Evolution and Genetic Diversity Flashcards
how are most traits determined?
multiple genetic loci
plus the environment
quantitative or continuous variation
P = G+E
(or G*E?)
basics of evolution
H2 = Va/Vp
degree to which a trait is inherited from gen to gen
proportion of phenotypic variation explained by additive genetic variation
selection (S)
individuals vary in contribution to next gen - sometimes due to variation in particular traits (natural selection applies here)
results in change in trait mean from one gen to next
evolution/response to selection (R=H2*S, breeders equation)
P=G+E
use the variation between individuals to measure this
Vp=Vg+Ve
Vg components?
Va - additive genetic variance
Vd - dominance variance
Vi - epistatic variance
Va?
due to alleles at underlying locu that are codominant (both alleles expressed in heterozygote) and cause small differing increases and decreases in trait value
Vd
occurs when alleles at underlying loci are not codominant but show full dominance
inbreeding depression special case?
Vi
due to interctions between alleles at different loci
phenotype can take any particular value depending on the genotype at another locus
Va as main focus
-main determinant of transmission across gens
-largest component of Vg
-thus major determinant of repsonse to selection as allele frequencies at underlying loci change over generations
(only inherit one allele from each parent so additive effect is really the only effect passed down through gens - other ones muddied by mixing of parent genomes)
can estimate from covariance (similarity) between relatives compared to rest of population
-expressed often as heritability
-narrow sense - more common - H2=Va/Vp
-broad sense - clones and inbred lines H2 = Vg/Vp - as other genetic effects not so muddied in clones and inbred as genomes arent really mixed up between gens
parent offspring regression
mid parent bill depth against measured offspring depth
parent and offspring share 1/2 genes so covariance is partially due to shared genetics
bill depth of parents increases - increase in offspring
assume that only cause of covariance between parent and offspring is Va - covariance between parents and offspring tells us what 0.5 of genetic variance
slope of regression can give estimate of heritability - can calculate Va from it
stronger correlation and steeper slope = higher heritability
estimating Va
Va (and therefore H2, Va/Vp) is estimated by examining covariance in trait values between individuals of known relatedness - taking into account how related they are
e.g.
identical twins - share all Va and Vs (share dominant alleles, Ve too as share all alleles at all loci too (similar for clones/inbred lines too ig)
parent offspting and full sibs - coefficient of 0.5 - only share 1 allele at each locus - major cause of similarity is 1/2 of Va (also 1/2 Vd - share rouglhly 1/2 dominant alleles)
half sibs - 0.25 coefficient, 0.25xVa (1/16 Vd)
estimating Va and H2 in nature
examples before use single type of relative
sometimes useful in nature (PO regression e.g.)
however in natural populations:
-often have many types of relatives in population - need to use as many as possible for best data
-use ANIMAL MODEL
-allows all types of relatives to be used
-statistical model, uses big matrix of relatedness between relatives in population
-use this matrix and covariance between all these relatives to estimate Va for a trait
need to know:
relatedness of individuals in population (can be difficult to assign fathers sometimes, facilitated by DNA profiling to determine genetic similarity)
trait values for those individuals
animal model example - red deer isle of rum
tag fauns each year - tell individuals apart
use genetic analysis to assign father
create pedigree with this relatedness data (from mothers and fathers of everyone being known)
can see how related every individual is to each other
can use this to feed relatedness matrix into animal model
estimate heritability of specific traits by using covariance of different relatives
(some values differ by sex - birth weight less heritable in males)
animal model benefits
-can use whole pedigree to estimate Va - many different relative types at once can be used (maximise data therefore maximise power, e.g might not have so large a sample of single type e.g. full sibs)
well set up for the types of pedigree found in wild pops
-can control and estimate other sources of covariance other than Va (environmental…)
-can handle missing data - can still have individual in model even if phenotype values are missing
-can handle unbalanced designs e.g. unequal family sizes
selection differential?
difference between mean of indiviudals that breed and the mean of the whole population (incl non breeders)
useful in artificial selection, but in natural populations can’t always record accurately who’s breeding
measuring selection in nature
big population crash
selection event e.g. El nino event
can get data from population after crash and compare it to data from before the crash
selection differential for trait = difference in trait between group that survived vs whole pop before
e.g. el nino galapagos ground finches
could see what traits were selected for
selection differential problem in natural population
-(before) dont know whos breeding exactly
-differences not always discrete yes/no in natural populations
e.g. variation in offspring number produced often used as best measure of overall fitness
(also e.g. variation in age at maturity, variation in lifespan)
these traits are easily measured but are not discrete yes/no black/white
so not so easy to fit onto selection differential
use selection gradient instead
selection gradient approach
do regression of relative fitness on trait X (trait on x axis)
equation of this line:
w=a + beta*X (analogous to y=mx+c)
beta - selection gradient is the gradient of this line on this regression
beta = the degree to which that trait value predicts fitness
(covariance of trait and relative fitness)
slope of line (beta) = covariance of X and Y divided by variance in X
KEY - in nat populations - fitness not always discrete (survivors vs non survivors like el nino thingy) - selection gradient used instead of differential
predicting response to selection?
