- Email: [email protected]
Linking etiologies in humans and animal models: Studies of autism
Reproductive
Toxicology, Vol. I I, Nos. 213, pp. 417422, 1997 Copyright 0 1997 Elsevier Science Inc. Printed in the USA. All rights reserved 0890.6238/97 $17.00 + .OO
PI1 SOS90-6238(96)00156-6
ELSEVIER
LINKING
ETIOLOGIES
IN HUMANS AND ANIMAL MODELS: OF AUTISM
STUDIES
PATRICIAM. RODIER, JENNIFERL. INGRAM, BARBARA TISDALE, and VICTORIAJ. CROOG Department
of Obstetrics
and Gynecology,
University
of Rochester,
Rochester,
New York
Abstract - Thalidomide has been shown to lead to a high rate of autism when exposure occurs during the 20th to 24th d of gestation. Both the critical period and the neurological deficits of the autistic cases indicate that they have sustained injuries to the cranial nerve motor nuclei. To determine whether such lesions characterize other cases of autism, the brain stem of an autistic case was compared to that of a control. The autopsy case showed abnormalities predicted by the thalidomide cases and evidence of shortening of the brain stem, a defect that could have occurred only during neural tube closure. To test whether animals can be similarly injured but remain viable, rats were treated with 350 mg/kg of valproic acid on day 11.5, 12, or 12.5 of gestation. Neuron counts showed reductions of cell numbers in the cranial nerve motor nuclei. Rats with motor neuron deficits also had cerebellar anomalies like those reported in studies of autistic cases, supporting the idea that these animals may 0 1997 Elseviec Science Inc. be a useful model of the developmental injury that initiates autism. Keq’ Words: teratology;
thalidomide;
valproic
acid; brain stem; facial nucleus; superior olive; cerebellum.
INTRODUCTION
Histologic studies suggest surprisingly few abnormalities in the brains of human cases [reviewed in (6)]. The brains are slightly large and grossly normal. Several labs have reported a decrease in cell density in the cerebellum, with Purkinje cells, granule cells, and neurons of the deep nuclei all affected. One lab has reported an increase in cell density in structures of the limbic system. Several MRI studies support the idea that the cerebellum is altered, indicating a reduction in the volume of the posterior lobe of the vermis (7). A new, larger MRI study (8) confirms the cerebellar anomaly and suggests small decreases in each region of the brain stem, with normal volumes for all other brain regions viewed in the midline.
Autism is characterized by impairment of social interaction, deficiency or abnormality of speech development, and limited activities and interests (1). The last category includes such abnormal behaviors as fascination with spinning objects, repetitive stereotypic movements, obsessive interests, and abnormal aversion to change in the environment. Symptoms are present by 30 months of age. Autism occurs with a prevalence of 1.3/1000 in the most complete study of a North American population (2,3), making it one of the most common congenital defects. While there is a strong genetic factor in the disease, it is not heritable in a simple fashion. For example, the concordance rate in monozygotic twins is about 36% (4)-well above what would be expected by chance, yet far below the 100% that should be characteristic of an outcome controlled directly by a single gene. Piven and Folstein (5) have reviewed the twin results, the increased risk for autism in siblings of autistic cases, and the elevated incidence of social, language, and cognitive deficits among parents and siblings of autistic patients, and have concluded that the existing data suggest a genetic liability for autism, which may interact with other genetic factors as well as teratogenic exposures in determining the final outcome.
THALIDOMIDE
AND AUTISM
New information on the causes of autism has appeared from an unexpected source-a study of patients in the Swedish thalidomide registry (9-l 1). Of about 100 patients, five have autism, and all five are from a group of 15 with evidence of exposure during the 20th to 24th d of gestation. Thus, the rate of autism after exposure to thalidomide during this period was l/3. The 20th to 24th d of development fall during the closure of the neural tube. The period also coincides with the production of the first neurons-those that form the motor nuclei of the cranial nerves. Not surprisingly, the autistic patients in the study had evidence of injury to the motor nuclei or their projections. The neurologic abnormalities observed in the five thalidomide autistic cases included the following: three patients had Duane syndrome, a failure of
Address correspondence to Patricia M. Rodier, Department of Obstetrics and Gynecology, University of Rochester Medical Center, Box 668, 601 Elmwood Ave., Rochester, NY 14642. 417
418
Reproductive Toxicology
the VIth cranial nerve (abducens) to innervate the lateral rectus muscle of the eye with subsequent reinnervation of the muscle by the IIIrd cranial nerve (oculomotor); one had gaze paresis (oculomotor palsy); four had Moebius syndrome, a failure of the VIIth cranial nerve (facial) to innervate the facial muscles, often associated with other cranial nerve symptoms; two had abnormal lacrimation, a failure of the neurons of the superior salivatory nucleus (cranial nerve VII) to innervate the lacrimal apparatus with subsequent misinnervation of the structure by neurons that normally supply the submandibular glands. Each autistic patient had ear malformations and hearing deficits. In light of these findings, it is interesting to note that the cranial nerve effects of thalidomide were actually reported soon after the initial discovery of its limb effects (12,13) but failed to attract the interest of researchers. In addition, ear malformations (14) eye motility problems (IS), and Moebius syndrome (16) have been associated with autism before. What is new is that the thalidomide study links all these associations to a brain stem injury, and suggests that the same injury initiates both autism and cranial nerve symptoms. The thalidomide results indicate that autism can be caused by a very early injury to the developing brain. Such an injury could occur by antimitotic action or by disturbance of the normal pattern formation of the rhombomeres, controlled by the Hox genes. The focus of the injury must be the brain stem, because no other parts of the CNS are present to be injured so early. Alterations of later-forming regions, such as the cerebellum, might OCcur as secondary effects. The thalidomide results predict that at least some cases of autism should have injuries to the cranial nerve motor nuclei, but the brain stem tegmentum, where these nuclei lie, has not been evaluated histologically in autistic cases. The thalidomide results also predict that it should be possible to create an animal model of autism by disrupting CNS development during neural tube closure. NEUROANATOMY
OF A HUMAN
CASE
We prepared serial sections of the tegmentum from a control case and one with a diagnosis of autism (17). This case had been followed for 15 years at our institution, and her records show that she met the criteria of DSM IV. Her mother’s history of hospital admissions for addiction to alcohol and amphetamines, and later, other drugs, began after the birth. On interview, the mother could not be certain whether these problems began before the birth or not. An earlier study of the forebrain of the same case had focused on cell numbers in cortical areas related to speech and hearing, and had found no differences (18). The tissue remaining for study did not include the abducens nucleus but did include the region from the oral border of the facial nucleus to about the
