Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening

Article

Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening Highlights d

Facilitation of HCN2 channel opening by cAMP analogs was investigated

d

Some analogs bind but do not fully promote opening or C-terminal interactions

d

d

Authors Leo C.T. Ng, Igor Putrenko, Victoria Baronas, Filip Van Petegem, Eric A. Accili

Correspondence

Promotion of C-terminal interactions and facilitation of opening are correlated

[email protected]

Ligand bound structures identify regions important for facilitation and binding

Ng et al. show that direct binding of cyclic nucleotides to the HCN2 ion channel variably facilitates channel opening, and those nucleotides that are less effective openers are also unable to promote C-terminal interactions. Regions of the binding site that control facilitation and binding are identified.

In Brief

Accession Numbers 5KHG 5KHH 5KHI 5KHJ 5KHK

Ng et al., 2016, Structure 24, 1–14 October 4, 2016 ª 2016 Elsevier Ltd. http://dx.doi.org/10.1016/j.str.2016.06.024

Please cite this article in press as: Ng et al., Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening, Structure (2016), http://dx.doi.org/10.1016/j.str.2016.06.024

Structure

Article Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening Leo C.T. Ng,1 Igor Putrenko,1 Victoria Baronas,1 Filip Van Petegem,2 and Eric A. Accili1,3,* 1Department

of Cellular and Physiological Sciences of Biochemistry and Molecular Biology University of British Columbia, Vancouver, BC V6T 1Z3, Canada 3Lead Contact *Correspondence: [email protected] http://dx.doi.org/10.1016/j.str.2016.06.024 2Department

SUMMARY

Cyclic AMP is thought to facilitate the opening of the HCN2 channel by binding to a C-terminal domain and promoting or inhibiting interactions between subunits. Here, we correlated the ability of cyclic nucleotides to promote interactions of isolated HCN2 C-terminal domains in solution with their ability to facilitate channel opening. Cyclic IMP, a cyclic purine nucleotide, and cCMP, a cyclic pyrimidine nucleotide, bind to a C-terminal domain containing the cyclic nucleotide-binding domain but, in contrast to other cyclic nucleotides examined, fail to promote its oligomerization, and produce only modest facilitation of opening of the full-length channel. Comparisons between ligand bound structures identify a region between the sixth and seventh b strands and the distal C helix as important for facilitation and tight binding. We propose that promotion of interactions between the C-terminal domains by a given ligand contribute to its ability to facilitate opening of the full-length channel.

INTRODUCTION Hyperpolarization-activated cyclic nucleotide (HCN)-gated channels are activated by hyperpolarization of the membrane potential and some isoforms open more easily when cAMP binds to the cyclic nucleotide-binding domain (CNBD) of the intracellular C terminus (DiFrancesco and Tortora, 1991; Gauss et al., 1998; Ludwig et al., 1998; Santoro et al., 1998). HCN isoforms have significant sequence homology with channels in the voltage-gated potassium channel superfamily (Gauss et al., 1998; Ludwig et al., 1998; Santoro et al., 1998) for which several crystal structures have been solved (Doyle et al., 1998; Lee et al., 2005; Long et al., 2005). By extrapolation, HCN subunits are predicted to have six transmembrane helices (S1–S6) with a cytosolic N and C terminus, and to combine as tetramers to form the ion-conducting channel. A tetrameric structure predicts that each individual HCN channel has four cAMP binding sites.

A C-terminal region of HCN2 provides tonic inhibition on pore opening, which is relieved by binding of cAMP to the CNBD (Wainger et al., 2001). A study of a purified version of this HCN2 C-linker/CNBD using analytical ultracentrifugation shows that its tetramerization in solution is promoted by cAMP, which has led to the proposal that ligand binding promotes inter-subunit interactions and a gating ring in the full-length channel that relieves the inhibition and facilitates opening (Zagotta et al., 2003). The correspondence between facilitation of opening and the degree of in vitro tetramer formation is supported by a comparison of HCN1 and HCN2 channels; in the absence of cAMP, the C-linker/CNBD of HCN2 form oligomers less efficiently (Chow et al., 2012; Lolicato et al., 2011), and HCN2 opening is more inhibited than opening of HCN1 (Wainger et al., 2001). Correlation between facilitation of opening and degree of tetramerization is also supported by experiments showing that replacement of a tripeptide sequence in the C-linker domain of HCN2 with that of a cyclic-nucleotide-gated channel reverses the action of cAMP, causing it to inhibit both channel opening and tetramerization (Zhou et al., 2004). An alternative structural view of cAMP facilitation has also been proposed, which suggests that disruption of interaction between subunits, rather than promotion of interaction between them, facilitates the opening of the HCN2 channel (Craven et al., 2008; Craven and Zagotta, 2004). A salt-bridge triad was identified in the crystal structure of the HCN2 C-linker/CNBD piece in the C-linker region that connects the cAMP binding domain to the pore. One salt bridge is formed by residues within one subunit and another between residues in adjacent subunits. Mutations that break the salt bridges favor channel opening, which suggests that these salt bridges normally help to keep the channel closed. Together, the data suggest that the C-linker is in a bound but resting or closed conformation in the crystal structure (Zagotta et al., 2003) and that cAMP binding re-arranges the C-linker, which disrupts the salt bridges and facilitates opening of the HCN2 channel. Here, we provide evidence in support of the view that facilitation of the opening in the HCN2 channel by cAMP occurs by promoting inter-subunit interactions and formation of a gating ring. We examined a series of cAMP analogs and identified weak agonists of HCN2 based on their ability to promote the Structure 24, 1–14, October 4, 2016 ª 2016 Elsevier Ltd. 1

apparent molecular weight (kDa)

Please cite this article in press as: Ng et al., Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening, Structure (2016), http://dx.doi.org/10.1016/j.str.2016.06.024

100 80 60 40 no ligand cAMP cGMP

20 0 0

100 200 300 400 protein concentration (µM)

oligomerization of the isolated HCN2 C-linker/CNBD domains in solution. Most of the ligands tested, including both cyclic purine and pyrimidine nucleotides, bind to the CNBD, promote oligomerization of C-linker/CNBD domains in solution, and produce a depolarizing shift of the HCN2 activation curve, which is as large as that produced by cAMP. In contrast, cCMP and cIMP produce a smaller depolarizing shift of the HCN2 activation curve and bind but do not promote oligomerization of the C-linker/CNBD in solution. Our data support a model in which the degree of facilitation of opening in the full-length HCN2 channel by a given ligand depends upon its ability to promote oligomerization of the C-linker/CNBD in solution.

apparent molecular weight (kDa)

RESULTS

100 80 60 40 no ligand cCMP cUMP

20 0

apparent molecular weight (kDa)

0

100 200 300 400 protein concentration (µM)

80 60 40 no ligand cPuMP 2-NH2-cPuMP cIMP

20 0 0

50

100 150 200 protein concentration (µM)

Figure 1. Oligomerization of the HCN2 C-Linker/CNBD Is Promoted by cAMP, cGMP, and a Subset of Structurally Related Analogs Plots of estimated average molecular weight versus concentration of purified HCN2 C-linker/CNBD protein are shown. The estimated molecular weight was determined by dynamic light scattering and is proportional to the population of oligomers in solution. An increase in oligomerization occurs with higher protein concentrations and further increased in the presence of cAMP, cGMP (left), cUMP (right), cPuMP, and 2-NH2-cPuMP (middle). Cyclic CMP (right) and cIMP (middle) have no effect on the apparent molecular weight. Values represent means ± SEM, determined from two separate preparations of purified protein measured on different days.

