4.4: Some mutations may not have detectable phenotypes - Biology

4.4: Some mutations may not have detectable phenotypes - Biology

We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

Silent Changes

After mutagen treatment, the vast majority of base pair changes (especially substitutions) have no effect on the phenotype. Often, this is because the change occurs in the DNA sequence of a non-coding region of the DNA, such as in inter-genic regions (between genes) or within an intron region. Also, even if the change occurs in a base within a codon, it may not change the amino acid that it encodes (recall that the genetic code is degenerate; for example, GCT, GCC, GCA, and GCG all encode alanine) and is referred to as a silent mutation. Additionally, the base substitution may change an amino acid, but this doesn’t alter the function of the product, so no phenotypic change would occur.

Environment and Genetic Redundancy

There are also situations where a mutation can cause a complete loss-of-function of a gene, yet not produce a change in the phenotype, even when the mutant allele is homozygous. The lack of a phenotypic change can be due to environmental effects: the loss of that gene product may not be apparent in that environment, but might in another. Alternatively, the lack of a phenotype might be attributed to genetic redundancy, i.e. the encoding of similarly functioning genes at more than one locus in the genome. Thus the loss of one gene is compensated by another. This important limitation of mutational analysis should be remembered: genes with redundant functions cannot be easily identified by mutant screening.

Essential Genes and Lethal Alleles

Some phenotypes require individuals to reach a particular developmental stage before they can be scored. For example, flower color can only be scored in plants that are mature enough to make flowers, and eye color can only be scored in flies that have developed eyes. However, some alleles may not develop sufficiently to be included among the progeny that are scored for a particular phenotype. Mutations in essential genes create recessive lethal alleles that arrest the development of an individual at an embryonic stage. This type of mutation may therefore go unnoticed in a typical mutant screen because they are absent from the progeny being screened. Furthermore, the progeny of a monohybrid cross involving an embryonic lethal recessive allele may therefore all be of a single phenotypic class, giving a phenotypic ratio of 1:0 (which is the same as 3:0). In this case the mutation may not be detected.

Naming Genes

Many genes have been first identified in mutant screens, and so they tend to be named after their mutant phenotypes, not the normal function or phenotype. This can cause some confusion for students of genetics. For example, we have already encountered an X-linked gene named white in fruit flies. Null mutants of the white gene have white eyes, but the normal white+ allele has red eyes. This tells us that the wild type (normal) function of this gene is actually to help make red eyes. Its product is a protein that imports a pigment precursor into developing cells of the eye. Why don’t we call it the “red” gene, since that is what its product does? Because there are more than one-dozen genes that when mutant alter the eye colour; e.g. violet, cinnabar, brown, scarlet, etc. For all these genes their function is also needed to make the eye wild type red and not the mutant colour. If we used the name “red” for all these genes it would be confusing, so we use the distinctive mutant phenotype as the gene name. However, this can be problematic, as with the “lethal” mutations described above. This problem is usually handled by giving numbers or locations to the gene name, or making up names that describe how they die (e.g. even-skipped, hunchback, hairy, runt, etc.) .

Previously we examined how the genetic composition of a population is studied. In this tutorial we will examine the conditions that can alter genetic compositions. This theme is central to evolution. Genetically stable populations (those in Hardy-Weinberg equilibrium) do not evolve, however, genetically unstable populations do undergo evolutionary change. We will examine those conditions that affect genetic stability, and hence contribute to evolutionary change. By the end of this tutorial you should have a basic understanding of:

  • How the founder effect and bottleneck effect relate to genetic drift
  • How gene flow, mutations, and mating behavior can affect genetic stability
  • How selection can influence allele frequency

Ribosomal proteins: mutant phenotypes by the numbers and associated gene expression changes

Ribosomal proteins are highly conserved, many universally so among organisms. All ribosomal proteins are structural parts of the same molecular machine, the ribosome. However, when ribosomal proteins are mutated individually, they often lead to distinct and intriguing phenotypes, including specific human pathologies. This review is an attempt to collect and analyse all the reported phenotypes of each ribosomal protein mutant in several eukaryotes (Saccharomyces cerevisiae, Caenorhabditis elegans, Drosophila melanogaster, Danio rerio, Mus musculus, Homo sapiens). These phenotypes were processed with unbiased computational approaches to reveal associations between different phenotypes and the contributions of individual ribosomal protein genes. An overview of gene expression changes in ribosomal protein mutants, with emphasis on ribosome profiling studies, is also presented. The available data point to patterns that may account for most of the observed phenotypes. The information presented here may also inform future studies about the molecular basis of the phenotypes that arise from mutations in ribosomal proteins.

1. Overview

Ribosomes are the complex molecular machines that synthesize proteins as instructed from the genetic information on messenger RNAs (mRNAs) [1–4]. Most of the observed phenotypes in cells and organisms arise from the function of polypeptides that the ribosomes produce. Hence, ribosomes are at the critical junction of the genotype-phenotype relation in all species. Fully assembled ribosomes have large and small subunits. The small, and large, subunits in eukaryotes are referred to as the 40S, and 60S subunits, respectively, based on their sedimentation properties. 80S refers to fully assembled ribosomes. Each subunit is a ribonucleoprotein particle, composed of one (in the 40S), or three (in 60S), ribosomal RNA (rRNA) molecules, and many (79 in yeast, 80 in animals) proteins in 80S ribosomes. The ribosomal proteins are structural, non-catalytic components of ribosomes [5,6]. Bacterial ribosomes have similar architecture, but they are smaller and have fewer proteins.

The majority of ribosomal proteins are essential for ribosome function and life. In budding yeast, 15 out of a total of 79 cytoplasmic ribosomal proteins are not essential [7]. In several cases, more than one gene may encode a ribosomal protein. For example, in the budding yeast Saccharomyces cerevisiae, 59 ribosomal proteins are encoded in each case by a pair of highly similar paralogous genes. As will be described below (figure 1), cells carrying mutations in ribosomal proteins display a broad spectrum of phenotypes, depending on the locus and alleles involved. It is the objective of this review to systematically go over these phenotypes and examine how they might come about.

Figure 1. Summary of the most common phenotypes that arise from loss-of-function mutations in ribosomal proteins, in each of the organisms examined in this review. The number of mutant genes for which phenotypes have been described is shown in the second column, while the total number of phenotypes detected by mutations in any ribosomal protein gene in that organism is shown in the third column.

The focus in this review is on six eukaryotic organisms, three invertebrate ones (budding yeast, worm, fly) and three vertebrate (fish, mouse, human). The next sections will describe the following: (i) the generation of a complete matrix of ribosomal protein mutant loss-of-function phenotypes in each of the six organisms (ii) a computational approach to define and group those phenotypes and the genes that underpin them (iii) a similar analysis for gain-of-function phenotypes of ribosomal proteins in yeast (iv) a discussion of the evidence linking in some cases ribosomal protein mutants to increased proliferation, including cancer and (v) an examination of the observed phenotypes in the context of changes in gene expression, especially at the translational level, which may bridge the genotype–phenotype relationships. Lastly, it is worth pointing out that all the datasets generated here are provided in the attached files, with the hope of stimulating further analyses of the remarkable properties and consequences of ribosomal protein perturbations.

2. Data input

The gene names of the ribosomal proteins queried are listed in the electronic supplementary material, file S1, in separate spreadsheets for each species. To facilitate comparisons across species, next to each gene name is shown the new unified name of the ribosomal protein that gene encodes [8]. Note that yeast ribosomes lack the eL28 protein. The name of each gene was used to query the well-curated database for each species, to collect all the available phenotypes for that gene in that organism: SGD for S. cerevisiae ([9], WormBase for Caenorhabditis elegans ([10], FlyBase for Drosophila melanogaster ([11], ZFIN for Danio rerio ([12], MGI for Mus musculus ([13] OMIM for Homo sapiens ([14], For the most part, primary reports describing ribosomal protein mutant phenotypes were neither cited here nor used as input in the resulting phenotypic matrices. Instead, the collected phenotypes were only those included in each database, with their accompanying descriptors. Because the literature across all these species is expansive, this was the only practical, unbiased, and standardized way to build the phenotypic matrices. Hence, it is possible that additional phenotypes may exist, which were missing in curated databases queried at the time of preparing this review. Nonetheless, even if such missing cases exist, it is unlikely that they would have significantly influenced the outcome of the analyses, because of the large number of data-points already present in the database of each organism.

The phenotypic matrix for each species was assembled from the downloaded individual text files describing the reported phenotypes associated with each gene, as described previously [15]. Each matrix is shown in a sheet (species_phenotypes) of a separate supplementary file for each species (e.g. the yeast phenotypic matrix is in the electronic supplementary material, file2/sheet ‘yeast_phenotypes’ the one for worms in the electronic supplementary material, file3/sheet ‘worm_phenotypes’ and so on). For yeast, a separate phenotypic matrix was built for gain-of-function phenotypes, and it will be described separately later in this report.

3. General properties of ribosomal protein mutants

An overview of the phenotypes arising from loss-of-function mutations in ribosomal protein genes in all species is in figure 1. The number of the observed phenotypes was considerable, but they were not all observed in most ribosomal protein mutants. With all that information at hand, the first two obvious questions are: what are the most common phenotypes in loss-of-function ribosomal protein mutants, and are there any common patterns across species?

In yeast, 137 genes encoding ribosomal proteins lead to 111 loss-of-function phenotypes (electronic supplementary material, file2/sheet ‘yeast_phenotypes’). The three most common phenotypes in this single-celled organism were decreased competitive fitness, decreased resistance to chemicals, and decreased vegetative growth, observed in 90%, 89% and 80%, respectively, of all the reported loss-of-function mutants. In worms, out of 151 phenotypes observed when 77 loci were mutated, larval arrest, embryonic lethality, and maternal sterility were reported for greater than 84% of all ribosomal protein mutants (electronic supplementary material, file3/sheet ‘worm_phenotypes’). In flies, the most common phenotypes are not shared by as large a portion of mutants as in yeast and worms. Nonetheless, lethality during the larval stage, partially lethality, and the Minute phenotypes were observed in 49%, 40% and 35%, respectively, of all ribosomal protein mutants (electronic supplementary material, file4/sheet ‘fly_phenotypes’). The Minute phenotype has been studied extensively in Drosophila, and it has long been recognized to result from cell-autonomous, delayed cell cycle progression and impaired cell growth, leading to smaller cell size [16]. There is a dose-response relationship of the degree of ribosomal protein insufficiency and the strength of the Minute phenotype [17–19]. In zebrafish, there are mutants in about half of the ribosomal protein genes, leading to greater than 200 distinct phenotypes (figure 1). In greater than three-quarters of these mutants, the most common phenotypes were a decreased head size, reduced thickness of the yolk extension, and smaller eyes (electronic supplementary material, file5/sheet ‘fish_phenotypes’). In mice, there are mutants for only 23 ribosomal protein genes (figure 1). Although 289 distinct phenotypes have been observed in these mice, the most common ones, found in approximately one-quarter of these mutants, are decreased body weight, kinked tail and prenatal lethality (electronic supplementary material, file6/sheet ‘mouse_phenotypes’). Viewing this comprehensive data in its totality, it becomes clear that from yeast to mice, the most likely outcomes of loss-of-function mutations in ribosomal protein genes are: reduced or delayed cell proliferation reduced cell, organ or organismal size developmental delay, arrest or lethality (figure 1 electronic supplementary material, files 2–6).

