PREreview of Activating alternative transport modes in a multidrug resistance efflux pump to confer chemical susceptibility

Activating alternative transport modes in a multidrug resistance efflux pump to confer chemical susceptibility

Summary

EmrE continues to harbor mysteries, as this manuscript demonstrates. The authors have in the past productively and rigorously examined it as a model for promiscuous drug transport, as well as identifying where it fails to be well-behaved. They continue here by investigating a new substrate, harmane, for which EmrE potentially induces sensitivity rather than resistance. They explain this in the framework of their free-exchange model, in which all possible conformational states and transitions are allowed, with overall transport the result of differential kinetic accessibility. Their prior work has demonstrated the relevance (Robinson et al. 2017) and theoretical novelty (Hussey et al. 2020) of this model; the current manuscript is a validation of their prediction that certain combinations allow sensitivity. This could potentially be of wide interest, both to the transporter community, by showing a surprising reversal of expected phenotype, and to antibiotic research, where it may point towards a new class of antimicrobial.

The current framing is too mechanistically ambiguous about the role of harmane, which undercuts this potential significance. Some more aggressive speculation is probably warranted here. Additionally, I have some questions about the controls for the SSME experiments, as well as some confusion about the overall physiology of the E14Q mutant and how it interacts with their interpretation of the biology and harmane data.

Overall, I believe that harmane might be interacting in a unique way with EmrE, and that extending their SSME stoichiometry method to uncharged compounds could be very powerful, but both require some additional work to fully demonstrate.

Major issues

Framing and interpretation

My major issue is with the conclusions and overall framing. What exactly is harmane doing? Is EmrE a harmane-gated proton transporter? Or is it a symported toxin? The authors seem to suggest the former, but don’t explicitly say it. SSME might not the best tool to demonstrate this, of course. Perhaps their mass spectrometry technique (Robinson et al, 2018) would be appropriate?

Either possibility would be extremely interesting! By avoiding mechanism, however, they reduce the impact of this work. This also means they can’t use their free exchange framework, so this doesn’t fully serve as a validating case, nor can it provide boundaries to the biochemical parameters that could produce the observed outcome.

Proton leak

A second concern is the question of proton leak in WT Emre. SI Figure 1 b shows that in the absence of drug, WT EmrE has a substantial growth defect relative to the E14Q mutant, roughly 0.2 OD units at 14 hours. The authors explain this as resulting from WT leaking protons, thus reducing the overall fitness of expressing cells. However, elsewhere the authors state that WT EmrE does not spontaneously leak protons. This would seem to contradict these results. It could be that the leak is slow enough on the SSME timescale to be disregarded, but this should be clarified. But from Figure 5, E14Q transports essentially 0 charge in the no drug condition, but WT an appreciable amount. This could represent a single turnover, rather than a productive leak current, but again, this requires some explicit argument. Given the singular importance of proton currents here, as they are the only reporters of transport, absolutely nailing down the phenotype is essential.

SSME methods

Finally, I have some questions about the SSME experiments. Overall, the framework for measuring stoichiometry they have developed is elegant and well-suited to the specific questions studied here, where they could be extremely powerful. Adapting them to an uncharged substrate requires some additional controls to provide confidence, however.

As SSME measures only the electrogenic components of transporter activity, applying it to the uncharged harmane requires some care. First, as a hydrophobic compound, I would expect harmane to non-specifically interact with the lipid mono/bilayers in the sensor well. Without a charge, though, this could not be detected. Comparing WT vs E14Q liposomes is a good control, but also comparing liposomes with no protein should be done.

Any currents observed will necessarily arise strictly from the coupled protons rather than the substrate itself. The framework the authors developed for measuring transporter stoichiometry with SSME (Thomas et al, 2021) requires some adjustments to account for this. In particular, Figure 4 is not quite adequate to interpret results with harmane. It assumes a cationic drug, which harmane is not. It also assumes 2:1 antiport and 1:1 symport stoichiometries, which may not hold. The cartoon only shows charged drugs!

From looking at the raw currents in SI 4 C in addition to the total charges in 4 C, the outward-facing and no gradient traces are more or less overlapping, while the inward-facing case shows both a decreased initial magnitude and a much slower decay. This is in contrast to the MeTPP case, where the decay rates across all traces are roughly similar, with only overall magnitudes changing. This points towards more complicated behavior, as the current decay reflects steady turnover while the initial amplitude reflects Vmax. From this, the proton import rate has a complex harmane gradient dependence that requires more investigation.

Minor points

1 - How were the differences in growth capacity as determined using OD controlled in the phenotypic microarrays? The authors explain that “known non-hits” were used to calculate the standard deviation for determining significance of the delta area score, but this should still result in an overall negative bias to the scores? Was a delta score calculated for wells without any substrate, for example?

2 - Understanding that the biolog data is more of a screening tool than a quantitative one, I am hesitant to submit it to too much criticism, but some more quantification of the significance calculations would be useful. More statistical information should be reported, including the means and variances of both each areas and the delta area, and the concentration dependence should be clarified, too (did hits display the expected concentration dependence?).

3 - For the microplate growth assays, I would be very worried about evaporation during 20 hours of growth in 384-well plates. This would arguably affect each well similarly, and so not bias their overall results, but I wonder if this was noticed by the authors?

4 - The pTrc promoter is susceptible to catabolite repression. In their phenotyping assay, plates PM3-8 were supplemented with 20 mM glucose, which should be sufficient to inhibit expression. Was this noticed?

5 - Some additional details about SSME methods should be included. What was the flow rate for the SSME experiments? What was the specific conductance and capacitance range used for QC acceptance? How many times was each sensor used? How many sensors were used for replicates, how many perfusions were conducted per sensor, etc?

Figure 3

Panel A is very difficult for me to interpret. Certainly the spectra are different, but I don’t know what else to make from it. The text states that CSPs for “a subset of peaks” indicate specific harmane binding, but that is not at all discernable from 3 A.

Could the authors assay whether E14Q binds harmane using Trp fluorescence? That might provide some more mechanistic insight into the substrate/proton interaction going on. 

SI Figure 3

I also just want to point out how interesting this is! Each substrate seems to work a different way – harmane, as suggested, slows WT EmrE growth during later exponential growth, while 18-crown-6-ether extends lag phase and slows exponential growth. Chelerythine chloride seems to mainly extend lag phase (for E14Q), but could potentially lead to similar final OD for E14Q over a longer timescale. This is a really powerful feature of this dataset that the authors should emphasize!

SI Figure 4

It looks like there might actually be some concentration dependence to E14Q in panel F – the magnitudes are definitely low, but plotting the charge vs [harmane] to prove this should be done.

Competing interests

I have published on EmrE and the SMRs and was a recipient (via my advisor Randy Stockbridge) of a Nanion Research Grant with second author Nathan Thomas