(Received for publication, April 24, 1997, and in revised form, June 25, 1997)
From the School of Biological Sciences, University of
East Anglia, Norwich, NR4 7TJ, United Kingdom, ¶ Protein Structure
Group, Zeneca Pharmaceuticals, Alderley Park, Macclesfield, Cheshire,
SK10 4TG, United Kingdom, and
School of Chemical Sciences,
University of East Anglia, Norwich, NR4 7TJ, United Kingdom
The colicin DNase-specific immunity proteins
interact with the endonuclease domain of the bacterial toxin colicin E9
with dissociation constants that span the millimolar to femtomolar affinity range. Among the non-cognate interactions Im2 shows the strongest binding toward the E9 DNase domain with a
Kd of 108 M, 6 orders of
magnitude weaker than that of the cognate immunity protein Im9. Based
on a NMR structure of Im9 that shows it to be a 4-helix protein, we
have conducted a mutagenic scan in which elements of Im9 secondary
structure were substituted into Im2 to precisely delineate regions that
define specificity. Eleven chimeras were constructed, and their
biological cross-reactivity toward colicins E2 and E9 was evaluated.
From this set of mutants seven proteins were purified, and the
Kd for their interaction with the E9 DNase domain
was measured by a combination of stopped-flow fluorescence and subunit
exchange kinetics. Our results show that immunity specificity is
dominated by residues on helix II, accounting for 5 orders of magnitude
binding specificity relative to Im2, and that packing interactions of
helix II with its neighbor helix I and the loop connecting helix III
with helix IV play minor roles. The conformational stability of these
chimeric proteins was also determined. Proteins displaying an Im9
phenotype were all more stable than the parent Im2 protein, and
surprisingly some chimeras were significantly more stable than either
Im2 or Im9.
Understanding specificity in protein-protein recognition is fundamental to our knowledge of complex biological processes such as transcription, signal transduction, and immune recognition, yet little is known about how specificity is determined in protein-protein interactions. In this paper we describe the use of homologue-scanning mutagenesis to determine the main structural elements that define the specificity of the immunity protein Im9 in its interaction with the DNase derived from the bacterial toxin colicin E9 (ColE9).1
Colicins are plasmid-borne toxins produced by the Enterobacteriacae that have lethal action against other related strains and are classified into groups on the basis of the cell surface receptor to which they bind (reviewed by James et al. (1)). The focus of our work has been the E group colicins, which bind to the product of the chromosomal btuB gene, an essential component of the high-affinity transport system for vitamin B12 in Escherichia coli (2). Following receptor binding, the toxin translocates into the cell and initiates cell death. Each E colicin plasmid codes for the production of a ~61-kDa colicin toxin, a 9.5-kDa inhibitor (immunity) protein that protects the producing cell against the cytotoxic activity of the toxin, and a small lipoprotein (the lysis protein) that releases the resulting 71-kDa heterodimeric colicin complex from the bacterium.
Colicin-producing cells are naturally resistant to the action of their own toxin but sensitive to the action of other bacteriocins from the same family. This forms the basis for the so-called immunity test by which cells producing different colicins can be identified. Using this biological test the E group has been subdivided into 9 types (ColE1-ColE9), and these fall into three cytotoxic classes (3, 4); the periplasmic membrane-depolarizing (or pore-forming) toxin ColE1 (5, 6), the RNases ColE3, ColE4, ColE5, and ColE6 (7-9), and the DNases ColE2, ColE7, ColE8, and ColE9 (9-13).
We have been studying the DNase colicins and their interaction with
immunity proteins as a model system for investigating specificity in
protein-protein recognition. The four colicins are almost identical in
sequence in the N-terminal regions of the protein that are involved in
translocation and receptor binding, but share only ~80% sequence
identity in the C-terminal DNase domains. Hence, the specific immunity
proteins Im2, Im7, Im8, and Im9 have evolved to counteract the toxicity
of each DNase colicin, and these share ~50% sequence identity (9).
Im9 binds to either ColE9 or the isolated 15-kDa E9 DNase domain with a dissociation constant of 1016 M in the
absence of salt (pH 7.0 and 25 °C) rising to 10
14
M in 200 mM salt (14). Association between the
two proteins is diffusion controlled, wherein the proteins are
electrostatically steered toward one another to produce an encounter
complex, which then undergoes a conformational change to produce the
final stable complex (14). The other immunity proteins of this family
can also bind the E9 DNase, but with much weaker affinities; the
equilibrium Kd values for these complexes are
10
4 M for Im7, 10
6
M for Im8, and 10
8 M for Im2. In
each case, binding results in inhibition of enzymatic activity, and the
binding affinity is governed by the rate of complex dissociation
(15).
