The evolution and genetic basis of complex traits in D. melanogaster

Z. Forrest Elkins

University of Missouri - Columbia

January 9, 2023

Complex traits

Not a single gene with two alleles
Poly- or omni-genic
- many genes of small effect
Continuous phenotypic data

A quantitative, complex trait – polar bear weight

Genetic basis

Sample genome-wide association study (GWAS) with simulated data

Experimental evolution

Artificially select for some phenotype
- Offspring of parents with desired phenotype are mated to seed next generation
After >1 generations, the population phenotype will be vastly different
The genotype of organisms with the desired phenotype will be vastly different than ancestor genotype

Artificial selection

Evolve and resequence

Sequence ancestor DNA
Artificially select for n generations
Resequence organisms from the nth generation
Compare genetic differences between ancestor and nth generation

Bulk segregant analysis

Artificially select for opposing phenotypes (denoted as ‘high’ and ‘low’) in as low as 1 generation
Sequence ‘high’ bulks and ‘low’ bulks
- Bulk = pooled genetic sample
Compare genetic differences between bulks

Overview

Genetic basis of exploration tendency in a multiparent population of D. melanogaster

Phenotypic differences in starvation resistance between selection lines of an experimentally evolved population of D. melanogaster

Exploration tendency

Dispersal and exploration

Dispersal – complex trait
- Any movement with potential to lead to gene flow
Exploration – complex trait
- Sub-phenotype of dispersal
- Moving from known environment to unknown environment

Drosophila exploration

D. melanogaster can get all the resources they need from a piece of vegetable rot
What causes a fly to leave its current environment for a novel environment?
- Costly investment
- Trade-off with other traits like reproduction & lipid storage

Main questions

Is exploration a heritable trait?: I.e., how large a genetic component contributes to variation in exploration within the population?
If so, what on the genetic level contributes to exploration in Drosophila?: What are the differences in genetic architecture between exploring and non-exploring flies?

Heritability background

Heritability: The degree of variation in a phenotypic trait in a population due to genetic variation in the population
Broad-sense heritability: Total genetic variability over total phenotypic variability; Cannot calculate narrow-sense heritability in the DSPR due to lack of family data

Drosophila Synthetic Population Resource

Heritability experimental setup

Heritability results

\[\begin{gather*} H^2 = V_G / V_P = 0.4 \end{gather*}\]

Main questions

Is exploration a heritable trait?: I.e., how large a genetic component contributes to variation in exploration within the population?
If so, what on the genetic level contributes to exploration in Drosophila?: What are the differences in genetic architecture between exploring and non-exploring flies?

Bulk-segregant analysis population

G’ statistic

Stochasticity in sequence coverage
Variation in allele frequency estimates due to organism sampling during the formation of the bulks
Null hypothesis: no QTL, i.e., allele is equally expected in both bulks

Results

Randomization significance testing

Markov chain Monte Carlo (MCMC) method
- Resampling algorithm

G’ statistic

Simulated G’ value comparison for null (no QTL) and significant (QTL present)

Allele frequencies of explorers (E) and non-explorers (NE) at significant G’ QTL locations

Chromosome	Position	Gene	Associated phenotype
2L	17054316, 17054384	CG15136	Abnormal flight
	18515161	Ugt201D1	Enables UDP-glycosyltransferase activity
		CG10211	Involved in response to oxidative stress
3L	6818526	vvl	Specification of cell fates, patterning and immune defense
	12226612	app	Regulation of fat signaling, abnormal locomotive behavior
	14401683, 14401693, 14401703	Dscam2	Abnormal neuroanatomy, size, body color
3R	25454633	Men	Abnormal heat stress response, abnormal sleep
	27194130	G14369, CG14370	Little to no information
X	10174752, 10174756	CG32767	Expressed in wing hinge primordium and wing pouch
	12446014	Btnd	Flightless, abnormal heat stress response
		Efr	Manifests in wing vein
		sqh	Involved in cytokinesis and tissue morphogenesis
		dtn	Abnormal heat stress response

Conclusions

Exploration behavior is a heritable trait
Implicated QTL involved in stress & flight traits
Established a novel method of allele frequency estimation in a bulk-segregant experimental paradigm

Overview

Genetic basis of exploration tendency in a multiparent population of D. melanogaster

Phenotypic differences in starvation resistance between selection lines of an experimentally evolved population of D. melanogaster

Starvation resistance

Adaptation that confers resistance to environmental stressor
- ‘Hunker down’ tactic

Starvation resistance

Correlated with higher lipid storage
- More body lipid, higher starvation resistance¹^,²
Evolved quicker in female D. simulans than males²
- 5 generations vs. 15 generations
Alters feeding behavior in D. melanogaster³

Experimental evolution population

Selection experiment

Three selection lines:

Fluctuating availability (FA)
Deteriorating availability (DA)
Constant high availability (CHA)

Three diets:

Control (C)
Dietary restrictive (DR)
High sugar (HS)

Selection	Days post-oviposition
	8 - 13	14 - 17	18 - 21
FA	C	DR	C
DA	C	C	DR
CH	HS	HS	HS
Eclosure: Day 10 Egg collection: Day 21

Main questions

Does starvation resistance co-evolve with resource allocation strategies in D. melanogaster?: I.e., are certain selection lines more resistant to starvation than others?
Is starvation resistance phenotypically plastic?: I.e., can the same genotype give rise to various phenotypes depending on the environment?

