Mg2+ is a Missing Link in Plant Cell Ca2+ Signalling and Homeostasis—A Study on Vicia faba Guard Cells

Hyperpolarization-activated calcium channels (HACCs) are found in the plasma membrane and tonoplast of many plant cell types, where they have an important role in Ca2+-dependent signalling. The unusual gating properties of HACCs in plants, i.e., activation by membrane hyperpolarization rather than depolarization, dictates that HACCs are normally open in the physiological hyperpolarized resting membrane potential state (the so-called pump or P-state); thus, if not regulated, they would continuously leak Ca2+ into cells. HACCs are permeable to Ca2+, Ba2+, and Mg2+; activated by H2O2 and the plant hormone abscisic acid (ABA); and their activity in guard cells is greatly reduced by increasing amounts of free cytosolic Ca2+ ([Ca2+]Cyt), and hence closes during [Ca2+]Cyt surges. Here, we demonstrate that the presence of the commonly used Mg-ATP inside the guard cell greatly reduces HACC activity, especially at voltages ≤ −200 mV, and that Mg2+ causes this block. Therefore, we firstly conclude that physiological cytosolic Mg2+ levels affect HACC gating and that channel opening requires either high negative voltages (≥−200 mV) or displacement of Mg2+ away from the immediate vicinity of the channel. Secondly, based on structural comparisons with a Mg2+-sensitive animal inward-rectifying K+ channel, we propose that the likely candidate HACCs described here are cyclic nucleotide gated channels (CNGCs), many of which also contain a conserved diacidic Mg2+ binding motif within their pores. This conclusion is consistent with the electrophysiological data. Finally, we propose that Mg2+, much like in animal cells, is an important component in Ca2+ signalling and homeostasis in plants.


Introduction
Ca 2+ has long been recognized as an essential component in many plant cellular processes. In order for Ca 2+ to function as an intracellular signal, temporal-, spatial-, and stimulus-specific changes in [Ca 2+ ] Cyt need to be tightly controlled [1]. Ca 2+ influx into plant cells is achieved by three main types of Ca 2+ channels [2,3]: firstly, channels that show little or no voltage sensitivity, referred to as non-selective calcium channels (NSCCs), primarily for being active at all voltages and for being less selective for calcium over monovalent cations, such as K + and Na + [4]; secondly, channels that show high voltage-dependence, such as those activated by depolarization (DACCs) [5]; thirdly, others that are activated only by membrane hyperpolarization (HACCs) [6], which are the focus of this report. as well as other intracellular substrates, are required to fully activate a plasma membrane electrogenic ion pump capable of hyperpolarizing the membrane to around −140 mV, a potential well beyond the activation threshold for I K,in [35].
Given the importance of calcium channels in plant cellular signalling, including in guard cell aperture regulation, we address the question of whether internal Mg 2+ can affect the activity HACCs in guard cell protoplasts (GCPs) and show examples of Gd 3+ -sensitive Ba 2+ currents (I Ba ) activated by hyperpolarization with and without Mg 2+ in the patch pipet. We also assess important properties of I Ba in the absence of Mg 2+ , such as the permeability and sensitivity to some relevant inorganic compounds and physiological effectors, including abscisic acid (ABA) and cAMP.

Solutions
Protoplasts were placed in a 0.5 mL chamber, left to settle down, and then perfused continuously at a flow rate of ≈0.5-1 mL/min. To record I Ba currents through HACC, we used barium-containing solutions. The bath medium contained (in mM): 100 BaCl 2 , 10 Mes (pH 5.5 with Tris base); the pipette contained (in mM): 100 BaCl 2 , 4 EGTA, 10 Hepes (pH 7.5 with Tris). For experiments where I K,in and I Ba measurements were made on the same GCPs, a different bath and internal solutions as follows. Bath (in mM): 30 KCl, 10 Mes (pH 5.5 with Tris base) to measure I K,in , which was replaced by 100 BaCl 2 , 10 Mes (pH 5.5 with Tris base) to measure I Ba . Internal solution (in mM): 1 BaCl 2 , 18 KCl, 4 EGTA, 10 Hepes (pH 7.5 with Tris base). Mg-ATP, MgCl 2 , or K 2 -ATP were added as specified in the figures. Osmolality was adjusted with sorbitol to 310 mOsm.kg −1 . For classic solutions used to measure I K,in refer to [38]. ABA was added externally. All chemicals were from Sigma Chemical, Poole, Dorset, UK. The membrane permeable cAMP analog Bt 2 cAMP was solubilized in deionized water and stored in aliquots of 50-100 µL at a concentration of 0.1 M. Bt 2 cAMP was diluted to the final desired concentration just a few minutes before its use.

Current-Voltage Recording and Analysis
Patch pipettes (5)(6)(7)(8)(9)(10) were pulled from Kimax-51 glass capillaries (Kimble 34500; Kimble, Owens-Illinois) using a two-stage puller (Narishige PP-83, Japan). Experiments were performed at room temperature (20 to 22 • C) using standard whole-cell patch clamp techniques, with an Axopatch 200B integrating patch clamp amplifier (Axon Instruments, Inc., Union City, CA, USA). Voltage commands and simultaneous signal recordings and analyses were assessed by a microcomputer connected to the amplifier via a multipurpose input-output device (Digidata 1320A) using pClamp software (versions 8.0 and 10; Axon Instruments, Inc.). After giga ohm seals were formed, the whole-cell configuration was then achieved by gentle suction and the membrane was immediately clamped to a holding voltage (h v ) of −36 mV. GCPs were continuously perfused throughout the experiment and current recordings began only after at least 5-10 min from going into whole-cell mode to allow for intracellular equilibrium between the cytoplasm and patch pipet solution. All current traces shown were low-pass filtered at 2 kHz before analog-to-digital conversion and were uncorrected for leakage current or capacitive transients. Membrane potentials were corrected for liquid junction potential as described in [39]. Ionic activities were calculated using GEOCHEM-EZ [40]. Current-voltage (I-V) relationships for I Ba and I K,in were plotted as steady-state currents vs. test potentials when using the square pulse stimulations or utilizing the "trace vs. trace" feature of Clampfit analysis when using voltage ramps. Unless otherwise stated, every experiment reported here was repeated a minimum of three times and data were graphed as mean ± SEM.

