Oxygen-sensing persistent sodium channels in rat hippocampus

Abstract

1. Persistent sodium channel activity was recorded before and during hypoxia from cell-attached and inside-out patches obtained from cultured hippocampal neurons at a pipette potential (V_p) of +30 mV. Average mean current (I′) of these channels was very low under normoxic conditions and was similar in cell-attached and excised inside-out patches (-0.018 ± 0.010 and -0.025 ± 0.008 pA, respectively, n = 24).

2. Hypoxia increased the activity of persistent sodium channels in 10 cell-attached patches (I’ increased from -0.026 ± 0.016 pA in control to -0.156 ± 0.034 pA during hypoxia, n = 4, P = 0.013). The increased persistent sodium channel activity was most prominent at a V_P between +70 and +30 mV (membrane potential, V_m = -70 to -30 mV) and could be blocked by lidocaine, TTX or R56865 (n = 5). Sodium cyanide (NaCN, 5 mM; 0.5-5 min) increased persistent sodium channel activity in cell-attached patches (n = 3) in a similar manner.

3. Hypoxia also increased sodium channel activity in inside-out patches from hippocampal neurons. Within 2-4 min of exposure to hypoxia, I′ had increased 9-fold to -0.18 ± 0.04 pA (n = 21, P = 0.001). Sodium channel activity increased further with longer exposures to hypoxia.

4. The hypoxia-induced sodium channel activity in inside-out patches could be inhibited by exposure to 10-100 μM lidocaine applied via the bath solution (I′ = -0.03 ± 0.01 pA, n = 8) or by perfusion of the pipette tip with 1 μM TTX (I′ = -0.01 ± 0.01 pA, n = 3).

5. The reducing agent dithiothreitol (DTT, 2-5 mM) rapidly abolished the increase in sodium channel activity caused by hypoxia in excised patches (I’ = -0.01 ± 0.01 pA, n = 4). Similarly, reduced glutathione (GSH, 5-20 mM) also reversed the hypoxia-induced increase in sodium channel activity (I’ = -0.02 ± 0.02 pA, n = 5).

6. These results suggest that persistent sodium channels in neurons can sense O₂ levels in excised patches of plasma membrane. Hypoxia triggers an increase in sodium channel activity. The redox reaction involved in increasing the sodium channel activity probably occurs in an auxiliary regulatory protein, co-localized in the plasma membrane.

During a ‘brain attack’ or stroke, when arteries in the brain are blocked or leaking, cells are deprived of oxygen and the ischaemia/hypoxia eventually leads to cell death. In the initial stages of hypoxia, within minutes, cells undergo alterations in membrane potential and detrimental changes in their intracellular environment such as a rise in intracellular Ca²⁺ concentration, attenuated ATP levels and changes in pH (for review see Lipton, 1999). Sodium channel blockers such as TTX and lidocaine (lignocaine) and extracellular solutions containing a low Na⁺ concentration protect neurons from ischaemic damage, suggesting that voltage-operated Na⁺ channel activity is an early and important step in oxygen sensing and cell damage (Boening et al. 1989; Yamasaki et al. 1991; Stys et al. 1992; Friedman & Haddad, 1993, 1994a,b; Fried et al. 1995; Fung & Haddad, 1997; Toner & Stamford, 1997; Fung et al. 1999).

The increase in the persistent Na⁺ current (I_Na,P) that we have previously reported in intact cardiac (Ju et al. 1996) and hippocampal cells (Hammarstrom & Gage, 1998) is a likely trigger for abnormal electrical activity and accumulation of intracellular Na⁺ during hypoxia, and may eventually lead to damaging increases in intracellular Ca²⁺ concentration and cell damage (Astrup et al. 1981; Boening et al. 1989; Lucas et al. 1989; Haigney et al. 1994; Weber & Taylor, 1994; Fried et al. 1995; Fung & Haddad, 1997). In both heart muscle and neurons it is thought that the increases in intracellular Na⁺ and Ca²⁺ concentration are coupled by the Na⁺-Ca²⁺ exchanger because both are reduced by Na⁺ channel blockers or by removing extracellular Na⁺ (Stys et al. 1992; Friedman & Haddad, 1993, 1994a; Haigney et al. 1994). Hence, our initial finding provided a pathway or source for this sustained Na⁺ influx during hypoxia. To determine whether this effect of hypoxia on whole-cell neuronal I_Na,P (Hammarström & Gage, 1998) depended on a cytosolic second messenger system or was a direct effect on the Na⁺ channel protein itself (or a closely associated regulatory protein), we examined the effect of lowering O₂ tension on channel activity in excised inside-out patches from cultured hippocampal neurons. Preliminary reports of some of these results have appeared elsewhere (Hammarström & Gage, 1999b).