Direct measurement assay

Survival analysis

Survival data
- Time-to-event
- Clear start and end time
Time-to-death
- Start: placed onto nutritionless agar
- End: time of death

Censoring survival data

‘Censoring’ occurs if a subject doesn’t experience the event by the end of the experiment
Flies that die due to other factors
- Ex.: crushed by vial plug
Still provide valuable information
- Survival analysis appropriately accounts for this

Survivor analysis and censoring

\[Y_i = min(T_i,C_i)\]

\(Y_i\): observed time
\(T_i\): event time
\(C_i\): censoring time

Kaplan-Meier survival curves

The Kaplan-Meier estimate is the product of the survival probabilities:

\[\prod_{i=0}^n S(t) \] Where we estimate survival probability as the number of flies alive at the current time, divided by the number of flies alive at the previous time.

Cox proportional hazard model

Characteristic	HR¹	95% CI²	p-value
Selection
CH	—	—
DA	1.73	1.30, 2.31	<0.0001
FA	1.41	1.05, 1.88	.022
Sex
F	—
M	1.90	1.50, 2.42	<0.001
¹ HR = Hazard Ratio; ² CI = Confidence Interval

AICc model comparison

Top performing Cox proportional hazard model:

\[Selection + Sex + Batch,\] \[\Delta AIC = 0.00\]

Second-best Cox proportional hazard model:

\[Selection + Sex,\]

\[\Delta AIC = 2.31\]

Main questions

Does starvation resistance co-evolve with resource allocation strategies in D. melanogaster?: I.e., are certain selection lines more resistant to starvation than others?
Is starvation resistance phenotypically plastic?: I.e., can the same genotype give rise to various phenotypes depending on the environment?

Diet treatment assay

The risk (1 - survival probability) of death between males and females

The cumulative hazard (-log (survival probability)) over time for flies from our three diet treatments.

Survival probability in female flies predicted by diet and selection line

Cox proportional hazard model

Characteristic	HR¹	95% CI²	p-value
Diet
C	—	—
DR	0.72	0.54, 0.96	0.027
HS	0.49	0.37, 0.67	<0.001
Selection
CH	—	—
DA	1.16	0.87, 1.54	0.3
FA	1.27	0.94, 1.70	0.12
Sex
F	—
M	3.34	2.52, 4.43	<0.001
¹ HR = Hazard Ratio; ² CI = Confidence Interval

AICc model comparison

Top performing Cox proportional hazard model:

\[Diet + Selection + Sex,\]

\[\Delta AIC = 0.00\]

Second-best Cox proportional hazard model:

\[Diet*Selection + Sex,\]

\[\Delta AIC = 2.96\]

Conclusions

CHA & FA selection lines confer higher SR resistance
- ‘Thrifty’ spending vs. abundance of riches
Both CHA selection line and HS diet groups resisted starvation for a significantly longer period of time
- Flies from CH selection line on DR diet treatment had worse SR
In direct measurement flies, we saw significant differences in SR for the selection groups in both KM and Cox
In diet treatment flies, selection by itself was not a significant predictor, while diet was
- Diet + Selection, and Diet * Selection, however, were significant in our Cox model

Summary

Adaptations to environmental stress
- ‘Leaving’
- ‘Hunkering down’
Identified candidate genes underlying exploration tendency
- Some genes involved with other environmental stress traits
Observed the co-evolution of starvation resistance alongside resource allocation strategies
- Resource allocation strategies confer various levels of SR
- An organism’s environmental upbringing impacts SR
- Eating sugar is good for you!

Acknowledgements

The King Lab, past and present
Dr. Libby King

My committee
  Dr. Lauren Sullivan
  Dr. Rex Cocroft
  Dr. Greg Blomquist

My friends & family
  Dr. Arianne Messerman
  Mom, Dad, Liv, Sam & Haley
  Jo Moaton

Funding
Life Sciences Fellowship
NIH

DNA Core

Special thank you to Debbie Allen, Rebecca Ballew, Nila Emmerich, Melody Kroll, the IT office and the RSS group.

Extra slides

Different approaches

DROP THIS SLIDE, put in extra slides – put ‘main questions’ slide here again - QTL mapping - Requires family information - Maps linkage genes - Identification of QTL underlying a phenotype via analysis of polygenic inheritance - Genome-wide association study - Requires phenotyping of hundreds of RILs - Whole-genome sequencing data - Associates small changes in polygenic SNP frequencies with phenotype - Bulk-segregant analysis - Allows for rapid testing of extreme phenotypes - Need genetic variation for extreme selection to act on

Survival probability

The probability that a fly will survive past a given time, i.e., survival probability:

\[S(t) = Pr(T > t) = 1 - F(t),\] \[F(t) = Pr(T \le t)\]

Statistical error in BSA-seq

estAF <- function(cvg,trueAF){
  est <- rbinom(1,cvg,trueAF) / cvg
  return(est)
}

Function calculating estimated allele frequency
Takes in coverage and ‘true’ allele frequency
Outputs estimated allele frequency


# run this n times, scale up 
n <- 1000
tru <- 0.43
cvg <- sample(2:150,n,replace = TRUE)
af <- tibble(
  "cvg" = cvg,
  "trueAF" = rep(tru,times=n)
)

Variability in allele frequency estimations due to coverage.

Significance thresholds

False discovery rate (FDR)
- Expected proportion of false “discoveries,” i.e., Type I error
Family-wise error rate (FWER)
- Probability of making >= 1 false “discoveries”
- Stricter than FDR

Null hypothesis is…	TRUE	FALSE
Rejected	Type I error	Correct
Accepted	Correct	Type II error