Volume 11, Numbers 2/3, 1997
midpoint of the hypoglossal nucleus. The brain stem of the autistic case was abnormal in several ways. Most striking was a reduction in the motor neurons of the facial nucleus from over 9000 in the control to about 400 in the autistic brain. Small clusters of cells were present in the most oral part of the nucleus but no cells at all were present in the middle or caudal levels normally occupied by the nucleus. The superior olive, an auditory relay nucleus, was also missing. That these structures had been absent throughout development was apparent from the patterns of fibers within and around the regions where neurons were missing. The motor nuclei ordinarily are distinguished not only by the presence of giant motor neurons, but by the way axons traveling through the brain stem avoid these early-forming structures. When both neuron nuclei and fibers are stained, this pattern results in the appearance of the motor neurons on a pale, fiber-free field. In the autistic brain the density of fibers of passage was not reduced in the region that should have contained the facial nucleus. Even where small clusters of facial motor neurons were present, axons failed to respect the space around them. The superior olive ordinarily is outlined by distinctively shaped fiber bands. These were not visible in the autistic brain. Figure 1 shows the facial nucleus at the middle of the trapezoid body in the control case and the autistic case. We have measured the cephalo-caudal dimensions of the autopsy cases by counting the number of 10 sections between various landmarks. This exercise suggests that the autistic case is lacking the fifth rhombomere, the same rhombomere that fails to form if the Hoxa-I gene is lacking. The positions of structures in the autistic brain stem are comparable to those described for the Hoxa-I knockout mouse in Carpenter et al. (19) Hoxa-1 knockout mice (20) lack or have major reductions in the abducens nucleus (not available in our tissue), the facial nucleus (near-absent in our case), and the superior olive (absent in our case). All these form from the 5th rhombomere. In addition, Hoxa-1 knockout mice have malformed inner, middle, and external ears (not available for study in our case, but characteristic of the thalidomide cases), and they have a shortened hindbrain, with caudal structures such as the hypoglossal nucleus moved toward the pons. Thus, we propose that the autistic autopsy case may be explained as caused by a defect of Hoxa-I, and are investigating this gene in genomic DNA from the autopsy tissue. If we can identify a defective gene in the autistic patient we have described, that would suggest an identity for at least one of the genes that contributes to the genetic factor in autism. While our working hypothesis is that the autopsy case’s defects are genetic in origin, we must remember that her anomalies are similar to those of the thalidomide cases, and, thus, may reflect an exposure to some injury. Whatever the cause, the defect we have described could
Etiology of autism
l
P. M. RODIERET AL.
419
Fig. 1. The levels in these photographs were matched on the basis of being near the middle of the trapezoid body. The largest part of the facial nucleus lies caudal to this level. The facial nucleus in a control case (a) has more than 80 giant motor neurons at this level, and the region of the nucleus contains few fiber pathways. The same area in an autistic case has fewer than 20 motor neurons
(arrows), but many myelinated fibers are seen passing in all directions. Both cases were cut at 10 km and stained with cresyl violet and Luxol fast blue. have arisen only at the time of neural tube closure, because loss of a horizontal band of tissue like this could not have been repaired to reunite the separated segments of the brain stem at a later stage. AN ANIMAL
MODEL
In most studies of experimentally induced brain damage investigators expose animals to injury at times when neuron production is peaking in forebrain structures. This time period, around day 15 to 17 in the rat, is past the embryonic stage, and thus produces animals without somatic defects. In addition, exposure at this time produces types of brain damage already known to have many behavioral correlates. Another period of interest is early postnatal life, when the process of synaptogenesis is very active in rodents. Again, somatic anomalies are avoided and the investigator has a good idea of what behaviors might be affected. Many investigators assume that an injury as early as that specified by the thalidomide data would be fatal or recover completely. Thus, most of the experimental literature regard-
ing injuries during early brain development addresses problems of neural tube closure and not the condition of the brain in surviving animals. From the thalidomide studies, we suspected that it must be possible to injure the brain stem motor nuclei as the neural tube closes and have the offspring survive without major malformations. This hypothesis could not be tested in rats with thalidomide itself, because thalidomide does not have the same teratologic effects in rodents as in primates (21). Instead, we used valproic acid (VPA) to injure the brain stem in utero (17). VPA’s somatic effects are similar to thalidomide’s (22), its brain-damaging effects are evident in humans (23), and it has been linked to autism (24). Neuron numbers were reduced significantly by exposure at a variety of treatment times, but the animals were robust. They survived to maturity and the brain stem deficits persisted in adults. Counting the day of conception as embryonic day 1, VPA affected only the earliest forming nuclei on day 11 S, the day the neural tube closes. The motor nucleus of trigeminal and the hypoglossal were reduced significantly by a single dose of 350 mg/kg. The same treatment on day 12 resulted in significant loss of
420
Reproductive Toxicology
neurons in the abducens nucleus, in addition to the trigeminal and hypoglossal. Treatment with VPA on day 12.5 affected all the nuclei listed and the oculomotor, as well. The facial, which forms slightly later than the other motor nuclei, was not affected at any of the treatment times. Figure 2 compares the hypoglossal nucleus in a representative control and a rat exposed to VPA on day 11.5. Only a few other groups of neurons form as early as the somatic motor nuclei of the cranial nerves-these include the mesencephalic nucleus of trigeminal, the locus ceruleus, and the dorsal motor nucleus of the vagus. Counts of these groups showed no effect of VPA exposure at any treatment time. In a new unpublished study we have found that measures of distances between landmarks in the brain stems of VPA rats show significant shortening of the region between the facial nucleus and the hypoglossal in every treatment group, while some distances anterior to the facial nucleus are significantly increased. This effect is like the shortening observed in the human case. VPA’s mechanism of action is not known, but the results of these studies are consistent with an antimitotic action of the teratogen. This would explain the extremely specific cell loss observed at different treatment times. If this is the way VPA works, it is not uniformly effective on all cell types, but selective in some way for somatic motor neurons. The sparing of other neurons could be due to differences in the sensitivity of cells to VPA damage, or to differences in the ability of cells to recover from VPA damage. Another explanation of the pattern of
Volume 11, Numbers 213, 1997
