- Email: [email protected]
Can a population of suprachiasmatic nucleus neurons with different period lengths produce a stable circadian rhythm?
BRAIN RESEARCH ELSEVIER
Brain Research 670 (1995) 333-336
Can a population of suprachiasmatic nucleus neurons with different period lengths produce a stable circadian rhythm? Y o n a Bouskila, F. E d w a r d D u d e k * Department of Anatomy and Neurobiology, Colorado State University, Fort Collins, CO 80523, USA Accepted 16 November 1994
Abstract
The firing rate of a population of SCN neurons in vivo exhibits stable circadian oscillations, but the period length of individual neurons is not known and may be different or similar to the population rhythm. To address this question we used published data from Bos and Mirmiran [Brain Res., 511 (1990) 158-162] that reported different period lengths and amplitudes for individual neurons recorded in explant cultures of the SCN. We reconstructed the individual rhythms for several cycles, calculated the population rhythm, and then tested its stability. The period and amplitude of the rhythm of groups of neurons with different period lengths were unstable. Furthermore, the stability of the rhythm was reduced as the number of sampled neurons increased. These results suggest that the stable circadian rhythm reported for neuron populations in the intact SCN emerges from the identical period length of individual neurons. The possible intercellular interactions in the SCN that may underlie the stable circadian rhythm are discussed.
Keywords: Circadian rhythm; Suprachiasmatic nucleus; Biological clock; Synchronization; Electrophysiology
The averaged firing rate of neuronal populations in the SCN exhibits circadian oscillations (i.e. close to 24 h) with a stable period and amplitude for at least 35 days in vivo [4]. Even in acute SCN slices, the period appears to be stable [6]. Mathematical models have suggested that a precise aggregate rhythm can be produced by weakly coupled oscillators with faster or slower native rhythms [10], but the period length of individual neurons in the intact SCN is not known. The data showing a stable circadian rhythm of neuronal activity have been based on the averaged firing rate of the population. In slice preparations, these results have been obtained by calculating an average of relatively brief (i.e. 2 min) single-unit recordings from many SCN neurons and compiling their average firing rate in 2 h bins [6]. For in vivo studies, multiple-unit recordings have been obtained from a population of SCN cells simultaneously, which is essentially an averaging process. Individual ceils from explant cultures of the SCN, however, exhibit different period lengths that range from 16 to 32 h [1]. This preparation is the only one
* Corresponding author. Fax: (1) (303) 491-7907. 0006-8993/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved SSDI 0 0 0 6 - 8 9 9 3 ( 9 4 ) 0 1 3 5 6 - X
that allows continuous recordings of single neurons for long periods. One possibility consistent with the data of Bos and Mirmiran [1] is that the stable circadian rhythm obtained in vivo and in brain slices is a product of the averaging technique, while individual ceils exhibit rhythms that deviate considerably from 24 h. This is possible only if neurons with a 24-h period dominate the population by their higher rhythm amplitudes and prevalence. To test this hypothesis, we used a cosine function to reconstruct mathematically the rhythms of single neurons with known periods and amplitudes based on the published physiological data of Bos and Mirmiran [1] from explant cultures of the SCN. The rhythms from single cells were averaged, and the product of the population rhythm was then evaluated for the precision and stability of its 24 h period over several cycles. Reconstruction of firing rate rhythm was based on long-term recordings (36-66 h) from cultured SCN cells reported by Bos and Mirmiran [1]. From each explant a single cell was recorded, except for two cases in which a few ceils were recorded and analyzed simultaneously. In all cases significant periodicity was ob-
Y. Bouskila, F.E. Dudek / Brain Research 670 (1995) 333-336
334
A
served. The values of maximum firing rate (Fmax) , minimum firing rate (Fmin) and period were used to reconstruct individual rhythms of 17 cells using a simple cosine function (Eqn. 1), with an increment of 1 h for up to 200 h. This function was chosen because the original data were analyzed by fitting the waveform of firing rate to a cosine function [1]. The mean firing rate of the population at time t(F(t)) was calculated with Eqn. 2 using all 17 cells or four random samples of 4, 8 and 12 cells.