METHODS

Cell preparation

Cultures of hippocampal neurons were prepared as described previously (Premkumar et al. 1990). Briefly, newborn Wistar rats were decapitated (CO₂ anaesthesia); the hippocampus was then dissected and the tissue triturated to dissociate cells. The dissociated cells were grown on poly-l-lysine-coated glass coverslips and used in experiments after 5-12 days. The culture medium was Minimum Essential Medium to which was added fetal calf serum (10%), glucose (2%), penicillin-streptomycin (1%) and fungizone (1%) (Flow Laboratories). These methods were approved by the Animal Experimentation Ethics Committee at the Australian National University, Canberra, Australia (J.BM.51.00).

Solutions

The bath solution for cell-attached and inside-out patch recordings contained (mm): 150 potassium aspartate, 10 EGTA, 2 MgCl₂, 2 CaCl₂ and 10 Tes, pH adjusted with NaOH to 7.4. The pipette solution contained (mm): 135 NaCl (or 65 NaCl + 65 choline chloride for some cell-attached recordings where indicated in the text), 5 KCl, 1 MgCl₂, 1 CaCl₂, 5 CoCl₂, 5 CsCl and 10 Tes, adjusted to a pH of 7.4 with NaOH. These solutions are similar to those used previously (Ju et al. 1996; Hammarström & Gage, 1998).

Hypoxic solutions and those containing drugs were applied through a fine (200 μm i.d.) tube carefully positioned to cause a rapid change of solution close to the patch (Ju et al. 1996; Hammarström & Gage, 1998). Hypoxia was induced by rapid perfusion with a solution bubbled with 100% N₂ or 95% N₂-5% O₂ (Ju et al. 1996; Hammarström & Gage, 1998). When bubbling the perfusion solution with N₂ in this manner, the O₂ level in the bath near the outlet of the perfusion tube was reduced to ≈45 mmHg within ≈3 min (Fig. 1). The oxygen tension at the tip of the perfusion tube, where the patch would have been, was most probably lower, and likely to have reached equilibrium more quickly, than detected by the larger O₂ electrode (Diamond Electro-Tech 760 Microelectrode, Diamond General Development Corp., Ann Arbour, MI, USA and Polarographic Amplifier Model 1900, A-M Systems, Inc., Everett, WA, USA). The perfusion solution would not mix with the bath solution to the same extent around the very small diameter of the patch pipette as it would around the much larger O₂ probe. The diameter of the O₂ electrode (0.7 mm) was substantially larger than that of the patch.

Figure 1. Effect of bubbling the bath solution with 100% N₂ on O₂ tension.

The graph shows the level of O₂ recorded at various times after commencing the bubbling of the perfusing solution with 100% N₂. O₂ tension reached levels of less than 30% of normal air O₂ within 3 min (see Methods). This time corresponds well with the initial increase in persistent Na⁺ channel activity.

Appropriate concentrations of tetrodotoxin (TTX; Boehringer Mannheim), R56865 (gift from Janssen Research Foundation, Beerse, Belgium), lidocaine, dithiothreitol (DTT), reduced glutathione (GSH) (all Sigma) and NaCN (The British Drug Houses, Poole, UK) were applied in the same manner as the hypoxic solution.

Electrophysiological recording and data analysis

Tight giga-ohm seal patch clamp techniques were used to record currents in cell-attached and inside-out patches from rat hippocampal neurons. Pipettes (GC150F-15, Clarke Electromedical Instruments, UK) were pulled on a Flaming/Brown (P-87) puller and had resistances of 15-20 MΩ when filled with pipette solution. The pipettes were coated with Sylgard (Dow Corning) as close to the pipette tip as possible and tips were fire polished. Cells that had a swollen or grainy appearance were not used.

Potential was controlled by voltage commands from an IBM-compatible PC and digital-to-analog interface and recorded with an Axopatch-1D amplifier. Currents were analysed using computer techniques and ‘in-house’ software (Ju et al. 1992, 1996; Saint et al. 1992; Hammarstrom & Gage, 1998). Currents were generated by depolarizing steps from a holding potential of V_P =+70 mV (pre-pulsed to V_P =+120 mV for 200 ms). With each step, the depolarizing pulse was incremented by 5 or 10 mV, to a final V_P of -50mV. ‘Early’ Na⁺ channel openings were quantified as those occurring within the first 50 ms of the triggering of the depolarizing pulse. Na⁺ channel openings occurring later during the 400 ms depolarizing pulse were considered to be due to persistent Na⁺ channels (Ju et al. 1996). Potentials were corrected for liquid junction potential using the software program JPCalc (copyright Professor P. H. Barry, University of New South Wales, Sydney, Australia) which utilizes the generalized Henderson equation for its calculations. This program gave a liquid junction potential of -16.4 mV for the excised patch solutions and -18.8 mV for the solutions used for cell-attached patches. This value was used to obtain the reversal potentials presented in Results and the potentials in the current-voltage relationships shown in Fig. 2 and 7. Mean current was calculated from 0.5-4 min recordings digitized at 10 kHz. All values are expressed as means ± 1 s.e.m. and the number of cells (n) in each group is given. Statistical analysis was performed using Student’s two-tailed unpaired t test.

