©20

05 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

uren

euro

scie

nce

10 VOLUME 8 | NUMBER 1 | JANUARY 2005 NATURE NEUROSCIENCE

N E W S A N D V I E W S

the eyes stimulated prompt synaptic changes that strengthened the specificity of collicular inputs2 (Fig. 1b). Applying this synchroniza-tion procedure to the experimental design of Maffei et al. would nail down whether eye opening (or its deprivation) is sufficient to reorganize layer IV or whether instead a devel-opmental program independent of patterned vision also participates.

Would the huge changes in spontaneous activity measured by Maffei et al. in corti-cal slice be detectable in the intact brain? Extracellular recordings like those done11 in layer IV will be required to address this ques-tion. One can imagine combining the clever designs of the two new papers1,2 to evaluate local connectivity and activity in whole-cell recordings from layer IV in animals with one eye open and the other closed.

Cortex and superior colliculus are not the only structures in the visual pathway that eye opening might reorganize. Retinal ganglion cells were long believed to be immune to activ-ity-dependent modification. Yet we now know that a week of patterned vision stimulates the segregation of ganglion cell dendrites into on- and off-response specific layers12 and increases the rate of spontaneous synaptic events13.

Retinogeniculate connections are refined throughout this period, but we do not know to what extent the reorganization is driven by eye opening14. Synchronization experiments may tease out other prompt effects of eye opening on visual system maturation.

Though patterned vision through the opened eyes seems to trigger a maturation of the visual pathway, the effects of eye open-ing may be mediated instead or in addition by factors other than a change in activity. For example, the neurotrophic molecule BDNF is produced in the retina upon eye opening15. Transneuronal transport of BDNF injected into a visually deprived eye occludes the ocular dominance plasticity that would normally shift cortical responses toward the non-deprived eye (Mandolesi, G. et al., Soc. Neurosci. Abstr. 66.6, 2004. Neurotrophins may be an important prerequisite for allowing activity to mature the visual pathway.

It is gratifying to see that a dramatic event in development like eye opening has such strik-ing effects on the visual system. Using spon-taneous firing as a readout of rapid changes in local connectivity1 and synchronizing eye opening to measure prompt biochemical and synaptic changes2 seem like ideas too good

to have taken this long to appear. But maybe really good ideas always seem obvious once our eyes have been opened to them.

1. Maffei, A., Nelson, S.B. & Turrigiano, G.G. Nat. Neurosci. 7, 1353–1359 (2004).

2. Lu, W. & Constantine-Paton, M. Neuron 43, 237–249 (2004).

3. Blue, M.E. & Parnavelas, J.G. J. Neurocytol. 12, 599–616 (1983).

4. Desai, N.S., Cudmore, R.H., Nelson, S.B. & Turrigiano, G.G. Nat. Neurosci. 5, 783–789 (2002).

5. Pham, T.A., Impey, S., Storm, D.R. & Stryker, M.P. Neuron 22, 63–72 (1999).

6. Cancedda, L. et al. J. Neurosci. 23, 7012–7020 (2003).

7. Yoshii, A., Sheng, M.H. & Constantine-Paton, M. Proc. Natl. Acad. Sci. USA 100, 1334–1339 (2003).

8. Turrigiano, G.G. & Nelson, S.B. Nat. Rev. Neurosci. 5, 97–107 (2004).

9. Hubel, D.H. & Wiesel, T.N. The period of susceptibility to the physiological effects of unilateral eye closure in kittens. J. Physiol. 206, 419–36 (1970).

10. Berardi, N., Pizzorusso, T. & Maffei, L. Critical periods during sensory development. Curr Opin. Neurobiol. 10, 138–45 (2000).

11. Chiu, C. & Weliky, M. J. Neurosci. 21, 8906–8914 (2001).

12. Tian, N. & Copenhagen, D.R. Neuron 39, 85–96 (2003).

13. Tian, N. & Copenhagen, D.R. Neuron 32, 439–449 (2001).

14. Chen, C. & Regehr, W.G. Neuron 28, 955–966 (2000).

15. Seki, M., Nawa, H., Fukuchi, T., Abe, H. & Takei, N. Invest. Ophthalmol. Vis. Sci. 44, 3211–3218 (2003).

Bridging the gap: coupling single-cell oscillators in the suprachiasmatic nucleusChristopher S Colwell

Neurons in the mammalian master clock can maintain circadian rhythms in isolation, but must synchronize to function as a time-keeping system. A new study finds that gap junctions between neurons promote synchronous electrical activity and rhythmic behavior.

unambiguously demonstrate that SCN neurons are electrically coupled and that this coupling not only promotes synchronization of neural activity, but also is required for the maintenance of circadian rhythms in behavior.