Protein Structural Modelling
Models were built using Swiss-Model [41] and structures were visualized with PyMOL (PyMOL Molecular Graphics System, Version 1.5.0.4 Schrödinger, LLC). Homology modelling was supported by PSIPRED [42] and multisequence alignments were produced with MUSCLE [43]. Alignments and additional information of the models are provided in the respective legends.

Removal of Intracellular Mg-ATP Unveils a Larger Instantaneously-Activated, Inwardly-Directed, and Gd 3+ -Sensitive Ba 2+ Current
We recorded Ba 2+ currents (I Ba ) in guard cell protoplasts in either the presence or absence of Mg-ATP ( Figure 1A). At the end of the trials we also tested for I Ba sensitivity to Gd 3+ , a potent blocker of I Ba ( Figure 1B). The recorded currents were generated in response to a square pulse protocol from +64 mV to −256 mV in increments of −20 mV, with the holding voltage (h V ) set to −36 mV. In the presence of Mg-ATP ( Figure 1A; left panel), a small instantaneous inward-rectifying current of ≈−20 pA started to activate at around −200 mV, reaching a maximum of −60 pA at −256 mV. As expected, this current was sensitive to the addition of extracellular Gd 3+ (see Figure 1B; left panel). Meanwhile, in the absence of Mg-ATP ( Figure 1A; right panel), much larger (≥8-fold) instantaneous rectifying currents were recorded, which happened to also be Gd 3+ -sensitive ( Figure 1B; left traces). The current-voltage (I-V) relationships in the presence (•; n = 3) or absence ( ; n = 7) of Mg-ATP are represented in Figure 1C (left panel). We also plotted the effect of Gd 3+ on the I-V relationships in the presence ( ) or absence ( ) of Mg-ATP ( Figure 1C; right panel). The reversal potential (Erev) is close to the reversal potential for Ba 2+ (EBa). Typical current trace recorded in the absence of Mg-ATP (below) in response to a voltage ramp which consists of activating IBa using a hyperpolarization square pulse to −156 mV and immediately followed by a continuous depolarizing ramp to +64 mV (the voltage protocol is depicted above), with a slope of 0.7 V s −1 . Scale bars are shown below the current traces and to the left of the voltage protocol.
These results highlight that when omitting Mg-ATP from the intracellular medium, a larger Gd 3+ -sensitive, inwardly rectifying Ba 2+ current is unveiled, which activates at significantly less- Removal of intracellular Mg-ATP unveils a larger instantaneously-activated, inwardly-directed, and Gd 3+ -sensitive Ba 2+ current. (A) Typical examples of current traces in whole-cell mode obtained from two separate V. faba guard cells with the pipette solution either containing (left traces) or lacking (right traces) Mg-ATP (1 mM). The pulse protocol mostly used throughout this study consisted of 0.6 s long square voltage pulses ranging from +64 to −256 mV in −20 mV increments; the holding potential h v was set to −36 mV. In order to preserve the quality of the "giga" seals, guard cell protoplasts (GCPs) with no Mg-ATP in the pipette were not subjected to higher voltages beyond The reversal potential (E rev ) is close to the reversal potential for Ba 2+ (E Ba ). Typical current trace recorded in the absence of Mg-ATP (below) in response to a voltage ramp which consists of activating I Ba using a hyperpolarization square pulse to −156 mV and immediately followed by a continuous depolarizing ramp to +64 mV (the voltage protocol is depicted above), with a slope of 0.7 V s −1 . Scale bars are shown below the current traces and to the left of the voltage protocol.
These results highlight that when omitting Mg-ATP from the intracellular medium, a larger Gd 3+ -sensitive, inwardly rectifying Ba 2+ current is unveiled, which activates at significantly less-negative voltages (see the shift to the right of ≥−100 mV in the I-V plot). Furthermore, the currents recorded in 0 Mg-ATP seem to reverse near the calculated Nernst equilibrium potential for Ba 2+ (E Ba ≈ +28 mV), and are far removed from E Cl (−54 mV). Likewise, using fast depolarization ramps (0.07 V/s) after activating the current I Ba with a square pulse to −156 mV (see voltage protocol and current trajectories in Figure 1D; left), a reversal potential of +17 mV was measured, again close to E Ba rather than E Cl (see Figure 1D, zoomed I-V plot). This is also indicative of the higher permeability of this conductance to Ba 2+ as compared to Cl − .
To test whether changing external Ba 2+ concentration will affect current magnitude, as well as the I-V relationship, GCPs were patched in whole cell mode using the Mg-ATP-free internal solution ( Figure 2). negative voltages (see the shift to the right of ≥−100 mV in the I-V plot). Furthermore, the currents recorded in 0 Mg-ATP seem to reverse near the calculated Nernst equilibrium potential for Ba 2+ (EBa ≈ +28 mV), and are far removed from ECl (−54 mV). Likewise, using fast depolarization ramps (0.07 V/s) after activating the current IBa with a square pulse to −156 mV (see voltage protocol and current trajectories in Figure 1D; left), a reversal potential of +17 mV was measured, again close to EBa rather than ECl (see Figure 1D, zoomed I-V plot). This is also indicative of the higher permeability of this conductance to Ba 2+ as compared to Cl − . To test whether changing external Ba 2+ concentration will affect current magnitude, as well as the I-V relationship, GCPs were patched in whole cell mode using the Mg-ATP-free internal solution ( Figure 2). Once again, the current magnitudes recorded in the absence of Mg-ATP are substantial. For instance, at −196 mV, a −280 pA current is measured in 100 mM [Ba 2+ ] out (Figure 2A), while in 30 mM [Ba 2+ ] out ( Figure 2B) the same voltage gives rise to a current value of −220 pA. All current magnitudes at any given voltage are decreased when switching to lower [Ba 2+ ] in the bath. The corresponding I-V plots appear as shifting to negative values ( Figure 2C) when switching from 100 to 30 mM Ba 2+ , which is accompanied by a negative shift in the apparent reversal potential (E rev ) values. Indeed, when zooming in ( Figure 2C, inset), the apparent E rev shows a negative shift of ≈−14 mV as a result of this Ba 2+ concentration change. Furthermore, the apparent E rev in 30 mM [Ba 2+ ] out (≈+8 mV) is still closer to the calculated E Ba rather than E Cl , the values for which in this case are +17.7 and −25.7 mV, respectively. This experiment was repeated with 100, 30, and 10 mM BaCl 2 in the bath, with the same qualitative effects observed, i.e., decrease of current amplitude when decreasing the [Ba2+] o and negative shift in the apparent E rev .