VPA effects would be that the drug somehow alters pattern formation in the basal plate, disrupting the conditions necessary for neuron production in the rhombomeres, rather than injuring proliferating neurons directly. Rats treated with VPA share many features with both the thalidomide-induced cases of autism and the autopsy case we have described. In addition to their anatomic similarity, they were injured at the same stage of development with a drug that has recently been associated with human autism. They confirm that injuries like those characteristic of the thalidomide cases can be induced during closure of the neural tube. It would be interesting to compare the behavior of these animals to that of controls on measures indicative of autism. Unfortunately, translating the diagnostic criteria used in humans to animals is not an easy task. It is likely that VPA-exposed animals would be abnormal on many behavioral measures, but most measures can be affected by many different brain injuries. For example, social behaviors are altered by prenatal mercury exposure, prenatal lead exposure, maternal deprivation, and lesions of the amygdala. Because there is no reason to think that the injuries after these treatments are similar, we have to conclude that social behaviors are sensitive to many different brain injuries. Thus, it is not clear what a positive result in the VPA-exposed animals would mean, beyond indicating that the treatment causes brain damage. What is needed is some set of behavioral tests that discriminate between autism and other kinds of brain damage. This will require further study of human cases as well as animals. We are collaborating with clinicians to find appropriate tasks. CEREBELLAR EFFECTS IN THE ANIMAL MODEL COMPARED TO HUMAN CASES
Fig. 2. Exposure to VPA reduced the number of hypoglossal neurons at each treatment time employed in these studies. An example of a nucleus from a control (a) and one from an animal treated on day 11.5 (b) are shown. The rat brains were cut at 10 Pm and stained with cresyl violet and Luxol fast blue.
To test the adequacy of the VPA model as representing an injury like the one underlying autism, we turned to nonbehavioral features of the disease as sources of comparison. In rats exposed to VPA on day 12.5 and controls, we compared Purkinje cell number, granule cell number, neuron number in the deep nuclei of the cerebellar vermis, and linear measures of the size of the lobules. Because all cerebellar neurons form after day 12.5, the neurons counted in this study could not have been exposed to VPA. Thus, differences in cerebellar cell counts should occur only if the cerebellum is altered secondarily to the motor neuron lesion created by VPA exposure. In fact, cerebellar neuron numbers were reduced in the VPA-exposed rats. Purkinje cells were significantly fewer in lobules VI-VIII and IX, but not in the anterior lobes (IV-V) of the vermis. Among the deep nuclei, the nucleus interpositus (corresponding to the globose and emboliform nuclei of the human) was sig-
Etiology
421
of autism 0 P. M. RODIER ET AL.
nificantly reduced, but not the fastigial or dentate nuclei. In human cases, the dentate is spared and the fastigial, globose, and emboliform are injured (6). We found no significant reductions in cell density in vermian samples of the granule cells. A study of lobular volumes is not complete, but linear measures of the lobules show a reduction in the size of the most posterior ones, and normal size in the anterior ones. Taken together, these results suggest that: 1) most of the cerebellar effects observed in human cases of autism appear in the VPA-exposed rats as well. 2) In rats, as in humans, the effects are most prominent in the posterior lobe of the cerebellum. 3) The changes in cerebellar anatomy of human cases probably are secondary effects of loss of neurons in the cranial nerve motor nuclei. Most importantly, we believe that the presence of cerebellar effects in animals treated to injure the cranial nerve motor nuclei supports the contention that such animals are valuable as a model of autism. Not only do they model the effects on the brain stem nuclei observed to be involved in the thalidomide cases and our human case, but they resemble the histology of autopsy reports and the gross features of living patients as described by other investigators. SUMMARY Uncovering the etiologies of human developmental defects is always difficult. The case of autism has been particularly perplexing because the disease appears to involve both genetic and environmental factors and multiple etiologies. Under such conditions, studies of preand postnatal events in cases produce weak associations, at best. Screening the whole brain for abnormalities is just as problematic. Even when abnormalities appear, it may not be possible to determine whether they are related to the symptoms or incidental to the actual cause. A third approach-trying to guess the cause from the behavioral symptoms-is especially difficult when the human symptoms are described at a level so far removed from brain function. For example, disorders of communication, such as lack of facial gestures, can be interpreted to arise from almost any part of the CNS, from the brain stem to the frontal cortex. Without some information to guide the search, investigators cannot make use of all the wonderful techniques at their disposal, because they do not know what to test. The unexpected association of autism with thalidomide provides a new direction for research on the causes of the disease. It is the human equivalent of a laboratory experiment in which the time of exposure to the teratogen was varied. By telling us when an injury that initiates autism occurs, the thalidomide study tells us what was injured and thus, where to look. It not only allows us the
possibility of modeling the injury in animals but provides new ideas for evaluation of human cases. The thalidomide study suggests that the neuroanatomy of the autistic brain stem should be altered, and that hypothesis can be confirmed. The symptoms of the thalidomide cases and the anatomy of the autopsy case correspond to the deficits in Hoxa-I knockout mice, and, thus, suggest that mutations of developmental genes are likely to be the ones contributing to the genetic factor in autism. The thalidomide results suggest that we might want to examine much lower-level functions of the nervous system in autistic patients, and consider whether disturbances in communication and social behaviors might be mediated by more basic deficits in sensory and motor function. It is also important to investigate whether injuries to the closing neural tube generate developmental changes in later-forming parts of the brain. This appears to be the case in the cerebellum, and other regions might be affected as well. If investigators are to work back and forth between human and animal studies, one thing is clear: we cannot succeed if we abandon some levels analysis and concentrate on just a few. Autism makes a good example of a subject for which information has come from many levels of analysis. Only astute clinicians with knowledge of neurology and teratology could have carried out the thalidomide study. Psychiatrists and psychologists have defined the syndrome of autism by distinguishing its behavioral features from those of other mental illnesses. Classical embryologists have contributed to our understanding of the development of the hindbrain and geneticists and molecular biologists are building on that foundation. Teratologists have studied the effects of thalidomide and valproic acid for many reasons, and their results now turn out to be important in ways they could not have predicted. We need research on all levels to unravel the causes of developmental defects. Acknowledgments NS24287, ROl R824758.
