Address correspondence and reprint requests to Dr. D.M. Durand at Department of Biomedical Engineering, 112 Wickenden Building, Case Western Reserve University, 10900 Euclid Ave., Cleveland, OH 44106, U.S.A. E-mail: firstname.lastname@example.org
Summary: Purpose: It has been shown that a low-calcium high-potassium solution can generate ictal-like epileptiform activity in vitro and in vivo. Moreover, during status epileptiform activity, the concentration of [K+]o increases, and the concentration of [Ca2+]o decreases in brain tissue. Therefore we tested the hypothesis that long-lasting persistent spike activity, similar to one of the patterns of status epilepticus, could be generated by a high-potassium, low-calcium solution in the hippocampus in vivo.
Methods: Artificial cerebrospinal fluid was perfused over the surface of the exposed left dorsal hippocampus of anesthetized rats. A stimulating electrode and a recording probe were placed in the CA1 region.
Results: By elevating K+ concentration from 6 to 12 mM in the perfusate solution, the typical firing pattern of low-calcium ictal bursts was transformed into persistent spike activity in the CA1 region with synaptic transmission being suppressed by calcium chelator EGTA. The activity was characterized by double spikes repeated at a frequency ∼4 Hz that could last for >1 h. The analysis of multiple unit activity showed that both elevating [K+]o and lowering [Ca2+]o decreased the inhibition period after the response of paired-pulse stimulation, indicating a suppression of the after-hyperpolarization (AHP) activity.
Conclusions: These results suggest that persistent status epilepticus–like spike activity can be induced by nonsynaptic mechanisms when synaptic transmission is blocked. The unique double-spike pattern of this activity is presumably caused by higher K+ concentration augmenting the frequency of typical low-calcium nonsynaptic burst activity.
Excitatory chemical synaptic connections between neurons are considered critical in the development of epileptiform activity, especially for interictal spikes with brief periods (1). However, previous studies have shown that nonsynaptic mechanisms play an important role in producing ictal activity with longer burst duration. For instance, elevation of extracellular K+ concentration can induce brief interictal-like activity in hippocampal slices (2,3). Reducing extracellular Ca2+ concentration in addition to elevating K+ can induce long-lasting ictal-like seizure activity in hippocampal slices with synaptic transmission blocked (4–6). Furthermore, with normal Ca2+ levels, blocking synaptic transmission with neurotransmitter-receptor blockers can transform the interictal discharges induced by high K+, 4-aminopyridine, or other convulsants into long-duration ictal-like discharges (7). Nonsynaptic mechanisms, such as gap junction, have been found to be crucial in many ictal epileptic models as well. For instances, gap-junctional blockers can reduce or abolish seizure activity induced by different means such as 4-aminopyridine, zero-magnesium solution, or tetanic stimulation (8–10). Taken together, these data indicate that non-synaptic mechanisms are important for the generation of ictal-like activity and could prolong the burst duration.
Status epilepticus (SE), one of the most severe and damaging forms of epileptic seizure, can last for >30 min with a frequency of population burst >2 Hz (11). When SE is induced either by chemical convulsants or by electrical stimulation, K+ concentration increases, and Ca2+ concentration decreases significantly in brain tissue (12–14), suggesting that nonsynaptic mechanisms could be important in the generation of SE. The question arising is whether SE could be generated in the absence of functioning synaptic transmission. In human epilepsy, two common types of SE patterns have been observed: intermittent seizure activity without recovery and continuous seizures (11). Ictal seizure activity is a typical type of in vitro low-Ca2+ nonsynaptic epileptiform activity induced by high-K+ and low-Ca2+ solution when synaptic transmission is blocked (4–6,15,16). However, low-Ca2+ bursts usually have intervals lasting for tens of second to minutes. Persistent seizure activity has not been reported in the in vitro preparations of the high-K+, low-Ca2+ nonsynaptic epilepsy model. This situation might be due to the lack of enough intact connections in the in vitro preparations. Recently we observed that by perfusing the exposed hippocampus in vivo with a solution containing high K+ (12 mM) and a calcium chelator, a type of persistent spike activity could be induced when synaptic transmission was blocked (17,18). We present data concerning the relation between the persistent nonsynaptic spike events and the typical low-Ca2+ nonsynaptic ictal events by using in vivo preparations. Multiple unit activity will be recorded for the analysis of the possible mechanism underlying the persistent spike activity.
