Copying out Our Abcs the Role of Gene Redundancy in Interpreting Genetic Hierarchies

Genomic imprinting in mammals COMMENT Outlook

14 Nicholls, R.D. et al. (1998) Imprinting in Prader–Willi and 21 Feil, R. et al. (1997) Parental chromosome-specific chromatin 28 Macleod, D. et al. (1994) Sp1 sites in the mouse Aprt gene Angelman syndromes. Trends Genet. 14, 194–199 conformation in the imprinted U2af1-rs1 gene in the mouse. promoter are required to prevent methylation of the CpG 15 Hark, A.T. and Tilghman, S.M. (1998) Chromatin conformation J. Biol. Chem. 272, 20893–20900 island. Genes Dev. 8, 2282–2292 of the H19 epigenetic mark. Hum. Mol. Genet. 7, 1979–1985 22 Schweizer, J. et al. (1999) In vivo nuclease hypersensitivity 29 Brandeis, M. et al. (1994) SP1 elements protect a CpG island 16 Szabó, P.E. et al. (1998) Characterization of novel parent- studies reveal multiple sites of parental-origin-dependent from de novo methylation. Nature 371, 435–438 specific epigenetic modifications upstream of the imprinted differential chromatin conformation in the 150 kb SNRPN 30 Kirillov, A. et al. (1996) A role for nuclear NF-␬B in mouse H19 gene. Mol. Cell. Biol. 18, 6767–6776 transcription unit. Hum. Mol. Genet. 8, 555–566 B-cell-specific demethylation of the Igk locus. Nat. Genet. 13, 17 Khosla, S. et al. (1999) Parental allele-specific chromatin 23 Lyko, F. et al. (1998) Identification of a silencing element in 435–441 configuration in a boundary/imprinting-control element the human 15q11–q13 imprinting center by using 31 Hsieh, C-L. (1999) Evidence that protein-binding upstream of the mouse H19 gene. Mol. Cell. Biol. 19, transgenic Drosophila. Proc. Natl. Acad. Sci. U. S. A. 95, specifies sites of DNA demethylation. Mol. Cell. Biol. 19, 2556–2566 1698–1702 46–56 18 Boyes, J. and Felsenfeld, G. (1996) Tissue-specific factors 24 Ng, H-H. and Bird, A. (1999) DNA methylation and chromatin 32 Matsuo, K. et al. (1999) An embryonic demethylation additively increase the probability of the all-or-none formation modification. Curr. Opin. Genet. Dev. 8, 158–163 mechanism involving binding of transcription factors to of a hypersensitive site. EMBO J. 15, 2496–2507 25 Birger, Y. et al. (1999) The imprinting box of the mouse Igf2r replicating DNA. EMBO J. 17, 1446–1453 19 Lyko, F. et al. (1997) An imprinting element from the mouse gene. Nature 379, 84–88 33 Jones, P.L. et al. (1998) Methylated DNA and MeCP2 recruit H19 locus functions as a silencer in Drosophila. Nat. Genet. 26 Hatada, I. et al. (1997) Aberrant methylation of an imprinted histone deacetylase to repress transcription. Nat. Genet. 19, 16, 171–173 gene U2af1-rs1 (SP2) caused by its own transgene. J. Biol. 187–191 20 Shibata, H. et al. (1996) Inactive allele-specific methylation Chem. 272, 9120–9122 34 Nan, X. et al. (1998) Transcriptional repression by the methyl and chromatin structure of the imprinted gene U2af-rs1 on 27 Bird, A.P. (1992) The essentials of DNA methylation. Cell 70, CpG binding protein MeCP2 involves a histone acetylase mouse chromosome 11. Genomics 35, 248–252 5–8 complex. Nature 393, 386–389F Copying out our ABCs the role of gene redundancy in interpreting genetic hierarchies

The complete sequence of the Arabidopsis genome is scheduled to be determined by the end of the year 2000. While this goal could prove to be something of a moving target (the estimated size of the genome has grown from 120 Mb to 130 Mb over the last year1), it is clear that the majority of genes required for higher plant growth, reproduction and development will have been described within this time frame. Some of the implications of this landmark achievement are already becoming clear, even though less than a half of the genome has been sequenced.