Figure 2. Effect of hypoxia on Na⁺ channel activity in a cell-attached hippocampal patch.

Representative traces recorded from a cell-attached hippocampal patch at various potentials (-V_P) stepped from a holding potential (V_h) of +70 mV (pre-pulse to V_P =+120 mV) in control (A), hypoxia (B) and after addition of 100 μm lidocaine to the hypoxic perfusing solution (C). All traces start 100 ms after the beginning of the 400 ms test pulse, hence they only show persistent Na⁺ channel activity. D, the relationship between single channel current amplitude (I) and potential. The current amplitude is the mean of several measurements at each potential (n = 3-15) obtained from the same patch shown in A–C during normoxia (○) and hypoxia (•), and ±1 s.e.m. is shown where greater than the size of the symbol. The slope conductance was 9.7 pS and the extrapolated reversal potential (V_P = -12 mV) would suggest that these channels are also permeable to K⁺, as reported previously (Hammarström & Gage, 1999a). As persistent Na⁺ channel activity was very rare in control conditions, no reliable current-voltage realtionship could be established during normoxia. When Na⁺ channel activity was observed in control solution it did not appear to change in amplitude or reversal potential with hypoxia (see ○).

Figure 7. Effect of 5 mm NaCN on Na⁺ channel activity in a cell-attached hippocampal patch.

Representative Na⁺ channel traces recorded at various potentials (-V_P, all potentials corrected for a liquid junction potential of 18.8 mV) from a V_h of +70 mV (pre-pulse to +120 mV) in control (A), in the presence of 5 mm NaCN (B) and after addition of lidocaine to the NaCN-containing perfusing solution (C). All traces start 100 ms after the beginning of the 400 ms test pulse, and hence only show the persistent Na⁺ channel activity. D, current-voltage relationship established from the same cell-attached patch (n = 3-23 measurements at each potential) as in A–C, showing ±1 s.e.m. where greater than the size of the symbol. No channel activity was recorded around 0 ± 10 mV either in control (○) or in the presence of NaCN (•). The extrapolated reversal potential for this patch was +20 mV (-V_P), and the slope conductance was 10.1 pS. This reversal potential is less negative than the calculated reversal potential of +42 mV. This would suggest that these channels, like those activated by hypoxia, are also permeable to K⁺. Very few channel openings were recorded across the full voltage range in control, when no NaCN was present. The current amplitudes recorded during control, normoxic conditions (○) suggest that there was no change in current amplitude caused by NaCN.

RESULTS

Persistent Na⁺ channels in hippocampal neurons during normoxia

Cell-attached patches

Na⁺ channel activity was recorded in cell-attached patches held at a pipette potential (V_p) of +70 mV, and stepped to aV_P of +30 mV (preceded by a pre-pulse to V_P =+120 mV). As the bath solution contained 140 mm K⁺, it was assumed that the potential across the membrane (V_m) was approximately -V_P. Na⁺ channel activity was normally recorded at a V_p of +30 mV, since the classical, transient Na⁺ current (I_Na,T) is rapidly inactivated when patches are held at these potentials, whereas the persistent Na⁺ current (I_Na,P) is not (Saint et al. 1992). The activity recorded during normoxia in this manner was similar to that recorded previously in cardiac muscle (Ju et al. 1996). Na⁺ channel openings were rare, sometimes even absent under these conditions for long recordings of > 3 min. The open probability of persistent Na⁺ channels was very low between V_P =+70 and -40 mV (see Fig. 2A) and channel openings were hardly ever seen at voltages between V_P =+10 and -10 mV, as has been reported previously (Magistretti & Alonso, 1999). This channel behaviour made obtaining full current-voltage relationships very difficult without excessively long periods of recording (> 15 min). This low Na⁺ channel activity in cell-attached patches is also revealed in the very low average mean current (I′) of -0.018 ± 0.010 pA (n = 24) measured over a continuous 2-4 min period in cell-attached patches held at V_P =+30 mV.

Inside-out patches

The currents recorded during normoxia when holding the excised patch continuously at V_P =+30 mV resembled in amplitude, reversal potential and frequency those observed in the cell-attached patches described above. Hence, I′ in inside-out patches was also very low during normoxia (-0.025 ± 0.008 pA, n = 24, continuous 2-4 min recordings) and not significantly different from that recorded from the cell-attached patches (P = 0.54).

Effect of hypoxia on hippocampal neurons

Cell-attached patches

When cell-attached patches of hippocampal neurons were exposed to hypoxic perfusion solution, Na⁺ channel activity increased significantly from the very low or often absent activity in control (Fig. 2A) to frequent persistent Na⁺ channel activity, usually observed within ≈3-5 min of hypoxia (Fig. 2B). A similar result was obtained in nine other cell-attached patches. The increase in Na⁺ channel activity was most prominent at voltages between V_P =+70 and +30 mV, with only very rare openings at potentials between +10 and -10 mV. Channel activity resumed at more negative values of V_P (n = 5; see channel traces from one patch shown in Fig. 2B).