All procedures used in this study were approved by the Institutional Animal Care and Use Committee, Case Western Reserve University, Cleveland. Adult Sprague–Dawley rats (334 ± 45 g; n = 25) were anesthetized with urethane (1.25–1.5 gm/kg, i.p.) and placed in a stereotaxic apparatus. Body temperature was maintained at 37°C with a heating pad. The skull over the left cortex was opened, and the neocortex overlying the dorsal hippocampus was removed. Artificial cerebrospinal fluid (aCSF) was warmed to ∼37°C and perfused over the surface of the exposed dorsal hippocampus and was refreshed every 5 min throughout the experiment.
Solutions and drugs
Normal aCSF consisted of (in mM): 124 NaCl, 5 KCl, 1.25 NaH2PO4, 2 CaCl2, 1.5 MgSO4, 26 NaHCO2 and 2 g/L d-glucose. Calcium chelator ethylene glycol–bis (β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA), 5 mM, replaced CaCl2 to reduce [Ca2+]o in vivo. In the solution with a high concentration of K+, KCl was increased to 12 or 15 mM. Picrotoxin (PTX, 0.4 mM) was used to induce epileptic bursts by blocking the γ-aminobutyric acid (GABA)A receptor. All chemicals were obtained from Sigma or Fisher.
Recording of spontaneous and evoked potentials
Multichannel silicon recording probes provided by the Center of Neural Communication Technology, University of Michigan, were used for recording the spontaneous and evoked field potentials in the CA1 region of hippocampus. Two contacts located on the same probe, separated by a fixed distance of 400 μm, were used to record simultaneously the field potentials in the pyramidal layer (Pyr.) and stratum radiatum (S. rad.), respectively. The probe contacts were made of iridium with an area of 1,250 μm2. Stimulating electrodes were made from pairs of insulated nichrome wires (80 μm in diameter) with a 0.5-mm vertical tip separation. The recording probe was positioned in left hippocampal CA1 area (AP, −3.0; ML, 2.6), and the stimulating electrode was inserted into the same side (AP, −2.0; ML, 2.3) for Schaffer collateral stimulation. Patterns of the evoked potentials guided the vertical positions of both the recording probe and the stimulating electrode. Two stainless steel screws, fixed in the bone of the nose, served as ground and reference electrodes. Square-pulse stimulation with 0.1-ms duration and 0.30- to 0.35-mA amplitude was used to induce maximal orthodromic evoked potential in the CA1 region. Paired-pulse stimulus with a 25-ms interval was used to measure paired-pulse inhibition (PPI) caused mainly by the GABAergic inhibition.
CA1 field potential signals were amplified 1,000 times by model 1700 four-channel amplifiers (A-M System, Inc.) with filter frequency ranges of 0.1 Hz to 5 KHz for both spontaneous and evoked potentials. Signals were then sampled at a rate of 20 KHz by using a ML795 PowerLab/16SP data-acquisition system (ADInstruments, Castle Hill, NSW, Australia) and stored on hard disk for off-line analysis. Multiple unit activity (MUA) was separated from the pyramidal layer spontaneous field potential by a high-pass filter with a cutoff frequency of 500 Hz. The time period with a low level of MUA after evoked spikes or spontaneous spikes was referred as the inhibition period (IP), and its duration was measured as an indicator of the AHP period with few neural firings.
Data are expressed as mean ± standard deviation. A Student's t test (two-tailed) was used for statistical comparisons.
Immediately after the surgical exposure of the hippocampus, normal aCSF was perfused over the hippocampus for ∼2 h to allow recovery from surgery. Stimulating electrode and recording probe were placed, and baseline signals of both spontaneous and evoked potentials were recorded at the end of this period. Then the aCSF solution with EGTA, 5 mM, and K+, 6 or 7.5 mM, was used to suppress the synaptic transmission and induce nonsynaptic epileptiform activity in the CA1 region. After the appearance of nonsynaptic bursts, K+ concentration was increased to 12 or 15 mM (n = 12). In another experimental group, K+ concentration was increased to 12 mM with normal Ca2+, 2 mM, for 40 min, followed by elimination of Ca2+ for another 40 min and final addition of EGTA, 5 mM, in the aCSF solution to reduce the [Ca2+]o (n = 8).