2 Structure 24, 1–14, October 4, 2016

Cyclic CMP and cIMP Do Not Promote Self-Association of the HCN2 C-Linker/CNBD when Measured by Dynamic Light Scattering In solution, cAMP promotes oligomerization of isolated HCN2 C-linker/CNBD domains, suggesting that inter-subunit interactions are part of the alterations in structure induced by cAMP that lead to facilitation of opening in the full-length HCN2 channel (Chow et al., 2012; Zagotta et al., 2003). Because the propensity to the HCN2 C terminus to oligomerize in solution appears to correlate with facilitation of opening, we used an oligomerization assay to identify cAMP analogs that act as antagonists or weak agonists; we hypothesized that weak agonists or antagonists would bind to the HCN2 C-linker/CNBD but not promote oligomerization to the same extent as cAMP. The C-linker/CNBD of HCN2 that was used for this study is comprised of six a helices in the C-linker and the four a helices and eight b strands that make up the CNBD. We initially compared the extent that cCMP and cUMP, both cyclic pyrimidine nucleotides, promote self-association of the HCN2 C-linker/CNBD with that produced by cAMP, using dynamic light scattering (DLS) (Figure 1). Cyclic CMP is known to facilitate HCN2 channel opening (Zong et al., 2012) to a lesser extent than cAMP, and therefore, we expected that cCMP would also promote oligomerization of the HCN2 C-linker/ CNBD to a lesser extent than cAMP. As found previously, the apparent molecular weight of the HCN2 C terminus is increased with increasing amount of protein, even without the involvement of ligand, showing that tetramers are formed at high protein concentrations (Chow et al., 2012). By adding cAMP, the average weight is further increased at lower protein concentrations, corresponding to an increased proportion of larger oligomeric forms, which is consistent with the previous studies using DLS and ultracentrifugation (Chow et al., 2012; Zagotta et al., 2003; Zhou et al., 2004). Cyclic UMP produced an effect on molecular weight that was similar to that produced by cAMP, although larger amounts of the former were required (Figure 1). In contrast, cCMP had no effect on the apparent molecular weight of HCN2 C-linker/CNBD, even when present at very high ligand concentrations (Figure S1). The lack of effect of cCMP on oligomerization of the HCN2 C-linker/CNBD in solution suggests that its ability to promote a gating ring in the full-length channel is likewise limited,

Please cite this article in press as: Ng et al., Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening, Structure (2016), http://dx.doi.org/10.1016/j.str.2016.06.024

A

NH2

7N

5

N

8 O

9N

4

N

N

1 2

3

O

O HO

B

HO

2’

O

Time (min) 50

NH2

OH

Time (min) 50

100

0.0 μcal/s

-0.4 μcal/s

N

O

0

100

0.0

-0.8 -1.2 -1.6

cAMP

kcal/mol per injectant

kcal/mol per injectant

NH

P O

OH

0

N

O

P O

Figure 2. The Cyclic Purine Nucleotides cAMP and cGMP Bind to the HCN2 C-Linker/CNBD with Negative Cooperativity

O

6

0.0 -4.0 -8.0

-0.4 -0.8 -1.2 0.0

cGMP

(A) Chemical structure of the two ligands. The atoms in the purine ring are labeled for reference. (B) Plots of heat and binding isotherms produced by progressive injections of cAMP (left) and cGMP (right) to 200 mM HCN2 C-linker/CNBD as indicated, measured by ITC. (C) Bar graph showing the affinity and thermodynamics for the two binding events, a high-affinity and a low-affinity binding event, which were determined from the fit in (B). Values in the graph represent means ± SEM. Each mean was determined from independent ITC binding experiments as shown in Table 1; N in Table 1 refers to the number of independent experiments for each condition.

-2.0 -4.0 -6.0

Cyclic CMP and cIMP Bind to the HCN2 C-Linker/CNBD without -10.0 Negative Cooperativity 0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0 Molar Ratio Molar Ratio We next compared the binding of cyclic pyrimidine nucleotides and cyclic 10 purine nucleotides to the purified C HCN2 C-linker/CNBD using isothermal titration calorimetry (ITC). The protein 1 was tested at 200 mM, a concentration at which the C-linker/CNBD is predominantly in a tetrameric form in solution 0.1 as shown by DLS. We found that each cyclic nucleotide produced a clear and distinctive pattern of heat release (Fig0.01 ure 2B). Cyclic AMP, as we showed precGMP cAMP viously (Chow et al., 2012), produced two clear transitions, which were best low affinity low affinity fit with a two-independent-binding-site high affinity high affinity cAMP cGMP 10 10 model and yielded two Kd values of
0.13 mM and
1.8 mM (Figure 2B). 5 5 These two affinities have previously 0 0 been reported to indicate negative cooperativity (Chow et al., 2012) We found -5 -5 that cGMP binds to the HCN2 C-linker/ -10 -10 CNBD also with negative cooperativity, like cAMP, but with overall lower affinity -15 -15 (Kd values of
0.4 and
8.5 mM) (FigΔG ΔH -TΔS ΔG ΔH -TΔS -20 -20 ure 2B). The lower affinity of cGMP binding compared with cAMP binding consistent with its smaller effect on HCN2 channel opening is consistent with higher half maximal effective concentration (Zong et al., 2012). (EC50) values for cGMP measured by patch-clamp electroBy DLS, we also examined cGMP, which produces the same physiology in the full-length channel (Zagotta et al., 2003; maximum effect on the full-length HCN2 channel opening as Zhou and Siegelbaum, 2007). Cyclic UMP produced a pattern similar to cAMP, but, again, does cAMP but with lower potency (Zagotta et al., 2003; Zhou and Siegelbaum, 2007), and three cyclic nucleotide analogs that with lower values for affinity; high-affinity binding Kd of are intermediate in structure between cAMP and cGMP; cPuMP
1.3 mM and a low-affinity binding Kd of
23.7 mM (Figure 3B). (no exocyclic constituent), 2-NH2-cPuMP (2-NH2), and cIMP In contrast to cAMP and cUMP, the heat released by cCMP (6-keto). Of these three analogs, only cIMP was unable to promote could be fit best with a single-binding-site model, yielding an oligomerization, even when present in concentrations up to affinity of approximately 20 mM. We also found that cPuMP 10 mM, which is 200 times higher than the Kd (Figures 1 and S1). (
0.2 mM and
2.8 mM) and 2-NH2-cPuMP (
0.4 mM and -8.0

kcal/mol

kcal/mol

binding affinity (μM)

-12.0

Structure 24, 1–14, October 4, 2016 3

Please cite this article in press as: Ng et al., Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening, Structure (2016), http://dx.doi.org/10.1016/j.str.2016.06.024

A

NH2

4

5

N3

NH

2

6 O

N

1

O

O

O HO

O

Time (min) 50 100

0

μcal/s

μcal/s

0.0 -0.2 -0.4 F&03

-0.8 0.0

kcal/mol per injectant

kcal/mol per injectant

-0.6

-1.0 -2.0 -3.0 0.0

(A) Chemical structure of the two ligands. The atoms in the pyrimidine ring are labeled. (B) Plots of heat and binding isotherms produced upon progressive injections of 2 mM cCMP and cUMP to 200 mM HCN2 C-linker/CNBD measured by ITC. (C) Bar graph showing the affinity and thermodynamics of binding, which were determined from the fit in (B). Values in the graph represent means ± SEM. Each mean was determined from independent ITC binding experiments as shown in Table 1; N in Table 1 refers to the number of independent experiments for each condition.

O

0.0 -0.2 -0.4 -0.6 -0.8 -1.0

O OH

Time (min) 50 100

F803

0.0 -2.0 -4.0 -6.0 -8.0

0.5 1.0 1.5 2.0 Molar Ratio

C

P O

OH

0

B

N

O HO

2’

P O

Figure 3. The Cyclic Pyrimidine Nucleotide cUMP, but Not cCMP, Binds to the HCN2 C-Linker/CNBD with Negative Cooperativity

O

0.0

0.5 1.0 1.5 Molar Ratio

2.0

binding affinity (kd)

10

1

0.1

cUMP

cCMP

able entropic contribution for both cCMP and cIMP. Binding of the second cAMP or cUMP molecule occurs with lower affinity but more favorable enthalpy than binding of the first molecule, indicating that the entropic contribution is an important determinant of affinity and may underlie negative cooperativity. Together, the data suggest that cCMP and cIMP bind to the high-affinity binding site of the tetrameric C-linker/CNBD but that, upon binding, they are poor at inducing conformational changes and negative cooperativity. Interestingly, these are the same two molecules that failed to promote oligomerization (Figure 1), which suggests that negative cooperativity and ligand-induced tetramer formation are correlated.

