1.1Molecular Population Genetics

CHAPTER 1:

Introduction

During the past century some major achievements in biology divide it approximately in three periods: the introduction of the Modern Synthesis in the 30's, the emergence of Molecular Biology in the 50's and the start ofthe current 'Omics Era'in 90's. Genomics, being maybe the first 'omic' discipline, finally left behind decades of active struggling to studywhole genomesthoroughlyand opened the door to the complete genetic sequence of organisms.

During the past two decades "numerousderivatives of the basic concept of large-scale biological analyses" (Ellegren 2014) have emerged and started adding the suffix 'omics' to their name. One of them is Population Genomics.

But, what is Population Genomics? It's simply "a new term for a field of study as old as Genetics itself" (Charlesworth 2010). It's the 'old field' of Population Genetics when studies the amount and causes of variability in natural populations in a genome-wide fashion. Population Genomics, along with Bioinformatics, are the main topics of this Thesis project.

1.1Molecular Population Genetics

1.1.1. Mutations at the DNA molecule

Since Charles Darwin most universally known publication, "The origin of Species" (1859), it became accepted that heritable variation in any trait must exist before it can undergo any process of adaptation by Natural Selection. Hence the study of variation within individuals and populations is crucial to understand any process of evolutionary change.

But for many years during the 19th century and the start of the 20th, variation could only be studied in phenotypic traits, where discrete Mendelian variation is rare and quantitative traits are abundant. A phenotypic trait results from the interaction between a given genotype (which is heritable) and a specific environment. Observed phenotypes are the final result of many interactions difficult to discern. Therefore, direct study of the hereditary material is crucial to understand the processes that originated and lead to adaptation to any trait.

Avery et al. discovered in 1944 that the DNA is the molecule that carries the genetic information. Despite that, genetic variation was studied indirectly using gel electrophoresis of proteins(Johnson et al. 1966; Lewontin and Hubby 1966, Harris 1966). It was not until the late 70's that actual variation in theDNA molecule was analyzed using restriction enzymes (Avise et al. 1983). The final milestone in DNA studies was the development of sequencingtechnologies (Sanger and Coulson 1975, Maxam and Gilbert 1977).The automation and parallelization of the Sanger method was the key that provided us with the impressivenumber of sequenced genomes in practically 20 years. Nowadays, more advanced andhigh throughput next generation sequencing (NGS) methods are used to analyze several types of variation in the DNA sequence (Table 1.1).

Table 1.1 Common DNA mutation types.

[Adapted from Casillas 2007 and Freeman et al. 2006]

Mutation is the ultimate source of genetic variation. Once a new variant appears by mutation in the DNA it can be replicated and transmitted from generation to generation. During a while most studies of genetic variation focused on single-nucleotide differences among individuals. Although only one nucleotide is affected, their abundance in the genome makes them the most frequent source of inter-individual genetic variation in respect to other variation types. Single nucleotide polymorphisms (SNPs) were believed to account for >90% of the genomic variability in humans (Collins et al. 1998). However, variation in genomes range from single-nucleotide changes to large (>3 Mb), microscopically visible karyotypic alterations (Feuk et al. 2006, Conrad and Hurles 2007). Regardless of being in minor amounts than single-base variants, all this other non-SNP variants comprise a significant fraction of a genome since each one involve a longer segment of DNA.

Following SNPs, the next most abundant form of genetic variation are short (≤ 50bp) insertion and deletions (indels), at least in the human genome (Montgomery et al. 2013). But compared to SNPs and larger structural variants, their origins and functional effects were poorly understood until now. Different than SNPs, that can appear mostly by external mutagenic agents and/or replication/reparation errors, indels can have various mechanisms of origin. Most short indels seem to be caused by polymerase slippages (Montgomery et al. 2013, Streisinger et al. 1966, Levinson and Gutman 1987, Greenblatt et al 1996, Taylor et al. 2004), but they can be originated by other mechanisms also implied in other structural variants such as imperfect repairs of double-strand breaks (Chu 1997, McVey et al. 2004), fork stalling and template switching (FoSTeS) and microhomology-mediated break-induced replication (MMBIR) (Lee et al 2007, Hastings et al 2009) and hairpin loop formation due to presence of palindromic sequences (Greenblatt et al 1996, Hastings et al 2009). Longer indel mutations can be also caused by transposable elements or by complex meiotic events when related to other genome reorganizations.