Our knowledge of the regions of an immunity protein that determine its
specificity is limited to the results of homologous recombination
experiments between Im9 and Im8 (16). This early work focused on the
biological activity of chimeric immunity proteins and showed that the
primary determinants for specificity were located in the N-terminal
half of the protein. Moreover, substituting Val-34 in Im9 with aspartic
acid, the amino acid at this position in Im8 provided the resulting
mutant Im9 with some ColE8 cross-reactivity in in vivo
assays. Subsequent structural work has shown that Im9 is a 4-helix
protein in which position 34 is part of helix II (17, 18) (Figs.
1 and 2).
Chak et al. (19) have recently proposed, based on the
crystal structure of the Im7 protein, that helices I and II are both
involved in determining specificity as well as loop 2 and possibly loop
1 (Fig. 1). In large part their model is based on the observation that
these are the most variable sequences in the immunity protein family
(Fig. 2). However, this is at variance with recent isotope-edited NMR
experiments in which the amide resonances of 15N-labeled
Im9 bound to unlabeled E9 DNase (forming a 25-kDa complex) were
assigned. These data show that the major regions of perturbation on
binding the DNase are helices II and III, the latter being conserved in
the immunity protein family (20).
The present work set out to use homologue-scanning mutagenesis to identify unambiguously the elements of immunity protein secondary structure that determine colicin specificity. With Im2 as our template, we have substituted structural elements from Im9 into Im2 and characterized the resulting chimeras using both in vivo biological toxicity tests and in vitro binding assays. We find that specificity in this system is multifactorial but dominated by residues from a single helix.
E. coli strain
JM83 Hsd R, a colicin-sensitive restriction-deficient
derivative of E. coli JM83 (Ara+, Lac+, Pro, thi,
rpsL, 80
Lac ZM15), was used as the host strain. The
expression vector pTrc99a was purchased from Pharmacia Biotech Inc.;
pRJ345 and pCD01 were the resulting derivatives of this vector
containing the imm9 and imm2 genes encoding Im9
and Im2, respectively. Both constructs were used as templates for the
mutagenic scan reported in this study.
Mutagenesis was carried out
by modifying the megaprimer method of Sarkar and Sommers (21). Primers
were synthesized at each junction site between imm2 and
imm9. They were designed so that the 3 half of the primer
can anneal to one imm gene template (e.g. imm2),
while the 5
half is the upstream sequence of the other immunity gene.
Thus, in the first polymerase chain reaction using imm2 as
template, the resulting product will have an imm9 sequence
overhang. This product is then used as a "mega" primer for a second
polymerase chain reaction, although the sequence that actually primes
this reaction is only around 20-30 nucleotides in length. The second
polymerase chain reaction product was cloned into pTrc99a and
transformed into E. coli JM83. Transformants were screened
by streaking out on Luria-Bertani (LB) plates with 100 µg/ml ColE2 or
ColE9. The surviving clones were checked by SDS-polyacrylamide gel
electrophoresis after IPTG induction and double-stranded DNA sequencing
using an ALF sequencer (Pharmacia).
The biological activity of each chimeric immunity protein toward ColE2 and ColE9 was tested using a modified procedure of a plate assay described previously (15). 22 × 22-cm LB agar plates containing 100 µg/ml ampicillin were divided into lanes and each lane overlaid with a lawn of exponentially growing E. coli JM83 cells producing a different immunity protein. Two parallel plates were used, one induced with IPTG (1 mM) and the other without. Aliquots (2 µl) of 5-fold serially diluted ColE2 and ColE9 from 3 mg/ml to 1.5 ng/ml were spotted on top of the cells. The plates were incubated overnight at 37 °C and scored the following day for biological protection against the toxin.
