news and v iews

1048 volume 43 | number 11 | november 2011 | nature genetics

news and v iews

interactions whose signal is subtle (the prob-ability of trans contact is overall much lower than that of intrachromosomal (cis) contact6). In fact, the authors noticed that Hi-C data sets derived from two different enzymes produced trans contact maps that were not correlated, even though, in theory, the choice of restriction enzyme should not change the biological con-clusions. They identified four different sources of bias, among which the local GC content in ligated DNA was the dominant factor behind the enzyme-specific profiles. Here, a simple successive correction of biases would not work due to the intertwined error structures. Rather, they used a probabilistic model to account for existing biases by a maximum-likelihood method. The corrected trans-contact map is now concordant between the two replicate data sets from different restriction enzymes. From the trans-contact map, three clusters of genomic regions emerge: active loci, inac-tive and centromere-proximal loci, and inac-tive and centromere-distal loci. This refines the two spatial compartments reported in the original Hi-C study2. As an added assurance, the contiguity along the linear chromosomes was recovered solely from the trans-interaction patterns. However, it is unclear whether the centromere-based segregation of repressed

chromatin reflects an interphase organization or an artifact of actively dividing blasts.

Interchromosomal interactionsHistorically, interchromosomal interactions have been more elusive than cis contacts7,8. For one thing, the signal strength is modest regardless of the experimental assay, whether in a single-cell analysis or a 3C variant3,8. Furthermore, it has been difficult to determine the relevant epigenetic features that influence trans interactions, as multiple marks are appar-ently associated with contact regions. Discovery of the most directly associated genomic feature would undoubtedly shed light on the mecha-nisms driving nonrandom organizations of the nucleus. To address this, Yaffe and Tanay looked for global correlates of organization by relating the corrected Hi-C data to available epigenomic profiles from the ENCODE human lymphoblast database, including DNase I hypersensitivity, histone marks, RNA polymerase II occupancy and CTCF binding. Extensive enrichment analysis and comparisons singled out the pres-ence of DNase I hypersensitive sites (DHSs) as an important hallmark of trans interactions (Fig. 1). The DHS-associated effect on con-tact probability persists largely independently of other markers, whereas the other features’

effects depend on DHS presence. Because DHSs often represent regulatory sites that are created and/or maintained by transcription fac-tors and other chromatin-binding proteins9–11, the nuclear architecture may be shaped col-lectively by the cell type–specific ensemble of chromatin-modifying proteins.

Our knowledge is still based on a small number of studies. However, with a well-formulated methodology now in place, the genomics community is primed to extract robust nuclear architectural information from Hi-C experi-ments on more cell types and states.

COMPETING FINANCIAL INTERESTSThe authors declare no competing financial interests.

1. Irizarry, R.A., Wu, Z. & Jaffee, H.A. Bioinformatics. 22, 789–794 (2006).

2. Lieberman-Aiden, E. et al. Science 326, 289–293 (2009).

3. Yaffe, E. & Tanay, A. Nat. Genet. 43, 1059–1065 (2011).

4. Dekker, J. Nat. Methods 3, 17–21 (2006).5. Simonis, M., Kooren, J. & de Laat, W. Nat. Methods 4,

895–901 (2007).6. Hakim, O., Sung, M.H. & Hager, G.L. Curr. Opin. Cell

Biol. 22, 305–313 (2010).7. Kocanova, S. et al. PLoS Genet. 6, e1000922

(2010).8. Hakim, O. et al. Genome Res. 21, 697–706 (2011).9. Hesselberth, J.R. et al. Nat. Methods 6, 283–289

(2009).10. John, S. et al. Nat. Genet. 43, 264–268 (2011).11. Biddie, S.C. et al. Mol. Cell 43, 145–155 (2011).

a transposon in tb1 drove maize domesticationMiltos Tsiantis

a new study shows an inserted retroelement in the regulatory sequences of the maize tb1 gene, which controls shoot branching, was the target of human selection during the domestication of maize from its wild relative teosinte. The insertion allele was already present at low frequency in teosinte populations before selection, highlighting the significance of standing genetic variation in the evolution of morphological diversity.