The authors made intracellular recordings from pairs of neighboring SCN neurons. They found that about 25% of the neurons were electrically coupled and that these coupled cells showed synchronized spiking activity. The coupling strength and biophysical properties were similar to those measured in other types of coupled neurons4. Gap-junction channels are formed by a family of proteins called con-nexins. Connexin 36 (Cx36) is a major com-ponent of gap-junction-mediated electrical coupling in neurons4, and this seems to be the case in the SCN. Long et al. found that the electrical coupling between SCN neurons was lost in Cx36 knockout mice3. As compared to

regions like the inferior olive, the new study found that the percentage of coupled cells in the SCN was relatively low 3. This lower coupling frequency between SCN neurons seems to be consistent with our knowledge of SCN physiology. These clock cells do not show absolutely synchronized action potential generation; instead the population has coordi-nated firing rates that are high during the day and low during the night. However, it may be that some cell populations within the SCN are highly coupled and others not at all.

To determine whether gap-junction-mediated electrical coupling may also be involved in behavioral rhythmicity, the authors turned to the best-characterized behavioral output of the circadian system—namely, the wonderfully precise rhythms in wheel-running activity. In a light:dark cycle, both wild-type and Cx36 knockout mice synchronized to the

From daily sleep cycles to dinnertime, the cir-cadian system is responsible for the timing of behavior and physiology. In mammals, the conductor of this multifaceted timing system can be localized to a pair of structures in the hypothalamus known as the suprachiasmatic nucleus (SCN)1. Individual SCN neurons in isolation have the capacity to generate circadian oscillations in electrical activity, secretion and gene expression, but the cells drift out of phase with each other2. Understanding how individ-ual oscillators remain synchronized in the intact SCN has been a fundamental gap in our knowl-edge of SCN function. In this issue, Long et al.3

Christopher S. Colwell is in the Department of

Psychiatry and Biobehavioral Sciences at the

University of California, Los Angeles, 760 Westwood

Plaza, Los Angeles, California 90024, USA.

e-mail: ccolwell@mednet.ucla.edu

©20

05 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

uren

euro

scie

nce

NATURE NEUROSCIENCE VOLUME 8 | NUMBER 1 | JANUARY 2005 11

N E W S A N D V I E W S

lighting conditions and showed nocturnal activ-ity rhythms characteristic of rodents. However, in a light:dark cycle, photic input organizes the temporal pattern of activity by synchroniz-ing an endogenous clock to the period of the environmental signal (entrainment) as well as directly regulating activity (masking). To dis-tinguish between these two effects of light, the authors placed the mice in constant darkness and measured their activity rhythms without light cues. In these conditions, the Cx36-defi-cient mice showed rhythms that were weaker and less coherent than those of controls. These deficits seemed to be due to a greater tendency for the KO mice to be active at inappropriate times in their daily cycle. The cycle-to-cycle variability in the onset of the daily activity bout was also higher in the mutant mice. Thus, with-out Cx36, the circadian clock still keeps time but lacks the temporal precision that typically characterizes the behavioral output.

The Long et al.3 study helps to resolve a con-troversy about the presence and role of gap junctions in the SCN. The first suggestion that nonsynaptic mechanisms may link SCN neu-rons came from the observations that circadian rhythms in glucose utilization are present in the SCN before synapse formation5. In addi-

tion, when synaptic transmission is blocked by the removal of extracellular calcium, SCN neurons are still weakly coupled such that the activity of one cell increases the probability that a neighbor will generate an action poten-tial6. A tracer (biocytin, neurobiotin or Lucifer yellow) placed in one SCN neuron spreads to clusters of surrounding cells7–9. Dye coupling definitively marks the presence of gap junc-tions. However, because the dye-coupled cells in these studies were not physiologically char-acterized, it was unclear whether they were neurons, astrocytes or other non-neuronal cell types. Pharmacological gap junction blockers, such as halothane, disrupt circadian rhythms in SCN electrical activity and peptide secre-tion, as well as light-induced phase shifts of the circadian rhythm in wheel-running activity10. Unfortunately, these pharmacological tools are not very selective, and these agents have other effects besides blocking gap junctions. Anatomical studies have shown clear evidence for coupling between astrocytes and oligoden-drocytes in the SCN11, but proof of neuron-to-neuron coupling has proven elusive until recently. First, results from freeze-fracture and immunocytochemistry provided evidence for Cx36-containing gap junctions between SCN

neurons (Rash, J.E., et al., 749.11, Soc. Neurosci. Abstr., 2002). Now the new study3 demonstrates that SCN neurons are indeed electrically cou-pled and that this coupling is important for circadian rhythms in behavior (Fig. 1).

