Pacemaking kisspeptin neurons

New findings

• •

  What is the topic for this review?

  This review focuses on analysis of the electrophysiological properties of hypothalamic kisspeptin neurons.

• •

  What advances does it highlight?

  Molecular and cellular electrophysiological analysis of kisspeptin neurons reveals that they express the critical pacemaker channels for generating burst-firing activity.

Kisspeptin (Kiss1) neurons are vital for reproduction. Gonatotrophin-releasing hormone (GnRH) neurons express the kisspeptin receptor (GPR54), and kisspeptins potently stimulate the release of GnRH by depolarizing and inducing sustained action potential firing in GnRH neurons. As such, Kiss1 neurons may be the presynaptic pacemaker neurons in the hypothalamic circuitry that controls reproduction. There are at least two different populations of Kiss1 neurons; one in the rostral periventricular area (RP3V) that is stimulated by oestrogens and the other in the arcuate nucleus that is inhibited by oestrogens. How each of these Kiss1 neuronal populations participates in the regulation of the reproductive cycle is currently under intense investigation. Based on electrophysiological studies in the guinea-pig and mouse, Kiss1 neurons in general are capable of generating burst-firing behaviour. Essentially, all Kiss1 neurons, which have been studied thus far in the arcuate nucleus, express the ion channels necessary for burst firing, which include hyperpolarization-activated, cyclic nucleotide-gated cation (HCN) channels and the T-type calcium (Cav3.1) channels. In voltage-clamp conditions, these channels produce distinct currents that can generate burst-firing behaviour in current-clamp conditions. The future challenge is to identify other key channels and synaptic inputs involved in the regulation of the firing properties of Kiss1 neurons and the physiological regulation of the expression of these channels and receptors by oestrogens and other hormones. The ultimate goal is to understand how Kiss1 neurons control the different phases of GnRH neurosecretion, hence reproduction.

Introduction

Kisspeptin (encoded by the Kiss1 gene) and its cognate receptor (GPR54) are essential for gating the onset of puberty and regulating reproductive function in mammals (Seminara et al. 2003; de Roux et al. 2003). Kiss1 mRNA is expressed in the hypothalamic arcuate nucleus and rostral periventricular area (RP3V), which includes the anteroventral periventricular nucleus (AVPV) in rodents (Gottsch et al. 2004; Mikkelsen & Simonneaux, 2009; Bosch et al. 2012). Kiss1 neurons are thought to serve as relay neurons for mediating the negative and positive feedback effects of gonadal steroids on gonadotrophin-releasing hormone (GnRH) and luteinizing hormone (LH) secretion (Smith et al. 2005a,b; Dungan et al. 2007; Glidewell-Kenney et al. 2008; Clarkson & Herbison, 2009; Oakley et al. 2009). Kiss1 neurons may also act as the presynaptic pacemaker neurons for driving GnRH neurons based on several lines of evidence. First, virtually all GnRH neurons express the kisspeptin receptor, GPR54 (Parhar et al. 2004; Irwig et al. 2004; Bosch et al. 2013), which makes them a potential target of Kiss1 neuronal input (Clarkson et al. 2008; Mikkelsen & Simonneaux, 2009). Second, electrical stimulation of AVPV Kiss1 neurons increases the action potential firing of mouse GnRH neurons (Liu et al. 2011). Third, kisspeptin potently stimulates the release of GnRH (Gottsch et al. 2004) by depolarizing and inducing the sustained firing of action potentials in GnRH neurons (Han et al. 2005; Liu et al. 2008; Dumalska et al. 2008; Zhang et al. 2008; Pielecka-Fortuna et al. 2008) Lastly, kisspeptin antagonists inhibit pulsatile GnRH/LH secretion (Roseweir et al. 2009). Collectively, these findings suggest that Kiss1 neurons are proximal (presynaptic) to GnRH neurons and are capable of driving GnRH neuronal activity.