Both early (within the first 0-50 ms of the depolarizing pulse, presumed to be classical transient Na⁺ channels), and later Na⁺ channel openings (persistent Na⁺ channels) were observed during the 400 ms depolarizing pulses. An increased occurrence of late openings (persistent Na⁺ channels) was more prominent during hypoxia (see Fig. 2B). I′ increased 6-fold from -0.026 ± 0.016 pA in control to -0.156 ± 0.034 pA (P = 0.013) during hypoxia in four patches from which longer, continuous recordings of at least 2-4 min were made at V_P =+30 mV. It was difficult to quantify any changes in the transient currents, which were not closely examined.

The current-voltage relationship of channels activated by hypoxia can be seen in Fig. 2D. The current-voltage relationship for this patch gave a best fit for the Na⁺ equilibrium potential (E_Na) during hypoxia as V_P = -12 mV and a slope conductance of 9.7 pS. Figure 2D also illustrates that the current amplitude during hypoxia was not different from that during normoxia at the potentials shown, suggesting that there is no obvious change in Na⁺ channel conductance with hypoxia. The extrapolated reversal potential in Fig. 2D is more negative than the calculated reversal potential for Na⁺ in these solutions (V_P = -42 mV; pipette solution containing 65 mm Na⁺). This suggests that the channels are also permeable to K⁺, as reported previously (Hammarström & Gage, 1999a). That these channels were Na⁺ channels was confirmed by their inhibition by Na⁺ channel blockers such as lidocaine, and one example of this is illustrated in Fig. 2C. Similar block of the channels by 100 μm lidocaine (n = 1), 20-40 μm R56865 (n = 2) or 1 μm TTX (pipette perfusion, n = 1) was also recorded in four other cell-attached patches.

The effect of hypoxia on Na⁺ channel activity appeared to be irreversible, although the possibility that some cells had been damaged or died after the hypoxia and several minutes of reperfusion cannot be excluded. In three cell-attached patches the hypoxic effect was not reversed even after 4-12 min of perfusion with normoxic solution (I′ = -0.469 ± 0.313 pA, n = 3).

The rise in persistent Na⁺ channel activity appeared to follow the progressive decrease in the level of O₂. Levels of O₂ near the patch were reduced to about 45 mmHg after ≈3 min of hypoxia (see Fig. 1) and effects of hypoxia on Na⁺ channel activity in cell-attached patches were usually not observed earlier than this.

Inside-out patches

In order to determine whether this increase in persistent Na⁺ channel activity triggered by hypoxia involved cytosolic second messenger(s), we exposed excised inside-out patches to hypoxia. The effects of hypoxia on persistent Na⁺ channel activity in excised patches is illustrated by channel recordings from two patches shown in Fig. 3.

Figure 3. Effect of hypoxia on inside-out patches excised from hippocampal neurons.

Representative traces from a hippocampal patch held at a V_P of +30 mV in control (A) and after 30 s of hypoxia (non-continuous traces in B). Dotted lines show the closed channel current level. C–E, all-points histograms calculated from full 0.5-3 min recordings sampled at 10 kHz obtained from another hippocampal patch held at a V_P of +30 mV (P, probability). C, control; D, 0-30 s hypoxia (inset shows open peak more clearly); E, 1-3 min of hypoxia. F, increase in mean current during hypoxia measured in control patches (n = 21, ▪), and after 0-4 min (n = 21) and 4-8 min (n = 8) of hypoxia (). Histograms show mean data and the vertical bars ±1 s.e.m.

Before exposure to hypoxia there was little channel activity (Fig. 3A). After 30 s of hypoxia in this patch, channel activity had increased markedly (Fig. 3B) and continued to increase during the following minute. In Fig. 3C–E, all-points, current amplitude histograms obtained from another excised patch show a similar time-dependent increase. The histogram in Fig. 3C shows only a baseline peak reflecting the paucity of Na⁺ channel activity under control, normoxic conditions. The all-points histogram in Fig. 3D obtained from the current recorded during the first 30 s of hypoxia shows a small amount of current activity that increased further after 1-3 min of hypoxia (Fig. 3E). Similar increases in mean current caused by hypoxia were recorded in 21 hippocampal patches. Before exposure to hypoxia, I′ was -0.02 ± 0.01 pA, whereas after 0.5-4 min of hypoxia, I′ was -0.18 ± 0.04 pA (n = 21, P = 0.001). There was an even greater increase (24-fold) in eight of these patches which were exposed to hypoxia for longer periods of 4-8 min (see Fig. 3F): I′ was -0.36 ± 0.09 pA (n = 8, P < 0.0001). In some patches, an increase in channel activity was observed as soon as 20-30 s after the commencement of hypoxia, but in most other patches an increase in activity was not seen until at least 2-4 min of hypoxia. The amount of channel activity progressively increased with the duration of hypoxia as illustrated by a significant increase in I′ shown in the histogram in Fig. 3F. The current amplitude and reversal potential for these channels (V_P = -16 ± 1 mV, n = 13) were not different from those recorded in cell-attached patches. In addition, the amplitude of the single Na⁺ channel current recorded during hypoxia was not different from that during normoxia (P = 0.398, n = 4).

