Materials and methods

We studied 10 families in which the index case fulfilled the diagnostic criteria for HLH4 and a careful family history was collected. Consanguinity was investigated and when not evident the parents were asked to obtain further information including the birth place of their ancestors. Clinical data were obtained from the attending physicians and from thorough evaluation of records.

Natural killer activity was determined in one of the two reference laboratories (Dr Rita Maccario, Pavia, Italy and Professor Marion Schneider, Ulm, Germany) as previously reported.12 13 Molecular analyses were performed as reported by Stepp et al,10sequencing exons 2 and 3 of the PRF1 gene. The sequences obtained were compared to the reported gene structure (gene number 190339 NCBI) using the BLASTN program (http://www.ncbi.nlm.nih.gov/BLAST). In order to confirm the mutations found, the parents were also tested in all but three families, in which the mutations were confirmed by repeated experiments or by identification of the same mutation in the affected sib. In families in which consanguinity was considered possible on the basis of available information, polymorphic markers (D10S537, D10S676) were tested to confirm this.

STATISTICAL ANALYSIS

The median age at diagnosis and quartiles were calculated for each group of children. The cumulative probability of diagnosis free survival was computed by means of Kaplan Meier estimation. Incidence rates expressed as events per person month were calculated for each group. Time to diagnosis distributions were compared between mutated and non-mutated subjects by means of the log rank test.

Results

We have identified six novel mutations in thePRF1 gene; three additional mutations that we observed had been previously reported by Steppet al.10 Two novel mutations (C657A in case 3 and 1182 ins T in case 6) (table 1) introduced a stop codon in the sequence which resulted in a truncated protein. The other novel mutations (T283C in case 2, G658A in case 4, C662T in case 5, and C694T in case 6) caused an amino acid change. The mutations we observed are scattered along exons 2 and 3 without any obvious clustering. Four mutations were located in the second transmembrane domain while two occurred within or close to the EGF-like domain of the protein.14 It is remarkable that all the four patients of Turkish origin had the same mutation, G1122A. In 21 additional caucasian patients with HLH a mutation was not found. In particular, none of the patients of German origin from this group harboured the mutation9 (zur Stadt et al, manuscript in preparation).

Parental consanguinity was documented in five families and was very likely in three additional families, in which the ancestors originated from the same small villages or geographical regions. These patients (cases 1, 2, and 4, tables 1 and 2) were homozygous for loci D10S537 and D10S676, thus supporting parental consanguinity. Thus, eight of the 10 patients had related (or very probably related) parents. In 231 cases enrolled in the International HLH Registry3 (M Aricò, unpublished data) with information on this, 56 (23.2%) had related parents. Among the 10 patients withPRF1 mutations, seven had one or more sibs. Of a total of 21 sibs, six were affected. Four of 10 patients had one or more affected sib.

The clinical and laboratory features of the 10 patients withPRF1 mutations fit the diagnostic criteria for HLH (fever, splenomegaly, cytopenia, hypertriglyceridaemia or hypofibrinogenaemia, and haemophagocytosis).4Additional features, like CNS alterations, skin rash, lymphadenomegaly, and oedema, were also present in some cases (table 2). Eight of the 10 patients with PRF1 defects developed HLH by the third month of life (median 2.2 months). This age was lower than that of our additional 21 patients in whom noPRF1 mutations were found (median 5 months, range 0.5-86). In 209 cases enrolled in the Registry with information on this, the median age at diagnosis was 5.3 months (range 0-254 months) (p=0.05); 32% were diagnosed within three months and 80% within two years. One of our patients withPRF1 mutations remained asymptomatic until the age of 6 years when he developed full blown HLH. Among the Registry patients, 17 (8.1%) presented when older than 5 years, with an estimated risk of being diagnosed when older than 5 years of 8.1% (SE 1.8).