N=17
10
8
"~ L,
4 2
B
N=12
12 10
"6
8
L,-
6
•
•
•
-
"
• •
:
v
•
-:
V
•
'It
•
F(,)i = M + A* cos(27r t/~- + ~b)
A
. .
qt9
N
~2
(2)
flu) = 1/n E Fu,i i
4
C 10
6 4 2
D
N=4
8
6 4 T
'C_ 2 L~ i
L
I
I
I
t
I
I
I
I
0
20
40
60
80
100
120
140
160
180
Time
200
(h)
Fig. 1. Mean firing rate rhythm of different cultured SCN population sizes. A: averaged rhythm produced by 17 cells. B-D: averaged rhythms produced by different population sizes sampled randomly from the same pool of 17 cells. The number of cells averaged for each rhythm is denoted above it. Notice the increasing instability of rhythm amplitude and period as the population size was increased.
A
B
1.0
1.0
o "C.
t
t
0.8 0.6
l
F = firing rate (Hz), M = (Fmax + Fmi,)/2 (Hz), A = amplitude ( M - Fmi.) (Hz), t = time (h), ~"= period (h), n = number of cells, ~/, = phase (assumed to be 0 for simplicity and in order to maximize synchronization at t = 0). The resulting rhythms were further analyzed by calculating the coefficient of variation (C.V.) of the period and amplitude of the population oscillations to evaluate rhythm stability. Period length was measured as the time between adjacent peaks, and rhythm amplitude as the difference in frequency between peak and adjacent trough. The mean firing rate of a population of cultured SCN neurons with different periods is depicted for 200 h in Fig. 1A. This result is based on the maximal and minimal firing rate and period length of all 17 neurons [1], and assumes a perfect cosine function for the oscillations of each cell (see Eqn. 1). The rhythm is unstable in amplitude and period length. Periods ranged from 15.5 to 26.5 h (mean = 21.9, S.D. = 3.1)
N=8
8
0
(1)
~ 0.8 2
0.6
E "~
o
0.4
i
0.2
0.4
c~ 0.2
+
0
+
0.0
0.0 4
i
i
i
i
i
i
i
r
i
6
8
10
12
14
4
6
8
10
Number
of
cells
Number
of
i
12
14
cells
Fig. 2. Stability of rhythm as a function of number of cells sampled. A: coefficient of variation (C.V.) of amplitudes in compound rhythms produced by different population sizes. B: coefficient of variation of the periods in compound rhythms produced by different population sizes. Each point represents an average of four random samples, and the vertical bars represent standard errors of the mean.
Y. Bouskila, F.E. Dudek /Brain Research 670 (1995) 333-336
and amplitudes ranged from 0.1 to 4.7 Hz (mean = 1.9, S.D. = 1.2). To test the possibility that the instability may be the result of the small number of cells, we reduced the sampled population. As the number of cells was decreased (Fig. 1B-D), a progressive increase in the regularity of the rhythm amplitude and period was introduced. The average of four cells, selected randomly from the original population of 17, showed regular oscillations with a slight change in rhythm amplitude (Fig. 1D). Quantitative analysis of rhythm stability produced by different population sizes confirmed that the averaged rhythm became progressively more perturbed with increasing number of cells. Accordingly, the coefficients of variation of period length and of amplitude showed a tendency to increase with larger population sizes (Fig. 2A and B). The relative contribution of cells with different period length to the population rhythm was also tested. The deviation of individual cells from a period of 24 h was not correlated with the rhythm amplitude (r = - 0 . 1 2 , df = 15, P > 0.1). This result indicates that there is a random relation between the period length of a cell and its contribution to the average rhythm of the population. Based on data obtained from SCN neurons in a cultured explant, the only preparation that has so far allowed sufficiently long recordings from individual neurons, it appears that a simple average of activity from oscillators with different periods and amplitudes produces an unstable rhythm. This contrasts with the stable circadian rhythm reported in vivo. The clear difference in rhythm stability was present despite the fact that in both cases the recordings were extracellular and the rhythms were obtained by averaging firing rates. A strong period component of 24 h rhythm was detected using periodogram analysis in acute brain slices by sampling as few as 189 SCN neurons [3], while a very stable circadian rhythm was observed for over a month from a small population of SCN cells in vivo [4]. In the Bos and Mirmiran [1] study, only 17 cells were described, but our results demonstrated that an increase in sampling size only further perturbed the population rhythm (Figs. 1 and 2). Therefore, the differences in rhythm stability cannot be explained by different sample size. Oscillators with different periods and without interactions may produce a stable averaged rhythm, if those with a large deviation from 24 h contribute less, while precise 24 h oscillators dominate the rhythm by higher amplitudes and prevalence. The unstable rhythm we obtained from averaging the firing rate of individual SCN neurons, in addition to the random relation between period length and amplitude of these cells, suggest that individual neurons in the intact SCN have a
