Familial hemophagocytic lymphohistiocytosis (FHL) is a rare autosomal recessive lethal condition characterized by fever, cytopenia, hepatosplenomegaly and hemophagocytosis. The hallmark of FHL is defect apoptosis triggering and lymphocyte cellular cytotoxicity. Thus far three disease-causing genes (PRF1, UNC13D, STX11) have been identified. We performed a genotype-phenotype study in a large, multi-ethnic cohort of 76 FHL patients originating from 65 unrelated families. Biallelic mutations in PRF1, UNC13D and STX11 were demonstrated in 13/74 (18%), 6/61 (10%) and 14/70 (20%) patients, respectively. In 27/60 (45%) patients analyzed for all three genes, no molecular diagnosis was established. STX11 mutations were most common in Turkish families (7/28, 25%), whereas in Middle East families, PRF1 mutations were most frequent (6/13, 46%). No biallelic mutation was identified in most families of Nordic origin (13/14, 93%). Patients carrying PRF1 mutations had higher risk of early onset (age <6 months) compared to patients carrying STX11 mutations [adjusted odds ratio 8·23 (95% confidence interval [CI] = 1·20–56·40), P = 0·032]. Moreover, patients without identified mutations had increased risk of pathological cerebrospinal fluid (CSF) at diagnosis compared to patients with STX11 mutations [adjusted odds ratio 26·37 (CI = 1·90–366·82), P = 0·015]. These results indicate that the disease-causing mutations in FHL have different phenotypes with regard to ethnic origin, age at onset, and pathological CSF at diagnosis.
To date, three genes have been identified that cause familial hemophagocytic lymphohistiocytosis (FHL) (MIM 267700). More is yet to be discovered about how they differ from one another with regard to their phenotype. The term haemophagocytic lymphohistiocytosis (HLH) comprises the inherited form, FHL, and secondary HLH, which usually is virus- or malignancy-associated. FHL is an autosomal recessive immune deficiency characterized by intermittent or constant fever, cytopenia, moderate to massive hepatosplenomegaly, and hemophagocytosis demonstrated in the bone marrow, and/or other organs such as liver, spleen, or lymph nodes (Janka, 1983, 2007; Henter et al, 1998; Filipovich, 2006). Most patients also display a severe inflammatory response with hypertriglyceridaemia, hypofibrinogenaemia and elevated ferritin levels. In addition, many of the patients develop progressive cerebromeningeal symptoms during the course of the disease (Henter & Elinder, 1992; Haddad et al, 1997; Henter & Nennesmo, 1997). Typical for the pathogenesis of HLH is a defect in apoptosis triggering and in natural killer (NK) and cytotoxic T lymphocyte (CTL) cellular cytotoxicity (Arico et al, 1988; Fadeel et al, 1999; Schneider et al, 2002) associated with a prominent overproduction of several inflammatory cytokines (Henter et al, 1991a; Osugi et al, 1997). Without treatment including hematopoietic stem cell transplantation (SCT), FHL is inevitably fatal (Henter et al, 2002; Ouachee-Chardin et al, 2006).
Familial hemophagocytic lymphohistiocytosis is a genetically heterogeneous disease and four loci have been mapped by homozygosity linkage to 9q21.3-22 (FHL1), 10q21-22 (FHL2), 17q25.1 (FHL3) and, most recently, to 6q24 (FHL4). Stepp et al (1999) identified mutations in the perforin gene (PRF1) in FHL patients linked to chromosome 10, which was later found to be responsible for approximately 20–40% of the FHL cases (Stepp et al, 1999; Goransdotter Ericson et al, 2001). The perforin protein (PRF1) is a constituent of lytic granules in CTLs and considered important for release of the serine-protease granzyme B (GZMB) from the endolysosomal compartment into the cytosol of the target cell. Granzyme B is then apparently rapidly trafficked to the nucleus (Trapani et al, 1996; Smyth et al, 2001), where it induces target cell DNA fragmentation and apoptosis. Feldmann et al (2003) identified mutations in the UNC13D at 17q25.1; this gene is important for the membrane fusion events in vesicle exocytosis of the lytic granules of CTLs (Feldmann et al, 2003). Mutations in a third gene, syntaxin 11 (STX11) at 6q24, were revealed in a subgroup of FHL patients by zur Stadt et al (2005), however, the contribution of mutations in STX11 to FHL was considered ambiguous. We recently showed that freshly isolated, resting natural killer (NK) cells and CD8+ T cells express syntaxin 11 (STX11). In addition, NK cells from patients with mutations in STX11 failed to degranulate when encountering susceptible target cells, as also reported in patients with UNC13D defects (Marcenaro et al, 2006; Bryceson et al, 2007). However, the failure to degranulate was partially restored upon interleukin 2 (IL-2)–stimulation (Bryceson et al, 2007).
Despite the genetically heterogeneity of FHL, not many distinct features of the phenotype in correlation to the genotypes have yet been identified, which complicates the molecular diagnosis. A previous genotype-phenotype study (Ueda et al, 2006) compared patients carrying mutations in PRF1, UNC13D and a group where neither PRF1 nor UNC13D mutations had been identified. Their results indicated a significant difference regarding ferritin levels and hepatomegaly, both more common in patients with PRF1 mutations (Ueda et al, 2006). In addition, several studies have shown differences in age at onset when comparing FHL patients harboring non-sense and missense mutations in PRF1 (Clementi et al, 2002; Feldmann et al, 2002; zur Stadt et al, 2006; Trizzino et al, 2008) and recently, patients with at least one missense mutation were shown to have residual cytotoxicity (Trizzino et al, 2008).
