Genetic Inactivation of Kcnj16 Identifies Kir5.1 as an Important Determinant of Neuronal PCO2/pH Sensitivity*

The molecular identity of ion channels which confer PCO2/pH sensitivity in the brain is unclear. Heteromeric Kir4.1/Kir5.1 channels are highly sensitive to inhibition by intracellular pH and are widely expressed in several brainstem nuclei involved in cardiorespiratory control, including the locus coeruleus. This has therefore led to a proposed role for these channels in neuronal CO2 chemosensitivity. To examine this, we generated mutant mice lacking the Kir5.1 (Kcnj16) gene. We show that although locus coeruleus neurons from Kcnj16(+/+) mice rapidly respond to cytoplasmic alkalinization and acidification, those from Kcnj16(−/−) mice display a dramatically reduced and delayed response. These results identify Kir5.1 as an important determinant of PCO2/pH sensitivity in locus coeruleus neurons and suggest that Kir5.1 may be involved in the response to hypercapnic acidosis.

Both Kir5.1 and Kir4.1 were originally cloned from the brain and are expressed abundantly in the brainstem, especially in several CO 2 -chemosensitive nuclei involved in cardiorespiratory control. In particular, both subunits are coexpressed in locus coeruleus (LC) neurons (10,11). The LC is a CO 2 -chemosensitive region of the pons where more than 80% of neurons respond to hypercapnic acidosis with an increase in firing rate (12)(13)(14). This increase in firing rate of LC neurons during hypercapnia is primarily thought to involve changes in intracellular pH, rather than extracellular pH or molecular CO 2 (14). Kir4.1/Kir5.1 channels are therefore attractive candidates as potential chemoreceptors in these cells. However, the identity of the channels involved remains unclear as several other types of ion channels have also been implicated (15,16). One of the problems with dissection of these pathways is the lack of specific blockers for many of the channels involved. Therefore, to gain a greater understanding of the potential contribution of Kir5.1 to these chemosensitive pathways, we created a mutant mouse with a specific deletion of the Kir5.1 (Kcnj16) gene. Our results demonstrate that Kir5.1 plays a crucial role in defining the pH sensitivity of LC neurons and may therefore play an important role in their response to hypercapnic acidosis.

EXPERIMENTAL PROCEDURES
Creation of Kir5.1 (Kcnj16) Knock-out Mice-Kir5.1 (Kcnj16) is encoded by a single exon in the mouse genome. A Kir5.1 DNA probe was used to screen a phage library prepared using genomic DNA from a 129/SvJ mouse. Positive clones containing the Kir5.1 gene were isolated, and a 6.0-kb BamHI fragment encoding the last 59 amino acids of Kir5.1 and associated downstream sequence was cloned into the pBluescript SK ϩ vector. A neomycin resistance gene was then inserted to replace the remainder of the Kir5.1 open reading frame, and a 3.2-kb fragment encoding the sequence upstream of the Kir5.1 exon was then added to create the targeting vector. The linearized vector was then electroporated into ES cells derived from 129/SvJ mice, which were then cultured in the presence of G418. Positive clones were then analyzed by Southern blotting and PCR to identify successfully integrated constructs. Blastocyst-mediated transgenesis was then performed to produce chimeric mice. The generation of chimeric mice using the pKO5 vector was performed by Polygene AG (Rümlang, Switzerland). Chimeric mice were then bred with C57BL/6J mice, and a colony carrying the null allele was established by breeding heterozygous mutant mice. The strain was backcrossed with C57BL/6J for more than 10 generations to establish an isogenic strain. An illustration of the gene-targeting strategy is shown in Fig. 1. This strain is now available as a public resource through Mousebook on the Web.
Extracellular Recordings and Tight-seal, Whole-cell Recordings-Extracellular recordings were carried out by using aCSF-filled micropipettes. The action potentials of LC neurons were recorded using an Axopatch 200B amplifier, and acquired with a Pulse software. Patch clamp recordings were performed from LC neuron under visual control using Hamamatsu and Axioskop 2FS infrared optics and were recorded in the whole-cell voltage and current clamp configura-tions. Patch glass pipettes were pulled in several stages to a tip of about 1-m outside diameter, had resistances of 3-4 megohms, and were filled with an intracellular solution containing 1.5 mM potassium methylsulfate, 20 mM KCl, 1.5 mM MgCl 2 , 5 mM HEPES, 0.1 mM EGTA, 2 mM Mg-ATP, 0.5 Na-GTP, 10 mM phosphocreatine, pH ϳ7.4. The liquid junction potential was calculated to be ϳ10 mV (pipette negative relative to bath). All data were obtained using this solution and left uncorrected. The electrode was advanced into the brain slice and seals obtained by applying negative pressure. Seal resistances were 5-10 gigohms. The membrane was ruptured by further suction. The recordings were performed after Ն10 min of stable seal formation and were analyzed on condition that action potential amplitudes were Ն80 mV and that the resting membrane potentials were stable and more negative than Ϫ40 mV and the series resistance changed Ͻ20% throughout the entire recording period. NH 4 Cl was dissolved in aCSF to a final concentration of 10 mM, pH ϳ7.4. Complete exchange of the bath solution occurred in about 1-2 min.
Data Analysis and Statistical Evaluation-The firing rate was obtained by calculating the instantaneous firing frequency (IFF), as 1/spike interval and the time at the end of each interval was used to indicate the time for each IFF. The action potential height was measured from threshold to peak. The spike threshold was defined as the membrane potential at which the first derivative of the membrane potential exceeded 10 V/s. Data were acquired at 10 -20 kHz, filtered at 3 kHz, and analyzed with Pulse-fit, Origin7, and Igor. The statistical significance of the differences was calculated by using Student's t tests and ANOVA, and the difference was considered significant at p Ͻ 0.05 (*) and p Ͻ 0.01 (**).  ensure isogenicity. Breeding of heterozygous Kcnj16 (ϩ/Ϫ) mice generated live pups in a Mendelian ratio. Kcnj16 (Ϫ/Ϫ) mice were also viable, fertile, and exhibited no observable behavioral or physical abnormalities.

