Homogenization of Chromosomes Revealed by Oligonucleotide-Stickiness
Ayumu SAITO and Koichi NISHIGAKI
The genome can be considered to be a chronicle of the evolutionary history of any organism. By scrutinizing the genome, we are able to follow the events that occurred during evolution. Some of these events must have had a positive influence on the evolution of the organism, whereas other events had no effect and so are considered 'neutral' . Extensive gene-level analyses of homolog and analog proteins have elucidated the phylogenetic relationships between genes and organisms . This well-established approach can reveal the frequencies of point mutations, deletions, and insertions that occurred in genes of interest and allow the determination of evolutionary rate, genetic distance, and other variables. On the other hand, as a system of genes, the genome should be dealt with as a whole. Comparisons of genome sequences have recently disclosed a massive scale of gene arrangements that were caused by recombination events [3 - 5]. Furthermore, repetitive element analysis  and segmental duplication analyses [7, 8] of the human genome unveiled that they probably generated the current genome structures [6 - 8]. Thus, the genome is less static and less stable than previously believed, making it more important to understand these dynamic genome rearrangements as a whole. Recently, a measure called oligostickiness was introduced to characterize genomes [9, 10] and was useful for investigating the degree of recombination between chromosomes. Here, we applied this methodological tool to determine the similarity of chromosomes contained in the nucleus and we found that recombination phenomena occurred with high frequency.
2. 1 Calculation of Oligostickiness
Oligostickiness, s, was calculated as explained in the legend to Figure 1. It can be expressed as follows:
where l0 + 1 and n are the genome sequence positions at which the sampling region begins and the sampling size for oligostickiness, respectively, and d is a determinant that takes the value of 1 when the primer (p) binds stably to the i-th local sequence of the genome (T(i)), in other words, a fragmental sequence that has a fixed 5'-end at the sequence position i, or the value of 0 when not bound. In this formula, the p - T binding is determined based on the thermodynamic stability of the p - T complex under given conditions (25°C, 1M NaCl) . Figure 2 shows the stickiness of 12 oligonucleotides to Escherichia coli genome DNA in the mode of a spider-web chart . In this mode, the oligostickiness of each oligonucleotide is expressed as a value between 0 (the center) and 1 (the outermost circle) and represents the frequency of binding of the oligonucleotide to the genome DNA (where 1 means that the oligonucleotide can bind at any site of the genome DNA with required or higher stability and zero means no binding throughout). The oligonucleotides were arbitrarily selected except with respect to the following considerations: i) oligonucleotides with different properties (G+C contents, sequence complexities, and thermodynamic stabilities) were selected as much as possible and ii) representative oligonucleotides from the viewpoint of oligostickiness based on data collected from over 20 genomes were selected .
Figure 1. How to calculate oligostickiness. Oligostickiness is defined as the normalized frequency of binding to template DNA in various ways as presented here (also see Methods). Each binding structure indicates sufficiently stable binding to template DNA at a particular site. The stability of each structure is calculated thermodynamically  and shown with more stable structures placed on a lower layer. Frequency of probe-binding is accumulated within a sector of the angle , normalized by actual size of fractional template, and represented as a pillar (or spike), with the height proportional to the normalized frequency. For convenience, oligostickiness is usually defined regarding registered genome sequence (or database sequence).
In addition, the obtained results are available at the following URL. http://gp.fms.saitama-u.ac.jp/stickiness/
2. 2 Spider-web representation
The representation of chromosome/genome properties used here is called the 'spider-web presentation of global oligostickiness' . Each global oligostickiness value with respect to a chromosome probed by a particular oligonucleotide was plotted on an axis radially extended from a common center, following a logarithmic scale (Figure 2). In this paper 12 axes per round were adopted with 12 different oligonucleotide probes. The nearby plots were connected with a line to define a characteristic pattern for each chromosome. This type of representation appears to be more effective in presenting a feature of a chromosome in depth than reducing it to a single figure (or value).
Figure 2. Spider-web chart of oligostickiness of Escherichia coli genome. Probes were P1~P12 (oligodeoxyribonucleotides written from 5' to 3'): P1, dGGGGTCGAGGGG; P2, dTGGGTGGGTGGG; P3, dGAGAGAGAGAGA; P4, dGCTAAAAAAAAA; P5, dAAAAAAAAAAAA; P6, dATATATATATAT; P7, dGTGCTGGGATTA; P8, dCCAGGCTGGTCT; P9, dCCGGCCGGCCGG; P10, dGGGGTCGAGGCG; P11, dAGACCGCGCCTG; P12, dACGACGACGACG. Oligostickiness values are plotted on the radial axes.
