in Fig. 4.3, varying the boundary conditions of the loop or the assumed protein flexibility can lead to enormous differences in predicted looping J-factors. Some of these predicted J-factors, using canonical assumptions about DNA flexibility, are in fact consistent with the J-factors we measure, so perhaps it should not be surprising that short transcription factor-mediated loops can form readily in vitro.
decreased transcriptional noise [75]. In fact, it has been argued that poly(dA:dT) tracts are the major determinants of nucleosome positions in vivo, rather than nucleosome-preferring sequences such as TA [15].
However, the mechanism by which poly(dA:dT) tracts exclude nucleosomes and influence tran- scriptionin vivois as yet unclear [148]. It is known that DNA polymers with 4 or more A nucleotides in a row show unique structural and dynamical properties in a variety of assays [148], and generally it is thought that long stretches of poly(dA:dT) are relatively straight and inflexiblein vitro [154]
(though see [61] for experimental evidence that poly(dA:dT) is more flexible, not less; and also it should be noted that thephasedA-tracts discussed in Chapter 1 are known to induce intrinsic bends into DNA [155], rather than be intrinsically straight). A leading hypothesis for why poly(dA:dT) tracts disfavor nucleosome formation, then, is that their unique structural and dynamic properties lead them to be especially resistant to the deformations that are required for DNA wrapped in a nucleosome [148]. That is, just as the TA sequence favors nucleosome formation because of a high intrinsic flexibility (at least with regards to certain deformations), poly(dA:dT) has a high intrinsic inflexibility relative to the deformations involved in nucleosome formation. Again this high inflexibil- ity is thought not to arise from any particular stiffness to AA dinucleotide steps but rather from the special structures known to form when more than two AA steps are found in a row [148]; nevertheless, they should look “stiff” under comparable deformations to those required in a nucleosome.
To test this hypothesis that poly(dA:dT) tracts disfavor nucleosome formation because of a high intrinsic inflexibility in the context of nucleosome-like deformations (regardless of the molecular origin of this stiffness), and also to test the generality of our results with E8 and TA, we chose a poly(dA:dT)-rich promoter region from S. cerevisiae that was shown to exclude nucleosomes in vivo by microarray analysis [149], and inserted this sequence into both the no-promoter and with- promoter loops described in the previous sections. (See Fig. B.3 and Appendix B.2 for details of these sequences.) If the results from the E8- and TA-containing loops hold more generally, and poly(dA:dT) sequences disfavor nucleosomes due to a high intrinsic inflexibility in the context of nucleosome-like shapes, then we would expect these poly(dA:dT) sequences to yield the same amount
(A)
90 95 100 105 110 115
0 0.2 0.4 0.6 0.8 1
Loop Length (bp)
Looping Probability
120 125 Oid-E8-O1 Oid-TA-O1 Oid-polyA-O1
10-11 10-10 10-9 10-8
Loop Length (bp) Jloop (M)
Oid- E8- O1 Oid- TA- O1
(C)
90 95 100 105 110 115 120 125
Oid-polyA- O1
90 95 100 105 110 115
0 0.2 0.4 0.6 0.8 1
Loop Length (bp)
Looping Probability
Oid-E8-O1, M Oid-E8-O1, B Oid-TA-O1, M Oid-TA-O1, B
(B)
120 125
0
Oid-polyA-O1, M Oid-polyA-O1, B
WITH PROMOTER NO PROMOTER
95 100 105 110 115 120 125
0 0.2 0.4 0.6 0.8 1
Loop Length (bp)
Looping Probability
Oid-E8-(prom)-O2 Oid-TA-(prom)-O2
(D)
90
Oid-polyA-(prom)-O2
95 100 105 110 115 120 125
0.2 0.4 0.6 0.8 1
Loop Length (bp)
Looping Probability
Oid-E8-(prom)-O2, M Oid-E8-(prom)-O2, B Oid-TA-(prom)-O2, M Oid-TA-(prom)-O2, B
90
(E)
Oid-polyA-(prom)-O2, M Oid-polyA-(prom)-O2, B
90 95 100 105 110 115 120 125
10-12 10-11 10-10 10-9 10-8
Loop Length (bp) Jloop (M)
Oid- E8- O1 Oid- E8- (prom)- O2 Oid- TA- (prom)- O2
(F)
Oid-polyA-(prom)-O2
Figure 4.4: Looping probabilities as a function of loop length for a poly(dA:dT)-rich sequence known to exclude nucleosomes in yeast, superimposed on the E8 and TA data of Fig. 4.2, both with and without the promoter. (A) Total looping probabilities for the no-promoter constructs, as in Fig. 4.2(A). Data at additional loop lengths with the poly(dA:dT) sequence will be necessary to draw definite conclusions, but it appears that the poly(dA:dT) has a different period than the E8 and TA data, and that, in contrast to the E8 and TA data, the poly(dA:dT) sequence may alter the amount of looping compared to E8, even without the promoter. (B)Looping probabilities for the two looped states separately, as in Fig. 4.2(B), for the constructs in (A) here. Interestingly, even though the periods of these three sequences seem to be different, the pattern of which looped state predominates at a given length is consistent between all three sequences: note especially the 107 and 108 bp lengths. (C)J-factors corresponding to the total looping probabilities in (A). One of the poly(dA:dT) loops is almost as flexible as TA94, which is surprising given that we expected, based on nucleosome affinity assays, that poly(dA:dT) might belessflexible. (D)Total looping probabilities for the with-promoter constructs, as in Fig. 4.2(D). As with the data in (A), it appears that the poly(dA:dT) sequence has a different period relative to E8 and TA, so the conclusions we can draw from only four data points are limited. However, it seems that, unlike with E8 and TA, with the promoter in the loop the sequence dependence of poly(dA:dT) may not follow that of nucleosome formation, at least relative to E8: the TA sequence, which nucleosomes preferentially bind to over the random E8 sequence, loops more than E8; but the poly(dA:dT) sequence, known to exclude nucleosomes from a promoter regionin vivo, loops as much as E8 (though still less than TA) at some lengths. (E) Looping probabilities for the two states separately, for the constructs in (A). As was observed for the E8 and TA constructs in Fig 4.2(E), with the promoter in the loop, the middle state predominates at all four lengths, whereas without the promoter in (B), the bottom state is equally or more dominant at 107 and 108 bp. (F)J-factors for the constructs in (D). The presence of the promoter decreases the J-factor of the poly(dA:dT) sequence relative to the no-promoter constructs, though not to a value less than that of E8, contrary to what we would expect from nucleosome formation assays.
