Competition between the two eyes for synaptic space is thought to play a crucial role in the developmental plasticity of ocular dominance in the primary visual cortex. This competition should be disrupted if geniculocortical afferents from the two eyes are spatially segregated. In kittens, strabismus was induced in one eye before the onset of the critical period; the effects of a brief period of monocular deprivation (MD) at the height of the critical period and subsequent recovery were assessed in a longitudinal study employing optical imaging of intrinsic signals. Results were compared with those from a control group without strabismus. MD caused a substantial loss of cortical territory dominated by the deprived eye in all animals. However, in the strabismic animals this loss was smaller than in the control group for the hemisphere contralateral to the deprived eye. When the deprived eye was reopened, recovery of cortical territory was remarkably rapid in all kittens, and close to pre-deprivation responses were attained within 3–4 days of reopening. However, kittens without strabismus exhibited a greater rate of recovery from MD. Moreover, recovery of visual acuity, as assessed by visually evoked potential (VEP) measurements, was slower and less complete in animals with strabismus prior to MD. Therefore, strabismus does not provide lasting protection against the effects of MD.
Since the pioneering work by Wiesel and Hubel (1963), monocular deprivation (MD) by eyelid suture and its effect on visual cortex physiology, anatomy and function has been one of the key paradigms for study of cortical plasticity. If imposed during the so-called critical period of early postnatal life, even a brief epoch of MD causes a pronounced shift in ocular dominance (OD) of neurons in the primary visual cortex (V1) towards the experienced eye (Olson and Freeman, 1975), which is accompanied by a loss of territory in layer 4 occupied by geniculocortical afferents that represent the deprived eye (Shatz and Stryker, 1978). Moreover, a profound loss in visual acuity is observed, such that the deprived eye becomes virtually blind (Mitchell et al., 1977). By comparison, binocular deprivation by dark-rearing for a similar length of time has a less severe effect on cortical binocularity and visual function (Wiesel and Hubel, 1965). These observations led Wiesel and Hubel (1965) to propose the now widely held view that ocular dominance development and plasticity reflects competitive interactions between the two eyes for synaptic space in the visual cortex, a hypothesis most explicitly stated by Guillery (1988). More recent evidence suggests that inputs from the two eyes compete for limiting amounts of a retrograde factor such as a neurotrophin (Bonhoeffer, 1996), and active neurons representing the open eye can outcompete less active neurons representing the closed eye and induce weakening of deprived-eye synapses. According to this view, the relative levels of evoked activity in afferents representing the two eyes determine functional changes in response to altered visual experience. However, if the deprived eye of a monocularly deprived kitten is simply reopened, there is substantial physiological and behavioural recovery (Giffin and Mitchell, 1978; Mitchell, 1988; Mitchell and Gingras, 1998). As reopening of the deprived eye does not give it a competitive advantage over the experienced eye unless the latter is sutured at the same time (reverse occlusion), anatomical and physiological recovery would not have been expected on the basis of the competition hypothesis, and indeed no recovery is observed in the same paradigm in macaques (Blakemore et al., 1981, Swindale et al., 1981). Mitchell and Gingras (1998) suggested on the background of their results that absolute activity levels, or some other non-competitive mechanisms, determine the degree of recovery from MD.
Interestingly, an alternative theory of synaptic modification, the Bienenstock–Cooper–Munro (BCM) model (Bienenstock et al., 1982; Bear et al., 1987), predicts recovery of the deprived eye, provided inputs from the two eyes are temporally correlated (Clothiaux et al., 1991; Kind, 1999). Indeed, we have recently demonstrated that correlated binocular input is essential for such recovery: physiological and behavioural recovery is far less complete if the two eyes are surgically misaligned after a period of MD (Kind et al., 2002). The absence of recovery from MD in monkeys, if the deprived eye is simply reopened, can be explained by the fact that their much smaller V1 receptive fields make a decorrelation of binocular inputs through even a small misalignment of the visual axes more likely. Such misalignments are observed frequently in monkeys following MD (Quick et al., 1989).
Here we further probe the involvement of competitive interocular interactions both during and following a period of monocular deprivation. This is the first longitudinal physiological study of the effects of MD and subsequent recovery, an approach that overcomes the uncertainties inherent to inter-individual comparisons, and greatly increases statistical power. Animals, even more so if not littermates, can vary widely with respect to the layout of OD columns, their periodicity and their contralateral bias (Kaschube et al., 2002, 2003), and starting conditions differing between animals can therefore compound or obscure developmental changes. In contrast, in an individual animal, maps remain remarkably stable in the absence of any manipulation of visual input (Chapman et al., 1996; Gödecke et al., 1997).