In humans, mutations in 24 ribosomal protein genes are linked to disease (figure 1). Patients with mutations in 18 of these loci develop different types of Diamond-Blackfan anemia (electronic supplementary material, file7/sheet ‘human_phenotypes’). The remaining ribosomal protein loci are associated with poor hair cell proliferation (hypotrichosis), poor bone growth (leading to dysplasias and short stature), shorter skull (brachycephaly), absence of a spleen (asplenia), developmental delay, refractory macrocytic anemia, mental retardation, or autism. Although ribosomal protein mutations are associated with distinct types of Diamond-Blackfan anemia, in all cases, there is a failure of the bone marrow to develop properly and produce enough red blood cells [20]. There are also additional abnormalities [20], which are consistent with the most common phenotypes observed in the other model systems discussed above. For example, about half of the Diamond-Blackfan patients have physical abnormalities. These abnormalities are manifested as an unusually small head (microcephaly), small lower jaw (micrognathia) and other malformations. About a third of affected individuals also grow slowly and have short stature. Hence, in humans, as in the different organisms discussed above, the typical phenotypic manifestations of ribosomal protein loss-of-function mutations are, in essence, consequences of hypo-proliferation.

But as satisfying as the congruence of the most common phenotypes of ribosomal protein mutants may be from yeast to humans, this view oversimplifies the underlying biology. It would be erroneous to conclude that ‘you‘ve seen one ribosomal protein mutant, you‘ve seen them all’. After all, there is still such a broad spectrum of additional phenotypes in each organism (electronic supplementary material, files 2–7). The apparent multitude of these phenotypes raises further questions, such as: to reduce this complexity, can one identify phenotypes that cluster together in different groups? If so, what are the ribosomal protein genes that drive this classification? Answering these questions may offer new insights into phenotype–phenotype and gene-phenotype associations among ribosomal protein mutants.

4. Multiple correspondence analysis of ribosomal protein phenotypes

Treating the different phenotypes as distinct variables, one could apply widely used multivariate statistical techniques to simplify related phenotypic variables. Measuring the degree that the observed phenotypic variables correlate with each other, provides the basis for reducing them. If two or more phenotypic variables share some features, then based on the magnitude and direction of the relationship, the observed complexity may be simplified. Techniques implementing the above principles include factor analysis and principal component analysis [21]. For categorical data (e.g. the presence or absence of a phenotype), a related approach is that of correspondence analysis [22], to detect and group underlying structures in the phenotypic variables within a dataset [15]. As a result, one obtains a lower-dimensional view of the internal structure of the data. Such approaches can be applied to datasets where each mutant displays at least a few phenotypes. This is the case for the ribosomal protein mutants in the model organisms we discussed above, except in humans. The phenotypic terms associated with almost all ribosomal protein mutants in humans are unique to each mutant. For RPL10, there are two associated diseases: autism and spondyloepimetaphyseal dysplasia (electronic supplementary material, file7/sheet ‘human_phenotypes’). Note that although anemias are prevalent among ribosomal protein mutant patients, each locus leads to a unique type of Diamond-Blackfan anemia (electronic supplementary material, file7/sheet ‘human_phenotypes’), which were kept as separate phenotypic variables. Hence, there is a near one-to-one correspondence between a phenotypic variable and ribosomal protein locus, which precludes any attempt to reduce the dimensionality of the human dataset.

For the ribosomal protein mutant phenotypes for each of the other species, multiple correspondence analysis (MCA) was performed as described elsewhere [15]. The process is summarized in figure 2. The percentage of the variance by the first 20 dimensions in each species is shown in the scree plots in figure 3. In the next paragraphs, the following will be described for each species: (i) the number of the dimensions/clusters that explain most of the variance in the observed phenotypes (ii) the phenotypes that contribute the most to each dimension (discussion will be limited to those phenotypes with an arbitrarily chosen cutoff of correlation ≥ 0.4) and (iii) the individual genes contributing the most to each dimension (again, the discussion will be limited to genes with correlations ≥ 0.4). All the data for each organism can be found in the corresponding supplementary files. Separate displays (figures 4–9) for each organism show the dimensions that were significantly driven both by specific phenotypic variables and by specific ribosomal protein genes (i.e. correlations greater than 0.4 in both cases). Overall, this approach might offer valuable insight about the variance in the data and reduce the bewildering complexity of ribosomal protein mutant phenotypes.

Figure 2. Schematic of the process to reduce the complexity of the observed phenotypes among ribosomal protein (RP) mutants and identify the genes that contribute the most to specific groups.

Figure 3. Scree plots for the first 20 dimensions in each species, showing the percentage of the variance explained by each dimension in each organism. For the full list of all the dimensions, see the electronic supplementary material file for each organism, in the sheets denoted ‘*_eigen’.

Figure 4. Phenotypes that show the most significant association with specific dimensions/clusters among loss-of-function ribosomal protein mutants in yeast. The ribosomal proteins that drive these groupings are indicated in each case.

Figure 5. Phenotypes that show the most significant association with specific dimensions/clusters among loss-of-function ribosomal protein mutants in worms. The ribosomal proteins that drive these groupings are indicated in each case. All the proteins shown had significant contributions (correlation coefficients > 0.4).

Figure 6. Phenotypes that show the most significant association with specific dimensions/clusters among loss-of-function ribosomal protein mutants in flies. The ribosomal proteins that drive these groupings are indicated in each case.

Figure 7. Phenotypes that show the most significant association with specific dimensions/clusters among loss-of-function ribosomal protein mutants in zebrafish. The ribosomal proteins that drive these groupings are indicated in each case. All the proteins shown had significant contributions (correlation coefficients >0.4).

Figure 8. Phenotypes that show the most significant association with specific dimensions/clusters among loss-of-function ribosomal protein mutants in mice. The ribosomal proteins that drive these groupings are indicated in each case.

Figure 9. Phenotypes that show the most significant association with specific dimensions/clusters among gain-of-function ribosomal protein mutants in yeast. The ribosomal proteins that drive these groupings are indicated in each case. All the proteins shown had significant contributions (correlation coefficients >0.4).

4.1. Saccharomyces cerevisiae

In yeast, the 111 loss-of-function phenotypes could be reduced to 11 dimensions. Together, these 11 dimensions accounted for 72% of the variance in the phenotypic variables (electronic supplementary material, file2/sheet ‘yeast_eigen’). In most of the dimensions (listed in the electronic supplementary material, file2/sheets ‘yeast_Dim’), the associated phenotypes were broadly dispersed and did not strongly associate with a given dimension (i.e. the correlation coefficients were less than 0.4). Note that the most common phenotypes in this organism (e.g. reduced fitness figure 1) are displayed in greater than 80% of the ribosomal protein mutants. Nonetheless, we noted that increased autophagy and sensitivity to pheromone were significantly related to Dimension 2 (R 2 = 0.58 see the electronic supplementary material, file2/sheet ‘yeast_Dim2’). Dimension 2 accounts for 13% of the variance among all 111 phenotypes. Increased sensitivity to pheromone probably reflects the prolonged G1 phase [23] observed in ribosomal protein mutants [24]. Autophagy is a strategy to obtain the resources necessary to sustain some degree of proliferation when nutrients are limiting, or during other stresses [25]. Hence, it is reasonable to expect increased autophagy in ribosomal protein perturbations, which may genetically mirror a nutrient-poor, stress environment.

A valuable outcome of the multiple correspondence analysis outlined above is pointing to the mutant gene that drives the grouping of the various phenotypic variables in each dimension (listed in the electronic supplementary material, file2/sheet ‘yeast_genes_cos2’). Interestingly, mutations in the ribosomal protein Asc1p (RACK1 in the unified nomenclature) drives the grouping in Dimension 2, dominated by increased autophagy and pheromone sensitivity (figure 4). Asc1p/RACK1 prevents frameshifting in paused ribosomes [26]. Ribosome pausing often occurs when the supply of amino acids is limited [27]. The inability of ribosomes to properly pause, in cells lacking RACK1, may mimic conditions that induce autophagy.

Reduced resistance to X-rays was strongly associated with Dimension 4 (R 2 = 0.69 see the electronic supplementary material, file2/sheet ‘yeast_Dim4’). This dimension only accounts for 7% of the variance among all the ribosomal protein mutant phenotypes (electronic supplementary material, file2/sheet ‘yeast_eigen’), and mutations in eL20 drove that grouping (figure 4). Interestingly, however, this contribution was paralogue-specific (electronic supplementary material, file2/sheet ‘yeast_genes_cos2’), from RPL20A (R 2 = 0.77), but not from RPL20B (R 2 = 0.0009). We will return to the issue of paralogue-specific phenotypes later.

4.2. Caenorhabditis elegans

In worms, the 151 loss-of-function phenotypes were also reduced to 11 dimensions, accounting for 76% of the variance (electronic supplementary material, file3/sheet ‘worm_eigen’). At least three of the dimensions were driven strongly by specific phenotypes (figure 5). For example, the first dimension in this metazoan organism, accounting for 22% of the variance among all the phenotypic variables, is a mix of cellular, tissue, and organismal manifestations of hypo-proliferation, including small cells and nuclei, small or absent tissues, and a narrowing of the central body axis (figure 5 electronic supplementary material, file3/sheet ‘worm_Dim1’). Lastly, the ribosomal protein genes that were most significantly associated with these phenotypic groups (figure 5), encoded mostly proteins of the large ribosomal subunit (uL13, eL28, eL43, uL4, uL23 electronic supplementary material, file3/sheet ‘worm_genes_cos2’).

4.3. Drosophila melanogaster

In flies, grouping the 43 loss-of-function phenotypes into eight dimensions explained 78% of the variance (electronic supplementary material, file4/sheet ‘fly_eigen’). As in worms, hypo-proliferative manifestations, such as defective cell growth, cell cycle, reduced fertility, or lethality, dominated the different groups (figure 6). The exception was Dimension 7, accounting for 6% of the total phenotypic variance, which was dominated by the ability of some, but not all, uL10 alleles to suppress variegation (electronic supplementary material, file4/sheet ‘fly_Dim7’). It is worth noting that cell cycle defects dominated Dimension 5, driven by a RpS2/uS5 mutant (electronic supplementary material, file4/sheet ‘fly_Dim5’).

4.4. Danio rerio

In fish, the number of phenotypic variables observed in ribosomal protein mutants expands significantly (figure 1), reflecting the added complexity of vertebrate biology. Remarkably, however, all these phenotypes could be reduced to just five dimensions, capturing 86% of the variance (electronic supplementary material, file5/sheet ‘fish_eigen’). A detailed list of the phenotypes and genes that are most significantly associated with each dimension is in the electronic supplementary material, file5. They are also summarized schematically in figure 7. The typical hypo-proliferative phenotypes displayed in the other model systems discussed so far, are also evident in fish.

Moreover, disrupted definitive haematopoiesis and defective neurocranium morphogenesis were closely associated with Dimension 5 (figure 7 electronic supplementary material, file5/sheet ‘fish_Dim5’). As discussed above, these are phenotypes also seen in human patients with Diamond-Blackfan anemias (electronic supplementary material, file7/sheet ‘human_phenotypes’). The gene driving this grouping in fish is rpl5a/uL18 (electronic supplementary material, file5/sheet ‘fish_Dim5’). Mutations in the human orthologue, RPL5/uL18, lead to Diamond-Blackfan anemia type 6. Lastly, it is worth pointing out that at the cellular level, increased autophagy was significantly associated with Dimension 3 in fish (figure 7), as was seen for one of the dimensions in yeast (figure 3). Overall, the above observations offer remarkable examples of the conservation of the phenotypic manifestations of ribosomal protein perturbations across multiple species.

4.5. Mus musculus

In mice, there are 23 reported ribosomal protein mutants, displaying an astonishing 289 distinct phenotypic variables (figure 1 electronic supplementary material, file6/sheet ‘mouse_phenotypes’). However, all these phenotypes could be grouped in just three dimensions, explaining 83% of the observed variance (electronic supplementary material, file6/sheet ‘mouse_eigen’). The phenotypes and genes that are most significantly associated with each dimension are in the electronic supplementary material, file6 and shown schematically in figure 8. As discussed above for fish and humans, skeletal abnormalities are also prominent in mouse ribosomal protein mutants.