rabidopsis has one of the most simplified plant such genes via site-selected mutagenesis. The major con- Agenomes, with only limited evidence for segmental clusion emerging from studies of this sort is surprising: duplications, little repetitive DNA, and with good diploid most insertional mutants have no discernible phenotype. genetics. One of the most important observations to Sometimes, the corresponding gene might be non-func- emerge from genome studies, however, is that most tional, representing a pseudogene or other evolutionary Arabidopsis genes are not unique. Of 100 genes found relic. However, genetrap studies reveal that many between prolifera and GA1 on chromosome 4, for exam- expressed genes, when disrupted, can still lack a detectable ple, 65 are members of small multigene families. Based on phenotype2. Mutations in such genes might not be recov- these and other data emerging from the genome project, at ered because they have subtle or conditional phenotypes. least two-thirds of Arabidopsis genes have one or more Alternatively, such mutations might not be recovered closely related homologs (L. Parnell and W.R. McCombie, because multiple closely related genes encode that func- pers. commun.). This is especially true of key regulatory tion. In such cases, double, triple and sometimes even Rob Martienssen molecules such as transcription factors, receptor kinases, more redundant combinations of mutations might be [email protected] F-box proteins and cell-cycle regulators. For example, the required to reveal a mutant phenotype. Vivian Irish* MADS box gene family has at least 50 members, while What might the consequences of this redundancy be for [email protected] there might be more than 300 receptor kinases. What are developmental genetics? One important ramification is in the implications of this widespread gene duplication? the ordering of regulatory pathways by double-mutant Cold Spring Harbor It is possible that gene duplications have allowed each analyses. As any genetics textbook will point out, if single- Laboratory, Cold Spring family member to evolve a unique function, for example in mutant phenotypes are distinct and the double mutant Harbor, NY11724, USA. a specialized cell type. However, in many cases, family resembles one of the single mutants, then the mutations *Department of members have overlapping expression domains and so are generally interpreted to affect steps in a linear path- Molecular, Cellular, and might effect the same process in the same cell type. With way. If the double mutant shows an additive phenotype, Developmental Biology, the advent of the polymerase chain reaction, it has become then the two mutations are thought to affect separate, Yale University, New a relatively trivial matter to obtain insertional mutants in unrelated processes. Alternatively, the double mutant Haven, CT 06520, USA.

might display a novel, synergistic phenotype that does not AP1 encodes a product that has similarity to the resemble either of the single mutants. In this case, the MADS-box family of transcription factors8 and several mutations are thought to affect genes that act co-operatively homologs of AP1 have now been identified in Arabidopsis towards some final outcome (Fig. 1). (Refs 9, 10). Loss-of-function mutations in one of these A crucial parameter in the interpretation of such genes, CAULIFLOWER (CAL), greatly exaggerate the double-mutant analyses is the nature of the alleles phenotype of ap1 in double-mutant combinations, involved. If null alleles of two genes in the same pathway although cal mutations have no phenotype on their own6. are combined, then the double mutant will resemble one CAL has been postulated to positively regulate LFY, of the single mutants. By contrast, when weak mutant because LFY is downregulated in ap1 cal double mutants alleles of the same two genes are combined, their effects (although it is expressed normally in ap1 single are often enhanced relative to each single mutant, leading mutants)5,6. These and other observations suggest some- to a novel, often more extreme, phenotype (Fig. 1). If the what different roles for these two family members: CAL nature of the alleles is not known, considerable confusion appears to be involved in initially promoting the activity can ensue because synergistic interactions could be due to of LFY, while CAL and AP1 act in concert with LFY to the effects of null alleles in parallel pathways, or to the effect meristem identity6. However, it is possible that AP1 action of weak alleles in a single pathway3. and CAL actually have analogous roles in meristem iden- The discovery of high levels of gene duplication sug- tity6. AP1 might also activate LFY expression, but in ap1 gests that many genes might encode at least partially mutants this function is fully complemented by CAL and redundant functions. In turn, this means that the defi- only becomes apparent in the ap1 cal double mutant. nition of many gene hierarchies that are based on the The full spectrum of AP1 and CAL functions might be intepretation of double-mutant analyses might have to be obscured because there are other family members that are revised; null alleles in one member of a redundant gene still active in either ap1 or cal mutant backgrounds. For family might result only in partial elimination of the func- instance, AGL8 shows marked sequence similarity to AP1 tion encoded by that family. Therefore, double-mutant and CAL and is functionally redundant with AP1 and combinations involving such ‘null’ alleles might only CAL (Refs 10, 11; C. Ferrandiz, Q. Gu, R. Martienssen represent partial losses of function and, consequently, pro- and M. Yanofsky, unpublished). Although AGL8 expres- duce a synergistic phenotype even when the corresponding sion does not overlap with that of AP1, in an ap1 mutant genes are in fact in the same pathway (Fig. 1). background AGL8 expression expands into the AP1 An example of one such hierarchy that has received domain and can partially compensate for loss of AP1 considerable attention in the last few years involves the function. The overlapping roles of AP1, CAL, AGL8 and homeotic genes that control floral development in potentially other, as yet undescribed family members, sug- Arabidopsis. These have been grouped into meristem- gest that the complete loss of AP1-family function has not identity genes and organ-identity genes responsible for yet been described. In turn, this interpretation could pro- specifying particular floral organs (the ABC genes). vide an alternative explanation for the synergistic pheno- Meristem-identity genes, such as APETALA1 (AP1) and type produced by lfy ap1 double mutants. If null ap1 LEAFY (LFY) are required for the formation of a florally mutants represent partial loss of function for a redundant determined meristem4–6. Strong mutations in each of these gene family, then double mutants with lfy can be exagger- genes leads to a partial loss of floral meristem identity, and ated phenotypically even though they might be in the same double-mutant combinations of such alleles result in an pathway. even more-severe phenotype, suggesting that AP1 and Redundancy might also cloud the interpretation of how LFY act in parallel pathways to effect the specification of a the ABC genes act to specify organ identity. In a now clas- floral meristem5,7. sic study, analyses of double mutants led to a model in