Cyclic CMP and cIMP Are Less Effective Facilitators of HCN2 F&03 cUMP Opening than the Other Cyclic Nucleotides Examined 5 5 The results from the oligomerization and binding experiments suggest that cCMP 0 0 and cIMP have weaker effects than cAMP on the full-length HCN2 channel. -5 -5 A weaker facilitation of HCN2 opening has been noted previously for cCMP -10 -10 (Zong et al., 2012). We compared the effects of all analogs examined bioΔG -TΔS ΔG -TΔS -15 ΔH ΔH -15 chemically above on the HCN2 channel expressed in Chinese hamster ovary
3 mM) bound with negative cooperativity and in a similar range (CHO) cells using the whole-cell patch-clamp approach with to the binding of cAMP (Figure 4B). In contrast, the heat released 1 or 2 mM of cyclic nucleotides in the pipette solution or no by cIMP could be fit best with a single-binding-site model, cyclic nucleotide at all. The concentrations chosen are well yielding an affinity of
49 mM (Figure 4B). above the maximum concentration that is required to produce The binding of cCMP and cIMP was driven by both favorable a maximum effect on HCN2 channels expressed in HEK enthalpy and entropy, similar to the high-affinity binding event cells (Zong et al., 2012) or on sinoatrial HCN (DiFrancesco for cAMP and the other ligands (Figures 3C and 4C). However, and Tortora, 1991), and well above the binding affinities deterthe affinity is lower, which seems to be the result of a less favor- mined by ITC. 4 Structure 24, 1–14, October 4, 2016

kcal/mol

kcal/mol

low affinity high affinity

Please cite this article in press as: Ng et al., Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening, Structure (2016), http://dx.doi.org/10.1016/j.str.2016.06.024

O

A

7N

5

9N

4

6 N

8 O

N

N

3

O

P O

B

O

HO

2’

Time (min) 50 100

NH2

O HO

O

Time (min) 50 100 0.0 -0.2

-1.0 -1.5

-10.0 -15.0 0.0 0.5 1.0 1.5 2.0 Molar Ratio

2-NH2cPuMP

kcal/mol per injectant

kcal/mol per injectant

-5.0

-20.0

μcal/s

0.0 -0.5

0.0

N

O OH

Time (min) 50 100

0

0.0

cPuMP

NH

P O

OH

-0.5

-1.5

N

O

0

-1.0

kcal/mol per injectant

N

P O

OH

μcal/s

μcal/s

0

N

O

O HO

NH

1 2

N

0.0 -5.0 -10.0 -15.0 0.0 0.5 1.0 1.5 2.0 Molar Ratio

-0.4 -0.6

cIMP

0.0 -2.0 -4.0 0.0 0.5 1.0 1.5 2.0 Molar Ratio

binding affinity (μM)

C 10

1

0.1 2-NH2-cPuMP

cPuMP

cPuMP

15

cIMP 15

10

10

5

5

5

0 -5 -10

kcal/mol

10 kcal/mol

kcal/mol

low affinity high affinity

2-NH2cPuMP

low affinity high affinity

15

cIMP

0 -5

0 -5

-10

-10

-15

-15

-15

-20

-20

ΔH

-TΔS

ΔG

ΔH

-TΔS

ΔG

-20

ΔH

-TΔS

ΔG

Figure 4. The Cyclic Purine Nucleotides cPUMP and 2-NH2-cPUMP but Not cIMP Bind to the HCN2 C-Linker/CNBD with Negative Cooperativity (A) Chemical structure of the three ligands. The atoms in the purine ring are labeled as a reference. (B) Plots of heat and binding isotherms produced upon progressive injections of cPuMP, 2-NH2-cPuMP, and cIMP to 200 mM HCN2 C-linker/CNBD measured by ITC. (C) Bar graph showing the affinity and thermodynamics of binding, which were determined from the fit in (B). Values represent means ± SEM. Each mean was determined from independent ITC binding experiments as shown in Table 1; N in Table 1 refers to the number of independent experiments for each condition.

The Ih activation curves were determined by plotting normalized tail-current amplitudes versus test voltage (Figure 5), which were then fit with Equation 1 to obtain values for half-activation voltage (V1/2) and slope factor (k) (Table 1). Compared with the control cells, those exposed to saturating concentrations of cCMP and cIMP produced the smallest shifts in V1/2 of activation and the smallest increase in conductance at three different volt-

ages (Table 2), i.e. they are weaker facilitators of opening than the other cyclic nucleotides examined. These data support a direct relationship between agonist-induced structural changes and oligomerization of the HCN2 C-linker/CNBD domains and facilitation of opening in the full-length HCN2 channel. Notably, cCMP produced a small significant shift in the activation curve at 90 mV but no shift at more negative voltages (see Table 2). Structure 24, 1–14, October 4, 2016 5

Please cite this article in press as: Ng et al., Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening, Structure (2016), http://dx.doi.org/10.1016/j.str.2016.06.024

Figure 5. Cyclic AMP and Analogs Variably Shift the HCN2 Ih Activation Curve to Less Negative Potentials Plots of degree of current activation, determined by normalizing tail-current amplitudes to their maximum value (I/Imax), versus test voltages. Values represent means ± SEM and the number of cells for each condition (n values) are shown in Tables 1 and 2. The solid lines through the values represent fitting to a Boltzmann curve by Equation 1 (methods), which yielded values for V1/2 and k (Tables 1 and 2). The shift in V1/2 is smaller for cCMP and cIMP.

6 Structure 24, 1–14, October 4, 2016

Please cite this article in press as: Ng et al., Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening, Structure (2016), http://dx.doi.org/10.1016/j.str.2016.06.024

Table 1. Values for Binding Affinity and Biophysical Parameters Describing Voltage-Dependent Activation with or without Ligand Binding Affinity Control (no ligand)

Electrophysiology

N

High Kd (mM)

Low Kd (mM)

n

–

–

–

7

V1/2 (mV) 97.15 ± 3.55

Slope (z)

DLS

11.11 ± 0.37

–

+cAMP

4

0.13 ± 0.03

1.83 ± 0.21

7

87.14 ± 3.77 (0.039)*

11.46 ± 0.89 (0.717)

U

+cGMP

4

0.43 ± 0.07

8.53 ± 0.70

6

87.45 ± 2.73 (0.029)*

12.61 ± 1.00 (0.162)

U

cAMP-cGMP intermediates +cPuMP

3

0.17 ± 0.03

2.76 ± 0.19

7

88.57 ± 4.55 (0.081)

12.61 ± 1.01 (0.187)

U

+2NH2-cPuMP

3

0.41 ± 0.04

3.08 ± 0.29

6

87.96 ± 2.81 (0.037)*

14.09 ± 1.05 (0.016)*

U

+cIMP

4

49.32 ± 6.45

7

92.73 ± 2.69 (0.170)

10.72 ± 0.69 (0.639)

7

Cyclic pyrimidine nucleotides Control (no ligand)

–

–

+cCMP

3

20.23 ± 0.66

+cUMP

3

1.33 ± 0.17

6

114.89 ± 3.76

13.52 ± 2.54

–

6

111.96 ± 4.77 (0.320)

20.06 ± 3.08 (0.133)

7

23.70 ± 1.82

6

104.72 ± 2.30 (0.022)*

14.78 ± 1.26 (0.769)

U

–

6

101.21 ± 2.59

10.88 ± 0.34

–

–

C-helix mutation R635A Control R635A (no ligand)

–

–

+cAMP

3

0.19 ± 0.01

+cUMP

3

183.8 ± 85.7

1.89 ± 0.05

6

87.75 ± 1.97 (0.001)*

11.46 ± 0.49 (0.352)

U

6

88.82 ± 2.24 (0.002)*

11.44 ± 1.17 (0.652)

U

The table summarizes values for binding affinity, half-activation voltage (V1/2) and slope factor (k) determined from fits of the data obtained from individual cells by Equation 1. For binding affinity, values for mean ± SEM are shown as determined from separate experiments (N = number of separate experiments). Kd values determined for cCMP and cIMP are italicized because only one binding site was observed. For the biophysical parameters, the values are means determined from individual cells (N = number of cells) ± SEM. The p values of V1/2, shown in parentheses, were calculated using onetailed unpaired Student’s t test comparing data obtained from cells examined using cyclic nucleotide in the pipette solution with respective control data obtained from cells examined without cyclic nucleotide in the pipette solution, which were part of the same HCN2 channel transfection. A one-tailed test was used for V1/2 values because all the treatments were expected to produce a depolarizing shift in the activation curve. Similarly, the p values of the slope factor were calculated with an unpaired t test. p Values <0.05 are highlighted with an asterisk. The DLS column shows whether the ligand is able to promote oligomerization, as an increase in apparent molecular weight of HCN2 C-linker/CNBD, upon adding ligand. The data show a correlation where a ligand that does not promote oligomerization in the soluble C-linker/CNBD also causes a smaller shift in the activation curve in the HCN2 full-length channel.