We surgically induced strabismus in one eye prior to MD and examined whether the ensuing decorrelation and enhanced segregation of inputs from the two eyes protected V1 from the OD shift typically associated with MD, as has been suggested in a brief report (Mustari and Cynader, 1981). Based on the competition hypothesis, such a protective effect would be expected, because left- and right-eye geniculocortical afferents no longer contact the same postsynaptic neurons. Since the eventual visual outcome is as important as the immediate deprivation result for the judgement of a potentially protective intervention, we also investigated the time-course of physiological recovery in order to test whether any advantage afforded by prior strabismus persists. We discuss our results in the context of competitive and associative mechanisms of synaptic plasticity.
Materials and Methods
Experiments were performed on 16 kittens bred in a closed colony; all procedures were carried out in accordance with UK Home Office regulations on animal experimentation [Animals (Scientific Procedures) Act 1986] and the European Communities Council Directive 86/609/EEC. Efforts were made to minimize animal suffering and to reduce the number of subjects used.
Three normally reared kittens served as controls. Five kittens were raised normally until about postnatal day 35 (P35). The right eye was then deprived by lid-suture for ten days, after which (at P45) it was reopened. In these animals (MDR) the ensuing binocular visual experience was presumably concordant, since they had no obvious squint and the projections of their visual axes, determined by direct ophthalmoscopy under paralysis during the optical imaging experiments, were similar to those of normal animals. We chose a period of MD long enough for maximal physiological changes to occur — these are almost saturated after 4 days (Olson and Freeman, 1975; Movshon and Dürsteler, 1977) — but brief enough to avoid irreversible anatomical degeneration: geniculocortical afferents serving the deprived eye shrink significantly after only a week of MD but partially regrow within 10 days following reverse lid-suture (Antonini and Stryker, 1993; Antonini et al., 1998).
In seven animals, strabismus was induced during the third postnatal week: either a convergent squint (six kittens) by myotomy of the lateral rectus muscle or a divergent squint (one kitten) by myotomy of the medial rectus muscle in the right eye. These animals were also monocularly deprived for 10 days, beginning at about P35. In four of them, the surgically deviated eye was lid-sutured (SR-MDR), while in the other three kittens the non-deviated eye was deprived (SR-MDL). All seven animals had binocular vision following the reopening of the deprived eye. We noted that two of the SR-MDL kittens actually appeared to fixate with the originally operated and non-deprived right eye, so that the deviating eye was the deprived left eye. In all animals, the angle of squint was determined under paralysis in the terminal experiment, by means of back-projecting the optic disks and areae centrales onto a tangent screen. The smallest angle was 4°, while for the remaining kittens it ranged from 7 to 15°.
For all kittens, ocular dominance and orientation maps in the primary visual cortex were obtained by optical imaging of intrinsic signals, before the onset of monocular deprivation, immediately after the MD period and at a further three time points over a total period of 3 weeks during binocular recovery.
Optical Imaging and Analysis
All surgical procedures were performed under sterile conditions. Anaesthesia was induced with an i.m. injection of ketamine (25–40 mg/kg) and xylazine (3–4 mg/kg). Animals were intubated and were placed in a stereotaxic frame. They were artificially ventilated (50–60% N2O, 40–50% O2, 2.0–2.5% isoflurane, decreased to 1.5% during imaging). ECG, EEG, end-tidal CO2 and rectal temperature were monitored continuously. A 4% glucose in saline solution was infused i.v. at 3 ml/kg/h throughout the experiment.
Optical imaging of primary visual cortex was performed as described previously (Bonhoeffer and Grinvald, 1996). In the initial imaging session, the scalp was incised and retracted. A circular craniotomy was performed above area 17 and a titanium chamber was cemented onto the skull. The cortical surface was carefully cleared and kept free from any traces of blood using Sugi sterile swabs (Kettenbach, Eschenburg, Germany). The chamber was filled with sterilised silicon oil (dimethylpolysiloxane, Sigma-Aldrich, Poole, UK) and was sealed with a glass coverslip. In order to reduce invasiveness to a minimum and to protect the underlying cortex from both mechanical damage and the risk of infection, the dura was usually left intact for all but the terminal imaging session.
