Learn more: PMC Disclaimer | PMC Copyright Notice
Introduction—Epiphanies in Epigenetics
Abstract
The combinatorial pattern of DNA and histone modifications and their associated histone variants constitute an epigenetic code that shapes gene expression patterns by increasing or decreasing the transcriptional potential of genomic domains. The epigenetic coding status, at any given chromosomal location, is subject to modulation by noncoding RNAs and remodeling complexes. DNA methylation is associated with histone modifications, particularly the absence of histone H3 lysine 4 methylation (H3K4me0) and the presence of histone H3 lysine 9 methylation (H3K9m). We briefly discuss four protein domains (ADD, CXXC, MBD, and SRA), and the functional implications of their architecture in linking histone methylation to that of DNA in mammalian cells. We also consider the domain structure of the DNA methyltransferase DNMT1, its accessory protein UHRF1, and their associated proteins. Finally, we discuss a mechanism by which methylation of DNA and of histones may be coordinately maintained during mitotic cell division, allowing for the transmission of parental methylation patterns to newly replicated chromatin.
I. What is an Epigenetic Code?
All cells face the problem of controlling the amounts and timing of expression of their various genes. In some cases, this control involves nonpermanent but relatively long-term and heritable modifications to the chromatin. Such modifications, that do not change the DNA sequence, are referred to as “epigenetic.” The resulting epigenetic effects maintain the various patterns of gene expression in different cell types.1 Epigenetic modifications include DNA methylation and posttranslational histone modifications. Nucleosomes consist of approximately 146 bp of DNA wrapped approximately 1.8 times around a histone octamer that is evolutionarily conserved.2 The combinatorial pattern of DNA and histone modifications constitutes an epigenetic code that shapes transcriptional patterns (Fig. 1A). The coding modification is “written” by sequence- and site-specific enzymes, and “interpreted” by effector molecules that mediate the assembly of higher-order chromatin structures involving remodeling complexes, histone variants, and noncoding RNAs (ncRNAs).12
A. DNA Methylation
In mammals and other vertebrates, DNA methylation occurs at the C5 position of cytosine, resulting in 5-methylcytosine (5mC), mostly within CpG dinucleotides (Fig. 1B). The Dnmt enzymes use a conserved mechanism13 that has been studied best in the bacterial 5mC MTase M.HhaI.14–18 Briefly, this mechanism involves MTase binding to the DNA, eversion of the target nucleotide so that it projects out of the double helix (base flipping), covalent attack of a conserved Cys nucleophile on cytosine C6, transfer of the methyl group from S-adenosyl-l-methionine (AdoMet) to the activated cytosine C5, and the various release steps.
B. DNA Hydroxymethylation
5-Hydroxymethylcytosine (5hmC) has long been noted in bacterial phage DNA.19–23 Its presence in mammalian cells24 was originally believed to be a byproduct of oxidative DNA damage.25 Recently, Kriaucionis and Heintz found that significant fractions (~40%) of cytosine nucleotides correspond to 5hmC, the amount of which inversely correlates with 5mC and nuclear heterochromatin, in isolated and relatively homogenous populations of Purkinje and granule neuronal nuclei of adult mouse brains.26 Even more fascinating, a conserved mammalian-specific family of TET (ten–eleven translocation) proteins was identified that converts 5mC to 5hmC (Fig. 1B).10 One of these proteins, TET1, is fused to the MLL protein in a subset of acute myeloid leukemia patients. Overproduction of TET1 in human cells led to the appearance of 5hmC. A concomitant reduction in DNA 5mC suggests that 5hmC results from enzymatic oxidation of 5mC. 5hmC was also detected in ES cells, and its amount decreased with RNAi knockdown of TET1. The surprising finding of a 5mC oxidation pathway raises numerous questions, such as whether oxidation of 5mC is an important epigenetic modification. 5hmC could be an end product, or an intermediate in active DNA demethylation, as supported by the existence of a 5hmC DNA excision repair glycosylase.27 It is intriguing that the bacterial 5mC MTase M.HhaI can promote the reverse reaction in vitro—the removal of formaldehyde from 5hmC to yield the unmodified cytosine11 (Fig. 1C). New lines of research will likely be catalyzed by the presence of 5hmC in mammalian DNA.
