Conflict between alleles and modifiers in the evolution of genetic polymorphisms Hans Metz & Mathematical Institute, Leiden University Quick Time™ en ee n TIFF ( ong ec omprimee rd) -de compr ess or zijn v ere ist o md ez e a fbee ldin g wee r te ge ven . (formerly ADN) IIASA QuickTime™ en een TIFF (ongecomprimeerd)-decompressor zijn vereist om deze af beelding w eer te geven. VEOLIAEcole Polytechnique QuickTime™ en een TIFF (ongecomprimeerd)-decompressor zijn vereist om deze af beelding w eer te geven. NCB naturalis the tool (Assumptions: mutation limitation, mutations have small effect.) the canonical equation of adaptive dynamics with Mendelian reproduction: dX 2 Ne dt C s X Y Y Y X T evolutionary stop =0 X: value of trait vector predominant in the population Ne: effective population size, : mutation probability per birth C: mutational covariance matrix, s: invasion fitness, i.e., initial relative growth rate of a potential Y mutant population. evolutionary constraints directional selection phenotype genotype Most phenotypic evolution is probably regulatory, and hence quantitative on the level of gene expressions. reading direction coding region regulatory regions DNA the canonical equation of adaptive dynamics The canonical equation is not dynamically sufficient as there is no need for C to stay constant. Even if at the genotype level the covariance matrix stays constant, the non-linearity of the genotype to phenotype map will lead to a phenotypic C that changes with the genetic changes underlying the change in X. additional (biologically unwaranted) assumption I only showed (and use) the canonical equation for the case of symmetric phenotypic mutation distributions saving grace? I have reasons to expect that my final conclusions are independent of this symmetry assumption, but I still have to do the hard calculations to check this. the canonical equation of adaptive dynamics s X Y dX 2 Ne C dt Y Y X dX N 2 dt Ts e2 C Tr N Ne Ts 2e R0 X Y Y Y X sX Y R0 : average life-time offspring number Ts : average age at death average age at reproduction T ln R0 X Y e2 : effective variance of life-time offspring number Tr : T Tr of the residents of the residents CE is derived via two subsequent limits individual-based stochastic process canonical equation limit branching t trait value limit type: system size ∞ successful mutations/time 0 mutational step size 0 this talk: evolution of genetic polymorphisms individual-based stochastic process canonical equation limit branching t trait value limit type: system size ∞ successful mutations/time 0 mutational step size 0 the ecological theatre Assumptions: but for genetic differences individuals are born equal, random mating, ecology converges to an equilibrium. equilibria for general eco-genetic models For a physiologically structured population with all individuals born in the same physiolocal state, mating randomly with respect to genetic differences, the equilibria can be calculated by (1) setting the average life-time offspring number over the phenotypes equal to 1, (2) calculating the genetic composition of the birth stream from equations similar to the classical (discrete time) population genetical ones, with those life-time offspring numbers as fitnesses. the eco-genetic model Organism with a potentially polymorphic locus with two segregating alleles, leading to the phenotype vector X G , with G aa, aA, AA. X : Xaa E: XaA XAA T instantaneous ecological environment (XG ; E) : expected expected per capita lifetime macrogametic output (= average number of kids mothered) (XG ; E): expected per capita lifetime