Aqueous fluoride ion is an omnipresent environmental xenobiotic that inhibits certain phosphoryl-transfer enzymes critical for energy production and nucleic acid synthesis (Marquis et al., 2003; Samygina et al., 2007). Two phylogenetically unrelated F^- efflux systems are now known that resist the toxicity of cellular F^-: the CLC^F family of F^-/H⁺ antiporters, and the Fluc family of F^- ion channels (Baker et al., 2012; Stockbridge et al., 2012). Fluc channels, present in all classes of life except higher animals, allow passive transit of F^- ions across the membrane down their electrochemical gradient. Bacterial Fluc channels are constructed as small, dual-topology dimers; the two subunits, each of ~120 residues, span the membrane in opposite orientations to form the functional complex (Stockbridge et al., 2013). These channels are notable for this unusual architecture as well as for exceptionally high F^-/Cl^-selectivity. Recent crystal structures of two Fluc channels (Stockbridge et al., 2015; Last et al., 2016) reveal four bound F^- ions whose transmembrane disposition imply two antiparallel ion-permeation pathways, each occupied simultaneously by two F^- ions (Figure 1a, Figure 1—figure supplement 1). Each pathway was postulated to be extremely narrow, largely anhydrous, and filled up by side chains that hand off F^- ions passing along it (Figure 1b). This unusual structural inference was confirmed functionally by substitutions of three conserved, ion-coordinating residues in this region – two phenylalanines and an asparagine – that fully inhibit F^- transport.

Figure 1 with 1 supplement see all

Download asset Open asset

Zooming in on a single pore of the Fluc homologue under study here (Figure 1b), we see two bound F^- ions separated by 12 Å and surrounded mainly by protein side chains. The lower ion approaches one of the wide aqueous vestibules on the two ends of the channel, while the other is deeply buried within the protein. The proposed critical pore-associated moieties are of two chemical types: aromatic rings of two conserved Phe residues contributed from different subunits, each of which coordinates a F^- ion in an electropositive edge-on orientation, and dipolar H-bond donors arranged to satisfy the F^- ion's H-bond accepting preference. These H-bonding residues were proposed (Stockbridge et al., 2015) to form a ‘polar track’ extending in both directions beyond the centrally located Phe rings, as indicated in the sequence alignment and structure of Figure 1. Most polar track residues emanate from TM4, where they appear at every 4^th position, and although not all are individually conserved, H-bond donors frequently appear here. Several of these coordinate F^- ions in the crystal structures, while others lie close to the aqueous vestibules where crystallographically ordered waters are also observed. We emphasize that the structures do not show water-filled aqueous tunnels lined by polar groups, as in many other ion channels; instead, the pores are filled with a web of side chains (Figure 1c) that must be negotiated by ions translocating at rates on the order of 10⁶–10⁷ ions/sec.

Our aim here is to further test this picture by mutating each of the residues thought to be involved in F^- transport and to determine their chemical requirements for channel function. Results confirm that the mutation-sensitive positions do indeed trace along the trajectory expected, but we were surprised to find that the polar character of several of these residues is not required for channel activity. Positions that are functionally sensitive to mutagenesis require a specific, finely tuned sidechain, and a conserved Thr residue unexpectedly uses its β-branched methyl group rather than its H-bonding hydroxyl. An additional surprise is that no other aromatic or canonical polar group substituted at the conserved Phe residues supports channel activity, but Met at one of these positions leads to a fully active channel; this seemingly strange result is rationalized by an electrostatic argument that the Met side chain mimics the Phe ring, a prediction confirmed by a crystal structure of the mutant.