Images were captured using an enhanced differential imaging system (Imager 2001, Optical Imaging Inc., Mountainside, NJ), with the camera focused ∼500 μm below the cortical surface and an illumination wavelength of 700 nm. Visual stimuli, produced by a stimulus generator (VSG, Cambridge Research Systems, Rochester, UK), consisted of high-contrast, sinusoidally modulated gratings (0.15–0.75 cycles/deg) of four different orientations, drifting at a temporal frequency of 2 Hz, presented to the two eyes separately in randomized sequence, interleaved with trials in which the screen was blank. Activity maps were analysed using IDL software (RSI, Boulder, CO). Single-condition responses (averages of 48–64 trials per eye and orientation) were divided (i) by responses to the blank screen and (ii) by the sum of responses to all four orientations (‘cocktail blank’; (Bonhoeffer and Grinvald, 1996)) to obtain iso-orientation maps. The actual signal used for subsequent quantitative analysis was reflectance change (ΔR/R) for each pixel, given at 16-bit precision. For illustrations, signals were range-fitted such that the 1.5% most responsive (least responsive) pixels were set to black (white), and signal amplitude was displayed on an 8-bit greyscale. Following each but the final experiment, the chamber was half-filled with agar containing an anti-inflammatory steroid (Dexafort, Intervet UK Ltd., Milton Keynes, UK). The rest of the chamber was filled with silicone oil and sealed with a glass cover-slip; anesthesia was then suspended. Kittens were given systemic antibiotics (Betamox, Norbrook Laboratories, Carlisle, UK; or Metacam, Boehringer-Ingelheim, Ingelheim, Germany) and analgesics (Ketofen, Merial, Harlow, UK); these injections were given prophylactically for 5 days. Animals were allowed to recover and were then returned to their mother and littermates. Completion of the final imaging session was followed by electrophysiological recording (see below); afterwards, animals received an overdose of barbiturate.
Ocular dominance maps were obtained by dividing the sum of responses to all four orientations through one eye by the similar sum of responses through the other eye. Relative rather than absolute responses through the two eyes were compared, since the latter are dominated by a rather variable global DC signal component which is superimposed on the spatially restricted stimulus selective (‘mapping’) signal (Bonhoeffer and Grinvald, 1996; Zepeda et al., 2004). Resulting maps were high-pass filtered (cut-off, 2.2 mm) in order to remove the DC signal component. Given that OD domains have an average size of 1 mm or less, the mapping signal was unaffected by the chosen high-pass filter. Moreover, repeated analyses with different cut-offs yielded essentially the same results as described below. In order to remove high-frequency noise, the images were also smoothed (Gaussian smoothing over 6 by 6 pixels). Within a region of interest that comprised the visually responsive part of the images in both cortical hemispheres, excluding blood vessel and other artefacts (as identified in an image of the surface taken with green illumination), pixels were assigned to the left and right eye, respectively, depending on whether their value was >1 or <1.
Orientation preference maps were calculated by vectorial addition of four blank-divided iso-orientation maps and pseudo-colour coded. In these ‘polar’ maps the vector angle is displayed as hue; the length of the vector is encoded as the brightness of the colour (Bonhoeffer and Grinvald, 1993). Dark regions indicate areas of weak orientation selectivity or areas where cells with very different orientation preference are found in close proximity, as is the case in pinwheel centres (Maldonado et al., 1999). Orientation selectivity indices were calculated for responses at each pixel as:
In four experimental and two normal control animals we determined quantitative orientation/direction tuning curves of single units, recorded with glass-insulated tungsten microelectrodes, and discriminated by their spike shapes using Brainware software (TDT, Alachua, FL). Left- and right-eye responses to drifting gratings (of optimum spatial frequency) of 16 different directions in 22.5° steps were averaged over 5 trials of 1.5 s duration. Smooth tuning curves were fitted to the data points based on Fourier analysis (Wörgötter and Eysel, 1987), and the preferred orientation and half-width of tuning at half-height (HWHH) were determined from these curves. Ocular dominance was calculated as the ratio of total responses to contra- and ipsilateral eye stimulation. Spatial frequency tuning curves were obtained for both eyes with gratings of optimal orientation and 12 spatial frequencies ranging from 0.1 to 4.52 cycles/deg in half-octave steps. Neuronal acuity was determined as the high-frequency cut-off.
Visually Evoked Potentials
In four animals we determined visually evoked potentials (VEPs) through either eye and each cortical hemisphere, recorded with a silver electrode placed on the dura near P4 / L1 (Horsley-Clarke coordinates). Left- and right-eye responses (low-pass, 300 Hz; high-pass off) to horizontal contrast-reversing gratings of a range of spatial frequencies from 0.05 to 6.4 cycles/deg and a reversal rate of 1 Hz were averaged over 20 trials of 3 s duration, using software written in LabVIEW (National instruments, Austin, TX). Results were analysed using IDL software (RSI, Boulder, CO). Responses to the six contrast reversals per trial were superimposed and then averaged across trials. Total VEP amplitude was defined as the peak-to-trough voltage difference of the averaged response.