microgametic output times fertilisation propensity ( average number of kids fathered) Abbreviations: G : (XG ; E), etc. (and similar abbreviations later on). the eco-genetic model C = classical discrete time model B: total birth rate density (C: total population density, N) pa , pA, qa , qA : allelic frequencies in the micro- resp. macro-gametic outputs ( pa 1 pA and qa 1 qA ) baa ,baA ,bAA : genotype birth rate densities (C: genotype densities, naa, etc) random union of gametes: baa pa qa B, baA ( pAqa pa qA )B, Point equilibria: with bAA pAqA B. qA naaAApTAs,aa qAbaa, 12etc. aA (pAqa paqA ), pA AA pAqA 12 aA (pAqa paqA ), : pAqAAA (pAqa paqA )aA paqaaa 1, : pAqA AA ( pAqa pa qA )aA pa qa aa . example ecological feedback loop: E F 1 (E, XG )nG , ... , k (E, XG )nG G G the evolutionary play Assumptions: no parental effects on gene expressions (mutation limitation, mutations have small effect) long term evolution Two models I. Evolution through allelic substitutions Xa , XA : allelic trait vectors : genotype to phenotype map: XaA (Xa , XA ), etc. Abbreviations: (Xa , X A ) : 1(Xa , X A ) : 1aA , Xa etc. II. Evolution through modifier substitutions b : original allele on generic modifier locus, X aa Xbb,aa XbB,aa B: mutant, changing X XaA Xbb,aA : Xb into XbB,aA : X B . X AA Xbb, AA XbB, AA smooth genotype to phenotype maps Model I (allelic evolution) If X XA Z A then Xa XaA 2aA ZA O( 2 ), XA XAA 2AA ZA O( 2 ), X XAA 1AA ZA 2AAZA O( 2 ) XAA 2 2AA ZA O( 2 ) Model II (modifier evolution) XbB,aa Xbb,aa X BB,aa Xbb,aa Z aa Z aa 2 If XbB,aA Xbb,aA Z aA then X BB,aA Xbb,aA 2 Z aA O( ) X X X X Z AA Z AA bB, AA bb, AA BB, AA bb, AA Model I: phenotypic change in the CE limit with dX a 0 21 aa dt dX dX dX dX a dX A 1 aA 2 aA dt dt a dt A dt dt dX A 0 2 2 AA dt R0, Xa XA (X ) R0, Xa XA (X ) dX A dXa A N A 2 AC A a N a 2 aCa , X dt X a dt A T with a 2 (Ts,a a2 ), A 2 (Ts,A A2 ), a , A the mutation probabilities per allele per birth, C a , C A the mutational covariance matrices, and a : 12 (pa qa ), A : 12 (pA qA ). T Model I: phenotypic change in the CE limit R0.Xa XA (Xa ) Xa R0, Xa XA (XA ) XA R0, Xaa , XaA , XAA (Xaa , XaA ) Xaa R0, Xaa , XaA , XAA (XAA , XaA ) XAA 1aa 2 AA R0, Xaa , XaA , XAA (Xaa , XaA ) XaA 1aA R0, Xaa , XaA , XAA (XAA , XaA ) XaA 2 aA Convention: Differentiation is only with respect to the regular arguments, not the indices. notation denotes the Kronecker product: a11B L AB M am1B L a1n B M amn B and I the identity matrix of any required size Model I: phenotypic change in the CE limit in matrix notation: structure matrix 0 dX 1 aa a 2 0 0 0 0 dt dX 1 aA 0 1 1 0 I 2 aA dX A dt 0 0 0 0 2 0 2 AA dt and (the allelic coevolution equations) dXa T 0 %a%a aCa 1aaT 1aA dt 0 N 2 allelic allelic T % C 0 dX % 0 0 A 2 aA A A A A dt with allelic a a A A , %a a allelic , R0, X , X , X (Xaa , XaA ) T aa aA AA X aa T R (X , X ) 0, Xaa , XaA , X AA aa aA XaA 0 T T 2 AA R 0, Xaa , XaA , X AA (X AA , X aA ) X aA T R (X , X ) 0, Xaa , XaA , XAA AA aA X AA allelic aa A A , %A A allelic , %a a allelic , %A A allelic . Model I: phenotypic change in the CE limit combining the previous results gives: dX allelic N dt 2 0 0 0 0 0 1 1 0 0 0 2 I 2 allelic C allelicGallelic with and %a% 0 0 0 a a 0 %a% 0 0 a a I : 0 0 %A% 0 A A 0 0 %A% 0 A A 1 aa 1 aA C allelic : 0 0 0 0 Ca 2 aA 0 2 AA T 0 1 aa C A 0 T 1 aA 0 0 T 2 aA 0 T . 