Thus, the effect of cCMP on opening appears to depend upon either voltage or on the proportion of channels open. The analog 2-NH2-cPuMP produced a significant effect on the slope of the activation curve that suggested a larger shift at less negative voltages (see Table 1). Cyclic UMP Makes Contacts with the HCN2 C Helix that Differ from Those Made by cCMP To directly determine how cCMP and cUMP bind to the C-linker/ CNBD of the HCN2 channel, we used X-ray crystallography and compared the solved structures (Table 3). The refined structure of HCN2 C terminus with cCMP and cUMP show that each forms a fully liganded tetramer with the cyclic pyrimidine nucleotide bound in the anti configuration (Figure 6 and S2) like cAMP and unlike cGMP, which were shown previously to bind to this fragment of HCN2 in the anti and syn configuration, respectively (Zagotta et al., 2003). The different complexes have near-identical folds, but the specific interactions in the nucleotide-binding pocket differ substantially and readily explain the ability of cUMP to bind with higher affinity than cCMP. This is particularly the case for residues E582, C584, and A593, found in a region containing the P helix and between b strands 6 and 7, as well as at residues R632, R635, and I636 found in the C helix near the end of the CNBD (Figure 6; Table S1). In both cCMP and cUMP, the side chain of residue E582 forms two hydrogen bonds with the OH group of the ribose and forms a salt bridge with the side chain of R632. A non-polar inter-

action of R632 with the pyrimidine ring and a hydrogen bond between the side chain and 2-keto group of the pyrimidine ring are also found in both cCMP and cUMP structures. However, the nitrogen of the main chain of E582 interacts more closely (within 4 A˚) with the 20 -OH of the ribose of cCMP. The nitrogen of the main chain of C584 also forms a hydrogen bond with the axial oxygen of the cyclic phosphate group of cCMP but not cUMP. The carbonyl group of the R632 main chain forms a hydrogen bond with the 4-NH2 group of cCMP, whereas this group is also close to the 4-keto group of cUMP and may experience some repulsive electrostatic interaction between the negative partial charges. The 4-keto group of cUMP is also close to the side chain of R635, which is positioned near enough to form a hydrogen bond in the cUMP bound structure but not in the cCMP-bound structure. The pyrimidine ring of cUMP, but not cCMP, makes a non-polar interaction with I636, found adjacent to R635 in the C helix and A593, located just before the seventh beta sheet. To assess the contribution of R635 to the binding affinity and maximum effect of cUMP, we mutated this residue to an alanine (R635A) in the C-linker/CNBD as well as the full channel to abolish the newly observed bond. The alanine substitution had little effect on the binding affinity of cAMP but produced a more drastic effect on the binding of cUMP, eliminating negative cooperativity and reducing binding affinity. Nevertheless, alanine substitution of arginine 635 did not alter the ability of cUMP, as well as cAMP, to promote oligomerization of the HCN2 C-linker/CNBD or their maximum effect on opening (Figure 7). Structure 24, 1–14, October 4, 2016 7

Please cite this article in press as: Ng et al., Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening, Structure (2016), http://dx.doi.org/10.1016/j.str.2016.06.024

Table 2. Values for Half-Activation Voltage and Fraction of Channel Activation at Three Voltages n

V1/2

Conductance at 75 mV

Conductance at 90 mV

Conductance at 105 mV

0.0752 ± 0.0229

0.3513 ± 0.0660

0.6424 ± 0.0653

Wild-Type (HCN2) with Lower KCl Extracellular Solution Control

7

97.15 ± 3.55

+2 mM cAMP

7

87.14 ± 3.77 (0.0386)*

0.2251 ± 0.0730 (0.0368)*

0.5371 ± 0.0574 (0.0275)*

0.7902 ± 0.287 (0.0302)*

+2 mM cGMP

6

87.45 ± 2.73 (0.0294)*

0.2230 ± 0.0478 (0.0069)*

0.5289 ± 0.0416 (0.0257)*

0.7769 ± 0.0188 (0.0462)*

+2 mM 2-NH2-cPuMP

6

87.96 ± 2.82 (0.0368)*

0.2389 ± 0.0464 (0.0035)*

0.5295 ± 0.0468 (0.0282)*

0.7550 ± 0.0757 (0.0815)

+2 mM cIMP

7

92.73 ± 2.69 (0.1701)

0.1463 ± 0.0350 (0.0575)

0.4338 ± 0.0582 (0.1835)

0.7451 ± 0.0426 (0.1061)

+2 mM cPuMP

7

88.57 ± 4.55 (0.0814)

0.2574 ± 0.0582 (0.0065)*

0.5389 ± 0.0651 (0.0329)*

0.7633 ± 0.05118 (0.0852)

Mutant R635A-HCN2 with Lower KCl Extracellular Solution 101.21 ± 2.59

Control

6

0.0615 ± 0.0227

0.2703 ± 0.0543

0.5964 ± 0.0419

+2 mM cAMP

6

87.75 ± 1.97 (0.0010)*

0.2233 ± 0.0331 (0.0012)*

0.5445 ± 0.0379 (0.0010)*

0.8050 ± 0.0277 (0.0010)*

+2 mM cUMP

6

88.82 ± 2.244 (0.0024)*

0.1954 ± 0.0413 (0.0087)*

0.5145 ± 0.0484 (0.0036)*

0.7844 ± 0.0228 (0.0014)*

Conductance at 90 mV

Conductance at 105 mV

Conductance at 120 mV

Wild-Type (HCN2) with Higher KCl Extracellular Solution Control

6

114.89 ± 3.76

0.1084 ± 0.0224

0.3559 ± 0.0473

0.6528 ± 0.0365

+1 mM cCMP

6

111.96 ± 4.77 (0.3199)

0.2355 ± 0.0478 (0.0184)*

0.4429 ± 0.0437 (0.1033)

0.6556 ± 0.0339 (0.4780)

+1 mM cUMP

6

104.72 ± 2.30 (0.0219)*

0.2611 ± 0.398 (0.0037)*

0.5068 ± 0.0459 (0.0226)*

0.7579 ± 0.0272 (0.0218)*

The table summarizes the values for half-activation voltages and compares the fraction of activation at three voltages for each condition. Values are means ± SEM and the number of cells for each condition (n values) are shown. The p values of V1/2, shown in parentheses, were calculated using onetailed unpaired Student’s t test comparing data obtained from cells examined using cyclic nucleotide in the pipette solution with respective control data obtained from cells examined without cyclic nucleotide in the pipette solution, which were part of the same HCN2 channel transfection. A one-tailed test was used for V1/2 values because all the treatments were expected to produce a depolarizing shift in the activation curve. p Values <0.05 are highlighted with an asterisk.

Because cUMP retains its full ability to promote oligomerization of the R635A HCN2 C-linker/CNBD and to facilitate opening of the full-length R635A channel, it is possible that negative cooperativity is still present, but simply unresolved in the ITC because of insufficient differences in affinity. Crystal Structures Show that cAMP-cGMP Intermediates Bind in the Anti Configuration and Identify Unique Contacts Made by cIMP with the CNBD We also examined the structure of the HCN2 C-linker/CNBD liganded to the cAMP-cGMP intermediates, knowing that cAMP binds in the anti configuration while cGMP binds in the syn configuration (Zagotta et al., 2003). Crystals of HCN2 saturated with cPuMP or 2-NH2-cPuMP diffracted at a resolution of
2 A˚ and both ligands bind in the anti configuration as for cAMP (Figures 6 and S2; Table 3). Cyclic IMP also bound in the anti configuration, which was somewhat surprising because none of the cAMP-cGMP intermediates bound in the syn configuration. Together, these data suggest that both the NH2 group at C2 and the keto-oxygen at N6 are required for cGMP binding in the syn configuration. Because the analogs bind in the anti configuration, the structure of the C-linker/CNBD bound with cIMP, cPuMP, and 2-NH2-cPuMP can be compared directly with the cAMP-bound structure. Knowing the binding configurations, the binding pocket of cIMP can be compared with strong agonists, cAMP, cPuMP, or 2-NH2-cPuMP, to understand its low binding affinity and reduced ability to promote tetramerization and facilitate opening. Although many of the specific interactions are similar between cIMP and the other cyclic nucleotides, major differences in interactions occur, again, at residues E582 and C584 8 Structure 24, 1–14, October 4, 2016

found in the P helix between b strands 6 and 7, and at residues R632 and R635 found in the C helix (Figure 6; Table S1). The nitrogen of the backbone of C584 is farther from the axial oxygen of the cyclic phosphate group of cIMP, 2NH2-cPuMP, and cPuMP than that of cAMP and will not form a hydrogen bond. The nitrogen of the backbone of E582 is farther from the equatorial oxygen of the cyclic phosphate group of cIMP and 2-NH2-cPuMP than that of cAMP and cPuMP, again reducing the probability of forming a hydrogen bond. The E582 backbone nitrogen also is farther from the 20 -OH group of the cIMP ribose and, unlike this interaction for the other three cyclic nucleotides, will also not form a hydrogen bond. The main chain carbonyl of R632 is close and forms a hydrogen bond with the 6-NH2 group of cAMP, but it creates an electrostatic repulsion with the 6-keto group of cIMP instead due to the partial negative charges; this main chain carbonyl does not interact with the other two ligands, which lack an exocylic group at position 6 of the purine ring. The side chain of R635 forms a close interaction with the 6-keto group of cIMP, likely forming a hydrogen bond not found in the other cyclic nucleotides. In summary, the keto-carbonyl repulsion causes a greater distance between cIMP and the P helix, while the distal C helix, and especially the R635 side chain, is more closely associated with cIMP than with the other three cyclic nucleotides. In cases for both weak agonists, there are differences in the binding interactions between the ligands and the binding pocket, especially with the P helix and the distal C helix. DISCUSSION To better understand the nature of cyclic nucleotide agonism in the HCN2 channel and to identify weak agonists, we examined