C. Histone Posttranslational Modifications
Histones are subject to a multitude of posttranslational modifications, including acetylation, methylation, ubiquitination and sumoylation of lysines, and methylation of arginines.28 There is extensive crosstalk among histone modifications.29 As a result, different histone markings often have nonadditive effects. Structural and biochemical characterization of more of the players in histone modification are beginning to clarify the underlying mechanisms of crosstalk, and the interplay with other pathways, such as chromatin remodeling and DNA methylation and repair.30–33 Mechanistically, crosstalk occurs when one or more binding modules and catalytic domains reside in the same complex or polypeptide, allowing coordination of different activities. Crosstalk can occur prior to catalysis, in which case the recognition of one mark (or its absence) can serve to recruit an enzyme to its substrate in the generation or removal of a second mark. For example, phosphorylation of H3 serine 10 by Ipl1/aurora kinase interferes with methylation of H3 lysine 9 by SUV39H1, whereas acetylation of lysine 9 or lysine 14 enhances serine 10 phosphorylation by Ipl1/aurora kinase.34 Ubiquitination of lysine 123 of histone H2B by Rad6 regulates methylation of H3 lysines 435 and 79 in Saccharomyces cerevisiae.36,37 PHF8 and KIAA1718 harbor two domains in their respective N-terminal halves: a PHD domain that binds trimethylated histone H3 lysine 4 (H3K4me3; a modification associated with transcriptional activation), and their linked Jumonji domains that remove methyl marks associated with transcriptional repression (di- and monomethylated histone H3 lysine 9, H3K9me2/1) via PHF8,38–41 H3K27me2/1 via KIAA1718,38,42,43 or H4K20me1 via PHF8.44,45
D. Remodeling Complexes
In eukarya, the packaging of DNA into chromatin provides a barrier that limits access to the genome. Condensed chromatin is refractory to processes acting on DNA, including transcription, replication, and repair.46,47 ATP-dependent chromatin-remodeling machinery can overcome this barrier—or strengthen it in silencing processes—by sliding or displacing nucleosomes and by altering their histone content.48
The modifications that constitute the histone code show significant interplay with the ATP-dependent chromatin-remodeling machinery. The PHD finger of a component of the nucleosome remodeling factor (NURF) complex binds H3K4me3.49,50 INO80 and SWR1 chromatin-remodeling complexes are recruited through H2A phosphoryation at serine129,51 and SWI/SNF associates with acetylated histones via bromodomains.52–54 It is unclear to what extent histone code modifications and their binding modules serve as adaptors for the chromatin-remodeling machinery, and to what extent they regulate chromatin structure directly.
E. Histone Variants
The H3 variants (H3.1, 3.2, and 3.3) differ at just five positions.55,56 Importantly, the N-terminal 31 residues are identical, so there is no difference in the immediate contexts of K4 and K9; but other residues in the core histones affect H3 methylation—at least of K4.57 Even before incorporation into nucleosomes, some methylation at H3K9 has been reported58; and this methylation is substantially more abundant on H3.1 than on H3.3 (which may only play a role in gametogenesis59).
F. Noncoding RNAs (Including lincRNAs Like HOTAIR)
ncRNAs are transcribed from DNA but are not translated into proteins. Many are functional, and are involved in the processing and regulation of other RNAs. Small ncRNAs, such as microRNAs (miRNAs) and short-interfering RNAs (siRNAs that cause RNA interference RNAi), are involved in the regulation of target mRNAs and chromatin. The ncRNAs can increase or decrease transcription, inhibit translation, or guide methylation.60–63 Longer ncRNAs (typically >200 nt) have also been implicated in gene regulation64,65 and have roles in epigenetic processes.66 Some of these roles have been known for some time (e.g., X-chromosome inactivation67–70 and gene imprinting71–74). Further, long ncRNAs are components of histone-modifying complexes. The male-specific lethal (MSL) complex, responsible for dosage compensation in Drosophila, is an example of a class of complexes with five protein subunits (including a histone acetyltransferase, MOF) and one of the two ncRNAs, RNA on the X1 and 2 (rox1 and rox2).75 HOTAIR, a long ncRNA, serves as a scaffold for at least two distinct histone modification complexes.76 A 5′ domain of HOTAIR binds polycomb repressive complex 2 (PRC2) for histone H3 lysine 27 methylation, whereas a 3′ domain of HOTAIR binds the LSD1/CoREST/REST complex for coupled lysine 4 demethylation.