Similar to the Na⁺ channel activity triggered by hypoxia in cell-attached patches, that recorded in excised patches (-0.361 ± 0.216 pA, n = 3, after about 10 min of perfusion with normoxic solution) also appeared to be irreversible.

Effect of Na⁺ channel blockers on hypoxia-initiated channel activity in excised patches

The single channel currents that appeared during hypoxia in inside-out patches were similar to the persistent Na⁺ channel currents we recorded in cell-attached patches (see above). However, in order to confirm that they were Na⁺ channels, we examined the ability of Na⁺ channel blockers to inhibit the activity. Examples are illustrated in Fig. 4 and 5.

Figure 4. Effect of TTX on I_Na,P induced by hypoxia.

A–E, effect of pipette perfusion with TTX in one inside-out hippocampal patch. Representative traces recorded at a V_P of +30 mV during control (A), after 5 min of hypoxia (B), after 10 min of hypoxia (C), after 1 μm TTX has reached the tip of the pipette and the membrane patch abolishing activity (D), and 12 min later (continuous hypoxia) when it can be assumed that TTX has diffused away and channel activity has resumed (E). The mean current (I′) from a full 0.5-4 min recording is shown below each set of traces. Dashed lines indicate closed current levels. F, mean data from excised patches exposed to extracellular TTX in the continued presence of hypoxia. Histograms show I′ (pA) and the vertical bars ±1 s.e.m. during control (), hypoxia (□) and exposure to TTX (▪).

Figure 5. Effect of lidocaine on I_Na,P induced by hypoxia.

A–F, effect of lidocaine in two inside-out hippocampal patches. Three representative traces (V_P =+30 mV) during control (A), after 5-6 min of hypoxia (B) and after 2.5 min exposure to 100 μm lidocaine added to the hypoxic perfusing solution (C). The mean current (I′) from a full 0.5-4 min recording is shown below each set of traces. Dashed lines indicate closed current levels. D–F, all-points histograms obtained from 4 min recordings sampled at 10 kHz from another inside-out patch held at a V_P of +30 mV during control (D), 4 min of hypoxia (E) and after 2 min the the presence of 100 μm lidocaine (F). G, cumulative data for the effect of lidocaine on hypoxia-elicited activity. Histograms show I′ (pA) and the vertical bars ±1 s.e.m. during control (), hypoxia (□) and exposure to lidocaine (▪, n = 8).

TTX

In three inside-out hippocampal patches, 1 μm TTX was applied via the pipette solution after an increase in channel activity had been established during hypoxia (Fig. 4B and C). Within minutes of injection of 1 μm TTX into the pipette tip, channel activity was completely abolished (Fig. 4D) but could then be seen to return within the next 10 min as the TTX diffused away from the pipette tip (Fig. 4E). Addition of TTX abolished the hypoxia-elicited activity in another two patches. Before exposure to TTX, I′ was -0.42 ± 0.13 pA and after exposure to TTX I′ was -0.01 ± 0.01 pA (n = 3, P = 0.03, Fig. 4F).

Lidocaine

The effect of lidocaine is illustrated in Fig. 5. The three traces in Fig. 5A were taken from one hippocampal patch before exposure to hypoxia. The low channel activity observed under normoxic conditions can also be seen from an all-points histogram obtained from another excised hippocampal patch (4 min recording, Fig. 5D). After 4-6 min of hypoxia (Fig. 5B and E), there was a clear increase in channel activity in both patches. This activity was blocked by adding 100 μm lidocaine to the hypoxic solution (Fig. 5C and F). Lidocaine, tested at concentrations of 10-100 μm, reduced the frequency of channel openings and the elevated mean current triggered by hypoxia to -0.03 ± 0.01 pA (n = 8, Fig. 5G).

The block by TTX and lidocaine confirmed that the channels were Na⁺ channels.

Effects of redox reagents on hypoxic channel activity

Lowering the oxygen tension could have a reducing effect on the Na⁺ channel protein. A reducing agent, dithiothreitol (DTT), has been reported to reverse thiol oxidation (Xu et al. 1998) and to mimic the effects of hypoxia on ion channel activity (Jiang & Haddad, 1994; Park et al. 1995). We have not, however, observed an increase in neuronal I_Na,P in the presence of reducing agents such as DTT under normoxic conditions (authors’ unpublished observations). In contrast, we have recently found that DTT reversed the increase in persistent Na⁺ channel activity caused by NO donors and oxidizing agents in inside-out patches (Hammarström & Gage, 1999a). We examined, therefore, whether reducing agents such as DTT and the reduced form of glutathione (GSH) could reverse the hypoxia-initiated activity in hippocampal neurons.