2. 3 Genome sequences
The genome sequences were retrieved from databases published in relation to genome sequencing projects about each genome. The source of A. thaliana chromosome sequences was the Kazusa DNA Research Institute (CONTIGs: Chr1, pseudo v211200b; Chr3, MEC18 T15D2). The E. coli genome sequence was from the National Institute of Genetics and H. sapiens chromosome sequences were from the National Center for Biotechnology Information (CONTIGs were as follows: chromosomes Chr1, NT_004359.5/ NT_004873.5; Chr2, NT_005194.5/ NT_005204.5; Chr3, NT_005718.5/ NT_005787.5/ NT_005997.5/ NT_006022.5/ NT_022517.5; Chr4, NT_006051.5/ NT_006052.5/ NT_006098.5/ NT_006316.5; Chr5, NT_006455.5/ NT_006547.5/ NT_006576.5/ NT_023115.5; Chr6, NT_027049.2/ NT_029309.1; Chr7, NT_007867.5/ NT_007918.5; Chr8, NT_007988.5/ NT_008157.5/ NT_008227.5; Chr9, NT_008484.5/ NT_023967.5; Chr10, NT_008682.5/ NT_008895.5; Chr11, NT_009107.5/ NT_009243.5/ NT_009325.5/ NT_009368.5; Chr12, NT_009711.5/ NT_009775.5; Chr13, NT_009799.5; Chr14, NT_010101.5; Chr15, NT_010178.5/ NT_010310.5/ NT_010351.5/ NT_024680.5; Chr16, NT_010530.5/ NT_010552.5/ NT_024827.5/ NT_027182.2/ NT_029459.1; Chr17, NT_010641.5/ NT_010783.5; Chr18, NT_010874.5/ NT_010990.5/ NT_011054.5/ NT_024981.5; Chr19, NT_011098.5/ NT_011141.5/ NT_011157.5/ NT_011196.5/ NT_011225.5; Chr20, NT_028391.3; Chr21, NT_011515.6;+A24 Chr22, NT_011523.7; ChrX, NT_011584.5/ NT_011618.5/ NT_011657.5/ NT_011793.5; ChrY, NT_011896.6). For other information, see Table 1.
Table 1. Size and source of genome DNA sequences.
a Source: 'CONTIG. p1 contig 1'
|Sample||Size (base)||Reference and note|
|A. thaliana Chr 1||22,743,551|||
|A.thaliana Chr 3||9,821,447||- Same as above -|
|C. elegans Chr I||13,467,562|||
|C. elegans Chr II||15,116,321||- Same as above -|
|C. elegans Chr III||12,476,799||- Same as above -|
|C. elegans Chr IV||15,919,001||- Same as above -|
|C. elegans Chr V||20,668,416||- Same as above -|
|C. elegans Chr X||17,432,311||- Same as above -|
|H. sapiens Chr 1||6,906,110||NCBIf|
|H. sapiens Chr 2||6,163,434||NCBI|
|H. sapiens Chr 3||7,256,523||NCBI|
|H. sapiens Chr 4||7,611,217||NCBI|
|H. sapiens Chr 5||6,506,832||NCBI|
|H. sapiens Chr 6||6,705,352||NCBI|
|H. sapiens Chr 7||6,270,794||NCBI|
|H. sapiens Chr 8||7,083,280||NCBI|
|H. sapiens Chr 9||6,270,222||NCBI|
|H. sapiens Chr 10||6,709,963||NCBI|
|H. sapiens Chr 11||6,717,522||NCBI|
|H. sapiens Chr 12||6,308,324||NCBI|
|H. sapiens Chr 13||6,050,474||NCBI|
|H. sapiens Chr 14||6,050,829||NCBI|
|H. sapiens Chr 15||6,319,005||NCBI|
|H. sapiens Chr 16||6,389,377||NCBI|
|H. sapiens Chr 17||6,147,036||NCBI|
|H. sapiens Chr 18||6,408,213||NCBI|
|H. sapiens Chr 19||6,229,141||NCBI|
|H. sapiens Chr 20||2,426,223||NCBI|
|H. sapiens Chr 21||3,427,677||NCBI|
|H. sapiens Chr 22||2,728,705||NCBI|
|H. sapiens Chr X||6,810,333||NCBI|
|H. sapiens Chr Y||6,386,620||NCBI|
|Lambda phage||48,502 b||NCBI|
|P. falciparum Chr 3||1,060,106 c||SCg|
b Source: 'NUCLEOTIDE. NC_001416'
c Source: 'GENOME. NC_000521'
d Artificially generated at random (equivalent to the size of E. coli genome)
e National Institute of Genetics
f National Center for Biotechnology Information
g Sanger Centre
3 Results and Discussion
3. 1 Chromosome homogenization
Genomes of various species can be characterized by oligostickiness maps using a number of oligonucleotide probes . Typical examples of oligostickiness maps are shown for the E. coli and C. elegans genomes using three probes (P4, P5, and P11) in Figure 3. As the length of the spike is proportional to oligostickiness s (for definition see Methods), the probe, P11, can be considered to be of higher oligostickiness to the E. coli genome than the other probes, P4 and P5. Aside from the E. coli genome, we have examined the genomes of other bacteria, such as B. subtilis, H. influenzae, H. pilori, and S. aureus (see Table 1. Data partly shown in Ref. 10). A clear advantage of oligostickiness maps is that they can represent local oligostickiness, g, with ease, which can not be achieved by numerical and statistical approaches. Another benefit of this analysis is the fact that any size of genome (or chromosome) can be depicted by a circle of the same diameter, preserving Equation 1 when the whole genome (or chromosome) is divided into the same number of portions (e.g., 3600). Local oligostickiness can be expressed as follows:
where a is a constant arbitrarily set for the sake of clear representation (we adopted 0.36 for a throughout our representations). The size of the genome can be shown in proportion to its real genomic size (if so, Eq. 2 does not yet hold). The same presentation is used to represent six chromosomes of Caenorhabditis elegans (Figure 3). It is clear that the tendency of oligostickiness is quite different between E. coli and C. elegans. However, it is remarkable that all six chromosomes of C. elegans have the same tendency (or texture ) against the three probes used here. We examined this further using 42 probes and with reference to the genomes of Saccharomyces cerevisiae and Homo sapiens. These results are shown as spider-web mode representations in Figure 4, where 12 probes were used in common. Figure 4B to D shows superimposed spider-web charts of all 16 chromosomes of yeast, 6 of nematode and 24 of human, respectively. All the spider-web charts are genome-specific except for the common dent around P4 to P6 corresponding to the oligostickinesses probed by AT rich sequences. Apart from the fact that each eukaryotic organism has a species-specific shape of spider-web chart, all of the chromosomes of each organism have essentially the same shape of spider-web chart. Especially, the 24 human chromosomes (22 autosomes and 2 sexual chromosomes) have very similar shapes (Figure 4D), considering that some vertices such as P1, P2, and P9-12, which are rather widely distributed, are connected in parallel to each chromosome (thus, providing a proportionally expanded shape of the spider-web chart). In addition, we confirmed the same tendency for all the other probe oligonucleotides tested (additional 20 sequences). This is surprising since each chromosome must contain distinct genes consisting of different sequences. This indicates that chromosomes in a cell (i.e. in the nucleus) have a similar texture of sequence, or in other words possess a monochronous sequence. The rational explanation for this can come from only two sources: (i) all chromosomes originate from a common ancestor and (ii) frequent recombinations occur between the chromosomes. The former asserts a rather unnatural consequence, which is that all chromosomes evolved from the same sequence to different sequences and acquired novel functions. Segmental duplication  and chromosomal duplication must have occurred during the course of evolution, as originally suggested by Ohno . However, it does not mean that the 24 human chromosomes evolved from a single ancestral chromosome confined in a cell, since this would be highly unrealistic. On the other hand, chromosome recombinations have been observed on various occasions, such as meiotic segregation  recombinational repair , SOS responses , and movable elements [6, 27]. Furthermore, recent observations of segmental duplications [7, 8] must have also contributed to generating the current genome state (it is necessary to note that chromosomes might have gathered and evolved prior to the establishment of the current species). Actually, the result shown here clearly demonstrates the possibility of frequent non-homologous recombinations between chromosomes. The 'Frequent recombination' hypothesis can be further supported by the fact that eukaryotic genomes