Elevation of potassium concentration transforms nonsynaptic ictal-like activity into status epilepticus–like activity
About 20 min after the onset of perfusion with aCSF containing EGTA, 5 mM, and K+, 6 or 7.5 mM, over the exposed hippocampus, synaptic transmission was suppressed in the CA1 region, and a type of ictal-like slow-wave activity appeared with (n = 7/12) or without (n = 5/12) superimposed spikes. This slow-wave activity was similar to the nonsynaptic epileptiform activity both in vitro (16) and in vivo (19). Large-amplitude spontaneous spikes recorded from the pyramidal layer represent the simultaneous firing of a population of neurons. To distinguish between spontaneous spikes and evoked population spikes (PSs), “spike” is used to indicate spontaneous population spikes. Figure 1 shows an example of slow-wave activity with superimposed burst spikes. During baseline recording (Fig. 1A), orthodromic Schaffer collateral stimulation evoked large field PSs in the CA1 pyramidal layer and field excitatory postsynaptic potentials (fEPSPs) in the CA1 stratum radiatum. No epileptiform activity was observed in the spontaneous potentials. Thirty minutes after the application of EGTA, 5 mM, and K+, 7.5 mM, slow-waves superimposed by burst spikes appeared with suppressed evoked potentials, indicating a decrease of synaptic transmission (Fig. 1B). Increasing K+ to 15 mM changed the slow-waves into persistent double-spike activity repeated at a frequency of ∼4 Hz lasting >30 min (Fig. 1C). The slow waves with superimposed spikes returned when K+ in the aCSF solution was changed back to 7.5 mM (Fig. 1D). EGTA, 5 mM, was applied through the whole period. The suppression of synaptic transmission indicated by a decrease of the evoked responses was persistent during the switches of epileptic patterns caused by the changes of K+ concentrations (Fig. 1B–D).
The application of aCSF with EGTA, 5 mM, and K+, 6 or 7.5 mM, sometimes induced slow-waves without superimposed spikes (Fig. 2). Baseline recordings in Fig. 2A show normal evoked responses and normal spontaneous potentials in both the pyramidal layer and the stratum radiatum. Thirty-five minute application of a solution containing EGTA, 5 mM, and K+, 7.5 mM, induced low-frequency high-amplitude slow waves without superimposed spikes in the spontaneous potentials. This low-frequency activity occurred when the evoked fEPSP decreased significantly and the evoked PS disappeared, thereby indicating suppression of synaptic transmission (Fig. 2B). The absence of spontaneous spikes in slow-waves did not necessarily mean absence of firing. It could also be the absence of phase-locked firing. After the elevation of K+ in the solution to 15 mM, persistent double-spike activity replaced the slow waves with a frequency of 3 to 4 Hz and large amplitude in the pyramidal layer (Fig. 2C). The spiking activity did not stop until the K+ concentration was returned to 7.5 mM, and the lower frequency slow-waves reappeared (Fig. 2D).
These results indicate that the slow-wave burst pattern induced by EGTA, 5 mM, and K+, 6 or 7.5 mM, was transformed into persistently recurring double population spikes when K+ was increased to 12 or 15 mM. Similar nonsynaptic persistent spike activity could also be generated directly from a nonepileptic condition. As shown in Fig. 3, the application of K+, 12 mM, with Ca2+, 2 mM, could not induce epileptiform activity despite an increase in neuronal excitability indicated by multiple-PSs in evoked potentials (Fig. 3A). Even the removal of Ca2+ from the perfusion solution with K+, 12 mM, did not induce epileptiform activity in the spontaneous potentials, although the evoked PSs were larger in the pyramidal layer (Fig. 3B). However, continuous spikes appeared almost immediately after the addition of EGTA, 5 mM, in the solution and lasted >1 h. At the beginning, the spike activity was observed as separate spikes with higher frequencies and higher amplitudes (Fig. 3C and E). About 20 min later, the activity appeared as paired spikes repeated at a frequency of ∼4 Hz (Fig. 3D and E, diamond line) while synaptic transmission was suppressed, as indicated by the disappearance of PSs and very low amplitude fEPSPs in the evoked potentials.