Miltos Tsiantis is in the Department of Plant Sciences at the University of Oxford, Oxford, UK. e-mail: miltos.tsiantis@plants.ox.ac.uk

changes contributed to the differences in form between maize and teosinte. In this issue of Nature Genetics, Doebley and colleagues make a key advance in understanding the precise genetic basis for the origin of maize by showing that a transposon insertion 60 kb upstream of the tb1 ORF underlies species-specific tb1 expression and consequent diversification of morphology5.

Hunting causal variantsTo delimit the genetic interval that contains the variants responsible for morphological diversity, the authors followed two comple-mentary approaches. First, they mapped recombination breakpoints in maize-teosinte introgression lines to determine the pheno-typic consequences of harboring different

maize and teosinte. They crossed the two spe-cies and found that allelic variation at five loci accounted for their major differences in form2. One of these genes was teosinte branched 1 (tb1), which encodes a transcriptional regu-lator involved in growth repression. In maize, tb1 expression is elevated relative to teosinte, correlating with repressed branch outgrowth3. A signature of selection was also identified at the tb1 locus in the form of a selective sweep, suggesting that tb1 was targeted by human selection4. That is, the tb1 gene showed much less nucleotide diversity in maize relative to teosinte compared to genes that have not been targeted by human selection. Reduced levels of polymorphism were found in upstream regions but not in the coding sequence of tb1, suggesting that regulatory

A major challenge in biology is to understand the origins of morphological diversity—how form changes through evolution. As Darwin noted, domestication offers valuable insight into this problem by providing a direct path between ancestral and descendant species1. For example, maize was domesticated from its wild ancestor teosinte, and this process resulted in reduced branching (Fig. 1). Twenty years ago, in a search for the genes underpin-ning maize evolution, John Doebley and col-leagues conducted a simple, yet powerful, experiment that exploited the interfertility of

npg

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

news and v iews

nature genetics | volume 43 | number 11 | november 2011 1049

portions of the teosinte tb1 upstream region in maize. Second, they mapped the extent of the selective sweep that fixed tb1 alleles in maize by using polymorphism data from 300 diverse maize and teosinte strains. These approaches allowed them to narrow the causative variant to an 11-kb DNA fragment positioned 60 kb upstream of the tb1 ORF that contained four sequence variants fixed in the maize allele: two single-nucleotide polymorphisms (SNPs) and two transposon insertions. The authors then assayed the function of these four variants in transcriptional assays in maize leaf protoplasts. They found that only one of them, a Hopscotch retroelement, resulted in a twofold increase in transcription, which is comparable to the dif-ference in tb1 expression seen between maize and teosinte. Thus, the authors conclude that the Hopscotch insertion enhances tb1 tran-scription, thereby contributing to the differ-ence in plant architecture between maize and teosinte. It will be interesting to test whether these results can be recapitulated in trans-genic plants and to determine at what stage of development and in which cells the Hopscotch retroelement influences tb1 transcription.

These experiments might help to identify the precise point at which domestication influenced the developmental program controlling branching in maize.

Standing variation as raw material for diversityMolecular dating of the Hopscotch tb1 allele indicates that it was present in teosinte as a standing variation and that human selection rapidly increased its frequency, thereby con-tributing to the development of modern maize. Thus, in this instance standing variation (not new mutation) fueled the domestication process. From a theoretical perspective, it is interesting that the authors were able to iden-tify a selective sweep at tb1, as selection from standing variation is often difficult to detect. This difficulty arises because standing varia-tion does not necessarily leave a characteristic signature of reduced polymorphism around the selected allele. By definition, the variant existed in the population before selection and had time to recombine into diverse neighbor-ing sequences. However, if the selected allele is present at low frequency and selection is

strong, as it is in domestication, then a sweep can be detected6.

Theory aside, a similar instance of mor-phological evolution via selection acting on standing variation was previously described in stickleback fish. Low-frequency Ectodysplasin (Eda) alleles that were present in heterozygotes in marine populations experienced strong selection during colonization of freshwater environments, resulting in a major morpho-logical shift from marine fish with many body armor plates to freshwater fish with few armor plates7 (Fig. 1). Standing genetic variation can therefore lead to morphological evolution in both plants and metazoa and in response to both artificial and natural selection. The find-ing that a modest degree of regulatory variation at tb1 contributed to the evolution of maize architecture is consistent with recent results suggesting that regulatory changes at key developmental genes often underlie morpho-logical variation between metazoan taxa—whether they diverged thousands or several million years ago8. Thus, genetic studies are helping to define unifying principles for the generation of morphological diversity in both plants and animals, despite the independent evolution of multicellularity and considerable differences in their life cycles.

