0X8-3932 94 s3 OO+ooo ,_ 1990 Pergamon Press plc
,~~ur,,p\,choloq,u. Vol. 28. No 6. pp 517-527. 1990 Pnnted in Great Bntam
MORPHOMETRIC
STUDY OF HUMAN CEREBRAL DEVELOPMENT
CORTEX
PETER R. HUTTENLOCHER* Departmentsof Pediatricsand Neurology
and Committee on Neurobiology, Chicago, IL 60637, U.S.A.
University
of Chicago.
Abstract-Morphometric studies of immature cerebral cortex in humans show developmental changes extending up to the time of adolescence. Growth of dendrites and of synaptic connections occurs during infancy and early childhood. Excess synaptic connections are eliminated during later childhood years. The exuberant connections that occur during infancy may form the anatomical substrate for neural plasticity and for certain types of early learning.
INTRODUCTION RECENT morphometric studies of human cerebral cortex during development provide interesting insights into the anatomieal substrate of emerging cortical functions in the infant and child. They give evidence for a dynamic, changing system, at least into the late childhood years, an age at which cortical structure had been thought to be relatively stable. The quantitative approach has been especially productive in the area of synaptogenesis. However, quantitative data on cortical volume. neuronal number, and dendritic growth also have yielded important findings. As yet, there has been little attempt to correlate developmental changes in these different components of the neuropil. Extensive data are now available for specific cortical areas, especially for primary visual cortex. These make it possible to obtain a fairly comprehensive picture of the changes that occur in cortical neurons during pre- and postnatal development. The present report is concerned with an analysis of extant data, emphasizing those in visual cortex, comparing them to findings in frontal cortex and stressing relationships between dendritic and synaptic growth.
VOLUME
OF VISUAL
CORTEX
(AREA
17)
Most quantitative morphologic data are obtained in terms of densities, i.e. number of structure per unit volume. For these to be meaningful, they have to be correlated with the total volume of the tissue. This is especially important during early development, when
*Address U.S.A.
for correspondence:
Department
of Pediatrics.
517
Box 228. University
of Chicago,
Chicago,
IL 60637,
518
P. R. HUTTENLOCHER
volume expansion of brain is an important factor. Measurements of volume of most cortical regions are difficult if not impossible, due to the fact that structural changes from one area to the other tend to be subtle, and transitions gradual. The primary visual cortex (area 17) forms a notable exception. It is clearly demarcated by its characteristic anatomy. with a very large granular layer, transsected by a prominent bundle of fibers, the stria of Gennari. Measurements of volume of area 17 during development have been made by HUTTENLOCHER et al. [22, 231 and by SAUER et al. [37]. The two studies show similar findings, i.e. rapid expansion of cortical volume during fetal life and during the first four postnatal months. During late childhood there is a small but statistically significant contraction in cortical volume. The values used for the present calculations of total number of neurons (Table 1) and synapses (Fig. 1) are those published by HUTTENLOCHER et al. [22,23]. Some ofthe other quantitative data analyzed in this report, including synapse counts and counts of neuronal number were carried out on the same brains. Tissue shrinkage or swelling during the agonal period and during fixation and embedding therefore is apt to be approximately the same for the volume and synaptic and neuronal density data, making it possible to examine relationships between them.
Table
1. Volume, neuronal density, and total number striate cortex (area 17) at various ages Neurons/mm” Age
28 wk GA Newborn 6 days 2 wk 2 mon. 4 mon. 19 mon. 3& yr 5 yr II yr 13 yr 26 yr 71 yr
Volume
(mm3)
230 920 1270 1330 2720 3800 4170 4430 4310 3620 3620 3790 3850
(x 104) 61.9 9.6 9.4 10.9’ 5.2 3.5 3.0 3.1 2.5’ 2.5 2.9* 3.6 4.2
of neurons
in
Total neurons
( x IO’) 14.2 8.8 11.9 14.5 14.1 13.5 12.5 13.7 10.8 9.1 10.5 14.6 16.2
The cortical volume and neuronal density data are not corrected for tissue shrinkage, since the neuronal density values published by LE~JHA and GAREY [26] do not contain such corrections. Readers interested in volume data corrected for tissue shrinkage are referred to Ref. [23]. *No neuronal density value was available for the exact age. Values for the nearest available ages are shown.
The very early growth of visual cortex, with maximum volume reached at about age 4 months, is somewhat surprising when one considers the size of the brain as a whole, which at 4 months is only about half that of the adult. It is likely that cortical association areas, especially frontal cortex, grow more slowly. Much of the later gain in brain weight may also be related to myelination of subcortical white matter which continues throughout the childhood years [47]. The slight decrease in volume of visual cortex observed during late childhood may reflect certain regressive changes in cerebral cortex that occur late in development, and that are subsequently detailed.
MORPHOMETRIC
STUDY
OF HUMAN
CEREBRAL
5 MONTHS
Fig. 1, Synaptic
density
NEURONAL
CORTEX
DEVELOPMENT
l
= SYNAPSES/mm3
0
= TOTAL
10
20
SYNAPSES
30
50
70
YEARS
and total synapses
NUMBER
in visual cortex as a function
IN VISUAL
of age.
CORTEX
Extensive data on the density of neurons in visual cortex during development have recently been reported by LEUBAand GAREY [26]. These data are especially useful for comparisons of synaptic and neuronal numbers during development (see below), since many of the cell counts were on semithin sections fixed in glutaraldehyde and embedded in Spurr’s medium. with adjacent, identically prepared blocks of tissue used for synapse quantitation [22. 231. Values obtained at ages when cortical volume measurements were also available are listed in Table 1. The density of neurons decreases markedly during fetal and early postnatal age to about development, from about 62 x lo4 neurons per mm3 at 28 weeks gestational 4 x lo4 in the adult. These values become meaningful only when they are related to cortical volume. Comparison with volume data for visual cortex show that total neuronal number remains constant through the life span, at least from gestational age (GA) 28 weeks to about age 70 years (Table 1). By GA 28 weeks all cortical neurons have arrived at the cortical plate where they are densely crowded together. The decrease in packing density observed during development appears to be entirely secondary to expansion of the volume of cortex, which in turn is related to the growth of axons, dendrites, and glia. The data provide no evidence that cortical neurons are lost during normal development or during the adult years. The total numbers of neurons in visual cortex in different brains show fluctuations from about 9 to 16 x 10’ which are not age related. Some of this variability may be related to inaccuracies in cell counts and volume measurements. It may also reflect normal variations in brain size. It is well known that the human brain can be organized into normal functioning units over a fairly large range in brain size, with total brain weight ranging from about 1000 to 1800 g in normal adults.
