Modulation of a Ca2’ Signaling Pathway by GM1 Ganglioside in PC 12 Cells*

The effects of exogenous G M l ganglioside on depolar- ization and ligand-induced Ca2+ signaling were inves-tigated in PC 12 cells. Cellular responses to K+ depolar- ization and bradykinin application in control and GMI-treated cells were examined with respect to: 1) changes in the intracellular Ca2+ concentration ([Ca”],) measured using fura-2 fluorescence in single cells, and 2) changes in Ca2+-dependent protein kinase activity as assayed by two-dimensional phosphopeptide analysis of the site-specific phosphorylation of tyrosine hydrox- ylase. Pretreatment of cells with G M ~ (10 or 100 pM) enhanced K+ depolarization-stimulated increases in [Ca2+I1 and in 32P04 incorporation into tyrosine hydroxylase phosphopeptide T2, a Ca’+/calmodulin-de- pendent protein kinase I1 substrate. In contrast, G M 1 treatment had no effect on the transient increases in [Ca2+]i evoked by bradykinin or on bradykinin-induced increases in the site-specific phosphorylation of tyro- sine hydroxylase. The depolarization-induced and GM1-enhanced increases in [Ca2+Ii and T2 phosphorylation were prevented by removal of external Ca” or pretreatment with 1 p~ nitrendipine, suggesting that these increases result from Ca2+ entry through dihydropyr-idine-sensitive Ca2+ channels. The ability of exogenous gangliosides to potentiate increases in [Ca2+]i

nervous system damage (6,7). Second, gangliosides promote neurite outgrowth from a variety of clonal cell lines (Z), including PC12 cells (8,9), and in primary cultures of central nervous system neurons (10,11). The molecular basis of the neuritogenic effects of exogenous gangliosides applied either alone or in combination with growth factors such as nerve growth factor (NGF)' is unknown.
Recent work has focused attention on gangliosides as modulators of membrane-associated protein kinases and transmembrane signal transduction events. For example, gangliosides GMl and GY3 inhibit growth factor receptor tyrosine kinase activity (12, 13) and several ganglioside species can inhibit the activity of the Ca2+/phospholipid-dependent protein kinase (C kinase) (14). In addition, gangliosides can stimulate a Ca2+-dependent protein kinase activity (15) and several other kinases associated with brain (16,17). In PC12 cells, GMl together with NGF stimulates a Ca2+-dependent protein kinase that is not normally activated in the NGF signal transduction cascade (18). Thus, the modulation of Ca2+-dependent signaling events by gangliosides may underlie their neuritogenic properties in neural cells. The relationships between ganglioside effects on Ca2+ signaling pathways and changes in intracellular calcium concentration ([Ca"'];) known to be important in the control of neurite extension and growth cone motility (19) have not been examined.
To explore the basis of ganglioside actions on Ca2+ signaling pathways in PC12 cells, the effects of exogenous gangliosides on [Ca2+]; were examined during K+ depolarization, which increases [Ca2+]; by promoting Ca2+ influx (20-22), and during bradykinin treatment, which causes Ca2+ release from internal stores (23, 24). Here we show that short term treatment of PC12 cells with micromolar concentrations of GMl ganglioside can enhance depolarization-induced entry of extracellular Ca2+ through nitrendipine-sensitive Ca2+ channels. The increase in [Ca2+Ii after ganglioside treatment is sufficient to modulate Ca2+-dependent processes in these cells, as GMl treatment enhances K+ depolarization-induced increases in the activity of a Ca2+-dependent protein kinase. In contrast, ganglioside treatment had no effects on the activation by bradykinin of signaling pathways dependent upon the release of Ca2+ from intracellular, inositol 1,4,5-trisphosphate (IP3)sensitive stores. The ability of exogenous gangliosides to enhance depolarization-induced entry of Ca2+ may underlie their diverse trophic and neuritogenic effects on neuronal cells. 24789 from ICN Radiochemicals. Fura-2AM and 4Br-A23187 were purchased from Molecular Probes (Eugene, OR). Gangliosides GD~. and

G~l b
were a gift of R. K. Yu (Dept. of Biochemistry and Molecular Biophysics, Medical College of Virginia). Omega conotoxin was a gift from Dr. P. Adams (Howard Hughes Institute, Stony Brook, NY). All other reagents and chemicals were of the highest grade commercially available.