HACC Permeability Sequence for Divalent Cations in the Absence of Mg-ATP: Ba > Ca ≈ Sr ≈ Mn >> Mg
Guard cell permeability to other divalent cations such as Ca 2+ , Sr 2+ , Mn 2+ , and Mg 2+ in the absence of Mg-ATP was also tested ( Figure 3). respectively. This experiment was repeated with 100, 30, and 10 mM BaCl2 in the bath, with the same qualitative effects observed, i.e., decrease of current amplitude when decreasing the [Ba2+]o and negative shift in the apparent Erev.

HACC Permeability Sequence for Divalent Cations in the Absence of Mg-ATP: Ba > Ca ≈ Sr ≈ Mn >> Mg
Guard cell permeability to other divalent cations such as Ca 2+ , Sr 2+ , Mn 2+ , and Mg 2+ in the absence of Mg-ATP was also tested ( Figure 3). As expected, HACC was permeable to Ca 2+ ( Figure 3A) and found to have similar permeability to both Sr 2+ ( Figure 3A) and Mn 2+ ( Figure 3B). Meanwhile, Mg 2+ did not permeate HACC ( Figure 3A). The I-V plots ( Figure 3A,B) summarize the permeability data, i.e., the lack of HACC permeability to Mg 2+ , as well as the much larger permeability to Ba 2+ when compared to either Ca 2+ , Sr 2+ , or Mn 2+ . The effect of Mn 2+ ions over time is reported in Figure 3B and highlights the unique behavior of this ion. Unlike Ca 2+ or Sr 2+ , Mn 2+ ( Figure 3B) triggered a transient blockage of HACC followed by some current recovery, while still washing out the Ba 2+ and replacing it with 100 mM Mn 2+ . This transient blockage As expected, HACC was permeable to Ca 2+ ( Figure 3A) and found to have similar permeability to both Sr 2+ ( Figure 3A) and Mn 2+ ( Figure 3B). Meanwhile, Mg 2+ did not permeate HACC ( Figure 3A). The I-V plots ( Figure 3A,B) summarize the permeability data, i.e., the lack of HACC permeability to Mg 2+ , as well as the much larger permeability to Ba 2+ when compared to either Ca 2+ , Sr 2+ , or Mn 2+ . The effect of Mn 2+ ions over time is reported in Figure 3B and highlights the unique behavior of this ion. Unlike Ca 2+ or Sr 2+ , Mn 2+ ( Figure 3B) triggered a transient blockage of HACC followed by some current recovery, while still washing out the Ba 2+ and replacing it with 100 mM Mn 2+ . This transient blockage effect was repeated on two other guard cell protoplasts but was never seen with either Ca 2+ or Sr 2+ , nor was it seen with Mg 2+ , even after 20 min of washing out the Ba 2+ .

HACC Permeability Sequence for Monovalent Cations in the Absence of Mg-ATP
We observed that this Gd 3+ -sensitive HACC is also permeable to some physiologically relevant monovalent cations, such as K + (Figure S1A), Na + (Figure S1B), and Cs + ( Figure S1C), but not tetraethylammonium (TEA + ) ( Figure S1D). These data indicate that the Gd 3+ -sensitive current characterized in this work does not select for small mono or divalent cations (except for the case of Mg 2+ and the bigger cation TEA + ). Chloride also does not seem to permeate through this HACC. Indeed, when 100 mM Clwas added at the same time as TEA + (see Figure S1D), no current could be detected, indicating that Clis as impermeable as TEA + . Qualitatively, the same effect was seen in all patched GCPs (n = 5 for K + and Na + ; n = 2 for Cs + and TEA + ).