520
P. R. HUTTEKLOCHER
DENDRITIC
DEVELOPMENT
IN VISUAL
CORTEX
Human cortical neurons are characterized by great complexity and length of dendritic arborizations. The function of cerebral cortex is thought to be closely linked to dendritic morphology. Stunted growth of cortical dendrites has been found in some forms of mental retardation [20,42]. Defective morphology of dendritic spines-the specialized regions on dendrites in which many synaptic contacts are located-has also been reported in brains from the mentally retarded [29, 30, 34, 351. Several aspects of dendritic development have been quantitated in human cerebral cortex, including dendritic spine formation, dendritic length, and complexity of branching patterns MICHEL and CAREY [31] examined the density of spines on apical dendrites of pyramidal neurons in layer III of visual cortex (area 17). They found that the mean number of dendritic spines increases to a maximum of about 79 per 50 pm of dendritic length at about age 5 months. It declines thereafter, and the adult value of about 50 spines/50 pm is reached by age 2 I months. These data per se do not provide information as to whether there is actual loss of spines-and hence presumably of synapses related to spines-after age 5 months, or whether the decrease in spine density is related to progressive elongation of dendrites, the number of spines per neuron remaining constant. Data published by BECKER et al.[4] show that progressive elongation ofdendrites occurs between ages 5 and 21 months. These authors obtained measurements of total dendritic length of pyramidal neurons in layers 111 and V of area 17. Total length of apical dendrites in layer III, the structures studied by Michel and Carey, increased from about 800 pm at age 5 months to 1400 pm at age 24 months. The increase in dendritic length is very close to the figure that would have been predicted on the assumption that the decrease in dendritic spine density that occurs during the same period is due to increase in dendritic length alone. The combined data therefore do not provide evidence for actual loss of spines. The study of BECKER et al. [4] suggests that the time course of dendritic growth varies in different cortical layers. The dendritic trees of pyramidal cells in layer V grow earlier than those in layer III. The mean total length of dendrites for pyramidal cells in layer III is only about 30% of the maximum at birth, while in layer V it is already about 60% of maximum. In addition, a higher degree of dendritic branching was observed in layer V neurons at birth than in layer III neurons. Dendritic development appears to reflect differences in the time course of neurogenesis in general. The birth of neurons and neuronal migration in the lower layers ofcortex precede these events in the upper layers, in what is referred to as the inside-out pattern ofcortical development [41]. A pattern ofearlier dendritic development in the lower cortical layers has also been reported in human motor cortex [28]. The study of BECKER or ul.[4] provides no evidence for regressive changes in dendritic development. There appears to be no significant overall pruning of dendrites on cortical neurons as development progresses, at least up to age 7 years. Elongation of dendrites on pyramidal neurons in visual cortex appears to end by about age 18 months, and dendritic length shows no significant change between that age and 7 years. The study leaves unanswered whether changes in dendritic length occur after 7 years. the oldest age examined by these authors. SYNAPTOGENESIS Dendritic interactions
IN VISUAL
CORTEX
geometry of neurons relates to function of the system through cellular that occur along the dendritic membranes at synapses. Data on synaptogenesis
MORPHOMETRIC
STUDY
OF HUMA&
CEREBRAL
CORTEX
521
DEVELOPMENT
in area 17 are available for all cortical layers, from GA 28 weeks to the adult [22. 231. Synaptic profiles are detectable already in the fetal brain at GA 28 weeks. but synaptic density at that stage is low, about 1 x lo* synapses per mm3 of tissue, or l/3 of the adult value. Synaptic density increases during late fetal life and in early infancy. A sudden spurt occurs between ages 2 and 4 months, when synaptic density almost doubles. After age 1 year synaptic density begins to decline to an adult value about 60% of the maximum, which is reached by age 11 years (Fig. 1). Analysis of synaptic density by cortical layers does not shov+ an inside-out pattern of development, i.e. synaptogenesis in the lower cortical layers does not precede that in the upper ones. This has also been noted by RAKIC et al. in the monkey [36]. In humans, there is some difference between layers. with a trend toward later synaptogenesis in the lower layers: in layer V maximum synaptic density is reached by age 11 months, and in layer VI by 19 months, while density in layer III is maximum already at 4 months. In visual cortex, synaptogenesis in regions concerned with processing of afferent inputs appears to lead that in efferent systems. Availability of synaptic density and volume data on the same brains has made it possible to calculate the approximate total number of synapses in area 17 at various ages (Fig. 1). Only about 1% of the maximum are present at GA 28 weeks, and about 10% at birth. The maximum is reached at postnatal age 8 months. Subsequently, there is a decrease to X&60% of maximum, which is reached by age 11 years. The data indicate a very significant loss of synapses during postnatal development of cerebral cortex. Several derived values can be calculated from extant data on synaptic density, neuronal density, and dendritic length. The average calculated number of synapses per neuron at various ages is given in Fig. 2. This value reaches a maximum of over 15 000 at age 8 months
20,000
Cesl.
Fig. 2. Mean number
MONTHS
of synapses
H w t-m
MEAN FOR ALL LAVER III LAYER IVc
w
LAYER
LAYERS
V
VEARS
per neuron in visual cortex as a whole and in selected cortical as a function of age.
layers
and declines to about 10 000 in the adult, for area I7 as a whole. It is lower for layer IVc, as would be expected from the fact that small neurons predominate in this cortical layer. The average number of synapses per micron of dendritic length also declines during childhood. These values have been calculated, using data for total dendritic length in layers 111 and V
P. R. HIJTENLOCHER
522
provided by BECKERet uI. [4], and synaptic density and neuronal density values for the same layers, obtained from HUTTENLKHER and DE COURTEN [23] and LEUBA and GAREY [26]. The results are shown in Fig. 3. The findings appear to differ between the two cortical layers, with greater crowding of synapses on dendrites during development of layer III than of layer V. Maximum density of synapses on dendrites is reached earlier in layer III, at about age 8 months, vs about 18 months in layer V. However, the values calculated at age 7 years are almost identical for the two cortical layers. These derived values are very gross approximations, since the calculations include the incorrect assumptions that all synapses are axodendritic, and that all neurons in layers III and V are pyramidal cells. However, the errors introduced by these assumptions are likely to remain constant across age, and are unlikely to affect the age-related trends indicated in Fig. 3.
C
= LAYER
III
0
= LAYER
V
0 028 Weeks
NB
2
4
6
8
10
12
MONTHS
Gesl Fig. 3. Estimate of distribution
2
5
10
YEARS
of synapses on dendrites as a function of age
FRONTAL
CORTEX
Available information on the development of cortical areas other than 17 in humans is limited. The only other area for which combined data on neuronal number, synaptic density and dendritic development are available is the middle frontal gyrus. The findings in frontal cortex form an interesting contrast to those in visual cortex. in that they indicate different developmental schedules in the two cortical areas. In frontal cortex, neuronal density in layer III at birth is about 100 000jmm3, similar to that in area 17. The subsequent decline in density is slower. While neuronal density in visual cortex reaches the adult value by about postnatal age 5 months, in frontal cortex it is still 55% above adult mean at age 2 years, and IO”/0 above adult mean at 7 years. No volume data are available for frontal cortex, and it is
MORPHOMETRIC
STUDY
OF HUMAN
CEREBRAL
CORTEX
DEVELOPMENT
53
therefore unknown whether the decrease in density is related to expansion of cortex or to neuronal loss. The adult value for neuronal density in layer III of middle frontal gyrus is about 13 000/mm3 [19, 381. Published values for motor cortex vary from about 10 000 to 26 000 neurons/mm3. Values for both of these frontal regions are considerably smaller than that for visual cortex, which is about 40 000 for layer III [ 12, 26. 391. Dendritic development in middle frontal gyrus also progresses considerably more slowly than in visual cortex. In area 17, the total dendritic length of pyramidal neurons approaches the adult value by age 4 months in layer V and by 18 months in layer III [4]. In contrast. dendritic length in layer III of middle frontal gyrus is only about 50% of the adult at age 2 years [38]. By age 2 years, mean total dendritic length of pyramidal neurons in layer III of middle frontal gyrus is already more than twice that of visual cortex. This. as well as the lower density of neurons undoubtedly reflects the greater complexity and size of pyramidal neurons in the frontal areas. In general, a decrease in density of cortical neurons has been observed as one progresses from simpler to more complex cortical systems, both within a species and phylogenetically [12, 18,32,39,40,44]. The limited available data also suggest that growth of some neural elements, especially dendrites, continues for a longer period of time in frontal cortex. BUELL and COLEMAN [6,73 have found that dendritic elongation continues even in old age, where it may provide compensation for neuronal loss. Synaptic density data during development of middle frontal gyrus are available for layer III only [21]. They show a slower postnatal increase in synaptic density than in visual cortex. Maximum density is reached at about age 1 year, vs age 4 months in visual cortex. Subsequently, synaptic density declines, but again more slowly than in visual cortex. The decrease in synaptic density does not become evident until after age 7 years, at which age synaptic density in visual cortex is already near the adult value, i.e. at about 60% of maximum. Synaptic density in frontal cortex is at adult level by age 16 years. The synaptic density data therefore show regional differences in the time course of cortical synaptogenesis in the human. In that respect, synaptogenesis in human brain appears to differ from that in the monkey, where concurrent overproduction of synapses in diverse regions of cerebral cortex has been found [36]. Available data indicate that synaptic density in adult frontal cortex exceeds that in visual cortex. CRAGG [12, 131, who obtained synapse quantitations in motor and visual cortex in the same brains, found synaptic density in area 4 to be about 50% greater than in area 17. Calculated values derived from neuronal density, dendritic length and synaptic density figures in frontal cortex show trends similar to those in visual cortex, but with a slower time course. The number of synapses per neuron appears to decrease between late childhood and age 16 years. The number of synapses per micron of dendritic length also decreases, the adult value being about 50% of the value obtained at age 2 years. It is not possible to be certain that actual loss of synapses occurs during development of middle frontal gyrus. since no volume data are available. However, the fact that synaptic density declines after neuronal density has stabilized, i.e. after age 7 years, suggests that loss of synapses does occur, similar in extent but later in age than in visual cortex.