335
uniform 24 h period. Uniformity of period length may be achieved by intercellular communication between the neuronal oscillators in the SCN, which may have been disturbed in culture. Different mathematical models of weakly interacting oscillators predict that, even though individual oscillators may have variations of 10% to 20% in their native cycle, the group rhythm exhibits a stability several orders of magnitude greater than that of any constituent oscillator [10]. Both synaptic and non-synaptic mechanisms of synchronization are present in the SCN, and may play a role to produce the highly stable circadian rhythm observed. Physiological and immunocytochemical evidence suggests that G A B A is the major inhibitory neurotransmitter in the SCN. Axons of many SCN neurons terminate within the nucleus while most cell bodies and half of the terminals contain G A B A [8]. Furthermore, preliminary results suggest that glutamate application to the SCN during whole-cell voltage-clamp recordings results in an increase in GABA-mediated inhibitory postsynaptic currents [7]. Taken together, these results suggest a role for GABA in local circuit interactions. SCN neurons, however, can also loosely synchronize their firing activity in the absence of detectable chemical synaptic transmission via an unknown mechanism [2]. In SCN cultures, communication between glial cells has been reported as waves of elevated intracellular calcium moving from cell to cell [9]. A recent study has shown that stimulation of a single astrocyte in a mixed glial-neuronal culture from the forebrain triggers a wave of cytosolic calcium increase in neighboring neurons [5]. This glial-neuronal interaction was unaffected by synaptic transmission blockers or the Na+-dependent action potential blocker, tetrodotoxin. Some degree of coordination between oscillators with different periods seems to be necessary to produce the precise and stable circadian activity rhythm of the population seen in the intact SCN. The synaptic and non-synaptic mechanisms that are present in the SCN may provide the necessary neuronal coupling between the individual oscillators in the intact SCN.
We thank Dr. R.D. Traub for comments on the manuscript. This work was supported by a grant from the United States Air Force Office of Scientific Research.
[1] Bos, N.P.A. and Mirmiran, M., Circadian rhythms in spontaneous neuronal discharge of the cultured suprachiasmatic nucleus, Brain Res., 511 (1990) 158-162. [2] Bouskila, Y. and Dudek, F.E., Neuronal synchronization without calcium-dependent synaptic transmission in the hypothalamus, Proc. Natl. Acad. Sci. USA, 90 (1993) 3207-3210. [3] Green, D.J. and Gillette, R., Circadian rhythm of firing rate
336
Y. Bouskila, F.E. Dudek / Brain Research 670 (1995) 333-336
recorded from single cells in the rat suprachiasmatic brain slice, Brain Res., 245 (1982) 198-200. [4] Inouye, S.T. and Kawamura, H., Characteristics of a circadian pacemaker in the suprachiasmatic nucleus, J. Comp. Physiol., 146 (1982) 153-160. [5] Nedergaard, M., Direct signaling from astrocytes to neurons in cultures of mammalian brain cells, Science, 263 (1994) 17681771. [6] Prosser, R.A. and Gillette, M.U., The mammalian circadian clock in the suprachiasmatic nuclei is reset in vitro by cAMP, J. Neurosci., 9 (1989) 1073-1081.
[7] Strecker, G.J. and Dudek, F.E., Local synaptic circuits in the suprachiasmatic nucleus, Soc. Neurosci. Abstr., 20 (1994) 1439. [8] van den Pol, A.N. and Dudek, F.E., Cellular communication in the circadian clock, the suprachiasmatic nucleus, Neuroscience, 56 (1993) 793-811. [9] van den Pol, A.N., Finkbeiner, S.M. and Cornell-Bell, A.H., Calcium excitability and oscillations in suprachiasmatic nucleus neurons and glia in vitro, J. Neurosci., 12 (1992) 2648-2664. [10] Winfree, A.T., Biological rhythms and the behavior of populations of coupled oscillators, J. Theor. Biol., 16 (1967) 15-42.