Please cite this article in press as: Ng et al., Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening, Structure (2016), http://dx.doi.org/10.1016/j.str.2016.06.024

Table 3. Data Collection and Refinement Statistics for mHCN2 C-Linker/CNBD Co-crystallized with Five Different Ligands HCN C-Linker/CNBD Co-Crystallized with cCMP (Ligand Code CC7)

cUMP (Ligand Code 6SY)

cPuMP (Ligand Code 6SX)

2-NH2-cPuMP (Ligand Code 6SZ)

cIMP (Ligand Code 6SW)

P 4 21 2

P 2 21 21

P 4 21 2

P 4 21 2

I422

Data Collection Space group Cell dimensions a, b, c (A˚)

96.204, 96.204, 46.68

46.486, 89.925, 98.421

96.278, 96.278, 46.763

96.429, 96.429, 46.423

96.037, 96.037, 115.913

a, b, g ( )

90, 90, 90

90, 90, 90

90, 90, 90

90, 90, 90

90, 90, 90

Resolution (A˚)

30.42–2.24 (2.32–2.24)

28.68–2.01 (2.08–2.01)

31.79–2.10 (2.18–2.10)

38.40–2.07 (2.14–2.07)

44.12–1.77 (1.88–1.77)

Rsym or Rmerge

7.1 (47.6)

8.8 (49.7)

6.3 (89.3)

7.0 (79.7)

7.4 (120.9)

I/sI

12.74 (2.54)

7.35 (2.00)

18.60 (2.68)

26.99 (4.08)

19.06 (2.10)

Completeness (%)

96.19 (87.56)

95.1 (94.3)

99.5 (99.0)

94.4 (99.1)

97.5 (95.2)

Redundancy

5.34 (5.24)

1.54 (1.52)

7.04 (7.09)

13.86 (13.18)

9.83 (9.92)

Resolution (A˚)

30.42–2.24

28.68–2.01

31.79–2.10

38.40–2.07

44.12–1.77

No. reflections

10,551 (1,657)

50,521 (8,055)

13,266 (2,090)

13,521 (2,028)

26,044 (4,035)

Rwork/Rfree

0.2231/0.2643

0.1978/0.2440

0.2006/0.2489

0.2161/0.2614

0.2144/0.2628

No. atoms

1,645

3,517

1,683

1,713

1,803

Protein

1,599

3,163

1,623

1,639

1,616

Ligand

20

40

21

22

21

Water

26

314

39

52

166

B factors

66.9

26.10

59.58

49.40

39.37

Protein

66.95

25.50

59.78

49.40

37.21

Ligand

68.88

20.30

54.83

50.50

33.65

Water

59.67

33.60

53.68

47.00

44.84

Refinement

Root-mean-square deviations Bond lengths (A˚) 0.008

0.011

0.0113

0.016

0.011

Bond angles ( )

1.14

1.437

1.77

1.50

1.01

Data, each set from a single crystal of HCN2 C-linker/CNBD liganded with respective cyclic nucleotides, was integrated with HKL2000 and XDS, and refinement was completed with Refmac5 of CCP4 suite and phenix.refine of PHENIX. Values in parentheses are for the highest-resolution shell.

the ability of cAMP analogs to induce changes in the structure of the C-linker/CNBD and correlated these changes with the degree of facilitation of opening. We find that the degree of oligomerization correlates with degree of facilitation of opening. Unlike the other analogs, cCMP and cIMP were unable to promote oligomerization of the HCN2 C-linker/CNBD in vitro. Because oligomerization in solution is thought to represent the formation of a gating ring in the full-length channel to facilitate opening (Chow et al., 2012; Lolicato et al., 2011; Zagotta et al., 2003), the inability of cCMP and cIMP to do so suggests that a gating ring is inefficiently promoted in the full-length HCN2 channel by these molecules. Both cCMP and cIMP are less effective facilitators of HCN2 opening compared with the other ligands. Cyclic CMP is a cyclic pyrimidine nucleotide that has been shown previously to act as a weak agonist of the HCN2 and HCN4 channels and of sinoatrial HCN channels (DiFrancesco and Tortora, 1991; Zong et al., 2012). We found that the effect of cCMP on the HCN2 channel was indeed limited compared with most of the other agonists. In contrast, cUMP, another cyclic pyrimidine nucleotide, was able to promote oligomerization of the HCN2 C-linker/CNBD and bind with negative cooperativity, and it produced stronger facilitation of opening. Unlike cIMP, cPuMP and 2-NH2-cPuMP

were able to promote oligomerization of the HCN2 C-linker/ CNBD and bind with negative cooperativity, and they also produced greater facilitation of HCN2 opening than cIMP. Together with the actions of cAMP and cGMP, these data suggest that the degree of facilitation for a given ligand depends, at least in part, on its ability to promote a gating ring in the full-length channel. The three cAMP-cGMP intermediates bound to the HCN2 C-linker/CNBD in the anti configuration like cAMP and not like cGMP, which is known to bind in the syn configuration (Zagotta et al., 2003). Because the three cAMP-cGMP intermediates do not bind to the HCN2 C-linker/CNBD in the syn configuration, less information can be gleaned from the binding data of these analogs about the atoms important for controlling cGMP binding affinity and effect. An important finding that does arise from these data is that both exocyclic groups, the 2-keto and 6-NH2 groups, are required in order for cGMP to bind in the syn conformation. These data appear to follow the syn/anti configurations of the cyclic nucleotides in water, with cGMP existing in syn configuration to a greater extent than cAMP (Zhou and Siegelbaum, 2008). The resolved crystal structures of the HCN2 C-linker/CNBD in complex with the ligands suggest that interactions in regions within the sixth and seventh b strands and in the distal C helix Structure 24, 1–14, October 4, 2016 9

Please cite this article in press as: Ng et al., Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening, Structure (2016), http://dx.doi.org/10.1016/j.str.2016.06.024

Figure 6. Crystal Structures of Five Analogs Bound to the HCN2 C-Linker/CNBD Reveal their Ligand-Binding Configuration and Important Interactions The binding behavior of the five analogs to HCN2 C-linker/CNBD is visualized using X-ray crystallography with PyMOL. The PyMOL structures consist of the protein represented with a ribbon backbone and the ligand represented as sticks; residues that interact with the ligand are also represented as sticks. The blue electron density is generated by the 2mFo  DFc map of the structure, sigma cutoff at 1.0, showing how each ligand fits. Refer to Figure S2 for the 2mFo  DFc omit map. The interactions inside the five binding pockets are also presented in a simplified 2D schematics, and the pockets of cAMP (PDB: 1Q5O) and cGMP (PDB: 1Q3E) are added as a reference. Dashed lines are hydrogen bonds determined by PISA; double-headed arrows are salt bridges, and curved lines are hydrophobic interactions. The numbers categorize where the bonds are found and correspond to Table S1, which contains intermolecular distances.

of the CNBD contribute to the allosteric changes that lead to facilitation of opening. These interactions are notable at or near residue R632 in the distal C helix, which has been proposed as a residue critical for allosteric signaling between ligand binding and channel opening (Zhou and Siegelbaum, 2007); differences were found near this residue when the cCMP-bound and cUMP structures were compared as well as when the cIMP bound structure was compared with the cAMP, cPuMP, and 2-NH2cPuMP bound structures. Our data fit with previous data in HCN2 and HCN4 using nuclear magnetic resonance, fluorescence resonance energy transfer, and double electron-electron resonance suggesting that cAMP binding shifts the B and C helices toward the b roll (Akimoto et al., 2014; Puljung et al., 2014; Saponaro et al., 2014; Taraska et al., 2009; VanSchouwen et al., 2015), and orders the P helix and the distal C helix, and that the shifts induced by cCMP are less stable than those produced by cAMP (Akimoto et al., 2014). Direct elucidation of the structural changes in the full-length channel induced by strong agonists such as cAMP and weak agonists such as cCMP will require other approaches, e.g., X-ray crystallography or cryo-electron microscopy at high resolution, which may capture the channel in cyclic-nucleotide-bound and activated conformations. Finally, 10 Structure 24, 1–14, October 4, 2016

as demonstrated by the binding and action of cPuMP, the interactions made by exocyclic groups are not required for cyclic purine nucleotide promotion of tetramerization of the C-linker/ CNBD and strong facilitation, and make only minor contributions to binding affinity. However, as shown by cIMP, the addition of a keto group at position 6 is sufficient to weaken binding, limit tetramerization of the C-linker/CNBD, and render facilitation of channel opening less effective. An understanding of how cAMP and its analogs act on HCN channels is important for several reasons. First, cCMP, cIMP, and cUMP are experiencing a re-birth as potential cellular second messengers (Bahre et al., 2015; Schlossmann and Wolfertstetter, 2016; Seifert, 2015), and knowing their mechanism of action is a necessary prerequisite for understanding their role in vivo. Second, understanding how these molecules act will help to understand cyclic nucleotide agonism in general in HCN channels, which may be useful for the development of therapeutic molecules that control their activity. Currently, the drug ivabradine, an HCN inhibitor, is available in the clinic for the treatment of angina and heart failure due to its ability to limit the increases in heart rate that exacerbate those conditions (DiFrancesco and Borer, 2007; Roubille and Tardif,