These data suggest that the 12 mM K+ concentration with normal Ca2+ concentration applied to the exposed surface of hippocampus cannot evoke epileptiform activity, whereas the combination of low Ca2+ concentration and elevated K+ concentration can induce sustained nonsynaptic double-spike activity that is different from low-Ca2+ slow-wave activity. The amplitude, frequency, and duration of both the nonsynaptic slow-wave activity and the double-spike activity are compared in Table 1. The duration of a single double-spike event (0.038 ± 0.039 s) was significantly shorter than that of a slow-wave event (1.24 ± 0.31 s; n = 12; p < 0.01). The frequency of the double spike (3.77 ± 1.28 Hz) was significantly higher than that of the slow wave (0.34 ± 0.14 Hz; n = 12; p < 0.01).
Table 1. Comparison among nonsynaptic slow waves, nonsynaptic double-spike activity, and PTX-induced interictal and ictal activity
Amplitude (mV) Pyramidal layer
Amplitude (mV) Stratum radiatum
ap < 0.01, double spikes vs. slow waves.
bp < 0.001, PTX interictal vs. double spikes.
cp < 0.001, PTX ictal vs. double spikes; Student's t test.
0.34 ± 0.14
1.24 ± 0.31
1.20 ± 0.48
3.25 ± 1.10
3.77 ± 1.28a
0.038 ± 0.039a
1.78 ± 1.21
1.00 ± 0.37a
0.28 ± 0.07b
0.14 ± 0.01b
2.19 ± 0.71
4.11 ± 1.34
0.09 ± 0.03 c
0.69 ± 0.21 c
6.79 ± 1.36
9.57 ± 1.79
The nonsynaptic continuous double spike repeated at a frequency of ∼4 Hz was quite different from the traditional synaptic epileptiform activity induced by convulsants, usually with long silent intervals between bursts. It has been shown that elevating K+ concentration can increase the frequency of burst activity (2,20–22) because of a decrease in the potassium AHP current. In addition, reducing Ca2+ concentration can also decrease inhibitory activity (15). Therefore we hypothesized that a decrease in inhibitory activity, such as GABAergic inhibition and AHP activity, was the underlying mechanism generating the high frequency of continuous double-spike activity observed in this investigation. This hypothesis was investigated by measuring multiple unit activity (MUA) and by comparing these waveforms with picrotoxin (PTX)-induced synaptic epileptiform activity.
Decrease of inhibitory activity by high-concentration K+ and low-concentration Ca2+
To study the gradual changes of neural activity, the exposed hippocampus was perfused first with an aCSF solution containing a 12 mM concentration of K+ for 40 min, and then with a zero-Ca2+ aCSF for another 40 min, and finally, EGTA (5 mM) was added.
During baseline recording with a perfusion solution of K+, 5 mM, and Ca2+, 2 mM, paired-pulse stimulation with an interval of 25 ms evoked a large PS in the first response (PS1) but no PS in the second response (PS2), indicating strong GABAergic inhibition (e.g., paired-pulse inhibition, PPI) in the CA1 region. After paired-pulse stimulation, a long inhibition period (IP) with a low level of MUA activity indicated a long period of AHP with limited neural activity (Fig. 4A). Elevating K+ to 12 mM in the solution changed the single PS1 into multiple PSs and decreased paired-pulse inhibition, as indicated by the appearance of a second population spike (PS2; Fig. 4B). Although the IP duration did not change significantly in the MUA recording, the appearance of an afterdischarge within the IP immediately after the paired-pulse stimulation (open arrow in Fig 4B) indicates an increase in neural excitability by high K+. After the removal of Ca2+ in the high-K+ perfusion solution, the amplitude of PS2 increased, and IP duration decreased significantly, indicating a further decrease of both GABAergic inhibition and AHP (Fig. 4C). The EGTA solution induced single spontaneous spike activity, reduced synaptic transmission, as indicated by a smaller PS1 amplitude with increased PS1 latency, and significantly decreased PPI and IP duration (Fig. 4D). Thirty minutes after the application of EGTA, spontaneous double-spike activity continued, and synaptic transmission was mostly blocked (Fig. 4E) with small IP duration.