Role of pacemaker current (I_h) and T-type calcium current (I_T) in burst firing

If Kiss1 neurons are the presynaptic pacemaker neurons that drive GnRH neurons, then they should exhibit the biophysical properties shared by other pacemaker-type neurons in hypothalamus, thalamus and subthalamic nucleus (Loose et al. 1990; McCormick & Huguenard, 1992; Erickson et al. 1993a,b; Bevan & Wilson, 1999; Lyons et al. 2010; Qiu et al. 2011). In these areas, burst firing is generated primarily by the hyperpolarization-acitvated, non-selective cation ‘h’-current (I_h) and the low-threshold, transient ‘T’-type calcium channel current (I_T; for review, see Zagotta & Siegelbaum, 1996; Lüthi & McCormick, 1998). The non-selective cation h-current is mediated by the hyperpolarization-activated cyclic nucleotide-gated (HCN) channel family, which includes channel subtypes HCN1–4. The T-type calcium current, I_T, is mediated by the low-threshold voltage-gated calcium channels, Ca_V3.1–3.3 (for review see Perez-Reyes, 2003). The T-type Ca²⁺ channel subtypes are ordered according to their inactivation kinetics, where Ca_V3.1 has the fastest and Ca_V3.3 the slowest inactivation kinetics. In a bursting model, I_h depolarizes neurons from hyperpolarized potentials, raising the membrane potential into the range of I_T activation (Erickson et al. 1993a; Kelly & Rønnekleiv, 1994; Lüthi & McCormick, 1998). Once activated, I_T initiates a prolonged Ca²⁺-driven depolarization above threshold, which has been termed the low-threshold spike (Tsien et al. 1987; Llinás, 1988). This depolarization then drives the neuron to fire an ensemble of Na⁺-driven action potentials, which constitutes the burst firing. It is this burst firing that facilitates the peptide (kisspeptin) release that is so vital for activating GnRH neurons, pubertal development and reproductive competence.

Although single action potential-generated calcium influx is sufficient to spark the release of classical neurotransmitters, burst firing or tetanic stimulation is required for the release of neuropeptides (Bicknell, 1988; Shakiryanova et al. 2005; Masterson et al. 2010). Twenty years ago we showed that I_h and I_T contribute to the phasic burst firing of guinea-pig supraoptic vasopressin neurons (Erickson et al. 1993a,b). These findings were not limited to the magnocelluar neurosecretory neurons; guinea-pig tuberoinfundibular (A12) dopamine (parvocellular neurosecretory) neurons were also found to express these critical pacemaker conductances (Loose et al. 1990). Recently, these findings were corroborated by Swedish researchers who determined that rat tuberoinfundibular dopamine neurons (identified subsequently with dual neurobiotin and immunoctyochemical staining for tyrosine hydroxylase) also exhibit rhythmic burst firing (Lyons et al. 2010). One would predict that rat tuberoinfundibular dopamine neurons also express I_h and I_T, but Lyons and colleagues did not carry out voltage-clamp experiments to determine whether the rat dopamine neurons express I_h and I_T. It is well known that the parvocellular neurosecretory GnRH neurons exhibit burst firing and express both I_h and I_T (Suter et al. 2000; Zhang et al. 2007, 2009; Chu et al. 2010); therefore, expression of I_h and I_T may be endogenous properties of hypothalamic ‘bursting’ neurons. Most importantly, 17β-estradiol treatment (which produces an LH surge in mice) upregulates the expression of T-type calcium channel (Ca_V3.1, Ca_V3.2 and Ca_V3.3) transcripts in GnRH neurons, which translates into a greater whole-cell T-type current (Zang et al. 2009). In all of the studies mentioned above, the T-type calcium channels are blocked with low concentrations of Ni²⁺, which is a property of these low-voltage-activated Ca²⁺ channels (Catterall et al. 2005).

In order to characterize the conductances in guinea-pig and mouse kisspeptin neurons, a combination of whole-cell voltage-clamp recordings and single-cell RT-PCR (scRT-PCR) was used in the hypothalamic slice preparation to determine whether Kiss1 neurons in arcuate nucleus exhibit burst firing and express the channels and currents necessary for generating this activity (Qiu et al. 2011; Gottsch et al. 2011).