A Fluc homologue from an E. coli virulence plasmid, nicknamed ‘Ec2’ (Stockbridge et al., 2013), was used for all experiments. Three strongly conserved, deeply buried residues spanning ~7 Å along the pore were previously found to be required for transport (Stockbridge et al., 2015; Last et al., 2016) in this and a different homologue: two Phe (F80, F83) and an Asn positioned between them (N41). To initially gauge the functional importance of the entire polar track, we mutated each of the four track residues on TM4 (S102, H106, S110, T114) to Ala as well as an additional residue, S84, which in the structure projects into the pore-region from TM3 of the partner subunit. Channel activity was assessed with an anion-efflux assay, wherein F^- or Cl^- transport by liposome-reconstituted Fluc was followed continuously with ion-specific electrodes (Stockbridge et al., 2013). Because F^- flux through WT channels is far faster than the electrode's time-response, this is effectively a binary assay of whether a construct is broadly functional or severely impaired (Figure 2a); even a channel with a 95% inhibited transport rate will display a WT phenotype, completely emptying its liposomes of F^- within the electrode's response time. For this reason, all mutants appearing active in the flux assay were also electrophysiologically examined in planar phospholipid bilayers for direct determination of single-channel conductance. Surprisingly, we find (Figure 2b) that F^- robustly permeates three track mutants denuded of their hydroxyl groups - S84A, S102A, and S110A - with single-channel conductances within 25% of WT (Figure 2c, Table 1). In contrast, H106A and T114A, as well as alternative H-bonding substitutes H106S, H106Y, H106W, H106N, H106F, N41S, N41Q, and N41Y are completely nonfunctional (Figure 2b, Figure 2—figure supplement 1), representing >10,000 fold rate-inhibition relative to WT. In crystal structures (Figure 1b) H106 coordinates a partially hydrated F^- ion located close to bulk solvent at one end of the narrow pore, T114 lies beyond the other F^- ion at the opposite end, and the other residues tested reside between them. These functionally sensitive positions thus chart out a ~ 20 Å span connecting the deepest parts of the aqueous vestibules, as suggested from the structures alone. We now turn to the chemical requirements for three of the positions vital for transport.

Figure 2 with 2 supplements see all

Download asset Open asset

F^--coordinating aromatics

The two F^--coordinating Phe residues stand out due to their unusual anion-quadrupolar coordination and their strong conservation. Despite the residue's hydrophobicity, its π-electrons leave a partial positive charge on the ring hydrogens (Mecozzi et al., 1996), such that each ring approaches a F^- ion in an edge-on orientation. Removal of aromaticity by substitution of either Phe with Ile was previously shown to abolish both transport and F^- occupancy at the site near the substitution (Last et al., 2016). We examined a variety of mutations to survey the chemical requirements for F^- binding and transport at these central Phe positions. Aromatics (Y, W, H) were introduced as conservative substitutions, aliphatic H-bond donors (S, T, Q) as potential F^--coordination partners of alternative chemistry, and hydrophobic aliphatics (L, A) as additional controls. All eight mutations, most of them tested at both Phe residues, completely abolish channel activity (Figure 3a, Figure 3—figure supplement 1). The most conservative variant, Tyr, is particularly striking since it differs from Phe only by the presence of the ring hydroxyl. A crystal structure of the F80Y mutant, solved to assess F^- binding and structure of the pore region, is essentially identical to WT (C_α rmsd ~0.3 Å) with the Tyr ring congruent to that of WT Phe (Figure 3b, Figure 3—figure supplements 2 and 3). In contrast to previous structures of inactive Ile mutants (Last et al., 2016), F^- remains bound to the Tyr construct in precisely the same positions as in WT (Figure 3—figure supplement 4). Other than the added hydroxyl group, no significant structural changes between WT and the inactive F80Y mutant are apparent, nor does the structure suggest a clear reason why F^- fails to permeate the mutant.

Figure 3 with 4 supplements see all

Download asset Open asset

While the coordinating Phe residues are strongly conserved among Fluc homologues, they are not perfectly so, as illustrated in the alignment of Figure 1d. Bacterial Fluc sequences present a complication that must be appreciated in interpreting alignments. While many Fluc proteins are homodimeric like Ec2, others are known or inferred heterodimers as in Lactobacillus acidophilus (La1, La2), Staphylococcus aureus (Sa1, Sa2), and Mycobacterium tuberculosis (Mt1, Mt2). This circumstance offers the possibility of genetic drift, such that one of the two pores becomes inactivated. As long as the second pore remains active, the inactivated pore can lose residues vital for transport with no physiological effect. In such cases, non-canonical residues appearing in the alignment might either reflect a novel transport chemistry, or simply represent a functionally disrupted pore that no longer has strict residue requirements. Such an arrangement has been suggested from sequence analysis (Stockbridge et al., 2015) and indicated experimentally in a yeast Fluc channel (Smith et al., 2015; Berbasova et al., 2017). Thus, both La and Sa heterodimers retain three of the central phenylalanines, but have either Leu or Met appear at the fourth position. Given that one pore in these dimers would still possess both phenylalanines, the altered residues may line an inactivated pore. The Mt heterodimer surprisingly loses both phenylalanines at the second position, replaced with either methionine or tryptophan. This leads to the unexpected hypothesis that that either Met or Trp are capable of coordinating and transporting fluoride in that construct.

