Page 121 - Read Online

Page 10 of 16 Varming et al. Microbiome Res Rep 2024;3:15 https://dx.doi.org/10.20517/mrr.2023.50

To satisfy our curiosity, we produced a model of full-length TP901-1 CI by AF2 without using a template
structure and compared the CI-NTD region to the CI-NTD region in the experimental complex.
Interestingly, in the computational model, both Gln55 and Glu69 side chain positions are compatible with
complex formation with Mor, though clashes arise at the N-terminus of CI.

Purification of φ13 Mor
[28]
We have previously purified and partly characterized φ13 CI . Here, we produce, purify, and characterize
for the first time φ13 Mor [Figure 3]. As observed for TP901-1 Mor, φ13 Mor elutes as a monomer in gel
filtration [Figure 3B] based on a comparison with a standard calibration curve.

Interaction of φ13 CI and Mor in solution
The possible interaction between φ13 CI and Mor was studied using isoelectric focusing (IEF)
electrophoresis. CI and Mor have theoretical pI values of 5.24 and 8.63, respectively, and a complex between
CI and Mor would thus be expected to have an intermediate pI. Each protein was run alone (lanes 2 and 3
in Figure 3C) and in complex with its counterpart at different ratios (lanes 4-6 in Figure 3C). Both protein
bands are smeared, making it difficult to establish a specific experimental pI. CI shows a stronger band just
below the standard at pH 6.0, while Mor shows at least two strong bands between the standards at pH 6.0
and pH 6.9. This could indicate conformational heterogeneity, consistent with a highly dynamic molecule as
previously observed for TP901-1 Mor. We do, however, see a clear difference between CI and Mor alone
and in complex. The bands arising from Mor in lane 3 disappear when mixed with CI in a 2:1 CI:Mor
complex, indicating that Mor interacts with CI. However, this complex also indicates an excess of CI as the
bands arising from CI in lane 2 are still visible, albeit less intense. This excess decreases in the 1:1 ratio
complex and seems to disappear completely in the 1:2 CI:Mor complex, suggesting that most CI and Mor in
solution have formed a complex, and it is a well-defined complex, since a single band is observed. The
binding stoichiometry of the CI:Mor interaction cannot be concluded unambiguously from these results,
also due to the uncertainties in the estimation of protein concentration, but it is clear that φ13 CI and Mor
interact in solution, supporting the notion that CI and Mor in φ13 regulate the genetic switch in a similar
way to their homologs in TP901-1.

Computational modeling of the φ13 CI-NTD:Mor complex
After demonstrating through experiments that there is interaction between φ13 CI and Mor, our focus
shifted to modeling the complex. Given the high sequence conservation between CI-NTD and Mor in
TP901-1 and φ13 (> 60% sequence identity), the interactions may be expected to be strikingly similar. Thus,
we initially constructed a model of the complex, by superposing models of φ13 CI and Mor from the AF
Database to the experimental structure of the TP901-1 complex. However, this resulted in severe clashes at
the interface between φ13 CI-NTD and Mor [Supplementary Figure 2], a major one being CI Tyr75
(corresponding to Phe75 of TP901-1) clashing with several residues of Mor including Trp49 (TP901-1
Leu49). The overall Molprobity clash score calculated at the Swissmodel server (https://swissmodel.expasy.
org/assess) was 27, with most clashing residues situated at the interface. Given that the TP901-1 CI:Mor
interaction involves several termini and side chain rearrangements, these clashes are perhaps not surprising,
but they diminish confidence in the results.

Previous attempts to generate a heterodimer model using an earlier version of AF2 yielded a very different
interaction interface and heterodimers compared to the experimental CI-NTD:Mor heterodimer for TP901-
1. Given the continuous progress in the AF2 computational pipelines, we recently attempted again, both
using 6TRI as a template and in no template mode, in both cases resulting in complexes similar to the
experimental TP901-1 heterodimer. The clash scores of unrelaxed models were still high (22-24), but now
the clashes were not concentrated at the interface. Relaxation of the best-scoring AF2 model produced with