RESULTS
Reduced Response of Locus Coeruleus Neurons to pHi in Kir5.1-null Mice-LC neurons in the pons are confined to an easily identifiable anatomical area at the border of the IVth ventricle. Typical LC neurons are spontaneously active (0.5-5 Hz), display pacemaker-like firing, and possess consistent action potential parameters (17). Thus, the properties of LC neurons recorded in brain slices are remarkably uniform, as is their response to hypercapnic acidosis. Cultured LC neurons also retain their CO 2 chemosensitivity, indicating that external inputs are not essential for this process (18). Thus, LC neurons represent an ideal experimental model system to assess the potential contribution of Kir5.1 to PCO 2 /H ϩ chemosensitivity. Knockout of Kir5.1 means that the remaining Kir4.1 subunits can also still form homomeric channels but that this Kir conductance, although effective in controlling the resting membrane potential, will not be pH-sensitive within the physiological range (3).
We therefore compared the response of LC neurons to intracellular acidification using both Kcnj16 (ϩ/ϩ) and Kcnj16 (Ϫ/Ϫ) adult mice (P90 Ϯ 10 days). The electrical activity of LC neurons were recorded by means of whole-cell patchclamp recordings in both voltage-clamp and current-clamp modes. The resting membrane potentials of LC neurons recorded from either Kcnj16 (ϩ/ϩ) or Kcnj16 (Ϫ/Ϫ) slices oscillated between Ϫ49 mV and Ϫ58 mV, displayed similar input resistances (366 Ϯ 41 megohms and 402 Ϯ 40 megohms, respectively; n ϭ 9; p Ͼ 0.05) as well as basal firing frequencies (3.4 Ϯ 0.5 and 2.9 Ϯ 0.4 Hz, respectively; n ϭ 16; p Ͼ 0.05). Intracellular alkalinization and acidification were induced with ammonium chloride (NH 4 Cl, 10 mM). This "NH 4 Cl prepulse" technique is a well established method for intracellular acidification that has been shown to decrease the intracellular pH of many neuronal and nonneuronal cell types (19 -21). Bath perfusion of aCSF containing NH 4 Cl at pH 7.4 initially causes a transient intracellular alkalinization. However, the subsequent removal of NH 4 Cl causes a rapid intracellular acidification which reverses with time (19,21).
Interestingly, we also observed that that the firing rates transiently decreased ϳ20% in Kcnj16 (ϩ/ϩ) neurons (n ϭ 8) prior to withdrawal of the NH 4 Cl prepulse (Fig. 3, A and   C). This effect is similar to the reported effect of NH 4 Cl on LC neurons in rat brain slices due to transient intracellular alkalinization prior to withdrawal of the prepulse (12). However, in Kcnj16 (Ϫ/Ϫ) mice this initial decrease in firing rate was barely detectable (Ͻ5%; n ϭ 7; p Ͻ 0.01; Fig. 3, B and C).
Furthermore, we also found that the time course of the response to cytoplasmic acidification was delayed in Kcnj16 (Ϫ/Ϫ) mice. The firing frequency of Kcnj16 (ϩ/ϩ) neurons reached a maximum value 233 Ϯ 7 s after application of NH 4 Cl (latency of the effect). By contrast, Kcnj16 (Ϫ/Ϫ) neurons reached a peak firing frequency after 269 Ϯ 6 s (p Ͻ 0.01; n ϭ 8) (Fig. 3E). The degree of neuronal sensitivity to pHi was