Effect of Blockers of I Ba in the Absence of Mg-ATP and Comparison with the Effect on I K,in
In order to further characterize this HACC, which was readily unmasked when Mg-ATP was omitted from the patch pipet, the effect of some classical blockers such as the lanthanides (La 3+ and Gd 3+ ), Mn 2+ , Cs + , and TEA + were tested on I Ba , as well as on I K,in, (the other major conductance that activates upon hyperpolarization in guard cells) (Figure 4). effect was repeated on two other guard cell protoplasts but was never seen with either Ca 2+ or Sr 2+ , nor was it seen with Mg 2+ , even after 20 minutes of washing out the Ba 2+ .

HACC Permeability Sequence for Monovalent Cations in the Absence of Mg-ATP: K ≈ Na > Ba > Cs >> TEA.
We observed that this Gd 3+ -sensitive HACC is also permeable to some physiologically relevant monovalent cations, such as K + (Figure S1A), Na + (Figure S1B), and Cs + ( Figure S1C), but not tetraethylammonium (TEA + ) ( Figure S1D). These data indicate that the Gd 3+ -sensitive current characterized in this work does not select for small mono or divalent cations (except for the case of Mg 2+ and the bigger cation TEA + ). Chloride also does not seem to permeate through this HACC. Indeed, when 100 mM Clwas added at the same time as TEA + (see Figure S1D), no current could be detected, indicating that Clis as impermeable as TEA + . Qualitatively, the same effect was seen in all patched GCPs (n = 5 for K + and Na + ; n = 2 for Cs + and TEA + ).

Effect of Blockers of IBa in the Absence of Mg-ATP and Comparison with the Effect on IK,in.
In order to further characterize this HACC, which was readily unmasked when Mg-ATP was omitted from the patch pipet, the effect of some classical blockers such as the lanthanides (La 3+ and Gd 3+ ), Mn 2+ , Cs + , and TEA + were tested on IBa, as well as on IK,in, (the other major conductance that activates upon hyperpolarization in guard cells) (Figure 4).  One of the most conspicuous effects lies in the potent effect of Gd 3+ in blocking I Ba ( Figure 4A), even when used at relatively low concentrations (20 to 100 µM), while the same concentrations of Gd 3+ had no effect on I K,in ( Figure 4A). An even higher concentration of Gd 3+ (500 µM) did not affect I K,in (data not shown). La 3+ also blocked I Ba measured in 0 Mg-ATP, but we found that much higher concentrations of La 3+ (0.2 to 0.5 mM) are needed to achieve the same block as compared to Gd 3+ (data not shown). Experimenting with cesium, a blocker of I K,in , hardly any effect on I Ba was registered ( Figure 4B). Even though Cs + was used at concentrations up to 1 mM, it had no or only a small effect on I Ba , while one-tenth of this amount (0.1 mM) is sufficient to block a large proportion (≈ 80% or more) of I K,in ( Figure 4B). Furthermore, Mn 2+ used at 2 mM inhibited HACC by ≈37% (at V = −196 mV) when the charge carrier (in this case Ba 2+ ) was still present in the bath ( Figure 4C). Increasing Mn 2+ concentration to 10 mM shows that Mn 2+ is not an efficient blocker of I Ba as compared to Gd 3+ or La 3+ , and 10 mM Mn 2+ only causes an extra 20% I Ba inhibition (see IV plot in Figure 4C). Finally, 10 mM tetraethylammonium chloride (TEA), a concentration that was shown to block 70% to 80% of I K,in in intact guard cells [44], had no effect whatsoever on I Ba ( Figure 4D) measured in 0 Mg-ATP.

Rapid Enhancement of I Ba by ABA
Given that in V. faba guard cells a HACC was implicated downstream of ABA in stomatal movements, we tested whether ABA affects I Ba activated in the absence of internal Mg-ATP. We patch-clamped guard cells to measure I Ba currents under baseline conditions, i.e., zero Mg-ATP inside and no added ABA outside ( Figure 5). Gd 3+ had no effect on IK,in ( Figure 4A). An even higher concentration of Gd 3+ (500 µM) did not affect IK,in (data not shown). La 3+ also blocked IBa measured in 0 Mg-ATP, but we found that much higher concentrations of La 3+ (0.2 to 0.5 mM) are needed to achieve the same block as compared to Gd 3+ (data not shown). Experimenting with cesium, a blocker of IK,in, hardly any effect on IBa was registered ( Figure 4B). Even though Cs + was used at concentrations up to 1 mM, it had no or only a small effect on IBa, while one-tenth of this amount (0.1 mM) is sufficient to block a large proportion (≈ 80% or more) of IK,in ( Figure 4B). Furthermore, Mn 2+ used at 2 mM inhibited HACC by ≈ 37% (at V = −196 mV) when the charge carrier (in this case Ba 2+ ) was still present in the bath ( Figure 4C). Increasing Mn 2+ concentration to 10 mM shows that Mn 2+ is not an efficient blocker of IBa as compared to Gd 3+ or La 3+ , and 10 mM Mn 2+ only causes an extra 20% IBa inhibition (see IV plot in Figure 4C). Finally, 10 mM tetraethylammonium chloride (TEA), a concentration that was shown to block 70% to 80% of IK,in in intact guard cells [44], had no effect whatsoever on IBa ( Figure 4D) measured in 0 Mg-ATP.

