Genetics of Endocrinology: Amit R. Majithia - David Altshuler - Joel N. Hirschhorn

CHAPTER 4
Genetics of Endocrinology
AMIT R. MAJITHIA • DAVID ALTSHULER • JOEL N. HIRSCHHORN
The Role of Genetics in Endocrinology, 49 With the expanding reach of precision medicine—
Principles of Genetics, 49 individualized diagnosis and therapy informed by genetics—
we anticipate that increasing numbers of patients will have
Genetics of Endocrine Diseases, 54 clinical indications for exome or genome sequencing, and
Considerations for Clinical Use of Genetic Information and others will come to clinical encounters with their sequences
Sequencing in Endocrinology, 62 already in hand. Clinicians will be asked to interpret these
genetic data to shed light on an individual’s risk of devel-
oping disease, on diagnosis and prognosis for those already
KEY POINTS affected, on implications to family members, and on indi-
vidualization of therapy. As such, it is critical that clini-
• The genetic basis of each heritable endocrine disease/ cians be able to draw valid and clinically useful connections
trait is quantified by its genetic architecture: (1) the between DNA sequence variation and human traits and
number of genetic variants/genes, (2) their frequency in diseases. Perhaps even more important, it is critical that
the population, and (3) their respective contributions to clinicians understand the limits of such information.
disease risk/phenotypic variation. In this chapter we present a guide to help clinicians
• Mendelian endocrine disorders are caused by few appreciate and critically interpret the relationship between
variants in few genes, found rarely in the population, DNA sequence (genotype) and an individual’s clinical pre-
and each has a large individual effect on disease risk. sentation (phenotype). We first discuss principles of genet-
• Common endocrine diseases/traits such as stature, type ics to provide the framework for understanding and
2 diabetes, and serum lipids are polygenic—the result of interpreting DNA variation in patients. We then focus on
combined, simultaneous effects of many variants in many endocrine disorders, providing an overview of the genetics
genes, found frequently in the general population, and of endocrine diseases, with illustrative examples from both
with each variant contributing a small individual effect. mendelian disorders (caused by mutations in single genes)
• Genetic information enables endocrinologists to personal- and polygenic disorders (in which variation in many genes
ize therapy for patients. influences disease risk). Finally, we examine scenarios for
• Comprehensive genetic testing (i.e., genome sequencing) clinical uses of genetic information in endocrinology and
can be standardized and automated, but drawing valid provide recommendations.
and clinically useful conclusions requires integration with Most diseases, including endocrine disorders, are heri-
patient history, physical examination, and other laboratory table, meaning that genetic variation contributes to dis-
examinations. ease risk in a population. These diseases range across the
• Genetic information is most likely to be of direct clinical spectrum of rare, single-gene disorders, such as multiple
use in patients with suspected mendelian syndromes. endocrine neoplasia (Chapter 39), Carney complex (Chap-
ter 15), and congenital adrenal hyperplasia (CAH) (Chap-
ter 23), to polygenic diseases, such as type 2 diabetes
(Chapter 31), Graves disease (Chapter 12), and osteoporo-
sis (Chapter 28). The detailed discussions of the genetics
THE ROLE OF GENETICS of these and other disorders can be found throughout this
IN ENDOCRINOLOGY textbook; this chapter will provide illustrative examples
that illuminate key concepts and will refer the reader to
The sequencing of the human genome has ushered in an those appropriate chapters for additional detail.
era of genomic medicine. The catalog of protein-coding
genes in humans is essentially complete, and the number
of associations between genes and specific diseases is PRINCIPLES OF GENETICS
growing rapidly. Moreover, it is now feasible to identify
nearly every genetic variant in an individual’s protein-
coding genes (whole-exome sequencing) or in his or her
A Brief Historical Perspective
entire genome (whole-genome sequencing). The ability to In Western conception, the relationship between inheri-
interpret this variation is less advanced but is improving, tance and physical characteristics (disease and nonpatho-
as databases of variants and their clinical associations logic) has been recognized since the time of Aristotle (323
increase in both size and accuracy. BC). But it was not until 1865 that the Austrian abbot
49
50 SECTION I Hormones and Hormone Action
Gregor Mendel, after decades of careful experimentation in 2005 Phase I), to systematically identify the genetic causes
pea plants, posited and provided evidence for the modern for common polygenic diseases.4
genetic concept of genes (as coined by the botanist Wilhelm
Johannsen in 1909).1 Mendel deduced certain rules govern- Heritability: An Estimate of the Importance of
ing the passage of genotype (the collective versions of
multiple genes in an individual) from parent to offspring,
Genetic Factors to Disease Causation
enabling the prediction of the resulting physical character- Relatives resemble each other in many ways. Resemblance
istics (phenotype) of the offspring. It was recognized in the with respect to traits such as height or to diseases such as
early 20th century that certain human phenotypes, includ- multiple endocrine neoplasia type 1 (MEN1) could be
ing diseases, were inherited according to the same rules explained by shared genotypes passed down through gen-
that Mendel had described; these diseases are called erations, shared environments, and nonlinear interactions
mendelian. between genes and environment. Heritability quantifies, as
Over the course of the next century, numerous break- a proportion, how much of this familial resemblance is due
throughs established that genes were composed of DNA, to genetic factors. A trait that has no genetic influence
physically connected on chromosomes, and encoded pro- would have a heritability of 0%; a trait that is completely
teins. The first description of the molecular basis of a men- determined by inherited factors would have a heritability
delian disease was made for sickle cell anemia, which of 100%. Most clinically important traits have heritabilities
involved a mutation in a single gene. In the 1970s, the ranging from 20% to 80% (Table 4-1). Appreciating the
ability to sequence DNA revealed natural and heritable heritability of a trait is important when interpreting the
sequence variation (genetic polymorphisms) in any given contribution of genetic risk factors in disease: genetic
gene among different individuals. It was appreciated that factors are less influential for traits with low heritability
the molecular basis of variation in the genotypes of indi- and are likely to have more predictive or explanatory
viduals resulted from DNA sequence polymorphisms, power for traits with high heritability.
which in turn effected alterations in phenotype. By tracing In the past, the gold standard for heritability estimation
the transmission of these polymorphisms in families, it was the comparison of monozygotic and dizygotic twin
became possible to identify genes causing mendelian concordance rates for diseases/traits. Such studies relied on
human disorders (those caused by altered function in a the rationale that an excess of disease correlation between
single gene and that consequently show distinctive pat- genetically identical individuals (monozygotic twin pairs)
terns of inheritance in families).2 as compared to those who share only 50% of their genes
However, most human diseases and phenotypes are not (dizygotic twin pairs) pointed to the role of genetic factors.
mendelian. Biometricians had appreciated in the early However, the validity of comparing twin concordance rates
1900s that most continuous and commonly varying traits across different families relied on the assumption that the
(such as height and blood pressure) did not follow mende- effect of environment was the same for the twin pairs,
lian patterns of inheritance. In 1918, R.A. Fisher3 provided regardless of whether they were monozygotic or dizygotic
a general framework explaining continually varying traits twins. More recent methods for heritability estimation can
as the consequence of polygenic inheritance; that is, poly- overcome some of these limitations by leveraging subtle
genic phenotypes are a result of combined, small, and addi- fluctuations in genetic similarities between sibling pairs.5
tive effects of variation in many genes simultaneously. In Regardless of the methodology employed, it is critical
this framework, monogenic/mendelian traits were a special to appreciate that heritability is not a fixed property of a
case. Despite this recognition, only a few genetic variants disease/trait. The heritability estimate from any study must
were convincingly connected with polygenic diseases/traits be interpreted in the context of the population in which
over the next 80 years. It would take a series of technologic it is being measured, including the historical period, and
advances, including the sequencing of the human genome variability in environmental factors such as socioeconomic
(Human Genome Project 1990-2003) and the systematic status and nutrition. These factors likely explain the wide
cataloging of DNA sequence polymorphisms across diverse range of heritability estimates for type 2 diabetes, ranging
human populations (International HapMap Project 2002- from 40% in Finland6 to 80% in Japan.7 An illustrative
TABLE 4-1
Heritable Endocrine Traits and Diseases
Common Form Heritability Reference* Selected Mendelian Forms
Type 1 diabetes 80% 117 KCNJ11, ABCC8 (permanent neonatal diabetes)
Type 2 diabetes 40-80% 29, 34, 118 AGPAT2 (congenital generalized lipodystrophy), LMNA (familial partial lipodystrophy 1) HNF4A,
GCK, HNF1A (MODY1-3)
Obesity 40-70% 119, 120 MC4R, POMC
Hypertension 30-70% 121 MEN1, RET (MEN2A/B), VHL, SCNN1A (Liddle syndrome)
Height 80% 30, 61 GH1, FGFR3 (achondroplasia), SHOX1 (Ullrich-Turner syndrome), FBN1 (Marfan syndrome)
Pubertal timing 50-80% 122 KAL1, KISS1R, FGFR1 (hypogonadotropic hypogonadism)
Graves disease 80% 123 TSHR (familial nonautonomous hyperthyroidism)
Hypothyroidism 67% 124 TSHR, SLC5A5, TG, TPO, and TSHB (congenital hypothyroidism)
Osteoporosis 50-85% 125, 126 COL1A1, COL1A2, IFITM5 (osteogenesis imperfecta)
Serum calcium 40% 127, 128 CASR (familial hypocalciuric hypercalcemia), HRPT2 (hyperparathyroid jaw-tumor syndrome)
Lipids 40-60% 81, 82 LDL: LDLR (familial hypercholesterolemia)
HDL: CETP
Triglycerides: APOE (familial dysbetalipoproteinemia)
Kidney stones 56%129 130 CLCN5 (X-linked recessive nephrolithiasis), NKCC2 (Bartter syndrome)
*Numbers in this column indicate references listed at the end of the chapter.
CHAPTER 4 Genetics of Endocrinology 51
example of the importance of history can be drawn by usually does not affect function. SNPs can be missense
examining type 1 diabetes rates across the Scandinavian changes (alteration of a single amino acid in a protein-
region of Karelia. In 1940 this region was divided between coding gene) as is the case of the C282Y mutation in the
Finland and the former Soviet Union with little contact HFE gene responsible for autosomal recessive hereditary
between the two sections over the next 60 years. Finnish hemochromatosis (Chapter 19). Some missense SNPs
Karelians have a sixfold increased rate of type 1 diabetes greatly alter function, whereas others appear to have no
compared to Russian Karelians.8 As a result, heritability for consequences. SNPs can also alter splice sites, disrupting
type 1 diabetes will be different when estimated in the the structure of the mRNA that is transcribed from the
combined Karelian populations than when estimated in DNA during gene expression. For example, the most
Finnish or Russian Karelians alone. The difference in dia- common cause of autosomal dominant isolated growth
betes rate is likely due to environmental factors, because hormone (GH) deficiency is single-base mutations that
both Karelian populations recently originated from a inactivate a splice donor site of intron 3 in the GH1 gene,
common ancestry and therefore likely have similar genetic causing skipping of exon 3 in GH1 (Chapter 24). SNPs can
risk factors for type 1 diabetes.9 also introduce stop codons, leading to premature termina-
tion of translation and a truncated protein product. These
Human DNA Sequence Variation: Molecular nonsense variants typically dramatically impair or elimi-
nate the function of the protein.
Forms and Biologic Effects Changing the protein sequence is not the only way that
Each human has two versions of his or her genome (one SNPs (and other types of genetic variations) can alter gene
from each parent); each version consists of a sequence of function. Most of the human genome does not code for
approximately 3 billion DNA bases. When comparing two proteins (see Table 4-2) and most genetic variation occurs
versions of a human genome, either within the same in this noncoding portion of the genome. For example,
person or between two different people, about 1/1000 of noncoding variants can alter the level, timing, or location
these bases vary (that is, 99.9% of them are the same) of gene expression, without changing the sequence of the
(Table 4-2). There are many possible ways in which DNA encoded protein. Noncoding variants often result in more
sequences can vary; several specific types of DNA sequence subtle biologic effects, and the mechanisms are still being
variants are frequently observed (Fig. 4-1). uncovered. For example, some SNPs subtly influence type
The most frequent form of variation, the single nucleo- 1 diabetes risk and lie in enhancers (noncoding DNA seg-
tide polymorphism (SNP), refers to the situation in which ments that activate gene transcription at a distance) that
a single base in the sequence of one individual is different appear to affect gene expression only in lymphoid cells.10
from the base seen at the same position in the sequence of Insertions and deletions (collectively called indels) refer
another individual. SNPs can exert a wide range of biologic respectively to the addition or removal of one or more
effects, depending on where the variant occurs and whether bases in the DNA sequence. Indels in protein-coding
it alters the function of the DNA sequence. Some SNPs sequences are called frameshift mutations, as long as the
occur within the portions of genes that are transcribed into number of bases inserted or deleted is not a multiple of
RNA and then translated into proteins (protein-coding three. Because the genetic code is composed of triplets
regions). Synonymous SNPs occur in the protein-coding (every three bases encode one amino acid), a frameshift
portion of DNA but both versions (alleles) of the SNP mutation alters how every subsequent base in the sequence
encode the same amino acid, and so this sort of variation is translated into a protein, resulting in profound molecu-
lar and clinical consequences. For example, classic salt-
wasting CAH is often caused by frameshift deletions in the
CYP21A2 gene that ablate its function (Chapter 23). Repeat
TABLE 4-2 polymorphisms (often referred to as copy number variants,
Characteristics of Human Genome Sequence Variation or CNVs, if the repeats are large) are a special case of indels
in which DNA sequences are repeated in tandem and the
Characteristic Frequency number of copies of the repeated sequence varies. For
Length of the human genome sequence (base 3 billion example, the AR gene (encoding the androgen receptor)
pairs) contains a repeat polymorphism in which a CAG codon,
Number of human genes (estimated) 20,000 encoding glutamine, is repeated 11 to 31 times (Chapter
Fraction of base pairs that differ between the 1.3% (1 in 80) 23). Structural variation can include both insertions and
genome sequence of a human and a deletions as well as rearrangement of large chunks of DNA
chimpanzee
Fraction of base pairs that vary between the 0.1% (1 in 1000)
sequence (translocations and other complex forms of
genome sequence of any two humans genomic variation). Structural variation causes familial
Fraction of coding region base pairs that vary in 0.02% (1 in 5000) hyperaldosteronism type 1; the adrenocorticotropic hor-
a manner that substantially alters the mone (ACTH, corticotropin)-responsive promoter of the
sequence of the encoded protein CYP11B1 gene is incorrectly located adjacent to the aldos
Number of sequence variants present in each 3 million terone synthase gene (CYP11B2), causing aldosterone to be
individual as heterozygous sites produced by ACTH stimulation (Chapter 16).
Number of amino acid–altering variants present 12,000
in each individual as heterozygous sites
Number of sequence variants in any given 10 million Factors Influencing the Biologic Impact of
human population with frequency of >1% Genetic Variants in a Particular Gene
Number of amino acid polymorphisms present 75,000
in the human genome with a population As discussed previously, the impact of a genetic variant on
frequency of >1% gene function will depend on the type of variant and its
Fraction of all human heterozygosity attributable 98% location with respect to the gene. For example, frameshift
to variants with a frequency of >1% deletions in the CYP21A2 gene completely eliminate
Adapted from Altshuler D. The inherited basis of common diseases. In
21-hydroxylase activity, whereas missense variants in
Goldman L, Schafer AI, eds. Goldman’s Cecil Medicine, 24th ed. CYP21A2 often retain partial 21-hydroxylase activity
Philadelphia, PA: WB Saunders; 2012. (Chapter 23). However, even a single, specific variant may
Low-frequency
variant
Indel Recombination Repeat
polymorphism hotspot polymorphism
Common SNP
G C T (G) C A G A T C C ATTCATTC
G C T A (G) C T A G A T C C ATTCATTC
G C T (G) C A C C C T G ATTC
A C T (G) A A G A T C C ATTCATTC
A C T (G) A A G A T C C ATTCATTC
A C T (G) A A C C C T G ATTC
G G A ( ) C T G A T C C ATTCATTC
G G A ( ) C T G A C T C C ATTCATTC
G G A ( ) C ( ) G ATTC
G G A ( ) C T C C C T G ATTC
( )
1 2 3 4 5 6 7 8 9 10 11 12
Strong correlation
No correlation
Figure 4-1 DNA sequence variation in the human genome. Common and rare genetic variation in 10 individuals, carrying 20 distinct copies of the human genome. The amount
of variation shown here is typical for a 5-kb stretch of genome and is centered on a strong recombination hotspot. The 12 common variations include 10 single nucleotide
polymorphisms (SNPs), an insertion-deletion polymorphism (indel), and a tetranucleotide repeat polymorphism. The six common polymorphisms on the left side are strongly
correlated. Although these six polymorphisms could theoretically occur in 26 possible patterns, only three patterns are observed (indicated by pink, orange, and green). These
patterns are called haplotypes. Similarly, the six common polymorphisms on the right side are strongly correlated and reside on only two haplotypes (indicated by blue and
purple). The haplotypes occur because there has not been much genetic recombination between the sites. By contrast, there is little correlation between the two groups of
polymorphisms because a hotspot of genetic recombination lies between them. In addition to the common polymorphisms, lower frequency polymorphisms occur in the human
genome. Five rare SNPs are shown, with the variant nucleotide marked in red and the reference nucleotide not shown. In addition, on the second to last chromosome, a larger
deletion variant is observed that removes several kilobases of DNA. Such larger deletion or duplication events (i.e., copy number variants [CNVs]) may be common and segregate
as other DNA variants. (Redrawn from Altshuler D, Daly MJ, Lander ES. Genetic mapping in human disease. Science. 2008;322(5903):881-888.)
have different effects in different individuals. The effect of manifest increased ferritin levels) but only 2% penetrance
any given genetic variant (genotype) on phenotype can be for the clinical phenotype of liver cirrhosis. Temporal
modified by variants in other genes (gene-gene interac- context is also an important consideration, as disease inci-
tions) or by environmental factors (gene-environment dence often increases with age. Carriers of mutations
interactions) or by random chance. It is usually not pos- causing MEN1 have nearly 100% penetrance for parathy-
sible to measure or quantify these factors in any one roid adenomas by age 40 but only 20% penetrance at
person, but their combined effect can be quantified on a age 20.
population level as penetrance, the proportion of individu- A common observation in members of a family carrying
als carrying a genetic variant who exhibit the phenotype. the same disease-causing genetic variant is that not all
The penetrance of a genetic variant is highly context- members of the family are equally affected. This range
dependent with respect to phenotypic definition. For of phenotypic expression resulting from a particular geno-
example, the hemochromatosis-associated C282Y allele in type is referred to as variable expressivity and, as with pen-
the HFE gene exhibits high penetrance for the biochemical etrance, arises from the range of impacts of specific variants
phenotype of high ferritin (>60% of homozygous carriers as well as modifying influences of genetic background
(gene-gene interactions), environment (gene-environment TABLE 4-3