Protein Purification and Protein DeterminationsJM83 cells
that had been transformed with wild type and mutant immunity genes
cloned into pTrc99A were grown in LB broth containing ampicillin (100 µg/ml) and were induced by the addition of 1 mM isopropyl
-D-thiogalactoside at an optical density at 550 nm between 0.6-0.8. The immunity protein was purified as described previously for Im9 (22). The E9 DNase was purified according to the
protocol described by Garinot-Schneider et al. (23). Protein
concentrations were determined using the molar absorption coefficients
of 17,550 M
1 cm
1 for the E9
DNase and 11,400 M
1 cm
1 for
Im9, Im2, and the chimeric immunity proteins (14).
The masses of all of the purified proteins were verified by mass spectrometry using a VG platform electrospray mass spectrometer. The lyophilized proteins were dissolved in high pressure liquid chromatography grade water, and the concentrations were adjusted to about 0.1 mg/ml. Formic acid was added to a final concentration of 1% just before the injection into a mobile phase of water:acetonitrile:formic acid (1:1:0.001) at a flow rate of 5 µl/min. The scan range was from 700 to 1700 m/z, and 10 transients were collected and averaged for each sample with the raw data processed using the accompanying MassLynx software. Horse heart myoglobin (Sigma) was used as a calibrant. Each protein was analyzed at least twice.
E9 DNase Immunity Protein Dissociation ConstantsDissociation constants were obtained from the ratio of
the individual dissociation (koff) and
association (k1) rate constants at 25 °C in
50 mM Mops buffer, pH 7.0, containing 200 mM
NaCl and 1 mM dithiothreitol, as described previously by
Wallis et al. (14, 15) and Osborne et al. (20).
Association of the proteins was monitored by stopped-flow fluorescence
under pseudo-first-order conditions using a 4-20-fold excess of
immunity protein over E9 DNase (0.35 µM). The resulting
biphasic fluorescence traces were fitted to a double exponential
equation, and the bimolecular rate constant was obtained from linear
replots of the rate of initial fluorescence enhancement
versus the immunity protein concentration. Dissociation
kinetics for E9 DNase-Im2 complexes were obtained either from
radioactive subunit exchange (in which complexes were chased with
3H-Im9) for slow dissociation rate constants
(<103 s
1) or from fluorescence chase
stopped-flow (in which complexes were chased with a 6-fold excess of
Im9) for fast dissociation rate constants (>10
3
s
1).
Differences in binding energy between Im2 mutants and wild type Im2 binding to the E9 DNase were determined according to Equation 1,
![]() |
(Eq. 1) |
Immunity protein stabilities were determined by guanidine hydrochloride (GdnHCl) denaturation where protein denaturation was monitored by the change in tryptophan fluorescence. All of the experiments were carried out on a Shimadzu RF5000 spectrofluorimeter at 25 °C using an excitation wavelength of 295 nm. The excitation and emission bandwidths were both 5 nm. Each immunity protein (2 µM) was equilibrated at 25 °C in 50 mM potassium phosphate buffer, pH 7.0, containing GdnHCl (0-3 M) for a minimum of 2 h, and the fluorescence emission was measured at 354 nm. The relative stabilities of the immunity proteins were compared by the concentrations of GdnHCl required to cause 50% protein denaturation ([GdnHCl]50%).
Eleven chimeric immunity proteins were constructed and cloned into the expression vector pTrc99a, transformed into E. coli JM83, and their biological specificity toward colicins scored using an agar plate assay (see "Materials and Methods"). We have previously shown (15) that in the absence of IPTG, JM83 cells containing this expression vector and a non-cognate immunity gene cloned into it are completely sensitive toward the action of ColE9, whereas the same construct containing the imm9 gene is resistant toward the action of the toxin. Although SDS-polyacrylamide gel electrophoresis indicates the absence of any significant amounts of expressed protein under these conditions, the complete protection of cells containing the pTrc99A+imm9 construct is most likely due to "leaky" expression (15). On addition of IPTG, >25% cell protein is immunity protein, and now both the Im8 and Im2 containing cells begin to show non-cognate biological cross-reactivity with the latter being the strongest.
The sensitivity of JM83 cells (with and without IPTG induction)
containing each of the eleven chimeric immunity proteins generated in
this study were tested against serial dilutions of both ColE9 and
ColE2. The results for five of these constructs are shown in Fig.
3 along with the data for cells
expressing wild type Im2 and Im9. As in our previously published work,
both Im2 and Im9 are completely resistant to the action of their
cognate colicin with or without induction by IPTG (i.e. no
zones of clearance). However, whereas IPTG-induced Im2 provides
E. coli cells with partial protection toward ColE9,
IPTG-induced Im9 shows no such cross-reactivity toward ColE2 (Fig. 3).