5. Gain-of-function phenotypes of ribosomal proteins in yeast

In yeast, there are 24 reported phenotypes associated with the over-expression of 75 ribosomal protein genes (electronic supplementary material, file2/sheet ‘yeast_gof_phenotypes'). The most common phenotypes were changes in the rate of vegetative growth, which increased for 32 genes but decreased for 19 others. For four genes (RPL24B, RPL34A, RPL37B, RPS22B), there were conflicting reports that vegetative growth was either increased or decreased (electronic supplementary material, file2/sheet ‘yeast_gof_phenotypes’). The 24 phenotypes associated with the over-expression of ribosomal proteins could be grouped in three dimensions, explaining 61% of the observed variance (electronic supplementary material, file2/sheet ‘yeast_gof_eigen’). Dimension 1, accounting for 43% of the total variance, is driven by abnormal morphology and cell cycle progression in G2. Yeast displays characteristic patterns of polarized growth and budding when it proliferates, which were affected by ectopic ribosomal protein expression, especially of RPS7A/eS7 (figure 9 electronic supplementary material, file2/sheet ‘yeast_gof_genes_cos2’). Numerous genes, encoding proteins of both the large and small ribosomal subunits, contributed to Dimension 2, characterized by increased vegetative growth (figure 9). Invasive growth in yeast is also associated with polarized growth [28]. The absence of invasive growth drove the grouping in Dimension 3 (figure 9). Hence, there appears to be a general pattern of altered polarized growth when ribosomal proteins are over-expressed in yeast.

6. Over-proliferation in ribosomal protein mutants

The increased proliferation observed when at least some ribosomal proteins are over-expressed in yeast is intriguing, but also puzzling. It is not known if those effects are reflections of ribosomal output, or of some unknown, extra-ribosomal function. Even if the increased cell proliferation is associated with ribosomal functions and more protein synthesis, it is unclear how over-expression of a single component of a giant molecular machine made of many parts, could drive the formation of more such machines. However, a recent report in mice argued that over-expression of RPL15 (eL15) not only enhanced translation of other genes, including cell cycle regulators, but also promoted distant metastases in mice with breast cancer [29].

Increased proliferation and cancer have also been associated with loss-of-function ribosomal protein mutations. As discussed above, early in life ribosomopathies are consistent with hypo-proliferation, such as defective haematopoiesis in Diamond-Blackfan anemias [30]. Paradoxically, later in life, some of these patients are predisposed to cancer [30,31]. Ten per cent of primary human samples of T-cell acute lymphoblastic leukemia have loss-of-function mutations in RPL22/eL22 [32]. RPL22 mutations are also found in microsatellite-unstable colorectal [33], and endometrial cancers [33,34], at 77%, and 50% frequency, respectively. In addition, cancer-associated mutations have been described for RPL5 (uL18) [35], RPL10 (uL16) [35], RPL11 (uL5) [36], RPS15 (uS19) [37,38], RPS20 (uS10) [39] and RPS14 (uS11) [40,41]. The cancer-associated mutations in ribosomal proteins are hypomorphic ones, impairing ribosome biogenesis [30]. Even a missense R98S mutation in RPL10 (uL16) observed in T-cell leukemia, was shown to impair ribosome biogenesis and delay cell proliferation when introduced in yeast and mammalian cells [35]. Evidence that ribosomal proteins may function as haploinsufficient tumour suppressors has been reported in zebrafish [42] and flies [43–45]. However, these results do not necessarily support a direct, negative role of ribosome biogenesis in cell division. Indeed, such effects in flies were owing to cell non-autonomous routes [46–48]. Overall, in the context of their role in protein synthesis, most of the evidence suggests that the initial phenotype upon loss-of-function perturbations of ribosomal proteins is hypo-proliferative. How then could ribosomal protein perturbations account for uncontrolled cell proliferation in cancer? There are at least three possibilities, which are not exclusive of each other:

Ribosomal protein perturbations reduce the concentration of active ribosomes [49–51], which then disproportionately affects translation of specific transcripts [51,52]. Ribosomal proteins themselves may not be direct negative regulators of cell division, but in ribosomal protein mutants, translation of some mRNAs, perhaps some with tumour suppressor roles, could be repressed more so than other transcripts, setting the stage for cancer. The mathematical background for this type of regulation was articulated long ago by Lodish [53]. Briefly, the Lodish model predicts mRNA-specific effects because of the nonlinear relationship between translational efficiency and the available ribosomes. In decreasing ribosome content, e.g. upon perturbations of ribosomal proteins in ribosomopathies, mRNAs with features (e.g. secondary structure, upstream open reading frames) that impede ribosome access to the main start codon of an mRNA will have a disproportionately lower translational efficiency than other mRNAs. The proposition that mRNA-specific cases of translational control, as predicted by the Lodish model, may underpin at least some of the phenotypes in ribosomopathies [51,52], is reasonable and straightforward.

Ribosomal proteins could have extra-ribosomal, non-translational functions [54,55]. Disruption of ribosome biogenesis induces nucleolar stress because free ribosomal proteins accumulate. Loss of Rpl22 may lead to cancer in mice by activating the stress-induced NF-κB pathway, which in turn triggers the stemness factor Lin28B [32]. When ribosome assembly is disrupted, some of the released ribosomal proteins could bind other targets. For example, Rpl5, Rpl11 and Rpl23 have been reported to stabilize the p53 protein, by inhibiting the Mdm2 ubiquitin ligase that degrades p53 [55]. It is not clear, however, how this extra-ribosomal role could promote cancer, as stabilization of the p53 tumour suppressor would probably be hypo-proliferative. A recent study in human cells lacking Rps25/eS25 reported that cellular adaptation to ribosomal protein loss, rather than direct translation control, can drive phenotypes assumed to result from preferential translation [56]. In that scenario, upon eS25 loss, the cellular ribosome pool was under a stress relating to its biogenesis and turnover, eliciting a specific cellular state change, which itself drives phenotypes [56].

Lastly, impairing ribosomal proteins could alter the composition of active ribosomes [57,58]. Translation of mRNAs that rely on ‘specialized’ ribosomes has been reported, especially in neurons [59]. However, there are no examples of transcripts whose translation is carried out by ‘specialized’ ribosomes and affected in cancers owing to ribosomal protein perturbations.

Regardless of the validity of each of the above models, until recently, there was very little information about gene expression changes in ribosomal protein mutants and, specifically, about the translational efficiency of all mRNAs in those settings. Without such knowledge, it is difficult to bridge the genotype-phenotype relationship in ribosomal protein mutants mechanistically. However, in the last 2–3 years, some answers have emerged, based on recent findings from ribosome profiling in ribosomal protein mutants, which will be discussed in the next section.

7. Gene expression changes in ribosomal protein mutants

Before discussing ribosome profiling experiments in ribosomal protein mutants, it should be noted that a few changes in the expression of specific gene products in some of those mutants have been catalogued in zebrafish. These data are in the electronic supplementary material, file5/sheet ‘fish_gene_expression’. It covers the reported changes at the levels of 36 loci in three mutants (rpl11, rpl5a, rps19), which may offer some insight into the phenotypes observed. In these cases, however, how the changes in gene expression came about was not clear.

Ribosome profiling incorporates next-generation sequencing to quantify all the pieces of mRNAs bound to ribosomes [60–62]. From the accompanying RNAseq data, for each mRNA species, one can compute from the observed steady-state levels of that mRNA as a reference, if the fraction that is bound to ribosomes is higher than expected, or lower, indicating an increased, or decreased, translational efficiency, respectively. In human cells, Khajuria and colleagues mimicked a Diamond-Blackfan setting by suppressing RPS19 (eS19), RPL5 (uL18), RPS24 (eS24) and RPL11 (uL5) [51]. In all cases, haematopoietic cells had lower levels of ribosomes, but the composition of the ribosomes did not change. The consequences of RPL5 and RPS19 suppression were then analysed by ribosome profiling. Changes in transcription and translation were similar between RPL5 and RPS19 mutants, arguing that Diamond-Blackfan anemias lead to a common set of molecular changes in human haematopoietic cells. Importantly, translation of a subset of transcripts that are normally upregulated at the early stages of erythroid lineage-specification, including GATA1—which encodes a transcription factor that triggers the differentiation of immature blood cells, was disproportionately reduced when RPL5 and RPS19 were repressed [51]. Translation of mRNAs encoding ribosomal proteins was also lower in these settings [51].

A similar general conclusion that lower ribosome levels result in specific and dose-dependent changes in gene expression was also reached by an elegant study in yeast [50]. These authors analysed by ribosome profiling 14 rpl and 9 rps mutants, each lacking one of the paralogues that encode the corresponding ribosomal protein. The primary phenotypic readout used in that study was the rate of vegetative growth. Remarkably, the patterns of gene expression changes matched the growth rate of each mutant [50]. In other words, if an rpl and an rps deletion have a similar effect on the growth rate, then the associated gene expression changes would also be similar. Unlike the situation in human cells, Cheng and colleagues found that the translation of genes involved in ribosome biogenesis was increased (not decreased), especially in rps mutants [50].

Besides general effects on the growth rate, more nuanced and specific effects must also exist, for several reasons. First, the spectrum of the phenotypes observed in ribosomal protein mutants is varied and complex. Second, the growth rate is a simple, quantitative parameter, but using growth rate changes alone as a criterion to evaluate ribosomal protein phenotypes runs the risk of ‘missing the trees for the forest’. Different cellular pathways may be affected by different ribosomal protein mutants, but these different inputs may be missed if they have comparable impacts on growth rate. For example, some ribosomal protein mutants often exhibit an equivalent G1 cell cycle delay, but for different reasons [63]. At least some phenotypes strongly associated with ribosomal protein mutations do not correlate at all with dose-dependent effects on growth rate. Such an example is replicative longevity. Mutations in ribosomal proteins of the large (60S) subunit promote longevity in yeast [7,49,64,65]. The relationship between rpl mutants and longevity is complex. For example, the Rpl22 double paralogue deletion is viable, but not long-lived [7]. The single rpl22aΔ mutant is long-lived, but rpl22bΔ cells are not long-lived [7], and there is no relationship between the growth rate of rpl mutants and their longevity [66].

Yet another demonstration of the power of ribosome profiling to provide the mechanistic underpinning of translational effects and their phenotypic consequences comes from studies that examined paralogue pairs in yeast, including the Rpl22 pair [66]. The authors found a small set (less than 100) of mRNAs that were differentially translated. These mRNAs were significantly enriched for transcripts that encode enzymes of one-carbon metabolism. Metabolomic measurements supported the conclusion that one-carbon metabolism is specifically downregulated in cells lacking Rpl22Ap, but not Rpl22B, accounting for all the phenotypes of rpl22aΔ cells, including in longevity [66]. As in the previous studies mentioned above [50,51], there was no change in bulk ribosome composition in rpl22 mutants [66]. In agreement with Cheng et al. [50], compared to wild-type cells, translation of transcripts encoding other ribosomal proteins was increased in the paralogue deletants, even though overall protein synthesis was reduced [66]. It seems that yeast cells attempt to offset their reduced protein synthesis capacity by increasing the levels of individual components of the ribosome. But these efforts do not globally restore the protein synthesis defect, presumably because the production of ribosomal components is unbalanced.

