Ion channels are highly specialized to respond to a wide range of environmental stimuli,
including transmembrane voltage and chemical ligands (Hille, 2001). The response of
channels to external cues causes a change in their ion conduction, which, in turn,
modifies the behavior of excitable and nonexcitable cells. The hyperpolarization-activated
and cyclic nucleotide–modified (HCN) family of ion channels, which are important for
spontaneous, pacemaking behavior in the heart and neurons, are regulated by both transmembrane
voltage and direct binding of cyclic nucleotides (e.g., cAMP) ligands (Brown and DiFrancesco,
1980; Mayer and Westbrook, 1983; DiFrancesco and Tortora, 1991).
HCN channels are composed of four subunits, each containing six transmembrane domains
(6TM). HCN channels are similar in primary structure to other 6TM channels (Gauss
et al., 1998; Ludwig et al., 1998; Santoro et al., 1998), including voltage-activated
potassium (Kv) channels and cyclic nucleotide–gated (CNG) channels (Kaupp et al.,
1989; Warmke and Ganetzky, 1994). Similar to CNG channels, HCN contains a C-terminal
cyclic nucleotide–binding domain (Zagotta et al., 2003). The mechanisms underlying
the regulation of HCN channels and other 6TM channels are the subject of intensive
work. Some common mechanistic themes are shared by 6TM channels: they have a pore
region that is selective for particular ions, a gate domain that opens to allow the
flow of ions or closes to restrict the flow of ions, and sensory domains that interact
with stimuli. The mechanism by which sensory domains are coupled to gates is not well
understood. Two studies by the Yellen laboratory in the September 2012 issue (Kwan
et al., 2012) and in this issue (see Ryu and Yellen) of the JGP cast new light on
the structural mechanism and energetics of coupling of the voltage sensor to the gate
in HCN channels.
These new findings are particularly notable in that gating in HCN channels differs
markedly from that of closely related 6TM channels. A hallmark of the 6TM domain Kv
and Ca2+-activated K (BK) channels is that they are all activated (opened) by depolarizing
voltages and closed by hyperpolarizing voltages. In contrast, HCN channels are activated
(opened) by hyperpolarizing voltages and closed by depolarizing voltages.
The voltage sensor of HCN and K channels undergoes similar movements
What is the mechanism by which HCN channels exhibit the “opposite” voltage dependence
of Kv and BK channels? Initially, one potential explanation for this difference was
the nature of the voltage-sensor domain itself. But, like Kv channels, the voltage-sensor
domain of HCN is likely composed of the S1–S4 transmembrane domains and contains a
series of positively charged residues at approximately every third position in the
fourth transmembrane domain (S4) (Männikkö et al., 2002). Furthermore, HCN voltage-sensor
movement is similar to that of Kv channels. S4 movement in HCN (Männikkö et al., 2002)
and Kv channels (Larsson et al., 1996) was monitored by making site-directed cysteine
mutations at various positions in the S4 domain and measuring the reactivity of cysteine
residues to methanethiosulfonate (MTS) reagents at depolarized and hyperpolarized
voltages. The accessibility of cysteine residues to MTS reagents was similar for HCN
and Kv channels, suggesting that voltage sensors move outward with depolarization
and inward with hyperpolarization in both HCN and Kv channels. If not the voltage
sensor, then what is responsible for the opposite voltage sensitivity of HCN channels?
The answer most likely lies in differences in the way that the voltage sensor is coupled
to the opening and closing (gating) machinery of the HCN channel.
Molecular determinants of the gate
Voltage-sensor movement in Kv channels appears to be coupled to pore opening through
the S4–S5 linker region and the lower part of the S6 region beneath the K channel
pore (Lu et al., 2002; Tristani-Firouzi et al., 2002; Pathak et al., 2007) (Fig. 1,
top). Structural information based on Kv1.2 channels also shows proximity of the S4–S5
linker with the S6 region of the same subunit (Long et al., 2005). In HCN channels,
the S4–S5 linker, S6, and C-linker (a region linking the S6 domain to the cyclic nucleotide–binding
domain) regions have also been implicated in voltage-dependent gating (Chen et al.,
2001; Rothberg et al., 2003; Decher et al., 2004; Bell et al., 2009) (Fig. 1, bottom).
The structure that gates the flow of ions also appears to be in the same place (the
bottom of the S6 domain) in HCN channels and Kv channels (Liu et al., 1997; Doyle
et al., 1998; Rothberg et al., 2002). Thus, not only are the voltage sensor and voltage-sensor
movement similar in HCN and Kv channels, the molecular regions involved in voltage-dependent
gating are also similar. Thus, the prevailing hypothesis for the opposite voltage
sensitivity of HCN channels and Kv channels is that the voltage sensor must be coupled
to the gate differently. But, what exactly about the coupling is different?
Figure 1.
