has been expected that, in general, the level of methylation at each IRS should represent genome-wide levels. Here, our data suggest that different types of IRSs lose methylation via different age-dependent mechanisms. While age-dependent hypomethylation of Alu and HERV-K was observed, hypomethylation of LINE-1 was not. These data do not conflict with previous findings. Although Bollati et al. ( 6) reported a trend of LINE-1 hypomethylation with aging, they observed only borderline significant levels, unlike Alu. Several other reports did not identify an age-associated hypomethylation of LINE-1 ( 8, 18, 25). Instead of age-associated hypomethylation, LINE-1 hypomethylation is common in cancer cells ( 8, 10, 11, 13, 15, 19, 24, 25, 31, 39, 40, 43, 46, 4951, 55). Interestingly, head and neck squamous cell carcinoma risk factors such as smoking could lower levels of LINE-1 methylation in peripheral blood cells ( 25). Finally, hypomethylation of both Alu or HERV-K and LINE-1 elements often coexists in cancer cells ( 10, 13, 19). Therefore, age- and cancer-related global hypomethylation are two distinctive processes.

Age-dependent and -independent mechanisms of IRS methylation.

Age-dependent and -independent mechanisms of IRS methylation are complex. Because Alu and HERV-K losses of methylation do not occur simultaneously, age-dependent hypomethylation of Alu and HERV-K may be independent events. Moreover, there is IRS methylation that is not age-dependent. While hypomethylation of Alu and HERV-K is age-dependent, hypomethylation of LINE-1 is not. We found that Alu and HERV-K methylation was directly associated with LINE-1 methylation at older ages, which may be after age-related IRS loss of methylation has ended. Therefore, Alu methylation levels may be divided into two classes. The first reduced methylation in relation to age. The other persists and is directly correlated with age-independent LINE-1 methylation. Because LINE-1 hypomethylation is commonly associated with cancer ( 8, 10, 11, 13, 15, 19, 24, 25, 31, 39, 40, 43, 46, 4951, 55), it is intriguing to hypothesize whether age-independent mechanisms of IRS methylation may prevent cancer development. It is possible that the same IRS element, such as Alu or HERV-K, may be methylated by both age-dependent and -independent mechanisms. Recently, we reported the methylation statuses of several LINE-1 loci and found some were resistant to cancer hypomethylation processes ( 46).

IRS methylation as epigenetic marks of aging.

Several physiological and anatomical alterations occur during aging. These changes occur and progress at different rates at certain ages. We have demonstrated that Alu methylation decreased during ages 34–68 yr and HERV-K decreased during ages 40–63 yr and again during ages 64–83 yr. Both were also correlated within age intervals. This may be due to the loss of methylation processes of both IRSs occurring commonly at the period. Because genomic methylation was proposed to be related to several diseases ( 55), it would be interesting to determine whether hypomethylation of SINE and LTR is associated with age-related disabilities. At present, IRSs are no longer considered junk DNA. Instead, several lines of evidence suggest that SINEs, LINEs, and LTRs are gene cis regulatory elements ( 12, 17, 26, 44, 45, 47). Because SINEs, LINEs, and HERVs produce RNA, DNA methylation should alter their gene regulation activities. In addition, IRS hypomethylation patterns associated with cancer and aging are different. Therefore, global hypomethylation in cancer and aging should lead to different cellular and physiological consequences. Finally, modifying epigenetic marks that contribute to disease development has been proposed as a novel therapeutic approach for disease treatment and/or prevention ( 1, 21, 37). This study may provide crucial information on future therapeutic targets for the prevention of aging or age-associated disabilities.

GRANTS

This work was supported by the Thailand Research Fund (TRF) Grant MRG 5180072 and a TRF Senior Research Scholar grant.

DISCLOSURES

All authors have no potential conflict of interest.

ACKNOWLEDGMENTS

We thank all personnel in the Department of Tropical Nutrition and Food Science, Faculty of Tropical Medicine, Mahidol University; Genetic Analysis Professional Center, Faculty of Medicine, Chulalongkorn University; and Center for Excellence in Molecular Genetics of Cancer and Human Diseases, Department of Anatomy, Faculty of Medicine, Chulalongkorn University.

Footnotes

  • 1 The online version of this article contains supplemental material.

REFERENCES

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
View Abstract