interactions), and random chance. For example, the same Origins of DNA Sequence Variation in Human Populations:
mutation in the androgen receptor (AR, encoding an S703G Common Versus Rare Variants
substitution) resulted in a spectrum of clinical androgen
insensitivity such that some individuals were raised as The type of genetic variant (missense, frameshift, noncoding, etc.)
46,XY females and others as males; other mutations in AR provides clues to its possible consequences; in addition, the
have different ranges of phenotypic effects (Chapter 23). population frequency of a variant, whether it is common or rare, can
Mosaicism, whereby cells within a single individual also provide information about its likely impact on phenotype. The
relative balance between common and rare genetic variation is
have different genotypes, is another mechanism that leads strongly influenced by evolution and human demographic history.
to variable expressivity. Most mutations known to influ- Modern humans likely originated from a small population residing in
ence disease are germline mutations—inherited from the Africa that had been evolving over millions of years. Within the past
sperm or egg and present in every cell—but some diseases 50,000 years, members of this ancestral population migrated “out
can be caused by somatic mutations that occur after fertil- of Africa,” settled the globe, and only recently, over the past 5000
ization and are present in only some cells, leading to mosa- to 10,000 years, multiplied exponentially.12 As a consequence of
icism. In these cases, which tissues or organs carry the this demographic history, most of the 3 million genetic variants an
mutation will influence the clinical outcome. The most individual inherits from his or her parents are common (typically
familiar class of disease caused in large part by somatic >1% frequency in the population), can be traced back to the ancient
African population, and are shared in many unrelated individuals in
mutations is neoplasia, including endocrine tumor syn- the population. Individuals also inherit thousands of genetic variants
dromes such as Conn syndrome and Cushing disease. unique to themselves and their relatives. These rare genetic
Another classic example from endocrinology is the variants arose more recently from spontaneous mutation in the past
McCune-Albright syndrome, in which the same activating 10 millennia, after the migration of many humans out of Africa, and
mutation in GNAS1 exhibits variable expressivity because are typically observed infrequently (<0.1% of all chromosomes) in
of postzygotic mosaicism. The phenotype of patients the population.
with McCune-Albright syndrome depends on which tissues
and what fraction of cells carry the GNAS1 mutation. A
minority of affected individuals (24%) display the classic
triad of café au lait spots, polyostotic fibrous dysplasia,
and gonadotropin-releasing hormone (GnRH)-indepen-
dent precocious puberty; the majority express two or fewer thyroidism, Chapter 28). The same mutation, when mater-
features of the classic triad (Chapter 25). The mechanism nally inherited, manifests not only with AHO but also with
of variable expressivity likely maps to the zygotic stage in hypocalcemia secondary to parathyroid hormone resis-
which the mutation arose: a mutation earlier in embryo- tance (pseudohypoparathyroidism type 1b), because only
genesis is present in more tissue lineages. Because muta- the maternal copy of GNAS1 is expressed in renal proximal
tions in a mosaic individual are not present in every cell, tubules.
they can be hard to detect in DNA isolated from a blood Evolution influences the frequency of variants that
sample if the cell in which the mutation occurred does not affect human phenotypes (such as endocrine diseases)
give rise to blood leukocytes. The GNAS1 mutation respon- through the process of natural selection. Variants that
sible for the McCune-Albright syndrome is detected in only greatly increase the risk of a disease that is deleterious from
8% to 46% of blood samples from affected individuals but a reproductive standpoint are less likely to be passed on to
is found in 90% of affected tissue sampled irrespective of offspring and will be rare in the population (unless they
clinical presentation (Chapter 25). Conversely, blood cells have a compensatory benefit, such as malaria resistance in
can contain somatic variation that is absent in other tissues carriers of sickle cell disease). If a disease is at least mildly
or the germline.11 evolutionarily deleterious, then most common variants
It is important to remember that the base pair composi- associated with that disease will only modestly increase
tion of a DNA sequence is not the only molecular determi- disease risk. This is because those common variants, if they
nant of phenotypic expression (Table 4-3). DNA is subject had strongly increased disease risk, would have then been
to other forms of modification besides sequence variation subject to strong negative evolutionary selection and never
(termed epigenetic variation), such as cytosine methylation would have risen in frequency to become common in the
or packaging into nucleosomes with various biochemically first place. By contrast, it is more plausible for rare/recent
modified histones, each of which can alter gene expression variants to exert strong effects on phenotype and greatly
and function. Thus, the same molecular form of DNA increase disease risk.
sequence variation can vary in its cellular and phenotypic Finally, the number of genes that contribute to disease
effect through epigenetic modifications. Indeed, epigenetic in a single individual (mendelian or polygenic disease) will
modification is a normal part of development and is the be related to the strength of effect of any one variant on
reason why different cells have different properties even disease risk. By definition, variants that cause mendelian
though they share the identical DNA sequence. A striking disorders have strong effects, whereas variants contributing
example of the effect of epigenetics is imprinting, the to risk of polygenic diseases will typically have more
expression of a genetic variant in a parent-of-origin specific modest effects. Thus, most variants with strong effects on
manner. For paternally imprinted genes, the copy that is disease will be rare, especially for those diseases that are
inherited from the father is silenced, and only the mother’s clearly deleterious from an evolutionary standpoint (lethal
copy is expressed in the offspring. Imprinting can affect before reproductive age). By contrast, common polygenic
the impact of disease-causing mutations. Inactivating diseases and traits will have a much more substantial con-
mutations in SDHD cause familial paraganglioma type 1 tribution from common genetic variants, although these
(Chapter 16). SDHD is maternally imprinted, so the muta- considerations do not rule out an important role for rarer
tion does not cause disease when inherited from the variants in polygenic phenotypes. As we will see later in
mother but is highly penetrant when inherited from the this chapter, these patterns of genetic variation have
father. Imprinting can also be tissue-specific. A paternally important implications for identifying genetic variants
inherited inactivating mutation in GNAS1 causes Albright that underlie disease and also for interpreting the impact
hereditary osteodystrophy (AHO, pseudopseudohypopara- of genetic variation on disease.
Summary TABLE 4-4
To summarize this introductory section, we have briefly Performing and Interpreting Genetic Studies
described a number of basic principles of genetics. Herita- For any heritable disease, the success of genetic mapping efforts, the
bility describes the proportion of a disease/trait that can be strategy employed, and the clinical utility of any resulting genotype-
explained by genetic factors; the heritability of most endo- phenotype map depend on its genetic architecture: (1) the number
crine diseases ranges between 20% and 80% (see Table 4-1). of genetic variants/genes, (2) their frequency in the population, and
Genetic variants can take many forms ranging from single- (3) their respective contributions to risk (i.e., penetrance). On one
base changes (SNPs) to translocations of entire chromo- end of the spectrum lie mendelian diseases, such as multiple
somes (see Fig. 4-1). The biologic effect of these variants endocrine neoplasia type 1 (MEN1), characterized by (1) few
depends on the type of variant, where in the DNA they are variants often in a single gene, (2) extremely rare frequency in the
population (<1 : 1000), and (3) high penetrance (>50-fold risk). On the
located (e.g., within coding sequence, splice sites, enhanc-
other end of the spectrum lie the so-called common diseases, such
ers), how severely the variant affects function, and for as type 2 diabetes, characterized by (1) many variants in many
somatic mutations, the cells and tissues that carry the muta- genes (polygenic), (2) high frequency in the population (>1 : 20), and
tion. Biologic impact can also be modified by the presence (3) low penetrance (<1.5-fold risk) (Fig. 4-2).
of genetic variants in other genes (gene-gene interactions), Owing to their simple genetic architectures, mendelian endocrine
the individual organism’s environment (gene-environment disorders were ideally suited for genetic mapping using the
interactions), and random chance. The demographic his- techniques of familial linkage mapping developed in the 1980s.2
tory of modern human populations explains the presence Because they are rare and have strong effects on phenotype,
of common and rare genetic variants in the human genome mendelian variants were typically identified in families. As a result,
the genotype-phenotype correlations for these variants could not be
(see Table 4-2). Common variants are mostly ancient, and
generalized to the population at large. For example, penetrance of
typically have relatively modest clinical effects, whereas mendelian variants could be accurately estimated only if these
rare variants tend to have arisen more recently and can variants were ascertained in the general population, rather than in
exert larger clinical effects (Table 4-4). selected families with a specific genetic background. Large-scale
sequencing studies in the general population, which can identify
all variants, rare and common, are now enabling such estimates.
GENETICS OF ENDOCRINE DISEASES Such studies have found that, when ascertained in the general
population, the so-called mendelian variants are less penetrant that
As described earlier, heritable diseases and traits, including was estimated from family-based studies.13
By contrast, the variants for common, polygenic disorders have been
endocrine phenotypes, span a range of genetic architec-
identified through genetic association studies in the general
tures ranging from single-gene mendelian disorders to population. Genetic association studies do not require the
common, polygenic diseases and traits. Mendelian and identification of rare families segregating disease because they
polygenic disorders represent two ends of a spectrum (see simply compared the frequency of a given genetic variant in disease
Fig. 4-2) of genetic architectures. While we distinguish cases and controls. Thus, they can be applied to identify genetic
between these two extremes of genetic architecture, it is factors underlying diseases occurring in a population of unrelated
important to appreciate that many disorders lie between individuals (i.e., common diseases). Unlike clinical risk factors/
these two extremes: rare variants of moderate effect can biomarkers association studies, correlation in genetic association
affect the common form of the disease, and genetic and studies implies causation because genotype always precedes
phenotype. Through the 1980s, genetic association studies were
nongenetic modifiers can strongly influence the outcome
performed using single nucleotide polymorphisms (SNPs) at
of mendelian disorders. Furthermore, many polygenic candidate genes selected by educated guessing. Such studies
endocrine disorders also have rare mendelian forms (see yielded a number of common disease associations but were poorly
Table 4-1). reproducible and confounded by false-positive results arising from
The genes for a wide range of mendelian endocrine population stratification. The development of modern sequencing
diseases have been mapped, revealing great mechanistic and genotyping technologies along with the cataloging of over 10
insight. Although mendelian diseases have offered valuable million common variants (the International HapMap project14)
insights into pathophysiology, not all insights gained from enabled genome-wide association studies (GWAS), a systematic
mendelian forms of disease translate directly to the approach to simultaneously test all genes for associations that could
account for population-based confounding.4 GWAS have yielded a
common forms of disease. For example, mendelian obesity
large number of reproducible genetic associations for diverse
caused by recessive inactivating mutations in the leptin common/polygenic diseases/traits,15 yielding insight into disease
receptor could be well treated by exogenous leptin, but this biology and genetic architecture.
clinical insight did not apply to most obese individuals When interpreting a result from any genetic study, it is important to
who actually demonstrate elevated leptin levels and do not bear in mind that the actual variant (usually a SNP) tested in the
respond to exogenous therapy with leptin (Chapter 36). study marks a haplotype (a combination of genetic variants inherited
Obesity as a common trait is highly heritable (heritability together) that can span millions of bases. The causal variant, in the
40-80%),19 and genome-wide association studies (GWAS) sense that it is molecularly responsible for alteration in gene
analysis has begun to identify risk variants for the common function leading to cellular and disease phenotype, may lie
anywhere on this haplotype. As with the chromosomal linkage
form. Although some of the risk variants overlap with studies of the past; identifying the causal variants/genes on a
those causing mendelian syndromes (as is also true for haplotype necessitates a combination of further association analysis
other diseases), GWAS have pointed to additional genetic (fine-mapping16,17) and functional experimentation in model
contributions outside the mendelian genes. And, of course, systems.18
the variants that have strong effects on quite rare genetic
syndromes do not explain much, if any, of the risk of the
common forms of disease. Thus, genetics of both mende-
lian forms and common polygenic forms will have impor- of the impact of genetic variation on disease, and implica-
tant, often complementary impacts on our understanding tions for clinical care and insights into new biology. We
of disease and on patient care. discuss several classes of mendelian diseases and highlight
The sections that follow discuss representative examples three polygenic endocrine diseases/traits: (1) type 2 diabe-
of mendelian and polygenic endocrine disorders that illus- tes, (2) stature, and (3) serum lipids. In each section, we
trate important concepts in gene discovery, understanding discuss what is known about the underlying genetic
Penetrance
High
Highly unusual
Mendelian for common
disease diseases
Intermediate Low-frequency
variants with
intermediate
penetrance
Modest Most variants
Hard to identify identified by
genetically GWA studies
Low
Allele
frequency
Very rare 0.001 Rare 0.01 Uncommon 0.1 Common
Figure 4-2 Genetic architecture of common and mendelian diseases. At one end of the spectrum are mendelian diseases cause by few variants in few genes each with a
large individual effect on disease risk. On the other end of the spectrum are common diseases and traits that are caused by the combined effects of many variants, observed
frequently in the population, each with a modest individual effect. (Redrawn from McCarthy MI, Abecasis GR, Cardon LR, et al. Genome-wide association studies for complex
traits: consensus, uncertainty and challenges. Nat Rev Genet. 2008;9(5):356-369.)
contributors, the impact of genetics on our understanding mutation in MEN1. Thus, much of the genetic architecture
of disease biology, and the translation into clinical care in of mendelian diseases remains uncharted territory for
the short and long term. genetic mapping. Modern sequencing technologies have
facilitated a renaissance in mendelian disease gene mapping
Mendelian Endocrine Diseases and will help improve our understanding of the genetic
basis of mendelian disorders. For example, by exome
Genetic Architecture sequencing two individuals in a kindred with familial
Mendelian diseases represent one extreme of a spectrum of combined hypolipidemia (Chapter 37), investigators iden-
possible genetic architectures (see Fig. 4-2). The alleles tified two nonsense mutations in ANGPTL3 that segregated
causing mendelian diseases are found in a small number of with low serum lipoproteins when genotyped in other