2 AA Model I: phenotypic change in the CE limit with Gallelic R0, X , X , X (X aa , X aA ) T aa aA AA X aa T R (X , X ) 0, Xaa , XaA , X AA aa aA X aA T R0, X , X , X (X AA , X aA ) aa aA AA X aA T R0, Xaa , XaA , X AA (X AA , X aA ) X AA Model I: phenotypic change in the CE limit an explicit expression for the allelic (proxy) selection gradient: Gallelic Ê B1 ÊGcommon with a 0 0 0 0 0 0 a I B 0 0 A 0 0 0 0 A on the Hardy-Weinberg manifold (pA = qA) : I aa aA aA aa AA aA aA AA a ÊÊÊ1 ( pa qa ) ,ÊÊÊÊ A ÊÊÊ1 ( p A q A ) 4 4 4 4 Model I: phenotypic change in the CE limit Gcommon aA pa 2 p aa A 2 p AA a 2 aA p A 2 aa aa aA pA qA qA pA qa 2 2X aa 2 X aa T T aA aa aA pa qa q q p A a a 2 2X aA 2 X aA T T aA aA AA pA qA qA pA qa 2 2X aA 2 X aA T T AA AA aA pa qa qa pa qA 2 X 2 2 X AA AA T aa (Xaa ;EXaa XaA XAA ), etc. with effect a mutation in the and a pa--allele qaaa (paqA pAqa )aA Ap-allele AqAAA 1 T Model I: phenotypic change in the CE limit Gcommon aA pa 2 p aa A 2 p AA a 2 aA p A 2 aa aa aA pA qA qA pA qa 2 2X aa 2 X aa T T aA aa aA pa qa q q p A a a 2 2X aA 2 X aA T T aA aA AA pA qA qA pA qa 2 2X aA 2 X aA T T AA AA aA pa qa qa pa qA 2 X 2 2 X AA AA T (Xaachange ;EAX=aa XqaAAinX)AAthe ), etc. with effect the resulting phenotypic on theofHardy-Weinberg (p aamanifold and heterozygotes 2 AA-homozygotes aa-homozygotes pqp2 (pq2ppp q ) p p q 1 1 a a aaa aa a A a a AA aA aA A AAA A AA T summary of Model I (allelic trait substitution) on the Hardy-Weinberg manifold: 2 0 0 0 dX allelic N 0 1 1 0 I 2allelic Callelic B1 Gcommon 0 0 0 2 dt Gcommon pa aA 2 p aa A 2 p AA a 2 aA p A 2 T T aa aA aa pA qA qA pA qa 2 2X aa 2 X aa T T aA aA aa pa qa qa pa qA 2 2X aA 2 X aA T T aA aA AA pA qA qA pA qa 2 X 2 2 X aA aA T T AA AA aA pa qa qa pa qA 2 X 2 2 X AA AA Model II: phenotypic change in the CE limit dX 2 modifier N 2 haplotype CmodifiersGmodifier dt 2 2 (T with modifier s,modifier modifier ) , haplotype the mutation probability per haplotype per birth, Cmodifiers the covariances of the mutational effects of modifiers. Gmodifier 1 0 0 0 0 0 1 1 0 0 0 1 with on the Hardy-Weinberg manifold: I ÊGcommon a 0 0 0 0 0 0 1 ÊÊ a A A a a 0 a 0 0 I 0 0 0 0 A 0 0 I ÊÊÊÊÊÊ:Ê 0 a 0 0 0 0 A 0 A aA aA aA aA a ÊÊqa pa 1 A ,Ê 1 a 0ÊÊÊ 0A 0q A A pA 2 2 2 2 summary: model comparison Model I (allelic substitutions): dX allelic N dt 2 0 0 Model II (modifier substitutions): dX 2 modifier N 2 haplotypeC modifiers dt Gcommon qa aA 2 q aa A 2 q AA a 2 aA q A 2 0 0 0 1 1 0 0 0 2 1 0 0 I 2 allelic C allelic B1Gcommon 0 0 0 1 1 0 0 0 1 I Gcommon T T aa aA aa qA pA pA qA pa 2 X 2 2 X aa aa T T aA aA aa qa pa p p q A a a 2 2 X aA 2X aA T T aA aA AA qA pA pA qA pa 2 2 X aA 2X aA T T AA AA aA qa pa p qa pA 2 a 2 X AA 2X AA summary: model comparison Model I (allelic substitutions): dX allelic N dt 2 0 0 Model II (modifier substitutions): dX modifier N 2 haplotypeC modifiers dt Gcommon qa aA 2 q aa A 2 q AA a 2 aA q A 2 0 0 0 1 1 0 0 0 2 2 0 0 I 2 allelic C allelic B1Gcommon 0 0 0 2 2 0 0 0 2 I Gcommon T T aa aA aa qA pA pA qA pa 2 X 2 2 X aa aa T T aA aA aa qa pa p p q A a a 2 2 X aA 2X aA T T aA aA AA qA pA pA qA pa 2 2 X aA 2X aA T T AA AA aA qa pa p qa pA 2 a 2 X AA 2X AA on the Hardy-Weinberg manifold summary: model comparison Model I (allelic substitutions): dX allelic N dt 2 0 0 Model II (modifier substitutions): dX modifier N 2 haplotypeC modifiers dt Gcommon qa aA 2 q aa A 2 q AA a 2 aA q A 2 0 0 0 1 1 0 0 0 2 2 0 0 I 2 allelic C allelic B1Gcommon 0 0 0 2 2 0 0 0 2 I Gcommon T T aa aA aa qA pA pA qA pa 2 X 2 2 X aa aa T T aA aA aa qa pa p p q A a a 2 2 X aA 2X aA T T aA aA AA qA pA pA qA pa 2 2 X aA 2X aA T T AA AA aA qa pa p qa pA 2 a 2 X AA 2X AA summary: model comparison 2 0 0 2 haplotypeC modifiers 0 0 0 1 1 0 0 0 2 2 0 0 on the Hardy-Weinberg manifold I 2 allelic C allelic 0 0 0 2 2 0 0 0 2 I summary: model comparison 2 