OD Prior to Deprivation
When imaged prior to the 10 day period of MD, kittens exhibited a modest bias in ocular dominance towards the contralateral eye in each hemisphere of V1, regardless of whether the animals had previously been made strabismic or not. In the five normal kittens, the left eye dominated, on average, 63.6% of the right V1 and the right eye dominated 63.9% of the left V1. In the seven strabismic animals, the non-operated left eye dominated 56.0% of the right V1 and the deviated right eye dominated 58.7% of the left V1. Although the contralateral bias was larger in the normal than in the strabismic group, this is unlikely to be a consequence of the surgically induced strabismus prior to the period of MD.
We also repeatedly imaged three kittens reared normally for a similar period of time (from 5 weeks of age) in order to control for the reproducibility and stability of quantitative OD results. In all animals, the relative territory dominated by the two eyes changed only by a few per cent during this time, and no consistent trend was observed. Images and quantitative data for one kitten are shown in Figure 1A,B. Overall, the contra- to ipsilateral bias was at around 60% to 40% throughout the recorded period (Fig. 1C).
Extent of OD Shift Caused by MD and Time-course of Recovery
Immediately after a 10 day period of right-eye MD, the five MDR animals all showed a characteristic, pronounced OD shift towards the open eye. The magnitude of the shift was similar in both cortical hemispheres, with the deprived eye dominating 16.7 ± 2.1% (mean ± SEM) of cortical territory in the contralateral, left hemisphere and 15.6 ± 1.3% in the ipsilateral, right hemisphere. A typical example, kitten 282, is illustrated in Figures 2 and 3A,B. In this animal, the right-eye territory decreased from 58.6% to 19.6% in the left hemisphere and from 36.2% to 19.2% in the right hemisphere during the 10 day MD period. Remarkably, within just 2–4 days of reopening, the deprived eye recovered virtually all of the lost territory. In kitten 282, four days of binocular recovery resulted in the previously deprived eye dominating 54.6% of V1 in the left hemisphere and 32.3% in the right hemisphere. For all MDR kittens, the mean (± SEM) values were 55.4 ± 3.0% for the left (contralateral) and 41.4 ± 3.9% for the right (ipsilateral) hemisphere. Interestingly, in one of two animals that were reassessed within only 2 days of reopening, the recovery of the deprived eye was incomplete (Fig. 4).
When ocular dominance was assessed at weekly intervals thereafter, unsurprisingly, little change was observed, as illustrated for kitten 282 in Figures 2 and 3A,B. Notably, the overall pattern of the OD maps also changed very little throughout the experiment (Fig. 2A), apart from a slight expansion that is commensurate with the growth of the cortical surface during the same time (Sengpiel et al., 1998). For all five MDR kittens, the right (formerly deprived) eye dominated 63.2 ± 3.1% of territory in the contralateral and 38.8 ± 2.2% in the ipsilateral hemisphere after 2.5–3 weeks of recovery. These values were not significantly different from those obtained prior to deprivation (P > 0.3, t-test).
In the initial set of experiments on kittens with prior strabismus, strabismus was induced in the same eye that was to be deprived in order to exclude the possibility that strabismus in one eye might balance out the effect of MD in the other, since it could be argued that strabismus by itself might have a detrimental effect on vision (e.g. amblyopia) in the affected eye and could increase the likelihood that the other eye would dominate the cortex. Later, we found that there was no significant difference in the territory occupied by the deprived eye after MD (see Fig. 5B) between animals that had been subject to strabismus and deprivation in the same eye (SR-MDR) or in opposite eyes (SR-MDL); the two groups were therefore combined for statistical analyses. A typical example from the SR-MDR group is shown in Figures 6 and 7 (kitten 065). Notably, the majority of animals with prior strabismus displayed a smaller OD shift in the hemisphere contralateral to the deprived eye than the MDR kittens, while the OD shift was more or less identical in both groups for the ipsilateral hemisphere. The preserved deprived-eye responses are well visible in the left hemisphere of kitten 065 (Fig. 6B, P45). Overall, in the seven strabismic kittens, the deprived eye dominated 28.8 ± 3.1% of the contralateral hemisphere immediately after the 10 day MD period, but only 19.2 ± 2.6% of territory in the ipsilateral hemisphere (Fig. 5B). Squint angle was not significantly correlated with the area of cortex responding to the deprived eye in the contralateral hemisphere after the period of MD (r = −0.16).
Since the initial OD distributions differed between the MD and the SR-MD kittens, the ‘protective effect’ of strabismus on territory dominated by the deprived eye in the contralateral hemisphere becomes even more striking when one compares the reductions in cortical territory rather than percentages of territory (compare Figs 5A and 5B). In the MD animals, the deprived-eye territory decreased by 46.8 ± 2.5% in the contralateral and by 20.0 ± 2.0% in the ipsilateral hemisphere. In contrast, in the SR-MD kittens the reduction was just 27.5 ± 4.5% for the contralateral hemisphere, while it was almost identical for the ipsilateral hemisphere (21.5 ± 2.0%).