Perfusion of a patch with 2-5 mm DTT reversed (usually within 1-2 min) the increase in channel activity caused by hypoxia (Fig. 6). Mean results from four patches are shown in Fig. 6D. Similarly, GSH (5-20 mm) also reversed the increase in Na⁺ channel activity caused by hypoxia (n = 5, Fig. 6E). GSH did not appear to be as potent as DTT in inhibiting hypoxia-elicited activity, as higher concentrations were needed to obtain the same effect.

Figure 6. Reducing agents inhibit the effect of hypoxia.

A–C, effect of DTT in one inside-out hippocampal patch. Three representative traces (V_P =+30 mV) are shown during control (A), after ≈12 min of hypoxia (B) and within 2 min of exposure to 5 mm DTT in the perfusing solution (C). The mean current (I’) from a full 0.5-4 min recording is shown below each set of traces. Dashed lines indicate the closed current level. All-points histograms obtained from 4 min recordings sampled at 10 kHz from the same inside-out patch held at a V_P of +30 mV during control, after 23 min of hypoxia and within 2 min of adding 5 mm DTT to the perfusing solution accompany each set of traces. D and E, mean effect of reducing agents on hypoxia-elicited activity. Histograms show I′ and the vertical bars ±1 s.e.m. during control (), hypoxia (hypo, □) and exposure to drugs (D, DTT, n = 4; E, GSH, n = 5; ▪).

Effect of NaCN on inside-out patches

Cell-attached patches

It has previously been shown that sodium cyanide (NaCN) increases the whole-cell persistent Na⁺ current amplitude in intact ventricular myocytes (Ju et al. 1996) and hippocampal neurons (Hammarström & Gage, 1998). The aim of the present study was to investigate the effect of NaCN on persistent Na⁺ channels in cell-attached patches and compare it with the effect of hypoxia.

Cells were perfused with 5-10 mm NaCN. Within 4-5 min of exposure to NaCN, Na⁺ channel activity in cell-attached patches on perfused cells (n = 3) had increased significantly from a very low, or even absent activity in control (see results from one patch in Fig. 7A) to very frequent persistent Na⁺ channel activity in the presence of NaCN (Fig. 7B). Similar to the effect of hypoxia, the increase in Na⁺ channel activity triggered by NaCN appeared most prominent at potentials between V_P =+70 and +30 mV (Fig. 7C). The increased occurrence of late openings (persistent Na⁺ channels) was also the most frequent and prominent event triggered by NaCN (see Fig. 7B).

An example of the current-voltage relationship in the presence of NaCN, when Na⁺ channel activity had been increased, can be seen in Fig. 7D. The extrapolated reversal potential (V_P = -20 mV) is more negative than the calculated reversal potential for Na⁺ in these solutions (V_P = -42 mV; pipette solution with 65 mm Na⁺). The reversal potential of the channels activated by NaCN would suggest that these channels, like the Na⁺ channels activated by hypoxia, are also permeable to K⁺ (Hammarström & Gage, 1999a). The slope conductance was 10.1 pS, similar to that for the Na⁺ channels activated by hypoxia. In addition, we confirmed their identity as Na⁺ channels by applying the Na⁺ channel blocker lidocaine (Fig. 7C).

DISCUSSION

We have previously reported that a persistent, inactivation-resistant, TTX-sensitive Na⁺ current (I_Na,P) increases during hypoxia in intact ventricular myocytes and hippocampal neurons (Ju et al. 1996; Hammarström & Gage, 1998). This event may be responsible for the depolarization and prolonged Na⁺ influx that has been described during hypoxia (Lipton, 1999). Our previous observations were made on whole cells and cell-attached patches, where the surface membrane was exposed to intracellular metabolism and messengers. In the present study we show that persistent Na⁺ channels can sense hypoxia in patches excised from hippocampal cells, i.e. in the absence of any cytosolic products of metabolism or second messengers. This result suggests that the Na⁺ channels themselves, or a co-localized regulatory protein, can sense the O₂ levels directly and trigger an increase in Na⁺ channel activity. The rise in persistent Na⁺ channel activity appeared to follow a significant decrease in the level of O₂ and channels were most probably responding to a P_O2 of ≈45 mmHg, or less (see Methods). Hence, Na⁺ channels may not be as responsive to small changes in O₂ levels as some K⁺ channels, which appear to respond to gradual changes of P_O2 from 120 to 70 mmHg (Lopez-Lopez et al. 1989; Ganfornina & Lopez-Barneo, 1991; Zhu et al. 1996). In spite of this, our results may still allow persistent Na⁺ channels an important role in the damaging ischaemic events involving Ca²⁺ influx, which often eventually lead to cell death. A large increase in [Ca²⁺]_i, enough to trigger transmitter release, occurs when P_O2 falls below 40-60 mmHg (Lopez-Barneo, 1996).