0 0 0 0 0 1 1 0 0 0 2 on the Hardy-Weinberg manifold I 2 allelic C allelic 2 haplotypeC modifiers 2 0 0 0 0 0 2 2 0 0 0 2 I summary: model comparison 2 0 0 0 0 0 1 1 0 0 0 2 on the Hardy-Weinberg manifold I 2 allelic C allelic 2 haplotypeC modifiers 2 0 0 0 0 0 2 2 0 0 0 2 I %a% 0 0 0 a a 0 %a% 0 0 a a I : 0 0 %A% 0 A A % % 0 0 0 A A A summary: model comparison 2 0 0 0 0 0 1 1 0 0 0 2 I 2 allelicCallelic 2 haplotypeC modifiers B on the Hardy-Weinberg manifold 2 0 0 0 0 0 2 2 0 0 0 2 A I %a% 0 0 0 a a 0 %a% 0 0 a a I : 0 0 %A% 0 A A % % 0 0 0 A A A summary: model comparison Model I (allelic substitutions): dX allelic N dt 2 0 0 Model II (modifier substitutions): dX modifier N 2 haplotypeC modifiers dt a 0 0 0 0 0 0 a I B 0 0 A 0 0 0 0 A 0 0 0 1 1 0 0 0 2 2 0 0 I 2 allelic C allelic B1Gcommon 0 0 0 2 2 0 0 0 2 I a 0 a 1 0 a A A a 0 0 0 0 Gcommon 0 0 0 0 I A 0 0 A aa aA aA aa AA aA aA AA a ÊÊÊ1 ( pa qa ) ,ÊÊÊÊ A ÊÊÊ1 ( p A q A ) 4 4 4 4 a ÊÊqa aA pa aA 1 A ,ÊÊÊÊ A qA aA pA aA 1 a 2 2 2 2 in reality alleles and modifiers will both evolve combining Models I and II: dX dt dX dX dX dX dX dt modifier dt a dt A dt modifier dt allelic evolutionary statics genetical and developmental assumptions Cmodifiers uniformly has full rank and C allelic uniformly has maximal rank. In biological terms: there are no local developmental or physiological constraints. When there are developmental or physiological constraints, we can usually define a new coordinate system on any constraint manifold that the phenotypes run into, and proceed as in the case without constraints. So-called genetic constraints are rooted more deeply than in the physiology or developmental mechanics. Example: some phenotypes can only be realised by heterozygotes. IF: There are no constraints whatsoever, that is, any combination of phenotypes may be realised by a mutant in its various heterozygotes. (known in the literature as the “Ideal Free” assumption). evolutionary stops Evolutionary stops satisfy I: 2 0 0 0 0 0 1 1 0 I 0 0 2 II: 1 0 0 0 0 0 1 1 0 I 0 0 1 & 1 Callelic B Gcommon 0 Gcommon 0 that is, Gcommon should lie in the null-space of I: respectively II: 2 0 0 0 0 0 1 1 0 I 0 0 2 1 0 0 Callelic B1 0 0 0 1 1 0 I 0 0 1 dCallelic 0 dt evolutionary stops Allelic evolution for model I: dXa T T 0 0 dt 1aA Ca 0 1aa allelic 2N allelic T T Gallelic 2 aA 2 AA 0 CA 0 0 dX A dt Gallelic Ê B1 ÊGcommon Hence at the stops: T T 0 0 1 1aA Ca 0 1aa Gcommon 0 T T B 0 C 0 2 aA 2 AA 0 A or equivalently, CallelicB1Gcommon 0 when do the alleles and modifiers agree? The alleles on the focal locus and the modifiers agree about a stop only if 1 C B Gcommon 0 I allelic and 1 0 0 0 0 1 1 0 I G II common 0 0 0 0 1 If the dimensions of phenotypic and allelic spaces are n resp. m, then I is a system of min{4n, 2m}, II a system of 3n equations. In the case of modifier evolution, these have to be satisfied by 3n, in the case of allelic evolution by min{2m,3n} unknowns (since the Xa , XA act only through the Xaa , XaA , XAA ). The seemingly simpler Gcommon = 0, amounts to 4n equations. Hence, generically there is never agreement. (When 2m > 4n, the alleles cannot even agree among themselves!) exceptions to the generic case We have already seen a case where the alleles and modifiers agree: if pA = qA. This can happen for two very different reasons: 1. When (XG ; E) (XG ; E) (HW) (the standard assumption of population genetics). 