Cell Culture Conditions"PCl2 cells (25) were cultured in DMEM containing 10% horse serum and 5% fetal calf serum (JRH, Lenexa, KS). Cells were incubated at 37 "C in an atmosphere of 90% air, 10% COS and were grown to a density of 2 X lo6 cells/lOO-mm dish and passed every 5-7 days.
Fura-2 Cell Loading Conditions-PC12 cells were cultured in 10% horse serum and 5% fetal calf serum in DMEM on poly-L-lysine/ laminin-coated glass coverslips (22 X 22 mm) at a density of 1-2 X lo' cells/coverslip. Measurements were made on cells 2-3 days after plating. Cells were incubated for 2 h at 37 "C in serum-free DMEM with or without G M~, and the cells were loaded with the membranepermeable form of the Ca2+ indicator fura-2 (26) by incubation with 1-2.5 pM fura-2 acetoxymethyl ester in Hepes-buffered DMEM for 30 min at 25 "C. Loading was performed at 25 "C to minimize compartmentalization of the trapped dye. Cells were washed several times with Hepes-buffered DMEM and were incubated in media without fura-2AM for an additional 15 min to allow for complete deesterification of the dye.
Microfluorimetric Ca2+ Measurements-The equipment, microscope modifications, and software were essentially as described (27). Fluorescence intensities were measured from single cells after providing illumination with alternating excitation wavelengths of 360 and 380 nm. Fluorescence emitted at 505 nm was monitored by photon counting, and [Ca2+]; was calculated as described (26,28). Calibration of fura-2 fluorescence was done using PC12 cells loaded with fura-2 by intracellular patch pipette, and the calibration constants used were those as described (28). To verify the Ca2+ sensitivity of the fura-2 fluorescence, Mn2+ was loaded into fura-2-containing cells using the Ca2+ ionophore 4Br-A23187. The resulting fluorescence signal after Mn2+ loading was reduced by 92% (n = 4). For recording, cells were kept at 25 "C and bathed in Hepes-buffered saline, pH 7.2, containing 5.4 mM KC1, 120 mM NaC1, 20 mM Hepes, 25 mM glucose, 1 mM NaH2P0,, 1.5 mM MgS04, and 5 mM CaC12. Cells were depolarized by bath exchange of normal Hepes-buffered saline with saline containing concentrations of KC1 between 20 and 60 mM with a corresponding reduction in NaCl to maintain isosmolarity. Responses were quantitated by measuring the peak or maximal rise in [Ca2+]; evoked by each treatment. This method was chosen since the kinetics of the responses of single cells varied due to the time lag required for a complete change in bathing solutions.
Zn Situ Protein Kinase Actiuity Assay-Zn situ 32P04 labeling of proteins, peptide mapping procedures, and quantitation of 32P0, incorporation into tyrosine hydroxylase phosphopeptides were as described (18, 29). For labeling experiments, PC12 cells were grown to lo6 cells/60-mm tissue culture dish. Cells were incubated in a low phosphate DMEM (0.09 mM NaH2P04) containing 2.0 mCi/ml 32P04 for 2 h at 37 "C. Labeling was terminated by removing the culture media and resuspending cells in ice-cold lysis buffer (0.01 M NaHzP04, pH 7.1, 0.1 M NaF, 0.24 M sucrose, and 0.01 M EDTA). Cells were lysed in the same buffer containing 1.0% Nonidet P-40, and nonnuclear proteins were extracted and subjected to one-dimensional electrophoresis on 10% SDS-polyacrylamide slab gels (30). 32P-Labeled tyrosine hydroxylase was visualized by autoradiography, extracted from the gel, and subjected to two-dimensional tryptic peptide mapping.

Effects of G M l Pretreatment on K+ Depolarization-induced
Increases in [Ca2+li"The effects of K+ depolarization on [Ca2+Ii were first characterized in untreated PC12 cells. As shown in Fig. 1 and Table I (Fig. 1B). Similarly, if the cells were depolarized in Ca2+-free saline, there was no depolarization-induced increase in [Ca2+]i (data not shown). The pathway of Ca2+ entry during K+ depolarization was sensitive to the dihydropyridine family of channel blockers (Fig. IC) as application of 1 p t~ nitrendipine for 5 min prior to depolarization inhibited the peak rise in [Ca2+Ii by 87 f 5% (mean f S.E., n = 5). Thus, as reported previously (21), K+ depolarization of PC12 cells results in the rapid and sustained elevation of [Ca2+Ii due to the entry of extracellular Ca2+ through a dihydropyridinesensitive pathway.