When helix II of Im9 is inserted into Im2 (Im2-(30-44)Im9)
the resulting chimera behaves very much like Im9 in that it provides
complete protection toward ColE9 ± IPTG and is sensitive toward
ColE2 in the absence of induction. It differs from Im9, however, since
on induction with IPTG it still retains some residual ColE2
cross-reactivity. By contrast, when the N terminus of Im9 was inserted
into Im2 (Im2-(1-29)Im9) the resulting chimera behaved
exactly like Im2 toward both colicins with or without induction
suggesting that helix I is not directly involved in specificity. A
further chimeric protein in which the N-terminal half of Im9 was fused
to the C-terminal half of Im2 (Im2-(1-44)Im9) showed
exactly the same biological phenotype as Im2-(30-44)Im9.
This latter result is in agreement with our previous homologous recombination experiments between Im9 and Im8, which showed that residues in the N terminus of the protein were the most likely determinants of biological specificity (16). The present data show for
the first time that of the two helices in the N-terminal half of the
protein, helix II is responsible for these differences in
specificity.
An intriguing property of both the Im2-(30-44)Im9 and Im2-(1-44)Im9 chimeras is that they retain some residual cross-reactivity toward ColE2 even though Im9 itself does not show this behavior (Fig. 3). The region of the protein responsible for this cross-reactivity was identified by introducing further structural elements of Im9 into Im2-(1-44)Im9. It was not until the loop connecting helix III and helix IV was introduced (Im2-(1-64)Im9) that this residual cross-reactivity was finally lost, and the protein behaved like Im9 ± IPTG (Fig. 3). Interestingly, the same phenotype could be obtained by combining the helix II and loop 2 substitutions (Im2-(30-44, 56-64)Im9) but not helix I and loop 2 (Im2-(1-29, 56-64)Im9) or loop 2 on its own (Im2-(56-64)Im9) both of which behave essentially like Im2. A summary of the biological phenotypes of the bacterial cells expressing the chimeric constructs is given in Table I.
|
Although a total of eleven chimeric immunity proteins were constructed (Table I) and induced by IPTG to approximately similar levels as deduced from SDS-polyacrylamide gel electrophoresis (data not shown), only those which showed representative changes in biological specificity were characterized further. Seven chimeras were purified and their masses confirmed by electrospray mass spectrometry. In each case the observed mass of the protein corresponded to the predicted mass ± 1 Da (data not shown).
The Kd for each purified chimeric immunity protein
binding the E9 DNase was obtained from the ratio of the individual dissociation and association rate constants,
koff and k1,
respectively, as summarized under "Materials and Methods"
(14, 15). Association kinetics were monitored by stopped-flow
fluorescence, and dissociation kinetics were monitored by subunit
exchange methods. The association of Im2 and Im9 with the E9 DNase is
biphasic in stopped-flow fluorescence experiments in which a
fluorescence enhancement representing the bimolecular collision is
followed by a fluorescence quench that is thought to emanate from a
conformational change in the complex (14). The rate constants for both
of these processes (in Mops buffer at pH 7.0, 25 °C and containing
200 mM NaCl) are very similar for both Im2 and Im9 where
k1 ~ 6-9 × 107
M1 s
1, and the conformational
change is ~4-5 s
1 (15). Very similar values were
obtained for each of the seven Im2/Im9 chimeric immunity proteins (data
not shown) indicating that the association kinetics of these proteins
do not play a significant role in determining the stability of the
resulting complexes with the E9 DNase. However, the complexes differed
quite significantly in the rate of complex dissociation; some behaved like Im2 (for example, Im2-(1-29)Im9), which has a
koff of ~1 s
1, whereas others
displayed off rates similar to that of wild type Im9 (for example,
Im2-(19-44)Im9), which has a koff
of ~10
6 s
1. The equilibrium dissociation
constants for each of the seven purified chimeric immunity proteins are
shown in Table I along with data for wild type Im2 and Im9. Also shown
in Table I are values for
G, which have been
calculated relative to Im2 binding the E9 DNase. Hence the
G for Im9 is
8 kcal/mol, reflecting the difference
in specificity between these two proteins for the E9 DNase.
