A revised model of how MED12 activates CDK8

In summarizing our in vitro and in vivo data, we propose a new model of how MED12 activates CDK8 (Fig. 59) (Klatt et al., 2020). In a first step, CDK8 is bound by Cyclin C, which leads to a “pushed-in” conformation of the αC-helix of CDK8, generating the kinase active site (Schneider et al., 2011). At this step CDK8 inhibitors were shown to trap the kinase in both a DFG-in conformation (type I) as required for catalysis (PDB code 5BNJ) (Dale et al., 2015) or in a DFG-out conformation (type II) that impedes catalysis and engages the deep pocket of the kinase (Fig. 60) (PDB code 3RGF) (Schneider et al., 2011). Binding of ATP-competitive type

Figure 59: A revised model of how MED12 activates CDK8 and remodels the active site of CDK8.

(A) Step 1: Cyclin C binds to CDK8 and pushed the aC-helix of CDK8 into the “pushed-in” conformation. This binding event is crucial for the formation of the active site of CDK8 and results in basal kinase activity as demonstrated in Fig. 21. Step 2:

MED12 binding to CDK8/Cyclin C stabilizes and activates the entire ternary complex. In particular, an activation helix in MED12 contacts and stabilizes the T-loop of CDK8, thereby activating the kinase. Likely, this contact is established through an interaction of an acidic residue at the N-terminal tip of the MED12 activation helix (E33) and the CDK8 arginine triad (R65, R150 and R178). Moreover, MED12 binding favors the active site of CDK8 to adopt a DMG-in conformation of the active site and disfavors type II kinase inhibitors from binding and inhibiting CDK8 in ternary CDK8/CycC/MED12 complexes. Please note that this model includes data that were not generated by myself. Adapted from Klatt et al., 2020.

I inhibitors is promiscuous due to the well conserved ATP binding site within the CDK family, increasing the potential for off-target side effects (Echalier et al., 2010; Force and Kolaja, 2011;

Kufareva and Abagyan, 2008). CDK8 was found to be the only known CDK that provides access to its deep pocket, enabling selective type II inhibitor binding (Schneider et al., 2013).

Figure 60: Views of the CDK8 active site in DMG-out and DMG-in conformation

Views of the CDK8 active site in DMG-out and DMG-in conformation (PDB codes 3RGF and 5BNJ, respectively). The CDK8 residues that are part of the arginine triad are shown as yellow sticks. The residues making up the DMG motif in CDK8 (D173-M174-G175) are shown in blue color. K52 and E66, two residues that are part of the catalytic triad of CDK8 (together with D173) and form a salt bridge once type I inhibitors bind to CDK8 in a DMG-in conformation, are shown in wheat color. The salt bridge is depicted as a dashed line- We note here that we found E66 to be crosslinked to MED12 K32 (Fig. 65A) and that we measured difference (fofc) density for the αC-helix by X-ray crystallography (Fig. 67). Adapted from Klatt et al., 2020.

We envision that the active site of the basal active CDK8/Cyclin C binary complex can exhibit both a DFG-in, as well as a DFG-out conformation. Next, the N-terminal segment of MED12 wraps around CDK8, thereby placing its "activation helix" right next to the T-loop of CDK8 (Klatt et al., 2020). Our mutational data indicate that the exact placement of this helix is crucial for activation of CDK8 by MED12. Binding alone is insufficient (Figs. 44 and 46). Moreover, our data indicate that the fact that MED12 wraps around CDK8 is important for its activity. A minimal fragment of MED12, for which we were able to isolate a stable, active ternary complex, comprised MED12 residues 23 to 69 (Fig. 61) (Klatt et al., 2020). This correlates well with

sites I and II (see chapter 6.2, Figs. 65A and C) and suggests that MED12 has to at least bind to these two sites for any stable association with CDK8. In more detail, our data suggest a direct contact between the arginine triad of CDK8 and E33 in MED12 (Figs. 39A and 44A). In support

Figure 61: Purification of ternary CDK8 (1-403)/CycC/MED12 complexes comprising numerous MED12 truncations The minimal stable truncation is MED12 (23-69), comprising the MED12 interaction site I and II. Shortening or eliminating the MED12 interaction II leads to hindered complex expression and thereby formation. Please note that ternary complexes were only affinity-purified and therefore MED12 still carries the Strep-tag.

of this contact, we also detected a zero-length crosslink between K32 in MED12 and E66 in the aC-helix of CDK8 (see chapter 6.2, Figs. 65A and C). K32 is situated right next to E33, the residue critical for MED12 function, and E66 is close to R65, one of the three members of the CDK8 arginine triad. As E66 is pointing towards the interior of the CDK8 molecule in binary CDK8/Cyclin C complexes, a further rearrangement of the aC-helix upon MED12 binding to CDK8 is likely (Fig. 60). We note here that E66 is part of the CDK8 catalytic triad. In addition to our crosslinking data, we detected negative fofc difference density in proximity of the aC-helix in a preliminary structure solution of a ternary CDK8/CycC/MED12 complex (see chapter 6.3, Fig. 67), which further supports a MED12-dependent rearrangement of the CDK8 active site. Finally, we hypothesize that the contact between the arginine triad and the activation helix of MED12 leads to a stably "folded-away" conformation of the CDK8 T-loop, allowing unobstructed substrate binding to the active site (Klatt et al., 2020). We note here, that the CAK assembly factor MAT1 was recently shown to utilize an extended helix that positions the T-loop of CDK7 in a comparable stably “folded-away” and fully active conformation (Greber et al., 2020).

The low-resolution cryo-EM map of the entire kinase module from Saccharomyces cerevisiae (Tsai et al., 2013) was interpreted to show that MED12 contacts Cyclin C only (see chapter 1.3, Fig. 10A). In contrast, our structural, functional and mutational data indicate the essential role of the MED12 N-terminal segment and the CDK8 αC-helix for contact formation.

This apparent discrepancy may be due to the limited resolution (15 Å) of the cryo-electron microscopy map that cannot resolve the extended shape of the MED12 N-terminal segment and its detailed interactions with the kinase (Klatt et al., 2020).

3.2 Future CDK8 inhibitors need to be developed against MED12-bound CDK8