II. Inheritance
It is well accepted that DNA methylation patterns are replicated in a semiconservative fashion during cell division, via mechanisms discussed below (and in subsequent chapters). However, one of the fundamental unresolved questions is how, and indeed whether, histone modifications are similarly “inherited.” Recent work has shed light on this question. Through a combination of kinetic isotope labeling and mass spectroscopy studies, Xu et al. showed that histone H3.1–H4 tetramers remain intact and are partitioned evenly during DNA replication-dependent chromatin assembly.77 Considering that the well-studied lysine methylation events reside on histone H3 (K4, K9, K27, K36, K79) or H4 (K20), this result suggests a model in which histone methylation patterns may be copied onto newly deposited tetramers from neighboring parental nucleosomes. Indeed, many of the SET domain histone methyltransferases contain intrinsic or associated “reader” domains that recognize the same mark that they generate, theoretically allowing the copying of these marks from old to new nucleosomes. For example, G9a/GLP catalyzes formation of H3K9me1/2 and contains an ankyrin repeat domain that binds H3K9me1/2.78 Likewise, SUV39H1/2, the H3K9me3 writer interacts with HP1, the H3K9me3 reader.79 Similarly, yeast Clr4 methylates H3K9 and contains a chromodomain that binds H3K9me3.80 These observations suggest coordinated mechanisms of deposition and replicative transmission of repressive chromatin marks.
The identification of UHRF1 and its potential role in modulating the specificity of Dnmt1 for hemimethylated CpG sites (see Chapter by Jafar Sharif and Haruhiko Koseki) adds another layer to the mechanisms ensuring the faithful transmission of epigenetic information during DNA replication. Hemimethylated CpG sites, where only one DNA strand is methylated, are transiently generated during semiconservative DNA replication. Methylation patterns are faithfully preserved with somatic cell division by a “maintenance” methylation reaction. UHRF1 has the potential to interact with both hemimethylated CpGs (via the SRA domain) and H3K9me2/3 (via the Tudor domain), and is known to interact with a wide variety of epigenetic regulators including Dnmt1,81,82 the H3K9 di- and monomethyltransferase G9a,83 and a histone acetyltransferase Tip60.84 Thus it is possible that UHRF1 and the proteins in this larger complex play a more central role in coupling the transmission of DNA and histone (H3K9 in particular) methylation during mitotic cell division.
III. Modularity of Epigenetic Modifiers
One broad emerging theme is that a web of interactions tightly coordinates the modification of a segment of DNA and its associated histones, particularly histone H3. This section focuses on four protein domains (ADD, CXXC, MBD, and SRA) in characterizing the functional links between histone and DNA modification in mammalian cells. In particular, we consider here the relationship between DNA CpG methylation and histone H3 methylation on lysines 4 and 9.85–91 DNA methylation and histone lysine methylation are intimately connected with one another.85,87–89 In fact, genome-scale DNA methylation profiles suggest that DNA methylation is correlated with histone methylation patterns.86 Specifically, DNA methylation is associated with the absence of H3K4 methylation (H3K4me0) and the presence of H3K9 methylation; by comparison, it has little correlation with methylation of H3K27.90 In vivo studies support a molecular link between the mechanisms that maintain DNA methylation and H3K9 methylation. Studies in Neuropora and Arabidopsis have shown a strict dependence of DNA methylation on the H3K9 methyltransferases Dim-5 and KRYPTONITE.92–94 Functional linkage of the mechanisms defining genomic patterns of DNA methylation, H3K4 methylation, and H3K9 methylation is underscored by the finding that treatment with a DNA methyltransferase inhibitor (5-aza-2′ -deoxycytidine or 5azadC) leads to depletion of DNA methylation, with concomitant loss of H3K9 methylation and increase in H3K4 methylation.95
A. ADD Domains of Dnmt3L/Dnmt3a/Dnmt3b Link DNA Methylation to Unmethylated H3K4
DNA methylation in mammals is coordinately established and maintained by two DNA methyltransferase families, the so-called de novo enzymes of the Dnmt3 family and the “maintenance” methyltransferase Dnmt1. The Dnmt3 family includes two active de novo Dnmts, Dnmt3a and Dnmt3b, and one regulatory factor, Dnmt3-like protein (Dnmt3L; Fig. 2A). Dnmt3a and Dnmt3b have similar domain arrangements: both contain a variable region at the N-terminus, followed by a PWWP domain, an ADD (ATRX-Dnmt3-Dnmt3L) domain, and a C-terminal catalytic domain. The amino acid sequence of the Dnmt3L ADD domain is very similar to those of Dnmt3a and Dnmt3b (Fig. 2A). Dnmt3L associates in vivo not only with Dnmt3b and Dnmt3a2 (a shorter isoform of Dnmt3a99) but also with the four core histones.100 Peptide interaction assays and cocrystallization of Dnmt3L with the amino tail of H3 showed that Dnmt3L specifically interacts with the amino terminus of histone H3, only when H3K4 is not modified (H3K4me0).100 These observations suggest that Dnmt3L acts as a sensor for H3K4 methylation, such that H3K4m0 is bound by Dnmt3L, that consequently promotes de novo DNA methylation by docking Dnmt3a to the nucleosome (Fig. 2B).