Proposed interdomain interactions that make up the activation gate in Kv channels
and HCN channels. (Top) Schematic showing two subunits of a four-subunit Kv channel.
The S1–S4 transmembrane domains make up the voltage-sensing domain, which is linked
to the pore-forming regions (S5–S6 domains) via the S4–S5 linker region. Potassium
ions are conducted through the central pore region. Amino and carboxyl termini are
intracellular. The lower part of the S6 domain is depicted as a separate cylinder.
The lower part of the S6 domain is positioned beneath the S4–S5 linker from the same
subunit. In response to hyperpolarization, the S4 of the voltage sensor moves downward
and pushes down on the S4–S5 linker region, which, in turn, pushes down on the lower
part of the S6 domain, which brings the S6 domains closer together to narrow and close
the channel gate, which restricts ion flow. (Bottom) Schematic showing two subunits
of a four-subunit HCN channel. Like Kv channels, HCN channels have an S1–S4 voltage-sensor
domain linked via an S4–S5 linker to the pore-forming S5 and S6 domains. Like Kv channels,
ions are conducted through a central pore. N- and C-terminal regions are intracellular.
Distinct from Kv channels, the C-terminal region of HCN channels contains a post-S6/C-linker
domain and a cyclic nucleotide–binding domain. Cyclic nucleotide monophosphate (cNMP)
is depicted as bound to the channel. (The C-terminal region of the HCN subunit on
the right is cut away for clarity.) Intrasubunit interactions are depicted between
364 in the S4–S5 linker region and 472 of the post-S6/C-linker region. This interaction
takes place in the open state. The dashed line connected to the post-S6/C-linker domain
indicates that amino acid 476 of the same post-S6/C-linker domain makes an intersubunit
interaction with 364 on an adjacent subunit. This interaction takes place in the open
state. Cd2+ bridges indicate the close proximity of the S4–S5 linker and post-S6/C-linker
regions. The red arrow indicates movement from a lock-open (364 linked to 472) to
a lock-closed (364 linked to 476) state.
The structural mechanism of gating differs between HCN and Kv channels
New insights into the mechanism of coupling comes from “lock-open” and “lock-closed”
HCN channels, in which experimental conditions favor either the open or closed state
of the channel. To perform lock-open and lock-closed experiments, the Yellen group
introduced pairs of cysteine mutations at specific sites and recorded currents in
the absence and then the presence of nanomolar levels of Cd2+ ion. A functional change
in channel gating suggests that the Cd2+ ion forms a metal bridge between two cysteine
residues (Holmgren et al., 1998; Shin et al., 2004; Prole and Yellen, 2006), and implies
that the two cysteine residues come into close proximity.
Kwan et al. (2012) used this technique to examine the proximity between three sites
in the S4–S5 linker and eight sites in the post-S6 and adjacent A′ helix of the C-linker
domain of the sea urchin HCN channel, spHCN. The authors introduced one cysteine mutation
in the S4–S5 linker and a second cysteine mutation at a site in the S6/C-linker region
and added Cd2+ to induce metal bridge formation (Kwan et al., 2012). Extending their
earlier findings (Prole and Yellen, 2006), they found that multiple S4–S5 linker sites
were in close proximity to S6/C-linker sites. They also found that the S6/C-linker
sites that had lock-open and locked-closed effects were interleaved; for instance,
residue 364C (located in the S4–S5 linker) had a lock-open effect with 472C, a locked-closed
effect with 474C, and a lock-open effect with residue 482C. To explain these results,
the authors proposed the provocative idea that one part of the S4–S5 linker contacts
the S6/C-linker in the same subunit, and another part of the S4–S5 linker contacts
the S6/C-linker of an adjacent subunit (Fig. 1, bottom). Testing this proposal with
concatenated dimers, they found some sites that were consistent with intrasubunit
interactions and some sites that were consistent with intersubunit interactions (Fig.
1). They interpreted the pattern of intrasubunit and intersubunit interactions to
mean that movements of the S4–S5 linkers relative to the S6 region in HCN channels
were different from those proposed for Kv1.2 channels, where the S4–S5 linker of a
subunit sits above the lower S6 region of the same subunit, likely making only an
intrasubunit interaction (Fig. 1, top) (Long et al., 2005; Pathak et al., 2007). This
implies that the precise structural interactions among the very similar domains involved
in HCN and Kv channel gating may produce very different responses to voltage.
The functional data from the Cd2+ bridge experiments in HCN channels do not fit well
with existing structural and structural modeling results from Kv channels. Structural
and computational data suggest that to close Kv channels, the voltage sensor pushes
down on the S4–S5 linker region and that the S4–S5 linker pushes down (i.e., inward
toward the cytoplasm) on the lower part of the S6 domain, thereby restricting the
opening of the channel gate and reducing the flow of ions (Fig. 1, top). In contrast,
for HCN channels the S4–S5 linker (in particular site 364C) makes locked-open intrasubunit
interactions (with 472C), locked-closed intersubunit interactions (with 476), and
locked-open intersubunit interactions (with 482C and/or 485C) in the lower S6 (Fig.