Biophysical properties of guinea-pig Kiss1 neurons and response to glutamate

The initial studies were done in female guinea-pigs in the late follicular phase by targeting arcuate neurons using ‘blind patch’ recording and subsequently identifying the neurons using scRT-PCR or immunocytochemistry (Qiu et al. 2011). Of the 190 recorded neurons, 69 (36%) of these were identified as Kiss1 neurons. Guinea-pig Kiss1 neurons in the arcuate nucleus exhibited a resting membrane potential of −59 mV and input resistance (R_in) = 770 MΩ, and the vast majority (80%) of these cells expressed both I_h and I_T. Although adjacent pro-opiomelanocortin (POMC) and Neuropeptide Y/Agouti-related peptide (NPY/AgRP) neurons express one or both of these pacemaker conductances, neither express such large-amplitude currents (Kelly et al. 1990; Van den Top et al. 2004). To verify the channel types pharmacologically, the T-type calcium current was blocked with mibefradil (10 μm) and the h-current with ZD 7288 (50 μm). Given that the vast majority of Kiss1 neurons expressed these pacemaker currents, it was of interest to see whether these critical ‘relay’ neurons in the reproductive circuit would exhibit burst-firing behaviour. Indeed, the glutamate receptor agonist NMDA (20 μm) reliably induced burst-firing activity in 78% of guinea-pig Kiss1 neurons, essentially the same percentage that express I_h and I_T. The NMDA-induced bursting activity was similar to that reported for midbrain dopamine neurons (Seutin et al. 1994). Therefore, Kiss1 neurons in the guinea-pig arcuate nucleus express the endogenous pacemaker conductances (T-type calcium and h-currents) and can be driven via glutamatergic inputs to generate burst-firing activity.

Biophysical properties of mouse Kiss1 neurons in the arcuate nucleus

A substantial step forward was the development of a of Kiss1-CreGFP knock-in mouse by Robert Steiner and Richard Palmiter at the University of Washington (Gottsch et al. 2011), making it possible to target GFP-expressing Kiss1 neurons. As we would have predicted based on the guinea-pig studies, both I_h and I_T were prominent in Kiss1 neurons in the mouse arcuate nucleus, and these neurons also expressed the critical transcripts (HCN and Ca_V3). Based on quantitative PCR of pooled Kiss1 neurons, HCN2, which is the most sensitive to cyclic nucleotide gating (Craven & Zagotta, 2006), was expressed at 2.5-fold higher levels than HCN1 and ∼50% higher levels than HCN3. Moreover, Ca_V3.1 channel transcripts were expressed at higher levels than Ca_V3.2, and Ca_V3.3 was not detected in the mouse Kiss1 neurons. The mouse Kiss1 neurons in the arcuate nucleus of ovariectomized females exhibit a resting membrane potential of −64 mV (Gottsch et al. 2011), compared with −75 mV for castrated males (Navarro et al. 2011). The relatively more negative resting membrane potential for mouse versus guinea-pig Kiss1 neurons in the arcuate nucleus (−59 mV) probably contributes to the higher percentage (∼50%) of silent (no spontaneous activity) neurons in this species (Gottsch et al. 2011; de Croft et al. 2012).

Although mouse Kiss1 neurons in the arcuate nucleus are silent or display low, irregular firing patterns (Gottsch et al. 2011; de Croft et al. 2012), they are essentially ‘one synapse’ away from generating burst-firing activity. The glutamate agonist NMDA induced burst-firing behaviour in all of the Kiss1 neurons in the arcuate nucleus. This was demonstrated in both the loose cell-attached mode, in which there was no disruption of the intracellular constituents, and in whole-cell current-clamp recordings (Fig. 1). In the presence of tetrodotoxin to block the fast Na⁺ spikes, the membrane oscillations (up and down states) induced by NMDA were clearly visible. The oscillatory behaviour that underlies the burst of Na⁺ action potentials appears to be an endogenous property of these Kiss1 neurons. A recent report has shown that neurokinin B (NKB) will induce burst-firing behaviour in male Kiss1 (Kiss1-Cre) neurons in the arcuate nucleus (Navarro et al. 2011), and the actions of NKB were antagonized by a neurokinin 3 receptor antagonist (Alreja, 2013). Kisspeptin neurons co-localize NKB and dynorphin; hence, the origin of the ‘KNDy’ acronym (Lehman et al. 2010). The autoregulatory excitation by NKB may be a mechanism by which Kiss1 neurons synchronize to excite GnRH neurons. Certainly, in vivo multiunit recordings in the arcuate nucleus region of KNDy neurons in the goat would suggest that there is some synchronous burst-firing behaviour (Ohkura et al. 2009).