Kir5.1 and Neuronal pH Sensitivity
also estimated from the slope of the NH 4 Cl-induced response (Fig. 3F). This showed that the degree of chemosensitivity measured in Kcnj16 (Ϫ/Ϫ) mice was only 25% of that seen in WT mice.
To confirm that these results were not due to an indirect effect of altering the cytoplasm of LC neurons in the wholecell recording configuration, we also performed extracellular recordings of LC neuronal activity in response to NH 4 Cl (Fig.  4). This confirmed that the effect was not dependent upon the recording configuration used.
Reduced Outward and Inward Currents in LC Neurons from Kir5.1-null Mice-We next performed whole-cell voltageclamp recordings to examine the currents elicited by NH 4 Cl application. Consistent with previous reports (12) we found that NH 4 Cl induced an outward current during superfusion and an inward current following NH 4 Cl withdrawal in WT Kcnj16 (ϩ/ϩ) LC neurons when clamped at Ϫ60 mV (Fig. 5A). However, in Kcnj16 (Ϫ/Ϫ) neurons these outward currents were barely detectable, and the inward currents were markedly reduced (Fig. 5, C and D). Furthermore, their activation was slower compared with WT Kcnj16 (ϩ/ϩ) LC neurons; inward currents in WT mice reached a peak amplitude of 156 Ϯ 32 pA after 236 Ϯ 32 s (n ϭ 10), whereas in Kcnj16 (Ϫ/Ϫ) neurons peak currents of 78.6 Ϯ 29 pA were reached after 309 Ϯ 20 s (n ϭ 6) (Fig. 5E). Their slope of activation was also reduced (Fig. 5F).
Consistent with the idea that this initial alkalinization-induced outward current might be caused by activation of a potassium channel in WT mice, we found that it reversed at the potassium equilibrium potential (Ϫ106 Ϯ 1.9 mV; n ϭ 6). However, the subsequent inward currents seen upon NH 4 Cl withdrawal did not show a reversal potential (Fig. 6).

Decreased Response of LC Neurons to CO 2 in Kir5.1-null
Mice-We next examined whether the reported response of LC neurons to hypercapnic acidosis (11)(12)(13)(14)(15)(16) was similar to the effects we observed with NH 4 Cl. We therefore tested the effects of aCSF bubbled with either 5% CO 2 (control) or 15% CO 2 (hypercapnia) and analyzed the effect on the spontaneous discharge rate of these neurons. All WT Kcnj16 (ϩ/ϩ) neurons tested responded to hypercapnia with a significant increase in their spontaneous firing rate. This effect reversed quickly after returning to 5% CO 2 (Fig. 7A, upper). The ⌬IFF response of Kcnj16 (ϩ/ϩ) neurons to 15% CO 2 was 1.48 Ϯ 0.2 Hz (n ϭ 5) (Fig. 7D). By contrast, Kcnj16 (Ϫ/Ϫ) neurons responded to 15% CO 2 with a reduced ⌬IFF of 0.81 Ϯ 0.1Hz (p Ͻ 0.05) (Fig. 7 A, lower, and D). The rate of this response was also flattened compared with WT neurons (Fig. 7, B and  C), and Fig. 7E shows that the slope of this response was ϳ50% of that seen in Kcnj16 (ϩ/ϩ) neurons.

