METHODS

The study protocol was approved in November 1998 by the Multi Centre Research Ethics Committee for Wales and by the Local Research Ethics Committees with responsibility for each of the geographical locations in which patients were assessed.

Subjects

Subjects were recruited from a cohort of 157 unrelated index cases with TSC, who had been ascertained sequentially for molecular genetic research into TSC at the Institute of Medical Genetics, Cardiff. Cases ascertained were from adult and paediatric medical clinics, from regional genetics services, and learning disability services, with help from the Tuberous Sclerosis Association. All fulfilled the most recent criteria for definite clinical diagnosis of TSC.³¹ All had undergone comprehensive analysis for mutations at the TSC1 and TSC2 loci, using a range of methods for the detection of both small intraexonic and splice site mutations, and larger deletions and rearrangements.²⁷

Clinical assessments

All subjects were assessed in their own homes or at a local genetics clinic by a single researcher (JCL), and their hospital medical records were reviewed. Assessments at interview included a structured clinical history detailing the onset, pattern of evolution, and treatment of seizures, and renal, skin, cardiac, and pulmonary symptoms; and a structured clinical examination. In addition, a range of standardised and validated psychological measures were used to assess intellectual level, autistic disorder, anxiety, depression, and behaviour. This involved:

• assessment of intellectual level using the Wechsler Adult Intelligence Scale (WAIS-R),³² Wechsler Intelligence Scale for Children (WISC-III),³³ Raven’s Coloured Progressive Matrices (RCPM),³⁴ or Vineland Adaptive Behaviour Scale³⁵

• assessment of autistic disorder using Childhood Autism Rating Scale (CARS)³⁶

• assessment of problematic behaviours in children using Conner’s Parent Rating Scale (CPRS)³⁷

• Assessment of psychological symptoms using Strengths and Difficulties Scale,³⁸ General Health Questionnaire,³⁹ and Hospital Anxiety and Depression Scale (HADS).⁴⁰

Statistical analysis was undertaken using the Statistical Package for the Social Sciences (SPSS) Version 10 for Windows. Chi squared (χ²) tests examined the relationship between genotype and the variables of IQ (< 70 v ⩾ 70), autistic disorder, history of infantile spasms, anxiety disorder, and depression. Logistic regression analysis was employed in a multivariate analysis, taking low IQ as the outcome variable and genotype and infantile spasms as independent variables.

RESULTS

Sample characteristics

Of the original 157 participants, 11 had died since their original ascertainment, and a further 11 were living outside mainland UK. The 135 remaining subjects (or their carers) were approached by letter with an invitation to participate in the present study; 98 agreed (response rate 98/135  =  73%), 9 declined, and 28 did not respond or could not be traced. Reanalysis of data on intellectual handicap reported by Jones et al (1999)²⁷ for subjects recruited to the present study compared with subjects who had either died, declined, or could not be traced, revealed no difference between the two groups (χ²  =  0.155; df  =  1; p  =  0.690). Mutations had been characterised in 92 of the 98 recruited cases: 22 had TSC1 mutations (6 familial and 16 sporadic cases), and 70 had TSC2 mutations (5 familial cases, 64 sporadic cases, and one adopted case). The ages of participants at assessment for this study ranged from 6 to 70 years (mean 27 years; SD 14). TSC1 cases were significantly older than TSC2 cases (means 33 years v 25 years, respectively; t  =  2.17; p  =  0.038), consistent with their ascertainment at a greater age.

Of the 98 participants, 47 were male and 51 female. No gender difference was found for normal v low IQ (χ²  =  0.133; df  =  1; p  =  0.715), history of infantile spasms (χ²  =  0.006; df  =  1; p  =  0.936), or autistic disorder (χ²  =  0.022; df  =  1; p  =  0.881). There was no gender difference between TSC1 and TSC2 probands (χ²  =  0.915; df  =  2; p  =  0.633).

Of the 98 cases, 93 had a history of epilepsy, including all of those (54/54) with an IQ of less than 70. Surprisingly, 20 of the 36 (56%) adults who were able to complete the HADS had anxiety related scores of 8 or above, indicative of anxiety disorder, and seven (19%) had depression related scores of 8 or above, indicative of depression. An apparent excess of these disorders in females did not reach significance (χ²  =  0.1626; df  =  1; p  =  0.202; and χ²  =  3.201; df  =  1; p  =  0.074, respectively).