breeder’s equation
R=h2*S
heritability * strength of selection
S = difference of mean of population of selected parents, and the mean of the whole pop before selection
response (R) is basically the difference in mean of offspring from the selected parents, and mean of offspring if whole population had been allowed to reproduce
Works OK in Lab and breeding
pitfalls in nat pops
breeders equation in wild?
el nino event for darwins finches
for bill depth
calculated
R=h2*S
=0.54 * 0.82=0.44
BUT observed value looks more like 0.8
B equation predicts change of 0.44mm, observed change was about 0.8mm
observed response was much stronger than predicted
estimated only half of the change in nat pop
poor
a lot worse in other studies
why breeders equation bad in wild?
possible reason 1:
poor estimation of h2 in wild due to environmental variation (plasticity is cause???)
estimating Va from covariance between relatives
assumes all covariance is genetic
BUT relatives also share environements - can also cause covariance
environmental variation Ve is v difficult to control in nat pops
causing problems in estimating h2 in nat pops
plasticity and breeders equation?
traits of an individual respond to environmental conditions
P=G+E
can be one off change:
- (daphnia grow helmets in response to predators, stay for life)
-good early conditions -> can give high lifetime breeding success regardless of conditions experienced
can also be ongoing through life - reaction norms
shared environment aspects between relatives?
Shared spatial environment
shared year of birth
shared rearing environments
shared mother (maternal effects)
shared spatial environment
rum red deer
individuals have small home ranges
individuals in same home range experience same set of variable (weather, food,,,)
can influence phenotype in same way causing covariance
here maternal relatives share home range
so estimating Va and some relatives used are maternal relatives - some of covariance shared due to same environmental conditions
shared year of birth
rum red deer
mean temp against mean birth weight of females that year
temp increases - BW increases
individuals born in same year have similar birth weights
will cause issues if relatives in model born in same year - share environmental effect - share covariance due to that not Va - bias heritability estimate
shared rearing environments
same male meerkat fathers multiple offspring in a litter
assuming that all covariance in these litters is completely genetic - overestimate heritability due to shared covariance from shared rearing environment
(remember rearing conditions predict LBS)
shared mother
maternal effects
individuals sharing mothers liekly share same environment that mother provided/experienced
if mother is:
-consistent in milk production
-consistent in gestation time
-for birds - investment in eggs hormones,,,
some covariance due to shared maternal investment not genetics
issues with estimating heritability in nature:
many factors that can cause covariance in relatives
can’t really predict direction of these effects (e.g. what causes mother to provide more food) and thus hard to include as fixed effect
not accounting for them though will inflate estimates of Va and thus h2 if not accounted for
solution for estimating heritability in nature
treat them as another source of covariance as you would Va
can estimate them from data
can fit a matrix of shared environmental effects same way you would relatedness into animal model
values will be all 0 unless individuals share environmental effects
also dont need to know complete info of that shared environment e.g. whether it is a larger mother providing more milk or other way round
this partitions variance into Va and Ve
can provide most conservative estimate of Va
only possible to be fit in animal model that allows us to fit these matrices to remove shared effect of environment
animal model environmental effects constraints
animal model will only allow this if we have different types of relative that share different amounts of genetic and environemtnal variation
allowing us to separate their effects
e.g. if only have maternal relatives and no other
will never be anle to separate maternal effect from genetic effect - no data from individuals that share genes but not environments
model performance depends on data
examples of animal model accounting for environmental effects
Soay sheep on st kilda
build up multigenerational pedigree
investigate heritability of weight and leg length in adult
estimated alone was overall higher values than when estimated taking into account maternal effect
proportion of the phenotypic variance was due to mother’s identity (factors discussed earlier)
modelling various effects to take them into account
looked at great tit population
look at body mass heritability
additive genetic effects (cause similarity between relatives)
nest effects (includes maternal) - cause similarity between chicks in same nest
temporal and spatial variation - caused similarity between chicks in same year/area
and unexplained “residual” variation
this model found genetic variance accounted for about 25% of variance
nest effects accounted for about 40%
wouldve been included in genetic and inflated the value for Va and h2 if not accounted for
reasons for breeders equation not working for wild populations?