DISCUSSION The quantitative anatomy approach to cortical development has clear limitations which need to be spelled out in order to put the present data into proper perspective. Development
524
P. R. HUTTENLOCHER
is an ongoing, dynamic process. The anatomical method is a static one, providing a glimpse of only one point in time for any given case. In the human, the material that is available for study may be affected by such uncontrollable events as effects of the agonal illness on the tissue and postmortem changes [45]. Even at best. the quantitative method gives only mean values. Growth and regressive changes may occur at the same time and may balance each other, in which case quantitation would show no change at all, and would suggest a static rather than a dynamic system. This may well be the case in the adult brain. It is quite likely that new synapses are constantly formed and old connections disappear, a phenomenon that is not reflected in the data on total synapse number in adults, which show little evidence of change. Finally, the anatomical methods are tedious, and they require the availability of samples of brain tissue. Both of these factors make it impractical to study the large number of subjects that would be required for the statistical analyses needed for the detection of small changes. Nevertheless, important information has resulted from morphometric analysis of cerebral cortex. Some widely held notions have been put in question. For example, programmed neuronal death, i.e. the normal loss of neurons during development [I I], is often assumed to be a universal developmental event. In fact, the extant developmental data show no evidence that neuronal death is important during development of the largest and most complex aggregate of neurons, the human cerebral cortex. Interestingly, this could not have been predicted from animal studies. In the mouse, a 30% loss of cortical neurons has been shown during development [18, 191. A smaller neuronal loss, about 15%. has been reported between birth and adulthood in area 17 in the monkey [32]. The data suggest that programmed cell death may be more important in less complex neural systems. It is generally accepted that programmed cell death relates to the failure of developing neurons to find targets for innervation. Target finding may be easier in systems of great complexity, such as human cerebral cortex, where potential targets are available on large populations of nearby neurons. The data for synapse elimination during cortical development show phylogenetic changes which are opposite those for programmed neuronal death. In rodents, there is little evidence for overproduction of synapses. AGHAJANIAN and BLOOM [I] found that synaptic density in the rat brain reaches a maximum at about postnatal day 35, which is less than 10% greater than the adult value. More definite evidence for synapse elimination has been found in the kitten, where the maximum, about 50% above the adult value, is reached at about age 36 days [ 141. The findings in the monkey resemble those in humans, with synaptic density values during infancy (age 6 months) that are 75-95% above the adult value [32,36]. It therefore appears that synapse elimination in cerebral cortex becomes a more important developmental event as the complexity of the system increases. This result conforms to expectations that derive from a theory concerning synaptogenesis proposed by CHANGEUX [8, 91. It is argued that genetic determination of synaptogenesis in complex systems is incomplete, the genome being insufficient in size to allow for specification of all synaptic connections.“The genetic program directs the proper interaction between main categories of neurons. During development within a category several contacts form at the same site; in other words a significant redundancy of the connections exists” [S]. These early synaptic contacts are viewed as being labile (synaptic plasticity). Some connections are stabilized by being incorporated into functioning systems either through the establishment of intrinsic neuronal circuits or through afferent activity transmitted from sense organs. The stabilized synaptic contacts persist, while labile contacts that fail to be incorporated into
MORPHOMETFUC
STUDY
OF HUMAN
CEREBRAL
CORTEX
DEVELOPMENT
525
functioning units regress. The theory thus implies overproduction of synaptic contacts during development, followed by synapse elimination. Recent computer simulations of nerve nets provide evidence that randomly connected systems can develop considerable organization in terms of strength of internal connections and output, provided that they are subjected to non-random input [15]. The anatomical findings in humans are consistent with Changeux’ theory, and they also provide data as to the ages at which synapses are most numerous, as well as the time course of their regression. The postnatal occurrence of these events is of particular interest. It suggests that the anatomical changes may be directly related to the development of cortical functions. including learning, memory and language. Parallels between the anatomical changes and functional development are easily discernible in visual cortex, since the functions of this cortical region are well understood. Relatively few synaptic connections (about 10% of the maximum) are found at birth, a time when visual alertness is as yet very low and visual fixation and following are just beginning to be demonstrable. The rapid burst in synaptogenesis at age 4 months correlates well with a sudden increase in visual alertness at about that time. Binocular interactions, which are clearly dependent on cortical rather than primarily retinal or lateral geniculate function. first appear between ages 4 and 5 months. These include stereopsis, stereoacuity, equalization of the optokinetic nsytagmus to nasotemporal and temporonasal motion, and binocular summation of the light reflex [S, 17, 431. Clinical observations confirm the importance of afferent input for the formation of functional connections during this period. Strabismic amblyopia, i.e. blindness in the squinting eye, and amblyopia due to absence of formed visual inputs, as occurs in congenital cataract, have been observed to occur at this age [ 16.25.273. Synaptic density-and hence presumably the number of unspecified or labile synaptic contacts-continues to be high in visual cortex until at least age 4 years. Functionally, this correlates with persistence of plasticity in the visual system. Most notably, strabismic amblyopia can be reversed during this time, provided that the good eye is occluded and the child is forced to use the squinting eye [2, 31. The persistence of large numbers of labile synapses may provide the anatomical substrate for neural plasticity in the child [24]. If so, then one would expect plasticity to persist later in middle frontal gyrus, where synapse elimination does not begin until about age 7 years. Unfortunately, we have no clear information about the functions of this cortical area. However. it is known that plasticity for some functions persists into late childhood. An example is the ability to recover language functions after large dominant hemisphere lesions. This appears to persist until at least age 8 years [46] and appears to depend on changes in the functional organization of the non-dominant hemisphere. The presence of large numbers of synaptic contacts during postnatal development probably influences cortical function in ways other than through imparting plasticity to the system. Available evidence suggests that at least some of the exuberant synapses may bc active rather than silent. This evidence derives from studies of cerebral metabolism. Cerebral metabolic rate is closely linked to the synaptic activity of cortical neurons. Developmental data show a remarkable parallel between synaptic density and cerebral metabolic rate as determined by positron emission tomography [IO]. The metabolic rate in human cerebral cortex rises rapidly during the first postnatal years. remains above adult levels throughout childhood, and declines to adult values during adolescence. These changes show a rough parallel to the age-related variations in synaptic density observed in frontal cortex, but they correlate less well with changes in synaptic density in primary visual cortex. It may well bc
P. R. HUTENLOCHER
526
that area 17 has an unusual. early developmental schedule that differs from that of most other cortical regions. The presence of excess synaptic activity may well have certain negative effects on brain function in the child. Activity in large pools of unspecified synapses may to some extent interfere with cortical processing. It may account for the well known high susceptibility of the young child to seizures. The high metabolic demand of immature brain also may make it more susceptible to damage related to deficiency of metabolic substrates, including oxygen. Many components of the complex interactions between structure and function in developing cerebral cortex are as yet unknown. Newer methodologies, including in t’i~o imaging techniques [lo] and computerized quantitation [33] are likely to bring significant future advances in this field.
REFERENCES AGHAJANIAX, G. K. and BLOOM, F. E. The formatIon of synaptic junctions in developing rat brain: a quantitative electron microscopic study. Brain Res. 6, 716727, 1967. 2 ASSAF, A. A. The sensitive period: transfer of fixation after occlusion for strabismic amblyopia. Br. J. I
Ophthnlmol.
66, 6470,
1982.