Studer et al. present a clear picture of how maize was domesticated by events operating at different scales: biological and cultural5. First, a retroelement insertion created tb1 alleles with altered gene expression, and these alleles then experienced intense selection applied by Indian tribes in Mesoamerica. Results from two other grasses, pearl millet and barley, point to tb1 as a possible hotspot9 that was repeatedly subject to selection during domestication10,11, suggesting predictability in the evolution of altered branch-ing. Intriguingly, Deng et al. provide data that indicate that the maize tb1 haplotype is not fully fixed, as the Hopscotch insertion is present in 95% of the maize chromosomes assayed, whereas correspondingly ‘maize-like’ sequence variants can be found in <5% of teosinte chromosomes. It will be interesting to investigate to what degree such divergent alleles influence morphology and whether epistatic interactions can partially com-pensate for divergence from the species-specific allelic configurations described here.

In summary, this study gives us new insight into the genetic basis for evolutionary change by show-ing that a retroelement insertion caused morpho-logical diversity by exerting relatively modest effects on the regulation of a developmentally important transcription factor. This event also had considerable impact on human culture by contributing to the domestication of an important grain crop. These insights into the precise mecha-nism underlying evolution of a key morphological

Natural selectionDomestication

Ancestral teosinte populations were highly branched and contained a low frequencytb1 variant with a Hopscotch insertion.

The tb1 variant with a Hopscotch insertion swept to high frequency in modern maize

populations, yielding plants with few branches.

Ancestral stickleback marine populations had full body armor

and harbored a low frequency Eda variant.

The Eda variant swept to high frequency in freshwater populations

yielding a reduced-armor morphology.

Predominant form

Predominant form

Predominant form

Predominant form

Figure 1 Standing genetic variation drives morphological change in maize domestication and stickleback natural selection. Left, a tb1 allele containing a Hopscotch insertion was rare in ancestral teosinte. Plants carrying this allele are depicted as green in the background of pseudo-colored purple teosinte plants. The Hopscotch allele was a target of human selection during maize domestication, resulting in increased apical dominance (reduced branching) compared to teosinte5. Right, transition from marine to freshwater environments is correlated with selection for an Eda variant present at low frequency in marine populations (the half-blue and half-red fish depicts such a heterozygous carrier). Fixation of this allele in freshwater populations resulted in fish with reduced body armor (blue) compared to their marine ancestors (red)7. (Stickleback drawings courtesy of David Kingsley.)

npg

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

news and v iews

1050 volume 43 | number 11 | november 2011 | nature genetics

6. Innan, H. & Kim, Y. Proc. Natl. Acad. Sci. USA 101, 10667–10672 (2004).

7. Colosimo, P.F. et al. Science 307, 1928–1933 (2005).

8. Carroll, S.B. Cell 134, 25–36 (2008).9. Stern, D.L. & Orgogozo, V. Science 323, 746–751

(2009).10. Remigereau, M.S. et al. PLoS ONE 6, e22404

(2011).11. Ramsay, L. et al. Nat. Genet. 43, 169–172 (2011).

1. Darwin, C. The Origin of Species by Means of Natural Selection (John Murray, 1859).

2. Doebley, J. & Stec, A. Genetics 134, 559–570 (1993).

3. Doebley, J., Stec, A. & Hubard, L. Nature 386, 485–488 (1997).

4. Wang, R.L., Stec, A., Hey, J., Lukens, L. & Doebley, J. Nature 398, 236–239 (1999).

5. Studer, A. et al. Nat. Genet. 43, 1160–1163 (2011).

trait also underscore the importance of function-based approaches for understanding how diverse plant and animal forms were generated.

Corrected after print 27 April 2012.

COMPETING FINANCIAL INTERESTSThe author declares no competing financial interests.

in the local DNA sequence itself–that is, in cis. Several of the tested promoters con-tained multiple CGIs and were therefore GC-rich and hypomethylated in the genome. However, base composition per se was not the sole determining feature, as GC-rich sequences derived from bacteria mostly failed to adopt a hypomethylated state.

