The Mediator complex is specific to eukaryotic organisms and, despite significant divergence in subunit sequence identity, several basic functions of the complex (e.g., interaction with DNA-binding transcription factors and RNA polymerase II (pol II), and general requirement for genome-wide transcription) are conserved between yeast and mammals1. Mediator possesses a disproportionately large number of predicted intrinsically disordered regions in the sequences of its subunits2. Many of the intrinsically disordered regions have sequence signatures that suggest the presence of protein-protein binding domains although the exact nature of these putative interactions remains largely unknown.

The first structures of the S. cerevisiae Mediator complex were resolved with electron microscopy (EM)3. Previous data revealed not only some conservation of structural features between yeast and mammalian Mediator complexes but also a great degree of flexibility and conformational dynamics3,4. For example, binding of pol II to the S. cerevisiae Mediator complex triggered extensive structural rearrangements.

The designations of the S. cerevisiae Mediator head, middle, and tail modules were made in part from EM 2D projections, based on the organization of protein density in the complex3. Understandably, however, a thorough confirmation of the location of specific subunits within the structure has not been completed. Given the large number of subunits, it would be a monumental task to systematically locate each Mediator subunit via EM labeling techniques. Whereas the study by Wang et al. does not label and localize each subunit in the Mediator complex, they combined EM labeling of four subunits with other strategies to reveal a modified architectural model of the S. cerevisiae Mediator complex5. As S. cerevisiae Mediator is structurally flexible and large in size (approximately 1 MDa; 21 protein subunits), EM is a preferred technique for structural analysis of the complex. Although numerous EM studies have been completed for S. cerevisiae Mediator, the work presented by Wang et al. addresses some key issues and reveals some surprising and important insights.

A number of aspects of the study by Wang et al. stand out as being important for its success. First, a purification protocol was implemented that allowed isolation of yeast (S. cerevisiae) Mediator at high concentration (1 mg/ml). Second, the authors employed nanogold labeling strategies to enable localization of Med14, Med17, Med21 and Med22 within the complex. This represents the first time that specific subunits have been localized in EM maps of an entire yeast Mediator complex. Third, Wang et al. used a variety of approaches to isolate and structurally characterize the Mediator middle + head domains and middle domains only. This information helped validate and clarify the location of the middle, head and tail modules of the S. cerevisiae Mediator complex (Figure 1). The EM structural model for the middle module shown by Wang et al. indicates an extended and apparently highly flexible shape, in agreement with published studies with recombinantly expressed subunits6. Taken together, the findings presented by Wang et al. provide new and very basic information about the subunit architecture of S. cerevisiae Mediator; the fact that such details remained uncertain reflects the challenges associated with structural analysis of the entire Mediator complex.

Figure 1
figure 1

Schematic representation of a previous architectural model of S. cerevisiae Mediator (left) and the redefined model (right) based on data from the Cai group5. The S. cerevisiae Mediator complex contains 21 subunits: 7 in the head module (Med6, Med8, Med11, Med17, Med18, Med20, Med22), 9 in the middle module (Med1, Med4, Med7, Med9, Med10, Med14, Med19, Med21, Med31), and 5 in the tail module (Med2, Med3, Med5, Med15, Med16).

Full size image

Importantly, the redefined structural organization of S. cerevisiae Mediator is consistent with existing genetic and biochemical studies. The Kornberg group, for example, has shown that the S. cerevisiae Mediator head module does not bind pol II on its own, but requires also the middle module and/or other preinitiation complex (PIC) components such as TFIIF7. These findings are supported by the architectural model proposed by Wang et al., as yeast pol II appears to interact predominantly with the head and middle domains based on the new model. Previous EM studies showed that “tail-less” S. cerevisiae Mediator complexes (i.e., containing only head + middle modules) could interact with the pol II enzyme4. Furthermore, S. pombe Mediator naturally lacks subunits that comprise the tail domain, like S. cerevisiae Mediator, and it also interacts with pol II8. The revised structural model is even consistent with possible structural roles for S. cerevisiae tail subunits in pol II-Mediator interactions during transcription9.

Given the results from Wang et al., it will be interesting to see whether similar subunit organization is observed in mammalian Mediator complexes. Presently only a few subunits have been localized in the human Mediator complex, and this localization has been demonstrated indirectly via antibody labeling of transcription factors bound to the Mediator. Given the distinct subunit composition and poor sequence conservation, it is possible that Mediator subunit organization is distinct in mammals. It is also worth noting that the human Mediator complex undergoes structural changes that could significantly re-organize subunit architecture1.

It remains to be determined how the head, middle, and tail modules of the S. cerevisiae Mediator complex re-organize upon binding to pol II and other components of the PIC. Initial EM studies of S. cerevisiae Mediator showed evidence for a dramatic structural re-organization upon interaction with pol II3. It is also unknown which or how many Mediator subunits directly interact with the pol II enzyme upon PIC assembly. Recent success with reconstitution and cryo-EM 3D reconstruction of the S. cerevisiae PIC lacking Mediator suggests that answers to these and other fundamental questions will soon be revealed10.

Although the data presented by the Cai group5 convincingly establish an updated structural model of the S. cerevisiae Mediator complex, it is noteworthy that the structural resolution is not better than previous studies. This is certainly not a reflection of the biochemical purification or the EM data, but rather results from the intrinsic flexibility and disorder of the Mediator complex2. Low-resolution structural data for the 21-subunit S. cerevisiae Mediator complex continues to limit understanding of its mechanism of action. Recent breakthroughs in EM instrumentation, such as direct electron detectors, and increasingly sophisticated image processing capabilities, offer great promise for improved understanding of this large, dynamic complex. Through the efforts of the Cai group and others, the complexities of Mediator structure and function will continue to be unraveled.