To determine whether GM1 can modulate K+ depolarization induced increases in [Ca2+]i, the peak increases in [Ca2+Ii were compared between cells from untreated and GMl-treated cultures as described under "Experimental Procedures." Since K+ depolarization-induced increases in [Ca2+Ii display considerable variability at [K' ], 2 40 mM ( Table I), G M~ effects on [Ca2+Ii were examined at a submaximal level of [K'], (30 mM), such that peak increases in [Ca2+]i could be easily quantitated. Fig. 2A shows [Ca"Ii measurements taken from fura-2-loaded cells treated in the absence or presence of G M~ (10 or 100 p~) for 2 h immediately prior to dye loading. Pretreatment of cells with GM1 enhanced the increase in [Ca2+Ii caused by 30 mM KC1 compared to untreated control cells (Table I) These differences cannot be attributed to effects on the resting [Ca2+]i, as ganglioside pretreatment did not change the resting [Ca2+]i compared to untreated cells (Table I). These data demonstrate that GMl pretreatment causes a shift in the mean peak increase in [CaZ+li in response to a depolarization with 30 mM KC1 without significantly altering either the resting level of [Ca2+Ii or the maximal level of [Ca2+Ii that can be achieved.
The results shown in Fig. 2B reveal a broad range in the magnitude of the increases in [Ca2+Ii evoked in response to 30 mM KC1 in both treatment groups, suggesting that these cells are heterogeneous in terms of their ability to respond to a given depolarization. The cell-to-cell variability in these responses was examined by first comparing responses between cells of the same treatment group and then by comparing the responses of a single cell after repeated depolarization with 30 mM KC1. Increases in [Ca2+Ii were determined in several cells from the same culture dish by taking [Ca2+Ii measurements from single cells during depolarization with 30 mM KC1 for 90s with 3-5 min intervals between selection of new cells. These results demonstrate that although an individual cell responds in a highly reproducible manner to a given depolarization, comparison of the responses of cells within the same culture dish shows a significant degree of variability. Thus, comparison of responses between cells from GMl-treated or untreated cultures would likely reveal a degree of overlap in the peak increases in [Ca2+Ii in response to a depolarization and this overlap may account, in part, for the lack of effect of GM1 pretreatment in some cases.  Fig. 3 shows [Ca2+Ii measurements taken from fura-2-loaded cells during depolarization with 30 mM KC1 followed by washout with normal saline and application of either normal saline (Fig. 3A) or saline containing 10 ( Fig.  3 B ) or 100 pM G M~ (Fig. 3C) for 10 min. Comparison of the increases in [Ca2+Ii revealed that GMl application potentiated the response to 30 mM KC1 by =30-35% when compared to the response obtained prior to ganglioside treatment in the same cell (Table 11). Enhanced responses to GMl were observed in all cells examined. No differences in depolarizationinduced increases in [Ca2+Ii were observed in cells treated for 10 min with normal saline (Fig. 3A and Table 11). Treatment of cells with the diand trisialogangliosides GD~. and GTlb for 10 min caused a similar enhancement in K+ depolarizationinduced increases in [Ca2+]i in these assays, whereas treatment with other charged or uncharged glycolipids (sulfatide and galactocerebroside) had no effect on depolarization-induced Ca2+ entry (data not shown). These results demonstrate  that short term treatments with exogenous ganglioside are sufficient to enhance depolarization-induced increases in [Caz+li when measured in the same cell. We examined the effects of nitrendipine on the GM1-enhanced entry of Ca2+. Cells were initially depolarized with 30 mM K+, treated with 100 pM GMl for 10 min, and then depolarized again with 30 mM K+ to measure the gangliosideinduced enhancement of [Ca2+Ii. Following a 5-min treatment with nitrendipine, the same cell was again depolarized with 30 mM K+. As shown in Fig. 4, treatment with nitrendipine prevented the increases in [Ca"]i due to depolarization. In 8 separate experiments, the increase in [Ca2+Ii in GMl-treated cells was inhibited by 90 & 4% (mean & S.E.) by nitrendipine. In this same experimental paradigm, w-conotoxin (10 p~, 5 min) was ineffective in blocking the depolarization-induced increases in [Ca2+Ii (data not shown). Thus, the enhanced response to K+ depolarization in (&-treated cells is pharmacologically indistinguishable from the responses in untreated PC12 cells.