Brain Research 670 (1995) 333-336
Can a population of suprachiasmatic nucleus neurons with different period lengths produce a stable circadian rhythm? Y o n a Bouskila, F. E d w a r d D u d e k * Department of Anatomy and Neurobiology, Colorado State University, Fort Collins, CO 80523, USA Accepted 16 November 1994
Abstract
The firing rate of a population of SCN neurons in vivo exhibits stable circadian oscillations, but the period length of individual neurons is not known and may be different or similar to the population rhythm. To address this question we used published data from Bos and Mirmiran [Brain Res., 511 (1990) 158-162] that reported different period lengths and amplitudes for individual neurons recorded in explant cultures of the SCN. We reconstructed the individual rhythms for several cycles, calculated the population rhythm, and then tested its stability. The period and amplitude of the rhythm of groups of neurons with different period lengths were unstable. Furthermore, the stability of the rhythm was reduced as the number of sampled neurons increased. These results suggest that the stable circadian rhythm reported for neuron populations in the intact SCN emerges from the identical period length of individual neurons. The possible intercellular interactions in the SCN that may underlie the stable circadian rhythm are discussed.
Keywords: Circadian rhythm; Suprachiasmatic nucleus; Biological clock; Synchronization; Electrophysiology
The averaged firing rate of neuronal populations in the SCN exhibits circadian oscillations (i.e. close to 24 h) with a stable period and amplitude for at least 35 days in vivo [4]. Even in acute SCN slices, the period appears to be stable [6]. Mathematical models have suggested that a precise aggregate rhythm can be produced by weakly coupled oscillators with faster or slower native rhythms [10], but the period length of individual neurons in the intact SCN is not known. The data showing a stable circadian rhythm of neuronal activity have been based on the averaged firing rate of the population. In slice preparations, these results have been obtained by calculating an average of relatively brief (i.e. 2 min) single-unit recordings from many SCN neurons and compiling their average firing rate in 2 h bins [6]. For in vivo studies, multiple-unit recordings have been obtained from a population of SCN cells simultaneously, which is essentially an averaging process. Individual ceils from explant cultures of the SCN, however, exhibit different period lengths that range from 16 to 32 h [1]. This preparation is the only one
* Corresponding author. Fax: (1) (303) 491-7907. 0006-8993/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved SSDI 0 0 0 6 - 8 9 9 3 ( 9 4 ) 0 1 3 5 6 - X
that allows continuous recordings of single neurons for long periods. One possibility consistent with the data of Bos and Mirmiran [1] is that the stable circadian rhythm obtained in vivo and in brain slices is a product of the averaging technique, while individual ceils exhibit rhythms that deviate considerably from 24 h. This is possible only if neurons with a 24-h period dominate the population by their higher rhythm amplitudes and prevalence. To test this hypothesis, we used a cosine function to reconstruct mathematically the rhythms of single neurons with known periods and amplitudes based on the published physiological data of Bos and Mirmiran [1] from explant cultures of the SCN. The rhythms from single cells were averaged, and the product of the population rhythm was then evaluated for the precision and stability of its 24 h period over several cycles. Reconstruction of firing rate rhythm was based on long-term recordings (36-66 h) from cultured SCN cells reported by Bos and Mirmiran [1]. From each explant a single cell was recorded, except for two cases in which a few ceils were recorded and analyzed simultaneously. In all cases significant periodicity was ob-
Y. Bouskila, F.E. Dudek / Brain Research 670 (1995) 333-336
334
A
served. The values of maximum firing rate (Fmax) , minimum firing rate (Fmin) and period were used to reconstruct individual rhythms of 17 cells using a simple cosine function (Eqn. 1), with an increment of 1 h for up to 200 h. This function was chosen because the original data were analyzed by fitting the waveform of firing rate to a cosine function [1]. The mean firing rate of the population at time t(F(t)) was calculated with Eqn. 2 using all 17 cells or four random samples of 4, 8 and 12 cells.