8. Concluding remarks

The general picture that emerges from the detailed profiling studies is straightforward: loss-of-function ribosomal protein mutants → fewer ribosomes → lower protein synthesis → general hypo-proliferation and dose-dependent, disproportionate translational control of a subset of mRNAs. This is a broad view that corresponds very well with the most common phenotypes summarized earlier from yeast to humans (figure 1). Additional, more specific effects that are uncoupled from the growth rate can also be accounted for by translational control of relevant transcripts [66]. The stress associated with the lower ribosome pool in ribosomal protein mutants may also trigger secondary changes, leading to stress-associated phenotypes, with no direct translational basis [56]. Nonetheless, from the evidence collected thus far, it appears that the varied phenotypic landscape of ribosomal protein mutants, from the general to more peculiar phenotypes, mainly comes about from the canonical roles of ribosomal proteins in ribosomes. The profiling studies did not support additional mechanisms of specialized ribosomes with altered composition or extra-ribosomal functions, but it was also not explicitly evaluated. Hence, these conclusions need to be tested further and in more detail. Applying these methodologies to the analysis of more ribosomal protein mutants that display phenotypes of interest, will undoubtedly advance our knowledge in the relationship between genotype and phenotype in ribosomal protein perturbations, illuminating their fascinating biology and the broader roles of translational control.


Clinical Phenotype.

We studied 2 brothers with JRRP, herein identified as patient 1 (P1) and patient 2 (P2), of Belgian ancestry born to consanguineous (first-cousin) parents (Fig. 1A). There were no other siblings. P1 developed hoarseness and recurrent laryngitis at age 5 y. Direct laryngoscopy revealed papillomas in the glottis and supraglottis (Fig. 1B). He required 8 surgical ablations of laryngeal lesions over the next year and continues to require multiple ablations each year, with decreasing frequency. P2 developed hoarseness shortly after birth, and laryngeal papillomatosis was diagnosed at age 20 mo. His disease has been less severe than that of P1, and he has required 2 to 3 ablations per year. Careful retrospective review of their medical history revealed the same mild dermatologic abnormalities in both brothers, including a small number of palmar and plantar warts keratosis pilaris on the lower back, buttocks, and thighs and atrophoderma vermiculata on the face, none of which required medical treatment (Fig. 1B) see the case reports in SI Appendix for full details. Dermatologic abnormalities are not typically seen in other patients with JRRP, which is isolated therefore, these 2 patients had a syndromic form of JRRP. The parents did not have any notable medical history, specifically no history of RRP or dermatologic disease. Histologically, larynx lesions exhibited a papillomatous morphology with focal areas of koilocytosis and scattered binucleated cells, typical of lesions in RRP and pathognomonic of HPV infection (32). (Fig. 1C). Papillomas from P1 (9 specimens) and P2 (2 specimens) tested negative for HPV-6 and HPV-11, consistent with studies of cohorts in which HPV is not detected in every patient (16 ⇓ ⇓ –19).

A private homozygous missense mutation in siblings with JRRP and dermatologic abnormalities. (A) Pedigree showing NLRP1 genotype of individuals. (B) Clinical images of P1 showing (from left to right) larynx papillomas, atrophoderma vermiculata on cheeks, plantar warts, and keratosis pilaris on buttocks and thighs. (C) Micrographs of a larynx papilloma from P1 showing (from left to right) gross morphology of papillomas, areas of binucleated cells (Insets: enlarged), and focal areas of koilocytosis (arrows). (D) Schematic representation of NLRP1 protein showing functional domains, location of patients’ T755N mutations (red), and location of previously described NLRP1 mutations (blue) and their mode of inheritance (AD, AR, or codominant [CoD]). (E) Protein sequencing alignment of human NLRP1 to known orthologs, showing conservation of T755.

Genetic Analysis.

We performed whole-exome sequencing (WES) on P1, P2, and both parents (I.1 and I.2 Fig. 1A). WES showed a high homozygosity rate in P1 (3.56%) and P2 (5.13%) (33), consistent with the known parental consanguinity. Principal component analysis confirmed the patients’ European ancestry (33). In light of this consanguinity, we hypothesized that a rare variant, homozygous in both patients and heterozygous in both parents, might be responsible for the patients’ phenotype. We selected variants predicted to result in a missense, nonsense, indel, or splice site mutation with a minor allele frequency of <0.01 in public databases (ExAC, 1000 Genomes, and NHLBI-ESP6500). Finally, we excluded variants in genes with a Gene Damage Index (GDI) >13.38 (34), variants with a combined annotation-dependent depletion (CADD) score less than the mutation significance cutoff (MSC) (35), and variants in our blacklist with an in-house frequency of >0.01 (36) (SI Appendix, Fig. S1A). This yielded 5 homozygous variants across 5 genes (SI Appendix, Fig. S1B). Two of these were present in homozygosity in healthy individuals in ExAC, suggesting they are unrelated to the patient’s phenotype another was in an unknown gene (ZNF417) and another was in a gene implicated in cardiac conduction defects (KCNH2) (37). The best candidate was a homozygous missense mutation in Nucleotide-Binding Domain, Leucine-Rich Repeat Family Pyrin Domain-Containing 1 (NLRP1), c.2819C > A (for transcript variant 1 NBCI NM_033004), p.T755N (herein T755N). NLRP1 isoform 1, composed of 1,473 amino acids, acts as a sensor for the innate immune complex known as the inflammasome (38) and is expressed across a variety of tissues and cell types ( T755N is slightly N-terminal to the leucine-rich repeat (LRR) domain (Fig. 1D). Homozygosity of the NLRP1 T755N allele in P1 and P2 and its familial segregation with the disease was confirmed by Sanger sequencing (SI Appendix, Fig. S1C). Using a similar variant-filtering strategy for the less likely X-linked pattern of inheritance did not yield any candidate variants (SI Appendix, Fig. S1D). Similarly, there were no de novo mutations shared by the 2 patients. Thus, these findings suggested that homozygosity for NLRP1 T755N might be the genetic etiology of JRRP in P1 and P2.

Population Genetics of NLRP1.

The NLRP1 variant T755N is not found in any public database (gnomAD, Bravo/TOPMED) or in our in-house cohort of >5,000 unrelated individuals with a variety of infectious diseases. T755N is predicted to be damaging by CADD, with a high score of 23.1, above the 99% confidence interval MSC value of 3.313 (35). The T755 residue of NLRP1 is highly conserved across species (Fig. 1E). NLRP1 has a GDI of 9.374, indicating a medium amount of mutational burden in the general population (34), and is under low to moderate negative selection, with a McDonald–Kreitman Neutrality Index of 0.400 and a residual variation intolerance score in the 95th percentile of the least intolerant genes (39) however, previous studies have shown that autosomal recessive disease-causing genes are not under purifying selection (40). In gnomAD, there are 40 missense mutations found in homozygosity in 1 or more individuals, 23 of which have a CADD score greater than the MSC. No predicted loss-of-function (LOF) variants are found in homozygosity in gnomAD, and the pRec (probability of being intolerant of homozygous, but not heterozygous, LOF variants) is 0.95 (41). Collectively, these findings suggest that T755N is likely to be damaging to NLRP1 protein function.

NLRP1 T755N Is Gain of Function and Reduces the Threshold for Inflammasome Activation In Vitro.

Germline gain-of-function (GOF) mutations in NLRP1 have been recently discovered to cause 3 Mendelian diseases of the skin. Multiple self-healing palmoplantar carcinoma (MSPC), described in 3 families, follows an autosomal dominant (AD) pattern of inheritance, with all 3 mutations (A54T, A66V, and M77T) in the pyrin (PYR) domain (42, 43). Familial keratosis lichenoides chronica (FKLC), described in 1 family, follows a codominant pattern of inheritance with a mutation (F787_R843del) immediately N terminal to the LRR domain (Fig. 1D) (43). Autoinflammation with arthritis and dyskeratosis (AIADK), described in 2 families, follows autosomal recessive (AR) (R726W) or AD (P1214R) inheritance, with mutations in the N-terminal LRR and function-to-find (FIIND) domains, respectively (44). In vitro, MSPC and FKLC disease-causing alleles display a similar GOF magnitude despite their different modes of inheritance (AD vs. codominant) (43). Because P1 and P2 had skin abnormalities similar to those seen in FKLC, we hypothesized that NRLP1 T755N would also be GOF.

We first confirmed that both NLRP1 wild type (WT) and T755N cDNAs were expressed normally by transfecting them in HEK293T cells (Fig. 2A). Published NLRP1 GOF alleles spontaneously oligomerize and induce secretion of IL-1β in keratinocytes (43). When NLRP1 T755N was overexpressed in HEK293T cells that do not express other inflammasome components, it spontaneously oligomerized, similar to previously described NLRP1 GOF variants M77T and F787_R843del and in contrast to the WT NLRP1 (Fig. 2B). This oligomerization of T755N is partially dependent on the autocleavage site, amino acid F1212 within the FIIND domain, as the noncleavable mutation F1212A reduced the amount of oligomerized NLRP1 T755N (Fig. 2B, lane 6). Taken together, these results suggest that the T755N mutant behaves in a similar fashion biochemically with the other NLRP1 GOF mutants in causing increased inflammasome activation via spontaneous oligomerization. In addition, overexpression of NLRP1 T755N in immortalized keratinocytes led to elevated production of secreted IL-1β, similar to that in previously described GOF alleles (Fig. 2C), which was dependent on cleavage, as seen in other NLRP1 GOF alleles (Fig. 2C). The magnitude of functional gain was similar in alleles that follow AD (M77T), codominant (F787_R843del), and AR inheritance (T755N) (Fig. 2 B and C). In summary, these findings demonstrate that NLRP1 T755N can cause increased inflammasome activation in vitro, suggesting that this allele is GOF and thus probably pathogenic.

NLRP1 T755N is GOF for inflammasome activation. (A) Western blot showing similar expression of NLRP1 WT and T755N protein in HEK293T cells. A GAPDH Western blot is shown as a loading control. The image is representative of 3 independent experiments. (B) Western blot for HA-tagged NLRP1 after BN-PAGE or conventional SDS-PAGE and lysates from HEK293T cells overexpressing cDNA of NLRP1 WT (2 replicates), T755N, previously published GOF alleles (M77T and F878_R843del), or noncleavable NLRP1 T755N (T755N+F1212A), demonstrating oligomerization of T755N NLRP1 similar to other GOF mutations. GAPDH Western blot is shown as a loading control. (C) ELISA of IL-1β in supernatants of keratinocytes after transfection with NLRP1 alleles demonstrating that T755N is GOF for IL-1β production and that cleavage at the F1212 position is required for IL-1β production. NT, nontransfected cells EV, empty vector. Data are an average of 4 replicates. ***P < 0.001, 1-way ANOVA.

Primary Keratinocytes from P1 and P2 Demonstrate Spontaneous Activation of the Inflammasome.

We derived primary keratinocyte cell lines from skin biopsy specimens taken from P2 and I.1 (NLRP1 genotypes T755N/T755N and WT/T755N, respectively). NLRP1 mRNA and protein were expressed to similar levels in P2, I.1, and 3 healthy control primary keratinocyte lines (Fig. 3A), confirming that the NLRP1 T755N allele is expressed at normal levels in healthy heterozygous and patient-derived cells. We next confirmed that T755N and WT NLRP1 mRNA are expressed in keratinocytes in proportion to their genotype. Cloning of a partial NLRP1 cDNA encompassing T755 showed that in heterozygous cells from I.1, ∼50% of the transcripts were WT and ∼50% were T755N (Fig. 3B), suggesting that mRNA of the T755N NLRP1 allele is expressed at levels equal to WT. Keratinocytes from P2 and I.1 released IL-1β into the supernatants, suggesting baseline activation of the inflammasome at a functional level (Fig. 3C). In contrast, basal IL-1β release was not seen in control cells (Fig. 3C). When these cells were stimulated with Val-boroPro (talabostat, a DPP9 inhibitor shown to activate the NLRP1 inflammasome) (45, 46), control and heterozygous keratinocytes released large amounts of IL-1β, while release of IL-1β in keratinocytes from P2 was unchanged (Fig. 3C). Similar results were observed for IL-18 (Fig. 3D). The abrogation of talabostat responsiveness in keratinocytes from P2 suggests that the mechanism of GOF in this allele is due to a decrease in DPP9 inhibition. Taken together, these results demonstrate that keratinocytes homozygous for NLRP1 T755N display inflammasome activation at the basal level.