Please cite this article in press as: Ng et al., Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening, Structure (2016), http://dx.doi.org/10.1016/j.str.2016.06.024

A

Time (min) 50

0

Time (min) 50

0

100

0.00

Figure 7. Effect of cUMP on the Binding and Oligomerization R635A Mutation in HCN2 C-linker/CNBD

100

0.00

-0.50 -0.10

-1.50 R635A cAMP

-2.00

kcal/mol per injectant

kcal/mol per injectant

0.00 -4.00 -8.00 -12.00 0.0

B

0.5

1.0 1.5 Molar Ratio

2.0

-0.20

R635A cUMP

-0.40

-0.80

-1.20 0.0

2.5

cAMP/ R635A

10

0.5

1.0 1.5 Molar Ratio

low affinity high affinity

10

5

5

0

0

kcal/mol

kcal/mol

μcal/s

μcal/s

-1.00

-5

2.5

cUMP/ R635A

-5

-10

-10

-15

-15

C

2.0

apparent molecular weight (kDa)

120 100 80 60 40 no ligand with cAMP with cUMP

20 0 0

100

200

300

400

protein concentration

D 1.0

no ligand + 2mM cAMP

1.0

0.4

0.6 0.4

h G

G

0.8 (normalized)

0.6

h

(normalized)

0.8

no ligand +2 mM cUMP

(A) Plots of heat produced upon progressive injections of cAMP and cUMP to 200 mM of R635A HCN2 C-linker/CNBD measured by ITC. The inflections in the top plot arise from injections of cAMP where each inverted peak shows the heat difference between the sample and the reference compartment. The peaks decrease in magnitude as binding sites become saturated. The lower plot shows values determined by integration of the area under the peaks from the upper plot versus the ratio of injected ligand to protein. The solid line through the values represents either a singlebinding-site model (cUMP) or a two-independentbinding-site model (cAMP), which yielded values for affinity and energetics (DG, DH, and DS). (B) Bar graph showing the thermodynamics of binding, which were determined from the fit in (A). Values in the graph represent means ± SEM. Each mean was determined from independent ITC binding experiments as shown in Table 1; N in Table 1 refers to the number of independent experiments for each condition. (C) Shown are plots of estimated average molecular weight versus concentration of purified HCN2 C-linker/CNBD protein. The estimated molecular weight was determined by DLS and is proportional to the population of oligomers in solution. An increase in oligomerization occurs with increasing protein concentration and further increase in the presence of cAMP and cUMP. Values represent means ± SEM, determined from three separate preparations of purified protein measured on different days. (D) Activation curves determined by plotting tailcurrent amplitudes, which were normalized to their maximum value (I/Imax), versus test voltage. The solid lines through the values represent fitting of a Boltzmann curve by Equation 1 (methods). Fits yielded values for V1/2 and k, which are found in Table 1 along with other parameters. Both cAMP and cUMP produce a shift in the activation curve that is comparable with the shift they produce in the wild-type channel. Values represent means ± SEM and the number of cells for each condition (n values) are shown in Tables 1 and 2.

0.2

examples, keys for the development of HCN modulators may well be isoform specificity (Emery et al., 2012) and -160 -140 -120 -100 -80 -60 -40 -160 -140 -120 -100 -80 -60 -40 more subtle effects that could potentially E (mV) E (mV) be attained by targeting the CNBD, which modulates ion flow, rather than 2013). However, ivabradine, a pore blocker (Bucchi et al., 2002, by targeting the pore, which is required for ion flow. The 2006, 2013), is associated with side effects such as compro- approach outlined here — identify analogs that do not promote mised vision and profound bradycardia (Fox et al., 2013), oligomerization by DLS, check whether they bind to the C terrendering it unavailable for a subset of patients. Another poten- minus by ITC, identify key interactions with the binding site by tial application for HCN inhibition is in the control of inflamma- X-ray crystallography, and determine if they are able to protory and neuropathic pain, which is reduced or eliminated in mote opening by patch-clamp electrophysiology — could be mice that have had the HCN2 gene deleted from neurons of an efficient method to identify promising cyclic nucleotide anathe dorsal root ganglia (Emery et al., 2011). In these and other logs, which could be then optimized (Schwede et al., 2015) to 0.2 0.0

0.0

Structure 24, 1–14, October 4, 2016 11

Please cite this article in press as: Ng et al., Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening, Structure (2016), http://dx.doi.org/10.1016/j.str.2016.06.024

bind more tightly than cAMP and effectively inhibit channel activity. EXPERIMENTAL PROCEDURES HCN Protein Purification and Mutagenesis The HCN2 C-linker/CNBD from mouse (residues 443–645) was subcloned in a modified pET28 vector (Novagen), containing a His6-tag, maltose-binding protein, and a cleavage site for the tobacco etch virus (TEV) protease preceding the construct of interest. The mutation R635A was made by standard Quikchange protocol (Stratagene). Primers used were 50 -ATT GAC CGG CTA GAT GCC ATA GGC AAG AAG AAC-30 and 50 -GTT CTT CTT GCC TAT GGC ATC TAG CCG GTC AAT-30 . The protein was expressed in Escherichia coli Rosetta (DE3) pLacI cells grown in 2xYT medium at 37 C for 4 hr, after induction with 0.4 M isopropyl b-D-1-thiogalactopyranoside (IPTG) at an optical density of 0.6. The cells were harvested by centrifugation and resolubilized in lysis buffer. Cell debris was removed by centrifugation for 30 min at 35,000 3 g. The tagged proteins were separated by cobalt affinity columns (Talon, Clontech) and eluted with 500 mM imidazole, similar to the purification protocol previously employed (Chow et al., 2012). The cleavage of the HMT tag was performed by incubating the purified protein with TEV protease overnight at 4 C. The tag was removed by another round of Talon resin, and further purified on a ResourceS cation exchanger (GE Healthcare). Purified HCN2 C-linker/CNBD was dialyzed in ITC buffer (150 mM KCl, 20 mM HEPES [pH 7.4], 10 mM 2-mercaptoethanol). Concentration was determined by spectrophotometry, using the Edelhoch method. Protein purity was confirmed by SDS-PAGE. Ligand Preparation Cyclic AMP and cGMP were purchased from Sigma. The other analogs were from Biolog Life Science Institute. Ligands were dissolved in water to make a stock solution of 10 mM. They were further diluted to 2 mM with ITC buffer. The exact concentrations of these analogs were determined by spectrophotometry with their respective extinction coefficients. A negative background test was done on the ITC for each ligand to make sure heat was not generated by water-buffer dilution or by ligand-solvent interaction. The background produced minimal noise for each of the ligands (data not shown). Ligand-Induced Tetramerization by Dynamic Light Scattering The proteins and ligands were mixed and diluted to the correct ratio. Samples were centrifuged for 1 min to remove debris and 50 mL was loaded into each well in the 384-well microtiter plate. DLS experiments were performed with a DynaPro plate reader (Wyatt Technology) based on detecting the scattered waves due to a laser shining onto moving molecules. Ten acquisitions (5 s per scan) were performed at 22 C, and the radii were averaged and regularized, with Dynamics 7.0, to produce the apparent molecular weights. Only readings with polydispersity values lower than 30% were used.