Of the 98 cases, 86 were sporadic, 11 were familial, and one subject was adopted from biological parents of unknown genetic status. No significant differences in frequencies were found between familial index cases and sporadic cases for: low IQ (49/86 sporadic v 4/11 familial; χ² =  1.672; df  =  1; p  =  0.196), autistic disorder (30/86 sporadic v 1/11 familial; χ²  =  2.984; df  =  1; p  =  0.084), anxiety (17/30 sporadic v 3/6 familial; χ²  =  0.090; df  =  1; p  =  0.764), depression (6/30 sporadic v 1/6 familial; χ²  =  0.035; df  =  1; p  =  0.851), or infantile spasms (40/82 sporadic v 2/11 familial; χ²  =  3.667; df  =  1; p  =  0.056).

Locus and phenotypic heterogeneity

Phenotypic variables were considered as either present or absent. Intellectual level was categorised as either IQ of 70 or over, or IQ under 70. Intellectual level, history of infantile spasms, and presence of autistic disorder, anxiety, and depression were assessed in relation to TSC1 or TSC2 genotype. Low IQ, a history of autistic disorder, and infantile spasms were significantly more frequent in patients with TSC2 than TSC1 mutations (table 1). There was no significant difference between cases with TSC1 and TSC2 mutations in the frequency of current anxiety symptoms (p  =  1.000), nor in the frequency of current depressive symptoms (p  =  0.674).

Features of ADHD (assessed with the Conner’s Parent Rating Scale) and problem behaviours in children (assessed with the Strengths and Difficulties Scale) were not analysed in relation to TSC1 and TSC2 status, because only two participants under the age of 16 years had TSC1 mutations.

Both the presence of a TSC2 mutation, in comparison with the presence of a TSC1 mutation, and a history of infantile spasms significantly increased the likelihood of low IQ as an outcome (table 2). In view of the significant correlation between presence of a TSC2 mutation and a history of infantile spasms (table 1), logistic regression analysis was performed to examine further the relationship between genotype, history of infantile spasms, and IQ. The presence of a TSC2 mutation and infantile spasms were significant independent determinants of low IQ (table 2), with infantile spasms being most strongly associated with low IQ.

Both the presence of a TSC2 mutation, in comparison with the presence of a TSC1 mutation, and a history of infantile spasms as an outcome significantly increased the likelihood of low IQ (unadjusted rows in table 2). In view of the significant correlation between the presence of a TSC2 mutation and a history of infantile spasms (table 1), logistic regression analysis was performed to examine further the relationship between genotype, history of infantile spasms, and IQ, revealing that increased risk of low IQ could also be attributed independently to the presence of a TSC2 mutation and a history of infantile spasms (adjusted rows in table 2).

Mutation class and position, and phenotypic heterogeneity

The exact nature and location of mutations had been defined in 92 cases. These comprised a total of 81 different mutations, because six TSC2 mutations were shared by at least two individuals. Phenotypes varied between unrelated individuals sharing identical TSC2 mutations. A much larger series would be required to establish whether this variation was less marked than that seen between patients with different TSC2 mutations. Investigation of subgroups of subjects showing the most and least severe phenotypes, assessed where possible using quantitative measures (as we did here), might help in addressing this question.

All but one of the TSC1 mutations (an in frame deletion) was predicted to be truncating. Cases with TSC2 mutations included 53 with mutations that were predicted to be truncating and 17 with mutations that were predicted to be non-truncating. No significant difference was seen in the frequency of low IQ (χ²  =  0.237; df  =  1; p  =  0.627), autistic disorder (χ²  =  0.064; df  =  1; p  =  0.800), or infantile spasms (χ²  =  0.060; df  =  2; p  =  0.970) in cases with truncating as opposed to non-truncating TSC2 mutations in which less severe effects on protein function might be anticipated (table 3). Neither did analysis of the nucleotide position of truncating TSC2 mutations (considered as non-parametric continuous variables), using the Mann–Whitney U test, reveal a significant relationship with low IQ, autistic disorder, or infantile spasms (there were insufficient numbers of TSC1 cases for comparable analysis of mutations at that locus).

DISCUSSION

In this study we found that people with TSC2 mutations were significantly more likely than those with TSC1 mutations to have autistic disorder, a low IQ, and a history of infantile spasms. In addition, low IQ was found to be independently associated with both TSC2 mutations and a history of infantile spasms.

Participants in the current study were ascertained from a variety of clinical settings, and are likely to be broadly representative of unrelated index cases with TSC who come to medical attention. The frequency of medical, psychiatric, and behavioural disorders in this group is expected to be higher than in epidemiological samples of patients with TSC that include secondarily ascertained members of multiplex families, in whom milder disease might be anticipated. Also, our sample is biased towards sporadic rather than familial cases when compared with other large series.²⁸ This probably reflects our efforts to recruit sporadic cases in particular to the present study.