Since alignments show Met occasionally appearing at the central Phe positions, we wondered if this might be functionally meaningful, despite ambiguities about heterodimeric sequences. In this context, Met can be viewed as a uniquely plausible non-aromatic substitute for Phe, since the electron-withdrawing sulfur is expected to produce a partial positive charge on the γ-methylene and terminal methyl groups, either of which might act as an electrostatic surrogate to coordinate F^-. Accordingly, each central Phe was mutated to Met and tested for F^- transport. Remarkably, F80M results in an active flux-phenotype (Figure 3c) and displays clean single-channel recordings with near-WT conductance (Figure 3d, Table 1). In contrast, F83M is inactive.

We determined crystal structures of both Met mutants to test the prediction above regarding F^- coordination. As with the Tyr substitute, the Met structures are virtually identical to WT (Figure 3b, Figure 3—figure supplement 2), and all residues surrounding the substitutions are unmoved (C_α rmsd ~0.3 Å from WT, Figure 3—figure supplement 3). The Met side chains in both take on a twisted conformation, filling approximately the same space as the Phe ring in WT. As predicted, in both Met mutants F^- ions are coordinated by the γ-methylenes located in nearly the same positions as the coordinating Phe ring carbons in WT (Figure 3—figure supplements 3 and 4), as though the substituted linear side chains attempt to mimic Phe’s aromatic ring in occupied volume and F^--binding electrostatics. As with the Tyr mutant, the similar structures offer no explanation why F80M is functionally active and F83M is not.

A paradoxically polar position

As described above, the moderately conserved Thr114 lies at one end of the polar track, distant from bound F^- ions, and its substitution by Ala abolishes channel activity, as if its hydroxyl group is necessary for ushering the anion into the narrow pore from the aqueous vestibule. However, a stark refutation of this idea emerges from the complete inactivity of T114S (Figure 4a), despite a T114S crystal structure showing the Ser hydroxyl exactly overlaying its WT counterpart and no significant structural changes elsewhere (C_α rmsd ~0.2 Å, Figure 4b, Figure 4—figure supplements 1 and 2). This perplexing result motivated an examination of Thr's β-branched methyl group as a functional determinant, and indeed Val and Ile were found to be fully active substitutes here in both liposome F^- efflux and single-channel recording experiments (Figure 4a,c, Table 1).

Figure 4 with 4 supplements see all

Download asset Open asset

What could be the function of the β-branched methyl group? In the WT structure, this contacts the Gly49 C_α, V85 C_α, and F88 side chain, all from the partner subunit (Figure 4—figure supplement 3). Thus, it is possible that the T114 methyl participates in a nonpolar interaction critical for stabilizing the conducting structure of WT (a structure that nevertheless is still observed in T114S crystals). We attempted mutating Phe88 to Ala (mutation of the backbone C_α’s being unfeasible) to weaken the interaction with the opposite subunit. The height of the F88A flux kinetic is clearly smaller than that of WT (Figure 4—figure supplement 4), demonstrating that a significant fraction of the channels, although apparently well-folded and dimeric off the gel-filtration column, are inactive. However, those channels that remain active have near WT levels of single channel flux (Figure 4—figure supplement 4, Table 1). These results are difficult to interpret, as the changes in the fraction of active F88A may be independent of any possible interaction with T114. Additionally, given that T114 also makes close contact with the F88 beta carbon (which would remain in the F88A construct, Figure 4—figure supplement 3), interactions vital for flux may still remain even after mutation. We therefore conjecture, without compelling evidence beyond the crystal structure itself, that Thr114 contributes to the pore’s structural integrity but does not coordinate F^- during its transit.

A striking feature of Fluc channels is their extreme (>10,000 fold) selectivity against Cl^- ions (Stockbridge et al., 2013). This may simply reflect steric hindrance arising from a narrow bore of the conduction pathway, but no evidence currently informs this question pro or con. Whatever the mechanism of Cl^- exclusion, though, it is preserved in all functional mutants described here; in no case do we observe any Cl^- leakage through these channels (Figure 2—figure supplement 2), a result further supporting the structural propriety of the mutant channels independent of their crystal structures.