The thermodynamic data in Table I show clearly that helix I makes no
contribution toward E9 DNase binding specificity
(G = 0), and this matches the biological
phenotype of the cells expressing the Im2-(1-29)Im9
mutant. In contrast, helix II contributes ~6 kcal/mol toward E9 DNase
binding specificity, and this corresponds to a complete switch in the
biological phenotype of the protein. As further sequences toward the C
terminus of Im9 are included in Im2 (Im2-(30-64)Im9)
binding is reduced a little, and at the same time the residual cross-reactivity toward ColE2 is lost. Conversely, as helix II is
combined with sequences toward the N terminus, including loop 1 and
part of helix I, binding affinity increases and even seems to surpass
that of the cognate immunity protein Im9. Nevertheless, the residual
biological cross-reactivity toward ColE2 is retained in these
chimeras.
Little is known about the
thermodynamic stabilities of immunity proteins, and since the homologue
scan involved moving whole elements of secondary structure between
immunity proteins, it was of interest to compare the thermodynamic
stabilities of Im2 and Im9 with the chimeras. Guanidine hydrochloride
was used as a denaturing agent, and denaturation was monitored by
tryptophan fluorescence spectroscopy. The endonuclease-specific
immunity proteins possess a single conserved tryptophan at position 74, and the wavelength of maximum emission for Im2 and Im9 is the same (334 nm), increasing to 354 nm on denaturation. In the folded immunity
proteins the emission from Trp-74 is quenched, and when unfolded the
fluorescence increases 5-fold. This was used to construct denaturation
curves (as described under "Materials and Methods") and plotted as
the fraction of unfolded protein versus GdnHCl concentration. The denaturation curves for Im2 and Im9 along with some
of the chimeras are shown in Fig. 4. Both
immunity proteins and all of the chimeras exhibited single cooperative
unfolding transitions indicating the absence of any unfolding
intermediates. The midpoints of denaturation for all of the purified
chimeric and wild type immunity proteins are listed in Table I. The
data in Fig. 4 and Table I show that Im9 ([GdnHCl]50% = 1.71 M) is marginally more stable than Im2
([GdnHCl]50% = 1.52 M) but also show quite
dramatic differences between some of the chimeras and these two
immunity proteins. The only chimeric protein that did not show an
increase in E9 DNase binding affinity was Im2-(1-29)Im9,
and this was also less stable than Im2. The other chimeric Im2 proteins
that displayed Im9 phenotypes were all more stable than Im2. In many
cases, the chimeric immunity proteins were significantly more stable
than Im9 as well; for example, Im2-(19-44)Im9
([GdnHCl]50% = 1.81 M). The increasing
stability of the mutant proteins suggests that the Im9 sequences, which
have been grafted into the Im2 framework, have been accommodated better than the corresponding Im2 sequences. It is also worth noting that
residues 45-64 from Im9 have a stabilizing effect, which becomes
apparent from comparing the midpoints of denaturation for
Im2-(30-64)Im9 to Im2-(30-44)Im9 and
Im2-(1-64)Im9 to Im2-(1-44)Im9 (Table I).
The specificity of E group colicin immunity proteins for their target endonuclease toxins has yet to be explained, and so we chose to use homologue-scanning mutagenesis combined with biological and in vitro binding assays to determine the regions of the protein that govern its specificity. Homologue scanning was first used by Wells and co-workers (24) to identify the regions of human growth hormone that are required for specific binding to its receptor. Since then it has been used in a number of systems to localize functionally important regions in families of homologous proteins (25, 26). The basic premise of the technique is the substitution of sections of sequence from one protein with analogous sections from a related protein and then testing the resulting chimeras by a functional assay that discriminates between the two wild type starting proteins. Clearly for the strategy to succeed the two proteins must be sufficiently similar in three-dimensional structure to allow accommodation of the novel sequences and so yield chimeras that are stable and can be purified and characterized.
In the present work, we have used this technique to localize the elements of secondary structures of the immunity proteins Im2 and Im9, which define their colicin specificity. The solution structure for Im9 has been determined by NMR (18) and this structure was used as the basis for the scan (Fig. 1). Im2 and Im9 are identical in length (86 amino acids) and share 68.6% sequence identity (Fig. 2) and so it is likely that the structure of Im2 will be similar to that of Im9. Indeed, the structures for the other endonuclease specific immunity proteins Im7 and Im8, which show even less sequence identity to Im9 than Im2, have very similar 4-helix structures (19).2 A further point, which suggests that the Im2 and Im9 structures will be comparable, stems from the results of this work since a total of eleven chimeras were generated, and in each case a stable protein was made suggesting that there was no gross distortion of the protein fold.