3 AWAYA. S., MIYAKE. Y., IMAIZUMI,Y., SHOISE,Y., KANDA, T. and KOMUKO. K. Amblyopia in man. suggestive of stimulus deprivation amblyopia. Jun. J. Ophthabtd. 17. 69-82. 1973. 4 BECKER, L.‘E., ARMSTRONG, D. L.1 CHAN,‘F. and WOOD, M. M. Dendritic development in human occipital cortical neurons. Deal Brain Res. 13, 117-124, 1984. 5. BIRCH, E. E. and HELD, R. The development ofbinocular summation m human infants. Invest. Onhrhalmol. Vis. Sci. 14, 1103-I 107, 1983. 6. BUELL, S. J. and COLEMAN, P. D. Dendritic growth in the aged human brain and failure of growth in senile dementia. Science 206, 854-856, 1979. 7. BUELL, S. J. and COLEMAN, P. D. Quantitative evidence for selective dendritic growth in normal human aging but not in senile dementia. Brain RFS. 214, 2341, 1981. x. CHANC;EU~. J.-P. and DANCHIN, A. Selective stabilization of developing synapses as a mechanism for the specification of neuronal networks. Mature 264, 705-712. 1976. 9. CHANGEUX, J. P., HEIDMANN, T. and PATTE, P. Learning by selection. In The Biology q/‘learniny, P. MAKLEK and H. S. TEKKACE (Editors), pp. 115-137. Springer-Verlag, New York, 1984. 10. CHUGANI, H. T.. PHELPS. M. E. and MAZZIOTTA. J. C. Positron emission tomography study of human brain functional development. Ann. Neural. 22, 487497, 1987. II. COWAN. W. M. Neuronal death as a regulative mechanism in the control of cell number in the nervous system. In De~:e/npmenr ad Aqiny in fhe Nerrou.s Sysrem, M. ROCKSTEIN (Editor), pp. 1941. Academic Press, New York. 1973. 12. CHAGC, B. G. The density of synapses and neurons in the motor and visual areas of the cerebral cortex. J. Anar. 101, 639-654, 1967. 13. CHAC;G, B. G. The density of synapses and neurons in normal. mentally defective and aging human brains. Brain 98, 81-90, 1975. 14. CKAC;C;.8. G. The development ofsynapses in the visual system ofthecat. J. camp. New-d. 160, 147-166. 1975. 15. EIXLMAI., G. M. Nrurcrl Darwinism. The Theory ofNrurona/ Group Selwrion. Basic Books, New York, 1987. I6 GELBERT. S. S.. HOYT. C. S., JASTKEBSKI,G. and MARC. E. Long-term visual results m bilateral congenital cataracts. Ajn. J. Ophthulmol. 93, 61 S-621, 1982. 17 HELI). R.. BIRCH. E. and GWIAZL)A, J. Stereoacuity of human infants. Proc. natn. Acad. Sci. U.S.A. 77, 5572 5574. 1980. IX HI UMAVK, D.. LUJBA. G. and RABINOWICZ, T. Postnatal development of the mouse cerebral neocortex. IV. Evolution of the total cortical volume. of the population of neurons and glial cells. J. Hirr!fir.wh. 19. 385-393. 197x. IY HI:LMANX, D. and Lu:~A. G. Neuronal death in the development and aging of the cerebral cortex of the mouse. ~~curopurh. uppl. Neurohiol. 9, 297 3 I I. 1983. 20 HUTI:NLOCHI:K. P. R. Dendritic development in neocortex of children with mental defect and infantile spasms. Rirwdo~~~~ 24, 203 2 IO. 1974. changes and effects of aging. 21 Ht TTENLO(‘H~K. P. R. Synaptic density in human frontal cortex. Developmental Hrtrltt RK\. 163, I95 205. 1979. in human visual 22 HUTINLO(‘HTK. P. R.. LX COUKTEN. C.. GAKEY. L. G. and VAN I)FK Loos. H. Synaptogenesis cortex+vidence for synapse elimination during normal development. Nrurosci. LHI. 33, 247 252, 1982.
MORPHOMETRICSTUDYOF HUMAN CEREBRALCORTEXDEVELOPMENT
527
23. HUTTENLOCHER. P. R. and DE COURTER, C. The development of synapses in striate cortex of man. Hum. Neurobiol. 6, 1-9, 1987. 24. HUTTENLOCHER. P. R. Synapse elimination and plasticity in developing human cerebral cortex. Am. J. Mertral Deficiency g&488496, 1984. 25. JACOBSEN,S. G., MOHINDRA, I. and HELD, R. Age ofonset ofamblyopia in infants with esotropia. Dot. Ophfhal. Proc. Ser., Vol. 30, L. MAFFEI (Editor), pp. 21&216. Dr W. Junk. Publishers, The Hague, 1981. 26. LEUBA, G. and GAREY. L. J. Evolution of neuronal numerical density in the developing and aging human visual cortex. Hum. Neurobiol. 6, 11-18, 1987. 27. LEWIS, T. L.. MAURER, D. and BRENT. H. P. Effect on perceptual development of visual deprivation during infancy. Br. J. Ophthalmo/. 70, 214220, 1986. 28. MARIN-PADILLA. M. Prenatal and early postnatal ontogenesis ofthe human motor cortex: a Golgi study. I. The sequential development of the cortical layers. Brain Res. 23, 167-183. 1970. 29. MARIN-PADILLA. M. Structural abnormalities of the cerebral cortex in human chromosomal aberrations. A Golgi study. Brain Res. 44, 625629. 1972. 30. MARIN-PADILLA, M. Pyramidal cell abnormalities in the motor cortex of a child with Down’s syndrome. A Golgi study. J. camp. Neural. 167, 63-82. 1976. 31. MICHEL, A. E. and GAREY. L. J. The development of dendritic spines in the human visual cortex. Hum. Neurobiol. 3, 223-227, 1984. 32. O’KUSKY, J. and COLONNIER, M. Postnatal changes in the number of neurons and synapses in the visual cortex (A17) of the macaque monkey. J. camp. Neural. 210,291-296. 1982. 33. PALDINO, A. M. and PURPURA. D. P. Quantitative analysis of the spatial distribution of axonal and dendritic terminals of hippocampal pyramidal neurons in immature human brain. Expl Neural. 64, 604619, 1979. 34. PURPURA, D. P. Dendritic spine ‘dysgenesis’ and mental retardation. Science 186, 11261128. 1974. 35. PURPURA, D. P. Normal and aberrant neuronal development in the cerebral cortex of human fetus and young infants. In Brain Mechanisms in Mental Retardation. N. A. BUCHWALD and M. A. B. BKAZIEK (Editors). DD. .. 141-169. Academic Press, New York, 1975. 36. RAKIC, P., BOURGEOIS, J.-P., ECKENHOFF. M. F.. ZECEVIC, M. and GOLMAN-RAKIC, P. S. Concurrent overproduction of synapses in diverse regions of primate cerebral cortex. Science 232, 2322234, 1986. 37. SALJER, N., KAMMARADT, G.. KRAUTHAUSEN, I., KRET~CHMANN, H.-T., LANGE, H. W. and WINGEKT, F. Qualitative and quantitative development of the visual cortex in man. J. camp. Neural. 214.44450. 1983. 38. SCHADE, J. P. and VAN GROENIGEN. W. B. Structural organization of the human cerebral cortex, 1. Maturation of the middle frontal gyrus. Acra Anat. 47, 74111. 1961. 39. SHARIFF, G. A. Cell counts in the primate cerebral cortex. J. camp. Neural. 98, 381400. 1953. 40. SHOLL. D. A. A comparative study of the neuronal packing density in the cerebral cortex. J. Anat. 93, 143.158. 1959. 41. SIDMAN, R. L. and RAKIC, P. Neuronal migration, with special reference to developing human brain: a review. Brain Res. 62, l-35, 1973. 42. TAKASHIMA,S., BECKER. L. E., ARMSTKONG.D. A. and CHAN, F. Abnormal neuronal development in the visual cortex of the human fetus and infant with Down’s syndrome. A quantitative and qualitative Golgi study. Brain Res. 225, l-21. 1981. 43. TELLER, D. Y. Scotoptc vision, color vision and stereopsis in infants. Curr. Eye Res. 2, 199-210. 1983. 44. TOWER, D. B. Structural and functional organization of mammalian cerebral cortex: the correlation of neurone density with brain size. J. camp. Neural. 101, 19-51. 1954. 45. WILLIAMS, R. S., FERRANTE, R. J. and CAVINESS.V. S. The rapid Golgi method in clinical neuropathology: the morphologic consequences of suboptimal fixation. J. Neuropafh. exp. Neural. 37, 13-33. 1978. 46. WOODS, B. T. and TEUBER. H. L. Changing patterns of childhood aphasia, Ann. Neural. 3, 273 280. 1978. 47. YAKOVLEV, P. 1. and LECOL‘RS.A.-R. The myelogenetic cycles of regional maturation of the brain. In Rrqionul Derelopmenr oflhe Brain in Ear/x Life. A. MINK~WSKY (Editor), pp. 3-70. Blackwells, Oxford. 1967.