Histone–Dnmt3L–Dnmt3a–DNA interactions have recently been studied in the budding yeast S. cerevisiae,101 which has no detectible native DNA methylation102 and lacks Dnmt orthologs. Introduction of the murine methyltransferase Dnmt1 or Dnmt3a into yeast leads to detectible, but extremely low levels of DNA methylation.103 In contrast, a substantially higher level of de novo methylation could be achieved in yeast by coexpressing murine Dnmt3a and Dnmt3L.101 This induced DNA methylation was found preferentially in heterochromatic regions where H3K4 methylation is rare. When genes for components of the H3K4-methylating complex COMPASS/Set1 were disrupted in the context of Dnmt3a/Dnmt3L overexpression, a greater level of genomic DNA methylation was observed. Deletions or targeted mutations in the ADD of Dnmt3L inhibited both global DNA methylation and the ability of Dnmt3L to associate with an H3K4me0 peptide. These same Dnmt3L mutants failed to restore normal DNA methylation to a specific promoter when introduced into ES cells from Dnmt3L−/− mice.101 That H3K4 methylation status plays an important role in establishing DNA methylation is further supported by studies with mammalian LSD1 and LSD2, two related lysine-specific demethylases whose substrates include mono- and dimethylated H3K4 (H3K4me1/2; Fig. 2B). LSD1 and LSD2 are essential for maintaining global DNA methylation,104 or establishing maternal DNA genomic imprints,105 respectively. Indeed, disruption of LSD1 results in earlier embryonic lethality and a more severe hypomethylation defect than disruption of the Dnmt genes themselves.104
The above data suggest that perhaps Dnmt3L binds to H3K4me0 (via its ADD domain) and recruits Dnmt3a to regions of chromatin where H3K4 is unmethylated. Such a model could explain part of the puzzle of how DNA methylation patterns are established de novo during embryonic and germ cell development, when both Dnmt3a and Dnmt3L are expressed.106 However, while Dnmt3a and Dnmt3b expression occurs in somatic cells, Dnmt3L is expressed poorly if at all in differentiated cells. This raises the questions of how de novo DNA methylation is restricted in somatic cells, whether Dnmt3a and Dnmt3b alone are capable of discriminating H3K4 methylation status, and (if so) the structural basis for that discrimination. To this end, recent work by Jeong et al. showed that in nuclei from HCT116 human colon cancer cells (which do not express DNMT3L), almost all of the cellular DNMT3a and DNMT3b (but not DNMT1) were associated with nucleosomes.107 Chromatin binding of DNMT3a and DNMT3b required an intact nucleosomal structure, though no other chromatin factors (such as the HP1 or MBD proteins), suggesting that DNMT3a and DNMT3b themselves can interact directly with chromatin components in addition to DNA. Further, recent in vitro studies indicate that the ADD domains of Dnmt3a or Dnmt3b possess the same H3 tail binding specificity as that of Dnmt3L.108 Indeed, a structure of the Dnmt3a ADD domain in complex with an amino-terminal tail peptide from histone H3 indicates that the ADD domain is sufficient to recognize H3K4me0.109
B. The CXXC Domain Links H3K4 Methylation to Unmethylated CpGs
In humans, there are at least eight histone lysine methyltransferases that act on H3K4. These include products of the mixed lineage leukemia (MLL) genes, MLL1 through 5, hSET1a, hSET1b, and ASH1. MLL1/SET1-associated methyltransferase activity appears to be functional only in the context of multiprotein complexes; characterization of these reveals distinct multiprotein complexes for each of the eight, though with several shared components.110 The MLL family plays an important role in embryonic development and is necessary for methylation of H3K4 at a subset of genes in the human and mouse genomes, particularly the HOX gene clusters.111 Translocations