Figure 1

Electrophysiological characteristics of arcuate Kiss1 neurons in oil-treated, ovariectomized Kiss1-CreGFP mice using whole-cell patch recording
Kisspeptin (Kiss1) neurons in the arcuate nucleus (of the female) rest at −63.8 ± 2.3 mV (n= 20). A–C, representative traces of action potentials recorded from arcuate Kiss1 neurons showing tonic (A), irregular (B) and silent firing patterns (C). D, summary pie chart of the firing pattern distribution in Kiss1 neurons of the arcuate nucleus (n= 20). E–H, endogenous conductances and NMDA-induced burst firing of an arcuate nucleus Kiss1 neuron. E, ensemble of currents in response to depolarizing/hyperpolarizing steps from −50 to −140 mV, illustrating the expression of a hyperpolarization-activated cyclic nucleotide-gated cation current (h-current) and a T-type Ca²⁺ current (arrow) in a representative Kiss1 neuron. Holding potential is −60 mV. F, current-clamp recording in a Kiss1 neuron of the arcuate nucleus, showing the response to NMDA (40 μm). The action potential firing was expanded to illustrate the pronounced effects of NMDA on burst-firing activity of Kiss1neurons, highlighting the ensemble of Na⁺ spikes riding on top of low-threshold spikes (arrows). G, in a current-clamped state close to the resting membrane potential, it is possible to see the bursting activity in the presence of NMDA. H, in the presence of tetrodotoxin to block the fast Na⁺ spikes, it is possible clearly to see the membrane oscillations (up and down states) induced by NMDA. Eighty per cent of arcuate Kiss1 neurons expressed the endogenous conductances (h-current and T-current) that are critical for generating burst-firing activity. Although Kiss1 neurons in the arcuate nucleus did not exhibit spontaneous burst-firing activity, glutamate (NMDA) was able to induce burst-firing activity in all of the cells. I, GABA (100 μm) inhibited firing in another arcuate kisspeptin neuron. Drugs were rapidly perfused into the bath as a 100 μl bolus.
(From Gottsch et al. Molecular properties of Kiss1 neurons in the arcuate nucleus of the mouse. Endocrinology 152: 4298–4309. Copyright 2011, The Endocrine Society.)

Kisspeptin neurons in the arcuate nucleus exhibit a relatively negative resting membrane potential (−64 mV in the mouse and −59 mV in the guinea-pig), but this membrane potential is not hyperpolarized enough to recruit a critical fraction of T-type calcium channels for generating burst firing (Zhang et al. 2009). The vast majority of the h-current is activated at even more negative potentials (Chu et al. 2010). Thus, some sort of inhibitory synaptic input is necessary for reaching these nadirs in the membrane potentials, and GABA hyperpolarizes Kiss1 neurons (Qiu et al. 2011; Gottsch et al. 2011), thereby providing a critical inhibitory feedback circuit for generating rhythmic burst firing (McCormick & Huguenard, 1992; Bevan & Wilson, 1999). GABA neurons are abundant in the arcuate nucleus (Wagner et al. 1999), and a subset of Kiss1 neurons expresses glutamic acid decarboxylase (GAD; Cravo et al. 2011). As such, Kiss1 neurons may themselves be an endogenous source of GABA, whose action could be autoinhibitory and perhaps be responsible for hyperpolarizing Kiss1 neurons via GABA_B receptors in addition to the actions of dynorphin via κ-opioid receptors (Navarro et al. 2009; Wakabayashi et al. 2010). Therefore, Kiss1 neurons in the Kiss1-Cre mouse express the same pacemaker conductances as the Kiss1 neurons in the female guinea-pig (identified by scRT-PCR) and, with the appropriate inputs, have the capacity to burst fire rhythmically.

Properties of mouse Kiss1 neurons in the anteroventral periventricular nucleus/rostral periventricular area

Kisspeptin neurons in the AVPV project to GnRH neurons in the preoptic area (Clarkson & Herbison, 2006) and have been identified as excitatory afferents to the GnRH neurons (Liu et al. 2011). Kisspeptins are the most potent agonist to induce GnRH neuronal firing (Han et al. 2005; Zhang et al. 2008; Pielecka-Fortuna et al. 2008). Although cell-attached recordings of Kiss1 neurons in the AVPV have been reported (Ducret et al. 2010; de Croft et al. 2012), studies are forthcoming that characterize the biophysical and molecular properties of these neurons. One recent study characterized some of the elementary properties, limited to the resting membrane potential, input resistance and cell capacitance, but did not elucidate any of the critical endogenous conductances for the generation of burst-firing activity (Frazão et al. 2013).