2. Phenotype space can be decomposed (at least locally near the ESS) into a component that influences only , and one that only influences (as is the case in organisms with separate sexes), and moreover the Ideal Free assumption applies. In that case at ESSes aa = aA = AA = 1 and aa = aA = A., Hence (HW) applies, and therefore pA = qA. inverse problem: find all the exceptions Assumption: 4m≥ n In that case there is only agreement at evolutionary stops iff at those stops Gcommon = 0. inverse problem: find all the exceptions For one dimensional phenotype spaces the individual-based restrictions on the ecological model that robustly guarantee that Gcommon = 0 are that (a) at evolutionary stops (HW) holds true, or (b) in their neighbourhood: (i) aA xaA 0ÊÊaA xaA 0 or aA xaA 0ÊÊaA xaA 0 or (ii) aa xaa 0 & aa xaa 0 or AA xAA 0 & AA xAA 0 If not (a), any individual-based restriction doing the same job implies (b). Examples: A priori Hardy Weinberg: . Ecological effect only through one sex: either x 0 or x 0. Sex determining loci: for AA females and aA males: AA 0, aA 0. The conditions for higher dimensional phenotype spaces are that after a diffeomorphism the space can be decomposed into components in which one or more of the above conditions hold true. biological conclusions Olof Leimar Quick Time™ e n e en TIFF ( LZW) -de compr ess or zi jn v ere ist om dez e a fbe eld ing we er te g ev en. When the focal alleles and modifiers fail to agree the result will be an evolutionary arms race between the alleles and the rest of the genome. This arms race can be interpreted as a tug of war between trait evolution and sex ratio evolution. Generically there is disagreement, with one biologically supported exception: the case where the sexes are separate. (Even though in all the usual models there is agreement!) Prediction Hermaphroditic species have a higher turn-over rate of their genome than species with separate sexes. QuickTime™ and a decompressor are needed to see this picture. Carolien de Kovel The end history basic ideas and first derivation (1996) hard proofs (2003) QuickTime™ en een TIFF (ongecomprimeerd)-decompressor zijn vereist om deze afbeelding weer te geven. Ulf Dieckmann & Richard Law Mendelian discrete diploids generations extensions (2008) QuickTime™en een -decompressor zij n vereist omdeze afbeeldi ng weer te geven. Nicolas Champagnat & Sylvie Méléard dX 122 N e C dt s Y X Y Y X hard proof for pure age rather unbiological dependence T Michel Durinx & me general with Poisson life histories # offspring only for Assumptions still (correspondingso to far a Lotkanot yet published equilibria Volterra type ODE model): individuals reproduce community clonally, have non-rigorous Tran exponentially distributed Chi lifetimes and give birth at constant rate non-rigorous (2006) from birth onwards in reality alleles and modifiers will both evolve in “reality”: dX dt dX dX dX dX dX dt modifier dt a dt A dt modifier dt allelic Generically in the genotype to phenotype map all three equations are incomplete dynamical descriptions as , C allelic and C modifier may still change as a result of the evolutionary process. C allelic and C modifier are constant when is linear and C a and C A resp. the C B are constant (two commonly made assumptions!). Otherwise constancy of C modifier and C allelic requires that changes in the various composing terms precisely compensate each other. rarely will be constant as a and A generically change with changes in X. the canonical equation of adaptive dynamics dX 2 ne C dt s Y X Y Y X T X: value of trait vector predominant in the population ne: effective population size, : mutation probability per birth C: mutational covariance matrix, s: invasion fitness, i.e., initial relative growth rate of a potential Y mutant population.