Despite functional experiments accompanied by crystal structures at resolution high enough to locate F^- ions, revealing two parallel and independent pores, the F^- permeation pathway of Fluc channels remains mysterious. The plug of sidechains clogging the pore leaves no clearly visible ion-diffusion pathway, thus provoking our proposal for an unconventional ‘channsporter’ permeation mechanism (Stockbridge et al., 2015) whereby F^- passage is accomplished by concomitant movements of side-chain rotamers on a submicrosecond timescale. Experiments here have now delineated the critical residues along the length of the pore (Figure 5, Figure 5—figure supplement 1); its overall trajectory, from the solvent-accessible binding site at F83/H106 to the terminus near T114, is as suggested from the structures, but several of the new findings are unexpected.

Figure 5 with 1 supplement see all

Download asset Open asset

The requirements for the central Phe positions, which stand out prominently in sequence alignments and crystal structures, mirror what is seen in nature: only rare replacements of Phe. The failure of even Tyr to functionally substitute for Phe at either Ec2 position is particularly surprising in light of the F80Y crystal structure. The Tyr and Phe rings and the F^- ions adopt identical positions, and the Tyr hydroxyl points away from the bound F^-, its presence leaving other parts of the protein undisturbed. Modification of ring electrostatics by the polarizable hydroxyl may lead to a more tightly bound ion, inhibiting flux. If any sidechain movement is required during transport, the additional hydroxyl may introduce either steric clash or competing polar interactions in those alternate conformers. In the case of F80Y, favorable hydroxide-phenyl interactions with the neighboring F83 could also inhibit sidechain movement.

Also surprising is the full activity of the F80M substitution. This may be rationalized by the F^--coordinating γ-methylene, whose weakly electropositive character recalls the electropositive, edge-on presentation of the Phe ring to its coordinated F^- ion. The Met side-chain configuration in the F80M structure is in striking agreement with this idea, but the functional impairment of F83M, which offers a similar configuration, remains puzzling. F83 coordinates its fluoride in conjunction with H106, which is the only other residue in this construct that will brook no substitution. This raises the possibility that more precise orientation is required here than at F80. While the Ec2 homologue studied here is inflexible at this location, different coordination schemes are clearly possible; M. tuberculosis (Figure 1d) replaces both the histidine and phenylalanine. It remains to be seen whether these alternate arrangements continue to be as finely tuned and unaccepting of modification.

Other aliphatic mutations at these critical Phe positions are known to eliminate transport (Figure 3a, Figure 3—figure supplement 1, Table 1), and this prohibition now extends to dipolar H-bond donors, with neither Ser, Thr, His, nor Gln supporting channel activity. Previous removal of the central phenylalanines (F80I, F83I) led to a loss of fluoride density in crystal structures (Last et al., 2016), suggesting that the F83M and F80Y mutants inhibit transport via a mechanism fundamentally different from the simple loss of a binding site. The results establish that the center of the pore has an exceptionally strict requirement for F^- coordination and transport, possibly reflecting a need for a geometrically precise, electrostatically weak binding partner for rapid transit of F^- in this anhydrous region.

Along the conserved polar track, only two H-bond donors are required: N41 and H106. The N41 residue, among the most strictly conserved in the Fluc family, is oriented between the two F^--coordinating Phe, well positioned to hand off the anion between them. In coordinating a partially hydrated F^- at the end of the pore, H106 could catalyze dehydration as the ion leaves bulk water in the vestibule. It is surprising, even alarming, that for the hydroxyl-bearing track residues - S84, S102, S110, and T114 - polarity and H-bonding are not directly involved in anion permeation, despite the conservation of polarity at these positions and their proximity to bound F^- ions in the pore. This apparent discrepancy between structural and functional facts presents a mechanistic puzzle calling out for future attack.

1. 1. Adams PD
   2. Afonine PV
   3. Bunkóczi G
   4. Chen VB
   5. Davis IW
   6. Echols N
   7. Headd JJ
   8. Hung LW
   9. Kapral GJ
   10. Grosse-Kunstleve RW
   11. McCoy AJ
   12. Moriarty NW
   13. Oeffner R
   14. Read RJ
   15. Richardson DC
   16. Richardson JS
   17. Terwilliger TC
   18. Zwart PH

   (2010) PHENIX: a comprehensive python-based system for macromolecular structure solution

   Acta Crystallographica Section D Biological Crystallography 66:213–221.

   https://doi.org/10.1107/S0907444909052925

   • PubMed
   • Google Scholar
2. 1. Baker JL
   2. Sudarsan N
   3. Weinberg Z
   4. Roth A
   5. Stockbridge RB
   6. Breaker RR

