SOME NOTES ON HUMAN GENETICS AND GENETIC EPIDEMIOLOGY

Overview of genetic epidemiology

The introductory slides came from here.

Recurrence Risks

Familial aggregation of a trait is suggestive but not proof of a genetic contribution.

The recurrence risk is the probability a person expresses a disease given they have an affected relative (of a given type). For example, the population lifetime risk of developing psoriasis is approximately 2%, but if one has an affected parent or sibling the recurrence risk is 20%.

The sibling recurrence risk ratio (PRR_sib, or often referred to as lambda_s) in this case is 10. The size of PRR determines how easily linkage studies can detect a gene underlying a trait.

Classical twin study

The classical twin design dates from the 1920's, when diagnosis of identical (MZ) and nonidentical (DZ) twins became reliable. It is assumed that twins in the same household are exposed to shared environmental risk factors to the same extent whether they are MZ or DZ. Then, MZ and DZ twins differ only in the proportion of genes they share. Recurrence risks of disease in a cotwin of a case can then be interpreted as follows:

Rpop = RDZ = RMZ	No familial factors involved
Rpop < RDZ = RMZ	Family environmental factors
Rpop < RDZ < RMZ	Genetic factors +/- family environment
Rpop < RDZ << RMZ	Epistatic polygenes and/or gene-environment interaction

Pooling two twin studies of psoriasis (Australian and Dutch):

Twin type	Recurrence Risk	95% Confidence Interval	PRR
MZ	60%	47-71%	30
DZ	22%	10-37%	11

Twin concordance and recurrence risk

Estimation of recurrence risk in families is confused by the ascertainment mechanism, that is the method of sampling or recruitment.

Example: The total population of twins (in the Australian NHMRC Twin Registry say)

		Twin 1
		Aff	UnA
Twin 2	Aff	a	b
	UnA	c	d

The recurrence risk is a/(a+b) or a/(a+c), or better still the ``average'' a/(a+0.5(b+c)).

In the usual situation, twins don't have a label, so one instead sees:

Aff-Aff	Aff-UnA	UnA-UnA
a	b+c	d
C	D	U

The correct formula for the recurrence risk is then:

R = C/(C+0.5 D) = 2C/(2C+D).

I belabour the point because the pairwise concordance (C/(C+D)) is often quoted in the literature, and does not estimate the recurrence risk in the same way as does this probandwise concordance.

The general rule for dealing with ascertainment

A proband is a family member whose disease status brought the family into the study. If a family contains two probands, then it should contribute twice to the estimation of recurrence risk. In the case of complete ascertainment, every affected person is a proband. This is the type of ascertainment one would see in a population survey where all members of a family are automatically examined.

For the twin example above, every family with two affected twins is counted twice, and every family with one affected twin is only counted once, so,

Number of times a cotwin is affected given the proband is affected	2C
Total number of families (counting those with two probands twice)	2C+D

If one is carrying out a twin study in a clinic, often only one twin will be the proband (patient). If the cotwin is subsequently found to be affected, s/he will be a ``secondary'' case, and this family will only be counted once. This is single ascertainment.

Caveats to the classical twin method

• Discordance of X-inactivation (MZ discordance for highly penetrant X-linked disorders
• Increased maternal hormone levels (breast and ovarian Ca possibly increased in twins
• Sharing a uterus (birth weight, complications, ``fetal programming'')
• Sharing a placenta (monochorionic versus dichorionic MZ twins)
• Inequality of parental treatment (MZ versus DZ).
• Greater MZ than DZ sharing of environmental exposures (``social contagion'', education, job choice)

Cotwin control study

One commonly matches cases and controls in observational and intervention studies to control confounding. MZ twins are well matched on genetic and family environment, so only small sample sizes are needed. Some examples:

• The shorter twin has a higher rate of atherosclerotic heart disease than the taller cotwin.
• The twin with tibiofemoral osteophytes weighs 4 kg more than the twin without radiographic changes.
• The smoking twin has lower bone density, larger carotid plaques etc.
• The twin who served in Vietnam has nine times the rate of PTSD.
• The schizophrenic twin has smaller hippocampi.
• The twin taking ascorbic acid gets just as many colds.