The mean IP duration in the final 20 min with 12 mM K+, zero Ca2+ (0.52 ± 0.08 s) was significantly shorter than that in the final 20 min with 2 mM Ca2+ (1.14 ± 0.44 s; F= 13.9; p = 0.003; n = 8; repeated measures ANOVA). The mean IP duration in the first 20 min of EGTA application was even shorter (0.30 ± 0.09 s). Because the amplitudes of both first and second evoked PS in the paired-pulse response were larger during the periods of K+ increase and Ca2+ reduction (Fig. 4B–D) than during the baseline recording (Fig. 4A), the decrease of IP could not be caused by a depression of evoked potentials. Therefore the decrease of IP could indicate a significant decrease of AHP duration by low-Ca2+ concentration. In contrast, IP durations after bursts induced by GABAA-receptor blockader PTX were significantly longer.
It has been shown in vitro that a PTX-induced epileptiform burst was followed by a long AHP because of a calcium-dependent potassium potential in hippocampal neurons (23–25). A similar phenomenon was also observed in this preparation. Application of PTX, 0.4 mM, with K+, 7.5 mM, over the surface of the exposed hippocampus, first, induced interictal bursts followed 30 min later by both interictal and ictal bursts (Fig. 5). Interictal activity in the CA1 pyramidal layer was characterized by a short positive wave superimposed with negative population spikes and followed by a long silent interval. The duration of the inhibition period with low MUA activity after the interictal bursts was 0.8 to 4.5 s (Fig. 5A). Ictal activity was characterized by a longer burst made up of several interictal bursts. A long IP duration (2.5–9.0 s) followed a long ictal burst (Fig. 5B). The frequencies of both PTX-induced interictal (0.28 ± 0.07 Hz) and ictal (0.09 ± 0.03 Hz) bursts were significantly lower than the frequency of EGTA-induced nonsynaptic double-spike activity (3.77 ± 1.28 Hz; p < 0.001; Table 1).
These data indicate that the decrease of GABAergic inhibition by PTX alone can induce bursts with long AHP inhibition intervals, thereby reducing the burst frequency. In contrast, high [K+]o combined with low [Ca2+]o can decrease both the GABAergic inhibition and the AHP inhibition, thereby shortening the inhibition duration after neural firing activity and producing the high burst frequency.
The major finding in this study is that elevating K+ concentration can transform typical low-Ca2+ nonsynaptic ictal events into persistent double-spike events in the hippocampal CA1 region in anesthetized rat when the synaptic transmission is suppressed. This double-spike activity repeats at a frequency of ∼4 Hz and can last >1 h, fitting the definition of status epilepticus activity that lasts >30 min with a frequency >2 Hz.
Mechanisms underlying the high-K+, low-Ca2+–induced SE-like epileptiform activity
A solution containing K+ (6 or 7.5 mM) with EGTA (5 mM) can induce ictal-like slow-wave activity with long intervals in the hippocampal CA1 region in vivo. This is due to the ability of high [K+]o and low [Ca2+]o solution to enhance neuronal excitability, reduce GABAergic inhibition, and increase the effects of ephaptic interactions and gap junctions (3,15,26), as well as enhance the persistent sodium current (27–29). Elevating K+ to higher concentrations such as 12 or 15 mM in the solution can generate long-lasting spike activity. This phenomenon could be explained by the fact that the increase of [K+]o can augment the frequency of the bursts. Studies have shown that a higher K+ concentration increases the frequency of interictal discharge because of the decrease of the potassium AHP current (2,20,30,31). The elevation of K+ concentration can also increase the frequency of nonsynaptic low-Ca2+ burst in vitro and in simulation models (21,22,32). Therefore in this study, the higher K+ concentration presumably decreased the interval between bursts and increased the frequency of nonsynaptic slow waves, resulting in continuous spike activity. Because of the significantly suppressed AHP activity by high-K+, low-Ca2+, as indicated by the decrease of inhibition duration after the paired-pulse response, the frequency of the nonsynaptic spike activity was significantly higher than PTX-induced synaptic interictal or ictal activity.