genes, are typically rare (<1 : 1000), are highly penetrant, and family members.20 These mutations and the ANGPTL3 gene
follow simple patterns of dominant and recessive inheri- were contained in the region identified by traditional
tance. They are considered monogenic in that a mutation linkage mapping20 and could be quickly identified because
in a single gene causes disease in an individual or family. the sequence of all exons in that region had been
But as different families segregating the same mendelian determined.
disease are identified and the causal genetic variants
mapped, genetic heterogeneity is often observed: different
alleles in different genes can cause the same disease. Some
Disease Biology
mendelian disorders (e.g., MEN2) demonstrate recurrent Every endocrine organ ranging from the pituitary to
mutations in the same gene, but of different molecular adrenal is affected by well-described and less-described
forms and locations, a phenomenon termed allelic hetero- mendelian disorders. Mechanistic insight into disease
geneity. For other disorders (e.g., familial paraganglioma) biology has been gained from discovering the identities
multiple genes across different chromosomes are impli- of the genes that lead to disease. When mutations in a
cated, each causing the same/similar disease in different number of different genes can all cause a disease (locus
individuals. This phenomenon, variants in different genes heterogeneity), additional mechanistic insight into molec-
causing the same disease, is termed locus heterogeneity. It ular pathophysiology becomes possible. This makes intui-
is important to bear in mind that locus heterogeneity tive sense in the context of a molecular understanding of
is intrinsically tied to the precision of disease definition. genes as encoding proteins that act in concert to accom-
For example, CAH can be caused by defects in multiple plish cellular functions. For example, Noonan syndrome
genes encoding steroid biosynthetic enzymes (CYP21A2, (characterized endocrinologically by variable short stature,
CYP11B1, CYP17A1, HSD3B2, POR, StAR; Chapter 15). delayed puberty, and cryptorchidism in the setting of dys-
However, if the CAH phenotype is refined to include bio- morphic features and variable cardiac defects; Chapter 23)
chemical measurements (mineralocorticoid, sex hormone, is typically caused by activating the RAS-MAPK (mitogen-
and electrolyte levels), individual subtypes emerge, each of activated protein kinase) signaling pathway. Dominant
which possesses a simpler genetic architecture (i.e., de- gain-of-function mutations in multiple pathway members
creased locus heterogeneity). (PTPN11, SOS1, KRAS, RAF1, BRAF, NRAS) have all been
When contrasted with common polygenic diseases, shown to cause Noonan syndrome. For other disorders, a
mendelian disorders exhibit relatively less locus heteroge- more complex picture emerges in which multiple molecu-
neity. In other words, an appreciable fraction of mendelian lar pathways are implicated. For example, Kallmann syn-
disease cases can be largely explained by mutations in one drome (Chapter 25), which arises from failure of migration
or a few genes. For example, recurrent mutations in a single of GnRH neurons during fetal development, demonstrates
gene (the eponymous MEN1) account for 70% of families X-linked (KAL1), autosomal dominant (FGFR1), and auto-
segregating the MEN1 clinical syndrome. Even in this somal recessive (PROK2) inheritance. The gene product
classic mendelian case, however, the genetic architecture of KAL1, a secreted protein called anosmin, is thought to
remains incompletely defined, as 30% of cases have no interact with the fibroblast growth factor (FGF) receptor,
whereas the gene product of PROK2, the secreted protein G protein (Gs). This mutation destabilizes the protein,
prokineticin 2, interacts with a different receptor. Both causing loss of function in most body tissues and thus
signaling pathways are required for GnRH neuronal hormone resistance. But these mutation carriers also exhib-
migration. ited paradoxical testotoxicosis consistent with gain of func-
At the level of a single gene/locus, genotype-phenotype tion in Gs. The paradox of the same mutation causing both
correlations mapping allelic heterogeneity to phenotypic loss and gain of function was resolved with a series of
heterogeneity can provide detailed insight into how altera- experiments showing that the Ala366Ser mutation caused
tions in gene function affect disease severity. CAH caused a temperature-sensitive effect physiologically relevant to
by CYP21A2 deficiency is a classic case. A number of genetic the normal temperature for Leydig cells. In most body
variants including frameshifting deletions, splice site alter- tissues maintained at 37° C, the mutant Gs protein was
ations, and missense mutations have been identified destabilized, whereas in Leydig cells (maintained at a 3-5° C
(Chapter 23) in CYP21A2. This spectrum of alleles has been lower temperature), the mutant Gs demonstrated increased
mapped to a biochemical spectrum of 21-hydroxylase activity. Elucidating this mechanism required an apprecia-
enzyme activity, which in turn maps to a spectrum of clini- tion of the discordant phenotype (testotoxicosis) and bio-
cal features along the axes of mineralocorticoid sufficiency, chemical characterization in the relevant physiologic
androgen excess, and ACTH elevation (Chapter 23). In this system (Leydig cells at a lower temperature).
disorder, it is possible to make predictions about clinical
phenotype (categorized as salt-wasting, simple virilizing,
and nonclassic) based on genotype. Notably, the positive
Clinical Translation
predictive value (PPV, the strength of the genotype- Target discovery, risk prediction, and the tailoring of phar-
phenotype correlation) is strongest for variants that severely macotherapy based on genotype are potential clinical
affect CYP21A2 gene function and are predicted to cause applications of genotype-phenotype correlation. The exis-
severe disease (salt-wasting, PPV ~100%). Predictive power tence of loss-of-function and gain-of-function variants in
is weaker for genetic variants that are expected to have an allelic series and their concordance with opposing phe-
more moderate effects on gene function and therefore notypes can provide a rationale for therapeutically modu-
result in milder disease (nonclassic, PPV ~60%). Some of lating gene function.23 For example, inactivating mutations
this complexity is due to the potentially compensatory in the KISS1R receptor cause hypogonadotropic hypogo-
21-hydroxylase enzyme activity of CYP2C19 and CYP3A4, nadism, whereas an Arg386Pro missense variant in KISS1R
a form of gene-gene interaction. is associated with central precocious puberty. Kisspeptin,
Genotype-phenotype correlations must be established the agonist ligand of the KISS1R receptor, has shown
empirically and are not possible in many cases. Even when promise as a fertility treatment.24
mutations of varying molecular severity are identified, they Genotype-phenotype correlation can be used to predict
may not predictably affect phenotype. For example, many risk of disease in asymptomatic carriers. Prior to the iden-
individuals with genetic variants in the SRDA2 gene (encod- tification and cloning of the RET proto-oncogene, MEN2
ing 5α-reductase) of differing molecular severity and loca- kindreds were monitored for evidence of medullary thyroid
tion have been identified (Chapter 23), but no correlation cancer (MTC) by calcitonin stimulation tests. Once muta-
between the genotype and the clinical degree of virilization tions in RET were established as causing MEN2A/B and
is apparent. familial MTC, it became apparent that specific mutations
The genetic causes of many mendelian disorders remain could be mapped to the different syndromes. The RET gene
unknown, but advances in sequencing technology have product encodes a cell surface receptor tyrosine kinase.
accelerated the pace of discovery. The identification of Mutations in the extracellular domain predispose to
gain-of-function mutations in the KCNJ5 and CACNA1D MEN2A (characterized by MTC, pheochromocytomas, and
genes as causes of hyperaldosteronism illustrates the po- hyperparathyroidism), whereas mutations in the intracel-
tential of combining modern genome scale-sequencing lular tyrosine kinase domain predispose to MEN2B (char-
with classical genetic study design.21,22 By sequencing acterized by MTC, pheochromocytomas, and mucosal
exomes in a series of aldosterone-producing adenomas neuromas) (Chapter 39). The clinical aggressiveness of
from individuals with primary hyperaldosteronism (Conn MTC, the sine qua non of all three syndromes, is greatest
syndrome) investigators identified missense mutations in in MEN2B, then less in MEN2A, with familial MTC dem-
the KCNJ5 gene in about one third of tumors from unre- onstrating the least propensity to grow and metastasize. A
lated individuals. They also identified a separate KCNJ5 well-defined genotype-phenotype correlation between spe-
missense mutation in a mendelian family with hyperten- cific RET mutations and clinical aggressiveness of MTC
sion, primary hyperaldosteronism, and massive adrenal now dictates the timing of lifesaving prophylactic thy
hyperplasia. Follow-up biochemical and electrophysiologic roidectomy in carriers of RET mutations.25 The key to
studies showed that this series of somatic and inherited establishing clinically robust risk prediction based on
missense mutations eliminated ion selectivity in the KCNJ5 genotype-phenotype correlations is a well-differentiated
gene product, a potassium channel. This caused mem- allelic series derived from multiple individuals/families.
brane depolarization of adrenal cortical cells in the zona The consensus genotype-phenotype correlation for pro-
glomerulosa, stimulating aldosterone release and cell phylactic thyroidectomy in RET mutation carriers was
proliferation. derived from analysis of over 200 individuals from over
Whereas modern genome sequencing technologies 100 families (Chapter 39).
facilitate the identification of causal mutations, translation A genetic diagnosis in a number of mendelian disorders
of mutations into a disease mechanism requires relevant can also directly inform pharmacotherapies. A classic ex-
experimental model systems and close correlation with ample includes obesity caused by leptin deficiency (Chap-
human phenotype. Pseudohypoparathyroidism 1a (PHP1a, ter 36), which can be treated by exogenous leptin injections.
Chapter 28) with gonadotropin-independent sexual pre- Other examples include HNF1A MODY (maturity-onset
cocity provides a classic example of the iterative rela diabetes of the young) and neonatal diabetes (discussed in
tionship between mutations, human phenotypes, and detail later) caused by genes whose properties predict excel-
laboratory experiments. Some individuals with PHP1a lent response to sulfonylureas. In the case of congenital
harbor the Ala366Ser missense mutation in a stimulatory hyperinsulinism, autosomal recessive mutations in ABCC8
or KCNJ11 correlate with diffuse disease on spectroscopic specific effects have also been identified. A T2D GWAS
imaging and lack of responsiveness to medical therapy performed in individuals of Latino and Mexican ancestry
(diazoxide); such individuals require near-total pancreatec- identified a common SNP at a locus containing the genes
tomy for control of hypoglycemia.26 SLC16A11/13 that confers a 1.25-fold increased risk of dia-
betes.38 The same locus was identified by another T2D
Type 2 Diabetes GWAS performed in Japanese individuals.39 Because the
associated SNPs were rare in Europeans, the locus had not
Genetic Architecture been detected in GWAS in European-ancestry populations.
Type 2 diabetes (T2D) is a multifactorial, polygenic disorder Similarly, a common variant in TBC1D4 in individuals
for which manifestation depends on multiple interacting from Greenland (present in 17% of Greenland’s popula-
genetic and environmental risk factors. Heritability esti- tion) strongly increases risk of T2D (10-fold increased
mates show strong evidence of familial clustering, ranging risk).40 This variant, which causes a premature truncation
from 40%6 to 80%.7 Approximately 5% of diabetes cases and associates with elevated muscle insulin resistance, is
that may be classified as type 2 (nonautoimmune) arise extremely rare in continental Europe and likely became
from a single-gene disorder, follow mendelian patterns of common in Greenland because it was present in founding
inheritance, and cluster into clinically defined syndromes. ancestors of Greenland’s current population.
These mendelian diabetes syndromes include neonatal dia- In summary, genetic mapping studies over the past 3
betes (Chapter 31), MODY (Chapter 31), and congenital decades have revealed a genetic architecture for T2D with
lipodystrophies (Chapter 31).27 To date, familial linkage widespread locus and allelic heterogeneity. With regard to
studies have successfully implicated approximately 30 effect size and allele frequency, T2D genetic architecture so
genes as monogenic causes of diabetes.28 far consists of some very rare variants of large effect, some
The genetic factors underlying the majority of T2D cases common variants with small to moderate effects (1.2 to
(95% of cases) fit a polygenic model; genetic variants in 1.5-fold increased risk), and a larger number of common
multiple genes independently contribute to disease risk, variants with even more modest effects on disease risk,
each with a modest effect. The partial elucidation of these with rare and common genetic variants spread out across
genetic risk factors required the advent of genetic associa- multiple loci in the genome. This genetic architecture has
tion studies/GWAS and the assembly of cohorts of thou- proved to be typical for other common diseases30 (see Fig.
sands of cases and control subjects.29 As of 2014 about 70 4-2) and reflects both the underlying genetic architecture
loci have been identified from aggregated analysis on about of the disease and the ability of large GWAS to more readily
150,000 case controls. Taken together, these loci account detect common variants of modest effect.
for about 6% of heritability for T2D.29 Of these loci, an SNP
at TCF7L2 (with the risk-increasing allele present at a fre-
quency of ~30%) has the largest overall effect on risk,
Disease Biology
conferring a 1.4-fold increase in risk per allele.29 The past three decades of genetic discoveries in T2D have
By contrast, type 1 diabetes shows a somewhat different nucleated a molecular understanding of disease mecha-
genetic architecture, with common loci of large effect (a nisms, highlighted differences between glycemia and T2D,
variant at the HLA locus found in 61% of the population and implicated previously unknown physiology in disease
confers a fivefold increase in risk,30 and a common variant pathogenesis.
at the insulin gene confers a threefold increase in risk). Supporting the current physiologic conception of T2D
This finding was consistent with prior studies from the as a disorder of decreased insulin production as well as
1980s estimating that 50% of the heritability of type 1 decreased insulin sensitivity, genetic mapping has pointed
diabetes was explained by common haplotypes at the HLA to a molecular basis for both axes. Prediabetic individu-
locus.31,32 als harboring T2D-associated variants in beta-cell genes
Notably, the genes implicated in monogenic causes of (SLC30A8, HNF1A) and cell survival genes (CDKAL1) mani-
diabetes also contribute to polygenic forms, but through fest with decreased insulin secretion (homeostatic model
distinct genetic variants. Genes associated with mendelian assessment [HOMA]-B; Chapter 31).34 On the other hand,
diabetes syndromes, such as KCNJ11 (neonatal diabetes, prediabetic individuals harboring T2D-associated variants
Chapter 31), HNF1A (MODY2, Chapter 31), and PPARG in adipocyte genes (PPARG, KLF14) tend toward increases
(familial partial lipodystrophy 3),33 were found to harbor in insulin resistance (HOMA-IR; Chapter 31).34 About 30%
common variants that conferred risk for common T2D.27,34 of T2D-associated SNPs point to insulin secretion/beta-cell
Conversely, genes first found associated with diabetes function, and 15% point to insulin resistance.28 Interest-
through GWAS have subsequently been identified as having ingly, the SNPs associating with insulin secretion in predia-
rare, highly penetrant alleles. For example, noncoding betic individuals predict incident T2D but the SNPs
common variants pointed to the MTNR1B gene (encoding associating with insulin resistance do not.41 These findings
the melatonin receptor) as a T2D-associated locus (1.15- from genetic epidemiology are consistent with beta-cell
fold risk).35 Subsequently, large-scale resequencing studies failure being a final common pathway for manifestation of
identified multiple, rare coding variants of the same gene hyperglycemia and diagnosis of T2D. Importantly, over
(present in <1 : 1000 individuals) that conferred a greater half of associated SNPs and the genes they point to cannot
than fivefold increased risk of T2D.36 be connected with either insulin secretion or sensitivity.
