In this section we focus on genes, and the repetition of gene segments and duplication of whole genes, forming gene families. DNA sequence repetition is generally rare in the small genomes of prokaryotes and mitochondria (which are typically tightly packed with genetic information that is presented in extremely economical forms). But in the large nuclear genomes of eukaryotes, and especially in complex multicellular organisms, there is less constraint in packing the genetic information, and repetitive DNA sequences are a striking feature of these genomes, both in terms of abundance and importance. Many human genes, both protein-coding genes and RNA genes, are members of multigene families that can vary enormously in terms of copy number and distribution. many non-genic sequences are repeated. As detailed below, repeated DNA sequences arise by one or more of a variety of different mechanisms that result in gene duplication.
Human genes show enormous variation in size and internal organization
Genes in simple organisms such as bacteria are comparatively similar in size and are usually very short (typically about 1 kb long). The genes of complex eukaryotes, however, can show huge size variation. As measured by the distance from the beginning of the first exon to the end of the last exon, the longest human protein-coding gene is CNTNAP2, which at 2.305 Mb just pips the dystrophin gene, the previous record holder at 2.242 Mb. Using the same definition of gene length, the mean length of a human protein-coding gene is 64.9 kb (but the median length is 26.0 kb), but there is enormous diversity as shown in Table 1. Although there is generally a direct correlation between gene and product sizes, there are some striking anomalies. For example, the 2.24 Mb dystrophin gene, 50 times the length of the apolipoprotein B gene, makes a protein that is almost 900 amino acids smaller than apolipoprotein B.
Table1. HUMAN PROTEIN-CODING GENES SHOW MARKED VARIATION IN SIZE AND ORGANIZATION
For human protein-coding genes, the mean exon size is 268 nucleotides (but median size is 129 nucleotides), with an average of nine exons per gene. As genes get larger, exon size remains fairly constant. Note, however, that the terminal exon can occasionally be very long; if we discount the first and last exons (which quite often are mostly untranslated sequence), the mean length of an internal exon in a human protein-coding gene is 147 nucleotides (median length 121 nucleotides). Unlike exons, the introns of human protein-coding genes vary enormously in size (from a few nucleotides to over 100,000 nucleotides long), and intron size generally correlates with gene length (but there are some notable exceptions—see Table 1).
Human protein-coding genes generally contain introns, but a small minority lack introns and are generally small genes (see the footnote to Table 1 for some examples). For those that do possess introns, there is an inverse correlation between gene size and fraction of coding DNA because large genes tend to have large introns while exons tend to be more uniform in size. Because transcribing long introns is costly in both energy and time (16 hours are needed to transcribe the 2.24 Mb dystrophin gene), very highly expressed genes often have short introns, or no introns at all.
Sizes and intron content of RNA genes
Many RNA genes have been identified very recently, and statistics on RNA gene size and intron content are being accumulated. Genes encoding lncRNAs can be tens of kilo bases long and often contain introns (the exons of these genes show a mean length of 683 nucleotides, but a median length of 129 nucleotides). At the transcript level, lncRNA genes show less diversity (a mean of 1.8 transcripts per gene) than protein-coding genes (close to 7 transcripts per gene). At the other end of the scale are tiny RNA genes, including those specifying miRNAs that often lack introns. However, as described above, there is some uncertainty as to how to define an RNA gene.
Variable gene density and overlapping transcription units in the nuclear genome
Simple genomes have high gene densities (roughly one per 0.5, 1, and 2 kb for the genomes of human mitochondria, Escherichia coli, and Saccharomyces cerevisiae, respectively) and often show examples of partly overlapping genes. Different reading frames may be used, sometimes from the same sense strand. In complex organisms, such as humans, genes are much bigger and there is less clustering of protein-coding sequences.
Gene density varies enormously from chromosome to chromosome. The constitutive heterochromatin regions of somatic cells are almost devoid of genes (but a few genes in these regions may be expressed in germ-line cells). Within the euchromatic portion of the genome, gene density can also vary substantially between chromosomal regions, and also between whole chromosomes.
The first general insight into how genes are distributed across the human genome was obtained when purified CpG island fractions were hybridized to metaphase chromosomes (CpG islands have long been known to be strongly associated with genes). The results indicated that gene density is high in subtelomeric regions, and that chromosomes with high %GC (for example, chromosomes 19 and 22) are gene-rich, whereas others with low % GC (for example, X and 18) are gene-poor. The predictions of differential CpG island density and differential gene density were subsequently confirmed by analyzing the human genome sequence.
