Fig. 2 Quadrant analysis of Chargaff differences for the W bases and the S bases in 1 kb windows from the "top" strand of the first 350 kb of the H. influenzae genome. (a) The 44 windows whose centres are closest to Chi-like sequences in the bottom strand. (b) The 14 windows whose centres are closest to Chi-like  sequences in the top strand. (c) The 1562 windows whose centres overlap leftward-transcribing ORFs. (d) The 1326 windows whose centres overlap rightward-transcribing ORFs. Other details are as in Fig. 1.

Similar results were obtained in the case of the first 350 kb of the H. influenzae genome (Fig. 2; Table 2). Reflecting the base composition of windows corresponding to rightward transcribing ORFs (Fig. 2d), windows corresponding to Chi sequences in the top strand tend to be GT-rich (Fig. 2b). Reflecting the base composition of windows corresponding to leftward-transcribing ORFs (Fig. 2c), windows corresponding to Chi sequences in the bottom strand tend to be AC-rich (Fig. 2a), the latter referring to the composition of the top strand. This means that bottom strand Chi sequences tend to be embedded in a locally GT-rich strand like their top strand counterparts. For the top strand Chi sequences (Fig. 2b) the average Chargaff difference for the W bases was -3.6  se   1.8 (i.e. T>A; Table 2). The corresponding value for the S bases was -9.3 se 2.7 (i.e. G>C). For the bottom strand Chi sequences (Fig. 2a) the average Chargaff difference for the W bases was 1.3 se 1.2 (i.e. A>T). The corresponding value for the S bases was 10.0 se 1.3 (i.e. C>G). For all windows transcribed to the right (Fig. 2d) the average Chargaff difference for the W bases was 2.1 se 0.2 (i.e. A>T). The corresponding value for the S bases was -3.6 se 0.2 (i.e. G>C). For all windows transcribed to the left (Fig. 2c) the average Chargaff difference for the W bases was 0.3 se 0.2 (i.e. A>T). The corresponding value for the S bases was 10.2 se 0.2 (i.e. C>G).

3.4. Chargaff difference analysis of transcription direction

   That Szybalski's transcription direction rule applies in local regions of DNA is evident from detailed studies of sequence features, and can be employed to assist the identification of functional ORFs (Dang et al., 1998). For example, Fig. 3 shows Chargaff differences for a 20 kb segment corresponding to the 100-120 kb region of the E. coli genome. If the line for the S bases (C and G) is followed from left to right (red symbols), it is  seen that G>C in the case of the rightwardly-transcribed first half of the  segment, where most of the genes are unambiguous.

Fig. 3. Variation of "Chargaff differences" ("G-C skew") along a 20 kb  segment from E. coli K-12. A 1 kb window was moved in steps of 100 nt and base compositions were determined in each window. Chargaff differences for the S bases (C and G; red squares) are calculated as (C-G)/S (where S = C+G). Chargaff differences for the W bases (A and T; yellow circles) are calculated as (A-T)/W (where W=A+T). Data points are located at the center of each window. The proposed locations of ORFs are shown as boxes. Horizontal blue arrows indicate the transcription directions assigned to each ORF by Blattner et al. (1997), with rightward ORFs being represented by filled boxes. Vertical arrows indicate the positions of Chi sequences either in the top strand (upward pointing), or in the bottom strand (downward pointing). Vertical dotted lines correspond to regions where the transcription direction switches. Abbreviated gene names are as in Blattner et al. (1997).

    The positions of Chi sequences in the "top" strand (the mRNasynonymous strand when transcription is to the right) are marked by upward pointing arrows, and one in the "bottom" strand (the mRNA synonymous strand when transcription is to the left) is marked by a downward-pointing arrow. The top strand Chi sequences are in regions where transcription is to the right, and the bottom strand Chi  sequence is in a region where transcription in to the left. The first two top strand Chi sequences are close to each other so that their "recombination islands" would slightly overlap. The region is purine-rich (A and G) rather than GT-rich. The third top strand Chi  sequence is in a GT-rich region. The bottom strand Chi sequence would be expected to be located preferentially in a region where the top strand would be AC-rich, but this particular Chi locates to a region where Chargaff differences are small.

3.5. Comparison of natural and shuffled  sequence

    In Fig. 4 average absolute Chargaff differences are plotted against the size of  sequence windows in the above 200 kb  segment from E. coli. With windows of only 200 nt, high differences would be expected since there would be great statistical fluctuations when base "coins" are "tossed" no more than 200 times. Indeed, average absolute differences for both the W bases and the S bases are high when windows are 200 nt. However, values for the natural  sequence exceed those of the shuffled natural sequence, implying an evolutionary selection pressure on base order favouring non-parity between the Watson-Crick base pairs.