A clustering of SNPs and indels appears to be ubiquitous in prokaryotes and eukaryotes (Tian et al. 2008, Hodgkinson and Eyre-Walker 2011, McDonald et al. 2011, Jovelin and Cutter 2013). Various mechanisms have been proposed to explain this: (i) possibly indels and repeats are mutagenic because they induce errors during the DNA polymerase replication near them (Tian et al. 2008; Jovelin and Cutter 2013,McDonald et al. 2011, Yang and Woodgate 2007), (ii) maybe the regions in which SNPs and indels occur are inherently mutagenic (Hodgkinson and Eyre-Walker 2011), (iii) indels may cause aditional point mutations when segregating due to pairing problems during meiosis (Hodgkinson and Eyre-Walker 2011) or, (iv) SNPs and indels occur simultenously and are subject to the same population genomic processes (Hodgkinson and Eyre-Walker 2011).

As a note, the term 'structural variation' is commonly used referring to variation of more than one nucleotide. This can lead to confusion since in the literature structural variation is the one involving segments of DNA longer than 1kb (Feuk et al. 2006). In the current work we'll try to simply distinguish between SNP and non-SNP variation, since most of the population genomics data presented in the next chapters comprises <1kb variants.

1.1.2 The dynamics of genetic variation

The Hardy-Weinberg principle, the mathematical model formulated in 1908 by G. H. Hardy and W. R. Weinberg served as a null model to explain the fate of genetic variation in a population during the first years of genetics. The principle explained that, in an ideal population in absence of any other evolutionary forces, allele frequencies would remain the same generation after generation. The forces that can affect allele frequencies in a population are principally mutation, migration, natural selection, recombination and random genetic drift.

A mutation is a random change in the genomic sequence of an individual. Most mutations will not will be lost the generation they appear either by genetic drift or because the change prevents the individual its reproduction. Sometimes, a mutation can increase its frequency in the population through generations, either again by chance or because it gives some advantage to the population. These allelic variants are now part of the variation within the population, this is called polymorphism. We define as fixation or substitution the process by which one allele segregating as polymorphism increases enough in frequency to replace allthe other alleles in the population.

Once in a while, two different populations of the same species may become isolated and after many generations of different independent allele fixations they can become two new different species (speciation). This accumulation of distinct allele fixations is referred to as divergence and is ultimately responsible for all the rich diversity in life forms we can see in Earth. Polymorphism and divergence can tell us different and complementary things about the history of a population: while polymorphism reflects recent changes, divergence provides information from more ancient times. The combined analysis of polymorphism and divergence is one of the most powerful approaches to understand the influence of different forces on the patterns of evolutionary change.

1.1.3 A short History of Population Genetics

The main aim of population genetics is the description and interpretation of genetic variation within and among populations (Dobzhansky 1937). Its mathematical foundations were established by R. A. Fisher, J. B. S. Haldane and S.Wright between 1910 and 1930. They figured out the consequences of chance and selection in populations with Mendelian inheritance, and turned population genetics into the explanatory core of the evolutionary theory. In the late 1930s and 40s, the integration of theoretical population genetics with other evolutionary research fields such as experimental population biology, paleontology, systematics, zoology and botany gave rise to the Modern Synthesis of evolutionary biology (Dobzhansky 1937; Mayr 1942; Simpson 1944; Stebbins 1950).The main difference between the modern synthetic theory and that of natural selection as set forth byDarwin is the addition of the Mendelian laws of heredity in a population genetics framework, therefore, this new theory is also called neo-Darwinism.

Neo-Darwinists thought that natural selection was the most important mechanism to explain evolution in the detriment of drift and other non-adaptive variation. In a first attempt to measure variation, two different models emerged. The ‘classical model’ supported the role of natural selection in purging the population of any new mutation and thus predicted that most gene loci are homozygous for the wildtypeallele (Muller and Kaplan 1966). On the other hand, the ‘balance model’ predicted that natural selection maintained high levels of genetic diversity in populations by favouring heterozigosity atmany gene loci (Dobzhansky 1970; Ford 1971). Between these two explanations, only the balance model can explain how a population can respond quickly to environmental changes by selecting variation already present in the population and changing its frequencies. This debate was not resolved even after the estimation of genetic diversity was available for the first time.

With the appearance of the electrophoretic techniques mentioned in the previous section, population genetics entered the molecular age, the so-called 'Allozyme era'. The results of such electrophoretic surveys revealed a large amount of genetic variation in most populations (Nevo et al. 1984) , much more than had been predicted, and seemed to better support the balance model rather than the classical model. Also,levels of genetic diversity were found to vary nonrandomly among populations, species, higher taxa and several ecological, demographic and life history parameters (Nevo et al. 1984).