This difference in gene density can also be seen with Giemsa staining (G-banding) of chromosomes. Regions with a low %GC content correlate with the darkest G-bands (G-positive bands), and those with a high %GC content correlate with pale bands. Pale G-negative bands are comparatively rich in genes; dark G-positive bands are gene-poor. In chromosomal regions with high gene density, overlapping genes may be found; they are typically transcribed from opposing DNA strands. For example, the class III region of the HLA complex at 6p21.3, located within a G-negative band, has an average gene density of about one gene per 15 kb and is known to contain several examples of partly overlapping genes. In striking contrast, the mammoth dystrophin gene extends over 2.24 Mb of DNA in a dark G-band at Xp21.2 without evidence for any other protein-coding gene in this region.
Genes-within-genes
Whole-genome analyses show that about 9% of human protein-coding genes overlap another such gene, and >90% of the overlaps involve genes transcribed from opposing strands. Sometimes the overlaps are partial, but in other cases small genes are located within larger genes. The NF1 (neurofibromatosis type I) gene, for example, has three small internal genes transcribed from promoters on the opposite strand (Figure 1A).
Fig1. Different types of organization for genes-within-genes. (A) Small protein-coding genes can be located on the antisense strand of a larger gene. The 60.5 kb intron 27b of the NF1 (neurofibromatosis type I) gene contains three small internal genes, all transcribed from the opposite DNA strand. The internal genes, each with two exons and not drawn to scale, are OGMP (oligodendrocyte myelin glycoprotein) and EVI2A and EVI2B (human homologs of murine genes thought to be involved in leukemogenesis and located at ecotropic viral integration sites). (B) Small RNA genes can be located on the sense strand of a larger gene. Intron 2 of the HTR2C gene (which encodes the 5-hydroxytryptamine receptor 2C protein) contains one snoRNA gene and four miRNA genes, each of which is co-transcribed as part of the primary HTR2C transcript, and then processed from the excised intron 2 RNA sequence.
Small RNA genes are also frequently found within both protein-coding genes and long non coding RNA genes, usually within introns. These genes are often located on the sense strand of the larger gene (usually described as a host gene), and their expression is dictated by the host gene’s promoter. For example, almost all snoRNAs, and many miRNAs and scaRNAs, are made by processing of the intron transcripts of larger host genes (see Figure 1B for an example).
Different origins for duplicated genes in the human genome
Gene duplication has been an important driver in the evolution of functional complexity and the origin of increasingly complex organisms. As a result, duplicated genes (and duplicated segments of coding sequence) are a common feature of animal genomes, especially large vertebrate genomes. In the human genome, the resulting multigene families have from two to many hundreds of gene copies that may be clustered together in one subchromosomal location, or dispersed over several chromosomal locations—we consider different types of gene-family organization below.
One type of DNA duplication arose at the birth of eukaryotes when the fusion of two cells gave rise to two genomes that would become the nuclear and mitochondrial genomes. A few of the original gene duplicates were retained. Subsequently, a variety of different DNA duplication mechanisms have led to more recent gene duplication and the formation of gene families, as listed below.
• Whole-genome duplication (WGD). The most recent WGDs in mammalian lineages appear to have occurred during the early evolution of chordates. A couple of WGD events occurring within the approximate interval of 525–875 million years ago can explain why there are four vertebrate HOX gene clusters, but many of the duplicated genes were not retained after whole-genome duplication.
• Tandem duplication. Crossover can occasionally occur between unequally aligned chromatids, either on homologous chromosomes (unequal crossover) or on the same chromosome (unequal sister chromatid exchange). The repeated segment may be just a few hundred nucleotides long or may be quite large and contain from one to several genes. Initial duplication events produce tandem repeats (that is, the head of one repeat joins the tail of its neighbor: → →). Figure 2 shows the general mechanism of tandem gene duplication. Over a long (evolutionary) timescale the duplicated sequences can be separated on the same chromosome (by a DNA insertion or inversion).
• Duplicative transposition by recombination. Certain euchromatic regions, notably those close to the centromeres and telomeres (pericentromeric and subtelomeric regions, respectively), are comparatively unstable, and are prone to recombination with other chromosomes. As a result, segments of DNA containing multiple genes can be transposed to another location (but with copies of the transposed genes retained on nonrecombining homologs). We give the examples of the NFI and PKD1 multigene families below.
• RNA-directed duplicative transposition. Reverse transcriptases within the cell are used to make a cDNA copy of an RNA transcript, whereupon the cDNA copy integrates into a new chromosomal location. The same type of mechanism can often lead to defective gene copies, and will be described below.