Patient-derived keratinocytes show normal expression of NLRP1 protein, baseline inflammasome activation, and unresponsiveness to further NLRP1 activation. (A) Western blot (Top) and qPCR (Bottom) of NLRP1 expression in keratinocytes from P2, heterozygous father (I.1), and 3 controls. The image is representative of 3 independent experiments. (B) Relative expression of NLRP1 WT, T755N transcripts as assessed by TA cloning and Sanger sequencing of an NLPR1 cDNA from keratinocytes from control (CTL), heterozygous father (I.1), and P2. (Inset) Numbers correspond to the number of unique clones sequenced. (C) IL-1β ELISA of supernatants from keratinocytes that were untreated or treated with 3 μM of talabostat (Val-boroPro) for 16 h. (D) IL-18 ELISA of supernatants from keratinocytes that were untreated or treated with 3 μM of talabostat for 16 h. Bars represent mean ± 1 SD. The results are representative of 3 independent experiments.

P1 and P2 Display Elevated Serum Cytokine Levels Consistent with Spontaneous Inflammasome Activation In Vivo.

We tested whether P1 and P2 had any clinical markers of spontaneous inflammasome activation. Patient serum was first tested for elevation of IL-1β and IL-18, the 2 cytokines that may be produced on activation of the inflammasome (47, 48). Both patients showed elevation of IL-18, but not of IL-1β, in repeated analyses (Fig. 4A), similar to previously reports of patients with homozygous NLRP1 GOF mutations near the LRR region of the protein (43, 44). Such a divergence of IL-18 and IL-1β elevation in the blood is also seen in other inflammasome activation disorders, such as NLRC4-mediated autoinflammation (49), and may underlie the phenotypic differences between inflammasome disorders and disease stage (50). Both P1 and P2 also showed elevated TNF-α, which is induced by IL-1β and IL-18 in many cell types and may mediate further up-regulation of inflammasome components (51), although IL-6 was not elevated (Fig. 4A), as was also seen in the previously described FKLC patient with a homozygous NLRP1 F787_R853del GOF mutation (43). Serum IL-1RA was also elevated in P1 and P2 (Fig. 4A), consistent with chronic inflammasome activation. Serum cytokine levels were not elevated in the heterozygous parents (SI Appendix, Fig. S2), consistent with the absence of any clinical manifestations. Stimulation of patient and healthy control peripheral blood mononuclear cells (PBMCs) with lipopolysaccharide (LPS) or heat-killed Listeria monocytogenes (HKLM), 2 Toll-like receptor agonists that trigger IL-1β production in an NLRP1-independent manner (52), led to similar levels of IL-1β production in patients and controls (Fig. 4B), suggesting a normal response to TLR ligands. These data demonstrate that the sera of P1 and P2 showed signs of inflammasome activation in vivo.

Inflammasome activation in P1 and P2. (A) Luminex measurements of serum levels of IL-1β, IL-18, IL-1Ra, IL-6, and TNF-α from P1, P2, and 3 healthy controls. Results are representative of 2 independent experiments. (B) ELISA for IL-1β after stimulation of PBMCs from P1, P2, and 2 controls with TLR ligands that induce IL-1β in an NLRP1-independent manner demonstrating normal regulation of production. Bars represent mean ± 1 SD. The results are representative of 2 independent experiments.

The Genetic Evaluation of a Child With Cancer

Nathaniel H. Robin MD , Anna C.E. Hurst MD, MS , in Pediatric Cancer Genetics , 2018

The Genetics Physical Examination

The genetics physical examination has several notable differences from the typical medical examination. The genetic examination focuses on identifying the subtle physical findings that represent clues to the underlying genetic syndrome, discussed below. This approach is termed “dysmorphology,” which is the study of abnormal form with an emphasis on structural developmental abnormalities. A dysmorphological evaluation of a child (or fetus or adult) looks for unusual physical (or behavioral) characteristics that might provide insight into errors in embryologic or fetal development, major or minor.

A geneticist’s thorough physical examination includes measurement of multiple structures and observational assessment for dysmorphic findings. When possible, measurements should be obtained and plotted against standardized growth charts. References such as the Handbook of Physical Measurements 5 contain descriptions of the methods by which to obtain precise measurements and accompanying growth charts.

The use of precise terminology is recommended in the description and documentation of the physical examination. Updated preferred terminology can be found in a collection of published articles, “Elements of Morphology: Human Malformation Terminology,” which were intended to develop accurate and clear definitions for the terms of the craniofacies, hands, and feet.

For geneticists, the first “vital signs” are growth parameters, including a careful assessment of length, weight, and head circumference. Undergrowth or overgrowth can be not just global, but regional growth differences are also associated with many cancer risks. Macrocephaly is an important marker of many genetic syndromes (such as PTEN-related disorders or nevoid basal cell carcinoma syndrome), and microcephaly is seen in many chromosomal deletion syndromes that can affect regions containing cancer predisposition genes.

The evaluation should include full skin examination assessing for congenital or acquired hyperpigmentation or hypopigmentation, growths (such as lipomas), and telangiectasias. The distribution of pigmentary changes should be closely examined, as streaking hyperpigmentation is a sign of tissue mosaicism, indicating genetic differences that may not be identified on routine blood testing. The hair, nails, and teeth should also be closely examined as part of the examination, and/or families should be asked about growth patterns or any irregularities that they have noted. For example, several dark spots around the mouth may not seem to be important in a teenager with newly diagnosed colon cancer, but they suggest the diagnosis of Peutz-Jeghers syndrome.

The remaining examination is best proceeded from head to toe, including the face and orbits, ears, nose, mouth, and jaw. Eye examination should include periorbital features (spacing of eyes, palpebral fissure size, and slanting) and ocular structures (pupils and fundus, if able). Care should be taken to assess for symmetry and the gestalt appearance of how each region “fits” in the overall appearance of the face. Measurements can be extremely useful in determining if a structure is objectively big or small. A common adage among clinical geneticists is “don’t say it’s big or small without measuring it first,” as a structure may appear small or large but merely be out of proportion to other structures.

Chest and abdominal examinations should include auscultation, as some congenital heart defects may not have been noted prior. (The cardiac auscultation is also a time when young patients in the room become quiet and still, and the careful examiner can use this time to silently observe the face in close detail.)

Genitourinary examinations can also reveal signs of hypogonadism or even dermatologic changes such as freckling of the glans penis as seen in PTEN-related disorders.

The extremities and musculoskeletal system exam can reveal obvious anomalies, such as absent or triphalangeal thumbs, or more subtle findings, such as broad digits, clinodactyly, or asymmetric palmar creases. Limb asymmetry may be hemihyperplasia, which could indicate a systemic syndrome or tissue mosaicism.

A full neurologic examination is also an important tool to assess for subtle deficits, although it is important to know the child’s prior baseline and what deficits may have been acquired during any cancer management or treatment (such as postsurgical changes).

DNA Polymorphisms: Meaning and Classes | Genetics

In this article we will discuss about the meaning an classes of DNA polymorphisms.

Meaning of DNA Polymorphisms:

Different alleles of a gene produce different phenotypes which can be detected by making crosses between parents with different alleles of two or more genes. Then by determining recombinants in the progeny, a genetic map can be deduced.

These are low resolution genetic maps that contain genes with observable phenotypic effects, all mapped to their respective loci. The position of a specific gene, or locus can be found from the map. However, measurements showed that the chromosomal intervals between the mapped genes would contain vast amounts of DNA.

These intervals could not be mapped by the recombinant progeny method because there were no markers in those intervening regions. It became necessary to find additional differential markers or genetic differences that fall in the gaps. This need was met by exploitation of various polymorphic DNA markers.

A DNA polymorphism is a DNA sequence variation that is not associated with any observable phenotypic variation, and can exist anywhere in the genome, not necessarily in a gene. Polymorphism means one of two or more alternative forms (alleles) of a chromosomal region that either has a different nucleotide sequence, or it has variable numbers of tandemly repeated nucleotides.

Thus, it is a site of heterozygosity for any sequence variation. Many DNA polymorphisms are useful for genetic mapping studies, hence they are referred to as DNA markers. DNA markers can be detected on Southern blot hybridisation or by PCR.

The alleles of DNA markers are co-dominant, that is they are neither dominant nor recessive as observed in alleles of most genes. DNA polymorphisms constitute molecularly defined differences between individual human beings.

Classes of DNA Polymorphisms:

There are some major classes of DNA polymorphisms.

1. Single Nucleotide Polymorphisms:

SNP is a single base pair change, a point mutation, and the site is referred to as SNP locus. SNPs are the most common type of DNA polymorphism, occurring with a frequency of one in 350 base pairs, and accounting for more than 90 per cent of DNA sequence variation. The majority of SNPs are found to be present in the non-coding regions of the genome, known as non-coding SNPs. SNPs in the coding regions, that is within genes, are known as coding SNPs (cSNPs).

Detailed studies of cSNPs in humans indicate that each gene has about four cSNPs, half of which resulting in missense mutations in the encoded protein, and half of which produce silent mutations. Whether a cSNP affects a phenotype, depends on the amino acid that is changed by the polymorphism.

About one-half of missense mutations that are SNPs are estimated to cause genetic disease in humans. A non-coding SNP can also affect gene function if it is located in the promoter region or in the gene regulatory region. A small number of SNPs can create a restriction site, or eliminate an already existing restriction site. SNP-induced alterations in restriction sites are detected by using the restriction enzyme followed by Southern blot analysis or PCR.

An individual SNP locus can be analysed by using the technique of allele-specific oligonucleotide (ASO) hybridisation. The search for one particular SNP locus in humans is a challenge, because this is one base pair that is polymorphic out of the three billion base pairs in the human genome.

In the ASO technique, a short oligonucleotide that is complementary to one SNP allele is synthesised and mixed with the target DNA. Hybridisation is performed under high stringency conditions that would allow only a perfect match between probe and the target DNA. That means, the oligonucleotide will not hybridize with target DNA that has any other SNP allele at that locus. Positive result of hybridisation indicates the SNP locus precisely.

A more recent technique of DNA Microarrays can be used for simultaneous typing of hundreds or thousands of SNPs. Details of this technique used for SNPs and genome wide gene expression are described later in this section.

A small number of SNPs can lead to changes in restriction sites either by creating a restriction site or eliminating one. Such SNPs can be detected by using the restriction enzyme for the site, and detection is done by Southern blot analysis or PCR. The different patterns of restriction sites in different genomes yield fragments of different lengths, called restriction fragment length polymorphisms (RFLPs) described below.

2. Restriction Fragment Length Polymorphisms:

RFLPs are restriction enzyme recognition sites that are present in some genomes and absent in others. Consider an organism heterozygous for an RFLP whose genotype we represent as Rr. This organism is backcrossed with another that is homozygous for the RFLP variation allele (rr). Genomic DNA from the progeny of this cross (Rr x rr gives progeny of which 50% is Rr and 50% is rr) is subjected to restriction enzyme digestion, and fragments separated on Southern blots.

The restriction fragments obtained are hybridised with a probe (a cloned DNA fragment) that will distinguish the various genotypes for an RFLP. The probe DNA is unique because it comes from only one DNA segment of the genome and that overlaps the restriction site. A key point of this technique, therefore, is the use of a specific cloned single-copy DNA probe that is specific for an individual marker locus.