Rapid Enhancement of IBa by ABA
Given that in V. faba guard cells a HACC was implicated downstream of ABA in stomatal movements, we tested whether ABA affects IBa activated in the absence of internal Mg-ATP. We patch-clamped guard cells to meas After about 10 min, the time usually necessary to reach steady-state conditions, we switched the perfusion solution to the one containing 20 µM ABA. A rapid and pronounced increase of I Ba currents is seen at all voltages between −100 and −200 mV (≥1.3-to 1.5-fold) after only 5 min of ABA treatment ( Figure 5B), and a near doubling of the size of the I Ba currents occurs at 10 min ( Figure 5A,B). The enhancement of I Ba in response to ABA, especially at 10 min, spans from −60 to −200 mV, and also appears to shift the activation threshold of I Ba ( Figure 5B) to the right. This suggests that ABA not only enhances calcium entry through HACCs but can also mobilize calcium entry at less negative voltages.

Characterization of the Effect of Mg 2+ on I Ba and I K,in
To answer whether internal Mg 2+ alone causes the block of I Ba when we add Mg-ATP, GCPs were patched either without Mg-ATP or without ATP, but with added Mg 2+ (added as MgCl 2 ) ( Figure 6). ( Figure 5A,B). The enhancement of IBa in response to ABA, especially at 10 minutes, spans from -60 to −200 mV, and also appears to shift the activation threshold of IBa ( Figure 5B) to the right. This suggests that ABA not only enhances calcium entry through HACCs but can also mobilize calcium entry at less negative voltages.

Characterization of the Effect of Mg 2+ on IBa and IK,in
To answer whether internal Mg 2+ alone causes the block of IBa when we add Mg-ATP, GCPs were patched either without Mg-ATP or without ATP, but with added Mg 2+ (added as MgCl2) ( Figure 6). Furthermore, and as a control for the "ion transport functionality" of the patched GCPs, we used conditions that allow measurements of not just IBa, but also to record IK,in. The experiments were started in conditions allowing to probe for IK,in with KCl (30 mM) in the bath, which was then followed by replacing the KCl with a solution containing BaCl2 (100 mM). This was done first in the absence of both intracellular Mg 2+ and ATP ( Figure 6A), and then repeated on another batch of guard cells with internal medium containing 1 mM MgCl2 but no ATP ( Figure 6B). Firstly, in zero Mg-ATP, the only current that activated in response to hyperpolarization was IBa ( Figure 6A). IK,in could not be activated. Secondly, when only Mg 2+ was included in the pipet solution (no ATP added), IK,in could then be activated in 30 mM K + . However, when switching the bath from K + -to Ba 2+ -containing media, IBa currents vanished, indicating that including only Mg 2+ in the patch pipet can cause the blockage of IBa at voltages where it is normally activated in 0 Mg-ATP.

Can cAMP Activate HACCs in V. faba Guard Cells Despite the Presence of Intracellular Mg 2+ ?
We know from our own previous work [15] that cAMP activated a Gd 3+ -sensitive HACC in guard cells while recording in Mg 2+ -and ATP-free media ( Figure 7A). However, since we had shown that this conductance discriminates poorly between divalent and monovalent cations, a hallmark Furthermore, and as a control for the "ion transport functionality" of the patched GCPs, we used conditions that allow measurements of not just I Ba , but also to record I K,in . The experiments were started in conditions allowing to probe for I K,in with KCl (30 mM) in the bath, which was then followed by replacing the KCl with a solution containing BaCl 2 (100 mM). This was done first in the absence of both intracellular Mg 2+ and ATP ( Figure 6A), and then repeated on another batch of guard cells with internal medium containing 1 mM MgCl 2 but no ATP ( Figure 6B). Firstly, in zero Mg-ATP, the only current that activated in response to hyperpolarization was I Ba ( Figure 6A). I K,in could not be activated. Secondly, when only Mg 2+ was included in the pipet solution (no ATP added), I K,in could then be activated in 30 mM K + . However, when switching the bath from K + -to Ba 2+ -containing media, I Ba currents vanished, indicating that including only Mg 2+ in the patch pipet can cause the blockage of I Ba at voltages where it is normally activated in 0 Mg-ATP.

Can cAMP Activate HACCs in V. faba Guard Cells Despite the Presence of Intracellular Mg 2+ ?
We know from our own previous work [15] that cAMP activated a Gd 3+ -sensitive HACC in guard cells while recording in Mg 2+ -and ATP-free media ( Figure 7A). However, since we had shown that this conductance discriminates poorly between divalent and monovalent cations, a hallmark characteristic of all animal and plant cyclic nucleotide gated channels (CNGCs) [13,45], we sought to check if the Gd 3+ -sensitive I Ba is also gated by cAMP in conditions where intracellular Mg 2+ is present and I Ba is already blocked ( Figure 7B). characteristic of all animal and plant cyclic nucleotide gated channels (CNGCs) [13,45], we sought to check if the Gd 3+ -sensitive IBa is also gated by cAMP in conditions where intracellular Mg 2+ is present and IBa is already blocked ( Figure 7B). In the absence of db2cAMP (the lipophilic permeable analog of cAMP), and as expected only a small background IBa current is seen (≤ 10 pA around −190 mV). After perfusing with db2cAMP (1 mM), a substantial increase in IBa amplitude between voltages from around −30 to −190 mV was observed (>60 pA around −190 mV). Keeping db2cAMP in the bath and adding Gd 3+ (50 µM) resulted in a total block of the current (≤ 4 pA around −190 mV). This may indicate that GCPs harbor CNGCs that can be activated by cAMP despite the blocking effect by Mg 2+ .