Fig2. Tandem gene duplication arising by unequal crossover. Homologous chromatids 1 and 2 have paired out of register, so that gene A on one chromatid is not directly opposite gene A on its partner chromatid. As shown here, the mispairing of the chromatids might sometimes be stabilized by closely related members of an interspersed repeated DNA family such as Alu repeats (small orange boxes). A crossover event (red X) in the mispaired region can result in tandem duplication of gene A, by joining of the left part of chromatid 2 to the right part of chromatid 1. (The other product of the crossover, formed by joining of the left part of chromatid 1 to the right part of chromatid 2, would lack gene A). A crossover event like this may result from unequal crossover (chromatids 1 and 2 are misaligned non-sister chromatids on homologous chromosomes) or from unequal sister chromatid exchange (chromatids 1 and 2 are misaligned sister chromatids).
Segmental duplication
More than 5% of the euchromatic portion of the human genome is accounted for by 400 large blocks of duplicated DNA where the sequence identity between the DNA copies is very high (often more than 95% identity). The very high degree of sequence identity across such long sequences is the result of evolutionarily recent DNA duplication, occur ring within the past 40 million years of primate evolution, and this type of recent duplication is often described as segmental duplication. It includes both intrachromosomal duplications (see Figure 3 for an example) and also interchromosomal duplications. Segmental duplications are important contributors to copy number variation and to chromosomal re-arrangements leading to disease and rapid gene innovation.
Fig3. An example of intrachromosomal segmental duplication in the human genome. Shown at the left are locations on human chromosome 16 of five types of chromosome 16 low-copy-number repeats (LCR16). The repeats vary in size from 19 kb (LCR16a) to 75 kb (LCR16t), and in copy number from 3 repeats (LCR16t, LCR16v) to over 20 repeats (LCR16a). Each type of repeat unit has a single counterpart in the baboon genome (where the LCR16t and LCR16a are neighboring sequences, as are the LCR16u and LCR16w repeats). Since divergence of the baboon lineage from lineages leading to humans and the great apes, approximately 30 million years ago, there has been a rapid series of duplications of the five sequences in the human and great ape genomes. In the human genome, the five LCR16 repeat families are believed to have arisen by duplication of single repeats originally located at 16p13.11, 16p12.3, and 16q24.2. Note that there has also been an expansion of these sequences in the chimpanzee and gorilla, but there are species differences regarding both the copy number of the repeats and also their chromosomal locations. (Adapted from Marques-Bonet T & Eichler EE [2009] Cold Spring Harb Symp Quant Biol 74:355–362; PMID 19717539. With permission from Cold Spring Harbor Laboratory Press.)
Functionally inactive copies of gene sequences
Gene duplication is driven by evolutionary pressure to develop specialized functions. We owe our developed sense of smell to having a family of close to 1000 olfactory receptor genes that collectively provide a large range of different olfactory receptor proteins. However, duplication is not always successful in producing additional functional genes. Some duplicated genes degenerate and become nonfunctional gene copies called pseudogenes (in the case of the olfactory receptor gene family, for example, the majority of the sequences are inactive—see Box 9.2 for an overview of human pseudogenes).
Repetitive functional sequences within genes
Having genes split into exons and introns facilitates tandem exon duplication using mechanisms such as unequal crossover. Tandem repetition of sequences encoding known or assumed protein domains is quite common, and it may be functionally advantageous by providing a more available biological target. Sometimes that leads to a series of repeated exons (or group of exons) that specify a protein domain or other functional unit.
Although the sequence identities between the repeated protein domains are often quite low, they can sometimes be high. A classic example is provided by tandemly repeated kringle domains present in the lipoprotein Lp (a) protein encoded by the LPA gene on chromosome 6q26. These domains, which are especially common in blood clotting and fibrinolytic proteins, are formed as multiple loop structures constrained by three disulfide bridges, and get their name from a Scandinavian pastry of the same shape. Each kringle domain in the Lp (a) protein is about 114 amino acids long and is encoded by a pair of exons; tandem repetition of the exon pair has produced a large number of almost identical kringle repeats (Figure 4). The high level of sequence identity between the two-exon repeats promotes unequal crossover and striking length polymorphism.