Like many good studies, this work raises as many questions as are answered by the experi-mental data. For example, we need to consider what signals are being spread from cell to cell via the gap junctions. Unlike chemical synapses, communication via gap junctions is bi-direc-tional and allows passage of small molecules (up to 1 kDa), thus linking cells both electrically and metabolically. Signaling molecules such as cyclic AMP, cyclic GMP, IP3 and calcium may be able to pass between neurons through these connec-tions. Future studies will have to consider the possibility that the passage of small molecules between cells may be as important as the direct passage of current. Gap-junction coupling also acts like an electrical filter in that some signals will pass more readily than others. During the day, SCN neurons undergo oscillations in mem-brane potential (2–8 Hz) that are driven by volt-age-gated calcium currents, among other ionic mechanisms12. These slower changes in mem-brane potential should pass more effectively through gap junctions than the fast voltage changes that occur during an action potential.

One of the more tantalizing observations in the new study was the suggestion that the electrical coupling between SCN neurons may itself be subject to diurnal variation. The authors found that coupling was greater in the middle of the day, when rhythmic neural activity in the SCN peaks, than in the late day or early night. This observation is consistent with previ-ous work demonstrating circadian variation in dye-coupling between SCN neurons8. In a few previous cases, changes in gap-junction permea-bility could be linked to changes in physiological function. For example, in the supraoptic nucleus of the hypothalamus, increased electrical cou-pling of oxytocin-secreting neurons may be a critical component of the milk-ejection reflex13. These types of observations raise the possibility that gap junctions do not just allow the passive spread of current, but instead form an actively regulated communication system whose proper-ties vary with the state of the organism.

Another unresolved issue concerns the rela-tive roles of electrical and chemical synaptic transmission in coupling SCN neurons. It is widely accepted that most SCN neurons express GABA and are likely to use this neurotransmit-ter for synaptic communication with other neurons in the SCN. In culture, GABA, acting through the GABAA receptor, can synchronize the electrical activity of SCN neurons9,14. Thus the synaptic release of GABA may act in concert with gap junctions to synchronize the neural

0 12 24 0 12 24

WT Cx36–/–

X

Time (h) Time (h)

Figure 1 Coupling of SCN neurons via gap junctions is important for the precision of circadian behavior. Top, schematic of pairs of SCN neurons (blue) from wild-type (WT) and Cx36–/– mice. Individual SCN neurons contain the molecular machinery necessary to generate circadian oscillations. One gap in our knowledge is the lack of understanding of how these single-cell oscillators are coupled. The new study3 demonstrates that SCN neurons are coupled through direct electrical connections. This coupling is lost in mice deficient in Cx36. Bottom, schematics of wheel-running activity records from WT and Cx36-deficient mice. Animals maintained in constant darkness show rhythms driven by the endogenous timing system. Each horizontal row represents the activity record for a 24-hour day. Successive days are plotted from top to bottom. The colored bars represent activity. The WT mice express robust circadian rhythms of locomotor activity with period shorter then 24 h. The onset of activity is typically under precise control. In contrast, the Cx36-deficient mice showed rhythms that were weaker and less coherent than controls. Without the Cx36, the circadian clock still keeps time but lacks the temporal precision that typically characterizes the behavioral output.

©20

05 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

uren

euro

scie

nce

12 VOLUME 8 | NUMBER 1 | JANUARY 2005 NATURE NEUROSCIENCE

N E W S A N D V I E W S

pling8,15. Sorting out the relative role of these interacting coupling mechanisms should keep SCN watchers busy for years to come. These mice and this new research3 should help us bridge the gap between cellular coupling and circadian behavior.

1. Reppert, S.M. & Weaver, D.R. Nature 418, 935–941 (2002).

2. Welsh, D.K., Logothetis, D.E., Meisterm, M. & Reppert, S.M. Neuron 14, 697–706 (1995).

3. Long, M.A., Jutras, M.J., Connors, B.W. & Burwell, R.D. Nat. Neurosci. 8, 61–66 (2004).

4. Connors, B.W. & Long, M.A. Annu. Rev. Neurosci. 27, 393–418 (2004).

5. Reppert, S.M. & Schwartz, W.J. J. Neurosci. 4, 1677–1682 (1984).

6. Bouskila, Y. & Dudek, F.E. Proc. Natl. Acad. Sci. USA 90, 3207–3210 (1993).

7. Jiang, Z.G., Yang, Y.Q. & Allen, C.N. Neuroscience 77, 1059–1066 (1997).

8. Colwell, C.S. J. Neurobiol. 43, 379–388 (2000).9. Shirakawa, T., Honma, S., Katsuno, Y., Oguchi, H. &

Honma, K.I. Eur. J. Neurosci. 12, 2833–2838 (2000).10. Michel, S. & Colwell, C.S. Chronobiol. Int. 18, 579–

600 (2001).11. Rash, J.E., Yasumura, T., Dudek, F.E. & Nagy, J.I.