However, the continuous spike activity was dominated by double-spike bursts at a frequency of ∼4 Hz, with the duration of a single event significantly shorter than that of slow depolarization waves (Table 1). The slow depolarization waves with or without superimposed spikes in this in vivo observation were similar to previous nonsynaptic low-Ca2+ slow waves mostly observed in vitro with frequency <0.5 Hz and duration >1 s (15,28). In contrast, this double-spike continuous bursting activity has not been reported in extracellular field potential recordings in any in vitro studies. Nevertheless, both intracellular recording and modeling results showed that in Ca2+-free solution, CA1 pyramidal cells could fire spontaneously with bursts of three or two action potentials riding on a depolarization wave (32). These bursts were characterized by an interburst interval of ∼270 ms (equal to a frequency of 3.7 Hz) and intraburst interval (e.g., burst duration) of ∼50 ms, close to the frequency of 3.77 ± 1.28 Hz and duration of 38 ±39 ms of the double-spike activity in this study (Table 1). Intracellular studies and modeling results have shown that the appearance of spontaneous multiple spike activity (doublet or triplet) in CA1 pyramidal cells within low-Ca2+ and high-K+ extracellular concentrations results from the enhancement of persistent sodium current (32–34). Therefore the extracellular double-spike bursts observed in this study are presumably generated by intracellular spontaneous double or triple action potentials.
Furthermore, a modeling study suggests that elevating [K+]o in Ca2+-free solution first increases the spike number in a burst to three or more and then decreases the spike number to double or single spike bursts until firing is suppressed by a depolarization block caused by [K+]o concentration >9 mM (32). This result indicates that the double-burst activity is a transitional period between a longer burst and depolarization block that occurs at increasing K+ concentration. The narrow K+ concentration range might explain why continuous extracellular double-spike activity is difficult to observe in the in vitro preparations because low-Ca2+ nonsynaptic burst activity in vitro is prone to depolarization block by high K+ concentration >9 mM (21,22). In vivo, the threshold K+ concentration for depolarization block might be higher than in vitro because a solution with K+ concentration as high as 15 mM, in this study, only induced occasional appearance of depolarization block. Therefore continuous and spontaneous double-spike activity could be induced within a wider K+ concentration in vivo.
The actual concentrations of extracellular potassium deeper in the hippocampus were not measured in the in vivo preparations. The K+ diffusion could form a gradient along the depth of hippocampal tissue. Further experiments measuring [K+]o at different depths within the hippocampus after the perfusion of different high-potassium solutions onto the exposed hippocampal surface should reveal some of the mechanisms of this persistent spike activity.
Comparison with other models of status epilepticus
The unique pattern of the persistent double-spike nonsynaptic activity observed in this study is different from common SE models. Most SE animal models are generated either by chemical convulsants, such as pilocarpine, pentylenetetrazol, and kainic acid, or by electrical stimulation in the afferent pathways (35–40). These types of SE activity usually appear as tonic–clonic activity with changes in spike amplitudes and in spike intervals (11,41,38). The most basic mechanisms of SE development are hyperexcitability of neurons and loss of GABAergic inhibition (11,35,42). Both mechanisms are also prerequisites in the nonsynaptic persistent spike activity reported here. The more uniform pattern of the nonsynaptic double-spike activity could be generated by nonsynaptic mechanisms, such as the decrease of AHP current, thereby shortening intervals between bursts and the enhancement of persistent sodium current. The in vitro low-Ca2+ nonsynaptic ictal bursts induced when synaptic transmission is blocked also appear as a more regular pattern with uniform periodic duration than other synaptic ictal bursts (4,5).
The double-spike nonsynaptic sustained epileptiform activity induced by high potassium, low calcium in vivo is a novel epilepsy model. Taken together with previous observations that SE induces significant [K+]o increases and [Ca2+]o decreases in brain tissue (12–14), this model provides new evidence for the importance of nonsynaptic mechanisms in SE.
A type of SE–like activity was induced in the hippocampus in vivo by high K+, 12 mM, with EGTA, 5 mM, and characterized by double-spikes repeated at a frequency of ∼4 Hz lasting >1 h. The mechanism of this long-lasting activity might be the combination of high K+ and low Ca2+ decreased AHP activity, thereby augmenting the frequency of low-Ca2+ nonsynaptic depolarization bursts. This novel epilepsy model suggests that nonsynaptic mechanisms could be important in the generation of SE.
Acknowledgment: This work was supported by the National Institute of Neurological Disorders and Stroke grant RO1 NS-40785, U.S.A. It also was supported by the National Natural Science Foundation of China (no. 30570585). The recording probes were provided by the University of Michigan Center for Neural Communication Technology sponsored by NIH/NCRR grant P41 RR09754, U.S.A. We also thank Alicia Jensen for proofreading of the manuscript.