Like the MD-only kittens, in animals with strabismus prior to MD, a brief period (3–5 days) of binocular recovery was sufficient to restore pre-deprivation OD in the hemisphere contralateral to the deprived eye (57.5 ± 5.0%). The deprived-eye territory in the ipsilateral hemisphere also recovered but did not quite reach pre-MD values (36.0 ± 2.5%); in particular, recovery was slower in the SR-MDR group than in the other groups at that time point (Fig. 5C), although this difference was not significant. Squint angle was inversely correlated with the percentage of cortical territory recovered after a brief period of binocular recovery (r = −0.79), therefore the magnitude of the squint did affect recovery of deprived-eye territory. However, this correlation did not reach significance (P = 0.11). An extended period (3 weeks) of recovery restored pre-deprivation OD values in both SR-MDR and SR-MDL groups (Fig. 5D).
Effect of MD on Orientation Selectivity and Time-course of Recovery
In all MDR kittens, the weak responses that were obtained through the deprived eye immediately after reopening were mostly lacking in orientation selectivity in both cortical hemispheres. One example is shown in Figures 2C and 3C for kitten 282. Following the reopening of the deprived eye, orientation selectivity recovered, but more slowly than OD territory, returning to close to pre-MD levels only by the second assessment, after 11 days of binocular experience.
We calculated the OSI (see Materials and Methods) for both the deprived eye and the non-deprived eye, and took their ratio as a relative measure of the deprived eye's orientation selectivity. Pooling all MDR kittens, we found this parameter to be consistently higher for the contralateral than for the ipsilateral hemisphere. However, the effects of MD and subsequent binocular recovery were qualitatively identical for both hemispheres (Fig. 8). OSI ratios averaged across both hemispheres dropped from 1.06 before MD to 0.39 after MD, and recovered to 0.86 after 2–4 days and 0.90 after 3 weeks of binocular experience. In summary, like OD territory, orientation selectivity recovered rapidly after reopening of the deprived eye, but recovery was not quite as fast and as complete.
Interestingly, the SR-MD kittens exhibited a similar loss of orientation selectivity as the MDR kittens during the MD period, despite a reduced OD shift in the hemisphere contralateral to the deprived eye. In particular, this also held true for the animals with the strongest remaining deprived-eye responses, such as kitten 065 (Figs 6C and 7C). There were no significant differences between the MDR and the SR-MD kittens at any time point. For all seven strabismic kittens, the relative OSI (averaged across both hemispheres) in the deprived eye dropped from 0.97 to 0.37. Again, recovery was rapid but not quite as fast or as complete as for ocular dominance, the OSI ratio being 0.82 after 3–5 days of recovery and 0.87 after 3 weeks (see Fig. 8).
Recovery of Neuronal Response Properties Following MD
Single-cell recordings were carried out in four animals at the end of the recovery period. In two MDR kittens, a total of 78 neurons were recorded, most of which (68, or 87.2%) were binocular, and only 10 cells (12.8%) were monocular (OD1 or OD7; Hubel and Wiesel, 1962). Overall, 48 (61.5%) cells responded more strongly to left- (non-deprived) and 30 (38.5%) more strongly to right-eye stimulation (Fig. 9A). We determined orientation selectivity in terms of HWHH of responses to monocular stimulation. For the cells dominated by the non-deprived eye, HWHH was 26.6 ± 1.7° (mean ± SEM), while for those dominated by the formerly deprived eye HWHH was slightly but not significantly (P = 0.09) greater, at 31.6 ± 2.4° (Fig. 9B).
In two SR-MDL animals, 60 neurons were recorded, of which 23 (38.3%) were strictly monocular (OD1 or OD7; Hubel and Wiesel, 1962) and a further 20 (33.3%) showed only a weak response through the non-dominant eye (OD2 or OD6). In total, 28 cells (46.7%) responded more strongly to stimulation of the left (formerly deprived) and 32 neurons (53.3%) more strongly to stimulation of the right (strabismic) eye (Fig. 9C). For the cells dominated by the deprived eye, orientation tuning HWHH was 35.5 ± 3.6°; this was slightly but not significantly (P = 0.27) greater than the value for the strabismic eye of 29.1 ± 4.4° (Fig. 9D).
By comparison, for 96 neurons recorded in two age-matched normal control kittens, the mean HWHH was 29.5 ± 1.4°. This value did not differ significantly from values obtained for either eye in SR-MDL or MDR animals.