,~~ur,,p\,choloq,u. Vol. 28. No 6. pp 517-527. 1990 Pnnted in Great Bntam
MORPHOMETRIC
STUDY OF HUMAN CEREBRAL DEVELOPMENT
CORTEX
PETER R. HUTTENLOCHER* Departmentsof Pediatricsand Neurology
and Committee on Neurobiology, Chicago, IL 60637, U.S.A.
University
of Chicago.
Abstract-Morphometric studies of immature cerebral cortex in humans show developmental changes extending up to the time of adolescence. Growth of dendrites and of synaptic connections occurs during infancy and early childhood. Excess synaptic connections are eliminated during later childhood years. The exuberant connections that occur during infancy may form the anatomical substrate for neural plasticity and for certain types of early learning.
INTRODUCTION RECENT morphometric studies of human cerebral cortex during development provide interesting insights into the anatomieal substrate of emerging cortical functions in the infant and child. They give evidence for a dynamic, changing system, at least into the late childhood years, an age at which cortical structure had been thought to be relatively stable. The quantitative approach has been especially productive in the area of synaptogenesis. However, quantitative data on cortical volume. neuronal number, and dendritic growth also have yielded important findings. As yet, there has been little attempt to correlate developmental changes in these different components of the neuropil. Extensive data are now available for specific cortical areas, especially for primary visual cortex. These make it possible to obtain a fairly comprehensive picture of the changes that occur in cortical neurons during pre- and postnatal development. The present report is concerned with an analysis of extant data, emphasizing those in visual cortex, comparing them to findings in frontal cortex and stressing relationships between dendritic and synaptic growth.
VOLUME
OF VISUAL
CORTEX
(AREA
17)
Most quantitative morphologic data are obtained in terms of densities, i.e. number of structure per unit volume. For these to be meaningful, they have to be correlated with the total volume of the tissue. This is especially important during early development, when
*Address U.S.A.
for correspondence:
Department
of Pediatrics.
517
Box 228. University
of Chicago,
Chicago,
IL 60637,
518
P. R. HUTTENLOCHER
volume expansion of brain is an important factor. Measurements of volume of most cortical regions are difficult if not impossible, due to the fact that structural changes from one area to the other tend to be subtle, and transitions gradual. The primary visual cortex (area 17) forms a notable exception. It is clearly demarcated by its characteristic anatomy. with a very large granular layer, transsected by a prominent bundle of fibers, the stria of Gennari. Measurements of volume of area 17 during development have been made by HUTTENLOCHER et al. [22, 231 and by SAUER et al. [37]. The two studies show similar findings, i.e. rapid expansion of cortical volume during fetal life and during the first four postnatal months. During late childhood there is a small but statistically significant contraction in cortical volume. The values used for the present calculations of total number of neurons (Table 1) and synapses (Fig. 1) are those published by HUTTENLOCHER et al. [22,23]. Some ofthe other quantitative data analyzed in this report, including synapse counts and counts of neuronal number were carried out on the same brains. Tissue shrinkage or swelling during the agonal period and during fixation and embedding therefore is apt to be approximately the same for the volume and synaptic and neuronal density data, making it possible to examine relationships between them.
Table
1. Volume, neuronal density, and total number striate cortex (area 17) at various ages Neurons/mm” Age
28 wk GA Newborn 6 days 2 wk 2 mon. 4 mon. 19 mon. 3& yr 5 yr II yr 13 yr 26 yr 71 yr
Volume
(mm3)
230 920 1270 1330 2720 3800 4170 4430 4310 3620 3620 3790 3850
(x 104) 61.9 9.6 9.4 10.9’ 5.2 3.5 3.0 3.1 2.5’ 2.5 2.9* 3.6 4.2
of neurons
in
Total neurons
( x IO’) 14.2 8.8 11.9 14.5 14.1 13.5 12.5 13.7 10.8 9.1 10.5 14.6 16.2
The cortical volume and neuronal density data are not corrected for tissue shrinkage, since the neuronal density values published by LE~JHA and GAREY [26] do not contain such corrections. Readers interested in volume data corrected for tissue shrinkage are referred to Ref. [23]. *No neuronal density value was available for the exact age. Values for the nearest available ages are shown.
The very early growth of visual cortex, with maximum volume reached at about age 4 months, is somewhat surprising when one considers the size of the brain as a whole, which at 4 months is only about half that of the adult. It is likely that cortical association areas, especially frontal cortex, grow more slowly. Much of the later gain in brain weight may also be related to myelination of subcortical white matter which continues throughout the childhood years [47]. The slight decrease in volume of visual cortex observed during late childhood may reflect certain regressive changes in cerebral cortex that occur late in development, and that are subsequently detailed.
MORPHOMETRIC
STUDY
OF HUMAN
CEREBRAL
5 MONTHS
Fig. 1, Synaptic
density
NEURONAL
CORTEX
DEVELOPMENT
l
= SYNAPSES/mm3
0
= TOTAL
10
20
SYNAPSES
30
50
70
YEARS
and total synapses
NUMBER
in visual cortex as a function
IN VISUAL
of age.
CORTEX
Extensive data on the density of neurons in visual cortex during development have recently been reported by LEUBAand GAREY [26]. These data are especially useful for comparisons of synaptic and neuronal numbers during development (see below), since many of the cell counts were on semithin sections fixed in glutaraldehyde and embedded in Spurr’s medium. with adjacent, identically prepared blocks of tissue used for synapse quantitation [22. 231. Values obtained at ages when cortical volume measurements were also available are listed in Table 1. The density of neurons decreases markedly during fetal and early postnatal age to about development, from about 62 x lo4 neurons per mm3 at 28 weeks gestational 4 x lo4 in the adult. These values become meaningful only when they are related to cortical volume. Comparison with volume data for visual cortex show that total neuronal number remains constant through the life span, at least from gestational age (GA) 28 weeks to about age 70 years (Table 1). By GA 28 weeks all cortical neurons have arrived at the cortical plate where they are densely crowded together. The decrease in packing density observed during development appears to be entirely secondary to expansion of the volume of cortex, which in turn is related to the growth of axons, dendrites, and glia. The data provide no evidence that cortical neurons are lost during normal development or during the adult years. The total numbers of neurons in visual cortex in different brains show fluctuations from about 9 to 16 x 10’ which are not age related. Some of this variability may be related to inaccuracies in cell counts and volume measurements. It may also reflect normal variations in brain size. It is well known that the human brain can be organized into normal functioning units over a fairly large range in brain size, with total brain weight ranging from about 1000 to 1800 g in normal adults.