As far as we are aware, our study is the first demonstration of direct O₂ sensing by TTX-sensitive Na⁺ channels. Chronic hypoxia (1-2 weeks) has been shown to augment voltage-dependent Na⁺ channel currents of rat carotid bodies, but this activity appeared to be triggered by elevated levels of the intracellular messenger cAMP (Stea et al. 1995). Interestingly, in this study we have seen a response to hypoxia in hippocampal cells cultured from newborn Wistar rats. Newborn mammals are more resistant to anoxia than adult mammals (Duffy et al. 1975; Ferriero et al. 1988; Cherubini et al. 1989; Haddad & Donnelly, 1990; Jiang et al. 1992) and neonatal CA1 neurons can withstand anoxia better than adult CA1 neurons (Ferriero et al. 1988; Cherubini et al. 1989; Choi & Rothman, 1990; Haddad & Donnelly, 1990). It will be interesting to compare our present results with those from cells from older animals. Previous results show that the whole-cell persistent Na⁺ current is regulated by O₂ levels in intact hippocampal cells from ≈20-day-old rats (Hammarström & Gage, 1998), but the underlying mechanism or trigger for this activity has not been investigated in the older animals. At present we do not know whether persistent Na⁺ channels in adult CA1 neurons can also be regulated by O₂ levels in excised patches.

The open probability of these persistent Na⁺ channels is very low during normoxia and full current-voltage relationships were very difficult to obtain, as found previously (Hammarström & Gage, 1999a). The persistent Na⁺ channel activity, when present, was usually most prominent between a V_P of +70 and +30 mV as also found in squid axon (Gilly & Armstrong, 1984). During hypoxia, persistent Na⁺ channel activity became very frequent and prominent at these potentials. However, even when Na⁺ channel activity was increased by hypoxia, openings were still rare around 0 ± 10 mV. This may suggest that a major effect of increased persistent Na⁺ channel activity during hypoxia is to depolarize the resting membrane potential, setting it closer to the threshold for action potential firing and in doing so probably triggering repetitive firing. A sustained current such as the persistent Na⁺ current would also allow a prolonged and significant influx of Na⁺ into the cell (Taylor, 1993) during hypoxia. This would put a lot of strain on the cell as energy stores and the supply of ATP are limited during hypoxia and the Na⁺-K⁺-ATPase is the major pathway for getting rid of excess Na⁺. The low persistent Na⁺ channel activity in normoxia may suggest that the hippocampal neuron is healthy and having ‘normal’ electrical firing activity.

NaCN caused a similar increase in persistent Na⁺ channel activity to that of hypoxia in cell-attached patches from hippocampal neurons. We initially thought that the effect of NaCN in the whole cell (Ju et al. 1996; Hammarström & Gage, 1998) was due to inhibition of re-oxidation of cytochrome a₃ of the mitochondrial electron transport chain. However, 5 mm NaCN was recently shown to cause an increase in I_Na,P in inside-out patches from hippocampal neurons (Hammarström & Gage, 1999a), suggesting that NaCN, like hypoxia, has a more direct effect on Na⁺ channels. This earlier study also showed that the increase in channel activity caused by NaCN could be reversed by 2-5 mm DTT, like the activity triggered by hypoxia. Hence, hypoxia and NaCN have similar effects even in excised patches. Therefore, the effect of NaCN cannot only be explained by an inhibition of electron transfer through the mitochondrial electron transport chain. Our result may be in part due to cyanolation as described by Arden and colleges (Arden et al. 1998), or some non-specific effect of cyanide, but may also suggest that a metal-containing plasma membrane moiety is involved in regulating the oxygen sensing by Na⁺ channels.

Our results suggest that hypoxia, quite unexpectedly, acts as an ‘oxidizing agent’, probably indirectly via modulation of or a conformational change in a Na⁺ channel regulatory protein, thereby triggering persistent Na⁺ channel activity. It is perhaps surprising that a lowering of O₂ levels translates into an oxidizing reaction. However, other experiments also support this idea. Firstly, DTT applied alone does not mimic the effect of hypoxia on whole-cell Na⁺ currents (authors’ unpublished observations), as has been reported by others (Jiang & Haddad, 1994; Park et al. 1995). Secondly, Na⁺ channel activity is not altered merely by excising the patch for > 4 min (see Results) and is therefore not due to the removal of the cytosol and cytosolic endogenous redox agents.

The increase in persistent Na⁺ channel activity during hypoxia appears to result from disulphide bond formation involving the Na⁺ channel protein or a closely associated regulatory protein, since it can be inhibited by reducing agents such as DTT and GSH. The activity is irreversible in their absence, both in cell-attached and in excised patches. This strengthens the suggestion that disulphide bond formation occurs during hypoxia. In line with this argument, oxidizing agents such as NO and 5’,5’-dithio-bis-2-nitrobenzoic acid (DTNB) have been shown to increase persistent Na⁺ channel activity in excised patches (Hammarström & Gage, 1999a). To determine whether hypoxia involves the release of an oxidizing mediator from the plasma membrane, oxidizing agents such as H₂O₂ could be added to the patch before, or during hypoxia. This kind of experiment would determine whether H₂O₂ and hypoxia act at the same site (no additive effects) or may suggest different sites of action (additive effects). However, the progressive increase in Na⁺ channel activity over time (see Fig. 3) would make results from such experiments very hard to interpret.