We also obtained spatial-frequency tuning curves and determined high-frequency cut-offs. In the two MD-only animals, the geometrical mean of the cut-offs through the non-deprived eye was 1.59 cycles/deg, while through the deprived eye it was 1.37 cycles/deg. For statistical testing, the cut-off values were log2 transformed. The values obtained through the deprived eye (0.45 ± 0.74, mean ± SD) and the non-deprived eye (0.67 ± 0.62) did not differ significantly (P = 0.15, t-test). In the two kittens with strabismus prior to MD, the geometrical mean of the cut-offs through the strabismic eye was 1.24 cycles/deg, while through the deprived eye it was 1.13 cycles/deg. Again, the log2 transforms of the cut-offs through the strabismic eye (0.31 ± 0.54) and the deprived eye (0.18 ± 0.43) did not differ significantly (P = 0.3, t-test).
In summary, following recovery, neurons dominated by the previously deprived eye did not differ from those dominated by the non-deprived eye in either the MDR or the SR-MD group; the only significant difference between the two groups was the prevalence of binocular cells in the MDR and of monocular neurons in the SR-MD group.
Visually Evoked Potentials: Effect of MD and Recovery
VEPs were recorded from both cortical hemispheres in two MDR and two SR-MDL animals, on the same days as the optical recordings. For each eye and hemisphere, VEP amplitude was plotted against the spatial frequency of the contrast-reversing grating stimuli, and high frequency cut-offs were determined.
VEPs recorded prior to MD did not reveal a clear contralateral eye bias in terms of amplitudes or cut-off frequencies. Stimulus-modulated VEPs were observed through either eye for spatial frequencies of up to ∼1.6 cycles/deg, in agreement with single-cells responses. MD all but eliminated evoked responses through the deprived eye, not only in the MD-only animals but also in kittens with strabismus prior to MD (Fig. 10). In contrast to the OD maps, no ‘protection’ from the effect of MD was observed in the hemisphere contralateral to the deprived eye (Fig. 10), and responses were depressed to similar extents as in the MDR kittens in both hemispheres of the SR-MDL animals (Fig. 11).
By comparison with OD maps, VEP responses through the deprived eye recovered more slowly and less completely in the SR-MDL kittens than in the MDR kittens. After 3 days of recovery, VEP amplitudes for the deprived eye were between one-third and one-half of the non-deprived eye responses, and cut-off frequencies were 1.5 octaves lower. Even after 3 weeks of recovery, VEP amplitudes remained clearly decreased and cut-off frequencies slightly reduced (Fig. 11). In contrast, in the MDR kittens, VEP cut-off frequencies for the deprived eye recovered to within half an octave of those for the non-deprived eye within just 2 days, and matched those for the non-deprived eye after 11 days. In addition, VEP amplitudes recovered faster and more fully than in the SR-MDL kittens (Fig. 11).
The main findings of our study are threefold. First, kittens with strabismus in one eye exhibit a smaller-than-normal loss of cortical territory, for an eye that is deprived near the peak of the critical period. However, this partial ‘protective’ effect of a prior strabismus is limited to the hemisphere contralateral to the deprived eye. Moreover, VEP recordings did not reveal a difference between the strabismic and control groups in this respect. Secondly, recovery of cortical territory in terms of OD maps was remarkably rapid in all kittens, the previously deprived eye attaining close to pre-deprivation responses within a few days of reopening. Single-cell responses after recovery were also close to normal in both groups. Kittens without strabismus prior to MD exhibited a greater rate of recovery from MD. Thirdly, recovery of visual acuity, as assessed by VEP measurements, was slower and less complete in animals with strabismus prior to MD than in those without. Overall, a beneficial effect of strabismus on the response to MD was only evident in some measures but not in others.
MD Effects and Binocular Competition
The fact that monocular deprivation has much more severe effects on the ocular dominance of neurons in V1 than binocular deprivation (Wiesel and Hubel 1965) has led to the prevailing viewpoint that competition between the two eyes for synaptic space plays a crucial role in development and plasticity of the mammalian visual system. This heterosynaptic interaction (i.e. between the two sets of synapses representing the two eyes) should be disrupted if the geniculocortical afferents from the two eyes are spatially segregated. Here, strabismus was induced surgically more than 2 weeks before the start of MD. Trachtenberg and Stryker (2001) have shown that both the physiological loss of binocularity among V1 neurons and the anatomical loss of horizontal connections between opposite eye-column in V1 are profound within 2 days, while 7–14 days after strabismus induction the remodelling of the geniculocortical afferents is complete and inappropriate connections to opposite-eye domains are entirely lost. We can therefore assume that the overlap of left- and right-eye afferents in layer 4 of our strabismic kittens was minimal at the onset of MD. Nevertheless, these animals were ‘protected’ from the effects of MD only to a limited extent, and only in the hemisphere contralateral to the deprived eye, where the OD shift was ∼60% of that in the control group. Orientation selectivity was not preserved.