Mapping studies performed across various populations Their pathogenic mechanisms remain to be elucidated by
reveal both similarities and differences in the genetic risk physiologic and functional investigation.
factors contributing to diabetes among different ancestral Even without a full understanding of their molecular/
groups. T2D GWAS performed in multiple populations/ cellular mechanism of causation, the large number of T2D-
ancestries (European, South Asian, East Asian, Latino, Afri- associated loci (~70 as of 2014) have been deployed to
can American)37 reveal that many common variants are refine disease classification. By examining quantitative gly-
shared across populations with equivalent effects on dis- cemic traits (insulin production, sensitivity, processing,
ease risk, regardless of ancestry. This pattern is consistent and fasting glucose) in nondiabetic individuals genotyped
with the origin of most common variants in an ancestral for 37 T2D-associated common genetic variants, investiga-
African population (see Table 4-3), but remarkable ancestry- tors were able cluster genes with glycemic traits to define
unique diabetes subtypes.42 For example, individuals har- from T2D and have no other deleterious phenotype offers
boring variants in MNTR1B and GCK manifested a combi- a tantalizing therapeutic hypothesis that a chemical or
nation of fasting hyperglycemia and decreased insulin antibody inhibitor of SLC30A8 could treat diabetes and
secretion, whereas those harboring variants in SLC30A8, minimize side effects.
CDKN2A/B, TCF7L2, and other genes manifested primarily In the arena of risk prediction, genetics has not yet
with decreased insulin secretion. Notably, many genes did made a major impact on T2D because existing clinical risk
not cluster with predefined glycemic traits, again suggest- factors already predict disease well. Common genetic vari-
ing that the current physiologic description of T2D remains ants, by virtue of their relatively high frequency, can
incomplete. explain a large part of heritability for a trait/disease but
Genetic mapping has also corroborated the epidemio- have small effects on the individual. For example, the
logically identified intersection between T2D and obesity. common P12A variant in PPARG is associated with 1.25-
A SNP in the second intron of the FTO gene was identified fold risk of T2D.55 Based on the high population frequency
in parallel in GWAS for T2D43 as well as obesity.44-46 The of the risk variant (85% P), if one were to theoretically
association signal for T2D entirely disappeared with correc- substitute every P to A in the population, 20% of diabetes
tion for body mass index (BMI), indicating that this SNP would be eliminated. Despite this incredible population-
increased T2D risk by increasing BMI. Interestingly, this attributable risk, any given individual carrying the P
locus illustrates some of the difficulties in proceeding from variant only has a 25% increased risk of diabetes when
GWAS signal to function. Although this SNP was initially compared to someone carrying the A variant. Given the
thought to exert its effect on BMI by affecting FTO gene high population frequency of common variants, many
function, detailed mechanistic studies have revealed that disease susceptibility variants are found in the same indi-
it may function by altering expression levels of IRX3, a vidual, each of which confers modest increases in risk.
gene over a million bases away.47 Although initial studies The progressive accessibility of genome sequencing makes
in mice showed that increasing Fto gene dosage increased it feasible to ascertain all known risk-conferring variants
food intake leading to increased fat mass,48 no connection in an individual at once and combine these for potentially
has been found between the disease-associated SNPs and more clinically meaningful risk prediction. Investigators
Fto expression level or function.49 have attempted to combine common variants into a
Whereas T2D is diagnosed on the basis of hyperglyce- genetic risk score with modest success. For T2D, a risk
mia, genetic mapping has revealed that the genes that score combining 18 common variants (including PPARG
determine fasting glucose are partly distinct from those P12A) demonstrated 2.6-fold elevated risk in the high-risk
that are associated with T2D. Comparison of GWAS per- versus low-risk group.56 By contrast, simply reporting a
formed for blood glucose levels in nondiabetics versus in family history of diabetes increases risk by threefold to
T2D case-control studies reveals that glycemia and T2D sevenfold.57
have distinct genetic associations.50 Some genes harbor Tailoring pharmacotherapy based on genotype has been
variants that increase blood glucose levels and T2D risk, successful in monogenic diabetes; the genome-sequencing
but others alter blood glucose levels but do not confer T2D era promises to bring this benefit to a wider group of indi-
risk. Thus, the two phenotypes have both common and viduals. The classic example of genotype guiding pharma-
distinct biology. Additionally, it is important to bear in cotherapy is that of individuals with MODY2 caused by
mind that the genetic basis for surrogate measures of gly- autosomal dominant mutations in the GCK gene. Such
cemia do not always point to genes specifically altering individuals meet diagnostic criteria for diabetes but are able
glycemic physiology. A particularly salient example is the to regulate glycemia at a higher set-point, thus avoiding
association of hexokinase 1 (HK1) with hemoglobin A1c all secondary complications (Chapter 31). Thus, a genetic
levels but not with fasting or dynamic glycemia.51 It is diagnosis of GCK-related diabetes can allow such individu-
thought that the genetic variant in HK1 alters hemoglobin als to avoid pharmacotherapy. Individuals with permanent
A1c levels as a result of alteration of red blood cell life span neonatal diabetes cause by ABCC8 or KCNJ11 mutations
and anemia.51 can be safely treated with high-dose sulfonylureas in place
of insulin.58,59 It is likely that individuals with functional
mutations in ABCC8 or KCNJ11 will be found who do
Clinical Translation not present with the classic neonatal syndrome and are
The principle of genetics pointing to important therapeutic classified as common T2D but who may still respond pref-
targets in T2D is well validated. Both rare and common erentially to sulfonylureas. Proving this will require iden-
genetic variants link PPARG, the drug target of thiazoli- tifying such individuals and performing prospective clinical
dines to syndromic and common T2D.33 Similarly, rare trials. Prospective genotype-based intervention trials have
variants in the sulfonylurea receptor (encoded by ABCC8) been performed in individuals with MODY3 caused by
cause neonatal diabetes.52 Although these oral hypoglyce- HNF1A mutations and have shown the superiority of sul-
mics were identified in a pregenetic era, they point to an fonylureas over metformin.60 Interestingly, exome sequenc-
optimistic future of genetically guided drug discovery, one ing of Latin American T2D case-control groups revealed the
that will require detailed mechanistic understanding of the HNF1A E508K variant, previously annotated as MODY3, as
genes mapped by studies to date. A particularly attractive conferring a fivefold risk of T2D in approximately 1 : 1000
target nominated by genetics is SLC30A8, a gene that individuals. These data demonstrate that the E508K is not
encodes a zinc transporter expressed almost exclusively in fully penetrant; nevertheless, the individuals who carry it
the endocrine pancreas. The common R325W missense may still benefit preferentially from sulfonylurea treat-
variant in the protein encoded by the SLC30A8 gene ment, as do their counterparts with clinically defined
(present in ~1 : 3 individuals in most continental popula- MODY3.
tions) was found to associate with protection from T2D
(1.18-fold decreased risk).53 Rare, protein-truncating vari- Short Stature
ants in SLC30A8 (present in ~2 : 1000 individuals) have also
been associated with protection from T2D with a larger Genetic Architecture
effect size (2.6-fold decreased risk).54 The finding of human Adult height is a polygenic quantitative trait with a heri-
heterozygous knockouts for SLC30A8 who are protected tability of 80%.5 Many mendelian syndromes manifest
(Chapter 24) large phenotypic variations in stature, and cartilage/bone development. These cellular processes were
more than 150 genes are associated with monogenic short first implicated by monogenic short stature syndromes
stature or overgrowth. The effect sizes for these rare, highly caused by mutations in genes such as FGFR3 (achondropla-
penetrant alleles are typically large, up to 300 mm (3 stan- sia), SHOX (Langer mesomelic dysplasia), and FBN1 (Marfan
dard deviations [SD]). The genetic factors underlying most syndrome).
of human height variation are polygenic. GWAS aggregat- GWAS for height as a polygenic trait, combined with
ing more than 250,000 European samples have identified systematic clustering of genes in associated loci, have
over 400 independent loci associated with height61 through refined and expanded insights gained from mendelian
common variant association. Common variants account genetics. Common variants associated with multiple mem-
for about 60% of heritability.61,62 The effect sizes for these bers of the GH/IGF-1 axis demonstrate the importance of
common alleles are in the range of 1 to 15 mm (0.01 to this pathway in height regulation within the normal
0.15 SD). Taken together, these genetic mapping studies range.67 The association of TGF-β itself and of its binding
show that the genetics of height consists of the additive proteins, LTBP1–3, complements the finding of FBN1
effects of common alleles with modest individual effects mutations in Marfan syndrome in highlighting TGF-β sig-
except in very short individuals (> 2 SD below the mean).63 naling.68 Similarly, the FGF signaling pathway is also high-
In extremely short individuals, it is likely that rare alleles lighted, as common variants near FGF4 are associated with
of large effect play a larger role.63 height, complementing the finding of monogenic muta-
Many genes harboring rare alleles causing monogenic tions in FGFR3 causing achondroplasia. Clustering of
alteration in stature also harbor common alleles that con- GWAS-associated genes has also implicated hedgehog sig-
tribute to polygenic stature variations. Up to 30% of the naling (GLI2, LAMA5), Wnt signaling (CTNNB1, FBXW11,
genes in height-associated loci identified by GWAS also WNT4, WNT5A), and mammalian target of rapamycin
contain rare, monogenic alleles.61 A classic example is GH1 (mTOR) signaling (SMAD3, MTOR) cascades in genetic con-
(isolated GH deficiency 1a, Chapter 24). This substantial trol of height.61 Many associated loci do not cluster with
overlap suggests that the genes in associated loci that have the pathways above or other known tissue/cellular pro-
yet to demonstrate rare, large-effect alleles may be prime cesses such as bone/cartilage formation, implicating previ-
candidates for resequencing studies to discover new mono- ously unknown biology in height regulation. A notable
genic causes of alterations in stature. example is the microRNA cluster MIR17HG, which was also
Short stature, but not tall stature, has been associated identified as a syndromic cause of short stature.69 By cross-
with deletions in the genome. By systematically comparing referencing GWAS-associated genes with gene-expression
CNVs across the genomes of over 4000 individuals with microarrays from thousands of human tissue samples,
developmental delay and congenital abnormalities with investigators have found height-associated genes are most
7000 population-based control subjects, investigators ob- highly expressed in cartilage and the growth plate and to
served that individuals with short stature harbored an a lesser extent in bone and endocrine organs.61
excess number of low-frequency (found in <5% of the
population) deletions.64 Individuals in the cohort clinically
diagnosed with short stature demonstrated an average loss
Clinical Translation
of 900,000 bp from their genomes.64 Gene mapping and functional characterization in mono-
In summary, the genetic architecture of height is con- genic stature disorders have motivated therapeutic advances
sistent with classical theory and animal models65 for a for diseases of overgrowth and short stature. A classic
polygenic, quantitative trait: thousands of genetic variants example is that of Marfan syndrome, in which excess
in hundreds of genes contribute to the genetic variability TGF-β signaling resulting from the effects of disruptive
in height. Most of the genetic variation in height (97%) mutations in FBN1 has led to the development of TGF-β
occurs from the additive effects of common variants, each antagonist therapies.70 Human trials of the angiotensin
contributing a modest effect. Rare variants of large effect receptor blocker losartan, which exhibit TGF-β antagonist
often causing mendelian syndromes cluster at the short properties in in vivo preclinical models, were begun in
extreme of height variation. cohorts of Marfan syndrome patients.71 Results three years
after treatment showed a decrease in aortic root diameter
but not more so than conventional therapy with beta
Disease Biology blockers.72 The identification of the activating G380N
Genetic mapping of genes influencing both normal varia- mutation in FGFR3 that causes 95% of achondroplasia has
tion and the extremes of stature have revealed diverse motivated the development of inhibitors of the tyrosine
molecular pathways operating in multiple tissue types kinase activity of FGFR3 through small molecules73 as well
that work through both endocrine and cell-autonomous as through analogues of C-type natriuretic peptide.74
mechanisms. Diagnosis of short stature is an important application of
Genetic mapping in individuals with low GH levels and genetic testing, particularly in pediatric populations. For
short stature have delineated multiple components of the example, the genotype-phenotype correlation of SHOX
GH/IGF-1 (insulin-like growth factor 1) axis as a key endo- deficiency shows a clinical spectrum ranging from homo-
crine pathway in human height regulation (Chapter 24). zygous loss-of-function mutations causing the severe syn-
These components include hormones, their receptors, drome Langer mesomelic dysplasia to heterozygous defects
binding proteins, and intracellular signaling proteins, such found in patients with milder syndromic disease (Léri-Weill
as GH1, GHR, GHRHR, STAT5B, IGFALS, IGF-1, and IGF-1R. dyschondrosteosis, Ullrich-Turner syndrome) or idiopathic
Genetics has also pointed to the importance of the pitu- short stature.75 Genetic diagnosis qualifies these patients
itary gland as a key endocrine organ regulating height, as for GH therapy. Genetic diagnosis also enables directed
mutations in genes that that encode transcriptional regula- screening for comorbid conditions. For example, males
tors of pituitary development such as HESX1, PITX1, PITX2, with the likely underdiagnosed 3-M syndrome (caused by
PROP1, POU1F1, and LHX3 also lead to short stature when mutations in CUL7, OBSL1, or CCDC8), clinically defined
mutated in humans.66 At the paracrine and cell-autonomous by severe postnatal growth retardation, characteristic
levels, genetic mapping has highlighted the importance of facies, and radiographic findings of skeletal abnormalities,
cellular proliferation, extracellular matrix deposition, and are at high risk of primary hypogonadism and need to
be monitored.75 Similarly, individuals with short stature binations of lipoproteins.81 These overlaps are consistent
secondary to Noonan syndrome are also often underdiag- with observations in mendelian disorders and corroborate
nosed and are at higher risk for cardiac disease.76 the shared metabolism and lipoprotein constituents of
Tailoring pharmacotherapy based on genotype is a LDL, HDL, and TG.
useful adjunct to biochemical testing in stature disorders. Many of the genes identified as monogenic causes of
Physiologic stimulation and serum biochemistry are the dyslipidemia also alter lipid levels in the general popula-
gold standards for assessing GH sensitivity and resistance tion through both common and rare variants. LDLR,
guiding its pharmacologic use. But genetic information encoding the LDL receptor, provides a case in point of how
plays an important role in solidifying diagnoses suggested the allelic spectrum of genetic variation ranges from rare,
by biochemical testing and suggesting alternative pharma- mendelian alleles to common, small-effect alleles with
cologic therapy. For example, children with defects in the effect sizes inversely proportional to variant frequency. Dis-
GH receptor or post-GH receptor signaling (STAT5B) will ruptive mutations (those that frameshift or prematurely
be candidates for treatment with biosynthetic IGF-1.77 terminate the protein) in LDLR like the ones found in
Children with defects in IGFALS, a serum protein that familial hypercholesterolemia are found in 2 : 1000 to