Several other kinds of ion channels such as K⁺ channels and Ca²⁺ channels have been shown to be modulated by hypoxia (Lopez-Barneo et al. 1988, 1994; Youngson et al. 1993; Fearon et al. 1999) but whether this is a direct effect on the ion channel itself or involves an auxiliary regulatory protein is uncertain. Several models have been proposed for how ion channels may sense O₂ levels (Lopez-Barneo, 1994). These include (1) ‘direct’ sensing, or the formation of co-ordination complexes between O₂ and amino acids at metal-containing sites situated in or near the ion channel itself. This binding may then alter the conformational state and gating of the channel. (2) There may be an involvement of other channel subunits in O₂ sensing, e.g. the Na⁺ channel β1 or β2 subunits, or (3) the existence of a separate metal-containing (haem or non-haem) subunit co-localized with the channel in the plasma membrane, which may or may not be able to release a mediator (such as free radicals) for the effect to occur.

It has been shown that potassium (Kv2.1) channels in patches from pulmonary vascular myocytes respond uniformly to hypoxia. In contrast, only some COS cells express Kv2.1 channels that are able to respond to changes in oxygen tension, although in all other respects the biophysical characteristics of the channels appear to be the same (Patel et al. 1997). In contrast, the α_1C subunit of the L-type Ca²⁺ channel expressed in HEK 293 cells always responded to hypoxia (Fearon et al. 1999). These results would suggest that, at least for these types of ion channels, an auxiliary protein is needed for the O₂ sensing. It has been reported that plasma membranes contain flavin, b-type cytochromes, non-haem iron, coenzyme Q, α-tocopherol, thiol groups and maybe copper (Crane et al. 1991). The haem-linked, flavoprotein NAD(P)H oxidase has been suggested as one possible candidate for a plasma membrane ‘O₂ sensor’ by several laboratories investigating rat carotid body and pulmonary airway chemoreceptor regulation (Cross et al. 1990; Youngson et al. 1993, 1997; Fu et al. 2000). NAD(P)H oxidase catalyses the production of H₂O₂ and may play a central role in O₂ sensing in rat carotid body (Cross et al. 1990). Hence, examining the effect(s) of H₂O₂ on neuronal Na⁺ channels would also be interesting from the point of view that it is thought to be released from the plasma membrane NAD(P)H oxidase (Cross et al. 1990). Recent observations have, however, suggested that NAD(P)H oxidase does not play a role in O₂ sensing in rat and rabbit carotid body chemoreceptor cells (Obeso et al. 1999) or in pulmonary artery smooth muscle cells (Archer et al. 1999). In addition, mice lacking the gp91 phox subunit of NAD(P)H oxidase have preserved O₂ sensing (Archer et al. 1999), although the conserved O₂ sensing in these mice may also be due to some gp91phox homologues still being present (Griendling et al. 2000). The use of diphenyleneiodonium (DPI) to identify NAD(P)H oxidase as an ion channel regulator may also be inappropriate since DPI has been shown to non-selectively inhibit Ca²⁺ and K⁺ channels in pulmonary smooth muscle and carotid body, without affecting H₂O₂ formation (Weir et al. 1994; Wyatt et al. 1994).

The effect of hypoxia on Ca²⁺-dependent K⁺ channels in central neurons could be mimicked by iron chelators such as 1,10-phenanthroline (PTL) but not by DPI (Jiang & Haddad, 1994). This led to the suggestion that there was an associated iron-containing moiety, but not NAD(P)H oxidase, needed for O₂ sensing by these channels. In view of the similar effects of cyanide, NO and low O₂ tension that we report, an auxiliary metal-containing protein may well be co-localized with the Na⁺ channel in the plasma membrane. All these agents have the ability to bind to a metal-containing protein and, in doing so, perhaps modulate the electron transfer through a plasma membrane electron transport chain. This may ultimately affect the redox state of specific amino acid residues in the Na⁺ channel protein. Although our results allow us only to speculate on possible regulatory mechanisms, they do provide an important and missing link in the cascade of events triggered by ischaemia in central neurons.

This increase in persistent Na⁺ current may be a general response of excitable cells to hypoxia since we have recorded it in both cardiac myocytes (Ju et al. 1996) and hippocampal neurons (Hammarström & Gage, 1998). It may be an early and very critical link between hypoxia and cell injury in the brain. This ‘O₂ sensor’ may provide a target for preventative agents aimed at cell protection during stroke, since they would limit the damage triggered by increases in the levels of intracellular Na⁺ and abnormal electrical activity. Such an agent would improve the outcome of an ischaemic attack.