-This work was supported by NIH Grants ROl AA08666, and P30 ES01247, and EPA Grant
REFERENCES 1. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 4th ed. Washington, DC; 1994. 2. Bryson SE, Clark BS, Smith IM. First report of a Canadian epidemiological study of autistic syndromes. J Child Psycho1 Psychiatry. 1988;29:43345. 3. Bryson SE. personal
communication.
4. Folstein SE, Rutter ML. Infantile autism: a genetic study of 21 twin pairs. J Child Psycho1 Psychiatry. 1977;18:297-321. 5. Piven J, Folstein SE. The genetics of autism. In Bauman ML, Kemper TL, eds. The neurobiology of autism. Baltimore: Johns Hopkins University Press: 1994: 1844. 6. Bauman ML, Kemper TL. Neuroanatomic
observations
in autism.
422
7.
8.
9. 10.
11.
12. 13. 14. 15. 16.
Reproductive
Toxicology
In Bauman ML, Kemper TL, eds. The neurobiology of autism. Baltimore: Johns Hopkins University Press; 1994: 11945. Courchesne E, Townsend J, Saitch 0. The brain in infantile autism: posterior fossa structures are abnormal. Neurology. 1994;44:21428. Hashimoto T, Tayama M, Murakawa K, Yoshimoto T, Miyazaki M, Harada M, Kuroda Y. Development of the brainstem and cerebellum in autistic patients. J Autism Dev Disord. 1995;25: l-18. Miller MT, Stromland K. Thalidomide embryopathy: an insight into autism? Teratology. 1993;47:387-8. Stromland K, Nordin V, Miller MT, Akerstrom B, Gillberg C. Autism in thalidomide embryopathy: a population study. Dev Med Child Neurol. 1994;36:351-6. Miller MT. Thalidomide embryopathy: a model for the study of congenital incomitant horizontal strabismus. Tram Am Ophthalmol Sot. 1991;89:623-74. d’Avignon M, Barr B. Ear abnormalities and cranial nerve palsies in thalidomide children. Arch Otolaryngol. 1964;80: 136-40. Lenz W. Thalidomide and congenital abnormalities. Lancet 1962; 1:271-2. Walker HA. Incidence of minor physical anomaly in autism. J Aut Child Schiz. 1977;7: 165-76. Scharre JE, Creedon MP. Assessment of visual function in autistic children. Optom Vis Sci. 1992;69:433-9. Gillberg C, Steffenberg S. Autistic behavior in Moebius syndrome. Acta Pediatr Stand. 1989;78:31&6.
Volume
11, Numbers
2/3, 1997
17. Rodier PM, Ingram, JL, Tisdale B, Nelson S, Roman0 J. An embryological origin for autism: developmental anomalies of the cranial nerve motor nuclei. J Comp Neurol. 1996;370:247-261. 18. Coleman PD, Roman0 J, Lapham LW, Simon W. Cell counts in the cerebral cortex of an autistic patient. J Autism Dev Disord. 1985; 15:245-55. 19. Carpenter EM, Goddard JM, Chisaka 0, Manley NR, Capecchi MR. Loss of Hex-AI (Hex-1.6) function results in the reorganization of the murine hindbrain. Development. 1993; 118: 1063-75. 20. Mark M, Lufkin T, Vonesch J-L, Ruberte E, Ohvo, J-L, DollC P, Gorry P, Lumsden A, Chambon P. Two rhombomeres are altered in Hoxa-I mutant mice. Development. 1993;119:319-38. 21. Schumacher HJ, Terapane J, Jordan RL, Wilson JG. The teratogenie activity of a thalidomide analogue, EM12, in rabbits, rats, and monkeys. Teratology. 1968;5:233%40. 22. Binkerd PE, Rowland JM, Nau H, Hendrickx AG. Evaluation of valproic acid (VPA) developmental toxicity and pharmacokinetics in Sprague-Dawley rats. Fundam Appl Toxicol. 1988;11:485-93. 23. Ardinger HH, Atkin JF, Blackston, RD, Elsas LJ, Clarren SK, Livingstone S, Flannery DB, Pellock JM, Harrod MJ, Lammer EJ, Majewski F, Schinzel A, Toriello HV, Hanson JW. Verification of the fetal valproate syndrome phenotype. Am J Med Genet. 1988; 29:171-85. 24. Christianson AL, Chesler N, Kromberg JGR. Fetal valproate syndrome: clinical and neurodevelopmental features in two sibling pairs. Dev Med Child Neural. 1994;36:357-69.