   (2012) Widespread genetic switches and toxicity resistance proteins for fluoride

   Science 335:233–235.

   https://doi.org/10.1126/science.1215063

   • PubMed
   • Google Scholar
3. 1. Battye TG
   2. Kontogiannis L
   3. Johnson O
   4. Powell HR
   5. Leslie AG

   (2011) iMOSFLM: a new graphical interface for diffraction-image processing with MOSFLM

   Acta Crystallographica Section D Biological Crystallography 67:271–281.

   https://doi.org/10.1107/S0907444910048675

   • PubMed
   • Google Scholar
4. 1. Berbasova T
   2. Nallur S
   3. Sells T
   4. Smith KD
   5. Gordon PB
   6. Tausta SL
   7. Strobel SA

   (2017) Fluoride export (FEX) proteins from fungi, plants and animals are 'single barreled' channels containing one functional and one vestigial ion pore

   PLoS One 12:e0177096.

   https://doi.org/10.1371/journal.pone.0177096

   • PubMed
   • Google Scholar
5. 1. Chen VB
   2. Arendall WB
   3. Headd JJ
   4. Keedy DA
   5. Immormino RM
   6. Kapral GJ
   7. Murray LW
   8. Richardson JS
   9. Richardson DC

   (2010) MolProbity: all-atom structure validation for macromolecular crystallography

   Acta Crystallographica Section D Biological Crystallography 66:12–21.

   https://doi.org/10.1107/S0907444909042073

   • PubMed
   • Google Scholar
6. 1. Emsley P
   2. Lohkamp B
   3. Scott WG
   4. Cowtan K

   (2010) Features and development of Coot

   Acta Crystallographica Section D Biological Crystallography 66:486–501.

   https://doi.org/10.1107/S0907444910007493

   • PubMed
   • Google Scholar
7. 1. Evans PR

   (2011) An introduction to data reduction: space-group determination, scaling and intensity statistics

   Acta Crystallographica Section D Biological Crystallography 67:282–292.

   https://doi.org/10.1107/S090744491003982X

   • PubMed
   • Google Scholar
8. 1. Kabsch W

   (2010) XDS

   Acta crystallographica. Section D, Biological crystallography 66:125–132.

   https://doi.org/10.1107/S0907444909047337

   • PubMed
   • Google Scholar
9. 1. Last NB
   2. Kolmakova-Partensky L
   3. Shane T
   4. Miller C

   (2016) Mechanistic signs of double-barreled structure in a fluoride ion channel

   eLife 5:e18767.

   https://doi.org/10.7554/eLife.18767

   • PubMed
   • Google Scholar
10. 1. Marquis RE
    2. Clock SA
    3. Mota-Meira M

    (2003) Fluoride and organic weak acids as modulators of microbial physiology

    FEMS Microbiology Reviews 26:493–510.

    https://doi.org/10.1111/j.1574-6976.2003.tb00627.x

    • PubMed
    • Google Scholar
11. 1. McCoy AJ
    2. Grosse-Kunstleve RW
    3. Adams PD
    4. Winn MD
    5. Storoni LC
    6. Read RJ

    (2007) Phaser crystallographic software

    Journal of Applied Crystallography 40:658–674.

    https://doi.org/10.1107/S0021889807021206

    • PubMed
    • Google Scholar
12. 1. Mecozzi S
    2. West AP
    3. Dougherty DA

    (1996) Cation-pi interactions in aromatics of biological and medicinal interest: electrostatic potential surfaces as a useful qualitative guide

    PNAS 93:10566–10571.

    https://doi.org/10.1073/pnas.93.20.10566

    • PubMed
    • Google Scholar
13. 1. Samygina VR
    2. Moiseev VM
    3. Rodina EV
    4. Vorobyeva NN
    5. Popov AN
    6. Kurilova SA
    7. Nazarova TI
    8. Avaeva SM
    9. Bartunik HD

    (2007) Reversible inhibition of escherichia coli inorganic pyrophosphatase by fluoride: trapped catalytic intermediates in cryo-crystallographic studies

    Journal of Molecular Biology 366:1305–1317.

    https://doi.org/10.1016/j.jmb.2006.11.082

    • PubMed
    • Google Scholar
14. 1. Smith KD
    2. Gordon PB
    3. Rivetta A
    4. Allen KE
    5. Berbasova T
    6. Slayman C
    7. Strobel SA