of looping as the E8 and TA sequences without the promoter, but for the with-promoter loops to follow the same sequence dependence as nucleosome formation, that is, with the poly(dA:dT) sequences looping less than E8 and TA.
As shown in Fig. 4.4, this is not what we find. As our preliminary results include only four loop lengths with and without the promoter, we can at the moment draw only limited conclusions.
But the most striking feature of the poly(dA:dT) data is that it appears the period of looping (that is, at what lengths looping is maximized or minimized), is different for the A-tract containing DNAs compared to that of E8 or TA. This is perhaps unsurprising, given that A-tract-containing DNAs are thought to adopt unique structures, and in fact some A-tract-containing DNAs have been shown to have shorter periods than other sequences [148]. What is also striking, however, is how flexible both the no-promoter and with-promoter poly(dA:dT)-containing loops appear to be: the 105 bp loop without the promoter is almost as flexible as the TA94 sequence (Fig. 4.4(C)), and even without the promoter the poly(dA:dT) loop is at least as flexible as E8 loops of comparable (though not identical, due to the period offset) lengths. Indeed, it appears that, in contrast to the results with E8 and TA, the poly(dA:dT) loop does show a sequence dependence in the absence of the promoter, as well as with the promoter, in that its looping probability is different from that of E8 in both cases. One aspect of the data is consistent across all three sequences, though: the relative probabilities of the different looped states. Without the promoter, the two looped states alternate in prevalence, including for the poly(dA:dT) constructs, but with the promoter, the middle looped state predominates.
We argued in the previous section that we suspect that whether or not there is a sequence de- pendence to looping depends strongly on the shape of the loop, with the promoter altering the preferred conformation such that the promoter-containing loops are more similar in shape to nu- cleosomes than the no-promoter loops, and therefore the patterns of sequence dependence seen in nucleosome formation hold only for promoter-containing loops, and not the no-promoter loops. We have now shown that that is not generally the case: if the with-promoter DNAs followed the sequence preferences of nucleosomes, then the poly(dA:dT) sequence with the promoter should have looped
less than E8 regardless of the period offset. However, the fact that now neither the with-promoter nor no-promoter constructs follow the sequence dependence trends of nucleosomes underscores even further our argument that“sequence flexibility” is not a general term.
We maintain our original hypothesis that the notion of sequence flexibility needs to be linked to the shape of the deformation induced to measure such flexibility. Because poly(dA:dT) tracts are known to possess unique structural properties, the fact that the poly(dA:dT)-containing loops do not match the sequence-dependence trends of E8 and TA is perhaps further evidence that the shape of the loop plays a large role in the observed flexibility trends. If that were the case, it would also demonstrate that the Lac repressor can accommodate a range of different looped structures, based on the deformation-dependent flexibilities of the loop sequence. Indeed, Haeusler and coworkers have recently shown that the Lac repressor can accommodate a surprisingly large range of designed loop topologies (made with phased A-tracts that introduce static bends in the DNA) [70]. We anticipate that the poly(dA:dT) loops form yet an additional shape, beyond the different shapes we have postulated for with-promoter versus no-promoter E8- and TA-containing loops, because of the unique structural requirements of A-tract DNAs.
As will be described in Chapter 7, rigorous testing of our deformation-dependent hypothesis will require testing a broader region of sequence space than can be accomplished by picking and choosing from sequences studied in the context of nucleosome formation, which addresses only a limited region of shape space (roughly circular) that is probably inaccessible to looping. We will therefore propose ade novosearch of sequence space to try to identify sequences that are especially good or especially poor looping sequences, to try to build up rules for the sequence dependence to loop formation, as has already been done with nucleosomes [14]. More importantly, as will be seen in the next section, the question of whether sequence flexibility is a “knob” that tunes loop formationin vivoremains a very important and outstanding one that cannot, we will argue, necessarily be addressed byin vitro techniques alone.