contain a high ratio of non-coding regions such as introns and intergenic sequences that are not only permissible to non-homologous recombinations (which do not always lead to gene disruption) but also to a high probability of recombination. Here, we propose a natural interpretation of the phenomenon of chromosome homogenization: the ancestral chromosomes that had been recruited by successive duplication of chromosomes and/or by cell-fusion of different organisms evolved to their present state through a highly sophisticated mechanism of evolution, which must have involved recombinations such as exon  and module  shuffling. Interestingly, since the overwhelmingly larger portion of eukaryotic chromosomes are composed of non-coding sequences such as introns and intergenic sequences, the monochronous property observed here must be attributed mostly to such sequences, thus indicating that such non-coding sequences retain the genome-wide common yet species-specific property detected by oligostickiness. This is not yet interpretable but what is clear is that it can not be attributed only to highly repetitive sequences such as the Alu and L1 families (over 40% of the whole genome sequence in the human genome ) since the nematode genome contains only 7% repetitive sequence [29, 30]. Repetitive sequences such as the Alu and L1 families do not show high oligostickiness to probes P2 and P5 and they can not substantially contribute to the high oligostickinesses of P2 and P5 in human chromosomes (Figure 4D). This is contrary to the hypothesis of repetitive sequence-caused homogeneity of chromosomes, although they may be responsible for a considerable contribution of homogeneity. Nevertheless, recombination seems to be the main cause of this phenomenon. From Figure 4, the featuring oligostickiness for each organism can easily be read. (i) P12 (dACGACGACGACG) is commonly sticky to yeast and nematode genomes but of very low oligostickiness to the human genome. (ii) P3 (dGAGAGAGAGAGA) is most sticky to the nematode genome. (iii) P5(dA12) is the best discriminator between bacteria and eukaryotes, since oligostickiness to the bacterial genome is much suppressed, whereas that to the eukaryote genomes is high. The advantage of this oligostickiness-based representation is that it allows such features to be easily recognized.
Figure 3. Map presentation of oligostickiness against the six chromosomes of C. elegans and E. coli genome. Chromosomes are shown as a circle beginning at the top and running clockwise. A complete circle is equivalent to the entire chromosome and the diameter of each circle is drawn equal to a oligostickiness value 0.36. Each oligostickiness value is expressed by the length of a spike at each position where the running average of oligostickiness along a 1/3600 part of the entire chromosome is shown. The chromosome nomenclature and the probes used are shown on the top and to the side, respectively. The sizes of the chromosomes are 4.6 (E. coli), 13.5 (Chr I), 15.1 (Chr II), 12.5 (Chr III), 15.9 (Chr IV), 20.7 (Chr V) and 17.4 (Chr X) in Mb. The sequences of the probes are: P4, dGCTAAAAAAAAA; P5, dAAAAAAAAAAAA; and P11, dAGACCGCGCCTG.
Figure 4. Chromosome structures of four organisms represented by a spider-web chart of oligostickiness. The same representation and symbols are used here as in Figure 2. In addition, oligostickiness values plotted on the radial axes are connected with a line to form a circle for each chromosome for multi-chromosomal genomes (B~D). These circles are superimposed. Oligostickiness is plotted on a logarithmic scale with three polygons crossing at 10-3, 10-2, and 10-1 (inner to outer). A, Escherichia coli genome; B, Saccharomyces cerevisiae (16 chromosomes); C, Caenorhabditis elegans (6 chromosomes); D, Homo sapiens (22 autosomes and X and Y chromosomes).