520
P. R. HUTTEKLOCHER
DENDRITIC
DEVELOPMENT
IN VISUAL
CORTEX
Human cortical neurons are characterized by great complexity and length of dendritic arborizations. The function of cerebral cortex is thought to be closely linked to dendritic morphology. Stunted growth of cortical dendrites has been found in some forms of mental retardation [20,42]. Defective morphology of dendritic spines-the specialized regions on dendrites in which many synaptic contacts are located-has also been reported in brains from the mentally retarded [29, 30, 34, 351. Several aspects of dendritic development have been quantitated in human cerebral cortex, including dendritic spine formation, dendritic length, and complexity of branching patterns MICHEL and CAREY [31] examined the density of spines on apical dendrites of pyramidal neurons in layer III of visual cortex (area 17). They found that the mean number of dendritic spines increases to a maximum of about 79 per 50 pm of dendritic length at about age 5 months. It declines thereafter, and the adult value of about 50 spines/50 pm is reached by age 2 I months. These data per se do not provide information as to whether there is actual loss of spines-and hence presumably of synapses related to spines-after age 5 months, or whether the decrease in spine density is related to progressive elongation of dendrites, the number of spines per neuron remaining constant. Data published by BECKER et al.[4] show that progressive elongation ofdendrites occurs between ages 5 and 21 months. These authors obtained measurements of total dendritic length of pyramidal neurons in layers 111 and V of area 17. Total length of apical dendrites in layer III, the structures studied by Michel and Carey, increased from about 800 pm at age 5 months to 1400 pm at age 24 months. The increase in dendritic length is very close to the figure that would have been predicted on the assumption that the decrease in dendritic spine density that occurs during the same period is due to increase in dendritic length alone. The combined data therefore do not provide evidence for actual loss of spines. The study of BECKER et al. [4] suggests that the time course of dendritic growth varies in different cortical layers. The dendritic trees of pyramidal cells in layer V grow earlier than those in layer III. The mean total length of dendrites for pyramidal cells in layer III is only about 30% of the maximum at birth, while in layer V it is already about 60% of maximum. In addition, a higher degree of dendritic branching was observed in layer V neurons at birth than in layer III neurons. Dendritic development appears to reflect differences in the time course of neurogenesis in general. The birth of neurons and neuronal migration in the lower layers ofcortex precede these events in the upper layers, in what is referred to as the inside-out pattern ofcortical development [41]. A pattern ofearlier dendritic development in the lower cortical layers has also been reported in human motor cortex [28]. The study of BECKER or ul.[4] provides no evidence for regressive changes in dendritic development. There appears to be no significant overall pruning of dendrites on cortical neurons as development progresses, at least up to age 7 years. Elongation of dendrites on pyramidal neurons in visual cortex appears to end by about age 18 months, and dendritic length shows no significant change between that age and 7 years. The study leaves unanswered whether changes in dendritic length occur after 7 years. the oldest age examined by these authors. SYNAPTOGENESIS Dendritic interactions
IN VISUAL
CORTEX
geometry of neurons relates to function of the system through cellular that occur along the dendritic membranes at synapses. Data on synaptogenesis
MORPHOMETRIC
STUDY
OF HUMA&
CEREBRAL
CORTEX
521
DEVELOPMENT
in area 17 are available for all cortical layers, from GA 28 weeks to the adult [22. 231. Synaptic profiles are detectable already in the fetal brain at GA 28 weeks. but synaptic density at that stage is low, about 1 x lo* synapses per mm3 of tissue, or l/3 of the adult value. Synaptic density increases during late fetal life and in early infancy. A sudden spurt occurs between ages 2 and 4 months, when synaptic density almost doubles. After age 1 year synaptic density begins to decline to an adult value about 60% of the maximum, which is reached by age 11 years (Fig. 1). Analysis of synaptic density by cortical layers does not shov+ an inside-out pattern of development, i.e. synaptogenesis in the lower cortical layers does not precede that in the upper ones. This has also been noted by RAKIC et al. in the monkey [36]. In humans, there is some difference between layers. with a trend toward later synaptogenesis in the lower layers: in layer V maximum synaptic density is reached by age 11 months, and in layer VI by 19 months, while density in layer III is maximum already at 4 months. In visual cortex, synaptogenesis in regions concerned with processing of afferent inputs appears to lead that in efferent systems. Availability of synaptic density and volume data on the same brains has made it possible to calculate the approximate total number of synapses in area 17 at various ages (Fig. 1). Only about 1% of the maximum are present at GA 28 weeks, and about 10% at birth. The maximum is reached at postnatal age 8 months. Subsequently, there is a decrease to X&60% of maximum, which is reached by age 11 years. The data indicate a very significant loss of synapses during postnatal development of cerebral cortex. Several derived values can be calculated from extant data on synaptic density, neuronal density, and dendritic length. The average calculated number of synapses per neuron at various ages is given in Fig. 2. This value reaches a maximum of over 15 000 at age 8 months
20,000
Cesl.
Fig. 2. Mean number
MONTHS
of synapses
H w t-m
MEAN FOR ALL LAVER III LAYER IVc
w
LAYER
LAYERS
V
VEARS
per neuron in visual cortex as a whole and in selected cortical as a function of age.
layers
and declines to about 10 000 in the adult, for area I7 as a whole. It is lower for layer IVc, as would be expected from the fact that small neurons predominate in this cortical layer. The average number of synapses per micron of dendritic length also declines during childhood. These values have been calculated, using data for total dendritic length in layers 111 and V
P. R. HIJTENLOCHER
522
provided by BECKERet uI. [4], and synaptic density and neuronal density values for the same layers, obtained from HUTTENLKHER and DE COURTEN [23] and LEUBA and GAREY [26]. The results are shown in Fig. 3. The findings appear to differ between the two cortical layers, with greater crowding of synapses on dendrites during development of layer III than of layer V. Maximum density of synapses on dendrites is reached earlier in layer III, at about age 8 months, vs about 18 months in layer V. However, the values calculated at age 7 years are almost identical for the two cortical layers. These derived values are very gross approximations, since the calculations include the incorrect assumptions that all synapses are axodendritic, and that all neurons in layers III and V are pyramidal cells. However, the errors introduced by these assumptions are likely to remain constant across age, and are unlikely to affect the age-related trends indicated in Fig. 3.
C
= LAYER
III
0
= LAYER
V
0 028 Weeks
NB
2
4
6
8
10
12
MONTHS
Gesl Fig. 3. Estimate of distribution
2
5
10
YEARS
of synapses on dendrites as a function of age
FRONTAL
CORTEX
Available information on the development of cortical areas other than 17 in humans is limited. The only other area for which combined data on neuronal number, synaptic density and dendritic development are available is the middle frontal gyrus. The findings in frontal cortex form an interesting contrast to those in visual cortex. in that they indicate different developmental schedules in the two cortical areas. In frontal cortex, neuronal density in layer III at birth is about 100 000jmm3, similar to that in area 17. The subsequent decline in density is slower. While neuronal density in visual cortex reaches the adult value by about postnatal age 5 months, in frontal cortex it is still 55% above adult mean at age 2 years, and IO”/0 above adult mean at 7 years. No volume data are available for frontal cortex, and it is
MORPHOMETRIC
STUDY
OF HUMAN
CEREBRAL
CORTEX
DEVELOPMENT
53
therefore unknown whether the decrease in density is related to expansion of cortex or to neuronal loss. The adult value for neuronal density in layer III of middle frontal gyrus is about 13 000/mm3 [19, 381. Published values for motor cortex vary from about 10 000 to 26 000 neurons/mm3. Values for both of these frontal regions are considerably smaller than that for visual cortex, which is about 40 000 for layer III [ 12, 26. 391. Dendritic development in middle frontal gyrus also progresses considerably more slowly than in visual cortex. In area 17, the total dendritic length of pyramidal neurons approaches the adult value by age 4 months in layer V and by 18 months in layer III [4]. In contrast. dendritic length in layer III of middle frontal gyrus is only about 50% of the adult at age 2 years [38]. By age 2 years, mean total dendritic length of pyramidal neurons in layer III of middle frontal gyrus is already more than twice that of visual cortex. This. as well as the lower density of neurons undoubtedly reflects the greater complexity and size of pyramidal neurons in the frontal areas. In general, a decrease in density of cortical neurons has been observed as one progresses from simpler to more complex cortical systems, both within a species and phylogenetically [12, 18,32,39,40,44]. The limited available data also suggest that growth of some neural elements, especially dendrites, continues for a longer period of time in frontal cortex. BUELL and COLEMAN [6,73 have found that dendritic elongation continues even in old age, where it may provide compensation for neuronal loss. Synaptic density data during development of middle frontal gyrus are available for layer III only [21]. They show a slower postnatal increase in synaptic density than in visual cortex. Maximum density is reached at about age 1 year, vs age 4 months in visual cortex. Subsequently, synaptic density declines, but again more slowly than in visual cortex. The decrease in synaptic density does not become evident until after age 7 years, at which age synaptic density in visual cortex is already near the adult value, i.e. at about 60% of maximum. Synaptic density in frontal cortex is at adult level by age 16 years. The synaptic density data therefore show regional differences in the time course of cortical synaptogenesis in the human. In that respect, synaptogenesis in human brain appears to differ from that in the monkey, where concurrent overproduction of synapses in diverse regions of cerebral cortex has been found [36]. Available data indicate that synaptic density in adult frontal cortex exceeds that in visual cortex. CRAGG [12, 131, who obtained synapse quantitations in motor and visual cortex in the same brains, found synaptic density in area 4 to be about 50% greater than in area 17. Calculated values derived from neuronal density, dendritic length and synaptic density figures in frontal cortex show trends similar to those in visual cortex, but with a slower time course. The number of synapses per neuron appears to decrease between late childhood and age 16 years. The number of synapses per micron of dendritic length also decreases, the adult value being about 50% of the value obtained at age 2 years. It is not possible to be certain that actual loss of synapses occurs during development of middle frontal gyrus. since no volume data are available. However, the fact that synaptic density declines after neuronal density has stabilized, i.e. after age 7 years, suggests that loss of synapses does occur, similar in extent but later in age than in visual cortex.