involving MLL genes are involved in the etiology of myeloid and lymphoid leukemias. Considering the aforementioned inverse relationship between H3K4 methylation and DNA methylation, it is interesting that disruption of theMLL1 gene in mice results in loss of both H3K4 methylation and de novo DNA methylation at some Hox gene promoters,112,113 suggesting that MLLs directly or indirectly (through H3K4 methylation) prevent DNA methylation or perhaps stabilize unmethylated DNA. In fact, MLL proteins contain CXXC, an evolutionarily conserved domain that mediates selective binding to unmethylated CpGs (Fig. 3A).114–116 This interaction has now been confirmed by a solution structure of an MLL1-CXXC domain complexed with unmethylated DNA.117
The Set1 H3K4 methyltransferases also appear to interact with unmethylated DNA, although in this case it is via an accessory protein, as with Dnmt3a/Dnmt3L (discussed above). Set1 lacks a CXXC domain, but interacts directly with a protein that contains one—the CXXC finger protein 1 (Cfp1)118–120 (Fig. 3A). High-throughput sequencing of Cfp1-bound chromatin revealed a notable concordance between H3K4me3 and Cfp1 at unmethylated CpG islands, in the mouse brain.121 At loci that exhibit allele-specific DNA methylation (e.g., imprinted loci, Xist gene), Cfp1 bound specifically to the unmethylated allele. Depletion of Cfp1 resulted in a marked reduction in H3K4me3 genome wide. The targeting of Cfp1 to CpG islands was independent of promoter activity as the insertion of an unmethylated CpG-dense construct into the genome of ES cells was sufficient to nucleate Cfp1 binding and H3K4me3. This suggests that unmethylated CpGs recruit Cfp1, and the associated methyltransferase Set1 then creates new H3K4me3 marks on the local chromatin.
Like the histone H3K4 methyltransferases of the MLL/SET1 family, the Jumonji domain-containing histone demethylase JHDM1a (known as KDM2A) also has a CXXC domain122 (Fig. 3A). Recent work indicates that, like Cfp1, JHDM1 is recruited to unmethylated CpG islands on a genome-wide scale, and that this is dependent on the CXXC domain.123 The localization to CpG islands was independent of promoter activity and gene expression levels, and correlated with the selective depletion of H3K36me1/2 within the CpG island but not surrounding regions or the bodies of genes; knockdown of KDM2A resulted in the selective accumulation of H3K36me2 in these regions. Consistent with a role for DNA methylation in determining the localization of CXXC proteins in Dnmt1−/− mice, KDM2A was mislocalized to pericentric heterochromatin. Further, the lack of methylation alone does not appear to be sufficient for KDM2A recruitment in vivo, as KDM2A does not localize to unmethylated non-CpG island promoters. Although in vitro studies suggest that CXXC domains can bind a single CpG site with micromolar affinity, both the Cfp1 and KDM2A studies suggest that the targeting of CXXC proteins in vivo is dependent on CpG density as well as methylation status. It is possible that these proteins oligomerize and form nucleoprotein filaments on CpG-dense DNA, in a manner similar to that described for Dnmt3a and Dnmt3L.124
CXXC domains are also found in DNMT1, the enzyme responsible for postreplication maintenance of 5-methyl-Cyt,125MBD1, themethyl-CpG-binding protein,126 and Tet1, a Jumonji-like 2-oxoglutarate- and Fe(II)-dependent enzyme that catalyzes conversion of 5mC to 5hmC10 (Fig. 3A). Interestingly, a recurrent t19,20(q22;q23) translocation has been described in acute myelogeous leukemias, and results in a fusion transcript that juxtaposes the first six exons of MLL(containing AT hook and CXXC) to the C-terminal one-third of TET1 thus “replacing” the TET1 CXXC with the MLL CXXC.127,128 Whether this leads to altered targeting of methyl hydroxylation remains to be determined.