Our findings in part agree with and in part contradict a short report based on single-cell recordings in MD kittens (Mustari and Cynader, 1981), which described a normal OD distribution (i.e. complete protection) in the hemisphere contralateral to the deprived eye in animals with prior strabismus; orientation and spatial frequency tuning were not assessed in that study. Our results are not what one would have expected, if binocular interactions in V1 were entirely ‘competitive’. In that case, MD should have caused a much reduced OD shift in both cortical hemispheres, and orientation selectivity would have been expected to persist as well. However, the premise underlying that prediction is complete segregation of left- and right-eye domains in V1 following early-onset strabismus. But whilst thalamic inputs segregate clearly within layer 4 (Shatz et al., 1977), the dendrites of layer 4 cells are still capable of sampling from both sets of afferents (Kossel et al., 1995). Consequently, as both our single-cell recordings and several previous electrophysiological studies have shown, 20–30% of V1 neurons maintain at least some binocular input (e.g. Hubel and Wiesel, 1965; Sengpiel et al., 1994). Moreover, horizontal corticocortical connections between opposite-eye domains are much reduced, but still present (Löwel and Singer, 1992; Trachtenberg and Stryker, 2001). Given the incomplete anatomical and physiological segregation of the two eyes' inputs in V1, strabismus might not be expected to afford full protection from the MD-induced OD shift.
In recent years, a number of models have been developed that can account for the formation and developmental plasticity of ocular dominance and orientation preference maps in V1. The BCM model (Bienenstock et al., 1982) has been particularly successful in explaining the outcome of a variety of rearing conditions (Clothiaux et al., 1991; Kind, 1999).
The BCM model accurately predicts the rapid OD shift and loss of orientation selectivity associated with MD (Clothiaux et al., 1991), as well as the equally rapid recovery from those MD effects (see Sengpiel and Kind, 2002). For animals with strabismus prior to MD, BCM simulations predict a slower ocular dominance shift away from the deprived eye, for cells in which this eye is the dominant one (S.D. Faulkner and F. Sengpiel, unpublished observations). Depending on the length of the deprivation period, this would result in a reduced OD shift compared with MD-only animals, similar to what we observed experimentally. However, BCM simulations also predict that cells driven through the deprived eye will remain orientation tuned for as long as they remained dominated by that eye, whereas we found orientation selectivity to be largely lost in OD patches still responding to the deprived eye in SR-MD strabismic animals. Therefore, our findings can to a large extent, albeit not completely, be explained by BCM theory.
It is worth considering other homeostatic mechanisms that have been demonstrated in recent work to be active in addition to traditional Hebbian plasticity (for reviews, see Desai, 2003; Turrigiano and Nelson, 2004). For example, the apparent binocular competition in MD can be explained by synaptic scaling, an activity-dependent bidirectional change in quantal synaptic currents (Turrigiano et al., 1998). High activity causes a decrease in synaptic currents through a decrease in the number of receptors per synapse, and low activity has the opposite effect, while the relative strengths of left- and right-eye synapses remain the same. Since synaptic scaling requires integration of activity over long periods of time and affects all inputs to a particular neuron equally, it provides a mechanism of ‘metaplasticity’ that is analogous to the cell-wide modification threshold of the BCM model.
An alternative mechanism that does not require a global intracellular signal is spike-timing-dependent synaptic plasticity (STDP; Song et al., 2000). In contrast to the BCM model, where synaptic modification in the form of LTP and LTD is dependent principally on firing rates, in SDTP the exact timing of pre- and postsynaptic activity determines whether a synapse is potentiated or depressed (Markram et al., 1997; Bi and Poo, 1998; Feldman, 2000). In the context of our study, STDP is attractive, since the effects of both strabismus and MD can be explained on the basis of a change in timing of inputs from the affected eye. In the case of strabismus, all inputs from one eye would be systematically delayed or advanced, while in MD, their timing relative to those from the other eye and to postsynaptic activity would be essentially random. However, for strabismus followed by MD, a similar protective effect would be predicted as for traditional competitive models, since STDP replaces spatial with temporal competition.
In summary, our results are to some extent compatible with both competitive and associative mechanisms of synaptic plasticity, and indicate that either (or both) may be involved in the response of V1 neurons to monocular deprivation.
Why is the ‘Protective Effect’ Limited to One Hemisphere?