DISCUSSION The quantitative anatomy approach to cortical development has clear limitations which need to be spelled out in order to put the present data into proper perspective. Development
524
P. R. HUTTENLOCHER
is an ongoing, dynamic process. The anatomical method is a static one, providing a glimpse of only one point in time for any given case. In the human, the material that is available for study may be affected by such uncontrollable events as effects of the agonal illness on the tissue and postmortem changes [45]. Even at best. the quantitative method gives only mean values. Growth and regressive changes may occur at the same time and may balance each other, in which case quantitation would show no change at all, and would suggest a static rather than a dynamic system. This may well be the case in the adult brain. It is quite likely that new synapses are constantly formed and old connections disappear, a phenomenon that is not reflected in the data on total synapse number in adults, which show little evidence of change. Finally, the anatomical methods are tedious, and they require the availability of samples of brain tissue. Both of these factors make it impractical to study the large number of subjects that would be required for the statistical analyses needed for the detection of small changes. Nevertheless, important information has resulted from morphometric analysis of cerebral cortex. Some widely held notions have been put in question. For example, programmed neuronal death, i.e. the normal loss of neurons during development [I I], is often assumed to be a universal developmental event. In fact, the extant developmental data show no evidence that neuronal death is important during development of the largest and most complex aggregate of neurons, the human cerebral cortex. Interestingly, this could not have been predicted from animal studies. In the mouse, a 30% loss of cortical neurons has been shown during development [18, 191. A smaller neuronal loss, about 15%. has been reported between birth and adulthood in area 17 in the monkey [32]. The data suggest that programmed cell death may be more important in less complex neural systems. It is generally accepted that programmed cell death relates to the failure of developing neurons to find targets for innervation. Target finding may be easier in systems of great complexity, such as human cerebral cortex, where potential targets are available on large populations of nearby neurons. The data for synapse elimination during cortical development show phylogenetic changes which are opposite those for programmed neuronal death. In rodents, there is little evidence for overproduction of synapses. AGHAJANIAN and BLOOM [I] found that synaptic density in the rat brain reaches a maximum at about postnatal day 35, which is less than 10% greater than the adult value. More definite evidence for synapse elimination has been found in the kitten, where the maximum, about 50% above the adult value, is reached at about age 36 days [ 141. The findings in the monkey resemble those in humans, with synaptic density values during infancy (age 6 months) that are 75-95% above the adult value [32,36]. It therefore appears that synapse elimination in cerebral cortex becomes a more important developmental event as the complexity of the system increases. This result conforms to expectations that derive from a theory concerning synaptogenesis proposed by CHANGEUX [8, 91. It is argued that genetic determination of synaptogenesis in complex systems is incomplete, the genome being insufficient in size to allow for specification of all synaptic connections.“The genetic program directs the proper interaction between main categories of neurons. During development within a category several contacts form at the same site; in other words a significant redundancy of the connections exists” [S]. These early synaptic contacts are viewed as being labile (synaptic plasticity). Some connections are stabilized by being incorporated into functioning systems either through the establishment of intrinsic neuronal circuits or through afferent activity transmitted from sense organs. The stabilized synaptic contacts persist, while labile contacts that fail to be incorporated into
MORPHOMETFUC
STUDY
OF HUMAN
CEREBRAL
CORTEX
DEVELOPMENT
525
functioning units regress. The theory thus implies overproduction of synaptic contacts during development, followed by synapse elimination. Recent computer simulations of nerve nets provide evidence that randomly connected systems can develop considerable organization in terms of strength of internal connections and output, provided that they are subjected to non-random input [15]. The anatomical findings in humans are consistent with Changeux’ theory, and they also provide data as to the ages at which synapses are most numerous, as well as the time course of their regression. The postnatal occurrence of these events is of particular interest. It suggests that the anatomical changes may be directly related to the development of cortical functions. including learning, memory and language. Parallels between the anatomical changes and functional development are easily discernible in visual cortex, since the functions of this cortical region are well understood. Relatively few synaptic connections (about 10% of the maximum) are found at birth, a time when visual alertness is as yet very low and visual fixation and following are just beginning to be demonstrable. The rapid burst in synaptogenesis at age 4 months correlates well with a sudden increase in visual alertness at about that time. Binocular interactions, which are clearly dependent on cortical rather than primarily retinal or lateral geniculate function. first appear between ages 4 and 5 months. These include stereopsis, stereoacuity, equalization of the optokinetic nsytagmus to nasotemporal and temporonasal motion, and binocular summation of the light reflex [S, 17, 431. Clinical observations confirm the importance of afferent input for the formation of functional connections during this period. Strabismic amblyopia, i.e. blindness in the squinting eye, and amblyopia due to absence of formed visual inputs, as occurs in congenital cataract, have been observed to occur at this age [ 16.25.273. Synaptic density-and hence presumably the number of unspecified or labile synaptic contacts-continues to be high in visual cortex until at least age 4 years. Functionally, this correlates with persistence of plasticity in the visual system. Most notably, strabismic amblyopia can be reversed during this time, provided that the good eye is occluded and the child is forced to use the squinting eye [2, 31. The persistence of large numbers of labile synapses may provide the anatomical substrate for neural plasticity in the child [24]. If so, then one would expect plasticity to persist later in middle frontal gyrus, where synapse elimination does not begin until about age 7 years. Unfortunately, we have no clear information about the functions of this cortical area. However. it is known that plasticity for some functions persists into late childhood. An example is the ability to recover language functions after large dominant hemisphere lesions. This appears to persist until at least age 8 years [46] and appears to depend on changes in the functional organization of the non-dominant hemisphere. The presence of large numbers of synaptic contacts during postnatal development probably influences cortical function in ways other than through imparting plasticity to the system. Available evidence suggests that at least some of the exuberant synapses may bc active rather than silent. This evidence derives from studies of cerebral metabolism. Cerebral metabolic rate is closely linked to the synaptic activity of cortical neurons. Developmental data show a remarkable parallel between synaptic density and cerebral metabolic rate as determined by positron emission tomography [IO]. The metabolic rate in human cerebral cortex rises rapidly during the first postnatal years. remains above adult levels throughout childhood, and declines to adult values during adolescence. These changes show a rough parallel to the age-related variations in synaptic density observed in frontal cortex, but they correlate less well with changes in synaptic density in primary visual cortex. It may well bc
P. R. HUTENLOCHER
526
that area 17 has an unusual. early developmental schedule that differs from that of most other cortical regions. The presence of excess synaptic activity may well have certain negative effects on brain function in the child. Activity in large pools of unspecified synapses may to some extent interfere with cortical processing. It may account for the well known high susceptibility of the young child to seizures. The high metabolic demand of immature brain also may make it more susceptible to damage related to deficiency of metabolic substrates, including oxygen. Many components of the complex interactions between structure and function in developing cerebral cortex are as yet unknown. Newer methodologies, including in t’i~o imaging techniques [lo] and computerized quantitation [33] are likely to bring significant future advances in this field.
REFERENCES AGHAJANIAX, G. K. and BLOOM, F. E. The formatIon of synaptic junctions in developing rat brain: a quantitative electron microscopic study. Brain Res. 6, 716727, 1967. 2 ASSAF, A. A. The sensitive period: transfer of fixation after occlusion for strabismic amblyopia. Br. J. I
Ophthnlmol.
66, 6470,
1982.