Effects of G M~ Pretreatment on K+ Depolarization-induced
Phosphorylation of Tyrosine Hydroxylase-The results presented above demonstrate that GMl treatment can potentiate depolarization-induced increases in [Ca2+Ii. To determine whether ganglioside effects on [Ca2+]i are sufficient to modulate Ca2+-dependent processes in the cells, we examined the site-specific phosphorylation of tyrosine hydroxylase using the same treatment paradigms as in the fura-2 measurements. Short term depolarization with K+ (30 s) results in the specific phosphorylation of serine 19 within phosphopeptide T2 (31). Peptide T2 has been identified as a specific substrate for Ca2+/calmodulin-dependent protein kinase I1 (CaM kinase 11) in vitro (31, 32); thus, changes in the phosphorylation state of peptide T2 can be used to assess relative changes in Ca2+dependent protein kinase activity in intact cells.
Exposure of PC12 cells to 30 mM KC1 for 30 s caused a specific increase in the phosphorylation of peptide T2 compared to untreated, control cells ( Fig. 5 and Table 111). The phosphorylation states of peptides T1, T3, and T4 were not affected by K+ depolarization. Pretreatment of the cells with GM1 for 2 h potentiated K+ depolarization-induced increases in peptide T2 phosphorylation when compared to treatment with K+ alone (Fig. 5 and Table 111). Similar results were obtained with a 10-min GMl treatment (data not shown). Treatment with GM1 alone was previously shown to have no effects on the phosphorylation of tyrosine hydroxylase phosphopeptides (18). The ability of G M~ pretreatment to potentiate K+ depolarization-induced increases in peptide T2 phosphorylation is not likely to be due to an inhibition of phosphatase activity as the rate of loss of 32P0, from pulse-labeled tyrosine hydroxylase did not differ between K+-treated and GM1/K+-treated cells (data not shown). Treatment of the cells with 1 p~ nitrendipine for 15 min prior to depolarization

Effects of G,, on K+ depolarization and bradykinin-induced phosphorylation of tyrosine hydroxylase phosplwpeptides
Data are the means f S.E. of 2-4 separate experiments and are expressed as percent of untreated, control PC12 cells. Relative 32P04 incorporation into individual phosphopeptides was determined by scanning densitometry as described under "Experimental Procedures.'' Phosphate incorporation into peptide T4 was used as an internal reference to normalize for differences in labeling between samples (18, 29). Cells were pretreated with G M~ ganglioside at 10 p M for 2 h prior to treatment with the indicated agents for the indicated times.
Relative 32P0, incorporation prevented the depolarization-induced increases in the phosphorylation of peptide T2 in both GMl-treated and control cells (Fig. 5). Removal

Effects of G , , Pretreatment on Bradykinin-induced Ca2'
Signaling-The data presented above suggest that ganglioside treatment can modulate the entry of extracellular Ca2+ through dihydropyridine-sensitive Ca2+ channels after K+ de-polarization and that the enhanced levels of [Ca2+]i achieved after ganglioside treatment can modulate physiological processes within PC12 cells. To examine whether GMl can modulate Ca2+-dependent signaling events regardless of the mode whereby [Ca'+]i is increased, effects on increases in [Ca2+Ii and on protein kinase activity were assayed after exposure of cells to bradykinin. Bradykinin evokes a rapid and transient increase in [Ca2+Ii in PC12 cells via Ips-mediated release of Ca2+ from internal stores and also causes a delayed, slower increase in [Ca2+Ii by promoting Ca2+ entry through an unknown pathway (21, 24).