DISCUSSION
We have generated a Kir5.1 knock-out strain of mice and used these mice to investigate the role of Kir5.1 in the cellular pathways that underlie the chemosensitivity of locus coeruleus neurons. Our results clearly demonstrate that Kir5.1 plays a key role in determining the PCO 2 /pH sensitivity of these neurons.
Generation of Kir5.1-null Mice-In addition to being expressed in the brain, Kir5.1 is found in a wide variety of peripheral and epithelial tissues (1,5,22). However, Kcnj16 (Ϫ/Ϫ) mice exhibited no obvious physical or behavioral deficits. Furthermore, despite expression of Kir5.1 in both ovaries and  spermatozoa (22,23), male and female Kcnj16 (Ϫ/Ϫ) mice were fertile. The seemingly normal development of Kcnj16 (Ϫ/Ϫ) mice is in marked contrast to Kir4.1 knock-out mice, which die within 10 -14 days due to abnormal cerebellar development (24). The isogenic strain of Kcnj16 (Ϫ/Ϫ) mice we have created therefore now provides an excellent resource for future studies to address the potential role of Kir5.1 in other tissues where it is also expressed.
Kir5.1 as a Potential CO 2 Chemoreceptor in Locus Coeruleus Neurons and the Role of pHi-Previous studies have shown that intracellular acidification, induced by CO 2 , markedly increases the firing rate of LC neurons (12, 14 -16). We were also able to observe this effect in LC neurons from our Kcnj16 (ϩ/ϩ) mice. However, we found that this response was dramatically reduced and much slower in Kcnj16 (Ϫ/Ϫ) neurons. Furthermore, we also observed a reduction in H ϩ -induced inward currents (Fig. 5) in Kcnj16 (Ϫ/Ϫ) neurons and therefore propose that intracellular acidification, generated by either CO 2 or NH 4 Cl prepulse, increases the firing rate of LC neurons by inhibition of Kir4.1/Kir5.1 channels and subsequent depolarization of the cell membrane (see Fig. 8). Although the NH 4 Cl-evoked inward current does not reverse at E K , a concurrent decrease in potassium conductance and increase in other cation conductances would result in an inward current, as has been described for muscarine-and substance P-evoked inward currents in LC neurons (25,26) and proposed as a possible mechanism of action for NH 4 Cl (12). Indeed, a number of studies suggest that the firing rate response of LC neurons to acidic stimuli is complex and may involve multiple different ion channels including L-type Ca 2ϩ channels activation (27). NH 4 Cl acidification may also act predominantly at an electronically distant part of the cell, and such remote regions may not become sufficiently polarized to reverse the current flow. On the other hand, our conclusion is strongly supported by the fact that NH 4 Cl-induced outward current is due to an increase in K ϩ conductance that is almost absent in Kcnj16 (Ϫ/Ϫ) neurons. These findings indicate that the NH 4 Cl-induced outward current in LC neurons is mediated predominantly by the activation of Kir4.1/Kir5.1 channels.
In LC neurons it has been proposed that it is intracellular acidification that mainly underlies the increase in the firing rate (14,27). Interestingly, Kir4.1/Kir5.1 channels only respond to changes in intracellular pH and are insensitive to changes in extracellular pH (6). This is in marked contrast to the effect of hypercapnic acidosis on many K2P channels which have also been identified as chemoreceptors in other brainstem areas and which respond to changes in extracellular pH (28). In LC neurons which express both TASK-1 and Kir4.1/Kir5.1 channels (29, 30) mechanisms will therefore exist to sense changes in both extracellular and intracellular pH, and such redundancy may be functionally advantageous. The presence of a residual response in LC neurons lacking Kir5.1 is also perhaps not surprising, especially given the presence of other chemosensitive channels (15). However, our results demonstrate that despite this redundancy Kir5.1 clearly plays a major role in defining the response of these neurons to CO 2 .
Potential Role of Kir5.1 Subunits in the Ventilatory Responses to Hypercapnia-The activity of LC neurons is thought to play a critical role in attention, learning and memory, stress-induced responses (e.g. ''fight or flight" response), anxiety, and certain pain sensations (31,32). The LC also exerts an excitatory influence on the CO 2 stimulation of breathing (15). In healthy individuals, the primary function of pHsensitive neurons in the LC may be to produce an aversive or anxiety response to elevated CO 2 levels. Indeed, LC neurons receive and send inputs to the medullary respiratory network and, although controversial, contribute to the respiratory responses to hypercapnia. A number of studies have proposed that besides the LC, several other chemosensitive areas of the brainstem (including the nucleus of the solitary tract; the retrotrapezoid nucleus, the ventrolateral medulla, and the pre-Bötzinger complex) (15,33,34) may also need to be stimu- lated by hypercapnic acidosis to elicit a full ventilatory response. The precise contribution of other pH-sensitive K ϩ channels (e.g. TASK) to chemosensation in these tissues remains controversial (35)(36)(37). Kir5.1 (and Kir4.1) are also expressed in some but not all of these other chemosensitive areas, indicating that the response to CO 2 is probably complex and involves a large degree of functional redundancy. Interestingly, Kir4.1/Kir5.1 channels are also found in many glial cells (38), and a recent study has also demonstrated a role for astrocytes in central chemosensitivity although changes in astrocytic membrane potential are not thought to be involved (39). Future studies of the role of Kir5.1 in chemoreception will therefore undoubtedly help to provide an important insight into the complex mechanisms that control these processes.
In conclusion, the generation of a Kir5.1 knock-out strain has allowed us to dissect directly the role of this subunit in defining the chemosensitive response of distinct cells of the central nervous system. We conclude that the physiological pH sensitivity of heteromeric Kir4.0/Kir5.1 channels allows cells expressing these channels to link changes in CO 2 levels to changes in neuronal activity and therefore act as highly sensitive PCO 2 /H ϩ chemoreceptors.