Toxicology, Vol. I I, Nos. 213, pp. 417422, 1997 Copyright 0 1997 Elsevier Science Inc. Printed in the USA. All rights reserved 0890.6238/97 $17.00 + .OO
PI1 SOS90-6238(96)00156-6
ELSEVIER
LINKING
ETIOLOGIES
IN HUMANS AND ANIMAL MODELS: OF AUTISM
STUDIES
PATRICIAM. RODIER, JENNIFERL. INGRAM, BARBARA TISDALE, and VICTORIAJ. CROOG Department
of Obstetrics
and Gynecology,
University
of Rochester,
Rochester,
New York
Abstract - Thalidomide has been shown to lead to a high rate of autism when exposure occurs during the 20th to 24th d of gestation. Both the critical period and the neurological deficits of the autistic cases indicate that they have sustained injuries to the cranial nerve motor nuclei. To determine whether such lesions characterize other cases of autism, the brain stem of an autistic case was compared to that of a control. The autopsy case showed abnormalities predicted by the thalidomide cases and evidence of shortening of the brain stem, a defect that could have occurred only during neural tube closure. To test whether animals can be similarly injured but remain viable, rats were treated with 350 mg/kg of valproic acid on day 11.5, 12, or 12.5 of gestation. Neuron counts showed reductions of cell numbers in the cranial nerve motor nuclei. Rats with motor neuron deficits also had cerebellar anomalies like those reported in studies of autistic cases, supporting the idea that these animals may 0 1997 Elseviec Science Inc. be a useful model of the developmental injury that initiates autism. Keq’ Words: teratology;
thalidomide;
valproic
acid; brain stem; facial nucleus; superior olive; cerebellum.
INTRODUCTION
Histologic studies suggest surprisingly few abnormalities in the brains of human cases [reviewed in (6)]. The brains are slightly large and grossly normal. Several labs have reported a decrease in cell density in the cerebellum, with Purkinje cells, granule cells, and neurons of the deep nuclei all affected. One lab has reported an increase in cell density in structures of the limbic system. Several MRI studies support the idea that the cerebellum is altered, indicating a reduction in the volume of the posterior lobe of the vermis (7). A new, larger MRI study (8) confirms the cerebellar anomaly and suggests small decreases in each region of the brain stem, with normal volumes for all other brain regions viewed in the midline.
Autism is characterized by impairment of social interaction, deficiency or abnormality of speech development, and limited activities and interests (1). The last category includes such abnormal behaviors as fascination with spinning objects, repetitive stereotypic movements, obsessive interests, and abnormal aversion to change in the environment. Symptoms are present by 30 months of age. Autism occurs with a prevalence of 1.3/1000 in the most complete study of a North American population (2,3), making it one of the most common congenital defects. While there is a strong genetic factor in the disease, it is not heritable in a simple fashion. For example, the concordance rate in monozygotic twins is about 36% (4)-well above what would be expected by chance, yet far below the 100% that should be characteristic of an outcome controlled directly by a single gene. Piven and Folstein (5) have reviewed the twin results, the increased risk for autism in siblings of autistic cases, and the elevated incidence of social, language, and cognitive deficits among parents and siblings of autistic patients, and have concluded that the existing data suggest a genetic liability for autism, which may interact with other genetic factors as well as teratogenic exposures in determining the final outcome.
THALIDOMIDE
AND AUTISM
New information on the causes of autism has appeared from an unexpected source-a study of patients in the Swedish thalidomide registry (9-l 1). Of about 100 patients, five have autism, and all five are from a group of 15 with evidence of exposure during the 20th to 24th d of gestation. Thus, the rate of autism after exposure to thalidomide during this period was l/3. The 20th to 24th d of development fall during the closure of the neural tube. The period also coincides with the production of the first neurons-those that form the motor nuclei of the cranial nerves. Not surprisingly, the autistic patients in the study had evidence of injury to the motor nuclei or their projections. The neurologic abnormalities observed in the five thalidomide autistic cases included the following: three patients had Duane syndrome, a failure of
Address correspondence to Patricia M. Rodier, Department of Obstetrics and Gynecology, University of Rochester Medical Center, Box 668, 601 Elmwood Ave., Rochester, NY 14642. 417
418
Reproductive Toxicology
the VIth cranial nerve (abducens) to innervate the lateral rectus muscle of the eye with subsequent reinnervation of the muscle by the IIIrd cranial nerve (oculomotor); one had gaze paresis (oculomotor palsy); four had Moebius syndrome, a failure of the VIIth cranial nerve (facial) to innervate the facial muscles, often associated with other cranial nerve symptoms; two had abnormal lacrimation, a failure of the neurons of the superior salivatory nucleus (cranial nerve VII) to innervate the lacrimal apparatus with subsequent misinnervation of the structure by neurons that normally supply the submandibular glands. Each autistic patient had ear malformations and hearing deficits. In light of these findings, it is interesting to note that the cranial nerve effects of thalidomide were actually reported soon after the initial discovery of its limb effects (12,13) but failed to attract the interest of researchers. In addition, ear malformations (14) eye motility problems (IS), and Moebius syndrome (16) have been associated with autism before. What is new is that the thalidomide study links all these associations to a brain stem injury, and suggests that the same injury initiates both autism and cranial nerve symptoms. The thalidomide results indicate that autism can be caused by a very early injury to the developing brain. Such an injury could occur by antimitotic action or by disturbance of the normal pattern formation of the rhombomeres, controlled by the Hox genes. The focus of the injury must be the brain stem, because no other parts of the CNS are present to be injured so early. Alterations of later-forming regions, such as the cerebellum, might OCcur as secondary effects. The thalidomide results predict that at least some cases of autism should have injuries to the cranial nerve motor nuclei, but the brain stem tegmentum, where these nuclei lie, has not been evaluated histologically in autistic cases. The thalidomide results also predict that it should be possible to create an animal model of autism by disrupting CNS development during neural tube closure. NEUROANATOMY
OF A HUMAN
CASE
We prepared serial sections of the tegmentum from a control case and one with a diagnosis of autism (17). This case had been followed for 15 years at our institution, and her records show that she met the criteria of DSM IV. Her mother’s history of hospital admissions for addiction to alcohol and amphetamines, and later, other drugs, began after the birth. On interview, the mother could not be certain whether these problems began before the birth or not. An earlier study of the forebrain of the same case had focused on cell numbers in cortical areas related to speech and hearing, and had found no differences (18). The tissue remaining for study did not include the abducens nucleus but did include the region from the oral border of the facial nucleus to about the
Volume 11, Numbers 2/3, 1997
