To understand what determines the position of Tf1 integration, we examined the pattern of insertion in the promoter of fbp1. The insertion sites clustered into two distinct windows. The context of fbp1 within a target plasmid and the analysis of separate patches of cells allowed us to determine that positions 2840 and 2874 of TW2 and positions 3090 and 3100 of TW1 had many independent integration events. Our systematic deletion of the fbp1 sequences revealed that the target windows functioned independent of each other. More importantly, each target window was sufficient for establishing a specific set of coordinates as sites of repeated integration. These data, together with the observation that integration into TW1 and not TW2 depended on Atf1p, indicate that integration into the two target windows relies on different factors.
The integration assays we conducted were based on plasmid-carried copies of the fbp1
sequence. The positions of nucleosomes on ura4
are the same regardless of whether the genes are carried on plasmids (2
). However, it is possible that the chromatin bound to the fbp1
promoter in the plasmid may be assembled differently from what exists in the chromosomal copy of fbp1.
As a result, the integration patterns observed within plasmids might not accurately represent the patterns that occur in the chromosomal copies of genes. However, the data from the genome-wide study of Tf1 integration showed insertions in the promoters of fbp1
that corresponded closely to the patterns observed for the plasmid versions of these genes (18
). In four independent sets of integration data from the genome-wide study, insertions in fbp1
clustered in TW1 and TW2. In TW1, three inserts occurred at position 3090 and two occurred at position 3100. In TW2, one insert was detected at position 2840 and four occurred at position 2874. This close correlation between the insertions observed in the chromosomal copies of fbp1
and the patterns detected in plasmid copies indicates that the target plasmid assay can faithfully reproduce authentic integration patterns. There are a few sites in the backbone of the plasmid that are frequent sites for integration. These include positions 340 and 384 in the bacterial gene bla
. Since these positions are in a bacterial sequence, they do not represent natural insertion events. However, position 340 is targeted regardless of which other sequences are present in the plasmid or which host genes are deleted. It is possible that this site was fortuitously recognized by a factor that directs integration or that the DNA sequence itself has a structure that is directly recognized by IN.
Integration assays with S. pombe
confirmed the previous finding that Atf1p is required for the pattern of integration associated with UAS1 (25
). In addition, our use of TW1 and TW2 miniplasmids revealed that Atf1p was specifically required for the integration at TW1 and did not contribute to the inserts in TW2. These data and the previous finding that the pattern of integration in the fbp1
promoter depends on the enhancer sequence of UAS1 suggest two types of models. Atf1p bound at its recognition sequence in UAS1 may simultaneously bind directly to integrase and direct integration to occur at the primary sites 30 nt and 40 nt downstream of UAS1. This possibility is supported by coimmunoprecipitation assays showing that IN is in a complex with Atf1p (25
). Alternatively, integrase may be directed to UAS1 by any number of other transcription factors that rely on Atf1p simply because it initiates transcription. The result that other factors that modulate fbp1
transcription, such as Pcr1p, Rst2p, Tup11p/Tup12p, and Pka1p, did not contribute to the pattern of integration indicates that Atf1p plays a direct role in targeting integration. It is particularly telling that Pcr1p and Atf1p must bind UAS1 together as a heterodimer to stimulate transcription. The lacZ
assays showed that deletion of pcr1
reduced expression of fbp1
to the level exhibited by the strain lacking atf1.
Nevertheless, deletion of pcr1
did not alter the pattern of integration. This indicates that Atf1p plays a direct role in mediating integration in TW1. Interestingly, cells lacking tup11
had three times more integration in the plasmid than did wild-type cells. It is known from micrococcal nuclease digestion that cells lacking tup11
have reduced chromatin structure in the promoter of fbp1
). This lack of structure may account for the corresponding increase in integration efficiency.
Factors other than Atf1p can direct integration. Cells lacking Atf1p accumulate insertions genome-wide, with frequencies similar to those of wild-type cells. The finding that integration in TW2 was independent of Atf1p also showed that factors other than Atf1p can direct integration.
Since it appears that factors other than Atf1p can direct integration, we asked whether any transcription activator bound to its enhancer could promote integration. The result showing that integration at fbp1 is directed to UAS1 and not to UAS2 indicates that binding of a transcription activator to its cognate UAS does not necessarily promote integration. The fact that UAS2 is an efficient enhancer of transcription in the context of our plasmid supports the conclusion that activated transcription is not sufficient to mediate integration. Additional support for this conclusion was that the transcription activator Rst2p promoted efficient transcription of fbp1 in the plasmid but did not contribute to integration. These results indicate that only specific transcription factors, such as Atf1p, are capable of mediating integration.
Further support for the model that only specific transcription factors are able to direct integration came from the study of an artificial promoter. By fusing the DNA binding domain of LexA to the activator domain of VP16, we were able to drive efficient transcription of a lacZ reporter that had eight upstream copies of the lexA binding site. Nevertheless, this artificial promoter was not a target of Tf1 integration, showing that the binding of a transcription activator and the initiation of transcription were not by themselves capable of mediating integration. This result also argues against the possibility that integration is mediated by a general component of the transcription machinery. The most likely explanation for how integration is targeted is that there is a specific set of transcription factors that include Atf1p and that these proteins have the specific abilities to recruit integrase to positions in promoters and to promote integration. Our future studies will focus on identifying this key set of transcription factors.