stabilizes IGFs, respond poorly to both medications.78 GH 7 : 1000 individuals and increase LDL levels by 150 to
therapy is used to treat short stature arising from certain 200 mg/dL.83 Missense variants that decrease function of
genetic defects outside the GH/IGF-1 axis but is not indi- LDLR in cell models are found in about 1 : 100 individuals
cated in many others. Given the variable expressivity in and increase LDL levels by about 100 mg/dL.84 A common,
many syndromic stature disorders, clinically distinguishing intronic SNP in LDLR found in 1 : 10 individuals decreases
among syndromes presenting with short stature can be LDL levels by 7 mg/dL.82
challenging and imprecise. Genetic diagnosis can resolve In summary, the genetic architecture of serum lipids
ambiguity, particularly in children before all the features consists of a full spectrum of rare and common alleles in
of a syndrome are present. For example, GH is contraindi- hundreds of genes throughout the genome acting in a
cated in chromosomal breakage disorders. Bloom syn- polygenic fashion. Different variants at the same locus,
drome (caused by loss-of-function mutations in BLM, based on varying impact on gene function, can alter lipid
which encodes a DNA helicase) is one such disorder pre- levels in a broad range. Variant frequency in the popula-
senting with short stature. In the absence of genetic diag- tion is inversely correlated with the magnitude of effect on
nosis with only clinical and biochemical testing, there are serum lipid levels. Many loci have an impact on multiple
case examples of children being treated for years with GH lipoprotein levels simultaneously.
until their clinical presentation evolved and Bloom syn-
drome was diagnosed.79
Disease Biology
Lipids and Coronary Artery Disease Genetic mapping for lipid traits has a rich history of
synergizing with biochemical and physiologic investiga-
Genetic Architecture tion to reveal the molecular mechanisms of lipoprotein
Serum lipid levels are a complex polygenic trait signifi- metabolism and its relationship to cardiovascular disease
cantly influenced by environmental factors such as diet. in humans.
Heritability estimates suggest a large role for genetic fac- Mendelian hyperlipidemia syndromes were the first to
tors: approximately 40% to 60% for high-density lipopro- yield pathophysiologic insight, starting with Brown and
tein cholesterol (HDL-C), about 40% to 50% for low-density Goldstein’s classic studies showing that LDL failed to sup-
lipoprotein cholesterol (LDL-C), and about 35% to 48% for press HMG-CoA (3-hydroxy-3-methyl-glutaryl coenzyme
triglycerides (TG).80 Monogenic disorders causing extremes A) reductase activity in fibroblasts from subjects with famil-
of dyslipidemia have been associated with around 20 ial hypercholesterolemia.85 Subsequent studies of families
genes. Mendelian dyslipidemia syndromes can present with severe hypercholesterolemia (LDLR, APOB, ABCG5,
with single or combined lipoprotein abnormalities (Chap- ABCG8, ARH, PCSK9) yielded insights into basic mecha-
ter 37). On the hyperlipidemic side these abnormalities nisms of cholesterol absorption and biliary excretion as
include syndromes of elevated LDL (familial hypercholes- well as contributing to basic biologic understanding of
terolemia, sitosterolemia), elevated TG (LPL deficiency, receptor-mediated endocytosis, recycling, and feedback
APOCII deficiency), elevated HDL (CETP deficiency), and regulation.86,87 The association between high LDL choles-
combined LDL/TG elevation (familial combined hyper terol and increased rates of myocardial infarction (MI)
lipidemia, dysbetalipoproteinemia). Monogenic disorders was also noticed in these families and complemented by
manifesting with extremely low lipid levels have also been observations of low rates of coronary disease in families
identified and include low LDL (familial hypobetalipo segregating unusually low LDL levels (familial hypobetali-
proteinemia, PCSK9 mutations), low HDL (familial hypoal- poproteinemia: APOB, PCSK9, ANGPTL3).87 Epidemiologic
phalipoproteinemia, lecithin cholesterol acyltransferase association and the success of statin therapy in preventing
[LCAT] deficiency, Tangier disease), and combined low coronary heart disease (CHD) extended the relationship of
cholesterol/TG (abetalipoproteinemia, chylomicron reten- LDL and CHD to the general population,88 as have common
tion syndrome). LDL-associated SNPs identified by GWAS.82
The genetic factors underlying most serum lipid varia- The large number of associated loci for LDL, HDL, and
tions are polygenic. GWAS aggregating approximately TG combined with clinical outcome data on population-
200,000 multiethnic samples have identified more than sized cohorts of genotyped individuals has enabled causal-
150 independent loci associated with serum lipids through ity testing for epidemiologic associations with important
common variant association accounting for about 15% of public health consequences. Elevated HDL cholesterol
heritability.81 The effect sizes of common variants on lipid levels have been associated with protection from MI, but
levels range from less than 1 mg/dL to about 15 mg/dL the causality of this association is controversial. Should
(SNP rs964184 at APOA1 and TG).82 public health efforts be made to raise HDL levels in the
Among the loci identified, many alter one of the lipo- population? Should drugs that increase HDL levels con-
proteins identified, a few (CETP, TRIB1, FADS1-2-3, APOA1) tinue to be developed after early clinical failures?89 By lever-
alter all lipoprotein levels, and a subset alters various com- aging common variants associated with HDL levels, the
causality of HDL in heart disease can be tested using an increasing gene/protein function, and (3) nature has per-
approach called mendelian randomization.90 As described formed a lifelong clinical trial of inhibiting gene function,
earlier, genetic associations imply causality because geno- and the side effects of doing so are known. In the case of
type precedes phenotype, mitigating epidemiologic issues PCSK9, individuals with loss-of-function mutations dem-
of confounding, bias, and reverse causation. Mendelian onstrated no phenotypic abnormalities other than low LDL
randomization can be conceived of as a clinical trial per- levels and decreased risk of heart attack. In a similar exam-
formed by nature in which the subjects are randomized at ple, naturally occurring mutations that disrupt NPC1L1
conception to genetic variants associated with a risk factor function, the inhibitory target of ezetimibe, were found to
(e.g., SNPs increasing levels of HDL cholesterol). The sub- be associated with reduced plasma LDL cholesterol levels
jects randomized to genotype are then assessed for the and a reduced risk of CHD.98
outcome (e.g., MI) and relative risk for the outcome com- Genetic risk predictors of CHD based on lipid-associated
pared in the treated and untreated arms. By utilizing a loci alone add little to the excellent risk prediction already
mendelian randomization approach with SNPs quantita- provided by clinical risk factors, but when combined with
tively associated with lipid levels, investigators found that CHD-associated loci they can provide clinically meaningful
genetic variants increasing HDL cholesterol were not pro- CHD risk prediction. An early study utilized nine SNPs at
tective for MI, whereas genetic variants decreasing LDL loci known to alter lipid metabolism (LDLR, APOB, HMGCR,
cholesterol were protective of MI.91 These findings are con- PCSK9, CTEP, LPL, ABCA1, APOE, and LIPC) to construct a
sistent with drug trials showing that LDL-lowering agents genotype risk score and predict time to first cardiovascular
protect from MI, whereas multiple agents aimed at increas- event in a Swedish cohort.99 Even when the baseline lipid
ing HDL do not.92 levels were corrected for the genotype, the risk score was
For TG, on the other hand, multiple lines of genetic statistically associated with cardiovascular disease, indicat-
evidence have supported a causal connection with CHD. ing that the genotype-based score contained information
First, rare, loss-of-function mutations in APOC3 are asso not captured by serum lipid assessment.99 However, when
ciated with low levels of TG and are also protective for this nine-SNP risk score was added to a score derived from
ischemic cardiovascular disease in European93 and Ameri- standard clinical risk factors, it did not improve risk predic-
can cohorts.94 Second, a mendelian randomization study tion further. A subsequent study added SNPs associated
showed that SNPs that raised serum TG levels also directly with CHD from GWAS unrelated to lipid levels to
increased rates of CHD.95 Finally, because of the inter create a 13-SNP genotype risk score.100 This 13-SNP risk
relatedness of TG, LDL, and HDL, investigators have sys- score also did not improve risk prediction over clinical risk
tematically examined all lipid-associated loci in cohorts factors alone.100 By contrast, a 27-SNP risk score consisting
phenotyped for CHD to dissect the contribution of TG of both lipid and CHD-associated loci provided a threefold
to CHD risk apart from LDL and HDL.81 By constructing reduction in the number needed to treat with statins to
a statistical framework to account for the pleiotropic prevent a CHD event as compared to clinical risk factors
effects of SNPs on all three lipoprotein levels, they dem- alone.101 This study was performed in high-risk primary
onstrated that (1) SNPs that alter LDL and TG in the same prevention cohorts.101
direction of effect are associated with CHD risk, (2) SNPs Given the broad indications for statins in primary
that exclusively alter TG levels are also associated with and secondary CHD prevention,102 genetics can help to
CHD, and (3) the strength of a SNP’s effect on TG levels predict and elucidate side effects. GWAS has identified a
independently correlated with the magnitude of effect on few reproducible loci (APOE, LPA, SLCO1B1, and SORT1/
CHD risk.81 CELSR2/PSRC1) for the trait of LDL response following
statin therapy.103 SLCO1B1 encodes the organic anion
transporter, OATP1B1, which has been shown to regulate
Clinical Translation the hepatic uptake of statins.104 When exposed to a single
In the area of therapeutics, mapping of genes that affect dose of simvastatin, individuals carrying a missense variant
lipid levels has exemplified a promising approach to drug in SLCO1B1 (V174A), which causes loss-of-function, have
target identification: genes that protect from disease when up to a 2.5-fold increase in plasma levels of statin.104 A
inactivated by nature might be useful pharmacologic tar- GWAS for statin-induced myopathy performed in a second-
gets. Statins, which inhibit HMG-CoA reductase (encoded ary prevention cohort receiving high-dose simvastatin
by HMGCR), are among the most therapeutically successful identified the same genetic signal (via a noncoding SNP
drugs in lowering LDL levels and decreasing risk for CHD within the same haplotype) at SLCO1B1, which conferred
in both primary and secondary prevention settings. GWAS a 4.5-fold risk of myopathy for one allele and a 16.9-fold
did identify HMGCR as a locus altering LDL levels (with an risk of myopathy in homozygous individuals.105 The inves-
effect size of about 3 mg/dL),82 but how would this particu- tigators estimated that 60% of the myopathy cases in their
lar locus be prioritized as a therapeutic candidate among cohort were due to this variant at SLCO1B1.105 They per-
over 100 other associated loci? In some cases, experiments formed GWAS on only 200 case controls for myopathy, and
of nature (i.e., an allelic series) can be used to infer a dose- it is likely that additional pharmacogenetically relevant
response curve of gene function that indicates how loci like SLCO1B1 will be identified in future studies with
enhancement or suppression of the encoded protein’s larger sample sizes. A number of observational and inter-
activity raises or lowers disease risk. In the well-known case vention studies have also associated statin therapy with
of PCSK9, for instance, loss-of-function mutations decrease increased T2D risk. An important question is whether this
LDL and cardiovascular disease risk, whereas gain-of-func- risk is mediated by an off-target effect of statins or an
tion mutations increase LDL and cardiovascular disease on-target effect via HMG-CoA reductase. An off-target
risk.96 Early clinical trials suggest that inhibiting the pro- effect suggests that new, more specific statins could be
tein encoded by PCSK9 is a promising strategy for lowering developed without this side effect, whereas an on-target
LDL and preventing cardiovascular disease.97 Identifying effect implies that the development of more potent and
genes that, when inactivated by nature, protect from dis- specific statins would increase the risk of T2D. Genetics has
ease offers several advantages for therapeutic targeting: (1) begun to shed light on this question. Taking a mendelian
the target is already validated in humans, (2) designing randomization approach, investigators have found that
inhibitors of gene/protein function is more tractable than SNPs at HMGCR that decrease LDL levels also increase BMI,
insulin resistance, and risk for T2D, suggesting an on-target Genome Screening in the General Population
mechanism for statin-mediated T2D risk.
Population-based screening for mendelian mutations
known to cause disease would seem to be an advantageous
CONSIDERATIONS FOR CLINICAL USE application of genome sequencing. As mentioned earlier,
certain mutations in the RET gene predispose to aggressive
OF GENETIC INFORMATION AND MTC with such high penetrance that the American Thyroid
SEQUENCING IN ENDOCRINOLOGY Association recommends prophylactic thyroidectomy in
infants under 1 year of age.25 As genome sequencing
As detailed previously, the clinical applications of genetic becomes commonplace, it seems reasonable that genomes
information include diagnosis, prognosis, risk prediction, from the general population (e.g., as part of newborn
and personalized therapy (e.g., pharmacogenetics). In the screening) might be examined for RET mutations causing
past, the cost of DNA sequencing placed limitations on the this rare but potentially fatal disease. Cases of cancer might
use of genetic testing. Revolutionary advances in sequenc- be prevented and lives saved. However, one must consider
ing technologies (collectively referred to as next-generation that the clinical data on RET gene mutations have been
sequencing [NGS]) have made it feasible to sequence every obtained from kindreds affected by MEN2A/B and familial
gene in the genomes of all individuals at low cost. With MTC (Chapter 39). Are genotype-phenotype correlations
the barriers of cost and feasibility diminishing every year, between RET mutations and MTC, or between other gene
we believe that the use of genetic information will become mutations and risk of other diseases, applicable to someone
commonplace in clinical practice. To maximize benefits to from the general population with no family history of
patients while minimizing the burden of false-positive and disease?
false-negative results, clinicians will have to select the right An investigation of mendelian diabetes mutations ascer-
patients, deploy the appropriate genetic testing technol- tained in the general population suggests that genotype-
ogy, and interpret the results. Specific clinical algorithms phenotype correlations identified in mendelian disease
for patient/genetic test selection are being proposed for families may not generally hold for the population at large.
various endocrine diseases (e.g., short stature),75 but they As a case in point, genome sequencing of a population-
will take time to validate. With this in mind, we present a based U.S. cohort phenotyped longitudinally for diabetes
series of broad patient scenarios ranging from those with identified 25 individuals with mutations previously known
no clinically apparent disease to affected individuals with to cause autosomal dominant diabetes (MODY). Despite
clinically identifiable genetic syndromes, summarizing harboring mutations from a curated catalog of disease
benefits and caveats of genetic testing in each scenario. We genes (Human Gene Mutation Database Professional106)
subsequently review issues related to targeted and genome- only one of these individuals met clinical criteria for
wide testing and provide some guidance for patient and MODY, and overall this group of MODY mutation carriers
test selection. Finally, we provide an overview of disease- developed diabetes at a rate no different from that in
relevant classification of genetic findings, examine the the general population. As population-based sequencing
interpretation of genetic information as presented in a becomes prevalent, it will become necessary to reexamine
clinical laboratory report, and make suggestions for clinical estimates of penetrance and genotype-phenotype correla-
decision making (Fig. 4-3). tions in the context of differing genetic backgrounds (e.g.,
Affected
individual
Defined Undefined Not likely