J. Neurosci. 21, 1983–2000 (2001).12. Pennartz, C.M.A., de Jeu, M.T.G., Bos, N.P.A., Schaap,

J. & Geurtsen, A.M.S. Nature 416: 286–290 (2002).13. Hatton, G.I. Annu. Rev. Neurosci. 20, 375–397

(1997).14. Liu, C. & Reppert, S.M. Neuron 25, 123–128

(2000).15. Shinohara, K., Hiruma, H., Funabashi, T. & Kimura, F.

Neuroscience 96, 591–596 (2000).

activity of individual SCN oscillators. The SCN is made up of several cell populations whose specific functions we are just beginning to understand. One appealing hypothesis is that gap junctions may be more important for linking cells within a cell population, and that synaptic mechanisms may be more important for communication between SCN cell popula-tions. Of course, it is also possible that SCN neurons are coupled by multiple, overlapping mechanisms, which may not be independent. Two studies looking at dye coupling within the SCN found that activation of GABAA recep-tors by muscimol actively inhibits the cou-

An individual sensory neuron can convey only a small amount of information about the world, but somehow large populations of neurons join forces to create complex perceptions and behavior. How does this happen? Much of our understanding of how activity in a population of neurons relates to perception comes from pio-neering studies by Newsome and colleagues1–3 in the middle temporal (MT) cortical area, which is highly specialized for motion analysis. They argued that for discrimination between very different directions (for example, leftward versus rightward motion) in the presence of noise, indiscriminately pooling the responses of a large population of MT neurons with widely varying properties can account for behavioral performance1. In this issue, Purushothaman and Bradley report that an alternate strategy is used when monkeys are asked to perform a different task—discriminating closely related directions. The authors propose that in this case, perceptual decisions do not arise from unselective pooling, but are determined mainly by those neurons that are most sensitive to small variations around a particular direction4.

A sensory stimulus, such as a moving patch of dots, causes a large population of neurons to respond. However, individual neurons typically

The authors are in the Department of Anatomy and

Neurobiology, Washington University School

of Medicine, Box 8108, 660 South Euclid Avenue,

St. Louis, Missouri, 63110 USA.

e-mail: nadlerj@msnotes.wustl.edu

respond only to a small fraction of possible sensory inputs —that is, they are tuned to a limited range of stimulus properties. Neurons may respond to a sensory stimulus presented as part of a perceptual task, but may not be sensi-tive to the stimulus properties that are criti-cal for task performance. Some neurons may be much better suited than others for a given task because their tuning properties match the range of relevant stimulus properties. The activity of the whole population must be ‘read out’ to inform perceptual decisions, and pre-sumably the same population contributes to multiple behaviors.

How does one determine which neurons provide the critical sensory information for a perceptual decision? One approach is to exam-ine how neural activity relates to perceptual decisions, independent of the sensory stimu-lus. For some MT neurons, firing rate across repeated presentations of a sensory stimulus is correlated with the perceptual decisions of the monkey, even when the stimulus is ambiguous. For example, a neuron that prefers motion of dots to the right may show slightly increased firing rates on trials when the monkey per-ceives rightward motion, even if the stimulus does not contain net rightward motion. These correlations—which can be quantified as ‘choice probabilities’5 —are generally modest, so it is often assumed that responses of a group of neurons are pooled together to contribute to the monkey’s decision. For tasks involving motion and depth perception, many neurons

in MT show choice probabilities significantly greater than expected by chance5–7.

Purushothaman and Bradley identified MT neurons that were potentially linked to fine motion discrimination by virtue of high choice probabilities and analyzed the potential readout of the population under different task condi-tions from those previously used by Newsome and colleagues. The authors used a two-interval direction-discrimination task in which the monkey had to decide whether the direc-tion of motion of one random-dot test stimulus was clockwise or counterclockwise relative to a reference stimulus. The difference in direction between the test and the reference was always very small (≤3°) and the reference direction was always straight up (90°). This new study dif-fered from previous work3,5 in two key respects (Fig. 1). First, the monkeys were asked to make much finer discriminations than in prior experiments, which involved choosing between opposite motion directions embedded in noise. Second, the stimulus was not matched to the preferred direction of each neuron (though the location, size and speed of the stimulus was tai-lored to the neuron’s tuning preference).

This design allowed the authors to measure choice probabilities for MT neurons with a wide range of sensitivities based on their direc-tion tuning relative to the reference direction, thereby identifying the neurons that conveyed the most information about the animal’s decision. This enabled them to distinguish between different possible schemes for pool-

Precision pooling predicts primate perceptual performanceJacob W Nadler & Gregory C DeAngelis

Moving stimuli evoke a response from a large number of neurons in cortical area MT. A new study investigates how perceptual decisions may arise from that population response, with important implications for theories of neural coding.