midpoint of the hypoglossal nucleus. The brain stem of the autistic case was abnormal in several ways. Most striking was a reduction in the motor neurons of the facial nucleus from over 9000 in the control to about 400 in the autistic brain. Small clusters of cells were present in the most oral part of the nucleus but no cells at all were present in the middle or caudal levels normally occupied by the nucleus. The superior olive, an auditory relay nucleus, was also missing. That these structures had been absent throughout development was apparent from the patterns of fibers within and around the regions where neurons were missing. The motor nuclei ordinarily are distinguished not only by the presence of giant motor neurons, but by the way axons traveling through the brain stem avoid these early-forming structures. When both neuron nuclei and fibers are stained, this pattern results in the appearance of the motor neurons on a pale, fiber-free field. In the autistic brain the density of fibers of passage was not reduced in the region that should have contained the facial nucleus. Even where small clusters of facial motor neurons were present, axons failed to respect the space around them. The superior olive ordinarily is outlined by distinctively shaped fiber bands. These were not visible in the autistic brain. Figure 1 shows the facial nucleus at the middle of the trapezoid body in the control case and the autistic case. We have measured the cephalo-caudal dimensions of the autopsy cases by counting the number of 10 sections between various landmarks. This exercise suggests that the autistic case is lacking the fifth rhombomere, the same rhombomere that fails to form if the Hoxa-I gene is lacking. The positions of structures in the autistic brain stem are comparable to those described for the Hoxa-I knockout mouse in Carpenter et al. (19) Hoxa-1 knockout mice (20) lack or have major reductions in the abducens nucleus (not available in our tissue), the facial nucleus (near-absent in our case), and the superior olive (absent in our case). All these form from the 5th rhombomere. In addition, Hoxa-1 knockout mice have malformed inner, middle, and external ears (not available for study in our case, but characteristic of the thalidomide cases), and they have a shortened hindbrain, with caudal structures such as the hypoglossal nucleus moved toward the pons. Thus, we propose that the autistic autopsy case may be explained as caused by a defect of Hoxa-I, and are investigating this gene in genomic DNA from the autopsy tissue. If we can identify a defective gene in the autistic patient we have described, that would suggest an identity for at least one of the genes that contributes to the genetic factor in autism. While our working hypothesis is that the autopsy case’s defects are genetic in origin, we must remember that her anomalies are similar to those of the thalidomide cases, and, thus, may reflect an exposure to some injury. Whatever the cause, the defect we have described could
Etiology of autism
l
P. M. RODIERET AL.
419
Fig. 1. The levels in these photographs were matched on the basis of being near the middle of the trapezoid body. The largest part of the facial nucleus lies caudal to this level. The facial nucleus in a control case (a) has more than 80 giant motor neurons at this level, and the region of the nucleus contains few fiber pathways. The same area in an autistic case has fewer than 20 motor neurons
(arrows), but many myelinated fibers are seen passing in all directions. Both cases were cut at 10 km and stained with cresyl violet and Luxol fast blue. have arisen only at the time of neural tube closure, because loss of a horizontal band of tissue like this could not have been repaired to reunite the separated segments of the brain stem at a later stage. AN ANIMAL
MODEL
In most studies of experimentally induced brain damage investigators expose animals to injury at times when neuron production is peaking in forebrain structures. This time period, around day 15 to 17 in the rat, is past the embryonic stage, and thus produces animals without somatic defects. In addition, exposure at this time produces types of brain damage already known to have many behavioral correlates. Another period of interest is early postnatal life, when the process of synaptogenesis is very active in rodents. Again, somatic anomalies are avoided and the investigator has a good idea of what behaviors might be affected. Many investigators assume that an injury as early as that specified by the thalidomide data would be fatal or recover completely. Thus, most of the experimental literature regard-
ing injuries during early brain development addresses problems of neural tube closure and not the condition of the brain in surviving animals. From the thalidomide studies, we suspected that it must be possible to injure the brain stem motor nuclei as the neural tube closes and have the offspring survive without major malformations. This hypothesis could not be tested in rats with thalidomide itself, because thalidomide does not have the same teratologic effects in rodents as in primates (21). Instead, we used valproic acid (VPA) to injure the brain stem in utero (17). VPA’s somatic effects are similar to thalidomide’s (22), its brain-damaging effects are evident in humans (23), and it has been linked to autism (24). Neuron numbers were reduced significantly by exposure at a variety of treatment times, but the animals were robust. They survived to maturity and the brain stem deficits persisted in adults. Counting the day of conception as embryonic day 1, VPA affected only the earliest forming nuclei on day 11 S, the day the neural tube closes. The motor nucleus of trigeminal and the hypoglossal were reduced significantly by a single dose of 350 mg/kg. The same treatment on day 12 resulted in significant loss of
420
Reproductive Toxicology
neurons in the abducens nucleus, in addition to the trigeminal and hypoglossal. Treatment with VPA on day 12.5 affected all the nuclei listed and the oculomotor, as well. The facial, which forms slightly later than the other motor nuclei, was not affected at any of the treatment times. Figure 2 compares the hypoglossal nucleus in a representative control and a rat exposed to VPA on day 11.5. Only a few other groups of neurons form as early as the somatic motor nuclei of the cranial nerves-these include the mesencephalic nucleus of trigeminal, the locus ceruleus, and the dorsal motor nucleus of the vagus. Counts of these groups showed no effect of VPA exposure at any treatment time. In a new unpublished study we have found that measures of distances between landmarks in the brain stems of VPA rats show significant shortening of the region between the facial nucleus and the hypoglossal in every treatment group, while some distances anterior to the facial nucleus are significantly increased. This effect is like the shortening observed in the human case. VPA’s mechanism of action is not known, but the results of these studies are consistent with an antimitotic action of the teratogen. This would explain the extremely specific cell loss observed at different treatment times. If this is the way VPA works, it is not uniformly effective on all cell types, but selective in some way for somatic motor neurons. The sparing of other neurons could be due to differences in the sensitivity of cells to VPA damage, or to differences in the ability of cells to recover from VPA damage. Another explanation of the pattern of
Volume 11, Numbers 213, 1997