3. 2 Similar oligostickiness between complementary probes
Another piece of evidence supports the frequent recombination hypothesis. Figure 5 shows that each pair of oligonucleotides of complementary sequences has very close oligostickiness values (statistically shown as correlation coefficients (R): most of these correlation coefficients are very close to one). This relationship is valid for all recombinations of chromosomes and complementary probes with the sole exception of human chromosomes examined with the probe dCGATGCTAGCAT(=X)/ dATGCTAGCATCG(=X-bar) (this exception is very prominent as can be seen by the vertically-stacked triangles around 0.006 of oligostickiness for probe X in Figure 5). This indicates the extremely abnormal phenomenon that probe X binds equally to all human chromosomes while X-bar differs from chromosome to chromosome. So far we are unable to find any artifact to explain this phenomenon. Therefore, this might indicate a specific unknown function of this sequence for human chromosomes. On the other hand, the R values regarding non-complementary pairs of probes, such as dCGATGCTAGCAT/ dTACGATCGTAGC or dTGCTGCTGCTGC/ dCGTCGTCGTCGT, which have a relationship in terms of mutually-inverted sequences, were not so high (0.51 and -0.1, respectively) (not included in Figure 5 for clarity). Another important point that should be noted is the wide range of oligostickiness (e.g. 0.000006~0.73 for probe dA12 and 0.002~0.3 for probe dC12). Considering the occurrence of frequent recombinations established above, this phenomenon can only be explained by introducing two modes (the parallel and antiparallel modes; if the orientation of either of two adjacent chromosomes is reversed (face about) just before recombination, then 'parallel' is converted to 'antiparallel') during recombination. These results must represent the well-shuffled state of those sequences, which is attained after vigorous recombinations between chromosomes irrespectively of the direction of DNA. This should not be taken simply as a matter of course because, in completely random artificial genome sequences (Table 1) as indicated in Figure 5 (the insert plotted with the circles outlined in red), the oligostickiness can be the same for a pair of complementary oligonucleotides that is, riding on the line of slope 1. Evidently, this is not valid for those sequences that have highly deviated from the random state like poly (dA), to which the probe dT12 bound well but to which its complementary probe dA12 did not. There is no need for probe X to have a similar oligostickiness as the probe X-bar with regard to the same chromosome. Therefore, this represents evidence that supports the frequent recombination hypothesis. The other wide distribution of oligostickiness among genomes is a type of representation of species diversity acquired during the course of evolution (leading to the concept of negentropy ) while equal oligostickiness for complementary primers (e.g. dCGATGCTAGCAT/ dATGCATGCATGC) represents the opposite entropic effect during evolution.
Figure 5. Correlations of oligostickiness values for pairs of complementary probes obtained with 45 chromosomes from 16 species and 2 random sequences. Seven pairs of complementary oligonucleotides (probes) are adopted: colored symbols for these pairs (pairs of X and X-bar) are shown as an Insert in this figure (upper-left). Correlation coefficients (R) calculated for each set, which is composed of values obtained with 47 chromosomes or similar, are shown beside the least-square line. The width of the lines corresponds to the range of distribution of values (thus, values for probes of A12/T12 give the highest correlation coefficient (0.99) and widest range (0.000006~0.73)). Insert shows range of lower values of oligostickiness (right-bottom). Global oligostickiness is referred to here (in other words, averaged oligostickiness over entire range).
Therefore, chromosomes have maintained their identity by rendering themselves isolated in the nucleus while they have been mangled to homogeneity with respect to their hyper-structure, as observed in this study, by entropic force due to confinement in the nucleus. The resultant chromosomes still retain the information essential for the organism, and the improvement of the quality of information occurs by rare chance events (evolution).
Finally, it is quite natural to suppose that any pair of chromosomes taken from two species in close relationship have mutually similar oligostickiness values since their genomes should have originated from a near, common ancestor. Oligostickiness analysis may be useful for estimating such an evolutional distance between close species.
Software for the analysis of genome oligostickiness was applied to detecting the homogeneity of chromosomes that could be interpreted in terms of random-mode frequent recombination of chromosomes with a considerable contribution of repeated sequences. The visual representation, known as the spider-web chart of oligostickiness, is useful in simplifying the huge amount of information from genome sequences to a manageable database and yet is sufficiently powerful to present the facts clearly. Several features of the human genome as well as those of the yeast and nematode genomes were depicted by oligostickiness, providing novel implications.
The authors are grateful for financial assistance from the Kouhi fund from the Japanese government.
[ 1] M. Kimura, The Neutral Theory of Molecular Evolution, Cambridge University Press, New York (1983).
[ 2] C. H. House and S. T. Fitz-Gibbon, J. Mol. Evol., 54, 539-547 (2002).
[ 3] K. Hiramatsu, L. Cui, M. Kuroda and T. Ito, Trends Microbiol., 9, 486-493 (2001).
[ 4] T. Baba, F. Takeuchi, M. Kuroda, H. Yuzawa, K. Aoki, A. Oguchi, Y. Nagai, N. Iwama, K. Asano, T. Naimi et al., Lancet, 359, 1819-1827 (2002).
[ 5] E. Dawson, G. R. Abecasis, S. Bumpstead, Y. Chen, S. Hunt, D. M. Beare, J. Pabial, T. Dibling, E. Tinsley, S. Kirby et al., Nature, 418, 544-548 (2002).
[ 6] W. H. Li, Z. Gu, H. Wang and A. Nekrutenko, Nature, 409, 847-849 (2001).