Mitochondrial inheritance

• Each cell and mitochondrion contains multiple copies of the mitochondrial genome (the latter 2-10 copies).
• Mitochondria are transmitted from mother to child, so that mitochondrial diseases are inherited matrilineally.
• The mitochondrial genome contains 37 genes (13 respiratory chain proteins, 22 tRNAs, 2 rRNAs) in 16569 bp.
• The mutation rate of mitochondrial DNA is high (repair mechanisms are inferior to those for nuclear DNA, and recombination is rare or absent).
• Several hundred mutations leading to disease are known: OMIM has 240 entries for mitochondrial inheritance, and lists 18 mutations leading to Leber's hereditary optic neuropathy.
• Heteroplasmy refers to the fact that multiple different mitochondrial genotypes can be present throughout one organism.
• Most disease mutations are ``recessive'' in nature, in that they have to make up 70% or more of all mtDNA before disease occurs in that organ (heteroplasmy can vary by tissue).
• The disease phenotype is quite variable within families due to heteroplasmy.
• Mitochondrial diseases can affect all systems, but have a predeliction for CNS and muscular (including cardiac muscle) manifestations.
• A number of mitochondrial mutations increase with age through replicative errors during mitosis eg the ``common deletion'' (nt 8470-13447).

Prototypic mitochondrial diseases

Leber's hereditary optic neuropathy (LHON, MIM 535000)

• Midlife central vision loss, cardiac conduction defects (WPW, LGL), multiple sclerosis, more widespread CNS degeneration (putamen, basal ganglia).
• Several hundred families worldwide (including Australia).
• Male preponderance in Europeans (confusing the mode of inheritance).
• Three point mutations explain 90% of cases: at nucleotides 11778, 3460, and 14484. A better visual prognosis is associated with the bp 14484 mutation (NADH dehydrogenase subunit 6, Met64Val).

Mitochondrial myopathy, encephalopathy, lactic acidosis, and strokelike episodes (MELAS, MIM 540000)

• Episodic vomiting, seizures, recurrent cerebral infarcts.
• Diabetes mellitus, sensorineural hearing impairment, hypertrophic cardiomyopathy, ataxia, basal ganglia calcification, opthalmoplegia, cognitive decline, short stature.
• Ragged-red muscle fibre changes on occasionally.
• The 3243A>G mutation (tRNA-leu) has an estimated frequency 16 per 10⁵ in northern Finland.
• 3243A>G mutation is present in 1% of unselected patients with diabetes mellitus, and 2-5% of those with a matrilineal history of diabetes.
• 3243A>G mutation is present in 0.07% of unslected patients with deafness, and 7% of those with a matrilineal history of sensorineural deafness.

Myoclonic epilepsy associated with ragged-red fibres (MERRF, MIM 545000)

• Familial myoclonic epilepsy.
• 90% of cases due point mutation nucleotide 8344A>G.
• Histological muscle changes.

Parent-of-origin effects and imprinting

• In crosses between breeds/species (eg horse with donkey, lion with tiger), it is known that the phenotype in the offspring depends on the specific parental phenotype: imprinting.
• Several clusters of autosomal genes act differently depending on which parent transmits them. Methylation of the CpG rich imprinting control region (ICR) upstream from the target loci leads to binding of methyl-CpG-binding proteins (eg MeCP2) and subsequent compacting of chromatin packing and thus gene silencing. In this context, imprinting refers to inactivation of that specific parental allele in the offspring. A side effect is asynchronous replication at the locus (the imprinted allele replicates early).
• Different parentally derived alleles at loci in the same imprinted region may be inactivated.
• Other mechanisms can give rise to the same parent-specific transmission of phenotype. Maternal effects mediated by the hormonal milieu of the fetus can reinforce a phenotype only when the child carries the same genotype as the mother at a locus.
• Parent-of-origin effects in the triplet expansion diseases are mediated by differential expansion of repeat number. For example in Huntingdon's disease (MIM 143100), paternally transmitted alleles expand more than maternal alleles, so that offspring inheriting the disease from a father are more severely affected, and at an earlier age.