The effects of GM1 treatment on [Ca2+Ii were compared after K+ depolarization and bradykinin application to fura-2loaded cells as shown in Fig. 6. The cells were treated consecutively with 30 mM KCl, 40 mM KCl, and 100 nM bradykinin with a 3-min washing interval between treatments. G M~ pretreatment had no effect on bradykinin-induced increases in [Ca2+Ii. In 5 separate experiments, the increase in [Ca2+Ii after bradykinin treatment in control, untreated cells was 508 f 122 nM (mean & S.E.). For cells that were pretreated for 2 h with 10 p~ G M~, the peak increase in response to bradykinin was 488 f 124 nM (mean f S.E.). In the experiment shown in Fig. 6, G M~ enhanced the rise in [Ca2+Ii evoked by depolarization with 30 mM KCl, but had no significant effect on the response to 40 mM KC1. These results demonstrate that GM pretreatment can act differentially to enhance the peak increase in [Ca2+Ii produced in response to a submaximal K+ depolarization while having no effect on the change in [Ca2+Ii induced by application of bradykinin. Thus, the effects of G M~ may be limited to a potentiation of increases in [Ca2+]i occurring via Ca2+ entry rather than by agonist-induced Ca2+ release from IP3-sensitive intracellular stores.
Treatment of PC12 cells with bradykinin increases the activity of CaM kinase I1 within 1 min of application (33) and increases the phosphorylation of tyrosine hydroxylase phos- phopeptides T2 and T3 when measured after 10 min of exposure to the drug (34). To investigate the effects of GMl treatment on bradykinin-induced increases in Ca2+-dependent protein kinase activity, the site-specific phosphorylation of tyrosine hydroxylase was assayed after 1-and 5-min exposure to the ligand. The effects of GMl were determined in combination with 100 nM bradykinin, a dose determined to be submaximal for bradykinin-induced increases in the phosphorylation of tyrosine hydroxylase (data not shown). As shown in Table 111, pretreatment of cells with 10 p~ GM1 for 2 h did not alter the site-specific phosphorylation of tyrosine hydroxylase produced by either a 1-or 5-min application of bradykinin. Thus, in contrast to the stimulatory effects of GMl on Caz+-dependent protein kinase activity during depolarization, GMl does not enhance Ca2+-dependent protein kinase activity after bradykinin treatment.

DISCUSSION
The results presented in this study demonstrate that short term exposure of PC12 cells to GM1 ganglioside leads to a potentiation of the .cellular responses to depolarization with elevated [K+],. Depolarization-induced increases in [Ca2+]i resulting from Caz+ influx through dihydropyridine-sensitive Caz+ channels were enhanced significantly in GMl-treated cells compared to control cells. Cellular responses to agents that raise [CaZ+li via promoting the release of Ca2+ from internal stores were unaffected by G M~ pretreatment. The enhanced levels of [Ca"Ii achieved after ganglioside treatment may mediate the neuritogenic and neurotrophic properties of gangliosides.
The ability of gangliosides to amplify increases in [Ca2+Ii during depolarization provides a basis for the potentiating effects of gangliosides on a Ca2+-dependent protein kinase activity in PC12 cells. CaM kinase I1 is a likely candidate for this kinase since it can phosphorylate tyrosine hydroxylase in vitro at serine 19 (32), which is contained within phosphopeptide T2 (31). Depolarization with varying levels of [K+l0 brings about a graded phosphorylation of tyrosine hydroxylase phosphopeptide T2, which is maximal at [K+l0 2 50 mM (18). A similar dose-response relationship was seen in the [Ca2+Ii measurements reported here. Both types of responses were enhanced when ganglioside pretreatment was combined with sub-maximal depolarization. The simplest explanation of these data is that ganglioside incorporation leads to increased Ca2+ entry after sub-maximal depolarization and that the increased levels of [Ca2+]i achieved by ganglioside treatment enhance the activation of CaM kinase 11.
Greater than 90% of the increase in [Ca2+Ii in untreated or GMl-treated cells was abolished by nitrendipine pretreatment implicating L-type Ca2+ channels in the responses measured here. The observations that L-type Ca2+ channels appear to be responsible for >90% of K+ depolarization-induced Ca2+ entry (35) and carry the majority of depolarization-induced Ca2+ current (36) support with this hypothesis. Additional support for this hypothesis comes from our analysis of the effects of GMl pretreatment on bradykinin-induced Ca2+ signaling events. Bradykinin binding to PC12 cells causes a rapid and transient increase in [Ca2+]i as a result of IPSinduced release of Ca2+ from intracellular stores and of Ca2+ entry through a nitrendipine-insensitive pathway (21, 24). Treatment of cells with GM1 ganglioside had no effect on levels of [Ca2+Ii or on Ca2+-dependent protein kinase activity stimulated in response to bradykinin. GM1 also also had no effect on tyrosine hydroxylase phosphorylation when assayed in combination with the nicotinic cholinergic agonists carbachol and l,l-dimethyl-4-phenylpiperazinium and or with muscarine, all of which stimulate phosphatidylinositol turnover and cause increases in [Ca2+Ii in part via IPS-induced Ca2+ release (data not shown). Thus, it is likely that G M~ exerts its effects on cellular physiology by an interaction with a Ca" signaling pathway involving influx of Ca2+ through dihydropyridine-sensitive, L-type Ca2+ channels.