FIGURE 1. Interpretation of double-mutant combinations

Actual pathwayDouble-mutant Phenotype Possible combination interpretation

Epistatic ABC ab C Mutant ABC

Synergistic ABC aNovelbC A C B

A1 B1 Synergistic a1 b1 ++C C Novel A1 A2 B2 C A2 B2 B1

Action of weak allele Action of strong (null) allele Normal (wild-type) action

trends in Genetics

Predicted outcomes are shown for combinations of strong and weak alleles in linear pathways, along with interpretations that assume only strong alleles. For simplicity, all regulatory interactions are positive. Mutant alleles are shown as lower-case letters.

436 TIG November 1999, volume 15, No. 11 Genetic redundancy COMMENT Outlook

which combinations of the ABC genes act as ‘selectors’ to might actually result from the disruption of a single linear determine the identity of the floral organs12,13. In other genetic pathway. Complete loss of all A-gene function words, action of A-class genes is sufficient to specify sepal could have more dramatic effects than those predicted by identity, A and B for petal, B and C for stamen, and C for the model; support for such an idea comes from the carpel identity. At the time the ABC model was first pro- phenotype of ant ap2 double-mutants, which lack most posed, only one A-gene function had been identified, floral organs15. It is possible that there is no A function per APETALA2 (AP2), but subsequently AP1 was also char- se, but, rather, that the action of all the ‘A-class’ genes acterized as having A-class gene function in addition to its (including the AP2 and AP1 family members) is required meristem identity role4,6. Further complicating the issue is for the formation of a floral meristem, which in turn acti- the fact that AP2 also encodes a product that has similar- vates the C- and B-class genes via LFY. While certainly specu- ity to a large family of transcription factors, and at least lative, this might account for a number of discrepancies one partially redundant gene, AINTEGUMENTA (ANT), with respect to the original model. has now been identified14–16. Of course, the idea that redundancy affects our inter- The genetic evidence for the combinatorial action of the pretation of genetics is hardly new17. Similarly, the formal ABC genes is largely based on the novel phenotypes dis- interpretations of double-mutant phenotypes are well played by double-mutant combinations. For instance, established18. However, it is only with the analysis of plants that are doubly mutant for the A-class gene AP2 whole genomes that the extent of the redundancy problem and the B-class gene PISTILLATA (PI) display a novel has become apparent genetically. If these musings reflect phenotype that is dissimilar from that of each of the single reality, we might need to rewrite much, if not most, of our mutants. However, redundancy in the AP2 gene family favorite models in developmental genetics that rely on implies that the ap2 pi ‘novel’ double-mutant phenotype mutations in gene families for their interpretation.

