When the studies described here were initiated, numerous in vitro data were known about transcription by NF-κB, but the actual mechanism of transcription downstream of p65 binding to endogenous genes in vivo was unclear. Since the Mediator complex was known to play an important role in most, if not all, transcription by pol-II, we set out to disrupt its interaction with p65, as a means to dissect p65-driven transcription. We found that expression of some NF-κB target genes depends on direct contact between p65 and Mediator, which occurs through the Trap-80 subunit and the TA1 and TA2 regions of p65. This contact is needed for the establishment of an active, Med-26–containing Mediator complex at promoters, recruitment of TFIIB and pol-II, and thereby the initiation of transcription. While this result is not surprising, it does provide important confirmation that events hitherto only described in minimalist in vitro experiments are necessary, and sufficient, for the expression of some genes in their natural context in vivo.
The finding that 3T3 fibroblasts remain viable even after depletion of Trap-80 to <10% of normal levels is remarkable, considering that its yeast homologue Srb4 is required for most, if not all, transcription by pol-II [15
]. Srb4 is essential for the integrity of the yeast Mediator complex, which, without it, dissociates at the boundary between the structurally conserved head and middle modules [36
]. In mammalian cells, like in yeast, the Mediator complex is critical for pol-II–driven transcription: its addition is required for the in vitro activity of various transcriptional activators [10
], and its depletion from mammalian nuclear extracts abolishes all transcription by pol-II [7
]. Hence, the survival of fibroblasts after Trap-80 has been knocked-down implies that some Mediator activity must still remain. One possibility is that the amount of cellular Mediator is not normally limiting for the expression of essential genes, and that the residual 5%–10% level of Trap-80 in knock-down cells suffices for these. However, we find that the expression of >96% of transcripts changes by less than 1.5-fold in Trap-80–deficient cells (B). This finding argues that, instead, Trap-80–deficient cells contain Trap-80–less, but otherwise functional, Mediator complexes. Proteomic analyses have so far identified Trap-80 to be part of a common core of subunits, shared by all Mediator species [23
]. It seems likely, then, that Trap-80 does not play the same essential structural role as yeast Srb4, and that Mediator complexes that normally incorporate Trap-80 are still able to at least partially assemble when this subunit is missing. This interpretation is supported by our finding that the Mediator complex, revealed by the Med-26 subunit, is recruited to the Mip-2
promoter even in Trap-80–deficient cells (E).
We have shown that the interaction of p65 with Mediator through Trap-80 is sufficient to drive transcription. However, the discovery that a subset of p65-dependent genes are transcribed normally even when the interaction of p65 with Mediator is abolished was completely unanticipated. Moreover, a mutant form of p65 that not only cannot interact with Trap-80, but that also lacks both previously identified transcriptional activation domains, can still activate the Trap-80–independent Mip-2 gene in vivo (p65ΔTA1&2, C). This finding prompted us to examine more closely the events that occur at promoters upon engagement of NF-κB. Remarkably, we found that without the binding of p65, NF-κB target promoters cannot be bound by many other transcription factors. Thus, it appears that a p65-containing NF-κB dimer binds to target promoters as a lone, “pioneer” transcription factor, and controls their subsequent co-occupancy by secondary transcription factors (illustrated in ).
One model for this, which we do not favour, could be that secondary transcription factors bind to promoters via direct, co-operative interactions with p65. Such a scenario has been previously shown in the context of particular promoters containing juxtaposed binding sites (e.g., HIV1-LTR [40
]), but this arrangement is not a general feature of NF-κB target promoters. Moreover, it seems unlikely that pairwise interactions with p65 could account for the binding of multiple transcription factors to each of many different promoters (and at non-NF-κB target promoters, co-operative binding with p65 is clearly not required; see Figure S13
A more plausible possibility is that p65 controls promoter accessibility by inducing local alterations to chromatin. In macrophages, NF-κB–driven activation is accompanied by nucleosome remodeling at target gene promoters [42
]. However, we could detect no differences in promoter accessibility to micrococcal nuclease digestion after stimulation of wild-type and p65-knockout fibroblasts (unpublished data). Alternatively, p65 binding may bring about changes to histone modifications, several of which have been described to be associated with the expression of NF-κB target genes in different systems (e.g., lysine acetylation [43
] and methylation [45
], serine phosphorylation [46
]). Further experiments are required to determine whether these could account for the control over secondary transcription-factor binding.
In p65-knockout cells, artificial recruitment of a secondary transcription factor is sufficient to restore gene expression (), indicating that regardless of the mechanism, the regulation of promoter occupancy constitutes a second, independent mode of transcriptional activation by p65.
What, though, could be the benefit of having a second mode of transcriptional activation? After all, in real, nonexperimentally manipulated cells, an intact Mediator complex containing Trap-80 is always present. We can envisage two situations in which the ability of p65 to control recruitment of secondary transcription factors to a promoter could be important. First, if the only means by which p65 could activate transcription was through its direct binding to Mediator, then the transcriptional output at every NF-κB−dependent promoter should be the same, and upon its release from promoters, transcription would necessarily halt. There are numerous mechanisms that control the longevity of promoter-bound p65, including nuclear export by resynthesized IκB molecules [48
], ubiquitination and proteasomal degradation [49
], and replacement by other NF-κB dimer species [50
]. Considering the tremendous diversity of NF-κB target genes, though, it seems inconceivable that the optimal biological window and level of expression for all of them can be identical (and experimentally this is not the case; compare, for example, expression of Mip-2
in A). By endowing p65 with the ability to license promoters for the binding of secondary transcription factors, there is a means to customize expression levels, and prolong transcription after the departure of p65. From the point of view of a promoter, this would be an attractive solution, since NF-κB–dependence can be retained at the same time as tailoring the expression pattern by selecting binding sites for appropriate secondary transcription factors.
Second, κB sites are not always located in promoters close to the transcriptional start sites, and in some cases can be several kilobases away (A; examples include Mcp-1 and JunB). At such a distance, looping of the intervening DNA would be required to bring bound p65 into the proximity of core promoter elements, and this may not allow a sufficiently stable interaction with Mediator to enable nucleation of the pre-initiation complex. However, the local presence of p65 is nevertheless adequate to regulate the recruitment of secondary transcription factors. In turn, these newly arrived transcription factors can stably bind to the promoter, and themselves interact with components of the pre-initiation complex to drive transcription. In this model, the Trap-80–independent mode of activation by p65 is critical to permit it to operate at enhancers.
A corollary of activation by p65 in this way is that the activity of a given target promoter, although entirely NF-κB–dependent, will depend on the availability of suitable secondary transcription factors. Since this is determined by both cell-type and stimulus, this mode of activation is likely to be essential to allow genes controlled by NF-κB to attain an appropriate pattern of expression in different biological contexts.