Crosses between the positive RFLP organism with other RFLP bearing organisms would yield parental combinations and re-combinations. From the frequency of recombinants, a detailed RFLP map can be produced. RFLPs were the first DNA markers that were in use for characterisation of plant and animal genomes. They have now been replaced by markers based on variation in the number of short tandem repeats (STRs) described below.

3. Short Tandem Repeats:

STRs are also known as microsatellites and simple sequence repeats (SSRs). A tandem repeat is a sequence that is repeated end to end in the same orientation. STRs are 2 to 6 base pair DNA sequences tandemly repeated a few times.

For example, the sequence TCACATCACATCACATCACATCACA is a five-fold repeat of the sequence TCACA. There are dinucleotide, trinucleotide, four-nucleotide, five-nucleotide and six-nucleotide STRs in the human genome.

Microsatellite analysis can be done using a single-copy DNA to serve as a PCR primer pair specific for each marker locus. In contrast with RFLPs that have only one or two alleles in a population, STRs have a much larger number of alleles which can be detected in a population analysis.

Consequently, STRs have a higher proportion of heterozygotes which makes them more suitable for mapping purposes. Polymorphisms in STRs is common in populations which makes them valuable tools in genetic mapping.

4. Variable Number Tandem Repeats:

VNTRs, also called minisatellite markers, the repeat unit is a little larger than in STRs, from seven to a few tens of base pairs long. The VNTR loci in humans are 1 to 5 kilo-base sequences containing repeat units about 15 to 100 nucleotides long. VNTR loci also show polymorphisms. Due to the greater length of VNTR repeats that makes PCR unsuitable, analysis of VNTRs relies on restriction digestion and Southern blotting.

The entire genomic DNA is cut with a restriction enzyme which cuts on either side of the VNTR locus, but does not have a target site within the VNTR arrays, followed by Southern blotting. The VNTR specific probe against a particular repeat sequence of the VNTR locus, will bind at all locations of the repeat sequence in the genome, resulting in a large number of different sized fragments.

The number of tandem repeats is variable from one individual to the other, therefore Southern blot provides a distinct distinguishing pattern of fragments for a single individual. These patterns are also referred to as DNA fingerprints. The technique finds useful application in identification of individuals and in deciding parentage.

5. Microsatellite Markers:

Variable numbers of di-nucleotides repeated in tandem, called microsatellite markers, are dispersed in the genome. The most common type are CA and the complementary GT repeats. Probes are designed for detection of DNA regions surrounding individual microsatellite repeats by using PCR.

The procedure is explained by taking the example of human DNA as follows. Human genomic DNA is subjected to restriction digestion by an enzyme such as Alu l, that will result in fragments about 400 base pairs in length. The fragments are cloned into a vector and Southern blotting is carried out.

To identify genomic inserts that contain CA/GT di-nucleotides, probes specific for these di-nucleotides are used. Sequence of the positive clones is determined, on the basis of which PCR primers are designed that will hybridise with single-copy DNA sequences flanking the specific tandemly repeated microsatellite sequences. PCR amplification is carried out using these primer pairs and genomic DNA.

Thus, if any size variation exists in the stretch of tandemly repeated microsatellite sequence, it would be detected through gel electrophoresis of the DNAs from different individuals. The size variations may differ among the different individuals, all these variations could be determined. A size variation results in amplification product of a different size and represents a marker allele.

6. Randomly Amplified Polymorphic DNA:

RAPDs are based on random PCR amplification. The procedure is carried out by randomly designing primers for PCR which will amplify several different regions of the genome by chance. Such a primer results in amplification of only those DNA regions that have near them, inverted copies of the primer’s own sequence.

The PCR products consist of DNA bands representing different sizes of the amplified DNA. The set of amplified DNA fragments is called randomly amplified polymorphic DNA (RAPD). Certain bands may be unique for an individual and can serve as DNA markers in mapping analysis.


Yeast Strains, Plasmids, and Antibodies

Yeast strains are listed in Table1. pRS305-SIR2, an integrating plasmid that contains sir2 driven by its native promoter, was used. Mutant SIR2 genes were also cloned into these vectors.SIR2 and mutant sir2 strains were generated by cutting pRS305-SIR2 within the LEU2 gene at anAflII site and integrated using standard yeast transformation protocols. SIR2 or mutant sir2cloned into the pET28a vector was used for the production of recombinant protein. The hemagglutinin (HA) tagging of SIR4was done with the pSF323-SIR4–3XHA vector (a gift from Steve Bell), which integrates a tagged version of SIR4 into the native SIR4 locus. Rabbit antibody to Sir2p and Sir3p have been previously described (Mills et al., 1999 Imaiet al., 2000). The 12CA5 antibody to the HA epitope was obtained from Covance and the acetylated histone H3 and acetylated histone H4 were obtained from Upstate Biotechnology (Lake Placid, NY).

Site-directed Mutagenesis of SIR2

Site-directed mutants were generated in pRS305-SIR2 as per Imaiet al. (2000) and subcloned into pET-28a (Imai et al., 2000). The mutations were sequenced to ensure that the mutagenesis was successful. Expression of Sir2p in yeast was monitored by Western blot analysis of whole cell extracts probed with anti-Sir2p antibody.

Purification of Recombinant Protein and Enzymatic Assays

Six-his–tagged recombinant Sir2p and mutant Sir2p were purified from BL21 bacteria that overexpressed the gene on a pET28a plasmid as described previously (Imai et al., 2000). ADP-ribosylation of histones was detected as described previously (Imai et al., 2000). Histone deacetylation activity was measured using a peptide corresponding to the N-terminal tail of histone H4 (SGRGKGGKGLGKGGAKRHRC) labeled with tritiated acetate by using the histone deacetylase assay kit from Upstate Biotechnology. The assay was performed by incubating 2 μg of recombinant protein with the labeled peptide in 1 mM NAD + overnight. Ethyl acetate was then used to separate acetyl groups freed by the reaction from those still bound to the peptide. Deaceytlation activity was then measured by counting the free tritiated acetate in a scintillation counter. Histone deacetylation assays were also measured by performing the reaction for 1 h and running the reaction products on a high-performance liquid chromatography (HPLC) as previously described (Imai et al., 2000).

Silencing and rDNA Recombination Assays

To test silencing at the telomeres, 10-fold dilutions of the derivatives of W303RT were spotted on media containing 5-fluoroorotic acid (5-FOA). To assay for HM silencing, W303R derivatives were patched onto YPD with the tester strain CKy20 and after overnight growth were replica plated to minimal media with no supplemented amino acids. rDNA recombination rates were measured as in Kaeberlein et al.(1999).

Immunoprecipitation of HA-Sir4 and Sir2

Whole cell extracts were prepared from cells grown in 100 ml of YPD to an OD of 1.0 (Strahl-Bolsinger et al., 1997). Extract (200 μl) was diluted to 500 μl with lysis buffer to which 3 μl of anti-HA antibody was added and incubated at 4°C overnight. Protein A beads were then added and further incubated at 4°C for 1 h. The beads were washed three times with lysis buffer and then boiled in 60 μl of SDS running buffer. Ten microliters was run on a 7.5% PAGE gel for Western blotting analysis.

Chromatin Immunoprecipitation

Yeast were grown in 100 ml of YPD to and OD of 1.0. Immunoprecipitation of cross-linked extract was performed essentially as described (Strahl-Bolsinger et al., 1997), by using 2.5 μl of anti-SIR3 polyclonal antibody or 5.0 μl of anti-SIR2 polyclonal, anti-acetylated histone H3 antibody, or anti-aceytlated histone H4 antibody. Polymerase chain reaction (PCR) analysis of immunoprecipitated DNA was performed in 50-μl reaction volumes by using 1:25, 1:75, and 1:225 of the total immunoprecipitated DNA. PCR reaction conditions were as described using the following primers: TEL-300.fwd, GGATATGTCAAAATTGGATACGCTTATG TEL-300.rev, CTATAGTTGATTATAGATCCTCAATGATC TEL-3000.fwd, TGATTCTGCTTTATCTACTTGCGTTTC TEL-3000.rev, AGAGTAACCATAGCTATTTACAATAGG XV-internal2.fwd, GTAGTTCGTTAGGTATGGACATTGATTTGGCC and XV-internal2.rev, AAATGAA-ATGTATTGGGGCCTAGGTTCGCA. Slot blot analysis was performed by blotting 10 μl of immunoprecipitated (IP) DNA or 5 μl of input DNA to a Zeta-Probe membrane by using aBio-Rad slot blot apparatus. The blot was then probed with a 32 P-labeled DNA fragment corresponding to the 5S rDNA sequence.

Frequency-dependent Selection

Figure 2. A yellow-throated side-blotched lizard is smaller than either the blue-throated or orange-throated males and appears a bit like the females of the species, allowing it to sneak copulations. (credit: “tinyfroglet”/Flickr)

Another type of selection, called frequency-dependent selection, favors phenotypes that are either common (positive frequency-dependent selection) or rare (negative frequency-dependent selection). An interesting example of this type of selection is seen in a unique group of lizards of the Pacific Northwest. Male common side-blotched lizards come in three throat-color patterns: orange, blue, and yellow. Each of these forms has a different reproductive strategy: orange males are the strongest and can fight other males for access to their females blue males are medium-sized and form strong pair bonds with their mates and yellow males (Figure 2) are the smallest, and look a bit like females, which allows them to sneak copulations. Like a game of rock-paper-scissors, orange beats blue, blue beats yellow, and yellow beats orange in the competition for females. That is, the big, strong orange males can fight off the blue males to mate with the blue’s pair-bonded females, the blue males are successful at guarding their mates against yellow sneaker males, and the yellow males can sneak copulations from the potential mates of the large, polygynous orange males.

In this scenario, orange males will be favored by natural selection when the population is dominated by blue males, blue males will thrive when the population is mostly yellow males, and yellow males will be selected for when orange males are the most populous. As a result, populations of side-blotched lizards cycle in the distribution of these phenotypes—in one generation, orange might be predominant, and then yellow males will begin to rise in frequency. Once yellow males make up a majority of the population, blue males will be selected for. Finally, when blue males become common, orange males will once again be favored.

Negative frequency-dependent selection serves to increase the population’s genetic variance by selecting for rare phenotypes, whereas positive frequency-dependent selection usually decreases genetic variance by selecting for common phenotypes.

5. RNAi feeding on agar plates

Julie Ahringer, The Gurdon Institute, University of Cambridge, Cambridge CB2 1QN, UK

Synopsis: Grow RNAi bacteria and seed plates. Feed worms with RNAi bacteria and score for phenotypes. The following is based on the protocol in Kamath et al. (2001).

5.1. Preparing feeding plates

Pour plates: make standard NGM agar and add carbenicillin to 25 μg/ml and IPTG to 1mM just prior to before pouring. Pour plates 4𔃅 days before seeding, to allow them to dry. If plates are too wet, the bacteria won't dry after seeding and RNAi phenotypes will be weaker. Feeding can conducted using any format plates (e.g., single plates, 6-well, 12-well).

Spot individual desired bacterial strain(s) from glycerol stock to an LB plate containing 50 μg/ml ampicillin (or 25 μg/ml carbenicillin) and 10 μg/ml tetracycline. Use a 96-pin replicator to spot on a rectangular flat plate if growing in 96-well format. Grow overnight at 37°C.

Grow cultures in LB medium containing 50 μg/ml ampicillin. If using 96-well format, add 800 μl medium to each well of a 96-well deep well plate. To inoculate the cultures, use individual yellow tips or tips in a multichannel pipettor to scrape bacteria from a row or column and eject tips into the correct row or column of medium. When finished innoculating, remove the tips and cover the plates with plastic microtitre lids. Grow cultures with shaking at 300rpm for 6𔃆 hours.