N=17
10
8
"~ L,
4 2
B
N=12
12 10
"6
8
L,-
6
•
•
•
-
"
• •
:
v
•
-:
V
•
'It
•
F(,)i = M + A* cos(27r t/~- + ~b)
A
. .
qt9
N
~2
(2)
flu) = 1/n E Fu,i i
4
C 10
6 4 2
D
N=4
8
6 4 T
'C_ 2 L~ i
L
I
I
I
t
I
I
I
I
0
20
40
60
80
100
120
140
160
180
Time
200
(h)
Fig. 1. Mean firing rate rhythm of different cultured SCN population sizes. A: averaged rhythm produced by 17 cells. B-D: averaged rhythms produced by different population sizes sampled randomly from the same pool of 17 cells. The number of cells averaged for each rhythm is denoted above it. Notice the increasing instability of rhythm amplitude and period as the population size was increased.
A
B
1.0
1.0
o "C.
t
t
0.8 0.6
l
F = firing rate (Hz), M = (Fmax + Fmi,)/2 (Hz), A = amplitude ( M - Fmi.) (Hz), t = time (h), ~"= period (h), n = number of cells, ~/, = phase (assumed to be 0 for simplicity and in order to maximize synchronization at t = 0). The resulting rhythms were further analyzed by calculating the coefficient of variation (C.V.) of the period and amplitude of the population oscillations to evaluate rhythm stability. Period length was measured as the time between adjacent peaks, and rhythm amplitude as the difference in frequency between peak and adjacent trough. The mean firing rate of a population of cultured SCN neurons with different periods is depicted for 200 h in Fig. 1A. This result is based on the maximal and minimal firing rate and period length of all 17 neurons [1], and assumes a perfect cosine function for the oscillations of each cell (see Eqn. 1). The rhythm is unstable in amplitude and period length. Periods ranged from 15.5 to 26.5 h (mean = 21.9, S.D. = 3.1)
N=8
8
0
(1)
~ 0.8 2
0.6
E "~
o
0.4
i
0.2
0.4
c~ 0.2
+
0
+
0.0
0.0 4
i
i
i
i
i
i
i
r
i
6
8
10
12
14
4
6
8
10
Number
of
cells
Number
of
i
12
14
cells
Fig. 2. Stability of rhythm as a function of number of cells sampled. A: coefficient of variation (C.V.) of amplitudes in compound rhythms produced by different population sizes. B: coefficient of variation of the periods in compound rhythms produced by different population sizes. Each point represents an average of four random samples, and the vertical bars represent standard errors of the mean.
Y. Bouskila, F.E. Dudek /Brain Research 670 (1995) 333-336
and amplitudes ranged from 0.1 to 4.7 Hz (mean = 1.9, S.D. = 1.2). To test the possibility that the instability may be the result of the small number of cells, we reduced the sampled population. As the number of cells was decreased (Fig. 1B-D), a progressive increase in the regularity of the rhythm amplitude and period was introduced. The average of four cells, selected randomly from the original population of 17, showed regular oscillations with a slight change in rhythm amplitude (Fig. 1D). Quantitative analysis of rhythm stability produced by different population sizes confirmed that the averaged rhythm became progressively more perturbed with increasing number of cells. Accordingly, the coefficients of variation of period length and of amplitude showed a tendency to increase with larger population sizes (Fig. 2A and B). The relative contribution of cells with different period length to the population rhythm was also tested. The deviation of individual cells from a period of 24 h was not correlated with the rhythm amplitude (r = - 0 . 1 2 , df = 15, P > 0.1). This result indicates that there is a random relation between the period length of a cell and its contribution to the average rhythm of the population. Based on data obtained from SCN neurons in a cultured explant, the only preparation that has so far allowed sufficiently long recordings from individual neurons, it appears that a simple average of activity from oscillators with different periods and amplitudes produces an unstable rhythm. This contrasts with the stable circadian rhythm reported in vivo. The clear difference in rhythm stability was present despite the fact that in both cases the recordings were extracellular and the rhythms were obtained by averaging firing rates. A strong period component of 24 h rhythm was detected using periodogram analysis in acute brain slices by sampling as few as 189 SCN neurons [3], while a very stable circadian rhythm was observed for over a month from a small population of SCN cells in vivo [4]. In the Bos and Mirmiran [1] study, only 17 cells were described, but our results demonstrated that an increase in sampling size only further perturbed the population rhythm (Figs. 1 and 2). Therefore, the differences in rhythm stability cannot be explained by different sample size. Oscillators with different periods and without interactions may produce a stable averaged rhythm, if those with a large deviation from 24 h contribute less, while precise 24 h oscillators dominate the rhythm by higher amplitudes and prevalence. The unstable rhythm we obtained from averaging the firing rate of individual SCN neurons, in addition to the random relation between period length and amplitude of these cells, suggest that individual neurons in the intact SCN have a
