RR:C19 Evidence Scale rating by reviewer:
Marqusee et al. present a very intriguing dynamics analysis of several stabilized SARS-CoV-2 spike proteins, indicating the existence of an otherwise undetected extended conformation of the trimer. This conformational change appears to involve an unhinging event around the S1/S2 domain boundary. The transition is suggested to be distinct from the Up/Down dynamics seen in the RBD (which is essential for engaging ACE2).
The study uses HX-MS technology in a most satisfying manner. Bimodality in several deuteration signals localizes to a region of structure that indicates the potential of a “loosened” state, and these signals are convincingly attributed to (slowly) interconverting states. Departures from unimodal exchange events can be difficult to explain with confidence, but the authors present several tests that confirm this state interconversion. The pulsed labeling/reversibility test and the binding of the 3A3 antibody are particularly informative in this regard and help discriminate the effect from a purely EX1 exchange scenario. However, I imagine that the state change is more complex than the simplified model shown in Figure 5. Note that many of the peptides showing bimodality don’t show further increases in deuteration in the B state distribution after the first labeling timepoint, strongly suggesting an unfolded B state (or sub-states). Further, some peptides do show a more characteristic EX1-like exchange pattern (eg. 662-673). Together, this suggests to me a concerted transition from the prefusion state to an open trimer that is much more unstructured than is depicted in Figure 5. Nonetheless, it’s clear that such a transition does occur, as the experiments elegantly demonstrate.
The authors use stabilized forms of the spike protein, which may explain the slow kinetics of the interconversion. A recent HX-MS study by Anand et al. did not detect this interconversion, which the authors attribute to different methodology. While this may be the case, it is also possible that the dynamics are affected by the choice of mutations used to stabilize the prefusion state (these appear different in the two studies). The Anand study pointed to the allosteric communication between the RBD and the S1/S2 region so the potential for altered kinetics is there. It raises the question: is the two-state system described in the Marqusee study an artifact of the engineered form? I do not think this is the case. Again, the data from the 3A3 antibody is convincing. If anything, the open trimer is likely more readily populated in wild type. A plausible model involves the proline-stabilized form sitting at slightly lower energy than the open trimer (and certainly at lower energy than the wild type). An increase in temperature may drive enough energy into the system to crest the activation barrier and return the protein to a more stable prefusion state. The model presented in Figure 5 is quite sound: a transient population of three RBD Up states may reduce the barrier to an expanded open trimer leading to the spike getting “caught out” in the open trimer.
The study clearly justifies wider exploration of the Spike protein’s topology in the pursuit of therapeutics (ligands and vaccines), raising the exciting possibility of a vaccine that is more resistant to antigenic drift. The claims are very well-supported by the data and methods used, present the work in the context of the most up-to-date findings, and are generally well written.