Identification of Candidate Mg 2+ -Dependent Cation Channels in Arabidopsis thaliana
Given the data from the physiological experiments, we undertook a search for candidate Mg 2+dependent cation channels in plants (Figure 8). In the absence of db 2 cAMP (the lipophilic permeable analog of cAMP), and as expected only a small background I Ba current is seen (≤10 pA around −190 mV). After perfusing with db 2 cAMP (1 mM), a substantial increase in I Ba amplitude between voltages from around −30 to −190 mV was observed (>60 pA around −190 mV). Keeping db 2 cAMP in the bath and adding Gd 3+ (50 µM) resulted in a total block of the current (≤4 pA around −190 mV). This may indicate that GCPs harbor CNGCs that can be activated by cAMP despite the blocking effect by Mg 2+ .

Identification of Candidate Mg 2+ -Dependent Cation Channels in Arabidopsis thaliana
Given the data from the physiological experiments, we undertook a search for candidate Mg 2+ -dependent cation channels in plants (Figure 8).  Figure S2A). Top panel: side view, with the transmembrane region colored in green, the linker region in yellow, and the cytoplasmic cAMP binding domain in grey. The diacidic motifs (residues E437 and E438 in AtCNGC8) are highlighted as red sphere models. Bottom panel: view from the cytoplasm into the channel (90˚ rotation with respect to top panel). (C) Electrostatic surface representation (color-ramped from negatively charged in red to positively charged in blue) of homology models of AtCNGCs containing the diacidic motif (AtCNGC8 and 6) and of AtCNGC2 (model) and human HCN1 (PDB 5u6o) lacking this motif.
Crystallographic and functional analyses of a strong inward-rectifying K + channel (Kir2.2) from chicken [46] showed that the rectification characteristic can be explained by Mg 2+ binding to negatively charged regions in the pore (formed by D173), with possible additional contributions from acidic residues within the cytoplasmic regulatory domains (D256 and E300/E225) ( Figure S2). To explore the possibility that a similar mechanism allows Mg 2+ inward rectification for a subset of candidate plant channels that display the electrophysiological properties described here (i.e., activation by hyperpolarization and cAMP, absence of cation selectivity, and inhibition by lanthanides), we built homology models of the pore-forming residues for AtCNGCs [47,48]. The models were built using structures of the human hyperpolarization-activated channels HCN1, based on a ~22% sequence identity. The obtained models clearly showed that AtCNGCs do not have a Mg 2+ Crystallographic and functional analyses of a strong inward-rectifying K + channel (K ir 2.2) from chicken [46] showed that the rectification characteristic can be explained by Mg 2+ binding to negatively charged regions in the pore (formed by D173), with possible additional contributions from acidic residues within the cytoplasmic regulatory domains (D256 and E300/E225) ( Figure S2). To explore the possibility that a similar mechanism allows Mg 2+ inward rectification for a subset of candidate plant channels that display the electrophysiological properties described here (i.e., activation by hyperpolarization and cAMP, absence of cation selectivity, and inhibition by lanthanides), we built homology models of the pore-forming residues for AtCNGCs [47,48]. The models were built using structures of the human hyperpolarization-activated channels HCN1, based on a~22% sequence identity. The obtained models clearly showed that AtCNGCs do not have a Mg 2+ binding site corresponding to the location of K ir 2.2 D173 inside the pore region ( Figure 8A, left arrow, and Figure S2). However, a tandem glutamic acid motif that could form a possible diacidic Mg 2+ binding site is found in most AtCNGCs located a little downstream of the pore, in the so-called C-linker disc (Figure 8). Akin to K ir 2.2 D173 and E300/E225, this diacidic AtGNCG motif forms a positively charged opening, with distances of 7.3-8.3Å between neighbouring charges and a diameter of~11Å (for K ir 2.2 D173, these values are 7.3-7.4Å and 10.4-11.8Å, respectively (PDB 3jyc)). As in K ir 2.2, these distances between carboxyl groups are too large for direct ion coordination, suggesting that Mg 2+ is bound through bridging water molecules [46]. Interestingly, this diacidic motif is not present in AtGNCG2, which has been shown to be an atypical family member with respect to ion selectivity [49]. The diacidic motif is also absent in HCN1, for which Mg 2+ inward rectification has not been documented ( Figure 8C).

Discussion
In order to record ion currents (for instance I K,in or I K,out ) from plant cells in the "whole-cell" patch configuration, for reasons highlighted in the introduction (also ass Methods section in [50]), it is standard to include ATP and Mg 2+ in either the form of [Mg-ATP] or [MgCl 2 + K 2 -ATP]. In contrast, the composition of the internal solution used to characterize the hyperpolarization-activated Ca 2+ current (HACC or I Ba ) is more variable, especially with regards to ATP and Mg 2+ . For example, one can notice that ATP and Mg 2+ are either both included [9,11,17,51,52] or completely omitted from the internal solution [10,16,[53][54][55]. Here, we summarize the role and consequences of Mg 2+ inclusion, either in the form of Mg-ATP or MgCl 2 . Incidentally, there are no data on cytosolic magnesium concentrations or the distribution of magnesium within the different cellular compartments of guard cells. However, we think it is reasonable to assume that intracellular free [Mg 2+ ] is between 400 and 600 µM, which the total magnesium, including the complexed form, may be as high as 10 mM.
An overview of some of the pharmacological properties related to the HACC (I Ba ) and I K,in are presented in Figure 9. binding site corresponding to the location of Kir2.2 D173 inside the pore region ( Figure 8A, left arrow, and Figure S2). However, a tandem glutamic acid motif that could form a possible diacidic Mg 2+ binding site is found in most AtCNGCs located a little downstream of the pore, in the so-called Clinker disc (Figure 8). Akin to Kir2.2 D173 and E300/E225, this diacidic AtGNCG motif forms a positively charged opening, with distances of 7.3-8.3Å between neighbouring charges and a diameter of ~11Å (for Kir2.2 D173, these values are 7.3-7.4Å and 10.4-11.8Å, respectively (PDB 3jyc)). As in Kir2.2, these distances between carboxyl groups are too large for direct ion coordination, suggesting that Mg 2+ is bound through bridging water molecules [46]. Interestingly, this diacidic motif is not present in AtGNCG2, which has been shown to be an atypical family member with respect to ion selectivity [49]. The diacidic motif is also absent in HCN1, for which Mg 2+ inward rectification has not been documented ( Figure 8C).