Examples of parent-of-origin effects

IGF2 (MIM 147470)

• Only the paternally derived IGF2 allele is expressed normally after the 8 cell stage in embryonic development: maternal imprinting.
• At the neighbouring H19 locus, paternal imprinting is the norm.
• Both Beckwith-Wiedemann syndrome (MIM 130650) and a pure somatic overgrowth syndrome are associated with biallelic expression of IGF2.

Prader-Willi and Angelman (``happy puppet'') syndrome

• Both loci are within a 2 Mb imprinting region on 15q11-13.
• Deletion of the promotor and first exon of SNRPN gene along with a small upstream ICR on the paternally derived chromosome leads to Prader-Willi syndrome (MIM 600161), and a similar deletion of the maternally derived chromosome to Angelman syndrome (MIM 105830). Small nuclear ribonucleoprotein polypeptide N (snRPN, MIM 182279) is predominantly expressed in brain.
• Uniparental disomy is the other mechanism whereby alleles derived from only one parent are present in the region.
• Some Angelman families have simple point mutations in SNRPN.

Atopic disease

• Loci on chromosome 11q13 (FCER1B), 13q, and 3q have been found to exhibit linkage only through one parent and not the other to atopy, atopic dermatitis and asthma.
• FCER1B is in a region homologous to mouse imprinting regions, but true imprinting has not been documented.

Linkage and association analysis

There is much material on this here.

Lod score linkage analysis

Lod score linkage analysis requires a roughly correct model for transmission of the trait. This is used to estimate the most likely trait locus genotype each member of a family carries.

The recombination distance between the trait locus and one or more known marker loci can then be estimated. The strength of evidence for linkage (recombination fraction < 50%) is traditionally expressed as the decimal log odds ratio: the lod score. Given the size of the human genome, a lod score of 3 (equivalent to a P-value of 2x10^-4) is regarded as significant evidence for linkage, and a lod score of -2 as definitively excluding linkage.

Nonparametric linkage analysis

A number of statistical methods have been developed that do not require a trait locus model to be specified. The most common approach one will encounter in the literature is the affected sib pair method. This requires families containing two affected siblings, and can be shown to be equivalent to a lod score linkage analysis assuming a completely recessive model for the trait locus.

Haseman-Elston regression and variance component linkage analysis are nonparametric methods that can be applied to continuous traits (eg blood pressure).

Transmission-disequilibrium test

The transmission-disequilibrium test (TDT) tests the transmission ratio of alleles at a marker locus from a heterozygous parent to an affected child. There is distortion of the ratio from the Mendelian 50:50 only if the marker locus is linked to the trait locus and exhibits allelic association. If a marker locus is close (tightly linked) to the trait locus, then allelic association may be observed, but association does not automatically imply tight linkage.

Genetic marker loci: SNPs, RFLPs, microsatellites

To be useful in the detection of cosegregation of alleles at a marker locus with a trait, the marker needs to be polymorphic. This is measured usually as the expected heterozygosity of the marker (proportion of individuals expected to be heterozygous based on the allele frequencies).

Simple Sequence Repeat polymorphisms (SSRs)

These are common polymorphic loci where alleles are differing number of repeated sequences of 2, 3, 4 or more nucleotides. Also known as microsatellites or Short Tandem Repeat Polymorphisms. After PCR amplification of a length of DNA that includes the SSR, the different size fragments are identified by gel electrophoresis. They are currently the most utilised genetic markers for linkage analysis.

Single nucleotide polymorphisms (SNPs)

These are simply point mutations of a sufficiently high frequency to be useful for linkage and association analysis. The nucleotide-wise mutation rate is approximately 10 @Sup{-6} and randomly occurs within coding or noncoding regions, so that there is approximately one SNP per 1000 base pairs. Because of this density, they are more useful for fine mapping of trait loci known to be within a certain chromosomal region.

Restriction Fragment Length Polymorphisms (RFLPs)

This refers to a method of genotyping SNPs. PCR is used to amplify a segment of DNA that includes the SNP. A restriction enzyme is chosen that recognises one of the two variants at a SNP, and will cleave the DNA strand only when that variant is present so either one or two fragments will be formed.