Seed NGM agar feeding plates the bacterial culture. Use two 30 μl drops if using 12-well plates, and three 50 μl drops if using 6-well or individual plates. Let dry and induce overnight at room temperature.

5.2. Preparing the worms

Grow desired worm strain on standard NGM plates seeded with OP50 bacteria. Carry out standard bleaching/washing protocol to obtain embryos, and leave to hatch into L1s overnight in M9 buffer. These starved L1s will be synchronized at the beginning of the L1 stage. If feeding will be done with larvae older than L1, then put hatched L1s onto standard NGM plates containing OP50 and grow to the desired stage.

Wash worms off plates using M9 buffer, then wash 3X to remove bacteria. It is critical to remove OP50 as residual non-RNAi bacteria will interfere with the feeding results. Resuspend final worm pellet in M9 buffer containing 0.1% Tween-20 to prevent them from sticking to plastic. Adjust the volume of buffer so that the number of worms you want to aliquot per plate is in 10󈝻 μl.

5.3. Feeding and scoring

This part of the protocol will differ slightly depending on your assay. After feeding, either the aliquoted worms or their progeny can be scored. For some assays, scoring is easier if progeny are synchronized. In this case, fed gravid mothers are allowed to lay eggs on a new plate, then removed, and the progeny subsequently scored. This step is time consuming and not always necessary.

5.3.1. Standard L3/L4 feeding protocol: scoring of synchronized progeny

In this protocol, a semi-synchronized population of progeny laid in a 24-hour window are scored.

Aliquot 10󈝻 μl of L3/L4 worms per plate or well (10󈞀 worms).

Leave 72 hours at 15°C (or 36󈞔 hours at 22°C) for RNAi to take effect, then replica plate single adults onto other plates or wells seeded with the same bacteria.

After 24 hours, remove the adults from the replica and score the progeny for phenotypes at appropriate time points.

5.3.2. Streamlined L3/L4 feeding protocol: scoring of asynchronous progeny

In this protocol, all the progeny laid by fed mothers are scored. It has the benefit of being quick as there is no replica plating involved. Progeny laid early and late in the feeding protocol are in a single well, producing a range of RNAi knockdowns, from weak to strong. This can be helpful if screening for post-embryonic phenotypes, where a strong knockdown might cause embryonic lethality. Low percentage embryonic lethality is difficult to score using this method.

Aliquot 10󈝻 μl of L3/L4 worms per plate or well. As the adults and all the progeny will remain in this initial plate or well, it is important not to have too many worms for the food available. The optimal number should be determined empirically. For 6-well plates, using 10 worms per well should allow scoring of adult progeny in most cases.

Score when the progeny reach the desired age.

5.3.3. Feeding L1s instead of L4s

L1s can be used instead of L4s in either of the above protocols. An advantage of using L1s is that some phenotypes can be scored in the fed worms instead of the progeny, allowing an easily scored synchronized population to be used. However for some genes, inherited maternal product will be sufficient for gene activity, preventing induction of a phenotype in the fed worms. Also, as many genes are required at multiple times in development, different phenotypes may be seen when using L1s compared to L4s. For example, RNAi of some genes induces sterility of the fed L1s whereas L4 feeding induces embryonic lethality of the progeny. This will preclude scoring of progeny for these genes if L1s are fed. In contrast, using L1s is beneficial if the assay is for any form of lethality (e.g., sterility, larval lethality, or embryonic lethality).

Notes: Aliquoting bleached embryos directly onto feeding plates instead of hatching them into starved L1s first is not recommended because it is more difficult to aliquot the same number of animals/well due to embryos sticking together. Also, variable hatching rates between bleached preparations will cause inter-experiment variation.

5.4. Feeding two bacterial strains at the same time

The protocol is identical to feeding a single bacterial strain except that the two bacterial cultures are mixed prior to seeding. Double feeding works much better and more reproducibly in an RNAi supersensitive strain (e.g., rrf-3 Simmer et al., 2002 eri-1 Kennedy et al., 2004 or eri-1 lin-15B Wang et al., 2005 S. Woods, D. Rivers, and J. Ahringer, unpublished A. Fraser, pers. comm.). In a test of 17 control double feedings, 4 were successful in eri-1 , 7 in rrf-3 , and 14 in eri-1 lin-15B (S. Woods, D. Rivers, and J. Ahringer, unpublished). Double feeding is less reliable than single feeding, so in some cases, only one gene may be significantly inhibited, or both genes may be only slightly knocked down. Controls should be carried out to test for knockdown of each gene.

5.5. Acknowledgements

I thank members of the lab for comments. This is a modification of the protocol developed by Ravi Kamath (Kamath et al., 2001), with modifications from Gino Poulin, David Rivers, and Shane Woods.


† Present address: Department of Biological Sciences, Xi'an Jiaotong-Liverpool University, 111 Ren'ai Road, Suzhou 215123, People's Republic of China

Electronic supplementary material is available online at

Published by the Royal Society under the terms of the Creative Commons Attribution License, which permits unrestricted use, provided the original author and source are credited.


. 2011 Why don't we get more cancer? A proposed role of the microenvironment in restraining cancer progression . Nat. Med. 17, 320-329. (doi:10.1038/nm.2328) Crossref, PubMed, ISI, Google Scholar

Fridman WH, Zitvogel L, Sautes-Fridman C, Kroemer G

. 2017 The immune contexture in cancer prognosis and treatment . Nat. Rev. Clin. Oncol. 14, 717. (doi:10.1038/nrclinonc.2017.101) Crossref, PubMed, ISI, Google Scholar

Vermeulen L, Morrissey E, van der Heijden M, Nicholson AM, Sottoriva A, Buczacki S, Kemp R, Tavaré S, Winton DJ

. 2013 Defining stem cell dynamics in models of intestinal tumor initiation . Science 342, 995-998. (doi:10.1126/science.1243148) Crossref, PubMed, ISI, Google Scholar

. 1889 The distribution of secondary growths in cancer of the breast . Lancet 133, 571-573. (doi:10.1016/S0140-6736(00)49915-0) Crossref, Google Scholar

. 1976 The clonal evolution of tumor cell populations . Science 194, 23-28. (doi:10.1126/science.959840) Crossref, PubMed, ISI, Google Scholar

. 1993 Mutations, evolutionary theory and cancer . Trends Ecol. Evol. 8, 107-110. (doi:10.1016/0169-5347(93)90062-T) Crossref, PubMed, ISI, Google Scholar

. 2001 The microenvironment of the tumour–host interface . Nature 411, 375-379. (doi:10.1038/35077241) Crossref, PubMed, ISI, Google Scholar

Merlo LM, Pepper JW, Reid BJ, Maley CC

. 2006 Cancer as an evolutionary and ecological process . Nat. Rev. Cancer 6, 924-935. (doi:10.1038/nrc2013) Crossref, PubMed, ISI, Google Scholar

Korolev KS, Xavier JB, Gore J

. 2014 Turning ecology and evolution against cancer . Nat. Rev. Cancer 14, 371-380. (doi:10.1038/nrc3712) Crossref, PubMed, ISI, Google Scholar

. 2005 Evolutionary biology of cancer . Trends Ecol. Evol. 20, 545-552. (doi:10.1016/j.tree.2005.07.007) Crossref, PubMed, ISI, Google Scholar

. 1993 The multistep nature of cancer . Trends Genet. 9, 138-141. (doi:10.1016/0168-9525(93)90209-Z) Crossref, PubMed, ISI, Google Scholar

. 2005 The genetic theory of adaptation: a brief history . Nat. Rev. Genet. 6, 119-127. (doi:10.1038/nrg1523) Crossref, PubMed, ISI, Google Scholar

. 2008 A microenvironmental model of carcinogenesis . Nat. Rev. Cancer 8, 56-61. (doi:10.1038/nrc2255) Crossref, PubMed, ISI, Google Scholar

Junttila MR, de Sauvage FJ

. 2013 Influence of tumour micro-environment heterogeneity on therapeutic response . Nature 501, 346-354. (doi:10.1038/nature12626) Crossref, PubMed, ISI, Google Scholar

. 2013 Microenvironmental regulation of tumor progression and metastasis . Nat. Med. 19, 1423-1437. (doi:10.1038/nm.3394) Crossref, PubMed, ISI, Google Scholar

. 2016 Targeting the microenvironment in advanced colorectal cancer . Trends Cancer 2, 495-504. (doi:10.1016/j.trecan.2016.08.001) Crossref, PubMed, ISI, Google Scholar

2017 Heterogeneous tumor-immune microenvironments among differentially growing metastases in an ovarian cancer patient . Cell 170, 927-938.e20. (doi:10.1016/j.cell.2017.07.025) Crossref, PubMed, ISI, Google Scholar

2018 Interfaces of malignant and immunologic clonal dynamics in ovarian cancer . Cell 173, 1755-1769.e22. (doi:10.1016/j.cell.2018.03.073) Crossref, PubMed, ISI, Google Scholar

. 2011 Hallmarks of cancer: the next generation . Cell 144, 646-674. (doi:10.1016/j.cell.2011.02.013) Crossref, PubMed, ISI, Google Scholar

Lan C, Heindl A, Huang X, Xi S, Banerjee S, Liu J, Yuan Y

. 2015 Quantitative histology analysis of the ovarian tumour microenvironment . Sci. Rep. 5, 16317. (doi:10.1038/srep16317) Crossref, PubMed, ISI, Google Scholar

2015 Microenvironment-induced PTEN loss by exosomal microRNA primes brain metastasis outgrowth . Nature 527, 100-104. (doi:10.1038/nature15376) Crossref, PubMed, ISI, Google Scholar

2016 T-cell acute leukaemia exhibits dynamic interactions with bone marrow microenvironments . Nature 538, 518-522. (doi:10.1038/nature19801) Crossref, PubMed, ISI, Google Scholar

. 2016 Characteristics and significance of the pre-metastatic niche . Cancer Cell 30, 668-681. (doi:10.1016/j.ccell.2016.09.011) Crossref, PubMed, ISI, Google Scholar

Dewhirst MW, Cao Y, Moeller B

. 2008 Cycling hypoxia and free radicals regulate angiogenesis and radiotherapy response . Nat. Rev. Cancer 8, 425-437. (doi:10.1038/nrc2397) Crossref, PubMed, ISI, Google Scholar

. 2016 Hypoxic control of metastasis . Science 352, 175-180. (doi:10.1126/science.aaf4405) Crossref, PubMed, ISI, Google Scholar

Altrock PM, Liu LL, Michor F

. 2015 The mathematics of cancer: integrating quantitative models . Nat. Rev. Cancer 15, 730-745. (doi:10.1038/nrc4029) Crossref, PubMed, ISI, Google Scholar

. 1997 Modelling the consequences of interactions between tumour cells . Br. J. Cancer 75, 157-160. (doi:10.1038/bjc.1997.26) Crossref, PubMed, ISI, Google Scholar

Waclaw B, Bozic I, Pittman ME, Hruban RH, Vogelstein B, Nowak MA

. 2015 A spatial model predicts that dispersal and cell turnover limit intratumour heterogeneity . Nature 525, 261-264. (doi:10.1038/nature14971) Crossref, PubMed, ISI, Google Scholar

2015 A Big Bang model of human colorectal tumor growth . Nat. Genet. 47, 209-216. (doi:10.1038/ng.3214) Crossref, PubMed, ISI, Google Scholar

Anderson AR, Weaver AM, Cummings PT, Quaranta V

. 2006 Tumor morphology and phenotypic evolution driven by selective pressure from the microenvironment . Cell 127, 905-915. (doi:10.1016/j.cell.2006.09.042) Crossref, PubMed, ISI, Google Scholar