VPA effects would be that the drug somehow alters pattern formation in the basal plate, disrupting the conditions necessary for neuron production in the rhombomeres, rather than injuring proliferating neurons directly. Rats treated with VPA share many features with both the thalidomide-induced cases of autism and the autopsy case we have described. In addition to their anatomic similarity, they were injured at the same stage of development with a drug that has recently been associated with human autism. They confirm that injuries like those characteristic of the thalidomide cases can be induced during closure of the neural tube. It would be interesting to compare the behavior of these animals to that of controls on measures indicative of autism. Unfortunately, translating the diagnostic criteria used in humans to animals is not an easy task. It is likely that VPA-exposed animals would be abnormal on many behavioral measures, but most measures can be affected by many different brain injuries. For example, social behaviors are altered by prenatal mercury exposure, prenatal lead exposure, maternal deprivation, and lesions of the amygdala. Because there is no reason to think that the injuries after these treatments are similar, we have to conclude that social behaviors are sensitive to many different brain injuries. Thus, it is not clear what a positive result in the VPA-exposed animals would mean, beyond indicating that the treatment causes brain damage. What is needed is some set of behavioral tests that discriminate between autism and other kinds of brain damage. This will require further study of human cases as well as animals. We are collaborating with clinicians to find appropriate tasks. CEREBELLAR EFFECTS IN THE ANIMAL MODEL COMPARED TO HUMAN CASES
Fig. 2. Exposure to VPA reduced the number of hypoglossal neurons at each treatment time employed in these studies. An example of a nucleus from a control (a) and one from an animal treated on day 11.5 (b) are shown. The rat brains were cut at 10 Pm and stained with cresyl violet and Luxol fast blue.
To test the adequacy of the VPA model as representing an injury like the one underlying autism, we turned to nonbehavioral features of the disease as sources of comparison. In rats exposed to VPA on day 12.5 and controls, we compared Purkinje cell number, granule cell number, neuron number in the deep nuclei of the cerebellar vermis, and linear measures of the size of the lobules. Because all cerebellar neurons form after day 12.5, the neurons counted in this study could not have been exposed to VPA. Thus, differences in cerebellar cell counts should occur only if the cerebellum is altered secondarily to the motor neuron lesion created by VPA exposure. In fact, cerebellar neuron numbers were reduced in the VPA-exposed rats. Purkinje cells were significantly fewer in lobules VI-VIII and IX, but not in the anterior lobes (IV-V) of the vermis. Among the deep nuclei, the nucleus interpositus (corresponding to the globose and emboliform nuclei of the human) was sig-
Etiology
421
of autism 0 P. M. RODIER ET AL.
nificantly reduced, but not the fastigial or dentate nuclei. In human cases, the dentate is spared and the fastigial, globose, and emboliform are injured (6). We found no significant reductions in cell density in vermian samples of the granule cells. A study of lobular volumes is not complete, but linear measures of the lobules show a reduction in the size of the most posterior ones, and normal size in the anterior ones. Taken together, these results suggest that: 1) most of the cerebellar effects observed in human cases of autism appear in the VPA-exposed rats as well. 2) In rats, as in humans, the effects are most prominent in the posterior lobe of the cerebellum. 3) The changes in cerebellar anatomy of human cases probably are secondary effects of loss of neurons in the cranial nerve motor nuclei. Most importantly, we believe that the presence of cerebellar effects in animals treated to injure the cranial nerve motor nuclei supports the contention that such animals are valuable as a model of autism. Not only do they model the effects on the brain stem nuclei observed to be involved in the thalidomide cases and our human case, but they resemble the histology of autopsy reports and the gross features of living patients as described by other investigators. SUMMARY Uncovering the etiologies of human developmental defects is always difficult. The case of autism has been particularly perplexing because the disease appears to involve both genetic and environmental factors and multiple etiologies. Under such conditions, studies of preand postnatal events in cases produce weak associations, at best. Screening the whole brain for abnormalities is just as problematic. Even when abnormalities appear, it may not be possible to determine whether they are related to the symptoms or incidental to the actual cause. A third approach-trying to guess the cause from the behavioral symptoms-is especially difficult when the human symptoms are described at a level so far removed from brain function. For example, disorders of communication, such as lack of facial gestures, can be interpreted to arise from almost any part of the CNS, from the brain stem to the frontal cortex. Without some information to guide the search, investigators cannot make use of all the wonderful techniques at their disposal, because they do not know what to test. The unexpected association of autism with thalidomide provides a new direction for research on the causes of the disease. It is the human equivalent of a laboratory experiment in which the time of exposure to the teratogen was varied. By telling us when an injury that initiates autism occurs, the thalidomide study tells us what was injured and thus, where to look. It not only allows us the
possibility of modeling the injury in animals but provides new ideas for evaluation of human cases. The thalidomide study suggests that the neuroanatomy of the autistic brain stem should be altered, and that hypothesis can be confirmed. The symptoms of the thalidomide cases and the anatomy of the autopsy case correspond to the deficits in Hoxa-I knockout mice, and, thus, suggest that mutations of developmental genes are likely to be the ones contributing to the genetic factor in autism. The thalidomide results suggest that we might want to examine much lower-level functions of the nervous system in autistic patients, and consider whether disturbances in communication and social behaviors might be mediated by more basic deficits in sensory and motor function. It is also important to investigate whether injuries to the closing neural tube generate developmental changes in later-forming parts of the brain. This appears to be the case in the cerebellum, and other regions might be affected as well. If investigators are to work back and forth between human and animal studies, one thing is clear: we cannot succeed if we abandon some levels analysis and concentrate on just a few. Autism makes a good example of a subject for which information has come from many levels of analysis. Only astute clinicians with knowledge of neurology and teratology could have carried out the thalidomide study. Psychiatrists and psychologists have defined the syndrome of autism by distinguishing its behavioral features from those of other mental illnesses. Classical embryologists have contributed to our understanding of the development of the hindbrain and geneticists and molecular biologists are building on that foundation. Teratologists have studied the effects of thalidomide and valproic acid for many reasons, and their results now turn out to be important in ways they could not have predicted. We need research on all levels to unravel the causes of developmental defects. Acknowledgments NS24287, ROl R824758.