Fig4. Tandem repetition of kringle domains in lipoprotein Lp (a) as a result of exon duplication. The LPA gene that encodes lipoprotein Lp (a) is implicated in coronary heart disease and stroke. The Lp (a) protein shows extensive size variation, having variable numbers of tandem 114-amino-acid repeats representing kringle IV domains. For simplicity we show just the first nine kringle domains, which are identical in sequence and span amino acids 17 to 1042, being encoded by exon pairs 2+3, 4+5, 6+7, and so on, up to 18+19. The first amino acid encoded by each exon is always an alanine (shown in bold red or bold blue), and the codon for each such alanine is interrupted by an intron. There are additional, subsequent kringle domains (not shown, but up to as many as 32 copies) and they can show some limited sequence variation.
Different organizations for protein-coding gene families
Protein-coding genes that duplicated recently in evolution have a strong tendency to be clustered at a subchromosomal location and produce proteins with clear structural and functional similarity. By contrast, genes that make proteins with little sequence similarity are typically found at distinct chromosomal locations, and usually on different chromosomes, even if they work very closely together in the same functional pathway, or as part of a common protein. The insulin gene is located on chromosome 11 but the insulin receptor gene is on chromosome 19, for example, and genes encoding the five subunits of RNA polymerase I are located at 2p, 2q, 6p, 9p, and 13q.
Different classes and subclasses of gene family can be distinguished according to the degree of sequence similarity of the protein products, the gene copy number, the locations of gene members in the genome, and whether duplication events have occurred by making DNA copies of genomic DNA or mRNA. Many protein-coding gene families have pseudogenes. If the genes are clustered at a subchromosomal region, pseudogene copies are of the nonprocessed type. A large-copy-number gene family dispersed on multiple chromosomes is often the result of retrotransposition, and may contain many processed pseudogenes. Some dispersed gene families arise by alternative methods, such as through recombination; in that case, nonprocessed pseudogenes are evident. Different classes of gene family are outlined in Table 2 and in Figures 5 and 6.
Fig5. The class I HLA gene family at 6p21.3 has multiple nonprocessed pseudogenes and gene fragments. (A) Structure of a class I HLA heavy-chain mRNA. The full-length mRNA contains a polypeptide-encoding sequence specifying a leader sequence (L), three extracellular domains (α1 , α2 , α3 ,), a transmembrane sequence (TM), a cytoplasmic tail (CY), and a 3′ untranslated region (3′ UTR). The three extracellular domains are each encoded essentially by a single exon. The very small 5′ UTR is not shown. (B) Labeled items are: six functional genes (blue boxes; A–C, E–G); four full-length nonprocessed pseudogenes (Ψ); and a variable number of partial gene copies (unfilled boxes), seven in this example, some truncated at the 5’ end (1,3,5,6), some truncated at the 3′ end (7), and some containing single exons (2 and 4).
Fig6. Dispersal of nonprocessed NF1 and PKD1 pseudogenes as a result of pericentromeric or subtelomeric instability. (A) The 58-exon NF1 neurofibromatosis type I gene is located close to the centromere of human chromosome 17—exons and introns are shown as, respectively, thin vertical boxes and chevrons (^). Highly homologous defective copies of the NF1 gene are found at nine or more other genome locations, mostly in pericentromeric regions. The pseudogene copies have segments of the full-length gene, with both exons and introns. Seven examples are shown here, such as two copies on 15p that have intact genomic sequences spanning exons 13 and 27b. Sometimes, rearrangements have caused deletion of exons and introns (shown by asterisks). (B) The 48-exon PKD1 polycystic kidney disease gene is located close to the telomere of 16p. As a result of segmental duplication events during primate evolution (see Figure 9.7), large components of this gene have been duplicated and six PKD1 pseudogenes (Ψ) are located at 16p13.11, with large blocks of sequence (shown as boxes) copied from the PKD1 gene (asterisks represent the absence of counterparts to PKD1 sequences).
Table2. DIFFERENT CLASSES OF HUMAN GENE FAMILY ACCORDING TO CHROMOSOMAL DISTRIBUTION AND WHETHER SEQUENCES WERE COPIED AT THE DNA LEVEL OR AT THE RNA LEVEL
Organization and distribution of RNA gene families
Long noncoding RNAs are highly heterogeneous, but even if we consider long regulatory RNAs, the examples that we know are divergent in sequence; families of evolutionarily related genes are not so obvious. Genes encoding small noncoding RNAs are a different matter. There are several well-recognized gene families with moderately large numbers of genes, often several hundred genes, that specify RNAs of a particular class with conserved structures.