3 AWAYA. S., MIYAKE. Y., IMAIZUMI,Y., SHOISE,Y., KANDA, T. and KOMUKO. K. Amblyopia in man. suggestive of stimulus deprivation amblyopia. Jun. J. Ophthabtd. 17. 69-82. 1973. 4 BECKER, L.‘E., ARMSTRONG, D. L.1 CHAN,‘F. and WOOD, M. M. Dendritic development in human occipital cortical neurons. Deal Brain Res. 13, 117-124, 1984. 5. BIRCH, E. E. and HELD, R. The development ofbinocular summation m human infants. Invest. Onhrhalmol. Vis. Sci. 14, 1103-I 107, 1983. 6. BUELL, S. J. and COLEMAN, P. D. Dendritic growth in the aged human brain and failure of growth in senile dementia. Science 206, 854-856, 1979. 7. BUELL, S. J. and COLEMAN, P. D. Quantitative evidence for selective dendritic growth in normal human aging but not in senile dementia. Brain RFS. 214, 2341, 1981. x. CHANC;EU~. J.-P. and DANCHIN, A. Selective stabilization of developing synapses as a mechanism for the specification of neuronal networks. Mature 264, 705-712. 1976. 9. CHANGEUX, J. P., HEIDMANN, T. and PATTE, P. Learning by selection. In The Biology q/‘learniny, P. MAKLEK and H. S. TEKKACE (Editors), pp. 115-137. Springer-Verlag, New York, 1984. 10. CHUGANI, H. T.. PHELPS. M. E. and MAZZIOTTA. J. C. Positron emission tomography study of human brain functional development. Ann. Neural. 22, 487497, 1987. II. COWAN. W. M. Neuronal death as a regulative mechanism in the control of cell number in the nervous system. In De~:e/npmenr ad Aqiny in fhe Nerrou.s Sysrem, M. ROCKSTEIN (Editor), pp. 1941. Academic Press, New York. 1973. 12. CHAGC, B. G. The density of synapses and neurons in the motor and visual areas of the cerebral cortex. J. Anar. 101, 639-654, 1967. 13. CHAC;G, B. G. The density of synapses and neurons in normal. mentally defective and aging human brains. Brain 98, 81-90, 1975. 14. CKAC;C;.8. G. The development ofsynapses in the visual system ofthecat. J. camp. New-d. 160, 147-166. 1975. 15. EIXLMAI., G. M. Nrurcrl Darwinism. The Theory ofNrurona/ Group Selwrion. Basic Books, New York, 1987. I6 GELBERT. S. S.. HOYT. C. S., JASTKEBSKI,G. and MARC. E. Long-term visual results m bilateral congenital cataracts. Ajn. J. Ophthulmol. 93, 61 S-621, 1982. 17 HELI). R.. BIRCH. E. and GWIAZL)A, J. Stereoacuity of human infants. Proc. natn. Acad. Sci. U.S.A. 77, 5572 5574. 1980. IX HI UMAVK, D.. LUJBA. G. and RABINOWICZ, T. Postnatal development of the mouse cerebral neocortex. IV. Evolution of the total cortical volume. of the population of neurons and glial cells. J. Hirr!fir.wh. 19. 385-393. 197x. IY HI:LMANX, D. and Lu:~A. G. Neuronal death in the development and aging of the cerebral cortex of the mouse. ~~curopurh. uppl. Neurohiol. 9, 297 3 I I. 1983. 20 HUTI:NLOCHI:K. P. R. Dendritic development in neocortex of children with mental defect and infantile spasms. Rirwdo~~~~ 24, 203 2 IO. 1974. changes and effects of aging. 21 Ht TTENLO(‘H~K. P. R. Synaptic density in human frontal cortex. Developmental Hrtrltt RK\. 163, I95 205. 1979. in human visual 22 HUTINLO(‘HTK. P. R.. LX COUKTEN. C.. GAKEY. L. G. and VAN I)FK Loos. H. Synaptogenesis cortex+vidence for synapse elimination during normal development. Nrurosci. LHI. 33, 247 252, 1982.
MORPHOMETRICSTUDYOF HUMAN CEREBRALCORTEXDEVELOPMENT
527
23. HUTTENLOCHER. P. R. and DE COURTER, C. The development of synapses in striate cortex of man. Hum. Neurobiol. 6, 1-9, 1987. 24. HUTTENLOCHER. P. R. Synapse elimination and plasticity in developing human cerebral cortex. Am. J. Mertral Deficiency g&488496, 1984. 25. JACOBSEN,S. G., MOHINDRA, I. and HELD, R. Age ofonset ofamblyopia in infants with esotropia. Dot. Ophfhal. Proc. Ser., Vol. 30, L. MAFFEI (Editor), pp. 21&216. Dr W. Junk. Publishers, The Hague, 1981. 26. LEUBA, G. and GAREY. L. J. Evolution of neuronal numerical density in the developing and aging human visual cortex. Hum. Neurobiol. 6, 11-18, 1987. 27. LEWIS, T. L.. MAURER, D. and BRENT. H. P. Effect on perceptual development of visual deprivation during infancy. Br. J. Ophthalmo/. 70, 214220, 1986. 28. MARIN-PADILLA. M. Prenatal and early postnatal ontogenesis ofthe human motor cortex: a Golgi study. I. The sequential development of the cortical layers. Brain Res. 23, 167-183. 1970. 29. MARIN-PADILLA. M. Structural abnormalities of the cerebral cortex in human chromosomal aberrations. A Golgi study. Brain Res. 44, 625629. 1972. 30. MARIN-PADILLA, M. Pyramidal cell abnormalities in the motor cortex of a child with Down’s syndrome. A Golgi study. J. camp. Neural. 167, 63-82. 1976. 31. MICHEL, A. E. and GAREY. L. J. The development of dendritic spines in the human visual cortex. Hum. Neurobiol. 3, 223-227, 1984. 32. O’KUSKY, J. and COLONNIER, M. Postnatal changes in the number of neurons and synapses in the visual cortex (A17) of the macaque monkey. J. camp. Neural. 210,291-296. 1982. 33. PALDINO, A. M. and PURPURA. D. P. Quantitative analysis of the spatial distribution of axonal and dendritic terminals of hippocampal pyramidal neurons in immature human brain. Expl Neural. 64, 604619, 1979. 34. PURPURA, D. P. Dendritic spine ‘dysgenesis’ and mental retardation. Science 186, 11261128. 1974. 35. PURPURA, D. P. Normal and aberrant neuronal development in the cerebral cortex of human fetus and young infants. In Brain Mechanisms in Mental Retardation. N. A. BUCHWALD and M. A. B. BKAZIEK (Editors). DD. .. 141-169. Academic Press, New York, 1975. 36. RAKIC, P., BOURGEOIS, J.-P., ECKENHOFF. M. F.. ZECEVIC, M. and GOLMAN-RAKIC, P. S. Concurrent overproduction of synapses in diverse regions of primate cerebral cortex. Science 232, 2322234, 1986. 37. SALJER, N., KAMMARADT, G.. KRAUTHAUSEN, I., KRET~CHMANN, H.-T., LANGE, H. W. and WINGEKT, F. Qualitative and quantitative development of the visual cortex in man. J. camp. Neural. 214.44450. 1983. 38. SCHADE, J. P. and VAN GROENIGEN. W. B. Structural organization of the human cerebral cortex, 1. Maturation of the middle frontal gyrus. Acra Anat. 47, 74111. 1961. 39. SHARIFF, G. A. Cell counts in the primate cerebral cortex. J. camp. Neural. 98, 381400. 1953. 40. SHOLL. D. A. A comparative study of the neuronal packing density in the cerebral cortex. J. Anat. 93, 143.158. 1959. 41. SIDMAN, R. L. and RAKIC, P. Neuronal migration, with special reference to developing human brain: a review. Brain Res. 62, l-35, 1973. 42. TAKASHIMA,S., BECKER. L. E., ARMSTKONG.D. A. and CHAN, F. Abnormal neuronal development in the visual cortex of the human fetus and infant with Down’s syndrome. A quantitative and qualitative Golgi study. Brain Res. 225, l-21. 1981. 43. TELLER, D. Y. Scotoptc vision, color vision and stereopsis in infants. Curr. Eye Res. 2, 199-210. 1983. 44. TOWER, D. B. Structural and functional organization of mammalian cerebral cortex: the correlation of neurone density with brain size. J. camp. Neural. 101, 19-51. 1954. 45. WILLIAMS, R. S., FERRANTE, R. J. and CAVINESS.V. S. The rapid Golgi method in clinical neuropathology: the morphologic consequences of suboptimal fixation. J. Neuropafh. exp. Neural. 37, 13-33. 1978. 46. WOODS, B. T. and TEUBER. H. L. Changing patterns of childhood aphasia, Ann. Neural. 3, 273 280. 1978. 47. YAKOVLEV, P. 1. and LECOL‘RS.A.-R. The myelogenetic cycles of regional maturation of the brain. In Rrqionul Derelopmenr oflhe Brain in Ear/x Life. A. MINK~WSKY (Editor), pp. 3-70. Blackwells, Oxford. 1967.