One possible explanation for the reduced MD effect in strabismic animals is the fact that in cat V1 the contralateral-eye OD bands are wider than the ipsilateral-eye bands (Shatz and Stryker, 1978). Since segregation of OD bands is enhanced in strabismic kittens in terms of both geniculocortical afferents and horizontal connections within the cortex (Löwel and Singer, 1992; Trachtenberg and Stryker, 2001), interactions with cells dominated by the non-deprived eye would have to occur over greater distances in columns where the deprived eye is the contralateral one, if indeed competitive heterocellular interactions played a part in OD plasticity and the response to MD.
A more likely explanation centres on the fact that during early postnatal development, stimulus selective responses first emerge through the contralateral eye. Orientation and OD maps in kittens aged 2–3 weeks are heavily dominated by the contralateral eye, and develop even in binocularly deprived animals (Crair et al., 1998, 2001). Beyond 3 weeks of age, roughly the beginning of the critical period, only a weak contralateral bias persists in normal animals (Crair et al., 1998). Therefore, it is possible that the contralateral visual pathway provides a ‘scaffold’ for the development of orientation columns that is shared by both eyes (Crair et al., 1998). It could be argued that the initially weaker ipsilateral-eye responses are more sensitive to the effects of visual deprivation. The induction of strabismus at P21 would have effectively disconnected the ipsilateral-eye domains from any contralateral-eye influence within days (Trachtenberg and Stryker, 2001), at a time when responses through the ipsilateral eye have not yet fully matured (Crair et al., 1998). Consequently, these responses may have remained rather immature when MD was started at P35, and may have been more susceptible to disruption than the stronger contralateral-eye responses.
Speed of Recovery from MD
While it was originally thought that only reverse occlusion could promote recovery of vision in a previously deprived eye (Hubel and Wiesel, 1970; Blakemore and Van Sluyters, 1974), more recent physiological and behavioural evidence demonstrates that simply reopening that eye is sufficient to allow recovery (Mitchell et al., 1977, 2001; Mitchell, 1988; Mitchell and Gingras, 1998) provided binocular vision has not become discordant, e.g. through strabismus (Kind et al., 2002). In fact, the initial rate of recovery (though not the final extent) is greater in animals with binocular recovery than in those with reverse lid-suture, as predicted by the BCM model of synaptic plasticity (Clothiaux et al., 1991); a visual acuity of 2 cycles/deg is reached within 3 days of reopening of the deprived eye (Mitchell et al., 2001). This finding is in good agreement with our observation that VEPs of MDR kittens recovered to a large extent within 2 days, and cortical territory as assessed by functional imaging within 3–4 days of reopening. This rapid rate of recovery is, if anything, an underestimate. It is highly unlikely that the repeated assessment of the same animals in our longitudinal study accelerated recovery; rather, anaesthesia may have impeded synaptic plasticity for the duration of the imaging session (Rauschecker and Hahn, 1987).
Our finding of a greater rate of recovery in the MD-only animals than in those with prior strabismus fits in with the report by Kind et al. (2002) that eye alignment, and hence correlated binocular activity, is critical for recovery from MD. However, in contrast to predictions based on BCM theory, we found that long-term recovery differed only slightly between the MDR and SR-MD groups. The incomplete recovery observed in the simulations is due to the lack of correlated binocular activity in the SR-MD group, which would have been present in the MDR group and would have aided recovery (Mitchell et al., 2001; Kind et al., 2002). One reason why the BCM theory, by its nature, may fall short of fully explaining recovery from MD in the SR-MD group is that the enhanced OD segregation in that group likely has left a proportion of neurons with afferent input from one eye only, leaving no scope for interactions between two sets of synapses on single neurons but only for heterocellular interactions. Our finding of substantial long-term recovery in the SR-MD animals suggests that such heterocellular interactions do take place.
Our results indicate that loss of responses during MD and recovery of responses upon restoration of binocular vision may involve different mechanisms. This should not be too surprising in light of recent findings concerning putative molecular substrates. Although both loss and recovery of responsiveness are thought to involve NMDA receptors (Gu et al., 1989; Bear et al., 1990), only the loss of responses during MD is dependent on the cAMP/Ca2+ response element-binding protein, which in turn regulates the transcription of plasticity-related genes (Deisseroth et al., 1996; Finkbeiner et al., 1997), while recovery is not (Pham et al., 1999; Mower et al., 2002; Liao et al., 2002). Moreover, protein synthesis is required for OD plasticity during MD (Taha and Stryker, 2002), but, as a preliminary report suggests, not for the recovery from MD (Krahe et al., 2004).
In summary, we have shown that disrupting binocular competition offers only weak protection from the effects of MD, and we have demonstrated very rapid recovery by simply restoring binocular vision. These findings are better explained by associative than by competitive learning rules and can help to direct the search for underlying cellular mechanisms of visual cortical plasticity.
We thank Christopher Howarth and Laurent Watroba for their help with some of the experiments. This work was supported by the Medical Research Council.