Using a Kiss1-Cre/GFP mouse to target Kiss1 neurons in the AVPV/RP3V (Cravo et al. 2011), Frazão et al. (2013) described two distinct populations of AVPV/RP3V Kiss1 neurons in males and females; type I neurons fired irregularly and exhibited a resting membrane potential of −50 mV, whereas type II neurons were silent and rested at −67 mV (Frazão et al. 2013). The negative resting membrane potential of type II neurons appeared to depend on the expression of the K_ATP channel, because blockade of the channel activity with the sulfonylurea drug tolbutamide depolarized the type II neurons to a membrane potential equivalent to the resting membrane potential of type I neurons. However, Frazão et al. (2013) did not report that any of the Kiss1-Cre/GFP neurons in the AVPV/RP3V exhibited spontaneous burst-firing activity in the presence of the K_ATP channel blocker. For comparison, Frazão et al. (2013) measured the firing activity of Kiss1 neurons in the arcuate nucleus and found comparable firing rates to those in the AVPV/RP3V, but the Kiss1 neurons in the arcuate nucleus did not exhibit a bimodal distribution of resting membrane potentials. In contrast, in loose cell-attached recording studies, Allan Herbison and colleagues reported that 20–25% of AVPV/RP3V Kiss1 neurons in wild-type (C57BL/6J) mice, identified post hoc by immunocytochemical staining, exhibited burst firing, with a small increase in burst-firing activity during pro-oestrus (Ducret et al. 2010). The differences between the studies of Frazão et al. (2013) and Ducret et al. (2010) could be due to the experimental approach (whole-cell versus cell-attached recording) or the fact that some of the recordings in the Kiss1-CreGFP mouse were done from ‘ectopic’ Kiss1-expressing neurons as originally indicated (Cravo et al. 2011). 17β-Estradiol treatment of ovariectomized females shifted the firing pattern of AVPV/RP3V Kiss1 neurons from quiescent (type II) to spontaneously firing (type I). However, steroid treatment also increased the IPSC amplitude, which is difficult to reconcile with an increase in firing, unless the GABA_A input is excitatory as has been reported for the actions of GABA_A agonists on GnRH neurons (Herbison & Moenter, 2011). In support of the excitatory GABA_A input is the fact that female prepubertal AVPV/RP3V Kiss1 neurons also exhibited more IPSCs, yet all of these neurons were spontaneously active (Frazão et al. 2013). Again, no spontaneous burst firing was reported in these initial recordings, and clearly, additional voltage-clamp studies are warranted.

Kisspeptin neurons respond to hormones

It has been hypothesized that mouse Kiss1 neurons in the arcuate neurons are a target for the actions of leptin based on the findings that these neurons express the leptin receptor (LRb; Smith et al. 2006); however, studies had not shown a direct action of leptin to activate (excite) Kiss1 neurons. We hypothesized that leptin would depolarize kisspeptin neurons via activation of a non-selective cationic current, as previously described for POMC neurons (Qiu et al. 2010). Indeed, leptin (100 nm) depolarizes and increases the firing frequency of Kiss1 neurons by twofold (Qiu et al. 2011). Similar to mouse POMC neurons (Qiu et al. 2010), the leptin-activated current (measured in voltage-clamp conditions) reverses near −15 mV, indicative that leptin activates a non-selective cationic current. Given that leptin activates canonical transient receptor potential (TRPC) channels in POMC neurons, it follows that these channels were also conducting the cation current in Kiss1 neurons.

In heterologous expression systems, micromolar concentrations of lanthanum (La³⁺) potentiate TRPC4 and TRPC5 channels (Strübing et al. 2001). In arcuate Kiss1 neurons, La³⁺ (100 μm) potentiates the leptin-induced current in Kiss1 neurons by about twofold. Also, the relatively selective TRPC channel blocker 2-aminoethyldiphenylborinate (100 μm; Clapham et al. 2005) completely abrogates the effects of leptin. Therefore, the physiological and pharmacological characterization indicates that TRPC4 and TRPC5 channels are involved in the leptin-mediated depolarization of Kiss1 neurons. The pharmacological experiments were followed by molecular studies in which the expression of TRPC1, TRPC4 and TRPC5 channel transcripts were measured in guinea-pig Kiss1 neurons using scRT-PCR. In the analysis of 114 guinea-pig Kiss1 neurons, identified by Kiss1 mRNA expression, it was found that TRPC1 mRNA was expressed in 60% of the cells, TRPC4 mRNA in 20% of the cells and TRPC5 mRNA in 64% of the cells, indicating that TRPC1 and TRPC5 are the primary channel transcripts in Kiss1 neurons. Moreover, dual immunocytochemical analysis of tissue sections through the arcuate nucleus of the guinea-pig revealed that TRPC5 protein was expressed in Kiss1 neurons (Qiu et al. 2011). Collectively, these findings suggest that arcuate Kiss1 neurons are sensitive to leptin activation via TRPC5 channels.