Whole-Cell Patch-Clamp Electrophysiology Cells expressing GFP were chosen for whole-cell patch-clamp recordings 24–48 hr post transfection. The pipette solution contained 130 mM K-Asp, 10 mM NaCl, 0.5 mM MgCl2, 1 mM EGTA, and 5 mM HEPES with pH adjusted to 7.4 using KOH. For experiments at saturating levels of cyclic nucleotides analogs, each was added to the pipette solution as a sodium salt at concentrations of R1 mM, as indicated. Control experiments were carried out on cells in the same set of transfections as those in which a given ligand was tested. Two different potassium extracellular recording solutions were used as indicated in the Results. The solution with lower extracellular potassium contained 60 mM KCl, 80 mM NaCl, 0.5 mM MgCl2, 1.8 mM CaCl2, 5 mM HEPES (pH 7.4). The high extracellular potassium solution contained 135 mM KCl, 5 mM NaCl, 0.5 mM MgCl2, 1.8 mM CaCl2, 5 mM HEPES (pH 7.4). Whole-cell currents were recorded using an Axopatch 200B amplifier and Clampex software (Axon Instruments) at room temperature. Patchclamp pipettes were pulled from borosilicate glass and fire polished before use (pipette R = 2.5–4.5 MU). Data were filtered at 2 kHz and analyzed using Clampfit (Axon Instruments) and Origin 8.0 (Microcal) software. The Ih activation curves were determined from time-dependent tail currents at a 2 s pulse to 35 mV following 3 s test pulses ranging from 150 mV to 10 mV, in 20 mV steps. Single tail-current test pulses were followed by a 500 ms pulse to +5 mV to ensure complete channel deactivation. The resting current was allowed to return to its baseline value before subsequent voltage pulses. Normalized tail-current amplitudes were plotted as a function of test voltage, and values from each cell were fit with a single-order Boltzmann function (Equation 1), fðVÞ = 

Cell Culture and Transfection CHO cells were maintained in Ham’s F-12 medium (Life Technologies) supplemented with 50 mg/mL penicillin/streptomycin (Life Technologies) and 10% fetal bovine serum (Life Technologies). Cells were grown on glass coverslips (22 3 22 mm) in 35 mm dishes at 37 C with 5% CO2. Transfection of a mammalian expression vector encoding mouse HCN2 (2 mg per dish) was performed using the FuGene6 transfection reagent (Roche Diagnostics) according to the manufacturer’s protocol. The DNA plasmid (1 mg per dish) encoding GFP was co-transfected for visualization of transfected cells.

12 Structure 24, 1–14, October 4, 2016

(Equation 1)

to determine the midpoint of activation (V1/2) and slope factor (k). The effective charge (Z) was calculated using the equation Z = RT/kF, where T = 295 K, R is the gas constant, and F is Faraday’s constant. Crystallization of HCN2-Ligand Complexes Protein complexes, 200 mM (
10 mg/mL) of HCN2 C-linker/CNBD with 5 mM of respective analog, were diluted with ITC buffer. Crystals were obtained using the hanging-drop vapor-diffusion method, after 1–2 weeks at 4 C. The well solutions contained 10%–26% PEG400, 0.1 M sodium citrate (pH 4.6–5.5), 0.2 M NaCl, with reagents obtained from FLUKA and Sigma. The crystals selected were from these well conditions: d

d

d

d

Direct Measurements of Cyclic Nucleotide Binding by Isothermal Titration Calorimetry Titrations were performed in ITC200 (MicroCal, Malvern) at 25 C in an adiabatic environment. Then 200 mM (diluted in ITC buffer) HCN2 C-linker/CNBD was added to the sample cell; 2 mM ligand was loaded on the syringe and increments of 1 mL were injected into the sample cell. Heat difference between the reference cell with ddH2O and the sample cell with ligand-protein mixture was recorded. Calorimetric data were analyzed and binding isotherms were generated with the software Origin 7.0 for MicroCal.

Imax  ; 1 + eðV1=2 V Þ= k

d

cCMP + HCN2 C-linker/CNBD 200 mM NaCl, 0.1 mM sodium citrate (pH 5.0), 16% PEG cUMP + HCN2 CLCNBD 200 mM NaCl, 0.1 mM sodium citrate (pH 4.6), 13% PEG cIMP + HCN2 CLCNBD 200 mM NaCl, 0.1 mM sodium citrate (pH 5.5), 14.5% PEG cPuMP + HCN2 CLCNBD 200 mM NaCl, 0.1 mM sodium citrate (pH 5.5), 18% PEG 2-NH2-cPuMP + HCN2 CLCNBD 200 mM NaCl, 0.1 mM sodium citrate (pH 4.6), 12% PEG

The crystals were cryo-protected in identical conditions from the crystallization mix, with the exception of the PEG400 concentration, which was at 26%. After flash cooling in liquid nitrogen, datasets were collected at either beamline 23-ID-B or 23-ID-D (GM/CA-XSD) of Advanced Photon Source, and processed using XDS (Kabsch, 2010) or HKL2000 (Otwinowski and Minor, 1997). The structures were solved via molecular replacement using Phaser with PDB: 1Q5O as the search model. Structure files for the ligands were generated with JLigand in the CCP4 suite (Lebedev et al., 2012; Winn et al., 2011). The structures were refined through alternating cycles of manual building with COOT (Emsley et al., 2010) and automated refinement in PHENIX (Adams et al., 2010). The Ramachandran outliers include 2-NH2-cPuMP co-crystal (1%), cIMP (0.5%), cCMP (0.5%), and cUMP (0.3%). All structure images were prepared using PyMOL (version 1.3, Schro¨dinger, LLC). Intersubunit interactions were analyzed with PISA (Krissinel and Henrick, 2007) and HBplus (McDonald and Thornton, 1994), and ligand interactions were determined by Contact in the CCP4 suite.

Please cite this article in press as: Ng et al., Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening, Structure (2016), http://dx.doi.org/10.1016/j.str.2016.06.024

ACCESSION NUMBERS Coordinates and structure factors for the HCN2 C terminus bound to the cyclic nucleotide analogs reported in this paper have been deposited in the PDB. The accession number for the HCN2-CNBD crystal structures reported in this paper are PDB: 5KHG (cCMP), 5KHH (cIMP), 5KHI (cPuMP), 5KHJ (cUMP) and 5KHK (2-NH2-cPuMP). New ligands were imported into the PDB component dictionary: 6SW (cIMP), 6SX (cPuMP), 6SY (cUMP), and 6SZ (2-NH2-cPuMP). SUPPLEMENTAL INFORMATION Supplemental Information includes two figures and one table and can be found with this article online at http://dx.doi.org/10.1016/j.str.2016.06.024. AUTHOR CONTRIBUTIONS L.N. and E.A.A conceived and designed the study. L.N., I.P., and V.B. conducted the experimental procedures, and analyzed and interpreted the data. F.V.P. provided guidance. L.N. and E.A.A. wrote the manuscript, and all authors reviewed and edited the manuscript. ACKNOWLEDGMENTS

Craven, K.B., Olivier, N.B., and Zagotta, W.N. (2008). C-terminal movement during gating in cyclic nucleotide-modulated channels. J. Biol. Chem. 283, 14728–14738. DiFrancesco, D., and Borer, J.S. (2007). The funny current: cellular basis for the control of heart rate. Drugs 67, 15–24. DiFrancesco, D., and Tortora, P. (1991). Direct activation of cardiac pacemaker channels by intracellular cyclic AMP. Nature 351, 145–147. Doyle, D.A., Morais Cabral, J., Pfuetzner, R.A., Kuo, A., Gulbis, J.M., Cohen, S.L., Chait, B.T., and MacKinnon, R. (1998). The structure of the potassium channel: molecular basis of K+ conduction and selectivity. Science 280, 69–77. Emery, E.C., Young, G.T., Berrocoso, E.M., Chen, L., and McNaughton, P.A. (2011). HCN2 ion channels play a central role in inflammatory and neuropathic pain. Science 333, 1462–1466. Emery, E.C., Young, G.T., and McNaughton, P.A. (2012). HCN2 ion channels: an emerging role as the pacemakers of pain. Trends Pharmacol. Sci. 33, 456–463. Emsley, P., Lohkamp, B., Scott, W.G., and Cowtan, K. (2010). Features and development of Coot. Acta Crystallogr. D. Biol. Crystallogr. 66, 486–501. Fox, K., Komajda, M., Ford, I., Robertson, M., Bohm, M., Borer, J.S., Steg, P.G., Tavazzi, L., Tendera, M., Ferrari, R., and Swedberg, K. (2013). Effect of ivabradine in patients with left-ventricular systolic dysfunction: a pooled analysis of individual patient data from the BEAUTIFUL and SHIFT trials. Eur. Heart J. 34, 2263–2270.

This work was supported by operating funds from the Canadian Institutes for Health Research (CIHR) (to E.A.A. and F.V.P.). E.A.A. is the recipient of a Tier II Canada Research Chair. S. Wong and B. Gardill provided feedback and validated the five crystal structures. This research used resources of the Advanced Photon Source, a US Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under contract no. DE-AC02-06CH11357.