Clinical assessment monogenic monogenic monogenic
syndrome syndrome
Single gene/ Exome Exome Whole Limited

Genetic testing 10 gene sequencing sequencing genome utility
panel sequencing
1–10
Raw results variants 1000s 10,000
Bioinformatic filtering None Benign VUS Pathogenic Benign VUS Pathogenic
All results Not all Reported, Reported Not all Reported, Reported
Clinical interpretation returned reported NOT and reported NOT and
actionable actionable actionable actionable
Figure 4-3 Suggested use of targeted and genome-wide genetic testing in individuals suspected of a harboring a monogenic/mendelian disease. VUS, variant of uncertain
significance (see text).
ancestry) and environment. The current evidence and state ism and should be screened more vigilantly as a result of
of knowledge do not support genome sequencing for the genetic diagnosis to prevent future infertility. Some-
population-based screening. times, a genetic diagnosis can allow a patient to avoid
lifelong treatment and testing. Individuals with diabetes
Genetic Information and Sequencing in from GCK mutations (MODY2) manifest hyperglycemia
Individual Patients but are still able to regulate their blood sugar levels such
that they neither require hypoglycemic therapy nor are at
Asymptomatic Individuals elevated risk for secondary complications of diabetes. More
Endocrinologists may be referred individuals with no generally, and especially for heterogeneous endocrine dis-
apparent symptoms in whom risk of disease may be eases such as diabetes, genetic classification has the poten-
increased: (1) those with a family history of known genetic tial to guide pharmacotherapy. An existing example of
disease who have not yet been tested and (2) those without genetically guided pharmacotherapy can be found in indi-
family history who were tested and found with an appar- viduals with diabetes caused by HNF1A mutations. In clini-
ently pathogenic mutation (i.e., a genetic incidentaloma). cal trials, these HNF1A mutation carriers were much more
As with any medical test, the implication of genetic testing sensitive to sulfonylureas than noncarriers and maintained
with regard to the likelihood of developing disease in the more durable glycemic control without additional agents.60
individual depends on a combination of inherent test char- As in the above-described cases, genetic testing of
acteristics (quantified by sensitivity/specificity) and the affected patients is most likely to be informative when a
pretest likelihood of disease. For individuals with a family high penetrance genetic variant is identified (i.e., for rare
history of genetic disease the pretest probability of disease mendelian-type diseases). We offer the following criteria
can be as high as 1 in 2 for a highly penetrant autosomal and mnemonic (SSSS), which should raise clinical suspi-
dominant disorder or 1 in 4 for a sibling of an individual cion for a mendelian disorder: (1) severe—more severe than
with a recessive disorder. Thus, for individuals with a usual for the disease, (2) segregating—multiple affected
family history of mendelian endocrine disease, genetic family members or family history of consanguinity, (3)
testing is often warranted (depending on the risks and syndromic—features such as dysmorphism, developmental
benefits for the individual patient at hand, including psy- abnormalities, unusual biochemical phenotypes, etc.), and
chosocial factors); this fits well with current clinical prac- (4) too soon—early age of onset. The diagnostic yield of
tice for asymptomatic individuals from families with genetic testing is highly variable both due to technologic
mendelian disorders. reasons and underlying genetic (locus) heterogeneity. Even
On the other hand, for individuals with no family for well-described mendelian syndromes such as MEN1,
history presenting with a mendelian-like genetic inciden- 30% of kindreds have no identifiable mutations identified
taloma, the pretest likelihood of disease is that of the popu- by MEN1 gene sequencing. For affected individuals with
lation (1 : 10,000 to 1 : 100,000). Even with a mutation unknown syndromes, or for whom targeted sequencing
conferring 50-fold increased risk, the individual is far more has failed to make a diagnosis, systematic gene sequencing
likely to remain free of disease. So what reassurance can be using NGS methods would be required, and diagnostic
offered to such individuals who carry mendelian muta- yields are typically lower. A National Institutes of Health
tions? Population-based sequencing surveys reveal that, on (NIH) study of exome sequencing for rare disorders found
average, the genome of an apparently healthy individual a diagnosis in 20% of cases.108
contains approximately 100 mendelian-like disruptive
mutations (i.e., frameshifting indels and SNPs resulting in Selection of Genetic Tests: Targeted Versus
premature stop codons) up to 20 of which are homozy-
gously inactivated.107 Thus, even the protein-coding
Genome-wide Approaches
portion of the human genome (comprising only 1-2% of If cost were no object, one might consider deploying whole
the total genome) contains unexpected redundancy that genome sequencing to maximize sensitivity and find the
protects from disease. Depending on the severity of clinical “smoking gun” mutation in every case. Using genome
consequences, and the estimate of penetrance in the sequencing in this way is tantamount to performing thou-
general population, watchful waiting can be a prudent sands of genetic tests at once and clinically analogous to
course of action. ordering every possible hormone level for every endocrine
patient. Intuitively, this approach is not pursued because
every test carries with it the possibility of false-positive re-
Symptomatic Individuals sults. The more tests one orders, the more likely at least one
Individuals with symptomatic disease can present with result will be false. The same logic applies to genetic tests
clinically defined or unknown syndromes. In both cases, a and genome sequencing. Even if genetic variants are identi-
genetic diagnosis can be of psychological benefit, inform fied with 100% analytical sensitivity and specificity, their
family planning, and sometimes can direct therapeutic clinical sensitivity and specificity for disease risk are much
screening/intervention. The genotype-phenotype correla- less, due to incomplete penetrance, variable expressivity,
tion of endocrine tumor syndromes (MEN1, MEN2; see and our incomplete knowledge of genotype-phenotype
previous discussion) are a classic example of the benefits correlations. Thus, targeted testing (either measuring varia-
of genetic diagnosis for directed screening and prophylac- tion in a requested set of genes or masking/only reporting
tic interventions in both affected individuals and close back variation in the requested set of genes) can decrease
relatives. For example, the genotype-phenotype correlation false-positive results. In addition, clinical laboratories that
of RET oncogene mutations dictates the urgency of pro- focus on testing specific sets of genes may be particularly
phylactic thyroidectomy (ranging from <1 year old to >5 well versed in interpretation of variation in these genes,
years old) in children.25 Thus, even with a clinical diagno- potentially improving sensitivity and specificity. Of course,
sis, genetic testing can have great prognostic value. Another when targeted testing is negative, then more comprehen-
example is the 3-M syndrome (clinically defined by short sive testing may be required to make a diagnosis.
stature, facial dysmorphism, and skeletal abnormalities) However, even beyond the issue of false-positive find-
caused by mutations in the CUL7 gene (among others). ings, targeted testing may offer superior analytic perfor-
Males with this syndrome are at high risk for hypogonad- mance. Detection of variants can be highly variable in
current versions of genome-wide sequencing, in which mil- variants in a haystack of benign ones. Based on the assump-
lions of genetic variants are identified simultaneously.109 tion that the testing was done for a highly penetrant,
There is often a tradeoff between the number of variants mendelian variant, most genetic variants can be filtered as
identified and the sensitivity/specificity for detecting and benign based on having been observed at a frequency
calling any individual variant. This analytic sensitivity/ greater than 1% (or even lower thresholds for rare disor-
specificity relates to coverage, which is the depth of ders) in suitable reference populations. Computational
sequencing performed, or the number of independent analysis can support a benign classification by showing a
times a particular nucleotide has been sequenced in a variant is silent (e.g., encodes the same amino acid as the
single test. Depth can vary greatly across the genome. For reference base). Experimental data can provide evidence
example, a clinical trial of whole-genome sequencing that a missense variant does not alter protein function in
reported clinical grade genome sequencing at “30× cover- disease-relevant bioassays. In addition, family-based data
age on average and at least 8× coverage for more than 95% can be extremely powerful for interpretation; for example,
of bases.”110 This means that of the 3 billion bases in the the vast majority of variants for dominant diseases can be
human genome, each base was sequenced independently classified as benign if they are also observed in healthy
30 times on average, and more than 95% of bases were relatives. Typically, the filtering of benign variants results
observed at least 8 times. However, 5% of the human in a 100-fold to 1000-fold reduction in the number of vari-
genome was still poorly observed or missing. Thus, if clini- ants, requiring further analysis to 30 to 300 variants.12,110
cal suspicion motivated examination of certain genes or Among these, a similar hierarchy of evidence can
genomic regions, a targeted approach might well have support a pathogenic designation, or the evidence can
higher analytic sensitivity and specificity. The above- remain inconclusive, resulting in a VUS designation. Com-
described clinical grade genome is sufficient for detecting putational analysis showing predicted gene disruption
variants at a frequency found with germline heterozygosity (premature stop or frameshift variants) in a gene known to
(on average 50% of the molecules sequenced would contain cause disease when inactivated provides strong evidence of
the variant base), but for disorders requiring detection pathogenicity. Experiments in disease-relevant bioassays
below germline heterozygosity, such as somatic mutation can demonstrate deleteriousness of a variant, and family-
testing in tumors, higher coverage would be required. based data can increase the conclusiveness of findings by
The molecular type of genetic variation (single-base showing that a proposed pathogenic variant segregates
changes or more complex variation such as insertions/ with disease or that it is absent from both parents in de
deletions/duplications) strongly affects the analytic sensi- novo cases of disease. Presence of the variant in databases
tivity and specificity. The most commonly used gene-panel such as the Human Gene Mutation Database (HGMD) can
and genome-wide tests rely on NGS technology, which is also provide supportive evidence, but the quality of evi-
currently well suited for detecting single or multiple base dence can vary widely among variants. Before final report-
variations, but poorly optimized for the detection of struc- ing, variants classified as pathogenic are typically validated
tural variants (chromosomal rearrangements), triplet repeat by resequencing using traditional methods (i.e., Sanger
expansions, and CNVs. For example, all false-positive sequencing),113 although this practice may shift depending
genetic variants identified in a pilot clinical-grade genome on the state of sequencing technology. In a recent pilot
sequencing study were indels or were near repetitive DNA case-control study of 20 cardiomyopathy patients and con-
stretches.110 As NGS technology and genome reconstruc- trols subjected to genome sequencing, 2 to 4 variants per
tion algorithms improve, so will the ability to accurately individual (in both cases and controls) were classified as
detect these more complex genetic variants, reducing both pathogenic,110 illustrating the challenges in interpreting
false-negative and false-positive results. Currently, alterna- genetic variation in isolation.
tive modes of testing, such as chromosomal microarrays
(array CGH), are used to identify large (>50 kb) structural Using a Genetics Laboratory Report to Make
and CNVs.111
Clinical Decisions
In the clinic, genetic testing is used to identify or confirm
Interpretation of Identified Genetic Variants the cause of disease and to help the physicians make indi-
Once genetic variants are identified (from sequencing or vidualized treatment decisions (Table 4-5). Given the com-
otherwise), they must be interpreted for their impact on plexity of genetic testing, especially at the genomic (exome
health and disease. This interpretation requires the integra- or genome) scale, physicians and clinical laboratories will
tion of population data (to know whether a variant is seen have to work collaboratively to achieve useful results. For
at higher frequency than expected for a pathogenic variant), example, when a laboratory finds a rare or novel variant
computational predictions, experimental evidence, and in the course of genomic sequencing, the director cannot
familial comparisons. Curated databases are being estab- assume it is relevant to a patient just because it is rare, or
lished to begin to accurately catalogue this information novel. The context of the patient’s history, physical exami-
and assist in interpretation. The ClinVar archive aggregates nations, and previous laboratory examinations are key to
information about genomic variation and its relationship distinguishing among causal variants for the patient’s dis-
to human health.112 order, incidental findings, and benign variants.
Identified variants can be classified into three broad How should identified variants be used in clinical deci-
clinical categories: benign variants, pathogenic variants, sion making? Variant analysis contains uncertainties that
and variants of uncertain significance (VUS), but interme- are implicitly defined in the three-part classification
diate categories have also been proposed.113 In single-gene described earlier: benign, pathogenic, and VUS. American
or disease-targeted testing, the number of identified vari- College of Medical Genetics (ACMG) guidelines113 recom-
ants is small enough to allow for the individual assessment mend that variants classified as pathogenic can be used in
of all variants in each patient, once common benign varia- clinical decision making. However, genetic evidence should
tions are curated. However, exome sequencing identifies not be the only evidence of disease and should be used
tens of thousands of variants, and genome sequencing in conjunction with complementary clinical information
identifies several million in each individual.114 Thus, auto- when possible, especially as some pathogenic variants may
mated filtering is necessary to point out a few pathogenic be misclassified. Examples of corroborative clinical data
TABLE 4-5 able incidental finding and how they may be requested or