We can only speculate on how this specificity of site of action is achieved. The incorporation of negatively charged gangliosides into the outer leaflet of the plasma membrane may modestly increase the negative electrostatic or surface potential (37,38). This increase in surface potential could effect the functioning of voltage-dependent ion channels in two ways. First, the voltage dependence of gating of channels could be shifted towards more negative potentials (39,40), such that a sub-maximal depolarization would lead to more channel openings in ganglioside-treated cells as compared to control, untreated cells. Second, the increased surface potential could concentrate cations, including Ca2+, in the extracellular space immediately adjacent to the plasma membrane and increase the driving force of current flow through open channels. Consequently, a greater amount of Ca2+ would enter the cell. Alternatively, the added gangliosides might associate specifically with L-type Caz+ channels via interactions between the channel and the ganglioside molecule and thereby alter channel function. Whatever the nature of the interactions between added ganglioside and L-type Caz+ channels, the effects we have observed indicate that L-type Caz+ channels are a major cellular target for ganglioside action.
While we cannot directly rule out the possibility that exogenous gangliosides interact with the active cellular processes that regulate calcium homeostasis, such as extrusion by Ca2+ ATPases or sequestration into organelles, it seems unlikely for several reasons. First, ganglioside treatment does not alter resting [Ca2+Ii; any effect of exogenous gangliosides on membrane pumps would require a voltage-dependent interaction with those pumps since gangliosides increase [Ca2+Ii only after depolarization. Second, the majority of the exogenously added gangliosides associate with the outer leaflet of the plasma membrane and less than 5% of the added ganglioside reaches the cytoplasm of the cell (41). Thus, it is unlikely that exogenously added gangliosides interact with intracellular organelles. A complete determination of the site of action of exogenous gangliosides will require either wholecell or single-channel recordings from Ca2+ channels in PC12 cells, an approach that is beyond the scope of this study.
Previous studies from our laboratory have demonstrated that GM1 treatment can stimulate a Ca2+-dependent protein kinase in NGF-treated PC12 cells (18). Since GM1 can potentiate K+ depolarization-induced increases in [Caz+Ii, we examined whether Gllll might effect [Ca"]i in combination with NGF. Using fura-2 measurements in single cells, we did not detect an increase in [Ca2+]i in response to NGF (50 ng/ml) in either control or &-treated cells ( n = 22). These findings are consistent with a previous account which showed no effect of NGF on [Ca2+Ii (42) but in are in disagreement with the work of others (43,44) who found small changes in [Caz+Ii in single cells (x100 nM) and in populations of cells (25-50 nM). This discrepancy may be due to differences in the sensitivity of the method used for measuring [Ca2+Ii or in assay conditions, or to variations existing among PC12 cells from different laboratories. Although the effects of NGF on [Ca? remain controversial, a subset of NGF actions appear to require Ca2+ influx (45,46). Although the [Caz+]i measurement data reported here do not support a mechanism in which gangliosides potentiate Ca2+ entry in the presence of NGF, the synergistic action of these treatments on the activity of a CaZ+-&pendent protein kinase nevertheless requires the entry 17. Cimino, M., Benfenati, F., Farabegoli, C., Cattabeni, F., Fuxe, K., Agnati, Of extracellular ca2+ through nitripendine-sensitive ca2+ 18. Hilbush, B. S., and Levine, J. M. (1991) Proc. Natl. A C~. Sci. U. S. A. 8 8 , channels.

5616-5620
Calcium appears to be a key second messenger involved in 20. Meldolesi, J., Huttner, W. B., Tsien, R. Y., and Pozzan, T. Laughlin for their comments on this manuscript.