Gauss, R., Seifert, R., and Kaupp, U.B. (1998). Molecular identification of a hyperpolarization-activated channel in sea urchin sperm. Nature 393, 583–587.

Received: April 8, 2016 Revised: June 22, 2016 Accepted: June 23, 2016 Published: August 25, 2016

Lebedev, A.A., Young, P., Isupov, M.N., Moroz, O.V., Vagin, A.A., and Murshudov, G.N. (2012). JLigand: a graphical tool for the CCP4 template-restraint library. Acta Crystallogr. D. Biol. Crystallogr. 68, 431–440.

REFERENCES Adams, P.D., Afonine, P.V., Bunkoczi, G., Chen, V.B., Davis, I.W., Echols, N., Headd, J.J., Hung, L.W., Kapral, G.J., Grosse-Kunstleve, R.W., et al. (2010). PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D. Biol. Crystallogr. 66, 213–221. Akimoto, M., Zhang, Z., Boulton, S., Selvaratnam, R., VanSchouwen, B., Gloyd, M., Accili, E.A., Lange, O.F., and Melacini, G. (2014). A mechanism for the auto-inhibition of hyperpolarization-activated cyclic nucleotide-gated (HCN) channel opening and its relief by cAMP. J. Biol. Chem. 289, 22205– 22220. Bahre, H., Hartwig, C., Munder, A., Wolter, S., Stelzer, T., Schirmer, B., Beckert, U., Frank, D.W., Tummler, B., Kaever, V., and Seifert, R. (2015). cCMP and cUMP occur in vivo. Biochem. Biophys. Res. Commun. 460, 909–914. Bucchi, A., Baruscotti, M., and DiFrancesco, D. (2002). Current-dependent block of rabbit sino-atrial node I(f) channels by ivabradine. J. Gen. Physiol. 120, 1–13.

Kabsch, W. (2010). XDS. Acta Crystallogr. D. Biol. Crystallogr. 66, 125–132. Krissinel, E., and Henrick, K. (2007). Inference of macromolecular assemblies from crystalline state. J. Mol. Biol. 372, 774–797.

Lee, S.Y., Lee, A., Chen, J., and MacKinnon, R. (2005). Structure of the KvAP voltage-dependent K+ channel and its dependence on the lipid membrane. Proc. Natl. Acad. Sci. USA 102, 15441–15446. Lolicato, M., Nardini, M., Gazzarrini, S., Moller, S., Bertinetti, D., Herberg, F.W., Bolognesi, M., Martin, H., Fasolini, M., Bertrand, J.A., et al. (2011). Tetramerization dynamics of C-terminal domain underlies isoform-specific cAMP gating in hyperpolarization-activated cyclic nucleotide-gated channels. J. Biol. Chem. 286, 44811–44820. Long, S.B., Campbell, E.B., and Mackinnon, R. (2005). Crystal structure of a mammalian voltage-dependent Shaker family K+ channel. Science 309, 897–903. Ludwig, A., Zong, X., Jeglitsch, M., Hofmann, F., and Biel, M. (1998). A family of hyperpolarization-activated mammalian cation channels. Nature 393, 587–591. McDonald, I.K., and Thornton, J.M. (1994). Satisfying hydrogen bonding potential in proteins. J. Mol. Biol. 238, 777–793. Otwinowski, Z., and Minor, W. (1997). Processing of x-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326.

Bucchi, A., Tognati, A., Milanesi, R., Baruscotti, M., and DiFrancesco, D. (2006). Properties of ivabradine-induced block of HCN1 and HCN4 pacemaker channels. J. Physiol. 572, 335–346.

Puljung, M.C., DeBerg, H.A., Zagotta, W.N., and Stoll, S. (2014). Double electron-electron resonance reveals cAMP-induced conformational change in HCN channels. Proc. Natl. Acad. Sci. USA 111, 9816–9821.

Bucchi, A., Baruscotti, M., Nardini, M., Barbuti, A., Micheloni, S., Bolognesi, M., and DiFrancesco, D. (2013). Identification of the molecular site of ivabradine binding to HCN4 channels. PLoS One 8, e53132.

Roubille, F., and Tardif, J.C. (2013). New therapeutic targets in cardiology: heart failure and arrhythmia: HCN channels. Circulation 127, 1986–1996.

Chow, S.S., Van Petegem, F., and Accili, E.A. (2012). Energetics of cyclic AMP binding to HCN channel C terminus reveal negative cooperativity. J. Biol. Chem. 287, 600–606. Craven, K.B., and Zagotta, W.N. (2004). Salt bridges and gating in the COOH-terminal region of HCN2 and CNGA1 channels. J. Gen. Physiol. 124, 663–677.

Santoro, B., Liu, D.T., Yao, H., Bartsch, D., Kandel, E.R., Siegelbaum, S.A., and Tibbs, G.R. (1998). Identification of a gene encoding a hyperpolarization-activated pacemaker channel of brain. Cell 93, 717–729. Saponaro, A., Pauleta, S.R., Cantini, F., Matzapetakis, M., Hammann, C., Donadoni, C., Hu, L., Thiel, G., Banci, L., Santoro, B., and Moroni, A. (2014). Structural basis for the mutual antagonism of cAMP and TRIP8b in regulating HCN channel function. Proc. Natl. Acad. Sci. USA 111, 14577–14582.

Structure 24, 1–14, October 4, 2016 13

Please cite this article in press as: Ng et al., Cyclic Purine and Pyrimidine Nucleotides Bind to the HCN2 Ion Channel and Variably Promote C-Terminal Domain Interactions and Opening, Structure (2016), http://dx.doi.org/10.1016/j.str.2016.06.024

Schlossmann, J., and Wolfertstetter, S. (2016). Identification of cCMP and cUMP substrate proteins and cross talk between cNMPs. Handbook Exp. Pharmacol. http://dx.doi.org/10.1007/164_2015_38. Schwede, F., Bertinetti, D., Langerijs, C.N., Hadders, M.A., Wienk, H., Ellenbroek, J.H., de Koning, E.J., Bos, J.L., Herberg, F.W., Genieser, H.G., et al. (2015). Structure-guided design of selective Epac1 and Epac2 agonists. PLoS Biol. 13, e1002038. Seifert, R. (2015). cCMP and cUMP: emerging second messengers. Trends Biochem. Sci. 40, 8–15. Taraska, J.W., Puljung, M.C., Olivier, N.B., Flynn, G.E., and Zagotta, W.N. (2009). Mapping the structure and conformational movements of proteins with transition metal ion FRET. Nat. Methods 6, 532–537. VanSchouwen, B., Akimoto, M., Sayadi, M., Fogolari, F., and Melacini, G. (2015). Role of dynamics in the autoinhibition and activation of the hyperpolarization-activated cyclic nucleotide-modulated (HCN) ion channels. J. Biol. Chem. 290, 17642–17654. Wainger, B.J., DeGennaro, M., Santoro, B., Siegelbaum, S.A., and Tibbs, G.R. (2001). Molecular mechanism of cAMP modulation of HCN pacemaker channels. Nature 411, 805–810.

14 Structure 24, 1–14, October 4, 2016

Winn, M.D., Ballard, C.C., Cowtan, K.D., Dodson, E.J., Emsley, P., Evans, P.R., Keegan, R.M., Krissinel, E.B., Leslie, A.G., McCoy, A., et al. (2011). Overview of the CCP4 suite and current developments. Acta Crystallogr. D. Biol. Crystallogr. 67, 235–242. Zagotta, W.N., Olivier, N.B., Black, K.D., Young, E.C., Olson, R., and Gouaux, E. (2003). Structural basis for modulation and agonist specificity of HCN pacemaker channels. Nature 425, 200–205. Zhou, L., and Siegelbaum, S.A. (2007). Gating of HCN channels by cyclic nucleotides: residue contacts that underlie ligand binding, selectivity, and efficacy. Structure 15, 655–670. Zhou, L., and Siegelbaum, S.A. (2008). Pathway and endpoint free energy calculations for cyclic nucleotide binding to HCN channels. Biophys. J. 94, L90–L92. Zhou, L., Olivier, N.B., Yao, H., Young, E.C., and Siegelbaum, S.A. (2004). A conserved tripeptide in CNG and HCN channels regulates ligand gating by controlling C-terminal oligomerization. Neuron 44, 823–834. Zong, X., Krause, S., Chen, C.C., Kruger, J., Gruner, C., Cao-Ehlker, X., Fenske, S., Wahl-Schott, C., and Biel, M. (2012). Regulation of hyperpolarization-activated cyclic nucleotide-gated (HCN) channel activity by cCMP. J. Biol. Chem. 287, 26506–26512.