Lloyd MC, Cunningham JJ, Bui MM, Gillies RJ, Brown JS, Gatenby RA

. 2016 Darwinian dynamics of intratumoral heterogeneity: not solely random mutations but also variable environmental selection forces . Cancer Res. 76, 3136-3144. (doi:10.1158/0008-5472.CAN-15-2962) Crossref, PubMed, ISI, Google Scholar

. 1930 The genetical theory of natural selection . Oxford, UK : The Clarendon Press . Crossref, Google Scholar

. 2005 Fisher's microscope and Haldane's ellipse . Am. Nat. 166, 447-457. (doi:10.1086/444404) Crossref, PubMed, ISI, Google Scholar

2015 Aberrant epithelial GREM1 expression initiates colonic tumorigenesis from cells outside the stem cell niche . Nat. Med. 21, 62-70. (doi:10.1038/nm.3750) Crossref, PubMed, ISI, Google Scholar

Castro-Giner F, Ratcliffe P, Tomlinson I

. 2015 The mini-driver model of polygenic cancer evolution . Nat. Rev. Cancer 15, 680-685. (doi:10.1038/nrc3999) Crossref, PubMed, ISI, Google Scholar

Barnett GC, West CM, Dunning AM, Elliott RM, Coles CE, Pharoah PD, Burnet NG

. 2009 Normal tissue reactions to radiotherapy: towards tailoring treatment dose by genotype . Nat. Rev. Cancer 9, 134-142. (doi:10.1038/nrc2587) Crossref, PubMed, ISI, Google Scholar

Temko D, Tomlinson IPM, Severini S, Schuster-Bockler B, Graham TA

. 2018 The effects of mutational processes and selection on driver mutations across cancer types . Nat. Commun. 9, 1857. (doi:10.1038/s41467-018-04208-6) Crossref, PubMed, ISI, Google Scholar

. 2013 How ageing processes influence cancer . Nat. Rev. Cancer 13, 357-365. (doi:10.1038/nrc3497) Crossref, PubMed, ISI, Google Scholar

Reeves MQ, Kandyba E, Harris S, Del Rosario R, Balmain A

. 2018 Multicolour lineage tracing reveals clonal dynamics of squamous carcinoma evolution from initiation to metastasis . Nat. Cell Biol. 20, 699-709. (doi:10.1038/s41556-018-0109-0) Crossref, PubMed, ISI, Google Scholar

. 2019 Critical behavior of spatial networks as a model of paracrine signaling in tumorigenesis . Appl. Netw. Sci. 4, 47. (doi:10.1007/s41109-019-0167-7) Crossref, Google Scholar

. 2001 Bose–Einstein condensation in complex networks . Phys. Rev. Lett. 86, 5632-5635. (doi:10.1103/PhysRevLett.86.5632) Crossref, PubMed, ISI, Google Scholar

. 1963 Analysis of the life-cycle in mammalian cells . Nature 198, 359-361. (doi:10.1038/198359a0) Crossref, ISI, Google Scholar

. 2019 Circadian clocks and cancer: timekeeping governs cellular metabolism . Trends Endocrinol. Metab. 30, 445-458. (doi:10.1016/j.tem.2019.05.001) Crossref, PubMed, ISI, Google Scholar

. 2019 Interplay between Circadian clock and cancer: new frontiers for cancer treatment . Trends Cancer 5, 475-494. (doi:10.1016/j.trecan.2019.07.002) Crossref, PubMed, ISI, Google Scholar

Gatenby RA, Silva AS, Gillies RJ, Frieden BR

. 2004 The anti-angiogenic basis of metronomic chemotherapy . Nat. Rev. Cancer 4, 423-436. (doi:10.1038/nrc1369) Crossref, PubMed, ISI, Google Scholar

Zhang J, Cunningham JJ, Brown JS, Gatenby RA

. 2017 Integrating evolutionary dynamics into treatment of metastatic castrate-resistant prostate cancer . Nat. Commun. 8, 1816. (doi:10.1038/s41467-017-01968-5) Crossref, PubMed, ISI, Google Scholar

. 2005 Therapeutic targeting of the tumor microenvironment . Cancer Cell 7, 513-520. (doi:10.1016/j.ccr.2005.05.024) Crossref, PubMed, ISI, Google Scholar

Pich O, Muinos F, Lolkema MP, Steeghs N, Gonzalez-Perez A, Lopez-Bigas N

. 2019 The mutational footprints of cancer therapies . Nat. Genet. 51, 1732-1740. (doi:10.1038/s41588-019-0525-5) Crossref, PubMed, ISI, Google Scholar

Williams MJ, Werner B, Heide T, Curtis C, Barnes CP, Sottoriva A, Graham TA

. 2018 Quantification of subclonal selection in cancer from bulk sequencing data . Nat. Genet. 50, 895-903. (doi:10.1038/s41588-018-0128-6) Crossref, PubMed, ISI, Google Scholar

. 2012 Evolution of adaptive phenotypic traits without positive Darwinian selection . Heredity 108, 347-353. (doi:10.1038/hdy.2011.97) Crossref, PubMed, ISI, Google Scholar

. 1932 The roles of mutation, inbreeding, crossbreeding, and selection in evolution . Proc. Sixth Int. Congr. Genet. 1, 356-366. Google Scholar

. 1932 The causes of evolution , New York, NY : Longmans, Green and Co . Google Scholar

. 2006 A general multivariate extension of Fisher's geometrical model and the distribution of mutation fitness effects across species . Evolution 60, 893-907. (doi:10.1111/j.0014-3820.2006.tb01169.x) Crossref, PubMed, ISI, Google Scholar

Matuszewski S, Hermisson J, Kopp M

. 2014 Fisher's geometric model with a moving optimum . Evolution. 68, 2571-2588. (doi:10.1111/evo.12465) Crossref, PubMed, ISI, Google Scholar

. 2014 The utility of Fisher's geometric model in evolutionary genetics . Annu. Rev. Ecol. Evol. Syst. 45, 179-201. (doi:10.1146/annurev-ecolsys-120213-091846) Crossref, PubMed, ISI, Google Scholar

Fraisse C, Gunnarsson PA, Roze D, Bierne N, Welch JJ

. 2016 The genetics of speciation: insights from Fisher's geometric model . Evolution 70, 1450-1464. (doi:10.1111/evo.12968) Crossref, PubMed, ISI, Google Scholar

Lourenco JM, Glemin S, Galtier N

. 2013 The rate of molecular adaptation in a changing environment . Mol. Biol. Evol. 30, 1292-1301. (doi:10.1093/molbev/mst026) Crossref, PubMed, ISI, Google Scholar

Razeto-Barry P, Diaz J, Vasquez RA

. 2012 The nearly neutral and selection theories of molecular evolution under the fisher geometrical framework: substitution rate, population size, and complexity . Genetics 191, 523-534. (doi:10.1534/genetics.112.138628) Crossref, PubMed, ISI, Google Scholar

Blanquart F, Achaz G, Bataillon T, Tenaillon O

. 2014 Properties of selected mutations and genotypic landscapes under Fisher's geometric model . Evolution 68, 3537-3554. (doi:10.1111/evo.12545) Crossref, PubMed, ISI, Google Scholar

. 1999 Selection, the mutation rate and cancer: ensuring that the tail does not wag the dog . Nat. Med. 5, 11-12. (doi:10.1038/4687) Crossref, PubMed, ISI, Google Scholar

. 1999 Sex and adaptation in a changing environment . Genetics 153, 1041-1053. PubMed, ISI, Google Scholar

. 2009 The genetic basis of phenotypic adaptation II: the distribution of adaptive substitutions in the moving optimum model . Genetics 183, 1453-1476. (doi:10.1534/genetics.109.106195) Crossref, PubMed, ISI, Google Scholar

. 2012 Fisher's geometrical model of fitness landscape and variance in fitness within a changing environment . Evolution 66, 2350-2368. (doi:10.1111/j.1558-5646.2012.01610.x) Crossref, PubMed, ISI, Google Scholar

. 2006 Fisher's geometrical model of evolutionary adaptation—beyond spherical geometry . J Theor. Biol. 241, 887-895. (doi:10.1016/j.jtbi.2006.01.024) Crossref, PubMed, ISI, Google Scholar

Sottoriva A, Spiteri I, Shibata D, Curtis C, Tavare S

. 2013 Single-molecule genomic data delineate patient-specific tumor profiles and cancer stem cell organization . Cancer Res. 73, 41-49. (doi:10.1158/0008-5472.CAN-12-2273) Crossref, PubMed, ISI, Google Scholar

Jones AG, Arnold SJ, Burger R

. 2004 Evolution and stability of the G-matrix on a landscape with a moving optimum . Evolution 58, 1639-1654. (doi:10.1111/j.0014-3820.2004.tb00450.x) Crossref, PubMed, ISI, Google Scholar

. 1993 Directional selection and the evolution of sex and recombination . Genet. Res. 61, 205-224. (doi:10.1017/S0016672300031372) Crossref, PubMed, ISI, Google Scholar

Elena SF, Cooper VS, Lenski RE

. 1996 Punctuated evolution caused by selection of rare beneficial mutations . Science 272, 1802-1804. (doi:10.1126/science.272.5269.1802) Crossref, PubMed, ISI, Google Scholar

2009 Hypersensitivity to contact inhibition provides a clue to cancer resistance of naked mole-rat . Proc. Natl Acad. Sci. USA 106, 19 352-19 357. (doi:10.1073/pnas.0905252106) Crossref, ISI, Google Scholar

Sulak M, Fong L, Mika K, Chigurupati S, Yon L, Mongan NP, Emes RD, Lynch VJ

. 2016 TP53 copy number expansion is associated with the evolution of increased body size and an enhanced DNA damage response in elephants . eLife 5, e11994. (doi:10.7554/eLife.11994) Crossref, PubMed, ISI, Google Scholar

2010 Accumulation of driver and passenger mutations during tumor progression . Proc. Natl Acad. Sci. USA 107, 18 545-18 550. (doi:10.1073/pnas.1010978107) Crossref, ISI, Google Scholar

. 2009 The genetic basis of phenotypic adaptation I: fixation of beneficial mutations in the moving optimum model . Genetics 182, 233-249. (doi:10.1534/genetics.108.099820) Crossref, PubMed, ISI, Google Scholar

. 2015 The fitness effect of mutations across environments: Fisher's geometrical model with multiple optima . Evolution 69, 1433-1447. (doi:10.1111/evo.12671) Crossref, PubMed, ISI, Google Scholar

2004 Unifying the epidemiological and evolutionary dynamics of pathogens . Science 303, 327-332. (doi:10.1126/science.1090727) Crossref, PubMed, ISI, Google Scholar

. 2005 On statistical tests of phylogenetic tree imbalance: the Sackin and other indices revisited . Math. Biosci. 195, 141-153. (doi:10.1016/j.mbs.2005.03.003) Crossref, PubMed, ISI, Google Scholar

. 2015 Measuring asymmetry in time-stamped phylogenies . PLoS Comput. Biol. 11, e1004312. (doi:10.1371/journal.pcbi.1004312) Crossref, PubMed, ISI, Google Scholar

Martin G, Elena SF, Lenormand T

. 2007 Distributions of epistasis in microbes fit predictions from a fitness landscape model . Nat. Genet. 39, 555-560. (doi:10.1038/ng1998) Crossref, PubMed, ISI, Google Scholar

. 2010 Colloquium papers: adaptive landscapes and protein evolution . Proc. Natl Acad. Sci. USA 107(Suppl 1), 1747-1751. (doi:10.1073/pnas.0906192106) Crossref, PubMed, ISI, Google Scholar

Szendro IG, Schenk MF, Franke J, Krug J, de Visser JAGM

. 2013 Quantitative analyses of empirical fitness landscapes . J. Stat. Mech.: Theory Exp. 2013, P01005. (doi:10.1088/1742-5468/2013/01/P01005) Crossref, ISI, Google Scholar