Discussion
In order to record ion currents (for instance IK,in or IK,out) from plant cells in the "whole-cell" patch configuration, for reasons highlighted in the introduction (also ass Methods section in [50]), it is standard to include ATP and Mg 2+ in either the form of [Mg-ATP] or [MgCl2 + K2-ATP]. In contrast, the composition of the internal solution used to characterize the hyperpolarization-activated Ca 2+ current (HACC or IBa) is more variable, especially with regards to ATP and Mg 2+ . For example, one can notice that ATP and Mg 2+ are either both included [9,11,17,51,52] or completely omitted from the internal solution [10,16,[53][54][55]. Here, we summarize the role and consequences of Mg 2+ inclusion, either in the form of Mg-ATP or MgCl2. Incidentally, there are no data on cytosolic magnesium concentrations or the distribution of magnesium within the different cellular compartments of guard cells. However, we think it is reasonable to assume that intracellular free [Mg 2+ ] is between 400 and 600 µM, which the total magnesium, including the complexed form, may be as high as 10 mM.
An overview of some of the pharmacological properties related to the HACC (IBa) and IK,in are presented in Figure 9. We report experiments designed to assess the effect of Mg 2+ on IBa current activated by hyperpolarization in GCPs, being aware that other cell types, e.g., root cells, might indeed show different responses. We show that omitting Mg-ATP unmasked the presence of an instantaneously activating, inwardly rectifying conductance. This "newly unveiled" conductance still retains most of the biophysical and pharmacological characteristics that are hallmarks of the classical ICa-type [9,10] and CNGCs, which we referred to here as either IBa or HACCs [13,14]. Similarly to IK,in, this conductance is activated by hyperpolarizing-going voltages, and shows some voltage-dependent rectification (but not as strongly as compared to IK,in, for example; see Figure 4A). Additionally, typical of HACCs, we found that besides Ba 2+ , other divalent cations such as Ca 2+ , Sr 2+ , and Mn 2+ also permeate this channel, but at a much slower rate than Ba 2+ does, hence the smaller currents resolved even though the same amounts of divalent cations in the bath were used. More importantly, this
We report experiments designed to assess the effect of Mg 2+ on I Ba current activated by hyperpolarization in GCPs, being aware that other cell types, e.g., root cells, might indeed show different responses. We show that omitting Mg-ATP unmasked the presence of an instantaneously activating, inwardly rectifying conductance. This "newly unveiled" conductance still retains most of the biophysical and pharmacological characteristics that are hallmarks of the classical I Ca -type [9,10] and CNGCs, which we referred to here as either I Ba or HACCs [13,14]. Similarly to I K,in , this conductance is activated by hyperpolarizing-going voltages, and shows some voltage-dependent rectification (but not as strongly as compared to I K,in , for example; see Figure 4A). Additionally, typical of HACCs, we found that besides Ba 2+ , other divalent cations such as Ca 2+ , Sr 2+ , and Mn 2+ also permeate this channel, but at a much slower rate than Ba 2+ does, hence the smaller currents resolved even though the same amounts of divalent cations in the bath were used. More importantly, this HACC is activated by cAMP, as well as permeable to monovalent cations (see Figure S1), which are definite attributes of CNGCs. HACCs, including CNGCs, are specifically blocked by low concentrations of extracellular Gd 3+ that are far less effective in blocking I K,in (see Figure 4A). Equally important, we found that the unmasked I Ba is also enhanced by ≈1.3to 1.5-fold in response to 5 min treatment with ABA and by up to 2-fold after 10 min. Similarly, it was shown that ABA increases a PM I Ca -type whole-cell current in Arabidopsis guard cells by ≈2to 3-fold within 5 min of treatment [9]. Note that the ABA effect reported here was obtained not only in Mg 2+ -free internal solution, but even more importantly in ATP-free internal solution, thus indicating that ATP is not as crucial for this channel as was suggested by an earlier report [56]. This is in agreement with many other reports showing that ABA can indeed increase cytosolic Ca 2+ levels through activation of PM calcium channels via hyperpolarization in Arabidopsis thaliana guard cells [10,57,58]. Furthermore, ATP was neither required for cGMP-nor cAMP-activated [49,50] Ca 2+ -permeable cation channels in many different plant cell types (mesophyll, guard cells, or pollen tubes) that also show many of the HACC characteristics. This observation is in contrast to the above-mentioned report, where ATP and subsequent protein (de)phosphorylation was described as a prerequisite for an ABA effect on calcium channels (56). This indicates that there may be more than one subtype of calcium channel co-existing in the PM or that additional modes of regulation of these Ca 2+ -permeable channels are in operation, which might require ATP-and protein-kinase-dependent signalling [56,59].