Some genes present at high copy number are used to make large amounts of essen tial noncoding RNAs. Take the example of genes making the four cytoplasmic ribo somal RNAs. The 28S, 5.8S, and 18S rRNAs are encoded by a single 13 kb multigenic (28S+5.8S+18S) transcription unit that, along with a 27 kb spacer region, is tandemly repeated to form large ribosomal DNA (rDNA) arrays on the short arms of each of the five human acrocentric chromosomes: 13, 14, 15, 21, and 22. These rDNA arrays (known as nucleolar organizer regions because they associate to form a chromosomal structure around which a nucleolus forms), vary significantly in repeat copy number. A haploid human genome may have around 400 rDNA repeats (of which only 20–50% may be transcriptionally active in most cells), but individual nucleolar organizing regions can have from as low as 1-3 copies up to 140 rDNA repeats (spanning 5.6 Mb). Because the sequence arrays are in close proximity, there are frequent sequence exchanges between arrays on different acrocentric chromosomes as well as between arrays on homologous chromosomes or sister chromatids. The sequence exchanges (unequal crossover or unequal sister chromatid exchange) cause the individual sequences to be almost identical. Tandemly duplicated 5S rRNA genes are clustered at 1p42 and can also exchange sequences by unequal crossover, but in this family there are also many dispersed pseudogenes.
As well as the mtDNA sequences specifying mitochondrial tRNAs, there are close to 600 members of the human tRNA gene family dispersed over several chromosomes (including multiple pseudogenes), but there are some large clusters, notably one at 6p22.1.
Gene families making short guide RNAs that work in RNA maturation or regulation
In addition to genes that make tRNAs and the short ribosomal RNAs, some other moderately large gene families make short noncoding RNAs that serve to modify RNAs during RNA maturation or to regulate gene expression at the RNA level. In each case, the RNA works as a guide RNA that binds to RNA target sequences (rather than ones in the genome). Like other guide RNAs, they have binding sites for executive proteins that they transport to their target sequences to carry out an action, in this case a desired RNA modification. They include the RNAs listed below (with the number of human genes specifying them shown in brackets–gene details can be found by querying individual gene families at https://www.genenames.org/cgi-bin/genefamilies/).
• Small nuclear RNAs (snRNAs) are needed for RNA splicing (37 genes).
• Small nucleolar RNAs chemically modify rRNA at specific nucleotide sites (537 genes). There are two subfamilies: C/D box snoRNAs guide site-specific 2′-O-ribose methylations, with 105–107 varieties of this methylation in rRNA (355 genes; the SNORD family); H/ACA snoRNAs guide site-specific pseudouridylations, where uridine is isomerized to give pseudouridine at 95 different positions in the pre-rRNA (182 genes; the SNORA family).
• Small Cajal body RNAs (scaRNAs) chemically modify snRNAs (27 genes).
• MicroRNAs (miRNAs) regulate the expression of certain target genes by binding to the mRNAs they produce (1776 genes).
The genes specifying these RNAs are dispersed across the genome, but are occasionally grouped in small clusters, notably in the case of snoRNA genes. Note that most snoRNA genes are located within the introns of larger genes transcribed by RNA polymerase II; in these cases, snoRNAs are produced by processing of the intronic RNA, and so regulation of snoRNA synthesis is coupled to that of the host gene. The recent burgeoning interest in RNA has led to establishment of a wide range of RNA databases (see Table 3 for examples).
Table3. EXAMPLES OF RNA DATABASES
RNA pseudogenes
Some RNA gene families have large numbers of pseudogenes—see Table 4 for some examples. Whereas all protein-coding genes are transcribed by RNA polymerase II using an upstream promoter, some genes specifying some types of small noncoding RNAs are transcribed by RNA polymerase III, in which case transcription often occurs from an internal promoter. That is, the promoter is located within the transcribed DNA. (That is possible because the job of the promoter is to be a landing ground for a protein complex that binds to the promoter and then binds other proteins whose job is to position the RNA polymerase at the required position upstream. By the time the RNA polymerase is correctly positioned upstream, the protein complex that was bound to the promoter has been removed, leaving the way clear for transcription.)
Table4. EXAMPLES OF RNA GENE FAMILIES HAVING LARGE NUMBERS OF PSEUDOGENES
Genes that have an internal promoter—including tRNAs, some snRNA genes, and some other types of RNA gene—are prone to making large numbers of copies by retrotransposition. That happens because the transcripts carry a copy of the promoter sequence and when they are converted into cDNA and then integrate into the genome elsewhere, the retrocopies can be transcribed from their promoter to make more transcripts, and so on. the Alu repeat, the most common repeat in the human genome, and the mouse B1 repeat originated, independently, by retrotransposition of 7SL RNA transcripts prior to the primate–rodent split about 90 million years ago. Other highly repetitive human and mouse DNA sequences arose by retrotransposition of tRNAs.