Examining a Clinical Genetics Laboratory Report
declined. Guidelines have been set forth in the ACMG
recommendations for reporting of incidental findings in
A clinician considers several issues when examining a clinical genetics clinical exome and genome sequencing.115
laboratory report:
Technical Summary of Genetic Testing Future Perspectives and Summary
For targeted tests, the list of genes tested and coverage for each
gene can be found at the Genetic Testing laboratory (http:// In the future, we anticipate that a genome sequence will
www.ncbi.nlm.nih.gov/sites/GeneTests). For genomic tests, average become a standard accompaniment to the medical chart;
coverage (e.g., >30×) and minimum coverage (e.g., >8× for 95% of thus, the question Should we sequence? will transmute to
bases) should be considered in evaluating test quality. Genes with What part of the sequence should we look at? A rational
poor coverage or no coverage should be listed. For both targeted clinical approach will require the discipline to not look at
and genomic testing, the specific limitations of detection for all of it, or at least to rigorously interpret sequence data in
different molecular classes of genetic variants given the technology the clinical context. As detailed previously, every human
utilized (e.g., copy number variants are not well captured by genome is littered with thousands of VUS and multiple
next-generation sequencing) should be described.
variants classified as pathogenic; clinical suspicion is essen-
Clinical Interpretation tial to help direct where to look and how to interpret
With regard to the clinical indication for testing, was the test positive genetic variation. This approach recapitulates current clini-
(an explanatory variant identified for the patient’s condition), cal algorithms for genetic testing. For example, familial
negative (no explanatory variants identified), or inconclusive (only paraganglioma shows locus heterogeneity as a mendelian
variants of uncertain significance [VUS] identified)? Does the disorder with cases attributable to mutations in multiple
potential explanatory variant fit with the clinical scenario, or at least genes encoding proteins for the succinate dehydrogenase
explain some of the patient’s phenotypes? With regard to incidental complex. The current genetic testing algorithm is hierar-
findings, is there a carrier risk to future progeny or future risk of
monogenic disease to the patient? A compendium of monogenic
chical, starting with sequencing of the SDHD gene where
diseases and their patterns of inheritance can be found at the the majority of causal mutations are found (Chapter 16).
Online Mendelian Inheritance in Man (OMIM) website If no mutations in SDHD are found, other complex members
(www.omim.org/). are tested (e.g., SDHC). In a genome-sequencing era, a clini-
cal algorithm might hierarchically look up mutations in
Variant Reporting and Classification
the succinate dehydrogenase complex members from a
When applicable, the gene name, transcript, molecular form of variant sequenced individual with familial paraganglioma syn-
(SNP, indel, etc.), base changes, amino acid change, zygosity, drome. No additional sequencing costs would be incurred
population frequency, and classification (benign, pathogenic, VUS)
as each gene is subsequently tested, but honing in on the
should be provided. Naming conventions are determined by the
HUGO Gene Nomenclature Committee (http://www.genenames appropriate and interpretable areas of the genome will
.org). Variant frequencies in the population can be found at the 1000 reduce the clinical burden of false-positive results.
Genomes Project (http://www.1000genomes.org) and for exomes at In summary, genetic information is most likely to be of
the Exome Aggregation Consortium (http://exac.broadinstitute.org/). clinical use in individuals with suspected mendelian syn-
The justification for variant classifications should be provided, taking dromes (see 4S criteria enumerated earlier). For individuals
into account family-based data if available and information from with a clinically defined syndrome, for which targeted
clinical databases. Previously reported relationships among variants panels exist and are well validated, a targeted approach
and phenotypes can be found at ClinVar (http://www.ncbi.nlm.nih (single-gene or gene panel testing) is currently recom-
.gov/clinvar).112 The analytic accuracy of each reported variant in
mended as an initial approach. For example, genome-wide
terms of coverage and validation (e.g., Sanger resequencing) should
be reported. sequencing is likely not needed when MEN2B is suspected
on clinical grounds; sequencing RET will usually make a
diagnosis. If results from targeted genetic testing are unin-
formative, and the suspicion of a genetic disorder is high,
exome or genome sequencing will make additional diag-
noses in some patients (see Fig. 4-3). We recommend
include enzyme assays, physical findings, and imaging primary genome-wide approaches that assess both struc-
studies. VUS should generally not be used in clinical deci- tural variation and sequence variation for individuals with
sion making. Efforts to resolve the classification of the clinically unclassifiable genetic syndromes or when tar-
variant as pathogenic or benign should be undertaken, and geted panels are not available or well validated. Depending
interpretation in the context of the patient’s clinical sce- on technologic progress, this may simply be an unmasking
nario is critical. Variants classified as benign can usually be of data that had not been reported back in a targeted test
assumed not to cause the patient’s disorder. or may require new sequencing. The exome comprises 1%
Detection of pathogenic variants incidental to the diag- to 2% of the genome yet contains nearly 85% of known
nostic motivation for sequencing, but of potential clinical disease-causing mutations.116 Thus, given current technolo-
relevance, will be an inevitable consequence of genomic gies, exome sequencing supplemented with or following
testing. This scenario is analogous to the inevitable detec- array CGH is a reasonable initial genome-wide approach.
tion of adrenal masses in computed tomography scans or Many other best practices will improve the outcome of
thyroid nodules on physical examination. A clinician genetic diagnosis through sequencing. Ideally, DNA from
ordering genomic testing should be aware of laboratory both parents should be obtained, if possible, and DNA
policies and current ethical guidelines regarding such inci- from additional relatives may also aid in interpretation.
dental or secondary findings. Current recommendations If unaffected and affected tissues are identified, paired
are to offer the patient the option not to receive such inci- affected tissue-blood samples should be obtained when
dental findings, and laboratories may vary in their report- possible. Identified variants can be classified as benign,
ing of such incidental findings. From both the clinician pathogenic, or VUS based on cross referencing with data-
and patient perspective, incidental findings can also be bases of diseased and undiseased individuals as well as
specifically requested or declined. The laboratory should family members, computational analysis, and experimen-
provide clear information about what constitutes a report- tal evidence. Indeed, in order to maximize interpretability
of any genomic approach, it will be vital to interpret varia- 26. Kapoor RR, James C, Hussain K. Advances in the diagnosis and man-
tion in the context of massive numbers of genome agement of hyperinsulinemic hypoglycemia. Nat Clin Pract Endocrinol
Metab. 2009;5(2):101-112.
sequences obtained in healthy individuals and in patients 27. O’Rahilly S. Human genetics illuminates the paths to metabolic
with disease. Finally, accurate classification will require disease. Nature. 2009;462(7271):307-314.
physicians and clinical laboratories to work collaboratively, 28. Bonnefond A, Froguel P. Rare and common genetic events in type 2
and the resulting genetic information should always be diabetes: what should biologists know? Cell Metab. 2015;21(3):
357-368.
used in conjunction with complementary data (chemis- 29. Morris AP, Voight BF, Teslovich TM, et al. Large-scale association analy-
tries, imaging, etc.) for clinical decision making. sis provides insights into the genetic architecture and pathophysiology
of type 2 diabetes. Nat Genet. 2012;44(9):981-990.
30. Wellcome Trust Case Control Consortium. Genome-wide association
REFERENCES study of 14,000 cases of seven common diseases and 3,000 shared
controls. Nature. 2007;447(7145):661-678.
1. Sturtevant AH. A History of Genetics. New York, NY: Harper & Row; 31. Todd JA, Bell JI, McDevitt HO. HLA-DQ beta gene contributes to sus-
1965:165. ceptibility and resistance to insulin-dependent diabetes mellitus.
2. Botstein D, White RL, Skolnick M, Davis RW. Construction of a genetic Nature. 1987;329(6140):599-604.
linkage map in man using restriction fragment length polymorphisms. 32. Rotter JI, Landaw EM. Measuring the genetic contribution of a single
Am J Hum Genet. 1980;32(3):314-331. locus to a multilocus disease. Clin Genet. 1984;26(6):529-542.
3. Fisher RA. The causes of human variability. Eugen Rev. 1919;10(4): 33. Jeninga EH, Gurnell M, Kalkhoven E. Functional implications of
213-220. genetic variation in human PPARgamma. Trends Endocrinol Metab.
4. Altshuler D, Daly MJ, Lander ES. Genetic mapping in human disease. 2009;20(8):380-387.
Science. 2008;322(5903):881-888. 34. Voight BF, Scott LJ, Steinthorsdottir V, et al. Twelve type 2 diabetes
5. Visscher PM, Medland SE, Ferreira MAR, et al. Assumption-free estima- susceptibility loci identified through large-scale association analysis.
tion of heritability from genome-wide identity-by-descent sharing Nat Genet. 2010;42(7):579-589.
between full siblings. PLoS Genet. 2006;2(3):e41. 35. Bouatia-Naji N, Bonnefond A, Cavalcanti-Proenca C, et al. A variant
6. Kaprio J, Tuomilehto J, Koskenvuo M, et al. Concordance for type 1 near MTNR1B is associated with increased fasting plasma glucose levels
(insulin-dependent) and type 2 (non-insulin-dependent) diabetes mel- and type 2 diabetes risk. Nat Genet. 2009;41(1):89-94.
litus in a population-based cohort of twins in Finland. Diabetologia. 36. Bonnefond A, Clement N, Fawcett K, et al. Rare MTNR1B variants
1992;35(11):1060-1067. impairing melatonin receptor 1B function contribute to type 2 diabe-
7. Diabetes mellitus in twins: a cooperative study in Japan. Committee tes. Nat Genet. 2012;44(3):297-301.
on Diabetic Twins, Japan Diabetes Society. Diabetes Res Clin Pract. 37. DIAbetes Genetics Replication And Meta-analysis (DIAGRAM) Consor-
1988;5(4):271-280. tium, Asian Genetic Epidemiology Network Type 2 Diabetes (AGEN-
8. Kondrashova A, Viskari H, Kulmala P, et al. Signs of beta-cell autoim- T2D) Consortium, South Asian Type 2 Diabetes (SAT2D) Consortium,
munity in nondiabetic schoolchildren: a comparison between Russian et al. Genome-wide trans-ancestry meta-analysis provides insight into
Karelia with a low incidence of type 1 diabetes and Finland with a the genetic architecture of type 2 diabetes susceptibility. Nat Genet.
high incidence rate. Diabetes Care. 2007;30(1):95-100. 2014;46(3):234-244.
9. Kondrashova A, Reunanen A, Romanov A, et al. A six-fold gradient in 38. SIGMA Type 2 Diabetes Consortium, Williams AL, Jacobs SB, et al.
the incidence of type 1 diabetes at the eastern border of Finland. Ann Sequence variants in SLC16A11 are a common risk factor for type 2
Med. 2005;37(1):67-72. diabetes in Mexico. Nature. 2014;506(7486):97-101.
10. Onengut-Gumuscu S, Chen WM, Burren O, et al. Fine mapping of type 39. Hara K, Fujita H, Johnson TA, et al. Genome-wide association study
1 diabetes susceptibility loci and evidence for colocalization of causal identifies three novel loci for type 2 diabetes. Hum Mol Genet. 2014;
variants with lymphoid gene enhancers. Nat Genet. 2015;47(4): 23(1):239-246.
381-386. 40. Moltke I, Grarup N, Jorgensen ME, et al. A common Greenlandic
11. Jaiswal S, Fontanillas P, Flannick J, et al. Age-related clonal hemato- TBC1D4 variant confers muscle insulin resistance and type 2 diabetes.
poiesis associated with adverse outcomes. N Engl J Med. 2014;371: Nature. 2014;512(7513):190-193.
2488-2498. 41. Vassy JL, Hivert MF, Porneala B, et al. Polygenic type 2 diabetes predic-
12. Tennessen JA, Bigham AW, O’Connor TD, et al. Evolution and function at the limit of common variant detection. Diabetes. 2014;63(6):
tional impact of rare coding variation from deep sequencing of human 2172-2182.
exomes. Science. 2012;337(6090):64-69. 42. Dimas AS, Lagou V, Barker A, et al. Impact of type 2 diabetes suscep-
13. Flannick J, Beer NL, Bick AG, et al. Assessing the phenotypic effects in tibility variants on quantitative glycemic traits reveals mechanistic
the general population of rare variants in genes for a dominant men- heterogeneity. Diabetes. 2014;63(6):2158-2171.
delian form of diabetes. Nat Genet. 2013;45(11):1380-1385. 43. Zeggini E, Weedon MN, Lindgren CM, et al. Replication of genome-
14. International HapMap Consortium. The International HapMap wide association signals in UK samples reveals risk loci for type 2
Project. Nature. 2003;426(6968):789-796. diabetes. Science. 2007;316(5829):1336-1341.
15. National Human Genome Research Institute. A Catalog of Published 44. Dina C, Meyre D, Gallina S, et al. Variation in FTO contributes to
Genome-Wide Association Studies. Available at: <http://www.genome childhood obesity and severe adult obesity. Nat Genet. 2007;39(6):
.gov/gwastudies/>. 724-726.
16. Farh KK, Marson A, Zhu J, et al. Genetic and epigenetic fine mapping 45. Frayling TM, Timpson NJ, Weedon MN, et  al. A common variant
of causal autoimmune disease variants. Nature. 2015;518(7539): in the FTO gene is associated with body mass index and predis-
337-343. poses to childhood and adult obesity. Science. 2007;316(5826):
17. Nejentsev S, Howson JM, Walker NM, et al. Localization of type 1 889-894.
diabetes susceptibility to the MHC class I genes HLA-B and HLA-A. 46. Scuteri A, Sanna S, Chen WM, et al. Genome-wide association scan
Nature. 2007;450(7171):887-892. shows genetic variants in the FTO gene are associated with obesity-
18. Musunuru K, Strong A, Frank-Kamenetsky M, et al. From noncoding related traits. PLoS Genet. 2007;3(7):e115.
variant to phenotype via SORT1 at the 1p13 cholesterol locus. Nature. 47. Smemo S, Tena JJ, Kim KH, et al. Obesity-associated variants within
2010;466(7307):714-719. FTO form long-range functional connections with IRX3. Nature.
19. O’Rahilly S, Farooqi IS. Human obesity: a heritable neurobehavioral 2014;507(7492):371-375.
disorder that is highly sensitive to environmental conditions. Diabetes. 48. Church C, Moir L, McMurray F, et al. Overexpression of Fto leads to
2008;57(11):2905-2910. increased food intake and results in obesity. Nat Genet. 2010;42(12):
20. Musunuru K, Pirruccello JP, Do R, et al. Exome sequencing, ANGPTL3 1086-1092.
mutations, and familial combined hypolipidemia. N Engl J Med. 2010; 49. Wahlen K, Sjolin E, Hoffstedt J. The common rs9939609 gene variant
363(23):2220-2227. of the fat mass- and obesity-associated gene FTO is related to fat cell
21. Choi M, Scholl UI, Yue P, et al. K+ channel mutations in adrenal lipolysis. J Lipid Res. 2008;49(3):607-611.
aldosterone-producing adenomas and hereditary hypertension. Sci- 50. Dupuis J, Langenberg C, Prokopenko I, et al. New genetic loci impli-