335
uniform 24 h period. Uniformity of period length may be achieved by intercellular communication between the neuronal oscillators in the SCN, which may have been disturbed in culture. Different mathematical models of weakly interacting oscillators predict that, even though individual oscillators may have variations of 10% to 20% in their native cycle, the group rhythm exhibits a stability several orders of magnitude greater than that of any constituent oscillator [10]. Both synaptic and non-synaptic mechanisms of synchronization are present in the SCN, and may play a role to produce the highly stable circadian rhythm observed. Physiological and immunocytochemical evidence suggests that G A B A is the major inhibitory neurotransmitter in the SCN. Axons of many SCN neurons terminate within the nucleus while most cell bodies and half of the terminals contain G A B A [8]. Furthermore, preliminary results suggest that glutamate application to the SCN during whole-cell voltage-clamp recordings results in an increase in GABA-mediated inhibitory postsynaptic currents [7]. Taken together, these results suggest a role for GABA in local circuit interactions. SCN neurons, however, can also loosely synchronize their firing activity in the absence of detectable chemical synaptic transmission via an unknown mechanism [2]. In SCN cultures, communication between glial cells has been reported as waves of elevated intracellular calcium moving from cell to cell [9]. A recent study has shown that stimulation of a single astrocyte in a mixed glial-neuronal culture from the forebrain triggers a wave of cytosolic calcium increase in neighboring neurons [5]. This glial-neuronal interaction was unaffected by synaptic transmission blockers or the Na+-dependent action potential blocker, tetrodotoxin. Some degree of coordination between oscillators with different periods seems to be necessary to produce the precise and stable circadian activity rhythm of the population seen in the intact SCN. The synaptic and non-synaptic mechanisms that are present in the SCN may provide the necessary neuronal coupling between the individual oscillators in the intact SCN.
We thank Dr. R.D. Traub for comments on the manuscript. This work was supported by a grant from the United States Air Force Office of Scientific Research.
[1] Bos, N.P.A. and Mirmiran, M., Circadian rhythms in spontaneous neuronal discharge of the cultured suprachiasmatic nucleus, Brain Res., 511 (1990) 158-162. [2] Bouskila, Y. and Dudek, F.E., Neuronal synchronization without calcium-dependent synaptic transmission in the hypothalamus, Proc. Natl. Acad. Sci. USA, 90 (1993) 3207-3210. [3] Green, D.J. and Gillette, R., Circadian rhythm of firing rate
336
Y. Bouskila, F.E. Dudek / Brain Research 670 (1995) 333-336
recorded from single cells in the rat suprachiasmatic brain slice, Brain Res., 245 (1982) 198-200. [4] Inouye, S.T. and Kawamura, H., Characteristics of a circadian pacemaker in the suprachiasmatic nucleus, J. Comp. Physiol., 146 (1982) 153-160. [5] Nedergaard, M., Direct signaling from astrocytes to neurons in cultures of mammalian brain cells, Science, 263 (1994) 17681771. [6] Prosser, R.A. and Gillette, M.U., The mammalian circadian clock in the suprachiasmatic nuclei is reset in vitro by cAMP, J. Neurosci., 9 (1989) 1073-1081.
[7] Strecker, G.J. and Dudek, F.E., Local synaptic circuits in the suprachiasmatic nucleus, Soc. Neurosci. Abstr., 20 (1994) 1439. [8] van den Pol, A.N. and Dudek, F.E., Cellular communication in the circadian clock, the suprachiasmatic nucleus, Neuroscience, 56 (1993) 793-811. [9] van den Pol, A.N., Finkbeiner, S.M. and Cornell-Bell, A.H., Calcium excitability and oscillations in suprachiasmatic nucleus neurons and glia in vitro, J. Neurosci., 12 (1992) 2648-2664. [10] Winfree, A.T., Biological rhythms and the behavior of populations of coupled oscillators, J. Theor. Biol., 16 (1967) 15-42.