C. A Role for MBDs in the Linkage Between Histone Modification and DNA Methylation
The methyl-CpG-binding domain (MBD) is present in a family of proteins conserved throughout the eukaryotic lineage. This domain, in some but not all cases, confers the ability to bind methylated CpGs. Mammals have five well-characterized members of this family, each with unique biological characteristics (reviewed in Ref. 129; Fig. 4A). Of particular interest are recent reports indicating that in zebrafish, MBD4 (containing both a N-terminal MBD and a C-terminal thymine glycosylase domain,130) and the cytidine deaminase AID cooperate to demethylate DNA4 (see also Chapter byMary G. Goll and Marnie E. Halpern). Consistent with a role in DNA demethylation, erasure of DNA methylation in primordial germ cells is less efficient in AID-deficient animals,131 and AID is required to demethylate pluripotency genes during reprogramming of the somatic genome in ES cell fusions.132 It is noteworthy that phosphorylation of MBD4 enhances DNA demethylation.133
A particularly interesting observation, for the purposes of this review, is the fact that MBD1 forms a stable complex with SETDB1,134,135 an H3K9me3 methyltransferase responsible for silencing of endogenous retroviruses136–138 as well as the Suv39h1/HP1 complex,139 a heterochromatin-specific H3K9me3 writer and reader. SETDB1 also contains an intrinsic putative MBD domain with two conserved DNA-interacting arginine residues (Fig. 4B) that make direct contact with DNA in the structures of the MBD domain from MBD1140 and MeCP2.141 It remains to be seen whether the putative MBD domain of SETDB1 is similarly able to selectively bind methylated DNA. The intrinsic or associated coupling of a DNA methylation “reader” with H3K9me3 “writers” implies an interdependent mechanism for the propagation or maintenance of these marks.
As MBD1 contains two DNA-binding domains—an MBD domain (Fig. 4), which recognizes methylated CpG, and a CXXC domain (Fig. 3), which binds unmethylated CpG—it is unclear whether these two domains function independently to facilitate the recruitment of MBD1 to repressive complexes or active complexes, or if they cooperate in some way. Recently, Clouaire et al. found that the MBD domain of MBD1 binds more efficiently to methylated DNA within a specific sequence context, and that a functional MBD domain is necessary and sufficient for recruitment of MBD1 to these loci, while DNA binding by the CXXC domain is largely dispensable.142
D. UHRF1 Links Hemimethylated CpGs to Histone Modifications During Replication
UHRF1 harbors five recognizable functional domains (Fig. 2A): a ubiquitin-like domain (UBL) at the N-terminus, followed by a tandem Tudor domain that binds H3K9me2/3,143,144 an ADD-like domain that binds the histone H3 tail,145,146 a SET- and RING-associated (SRA) domain that binds hemimethylated CpG-containing DNA,81,82,147–149 and a really interesting new gene (RING) domain at the C-terminus that may endow UHRF1 with E3 ubiquitin ligase activity for histones.145 As discussed above, Dnmt1 contains a CXXC domain (Fig. 3A) that binds unmethylated CpGs.125 The interaction with UHRF1 provides an SRA domain that recognizes hemimethylated CpGs, the substrate of Dnmt1. It is tempting to speculate that these two DNA-binding domains function synergically to maintain the fidelity of DNA methylation during replication, by preventing Dnmt1 from modifying unmethylated CpGs (through the masking of such sites by the CXXC domain) and directing methylation to newly synthesized hemimethylated CpG sites (via the SRA domain). Indeed, molecular modeling suggests that the Dnmt1 catalytic domain and UHRF1 SRA domain are very unlikely to bind simultaneously to the same hemimethylated CpG site.148 Somehow, the SRA domain must be displaced from the site to allow methylation by Dnmt1.150
We therefore suggest that the SRA–DNA interaction (through recognition of and flipping of the parental strand 5mC out of the helix147–149) serves as an anchor to keep UHRF1 at the hemimethylated CpG site where it then recruits Dnmt1 to methylate the opposing CpG in the daughter strand, which maintains the fidelity of DNA methylation.
Acknowledgments
The work in the authors’ laboratories is currently supported by the U.S. National Institutes of Health (GM049245-17, GM068680-06, and GM092035-01 to X. C.) and the National Science Foundation (MCB-0964728 to R. M. B.). X. C. is a Georgia Research Alliance Eminent Scholar. This chapter is modified and updated from Ref. 91, with permission from American Chemical Society Publications.