ence. 2011;331(6018):768-772. cated in fasting glucose homeostasis and their impact on type 2 dia-
22. Scholl UI, Goh G, Stolting G, et al. Somatic and germline CACNA1D betes risk. Nat Genet. 2010;42(2):105-116.
calcium channel mutations in aldosterone-producing adenomas and 51. Bonnefond A, Vaxillaire M, Labrune Y, et al. Genetic variant in HK1
primary aldosteronism. Nat Genet. 2013;45(9):1050-1054. is associated with a proanemic state and A1C but not other glycemic
23. Plenge RM, Scolnick EM, Altshuler D. Validating therapeutic targets control-related traits. Diabetes. 2009;58(11):2687-2697.
through human genetics. Nat Rev Drug Discov. 2013;12(8):581-594. 52. Vaxillaire M, Dechaume A, Busiah K, et al. New ABCC8 mutations in
24. Hameed S, Jayasena CN, Dhillo WS. Kisspeptin and fertility. J Endocri- relapsing neonatal diabetes and clinical features. Diabetes. 2007;56(6):
nol. 2011;208(2):97-105. 1737-1741.
25. Kloos RT, Eng C, Evans DB, et al. Medullary thyroid cancer: manage- 53. Sladek R, Rocheleau G, Rung J, et al. A genome-wide association study
ment guidelines of the American Thyroid Association. Thyroid. 2009; identifies novel risk loci for type 2 diabetes. Nature. 2007;445(7130):
19(6):565-612. 881-885.
54. Flannick J, Thorleifsson G, Beer NL, et al. Loss-of-function mutations 82. Teslovich TM, Musunuru K, Smith AV, et al. Biological, clinical and
in SLC30A8 protect against type 2 diabetes. Nat Genet. 2014;46(4): population relevance of 95 loci for blood lipids. Nature. 2010;466(7307):
357-363. 707-713.
55. Altshuler D, Hirschhorn JN, Klannemark M, et al. The common PPAR- 83. Do R, Stitziel NO, Won HH, et al. Exome sequencing identifies rare
gamma Pro12Ala polymorphism is associated with decreased risk of LDLR and APOA5 alleles conferring risk for myocardial infarction.
type 2 diabetes. Nat Genet. 2000;26(1):76-80. Nature. 2015;518(7537):102-106.
56. Meigs JB, Shrader P, Sullivan LM, et al. Genotype score in addition to 84. Thormaehlen AS, Schuberth C, Won HH, et al. Systematic cell-based
common risk factors for prediction of type 2 diabetes. N Engl J Med. phenotyping of missense alleles empowers rare variant association
2008;359(21):2208-2219. studies: a case for LDLR and myocardial infarction. PLoS Genet.
57. Hariri S, Yoon PW, Qureshi N, et al. Family history of type 2 diabetes: 2015;11(2):e1004855.
a population-based screening tool for prevention? Genet Med. 2006; 85. Brown MS, Goldstein JL. Familial hypercholesterolemia: defective
8(2):102-108. binding of lipoproteins to cultured fibroblasts associated with impaired
58. Sagen JV, Raeder H, Hathout E, et al. Permanent neonatal diabetes regulation of 3-hydroxy-3-methylglutaryl coenzyme A reductase activ-
due to mutations in KCNJ11 encoding Kir6.2: patient characteristics ity. Proc Natl Acad Sci U S A. 1974;71(3):788-792.
and initial response to sulfonylurea therapy. Diabetes. 2004;53(10): 86. Brown MS, Goldstein JL. A receptor-mediated pathway for cholesterol
2713-2718. homeostasis. Science. 1986;232(4746):34-47.
59. Rafiq M, Flanagan SE, Patch AM, et al. Effective treatment with oral 87. Kathiresan S, Srivastava D. Genetics of human cardiovascular disease.
sulfonylureas in patients with diabetes due to sulfonylurea receptor 1 Cell. 2012;148(6):1242-1257.
(SUR1) mutations. Diabetes Care. 2008;31(2):204-209. 88. Holmes MV, Harrison S, Talmud PJ, et al. Utility of genetic determi-
60. Pearson ER, Starkey BJ, Powell RJ, et al. Genetic cause of hyperglycae- nants of lipids and cardiovascular events in assessing risk. Nat Rev
mia and response to treatment in diabetes. Lancet. 2003;362(9392): Cardiol. 2011;8(4):207-221.
1275-1281. 89. Barter PJ, Caulfield M, Eriksson M, et al. Effects of torcetrapib in
61. Wood AR, Esko T, Yang J, et al. Defining the role of common variation patients at high risk for coronary events. N Engl J Med. 2007;357(21):
in the genomic and biological architecture of adult human height. Nat 2109-2122.
Genet. 2014;46(11):1173-1186. 90. Smith GD, Ebrahim S. Mendelian randomization: prospects, poten-
62. Yang J, Benyamin B, McEvoy BP, et al. Common SNPs explain a large tials, and limitations. Int J Epidemiol. 2004;33(1):30-42.
proportion of the heritability for human height. Nat Genet. 2010; 91. Voight BF, Peloso GM, Orho-Melander M, et al. Plasma HDL choles-
42(7):565-569. terol and risk of myocardial infarction: a mendelian randomisation
63. Chan Y, Holmen OL, Dauber A, et  al. Common variants show study. Lancet. 2012;380(9841):572-580.
predicted polygenic effects on height in the tails of the distribu- 92. Keene D, Price C, Shun-Shin MJ, Francis DP. Effect on cardiovascular
tion, except in extremely short individuals. PLoS Genet. 2011;7(12): risk of high density lipoprotein targeted drug treatments niacin,
e1002439. fibrates, and CETP inhibitors: meta-analysis of randomised controlled
64. Dauber A, Yu Y, Turchin MC, et al. Genome-wide association of copy- trials including 117,411 patients. BMJ. 2014;349:g4379.
number variation reveals an association between short stature and the 93. Jorgensen AB, Frikke-Schmidt R, Nordestgaard BG, Tybjaerg-Hansen A.
presence of low-frequency genomic deletions. Am J Hum Genet. 2011; Loss-of-function mutations in APOC3 and risk of ischemic vascular
89(6):751-759. disease. N Engl J Med. 2014;371(1):32-41.
65. Valdar W, Solberg LC, Gauguier D, et al. Genome-wide genetic associa- 94. TG and HDL Working Group of the Exome Sequencing Project,
tion of complex traits in heterogeneous stock mice. Nat Genet. 2006; National Heart, Lung, and Blood Institute, Crosby J, Peloso GM, et al.
38(8):879-887. Loss-of-function mutations in APOC3, triglycerides, and coronary
66. Mullis PE. Genetic control of growth. Eur J Endocrinol. 2005;152(1): disease. N Engl J Med. 2014;371(1):22-31.
11-31. 95. Holmes MV, Asselbergs FW, Palmer TM, et al. Mendelian randomiza-
67. Weedon MN, Lango H, Lindgren CM, et al. Genome-wide association tion of blood lipids for coronary heart disease. Eur Heart J. 2015;36(9):
analysis identifies 20 loci that influence adult height. Nat Genet. 539-550.
2008;40(5):575-583. 96. Cohen JC, Boerwinkle E, Mosley TH Jr, Hobbs HH. Sequence variations
68. Lango Allen H, Estrada K, Lettre G, et al. Hundreds of variants clus- in PCSK9, low LDL, and protection against coronary heart disease.
tered in genomic loci and biological pathways affect human height. N Engl J Med. 2006;354(12):1264-1272.
Nature. 2010;467(7317):832-838. 97. Blom DJ, Hala T, Bolognese M, et al. A 52-week placebo-controlled trial
69. de Pontual L, Yao E, Callier P, et al. Germline deletion of the miR-17 of evolocumab in hyperlipidemia. N Engl J Med. 2014;370(19):1809-
approximately 92 cluster causes skeletal and growth defects in humans. 1819.
Nat Genet. 2011;43(10):1026-1030. 98. Myocardial Infarction Genetics Consortium Investigators, Stitziel NO,
70. Neptune ER, Frischmeyer PA, Arking DE, et al. Dysregulation of TGF- Won HH, et al. Inactivating mutations in NPC1L1 and protection from
beta activation contributes to pathogenesis in Marfan syndrome. Nat coronary heart disease. N Engl J Med. 2014;371(22):2072-2082.
Genet. 2003;33(3):407-411. 99. Kathiresan S, Melander O, Anevski D, et al. Polymorphisms associated
71. Habashi JP, Judge DP, Holm TM, et al. Losartan, an AT1 antagonist, with cholesterol and risk of cardiovascular events. N Engl J Med.
prevents aortic aneurysm in a mouse model of Marfan syndrome. 2008;358(12):1240-1249.
Science. 2006;312(5770):117-121. 100. Ripatti S, Tikkanen E, Orho-Melander M, et al. A multilocus genetic
72. Lacro RV, Dietz HC, Sleeper LA, et al. Atenolol versus losartan in chil- risk score for coronary heart disease: case-control and prospective
dren and young adults with Marfan’s syndrome. N Engl J Med. cohort analyses. Lancet. 2010;376(9750):1393-1400.
2014;371(22):2061-2071. 101. Mega JL, Stitziel NO, Smith JG, et al. Genetic risk, coronary heart
73. Jonquoy A, Mugniery E, Benoist-Lasselin C, et al. A novel tyrosine disease events, and the clinical benefit of statin therapy: an analysis
kinase inhibitor restores chondrocyte differentiation and promotes of primary and secondary prevention trials. Lancet. 2015;385(9984):
bone growth in a gain-of-function Fgfr3 mouse model. Hum Mol Genet. 2264-2271.
2012;21(4):841-851. 102. Taylor F, Huffman MD, Macedo AF, et al. Statins for the primary pre-
74. Yasoda A, Kitamura H, Fujii T, et al. Systemic administration of C-type vention of cardiovascular disease. Cochrane Database Syst Rev. 2013;
natriuretic peptide as a novel therapeutic strategy for skeletal dyspla- (1):CD004816.
sias. Endocrinology. 2009;150(7):3138-3144. 103. Postmus I, Trompet S, Deshmukh HA, et al. Pharmacogenetic meta-
75. Dauber A, Rosenfeld RG, Hirschhorn JN. Genetic evaluation of short analysis of genome-wide association studies of LDL cholesterol
stature. J Clin Endocrinol Metab. 2014;99(9):3080-3092. response to statins. Nat Commun. 2014;5:5068.
76. Wang SR, Carmichael H, Andrew SF, et al. Large-scale pooled next- 104. Wilke RA, Ramsey LB, Johnson SG, et al. The clinical pharmacogenom-
generation sequencing of 1077 genes to identify genetic causes of short ics implementation consortium: CPIC guideline for SLCO1B1 and
stature. J Clin Endocrinol Metab. 2013;98(8):E1428-E1437. simvastatin-induced myopathy. Clin Pharmacol Ther. 2012;92(1):
77. David A, Hwa V, Metherell LA, et al. Evidence for a continuum of 112-117.
genetic, phenotypic, and biochemical abnormalities in children with 105. Link E, Parish S, Armitage J, et al. SLCO1B1 variants and statin-
growth hormone insensitivity. Endocr Rev. 2011;32(4):472-497. induced myopathy—a genomewide study. N Engl J Med. 2008;359(8):
78. Domene HM, Hwa V, Argente J, et al. Human acid-labile subunit defi- 789-799.
ciency: clinical, endocrine and metabolic consequences. Horm Res. 106. Stenson PD, Mort M, Ball EV, et al. The Human Gene Mutation Data-
2009;72(3):129-141. base: 2008 update. Genome Med. 2009;1(1):13.
79. Renes JS, Willemsen RH, Wagner A, et al. Bloom syndrome in short 107. MacArthur DG, Balasubramanian S, Frankish A, et al. A systematic
children born small for gestational age: a challenging diagnosis. J Clin survey of loss-of-function variants in human protein-coding genes.
Endocrinol Metab. 2013;98(10):3932-3938. Science. 2012;335(6070):823-828.
80. Weiss LA, Pan L, Abney M, Ober C. The sex-specific genetic archi- 108. Gahl WA, Markello TC, Toro C, et al. The National Institutes of Health
tecture of quantitative traits in humans. Nat Genet. 2006;38(2): Undiagnosed Diseases Program: insights into rare diseases. Genet Med.
218-222. 2012;14(1):51-59.
81. Global Lipids Genetics Consortium, Willer CJ, Schmidt EM, et al. 109. Rehm HL, Bale SJ, Bayrak-Toydemir P, et al. ACMG clinical laboratory
Discovery and refinement of loci associated with lipid levels. Nat standards for next-generation sequencing. Genet Med. 2013;15(9):
Genet. 2013;45(11):1274-1283. 733-747.
110. McLaughlin HM, Ceyhan-Birsoy O, Christensen KD, et al. A systematic 121. International Consortium for Blood Pressure Genome-Wide Associa-
approach to the reporting of medically relevant findings from whole tion Studies, Ehret GB, Munroe PB, Rice KM, et al. Genetic variants in
genome sequencing. BMC Med Genet. 2014;15:134. novel pathways influence blood pressure and cardiovascular disease
111. Kearney HM, Thorland EC, Brown KK, et al. American College of risk. Nature. 2011;478(7367):103-109.
Medical Genetics standards and guidelines for interpretation and 122. Dauber A, Hirschhorn JN. Genome-wide association studies in pediat-
reporting of postnatal constitutional copy number variants. Genet Med. ric endocrinology. Horm Res Paediatr. 2011;75(5):322-328.
2011;13(7):680-685. 123. Chu X, Pan CM, Zhao SX, et al. A genome-wide association study
112. National Center for Biotechnology Information. ClinVar website. identifies two new risk loci for Graves’ disease. Nat Genet. 2011;43(9):
Available at: <http://www.ncbi.nlm.nih.gov/clinvar>. 897-901.
113. Richards S, Aziz N, Bale S, et al. Standards and guidelines for the 124. Eriksson N, Tung JY, Kiefer AK, et al. Novel associations for hypothy-
interpretation of sequence variants: a joint consensus recommenda- roidism include known autoimmune risk loci. PLoS ONE. 2012;7(4):
tion of the American College of Medical Genetics and Genomics and e34442.
the Association for Molecular Pathology. Genet Med. 2015;17(5): 125. Estrada K, Styrkarsdottir U, Evangelou E, et al. Genome-wide meta-
405-424. analysis identifies 56 bone mineral density loci and reveals 14 loci
114. Rehm HL. Disease-targeted sequencing: a cornerstone in the clinic. Nat associated with risk of fracture. Nat Genet. 2012;44(5):491-501.
Rev Genet. 2013;14(4):295-300. 126. Richards JB, Zheng HF, Spector TD. Genetics of osteoporosis from
115. Green RC, Berg JS, Grody WW, et al. ACMG recommendations for genome-wide association studies: advances and challenges. Nat Rev
reporting of incidental findings in clinical exome and genome Genet. 2012;13(8):576-588.
sequencing. Genet Med. 2013;15(7):565-574. 127. Kapur K, Johnson T, Beckmann ND, et al. Genome-wide meta-analysis
116. Majewski J, Schwartzentruber J, Lalonde E, et al. What can exome for serum calcium identifies significantly associated SNPs near the
sequencing do for you? J Med Genet. 2011;48(9):580-589. calcium-sensing receptor (CASR) gene. PLoS Genet. 2010;6(7):e1001035.
117. Barrett JC, Clayton DG, Concannon P, et al. Genome-wide association 128. O’Seaghdha CM, Yang Q, Glazer NL, et al. Common variants in the
study and meta-analysis find that over 40 loci affect risk of type 1 calcium-sensing receptor gene are associated with total serum calcium
diabetes. Nat Genet. 2009;41(6):703-707. levels. Hum Mol Genet. 2010;19(21):4296-4303.
118. Diabetes Genetics Initiative of Broad Institute of Harvard and MIT, 129. Goldfarb DS, Fischer ME, Keich Y, Goldberg J. A twin study of genetic
Lund University, and Novartis Institutes of BioMedical Research, and dietary influences on nephrolithiasis: a report from the Vietnam
Saxena R, Voight BF, Lyssenko V, et al. Genome-wide association analy- Era Twin (VET) Registry. Kidney Int. 2005;67(3):1053-1061.
sis identifies loci for type 2 diabetes and triglyceride levels. Science. 130. Urabe Y, Tanikawa C, Takahashi A, et al. A genome-wide association
2007;316(5829):1331-1336. study of nephrolithiasis in the Japanese population identifies novel
119. Fall T, Ingelsson E. Genome-wide association studies of obesity and susceptible loci at 5q35.3, 7p14.3, and 13q14.1. PLoS Genet. 2012;8(3):
metabolic syndrome. Mol Cell Endocrinol. 2014;382(1):740-757. e1002541.
120. Loos RJ, Lindgren CM, Li S, et al. Common variants near MC4R are
associated with fat mass, weight and risk of obesity. Nat Genet. 2008;
40(6):768-775.

Genetics of Endocrinology: Amit R. Majithia - David Altshuler - Joel N. Hirschhorn

Uploaded by

Copyright:

Available Formats

Genetics of Endocrinology: Amit R. Majithia - David Altshuler - Joel N. Hirschhorn

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Genetics of Endocrinology: Amit R. Majithia - David Altshuler - Joel N. Hirschhorn

Uploaded by

Copyright:

Available Formats

CHAPTER 4

(gene-gene interactions), environment (gene-environment TABLE 4-3

Summary TABLE 4-4

Defined Undefined Not likely

Single gene/ Exome Exome Whole Limited

Bioinformatic filtering None Benign VUS Pathogenic Benign VUS Pathogenic

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.