Although difficult to quantify, all patients had a very severe presentation and clinical course. They had to be aggressively treated and showed early relapse after disease control was initially achieved. All six patients who underwent BMT remain asymptomatic.

Discussion

We have described six novel mutations in thePRF1 gene in children with HLH; two introduced a premature stop codon in the sequence which resulted in a truncated protein, while the other four caused an amino acid change. Caution should be exercised in interpreting a missense mutation which causes an amino acid change to be responsible for the disease and is not just a population polymorphism. In our cases, these mutations modified a conserved amino acid and were never found in other subjects tested. These mutations were scattered along exons 2 and 3 without any obvious clustering, in keeping with the previous report by Steppet al.10

The same mutation, G1122A, was observed in the four patients of Turkish origin. This mutation had been observed twice by Steppet al 10 in patients of unspecified origin and was also reported in patients of Turkish origin.15 Altogether these data indicate that, at least in a subset of patients of Turkish origin with HLH, a founder effect is possible. Further analysis of affected children from the same geographical region should be undertaken to confirm this. No founder effect can be hypothesised in patients of Italian origin.

In this small series of patients with HLH andPRF1 mutations, each patient presented the symptoms which form the diagnostic criteria for HLH, as well as some of the less frequent abnormalities (table 2). Comparison with the additional 21 patients in whom PRF1mutations were not found confirms that no striking difference based on clinical grounds is evident between the two groups. The presence of an associated infection emphasises the triggering role of common pathogens and confirms that infection associated with HLH is common in patients with PRF1 mutations. All these mutations are very likely to cause a severe impairment of perforin function and in fact NK activity was severely impaired or absent in all of these patients.

Delayed onset of HLH, beyond five years, was reported in 8% of the Registry patients and was also documented in one of our patients withPRF1 mutations (case 6), who remained asymptomatic until the age of 6 years. Since patients of relatively older age, although fitting the diagnostic criteria, have often been thought to be potentially misdiagnosed, this information is relevant in that it confirms that, at least in a minority of cases, HLH should be suspected even beyond the usual age range.16 Whether HLH resulting from PRF1 mutation may present during adulthood remains an issue to be addressed.

All 10 patients with PRF1 mutations had a very severe presentation and clinical course. In some cases, HLH, either apparently sporadic or familial, may present with an incomplete picture and/or a mild course, including repeated episodes of remission, which may be controlled with minimal or intermittent treatment, and may even undergo spontaneous remission at least for a certain time, occasionally up to some years. This was not the case in our patients, all of whom had to be treated aggressively, showed early relapse after control of the disease was initially achieved, and were considered candidates for early BMT. All six patients who underwent BMT remain asymptomatic, confirming the unique potential of BMT for long lasting remission and even cure in HLH patients withPRF1 mutations.^17–19

Our findings underline the need to redefine the diagnostic approach to HLH in children. In particular, evaluation of NK activity, which was severely impaired in all but one (low-normal) case withPRF1 mutations, should be included in the clinical diagnostic work up of HLH.

In conclusion, our data confirm that PRF1mutations can occur throughout the coding region of exons 2 and 3 and suggest a founder effect for HLH in Turkey but not in Italy. HLH resulting from PRF1 mutation usually presents in infancy but occasionally may occur in older patients. Identification of a genetic defect in patients with HLH has diagnostic, prognostic, and therapeutic implications and should be pursued whenever possible. Despite frequent concordance of the age at onset within each family, asymptomatic sibs (including potential stem cells donors) cannot be safely defined as unaffected, unless their genetic status for HLH is assessed. Lack of this information may risk BMT from an affected donor in a presymptomatic phase.18 Identification ofPRF1 gene mutations allows diagnostic confirmation, correct genotype determination in the family, confirmed indication for BMT even from alternative donors, proper genetic counselling, and prenatal diagnosis. A detailed genotype-phenotype correlation cannot be performed until a much larger number of patients with and without PRF1 mutations are identified.