One discrepancy that stands out in our report is that Mg 2+ was shown to permeate HACCs in guard cells [10] and in root hairs [51], whereas in our experiments external Mg 2+ did not appear to permeate this channel ( Figure 3A). This is even more intriguing when considering that cGMP was recently shown to activate an inward rectifying current (which was also lanthanide-sensitive), with Mg 2+ as a charge carrier [16]. Hence, this is another hint that we may be dealing with more than one subtype of calcium channel. In animal cells, Mn 2+ was described as a blocker of calcium channels if Ca 2+ is present in the bath, while in the absence of Ca 2+ the Mn 2+ permeates the channel [60,61]. These data might infer that HACCs, despite sharing many similarities in terms of their biophysical and pharmacological characteristics, might slightly differ from one cell type to another depending on the tissue type or plant species. Therefore, the first key finding is that omitting Mg-ATP from the intracellular medium unmasks a larger Gd 3+ -sensitive, non-selective cation conductance that is also regulated by cAMP and ABA. The mechanism consists of shifting the I-V characteristic to the right, where less negative voltages can mobilize cations (including Ca 2+ ) through the channel. The second key finding is the demonstration that Mg 2+ alone can block this conductance ( Figure 6).
The current activated by voltages lower than −200 mV is small but significant ( Figure 1A). At the present time, we have no evidence to support that this instantaneous, rectifying, Mg 2+ -resistant Ba 2+ current would be carried by a different population type of HACCs (the channel type that was unmasked when Mg 2+ was omitted). If anything, this current could still be carried by the same type of channel, since addition of 20 µM Gd 3+ to the bath was still able to swiftly and efficiently block this current.
Our data describe that in guard cells, Mg 2+ blocks I Ba by shifting the I-V relationship and its activation threshold to more negative voltages ( Figure 1C). This effect is reminiscent of the inhibitory effects by Mg 2+ on many ion channels, which have been interpreted as "charge screening effects" [18]. Indeed, the rectification of the inwardly rectifying K + (K ir ) channels is due to a voltage-dependent blockage by cytosolic Mg 2+ (and polyamines), thereby blocking outward K + -efflux. Upon hyperpolarization, Mg 2+ is ejected from the pore, which appears to result in a time-dependent opening of the channel [62,63]. Likewise, TRPV6 shows Mg 2+ -dependent gating that contributes to its strong inward rectification [20]; it was suggested that Mg 2+ can block the channel by binding to a site within the transmembrane electrical field, where it interacts with other permeant cations [20]. It is also conceivable that other mechanisms could be operating, such as electrostatic interaction between Mg 2+ and some PM lipids, such as phosphatidylinositol 4,5-bisphosphate (or PIP 2 ). It was demonstrated that increasing the amount of membrane PIP 2 results in decreasing the sensitivity of KCNQ channel to inhibition by Mg 2+ [64]. In addition, some Ca 2+ channels were found to require PIP 2 for their normal function [65]. This begs the question of whether guard cell PM Ca 2+ channels are also PIP 2 -sensitive. Our results also raise the question of whether Mg 2+ could be equally important for I K,in gating. Indeed, unlike earlier reports, we found that activation of I K,in was dependent on Mg 2+ being present inside the patch pipet ( Figure 6).
Furthermore, the activation of the above HACCs by hyperpolarization and cAMP, the absence of cation selectivity, and inhibition by lanthanides is consistent with the hypothesis that channels responsible for the observed effects are CNGCs. This is consistent with our structural modelling, which revealed the presence of diacidic motifs in the pore-forming helix of a subset of AtCNGCs. Acidic residues pointing towards the inner side of the pore, have previously been shown to confer Mg 2+ dependence to inward rectifying K + channels in animals [47] ( Figure S2B). Given the position of the diacidic motif in the cytoplasmic side of the pore, it is conceivable that Mg 2+ binding can be affected by changes in pore opening, for example introduced by cAMP binding to the cytoplasmic region of AtCNGCs. Such crosstalk would provide a mechanistic explanation for our observation that cAMP overrides the channel blockage produced by 1mM MgCl 2 (no added ATP) ( Figure S2C), thereby preventing uncontrolled Ca 2+ leakage.

Conclusions
Here, we firstly propose that in Vicia faba guard cells, Mg 2+ can limit or prevent continuous Ca 2+ -leakage, possibly through all HACCs (including CNGCs) at resting membrane potentials, thereby being part of the (intra)cellular calcium signaling processes [66]. Furthermore, CNGCs have also been recognized as having a critical role in Ca 2+ -dependent plant defense signalling and responses [67,68]. Secondly, we propose that the activation of these channels requires mechanism(s) by which Mg 2+ binding is altered. In CNGCs, such a mechanism is conceivably enabled by cAMP binding, thereby assigning Mg 2+ an important role in calcium homeostasis and calcium-dependent downstream processes. Finally, the effect reported here suggests that Mg 2+ has a role in cellular Ca 2+ homeostasis, similarly to in animal cells. Funding: This research has been supported by the King Abdullah University of Science and Technology (KAUST). We are indebted to Enid MacRobbie (Department of Plant Science, University of Cambridge, UK) for allowing us to use some of the data gathered by F.L.-C. whilst in her laboratory (research was supported by BBSRC Grant P05730 to E.M.). We also thank Mark Tester for his invaluable comments.