At the time, a new theory was developed to explain the patterns of molecular genetic variation within and among species. In contrast to the selectionistargument of the balance hypothesis, Kimura’s Neutral Theory of molecularevolution suggests that most polymorphisms observed at the molecular level are eitherstrongly deleterious or selectively neutral, and that their frequency dynamics in apopulation are determined by the rate of mutation and random genetic drift rather thannatural selection (Kimura 1968). Some of the principal implications of the neutral theory are:

1. Deleterious mutations are rapidly removedfrom the population, and adaptive mutations are rapidly fixed; therefore, most variationwithin species is the result of neutral mutations (Figure 1.1).
2. A steady-state rate at which neutral mutationsare fixed in a population (k) equals the neutralmutation rate:k = fneutralμ, where fneutral is theproportion of all mutations that are neutral andμ is the mutation rate.
3. The level of polymorphism in a population (θ)is a function of the neutral mutation rate andthe effective population size (Ne): θ = 4Ne μ.
4. Polymorphisms are transient (on their way toloss or fixation) rather than balanced byselection. Larger populations are expected tohave a higher heterozygosity, as reflected in thegreater number of alleles segregating at a time.

By extension, the hypothesis of selectiveneutrality would also apply to most nucleotide or amino acid substitutions that occurduring the course of evolution. However, Kimura emphasized the compatibility of hisneutral theory —mainly determined by mutation and drift— at the molecular level, withnatural selection shaping patterns of morphological variation. There have been some refinements to the neutral theory, specially the nearly-neutral and slightly deleterious mutation hypotheses of Tomoko Ohta (Ohta 1995), that modules the original theory considering that slightly deleterious variants could still segregate at low frequencies in the population (Figure 1.1). In any case, the Kimura’s neutraltheory is the theoretical foundation of all molecular population genetics.

A corollary of the neutral theory is the existence of a random molecular clock,which had already been previously inferred from protein sequence data byZuckerkandl and Pauling (1962).Since the rate at which neutralalleles are fixed in a population equals the neutral mutation rate, when twopopulations or species split, the number of genetic differences among them isproportional to the time of speciation. On that account, the number of differences amonga set of sequences from different species can be used as a molecular clock to allow sortingthe relative times of divergence among these species.

Figure 1.1 Diagram showing the trajectory of different types of alleles in a population.

New variants that appear in a population start segregating and, after a time, they can become fixed (freq. 1) or disappear from the population (freq. 0). In green are represented the dynamics of neutral alleles reaching fixation or disappearing by random mechanisms. In blue are represented advantageous variants that become fixed quickly. In brown are represented slightly deleterious mutations that cannot grow much in frequency but they can segregate in the population for a time before disappearing. Red dots represent strongly deleterious mutations that are removed quickly from the population. [Adapted from Hartl and Clark (1997)]

1.4 Estimation of genetic variation

The data desideratum for population genetics studies is a set of homologous and independent sequences (or haplotypes) sampled in a DNA region of interest. From a set of haplotypic sequences two types of nucleotide diversity can be estimated: one can take each nucleotide site independently (one-dimensional) or a segment of sites can be analyzed together taking into account associations between alleles (multi-dimensional) (Table 1.2). Nearby nucleotides are not independent from each other since they tend to be clustered in blocks of different lengths, for example, up to 2kb in Drosophila(Miyashita and Langley 1988) and over several megabases in the human genome (Frazer et al. 2007). Multi-dimensional estimators are important to describe the forces that shape haplotypes such as recombination, selection and demography. Both one and multi-dimensional diversity components are indispensable for a complete description of sequence variation.

Table 1.2 Common measures of nucleotide diversity

[from Casillas 2007]

1.1.5Detecting selection in the genome

Looking for evidence of selection is a widely-used strategy for finding functionalvariants in the genome (Bamshad and Wooding 2003). Natural selection can leave several types of signatures in the genome: (i) a reduction in polymorphism, (ii) a skew towards rare derived alleles, and (iii) an increase in linkage disequilibrium (LD)(Bamshad and Wooding 2003). As mentioned previously, hitchhiking events reduce local levels of variation. Over time, since common neutral variants have disappeared, new mutations begin to appear again and start segregating at low frequencies, leading to an excess of new rare derived alleles in the region. Also, a long region with high LD and low diversity can indicate recent positive selection over an allele if present at high frequency, since recombination still has not had enough time to reduce the LD.

Background selection is another process that can reduce genetic diversity in a genome region, due to removal of chromosomes with strongly deleterious mutations. However, it's different from a hitchhiking event as it does not create blocks of LD nor presents rare variants at low frequencies. The effect in this case is a reduction of the number of chromosomesthat contribute to the next generation, which is identical to that of a reduction in population size except that the reduction applies, not to the genome as a whole, but to atightly linked region (Charlesworth et al. 1993).

Variations in the local rate ofrecombination along the genome also make the detection of selection difficult, since thesignatures of selection highly depend on the local rate of recombination (Hudson and Kaplan 1995). In this regard, the effects of non-selective forces like demography and recombination should be taken into account when trying to identify regions showing true signatures of adaptive evolution. Several tests based on the level of variability and the distribution of alleles havebeen developed to identify the footprints of selection searching for the previous signatures of selection (Table 1.3).