-This work was supported by NIH Grants ROl AA08666, and P30 ES01247, and EPA Grant
REFERENCES 1. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 4th ed. Washington, DC; 1994. 2. Bryson SE, Clark BS, Smith IM. First report of a Canadian epidemiological study of autistic syndromes. J Child Psycho1 Psychiatry. 1988;29:43345. 3. Bryson SE. personal
communication.
4. Folstein SE, Rutter ML. Infantile autism: a genetic study of 21 twin pairs. J Child Psycho1 Psychiatry. 1977;18:297-321. 5. Piven J, Folstein SE. The genetics of autism. In Bauman ML, Kemper TL, eds. The neurobiology of autism. Baltimore: Johns Hopkins University Press: 1994: 1844. 6. Bauman ML, Kemper TL. Neuroanatomic
observations
in autism.
422
7.
8.
9. 10.
11.
12. 13. 14. 15. 16.
Reproductive
Toxicology
In Bauman ML, Kemper TL, eds. The neurobiology of autism. Baltimore: Johns Hopkins University Press; 1994: 11945. Courchesne E, Townsend J, Saitch 0. The brain in infantile autism: posterior fossa structures are abnormal. Neurology. 1994;44:21428. Hashimoto T, Tayama M, Murakawa K, Yoshimoto T, Miyazaki M, Harada M, Kuroda Y. Development of the brainstem and cerebellum in autistic patients. J Autism Dev Disord. 1995;25: l-18. Miller MT, Stromland K. Thalidomide embryopathy: an insight into autism? Teratology. 1993;47:387-8. Stromland K, Nordin V, Miller MT, Akerstrom B, Gillberg C. Autism in thalidomide embryopathy: a population study. Dev Med Child Neurol. 1994;36:351-6. Miller MT. Thalidomide embryopathy: a model for the study of congenital incomitant horizontal strabismus. Tram Am Ophthalmol Sot. 1991;89:623-74. d’Avignon M, Barr B. Ear abnormalities and cranial nerve palsies in thalidomide children. Arch Otolaryngol. 1964;80: 136-40. Lenz W. Thalidomide and congenital abnormalities. Lancet 1962; 1:271-2. Walker HA. Incidence of minor physical anomaly in autism. J Aut Child Schiz. 1977;7: 165-76. Scharre JE, Creedon MP. Assessment of visual function in autistic children. Optom Vis Sci. 1992;69:433-9. Gillberg C, Steffenberg S. Autistic behavior in Moebius syndrome. Acta Pediatr Stand. 1989;78:31&6.
Volume
11, Numbers
2/3, 1997
17. Rodier PM, Ingram, JL, Tisdale B, Nelson S, Roman0 J. An embryological origin for autism: developmental anomalies of the cranial nerve motor nuclei. J Comp Neurol. 1996;370:247-261. 18. Coleman PD, Roman0 J, Lapham LW, Simon W. Cell counts in the cerebral cortex of an autistic patient. J Autism Dev Disord. 1985; 15:245-55. 19. Carpenter EM, Goddard JM, Chisaka 0, Manley NR, Capecchi MR. Loss of Hex-AI (Hex-1.6) function results in the reorganization of the murine hindbrain. Development. 1993; 118: 1063-75. 20. Mark M, Lufkin T, Vonesch J-L, Ruberte E, Ohvo, J-L, DollC P, Gorry P, Lumsden A, Chambon P. Two rhombomeres are altered in Hoxa-I mutant mice. Development. 1993;119:319-38. 21. Schumacher HJ, Terapane J, Jordan RL, Wilson JG. The teratogenie activity of a thalidomide analogue, EM12, in rabbits, rats, and monkeys. Teratology. 1968;5:233%40. 22. Binkerd PE, Rowland JM, Nau H, Hendrickx AG. Evaluation of valproic acid (VPA) developmental toxicity and pharmacokinetics in Sprague-Dawley rats. Fundam Appl Toxicol. 1988;11:485-93. 23. Ardinger HH, Atkin JF, Blackston, RD, Elsas LJ, Clarren SK, Livingstone S, Flannery DB, Pellock JM, Harrod MJ, Lammer EJ, Majewski F, Schinzel A, Toriello HV, Hanson JW. Verification of the fetal valproate syndrome phenotype. Am J Med Genet. 1988; 29:171-85. 24. Christianson AL, Chesler N, Kromberg JGR. Fetal valproate syndrome: clinical and neurodevelopmental features in two sibling pairs. Dev Med Child Neural. 1994;36:357-69.