Helix II, Loop 2, and Packing InteractionsOur data show that of the four helices in an immunity protein, helix II is the major determinant of colicin specificity that in the context of the binding of Im2 relative to Im9 contributes over 6 kcal/mol E9 DNase binding energy from a total of 8 kcal/mol. However, other sequences have to be added for full E9 DNase binding (Table I). The most important of these appears to be the C-terminal end of helix I (compare the binding energy data for Im2-(25-44)Im9 with Im2-(19-44)Im9). Interestingly, these sequences in isolation (in the form of the Im2-(1-29)Im9 construct) do not affect E9 DNase binding. In other words, the effect of helix I is not additive to that of helix II but is felt only in the presence of helix II. We conclude from these observations that it is the buried residues of helix I that elicit this effect, probably by changing the orientation of helix II. Since helix I and II pack against each other in the Im9 structure, and since both helices are largely variable in sequence between immunity proteins, it seems reasonable that packing interactions between them may play a role in specificity. Tight binding protein complexes are characterized by highly complementary and well packed binding interfaces (27), and so it is quite conceivable that such intramolecular packing interactions could affect intermolecular specificity. It is also noteworthy that packing interactions between these two helices affect the stability of the protein; adding residues 19-30 from Im9 to the Im2-(30-44)Im9 mutant increases [GdnHCl]50% by >0.2 M (Table I).
Although neither of the variable loops play important roles in specificity, the results for chimeras containing loop 2 do show some relatively minor effects that require comment. Coincident with the loss of residual ColE2 cross-reactivity of Im9-like chimeras, the inclusion of loop 2 from Im9 also seems to reduce E9 DNase binding in these proteins (Table I). This affect can be seen in two cases. First, it is seen in chimeras that seem to bind slightly more tightly to the E9 DNase than Im9 itself (Im2-(19-44)Im9 and Im2-(1-44)Im9). This affinity is reduced to wild type Im9 levels when loop 2 from Im9 is added (Im2-(1-64)Im9). Second, this region also reduces the binding of the mutant protein that contains helix II alone (compare Im2-(30-44)Im9 with Im2-(30-64)Im9). Taken together these results suggest a rather paradoxical situation in which loop 2 in Im2 imparts some specificity for its cognate E2 DNase (hence the residual biological cross-reactivity) as well as providing some binding energy for the non-cognate E9 DNase. Substituting this loop in Im2 with that from Im9 removes both effects. Loop 2 also seems to affect the thermodynamic stability of the immunity proteins since the [GdnHCl]50% of chimeras containing this loop is increased by 0.2-0.3 M (see "Results" and Table I).
The observations on specificity concerning loop 2 could be the result of direct effects where loop 2 is involved in DNase binding or indirect effects where loop 2 alters the conformation of binding residues. Two pieces of evidence suggest that the latter is the most likely. First, few amide resonances from this loop in Im9 are perturbed on binding the E9 DNase in 15N edited NMR experiments (20). Second, the solution structure of Im8 shows that the conformation of Tyr-55, a conserved residue in helix III, is markedly different from that of Im9, and the reason for this seems to be the nature of the adjoining residues in loop 2.2 The importance of this observation stems from the fact that recent mutagenesis data have shown this residue to be critical for DNase binding.3
In conclusion, the specificity of an immunity protein for an E colicin DNase domain is complex. Although dominated by residues from a single helix, interactions with other parts of the immunity protein scaffold also play an indirect role. Protein-protein interactions are characterized by relatively large surface areas of each binding partner (typically >600 Å2) becoming buried in the complex (27, 28), and given the very tight binding for the E9 DNase-Im9 complex this is likely to be true for a colicin complex. Since it is clear that other parts of Im9 are also involved in binding the DNase (20), it is interesting that specificity can be controlled almost exclusively by residues from a single helix, a situation reminiscent of some DNA-protein interactions (29).
We thank Ann Reilly and Christine Moore for expert technical assistance, and Andrew Leech, Russell Wallis, Kit-Yi Leung, Ansgar Pommer, Theonie Georgiou, and Catriona Giffard for help and advice during the course of this work. We also thank Ruth Boetzel for Fig. 1.