[ 7] E. E. Eichler, Trends Genet., 17, 661-669 (2001).
[ 8] J. Cheung, X. Estivill, R. Khaja, J. R. MacDonald, K. Lau, L. C. Tsui and S. W. Scherer, Genome Biol., 4, R25 (2003).
[ 9] K. Nishigaki and Y. Sakuma, J. Chem. Software, 2, 96-107 (1994).
 K. Nishigaki and A. Saito, Bioinformatics, 18, 1153-1161 (2002).
 W. Salser, Cold Spring Harb. Symp. Quant. Biol., 42, 985-1002 (1978).
 Y. Kawarabayasi, M. Sawada, H. Horikawa, Y. Haikawa, Y. Hino, S. Yamamoto, M. Sekin, S. Baba, H. Kosugi, A. Hosoyama et al., DNA Res., 6, 83-101, 145-152 (1999).
 E. Asamizu, Y. Nakamura, S. Sato and S. Tabata, DNA Res., 7, 175-180 (2000).
 F. Kunst, N. Ogasawara, I. Moszer, A. M. Albertini, G. Alloni, V. Azevedo, M. G. Bertero, P. Bessieres, A. Bolotin, S. Borchert et al., Nature, 390, 249-256 (1997).
 S. Kaneko, S. Iwamatsu, A. Kuno, Z. Fujimoto, Y. Sato, K. Yura, M. Go, H. Mizuno, K. Taira, T. Hasegawa et al., Protein Eng., 13, 873-879 (2000).
 M. D. Adams, S. E. Celniker, R. A. Holt, C. A. Evans, J. D. Gocayne, P. G. Amanatides, S. E. Scherer, P. W. Li, R. A. Hoskins, R. F. Galle et al., Science, 287, 2185-2195 (2000).
 R. D. Fleischmann, M. D. Adams, O. White, R. A. Clayton, E. F. Kirkness, A. R. Kerlavage, C. J. Bult, J. F. Tomb, B. A. Dougherty, J. M. Merrick et al., Science, 269, 496-512 (1995).
 J. F. Tomb, O. White, A. R. Kerlavage, R. A. Clayton, G. G. Sutton, R. D. Fleischmann, K. A. Ketchum, H. P. Klenk, S. Gill, B. A. Dougherty et al., Nature, 388, 539-547 (1997).
 D. R. Smith, L. A. Doucette-Stamm, C. Deloughery, H. Lee, J. Dubois, T. Aldredge, R. Bashirzadeh, D. Blakely, R. Cook, K. Gilbert et al., J. Bacteriol., 179, 7135-7155 (1997).
 R. Himmelreich, H. Hilbert, H. Plagens, E. Pirkl, B. C. Li and R. Herrmann, Nucleic Acids Res., 24, 4420-4449 (1996).
 Y. Kawarabayasi, M. Sawada, H. Horikawa, Y. Haikawa, Y. Hino, S. Yamamoto, M. Sekine, S. Baba, H. Kosugi, A. Hosoyama et al., DNA Res., 5, 55-76 (1998).
 Nature The Yeast Genome Directory
 T. Kaneko, S. Sato, H. Kotani, A. Tanaka, E. Asamizu, Y. Nakamura, N. Miyajima, M. Hirosawa, M. Sugiura, S. Sasamoto et al., DNA Res., 3, 109-136 (1996).
 S. Ohno, Evolution by Gene Duplication, Springer-Verlag, New York (1970).
 F. W. Stahl, Genetic Recombination; Thinking About It in Phage and Fungi, W. H. Freeman and Company, San Francisco (1979).
 J. A. Nickoloff and M. F. Hoekstra, DNA Repair in Prokaryotes and Lower Eukaryotes, In DNA Damage and Repair, Humana Press, New Jersey (1998).
 A. I. Bukhari, J. A. Shapiro and S. L. Adhya, DNA Elements, Plasmids, and Episomes, Cold Spring Harbor Laboratory Press, New York (1977).
 J. A. Kolkman and W. P. Stemmer, Nature Biotechnol., 19, 423-428 (2001).
 T. C. e. S. Consortium, Science, 282, 2012-2018 (1998).
 M. V. Katti, P. K. Ranjekar and V. S. Gupta, Mol. Biol. Evol., 18, 1161-1167 (2001).
 N. Wiener, Cybernetics or Control and Communication in the Animal and the Machine, M.I.T. Press, Cambridge (1961).