Fig. 4. Variation of average Chargaff difference values with size of windows in E. coli. Windows of varying size were moved along the 200 kb segment in steps of 100 nucleotides, and base compositions were determined in each window. Average absolute Chargaff differences for each window size are plotted either as large grey symbols (natural  sequence), or as large black symbols (shuffled  sequence). Small yellow diamonds refer to the ratio of the se values (the average Chargaff difference for the natural sequence divided by the average Chargaff difference for the shuffled sequence). Small red diamonds refer to the difference between these values determined by subtraction. The horizontal dotted lines indicate Chargaff differences for the entire sequence (i.e. the largest possible window, of which there is only one copy). Thus, the total number of windows of a given size varies with sequence length. In a 100 kb  sequence there will be 999 windows of 0.2 kb, and one window of 100 kb.

    With increasing window size, average Chargaff differences for both natural and shuffled sequences decrease exponentially to approach the value for the entire  segment (horizontal dotted lines). Values for the natural and shuffled sequences were compared either as a ratio (open yellow diamonds), or by subtraction (filled red diamonds). The size of the window at which Chargaff differences for natural and shuffled  sequences diverge maximally depends on the method used. In the case of the W bases the maximum divergence by ratio occurs with 4 kb windows, but the maximum divergence by subtraction is with 1.1 kb windows. In the case of the S bases the divergence by the ratio method is high with 1 kb windows, but reaches at maximum with 1.5 kb windows. By the difference method, the divergence reaches a maximum level at 0.6 kb which is sustained to 1.2 kb. Thus, the base order of the natural  sequence has been under evolutionary constraint, and for E. coli the window size at which this is maximally evident is about 1 kb. This is consistent with a fundamental transcription/recombination "island" unit size of approx. 1 kb.

4. Discussion

    In 1951 Chargaff reported the famous equality of the Watson-Crick pairing-bases in duplex DNA (%A=%T, %C=%G; Chargaff's first parity rule), and in 1968 showed that, to a close approximation, this also applied to single strands of DNA (Chargaff's second parity rule; Rudner et al., 1968; Chargaff, 1979). In 1966 Szybalski et al. reported that in certain microorganisms RNA polymerase template strands contained pyrimidine clusters; hence, the corresponding mRNA synonymous DNA strands contained purine clusters.

    Combining the observations of Chargaff, Szybalski and their coworkers, it appeared possible that when transcription was to the right there would be a compositional bias, a small deviation from the  second parity rule, in favour of purines; similarly, when transcription was to the left the bias would favour pyrimidines (see the prescient Fig. 4 of Szybalski et al., 1966). Studies of Smithies et al. (1981) suggested that this might apply both to prokaryotes and eukaryotes, as has been confirmed (S. J. Bell and D. R. Forsdyke, [1999a,b; unpublished in 1998]; Forsdyke and Bell, 1997). This regularity, which we refer to as Szybalski's transcription direction rule, can assist the prediction of functional ORFs (Dang et al., 1998). In the case of low (C+G)% genomes, the purine predictor of rightward-transcription tends to be A, whereas in the case of average or high (C+G)% genomes, the purine predictor tends to be G. Assuming the examined segments to be representative, for both H. influenzae (C+G = 38%), and E. coli (C+G = 51%), the best predictor of rightward-transcription is G (Figs. 1-3; Table 2). However, for leftward-transcription, T is the best pyrimidine predictor for E. coli, and C is the best for H. influenzae.

     Since contiguous ORFs may be transcribed in different directions, in one section of DNA one may encounter regions enriched in purines (rightward-transcription), and pyrimidines (leftward-transcription; Fig. 3). If the order of bases in the region were shuffled, the bias would be decreased, and Chargaff differences would be minimized. Thus, by comparing Chargaff differences in natural and shuffled  sequences, one arrives at a window size (1 kb) which is most likely to be biologically informative (Fig. 4). Below this window size stochastic factors make correlations with ORFs difficult. Above this window size, purine-rich regions tend to cancel out pyrimidine-rich regions (S. J. Bell and D. R. Forsdyke [1999a,b; unpublished in 1998]; Dang et al., 1998). The critical window size approximates that of the average transcription unit. Remarkably, this corresponds closely to the size of Chi domains which Tracy et al. (1997b) refer to as "recombination islands".