heritability not accounting for environmental effects (need to use animal model with environmental varaince matrix to correct for)
also R=h2*S -> assumes that environment will be the same between generations
but systematic environmental change is common between generationd
example of changing environment affecting selection prediciton with breeders equation?
collared flycathcers in gotland - sweden
can build pedigree from observing nestboxes
social pedigree that assumes males and females visiting same box are parents
investigating selection on fledgeling condition index (body mass relative to skeletal size [leg length])
increase in FCI (body mass? ish) was predicted (from Va estimates and Selection differential estimates)
actually observed a decrease
because environment was deteriorating in quality across study period
SO the genetic response to selection was being masked by plastic response to decrease in environment quality
worse environment -> “worse” phenotype
“counter gradient variation”
multiple traits problem?
traits are not independent of each other
they can covary both at phenotypic and genotypic level
leads to 2 problems:
- natural selection will not act on single traits, acts on individuals who have particular combinations of trait values
- if traits covary genetically, selection on one will cause selection on the other
SO estimates on both heritability and selection differential/gradient need to factor in what happens at other traits
directions of covariance/correrlation between traits
+ve (e.g. body size traits)
none (e.g. size and coat colour) traits independent of each other
-ve (no. and size of offspring)
these covariances can occur at pheno- and genotypic levels
phenotypic covariance affecting selection?
nat selection does not operate on single traits in isolation
acts on individual with particular combo of trait values
so selection is likely to act on multple traits at same time
so phenotypic correlations between traits can mask true selection on that trait
(i.e. the direct relationship between that trait and relative fitness)
example of phenotypic covariance masking selection
finches
plot beak width vs beak depth
positively phenotypically correlated
plot both traits against relative fitness on z axis in 3d
looking at relationship on whole phenotypic level - looks like overall positive selection on both traits
when look at beak depth at fixed beak width - still positive selection - higher than overall 3d phenotypic level gradient
when look at width at fixed depth
actually shows negative selection (what brings the overall gradient down)
the stronger positive association between beak depth and fitness was MASKING the true negative effect of beak width and fitness
how to account for phenotypic correlation when estimating selection
multiple regression analysis
(opposed to earlier simple regression)
instead of one single selection gradient
can use partial selection gradients (gradient for that trait after taking all other traits in model into account)
instead of
w= a+betax
we have
w= a + beta1x1(trait value for trait 1) + beta2*x2 …
only really need to know that it uses partial selection gradients to take all traits into account
genetic covariance between traits
genes or genetic effects on one trait also have an effect on another trait
can be positive, negative, no effect like phenotypic
estimated as genetic correlations - rg - standardised so it can be compared among traits
phenotypic and genetic correlations between the same traits
phenotypic correlations are a product of genetic+environmental covariance/correaltion so if environmental covariance is strong it can mask genetic
SO genetic correlations between traits don’t HAVE TO be in the same direction as phenotypic correlations between them
causes of genetic covariance between traits:
pleiotropy:
one locus influences more than one trait
-e.g. allele that increases bone length also likely to increase mass (due to how development works)
genetic linkage:
alleles at two or more loci that are very close together so that they are unlikely to recombine (are LINKED)
always inherited together (in parental combination)
and so if an allele 1 at locus 1 increases body size
and an allele 1 at the very nearby locus 2 reduces litter size
these traits will be negatively genetically correlated
(until recombination happens to break recombination - less likely due to close linkage)
why genetic covariance matters?
if two traits are genetically linked
(e.g. tail length and body length)
even if only one trait is directly under selection (or if selection on one is stronger than the other)
since the other one is linked it will also respond to selection in the same direction as its correlation with that other trait
get changes in traits as a result to selection on another trait
how to deal with multiple trait problem in breeders equation?
EXTENDED/MULTIVARIATE breeders equation
take the normal univariate breeders equation (using the partial selection gradient for trait 1)
then add
product of:
-sqrt heritability of product of h2 of both traits
-rg
partial selection gradient for trait 2 (one that youre not calculating R for)
also matrix form
basically - take into account
heritability of the two traits
the GENETIC CORRELATION between them
and the partial selection on the other trait (which would be influencing the response to selection of trait 1)
example of multivariate breeders equation being used
G fortis
darwin finch
looked at many traits and the selection differential on them
estimated selection was all positive
association between individual trait and survivial positively correlated for all here
then did a multiple regression to look at all traits at the same time
most of the values remained relatively similar
except:
-no real selection actually on tarsus length
-selection on bill WIDTH was actually NEGATIVE (because in drought - less food available - so have to crack nuts that G fortis did not have to before - need narrower bill in order to apply pressure so narrower bill better in more generalist less picky feeding needed after drought)
deeper bill better after drought
this mis-prediction in indepentent selection differentials was because the genetic correlation between beak depth and width was very strong
so when actually observing changes in traits
since beak depth was selected for
and is so strongly correlated genetically with beak width
increase in depth caused increase in width
masking the negative selection on width
problem with multivariate breeders equation?