If base composition was not sufficient to determine methylation status (although it evi-dently helped), could binding sites for trans-acting factors again be the key? The answer is yes. The ability to exclude DNA methylation was compromised for promoter fragments bearing mutations within known factor-binding sites. Importantly, several factors in addition to Sp1 were implicated, including the Rfx winged-helix transcription factors. The discrete ‘methylation-determining regions’ (MDRs) also turned out to be portable, as they imposed an unmethylated state on hetero-logous flanking DNA.

factor–binding sites at a promoter, specifically those for Sp1, are essential to protect the entire CGI from DNA methylation3,4. It was difficult to generalize from these studies, particularly because both the groups that performed them coincidentally investigated the same house-keeping gene (albeit from different rodents). Lienert et al. have taken this line of inquiry to another level by inserting over 50 differ-ent DNA sequences into mouse embryonic stem cells and analyzing their DNA methyla-tion status in detail. It is well known that the functional state of an inserted DNA sequence is critically dependent on the chromosomal environment in which it happens to land5. Lienert and colleagues nicely eliminated this source of variability by targeting all insertions to the same precise location within a trans-criptionally quiescent region of the genome1. They also insured that the selectable marker gene, whose transcription might other-wise interfere with interpretation, was lost during integration.

Cis but also transThe initial experiment in Lienert et al. involved integration of ten promoter fragments that were either hypo- or hypermethylated at their natural sites in the mouse genome. All ten inserts exactly mimicked the DNA methylation pattern of the native DNA sequence, despite the alien integration site and relatively small fragment size (700–1,600 bp). This close adherence to the endogenous methylation pattern was maintained when the ES cells were differentiated along a neuronal lineage. For example, the Nanog promoter fragment showed an appropriate change from a hypo- to a hypermethylated state. All the information required to specify the DNA methylation pattern, therefore, resides

Our chromosomes are embellished by a complex array of covalent chemical signals that together make up the ‘epigenome’. These so-called epi-genetic marks broadly facilitate or stabilize local activity states within the genome, and their detailed functional significance is the subject of intense research in many labora-tories worldwide. As well as understanding how the signals are interpreted biologically, we need to know what causes this information to be written at specific chromosomal regions. The best-known epigenetic mark, and the only one known to be directly applied to unmodi-fied mammalian DNA, is cytosine methylation, which occurs in the self-complementary CG DNA sequence. The CG dinucleotide varies greatly in methylation status throughout the genome, but little is known about the mecha-nisms that underlie this pattern. An elegant study by Dirk Schübeler and colleagues now establishes several rules that govern the rela-tionship between DNA sequence and methyla-tion patterns1. The findings reinforce the view that DNA sequence itself has a dominant role in determining this feature of the epigenome, paving the way for the elucidation of the under-lying molecular mechanisms.

The most prominent features of the DNA methylome are CpG islands (CGIs), where CG dinucleotides are locally very dense and usually free of DNA methylation2. CGIs are located within most gene promoters, and their methylation insures gene silencing. Early work used a transgenic mouse assay to show that a small number of transcription

Putting the dna back into dna methylationAdrian Bird

The origin of dna methylation patterns has been a mystery for many years. a new study identifies the dna sequence itself as a key determinant and focuses attention on the role of transcription factors.

Adrian Bird is at the Wellcome Trust Centre for Cell Biology, University of Edinburgh, Edinburgh, UK. e-mail: a.bird@ed.ac.uk

TFs

DNMTDNMT

DNMT

Figure 1 Exclusion of DNA methylation from a domain (gray oval) by transcription factors (TFs, green and orange circles) bound to discrete sites within it. The unmethylated domain extends well beyond the protein-bound DNA sequences. DNA methyltransferases (DNMTs, blue circles) are prevented from accessing the protected domain by an unknown mechanism, but they can modify accessible flanking DNA with methyl groups (red circles).

npg

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

errata

nature genetics

Erratum: A transposon in tb1 drove maize domesticationMiltos TsiantisNat. Genet. 43, 1048–1050 (2011); published online 27 October 2011; corrected after print 27 April 2012

In the version of this article initially published, a study by Studer et al. was incorrectly cited as being by Deng et al. The correct sentence should read: “Studer et al. present a clear picture of how maize was domesticated by events operating at different scales: biological and cultural5.” The error has been corrected in the HTML and PDF versions of the article.

npg

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.