It is known that LRb is coupled to the janus kinase (Jak2)–phosphotidylinositide 3-kinase (PI3K) signalling pathway, and this pathway will activate TRPC channels in mouse POMC neurons (Qiu et al. 2010). Subsequently, it has been shown in guinea-pig Kiss1 neurons that the Jak inhibitor AG490 (10 μm) potently blocks the effects of leptin. In addition, PI3K is essential for leptin-induced neuronal activation (Hill et al. 2008; Qiu et al. 2010), and it is also critical for the membrane insertion of TRPC channels (Bezzerides et al. 2004). In fact, the selective PI3K inhibitor wortmannin (100 nm) completely abrogates the TRPC current in Kiss1 neurons. This indicates that LRb couples to Jak2 to activate PI3K in guinea-pig Kiss1 neurons. Moreover, the selective phospholipase C-γ inhibitor ET-18-OCH3 (15 μm) potently blocks the effects of leptin (Qiu et al. 2011), similar to its actions in POMC neurons (Qiu et al. 2010). Again, molecular studies followed the physiological/pharmacological analysis, and scRT-PCR revealed that LRb and phospholipase C-γ1 transcripts are expressed in guinea-pig Kiss1 neurons. Therefore, it appears that Kiss1 neurons in the arcuate nucleus of the guinea-pig are a target for the actions of leptin, which suggests that this population of Kiss1 neurons are critical relay neurons in conveying energy status to GnRH neurons and are vital for pubertal development (Dungan et al. 2006).

Kisspeptin–GnRH neuronal communication: from channels to circuits

A recent publication by Catterall et al. (2012) reminded readers of the Hodgkin–Huxley heritage and the fact that investigators must move from studying channels to exploring circuits. What is critical for the control of GnRH neuronal excitability and ultimately the control of fertility is the hypothalamic circuitry. This circuitry includes not only the excitatory and inhibitory synaptic input onto Kiss1 and GnRH neurons, but also the effects of circulating hormones, such as 17β-estradiol and leptin, that convey vital feedback information about reproductive and energy states, respectively. Over 20 years ago, we identified the conductances (before the channels were cloned), first in the tuberoinfundibular dopamine neurons and then in vasopressin neurons, that were critical for generating burst-firing behaviour (see Kelly & Rønnekleiv, 1994). These conductances (I_h and I_T) were also described in thalamic neurons at about the same time (McCormick & Huguenard, 1992). Although GnRH neurons express both of these endogenous pacemaker conductances (Zhang et al. 2007, 2009; Chu et al. 2010), GnRH neurons are not the sole pacemaker neurons in the circuitry. It is apparent that Kiss1 neurons, which express both I_h and I_T highly, will generate burst firing with very little provocation, i.e. excitatory glutamate input (Fig. 1). Therefore, the future challenge is to identify not only all of key players (channels) and how they are regulated (e.g. HCN and Ca_V3 channels are highly upregulated by 17β-estradiol treatment; Zhang et al. 2009; Bosch et al. 2013), but also how these channels and perhaps other channels fit into the Kiss1–GnRH neuronal circuitry for the generation of burst firing.

Appendix

Additional information

Note added in proof

A recent publication (Piet et al. (2013) J Neurosci 33, 10828–10839) has shown that AVPV/RP3V Kiss1 neurons express I_h, and the current is regulated by oestrogen.

Competing interests

None declared.

Funding

The work in the authors’ laboratories was supported by the National Institutes Health R01 grants NS 38809, NS 43330 and DK 68098. The content is solely the responsibility of the authors and does not necessarily represent the official view of the National Institutes of Health.

Acknowledgements

The authors would like to acknowledge the excellent technical support of Ms Martha A. Bosch.