    (2015) Yeast fex1p is a constitutively expressed fluoride channel with functional asymmetry of its two homologous domains

    Journal of Biological Chemistry 290:19874–19887.

    https://doi.org/10.1074/jbc.M115.651976

    • PubMed
    • Google Scholar
15. 1. Stockbridge RB
    2. Lim HH
    3. Otten R
    4. Williams C
    5. Shane T
    6. Weinberg Z
    7. Miller C

    (2012) Fluoride resistance and transport by riboswitch-controlled CLC antiporters

    PNAS 109:15289–15294.

    https://doi.org/10.1073/pnas.1210896109

    • PubMed
    • Google Scholar
16. 1. Stockbridge RB
    2. Robertson JL
    3. Kolmakova-Partensky L
    4. Miller C

    (2013) A family of fluoride-specific ion channels with dual-topology architecture

    eLife 2:e01084.

    https://doi.org/10.7554/eLife.01084

    • PubMed
    • Google Scholar
17. 1. Stockbridge RB
    2. Kolmakova-Partensky L
    3. Shane T
    4. Koide A
    5. Koide S
    6. Miller C
    7. Newstead S

    (2015) Crystal structures of a double-barrelled fluoride ion channel

    Nature 525:548–551.

    https://doi.org/10.1038/nature14981

    • PubMed
    • Google Scholar
18. 1. Turman DL
    2. Nathanson JT
    3. Stockbridge RB
    4. Street TO
    5. Miller C

    (2015) Two-sided block of a dual-topology F- channel

    PNAS 112:5697–5701.

    https://doi.org/10.1073/pnas.1505301112

    • PubMed
    • Google Scholar
19. 1. Walden M
    2. Accardi A
    3. Wu F
    4. Xu C
    5. Williams C
    6. Miller C

    (2007) Uncoupling and turnover in a Cl^-/H⁺ exchange transporter

    The Journal of General Physiology 129:317–329.

    https://doi.org/10.1085/jgp.200709756

    • PubMed
    • Google Scholar
20. 1. Winn MD
    2. Murshudov GN
    3. Papiz MZ

    (2003) Macromolecular TLS refinement in REFMAC at moderate resolutions

    Methods in Enzymology 374:300–321.

    https://doi.org/10.1016/S0076-6879(03)74014-2

    • PubMed
    • Google Scholar
21. 1. Winn MD
    2. Ballard CC
    3. Cowtan KD
    4. Dodson EJ
    5. Emsley P
    6. Evans PR
    7. Keegan RM
    8. Krissinel EB
    9. Leslie AG
    10. McCoy A
    11. McNicholas SJ
    12. Murshudov GN
    13. Pannu NS
    14. Potterton EA
    15. Powell HR
    16. Read RJ
    17. Vagin A
    18. Wilson KS

    (2011) Overview of the CCP4 suite and current developments

    Acta Crystallographica Section D Biological Crystallography 67:235–242.

    https://doi.org/10.1107/S0907444910045749

    • PubMed
    • Google Scholar
22. 1. Winter G

    (2010) xia2 : an expert system for macromolecular crystallography data reduction

    Journal of Applied Crystallography 43:186–190.

    https://doi.org/10.1107/S0021889809045701

    • Google Scholar

Author details

1. Nicholas B Last

   Department of Biochemistry, Howard Hughes Medical Institute, Brandeis University, Waltham, United States

   Contribution

   Conceptualization, Data curation, Formal analysis, Supervision, Validation, Investigation, Methodology, Writing—original draft, Writing—review and editing

   Competing interests

   No competing interests declared

2. Senmiao Sun

   Department of Biochemistry, Howard Hughes Medical Institute, Brandeis University, Waltham, United States

   Contribution

   Data curation, Formal analysis, Validation, Investigation, Writing—original draft

   Competing interests

   No competing interests declared

3. Minh C Pham

   Department of Biochemistry, Howard Hughes Medical Institute, Brandeis University, Waltham, United States

   Contribution

   Data curation, Formal analysis, Validation, Investigation, Methodology

   Competing interests

   No competing interests declared

4. Christopher Miller

   Department of Biochemistry, Howard Hughes Medical Institute, Brandeis University, Waltham, United States

   Contribution

   Conceptualization, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Methodology, Writing—original draft, Project administration, Writing—review and editing

   For correspondence

   cmiller@brandeis.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-0273-8653

• 2,066

  views

• 341

  downloads

• 16

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.