1, bottom). To explain these results, the authors propose a structural model of gating
in which, with hyperpolarization, the voltage sensor moves downward and moves the
S4–S5 linker, especially the lower S5 portion. Movement of the lower S5 allows S6
helices to rotate and move outward from the central axis of the channel, causing channel
opening. This fundamental new mechanism for HCN channel gating raises many testable
questions. Ideally, new structural information about the S4–S5 linker and S6 gate
region of HCN channels would help to test the structural models derived from functional
data.
The energetics of coupling differs between HCN and Kv channels
The energetics of coupling between the voltage sensor and the gate in HCN channels
are also not well understood. Voltage-sensor movement in Kv channels is strict; in
other words, voltage-sensor movement is tightly coupled to channel opening. For instance,
in Kv channels, it is thought that the activation (or “up” configuration) of all four
voltage sensors is necessary for the channel gate to open (Zagotta et al., 1994; Gagnon
and Bezanilla, 2009). Therefore, at very negative voltages, when the voltage sensors
are “down” or resting, the probability of Shaker K channels opening is low (Islas
and Sigworth, 1999). Strict coupling of the voltage-sensor movement to the gate implies
that when the gate is held open (as with an inactivation particle; Bezanilla et al.,
1991), gating charge movement (and thus voltage-sensor movement) is immobilized. Linear
gating models, that is to say systems in which there are a sequential number of closed
states before an open state, are often sufficient to describe gating behavior in strictly
coupled channels (Zagotta et al., 1994).
The Yellen group previously discovered that the link between the voltage sensor and
the activation gate might be weak. They found that, in response to a hyperpolarizing
voltage command (in the absence of cAMP), spHCN channels first activated and then
reclosed as a result of an uncoupling or slippage between the activated voltage sensor
and the gate (Shin et al., 2004). In other words, despite the HCN voltage sensor being
in a “down” conformation (the activated conformation for HCN channels) because of
hyperpolarization, the channel gate slipped closed and the channels did not conduct
ionic currents. Other laboratories have also tested the voltage dependence of HCN
channel gating and have had success with fitting a Monod–Wyman–Changeux (MWC) model
to the voltage-dependent activation kinetics similar to one used to investigate voltage-activated
gating in BK channels (Cox et al., 1997). In the MWC models for HCN channels (Altomare
et al., 2001; Wang et al., 2002; Bruening-Wright et al., 2007), the channels can undergo
an opening transition (i.e., go from the closed to the open state) that is stabilized
by a constant value for each voltage sensor that goes from a resting to an active
state. Likewise, channels in an open state stabilize the transition of voltage sensors
from a resting state to an active state. In MWC models, unlike the linear models for
Shaker K channels where voltage-sensor movement is obligatory for channel opening,
channels can also go from the closed to open state in the absence of voltage-sensor
movement. Using fluorophores to label the S4 domain of HCN channels, Bruening-Wright
et al. (2007) determined that movement of two voltage sensors was sufficient to activate
HCN channels, an observation inconsistent with sequential models of activation but
consistent with MWC models of activation.
In another recent study, Ryu and Yellen (2012) determined the coupling factor between
HCN voltage sensors and activation gates by measuring gating currents (currents associated
with the movement of the voltage sensor) from lock-open or lock-closed HCN channels.
This approach allowed them to isolate measurements from channels with deactivated
voltage sensors or channels with activated voltage sensors. Isolating the state of
the voltage sensors allowed them to simplify the 10-state MWC model and more directly
determine a coupling constant (a measure of the ability of the voltage sensor to affect
channel opening and the ability of opening to affect the state of the voltage sensor).
They found that gating charge moved more easily (was activated at less negative voltages)
in the locked-open channels and less easily (was activated at more negative voltages)
in locked-closed channels.
The coupling factor of the voltage sensor to the gate was much smaller in HCN channels
(7.2 ± 3.0–fold per voltage sensor in lock-open channels) than in Kv channels (>100-fold
per voltage sensor) (Islas and Sigworth, 1999) or BK channels (15-fold per voltage
sensor). The voltage-sensor coupling to the activation gate in HCN channels is weaker
than in Kv channels and may be more like that of ligand-binding coupling to the activation
gate of ligand-gated channels.
The significance of the much weaker coupling in HCN compared with voltage-gated K
channels likely means that other factors, such as cyclic nucleotides, can make an
energetic contribution to open the HCN channel gate. This is a fascinating idea, and
it will be interesting to determine whether, similar to models of voltage and Ca2+
activation of BK channels (Horrigan and Aldrich, 2002), cAMP activation of HCN can
be added as a module to the mathematical models of voltage-dependent HCN channel gating.