In this study, 55% of cases had an IQ of less than 70, compared with 44.5% in a recent epidemiological survey²; and 95% had a history of seizures, compared with 79% in the epidemiological sample. These differences are likely to reflect the medical ascertainment sources in the current study and the inclusion of probands only. In the present study, 33% of subjects were found to have autistic disorder, as assessed by the Childhood Autistic Disorder Rating Scale (CARS).³⁶ CARS is based on Kanner’s criteria⁴¹ for the diagnosis of autistic disorder, and not on the wider diagnostic category of pervasive developmental disorder. Previous estimates of the frequency of autistic disorder in TSC have varied, not only because of different modes of ascertainment, but also because of differences in the diagnostic criteria that have been employed, including DSM III-R⁴² and ICD-10.⁴³ Hunt and Shepherd⁴⁴ identified autistic disorder in 5/21 (24%) of TSC cases ascertained through hospital departments and the Tuberous Sclerosis Association (a UK patient support group); whereas Gillberg et al⁴² diagnosed autistic disorder in 17/28 (61%) of patients under 20 years of age, identified in a population based study.

Although anxiety and depression are common complications of chronic illnesses in general,^45,46 it was noteworthy that 20/36 (56%) of adults capable of completing the HADS assessments scored above the threshold indicative of anxiety disorder. Anxiety disorders appeared to be under recognised in TSC, because none of the patients involved had received advice or treatment for these disorders, perhaps owing to the frequent comorbid effects of autism and intellectual handicap.

Locus and phenotype heterogeneity

TSC2 mutations (70/92, 76%) were more frequent than TSC1 mutations (21/92, 24%) in our series, and the proportions were very similar to those reported by Dabora et al²⁸ in a large independent series of medically ascertained people with TSC. Our finding, that TSC2 mutations are more frequently associated with low IQ, infantile spasms, and autistic disorder than are TSC1 mutations, is consistent with two previous studies that suggested greater severity in disease determined by TSC2 than in disease determined by TSC1.^27,28 It is currently unclear whether the increased risk of low IQ and autistic disorder in persons with TSC2 mutations is mediated by an increased risk of infantile spasms alone, but our data suggest some independence of genotype and infantile spasms as risk factors. This issue is of great significance to the potential for therapeutic intervention in TSC. An experimental approach to resolving the question would be a randomised trial of pre-emptive antiepileptic drug medication in cases of TSC diagnosed antenatally. Detection of cardiac rhabdomyomas on routine antenatal ultrasound would now make such a trial technically feasible. Unfortunately, symptomatic CNS manifestations have not been noted in single locus knockout animal models of TSC, so precluding preclinical trials to address this issue. It is of interest, however, that murine Tsc2 mutations do appear to be associated with a more severe visceral phenotype than Tsc1 mutations on the same genetic background.⁴⁷

Since the TSC1 and TSC2 gene products, hamartin and tuberin, have been shown to form a functional complex^30,48,49 that acts to suppress cell growth via the PI3-kinase/AKT pathway,^50,51 it may not be immediately apparent why severity of disease determined by TCS1 should differ from that determined by TCS2. One possibility is that the somatic mutation rate is higher at the TSC2 locus, leading to more frequent “second hit” mutations and hence more frequent development of hamartomas. However, it remains unclear whether all the CNS manifestations of TSC result from a “two hit” mechanism. Careful molecular genetic studies of cortical tubers, including laser capture microdissection of cytomegalic neurones in one study, have failed to identify somatic “second hit” mutations in tubers, in contrast to a variety of visceral hamartomas.^23,24 Alternatively, or in addition, tuberin may have currently unidentified roles that are not shared by hamartin.

The increasing evidence for severity differences between disease determined by TSC1 and TSC2 presents issues for those involved in discussing prognosis for TSC that is diagnosed antenatally or in early childhood. However, the great range of phenotypic severity seen among unrelated persons sharing identical mutations, and even among different affected members of large families, will continue to demand a cautious approach. Further genetic studies are likely to identify loci that modify the phenotypic effects of TSC1 and TSC2 mutations, and clinical studies have the potential to clarify the extent to which long term outcomes can be modified by early (or possibly pre-emptive) treatment with antiepileptic drugs. Clarification of the roles of TSC1 and TSC2 in regulating cell growth through the P13-kinase/AKT pathway^42,43 has identified inhibitors of this pathway, such as rapamycin, as potential therapies when either gene is mutated.