   Our studies show that a relationship between recombination and transcription is more evident than a relationship between recombination and replication (Table 1). Correlations between the orientations of Chi (or Chi-like)  sequences, and transcription directions, support the case for a fundamental relationship between domains of recombination ("recombination islands"), and domains of transcription ("transcription islands"). Chargaff difference quadrant analysis further emphasizes the linkage between recombination and transcription (Figs. 1, 2). G-rich  sequences bind preferentially to the E. coli RecA protein and its yeast homolog (Rad51; Tracy et al., 1997a). The G-richness of strand-specific Chi-domains seems to reflect the fact that they occupy mRNA-synonymous DNA strands which are generally purine-rich (Table 2). However, in the case of the AT-rich H. influenzae genome, windows with Chi-like  sequences at their centres may be a distinct subset of purine-rich windows. This is particularly apparent in the case of rightward-transcribed ORFs where both A and G contribute to the general purine richness, whereas mainly G contributes to top strand windows with Chi-like sequences at their centres (Fig. 2; Table 2). A possible explanation for "purine-loading" of mRNA synonymous strands is considered elsewhere (Cristillo et al., 1998).

Acknowledgements

We thank J. T. Smith for statistical advice, J. H. Gerlach and A. D. Cristillo for assistance with computer configuration, and Queen's University for financial support.

References

Blattner, F. R. et al., 1997. The complete genome  sequence of Escherichia coli K-12. Science 277, 1453-1474.

Fleischmann, R. D. et al., 1995. Whole-genome random  sequencing and assembly of Haemophilus influenzae Rd. Science 269, 496-512.

Forsdyke, D. R., 1996. Different biological species "broadcast" their DNAs at different (C+G)% "wavelengths". J. Theor. Biol. 178, 405-417.

Forsdyke, D. R. 1998. An alternative way of thinking about stem-loops in DNA. A case study of the G0S2 gene. J. Theor. Biol. 192, 489-504.

Forsdyke, D. R., Bell, S. J., 1997. Deviations from Chargaff's  second rule correlate with direction of transcription and genome structure. Proc. Can. Fed. Biol. Socs. 40, 87 (abstract no. 260).

Freeman, J. M., Plasterer, T. N., Smith, T. F., Mohr, S. C., 1998. Patterns of genome organization in bacteria. Science 279, 1827.

Kleckner, N., 1996. Meiosis: how could it work? Proc. Natl. Acad. Sci. USA 93, 8167- 8174.

Kodadek, T., 1998. Mechanistic parallels between DNA replication, recombination and transcription. Trends Biochem. Sci. 23, 79-83.

Lobry, J. R., 1996. Origin or replication of Mycoplasma genitalium. Science 272, 745-746.

Meyer, R. K., Krueger, D. D., 1994. Minitab Computer Supplement. Macmillan College Publishing, New York, pp. 228-229.

Nicolas, A., 1998. Relationship between transcription and initiation of meiotic recombination: towards chromatin accessibility. Proc. Natl. Acad. Sci. USA 95, 87-89.

Rudner, R, Karkas, J. D., Chargaff, E., 1968. Separation of B. subtilis DNA into complementary strands, III. Direct analysis. Proc. Natl. Acad. Sci. USA 60, 921-922.

Smith, G. R., 1997. Chi sites and their con sequences. In: Bruijn, F.J. de, Lipski, J.R., Weinstock, G. M. (Eds.), Bacterial Genomics. Thomson Publishing, New York, pp. 49-66.

Smith, H. O., Tomb, J. -F., Dougherty, B, A., Fleischmann, R. D., Venter, J. C., 1995. Frequency and distribution of DNA uptake signal  sequences in the Haemophilus influenzae Rd genome. Science 269, 538-540.

Sourice, S., Biaudet, V., El Karoui, M., Ehrlich, S. D., Gruss, A., 1998. Identification of the Chi site of Haemophilus influenzae as  several sequences related to the Escherichia coli Chi site. Mol. Microbiol. 27, 1021-1029.

Stewart, S. E., Roeder, G. S., 1989. Transcription by RNA polymerase I stimulates mitotic recombination in Saccharomyces cerevisiae. Mol. Cell. Biol. 9, 3464-3472.

Szybalski, W., Kubinski, H., Sheldrick, P., 1966. Pyrimidine clusters on the transcribing strand of DNA and their possible role in the initiation of RNA synthesis. Cold Spring. Harb. Symp. Quant. Biol. 31, 123-127.

Tartof, K. D., Henikoff, S., 1991. Trans-sensing effects from Drosophila to humans. Cell 65, 201-203.

Tracy, R. B., Baumohl, J. K., Kowalczykowski, S. C., 1997a. The preference for GT-rich DNA by the yeast Rad51 protein defines a set of universal pairing sequences. Genes Devel. 11, 3423-3431.

Tracy, R. B., Chedin, F., Kowalczykowski, S. C., 1997b. The recombination hot spot Chi is embedded within islands of preferred DNA pairing  sequences in the E. coli genome. Cell 90, 205-205.

Wu, C, -t., 1993. Transvection, nuclear structure, and chromatin proteins. J. Cell Biol. 120, 587-590.