we cannot know if we have measured and included ALL relevant traits
cannot know whether this has been done in wild population so predictions will be off
inbreeding depression importance?
thought to have driven evolution of dispersal and mating systems
important to avoid in breeding
potential cause of population decline so of conservation concern
inbreeding
mating between relatives
inbreeding depression
lower fitness in inbred individuals (or lower values of traits related to fitness)
thought to be due to increased homozygosity
fitness is lower in more inbred individuals
dominance patterns in inbreeding depression
- deleterious recessives:
with recessive deleterious alleles
average fitness of homozygotes is lower than that of the average heterozygote
(heterozygote is just as fit as WT - has dominant allele, however recessive homozygous fitness muxh lower - brings down homo average)
- overdominance
heterozygotes are fitter than average of homozygotes
rare
e.g. sickle cell
having no allele is bad in areas with malaria
having both is bad as bad anaemia
having one protects against malaria without really bad anaemia
in both - average hetero fitter than average homo
why are deleterious alleles expected to be recessive?
if they were additive or dominant they are much more easily removed by selection
identical by descent?
if both alleles in an individual for a locus are identical by descent -
then they are identical due to both coming from the same common ancestor
inbreeding coefficient
measure if degree of inbred-ness in an individual
probability that an individual has 2 alleles at a locus that are identical by descent
f=0 - not inbred
f refers to a single individual - but determined by degree of relatedness of that individuals parents so NEED PEDIGREE DATA on the population
measuring inbreeding coefficient
trace segregation probabilities of allele through generations
probability that it will segregate into both offspring (50%)
0.5 in each gen for each descendent
do for both alleles
formula
Sigma:(1/2)^n
n=number of individuals in path via the common ancestor
sigma - because it is sum of all paths
e.g. full sib mating
sum:(1/2)^3
notes on inbreeding coefficient?
assumes that:
-base population alleles are distinct, not IBD (base population is outbred)
-inbreeding affects all loci equally
-f is a probability so 0<1<1 (or equal to)
why does inbreeding increase homozygosity?
hardy weinberg equation shows us that:
after one gen of selfing (most extreme inbreeding)
-AA individuals have AA offspring
-aa individuals have aa ofspring
(so no. of homozygotes doesn’t decrease)
-Aa individuals have 1AA;2Aa;1aa
so number of heterozygotes decreases and homo increases
homo increase by 50% in one gen of selfing
so overall increased freq. of homozygosity in F1
fitness and inbreeding
expect to see decreased fitness with increased inbreeding
generally true
correlations between negative effects on traits and inbreeding
evidence of inbreeding depression in natural populations
Song sparrows of mandarte island in B. columbia
overwinter in b columbia
looked at survival during population crash
individuals that survived likely had much lower f
whereas individuals that died likely to be higher f
suggests inbred individuals have lower fitness
collared flycatchers in gotland
no. of offspring that survived to recruit
signifivantly reduced for f-0.25 (full sib or parent offspring mating)
red deer calves rum isle
higher f = lower probability of surviving first year
f=0 -> 60%
f=0.25 -> 20%
pedigree free estimates of inbreeding coefficient
use homozygosity markers in genome
use as an estimate of f in absence of pedigree data
poor estimate if few markers
need huge number
SNPs can give 1000s of markers and therefore v good estimate of f
should be better f as it measures actual allele sharing instead of just probability of allele sharing
(showed even stronger inbreeding depression in rum red deer)
opens up possibility of pedigree free studying of inbreeding
ecology of inbreeding depression hypothesis
environmental stress increases inbreeding depression
in benign conditions increasing f causes a shallower downward slope than in harsh conditions
because some mildly deleterious alleles don’t cause big effect in benign conditions
but have amplified effect in haesher
examples of inbreeding depression ecology
inbreeding depression in rosepink
compared to outbred individuals
in field drop was 74%, in benign greenhouse was only 50%
so in lab conditions - do see an effect of environment
however not so clear in nat pops:
-Darwin’s finches best example
inbreeding coefficient had less of an association with survival in wetter years than harsher drier years
however other natural populations (incl rum red deers) did not show any evidence for environment effect on relationship of f and fitness
however could be because
-possibly environment is already so harsh that deleterious recessive alleles already at max effect
-so when environment gets harsher - effects don’t really get exacerbated
-because wild pops live in harsher environments normally - particularly stark difference to lab conditions environments