References Development 119, 721–743 13 Coen, E.S. and Meyerowitz, E.M. (1991) The war of the whorls: 1 Marra, M. et al. High throughput bacterial artificial 7 Huala, E. and Sussex, I.M. (1992) LEAFY interacts with floral genetic interactions controlling flower development. Nature chromosome fingerprinting of the Arabidopsis genome. Nat. homeotic genes to regulate Arabidopsis floral development. 353, 31–37 Genet. (in press) Plant Cell 4, 901–913 14 Weigel, D. (1995) The APETALA2 domain is related to a novel 2 Martienssen, R.A. (1998) Functional genomics: probing plant 8 Mandel, M.A. et al. (1992) Molecular characterization of the type of DNA binding domain. Plant Cell 7, 388–389 gene function and expression with transposons. Proc. Natl. Arabidopsis floral homeotic gene APETALA1. Nature 360, 15 Elliott, R.C. et al. (1996) AINTEGUMENTA, an APETALA2-like Acad. Sci. U. S. A. 95, 2021–2026 273–277 gene of Arabidopsis with pleiotropic roles in ovule 3 Chory, J. (1993) Out of darkness: mutants reveal pathways 9 Kempin, S. et al. (1995) Molecular basis of the cauliflower development and floral organ growth. Plant Cell 8, 155–168 controlling light-regulated development in plants. Trends phenotype in Arabidopsis. Science 267, 522–525 16 Klucher, K.M. et al. (1996) The AINTEGUMENTA gene of Genet. 9, 167–172 10 Mandel, M.A. and Yanofsky, M.F. (1995) The Arabidopsis AGL8 Arabidopsis required for ovule and female gametophyte 4 Irish, V.F. and Sussex, I.M. (1990) Function of the apetala-1 MADS box gene is expressed in inflorescence meristems and is development is related to the floral homeotic gene APETALA2. gene during Arabidopsis floral development. Plant Cell 2, negatively regulated by APETALA1. Plant Cell 7, 1763–1771 Plant Cell 8, 137–153 741–753 11 Hempel, F.D. et al. (1997) Floral determination and 17 Pickett F.B. and Meeks-Wagner, D.R. (1995) Seeing double: 5 Weigel, D. et al. (1992) LEAFY controls floral meristem expression of floral regulatory genes in Arabidopsis. appreciating genetic redundancy. Plant Cell 7, 1347–1356 identity in Arabidopsis. Cell 69, 843–859 Development 124, 3845–3853 18 Avery, L. and Wasserman, S. (1992) Ordering gene function: 6 Bowman, J.L. et al. (1993) Control of flower development in 12 Bowman, J.L. et al. (1989) Genes directing flower the interpretation of epistasis in regulatory hierarchies. Arabidopsis thaliana by APETALA1 and interacting genes. development in Arabidopsis. Plant Cell 1, 37–52 Trends Genet. 8, 312–316

Intron size and GC-rich regions GENOME ANALYSIS Outlook Small introns tend to occur in GC-rich regions in some but not all vertebrates

here exists considerable variation in the size of introns, A covariance of intron size with GC composition Tboth within and between species. It has been reported would be potentially informative of some of the forces that, for some mammals and birds, genes in GC-rich iso- affecting intronic dimensions because the proportional GC chores might be both shorter (in terms of total intron size content and recombination rate are known to exhibit ϩ total exon size) and more compact (in terms of total covariance within mammals2. Indeed, it has been hypoth- intron size Ϭ total exon size) than genes in isochores of esized that, if recombination induces deletions, introns lower GC content1. Does this mean that the introns are might be smaller in GC-rich regions owing to a mutational shorter in GC-rich regions and is this generally true within bias1. However, an alternative selectionist model can also the vertebrates? To address these issues we have analysed be imagined. If longer introns are slightly deleterious then, the covariance of intron size and local GC composition in because selection is more efficient when the local recombi- a mammal, a bird, a fish and an amphibian. nation rate is high3, small deletions and insertions are