In an effort to better characterize the clinical phenotype in patients with FHL and to analyze whether a correlation could be found to patients with the three genotypes PRF1, UNC13D, STX11 and the no known gene defect, we retrospectively analyzed a cohort of 76 patients originating from 65 unrelated families from the Nordic countries, Turkey, and the Middle East (which included one patient originating from Pakistan).
Patients and methods
Patient selection and collection of clinical data
Between January 2000 and December 2006, DNA from 78 patients originating from the Nordic countries, Turkey and the Middle East was collected. Patients were initially included in the study if the treating physician had regarded and treated their disease as FHL. One patient was later diagnosed with Griscelli syndrome and was therefore excluded from the genotype-phenotype analysis. In a second patient sequencing of PRF1 revealed an homozygous A91V alteration in PRF1. As the pathogenic contribution of this mutation is unclear (zur Stadt et al, 2004) and most probably only contributes to the development of FHL in the presence of other pathogenic mutations, this patient could not be classified to a genetic subgroup. For that reason, the patient was excluded from the genotype-phenotype analysis and 76 patients remained to study. Seventy-two patients were analyzed for PRF1 mutations, 34 for UNC13D mutations and 59 for STX11 mutations. Unfortunately, all patients could not be analyzed for all three mutations due to lack of DNA. All families were sequenced at the Department of Molecular Medicine and Surgery at Karolinska Institutet, except three families from Oman, which were analyzed at Sultan Qaboos University in Oman. For this study, detailed clinical data of these 76 patients were collected, either by retrospectively reviewing patients’ notes and/or by a questionnaire sent out to the physicians treating the respective patients. Information was collected on clinical and laboratory findings at onset of disease, treatment, response to treatment, and long-term outcome. The study was approved by the Ethics Committee at Karolinska Institutet.
HLH disease status
Haemophagocytic lymphohistiocytosis was defined by the diagnostic criteria established by the Histiocyte Society (Henter et al, 1991b) and/or a positive family history. Cerebrospinal fluid (CSF) was considered abnormal in case of pleocytosis or hyperproteinemia. Central nervous system (CNS) involvement was defined as abnormal neurological clinical examination and/or abnormal CSF. Non-active HLH disease was defined as absence of clinical signs of disease, i.e. no fever except if infection-induced, no hepatosplenomegaly, no clinical signs of active CNS disease and no cytopenia (unless drug-induced), in accordance with the HLH-94 treatment protocol (Henter et al, 2002).
Genomic DNA was obtained from all patients and/or parents from peripheral blood leucocytes or fibroblasts. The complete open reading frame for PRF1, UNC13D and STX11 were sequenced by direct sequencing using ABI 310 and ABI 3730 Genetic Analyzers (Applied Biosystems, Foster City, CA). Primers, polymerase chain reaction (PCR) and sequencing conditions are available upon request.
Differences in distribution were compared by the Chi-square test and, when frequencies were small, the two-tailed Fisher’s exact test was used. Tests for associations between genotype and phenotype were performed by exact Pearson chi-square tests for r×c tables using PROC FREQ in the SAS software. Subsequently, multivariate analysis using logistic regression was performed with age less than six months at diagnosis and pathological CSF as dependent variables. The covariates used were genetic mutation group and ethnicity. Logistic regression analyses were carried out using SPSS™ statistical software (version 11.5) (SPSS Inc., Chicago, IL, USA).
Patient baseline characteristics
Seventy-six patients from 65 unrelated families, originating from the Nordic countries, Turkey and the Middle East were included in the analyses. The baseline characteristics of these patients are detailed in Table I, which also defines the patients according to their genotype. The diagnostic criteria (Henter et al, 1991b) were completely fulfilled in 55/68 patients (81%) (nd = 8). Of the 13 children evaluated who did not fulfil all criteria, eight had a family history of the disease, four carried a mutation and seven were consanguineous. Similarly, of the eight with missing data on diagnostic criteria, three had a positive family history, three carried a mutation and four were consanguineous. Notably, all patients fulfilled the HLH-2004 diagnostic criteria, had a positive family history or a verified biallelic mutation in either of the three genes studied (Henter et al, 2007). The median age at diagnosis was 198 d (range 18 d – 12 years). CNS involvement was found in 42/69 patients (61%) at the time of diagnosis (nd = 7). Information on NK cell activity was only available for 18 of the patients and therefore not analyzed. The majority of the patients (50/76; 66%) were treated according to the HLH-94 protocol (Henter et al, 2002). Of the 26 patients not treated according to the HLH-94 protocol, six received treatment prior to 1994 but with a similar treatment (Henter & Elinder, 1991), nine were treated according to the slightly modified protocol HLH-2004 (Henter et al, 2007), three received other treatment combinations, five died before therapy was started and in three cases the parents refused therapy. Out of these three, one died of disease after 59 d and the other two were lost to follow up. At the last follow-up of all 76 patients, 31 were alive, and the mean follow-up time from diagnosis was 3·9 years (range 53 d – 21 years). Forty-two patients were deceased and three patients were lost to follow-up. Hematopoietic stem cell transplantation had been performed in 28 patients; 20/28 (71%) of these patients were alive.
Table I. Characteristics of the 76 HLH patients at diagnosis and their association with genotype groups.
No known mutation
None of PRF1, UNC13D, STX11
Not all genes studied
Values in parentheses are expressed as percentages.
CNS, central nervous system; CSF, cerebrospinal fluid.
Number of patients
Age at diagnosis
Sequencing of the different genes so far known to be involved in FHL enabled molecular diagnosis in 33 of the 76 patients (43%), corresponding to 24 of the 65 unrelated families (37%) (Fig 1). Involvement of PRF1 was demonstrated in 13/74 (18%) patients, corresponding to 12/63 (19%) families. Mutations in UNC13D and STX11 were identified in 6/61 (10%) and 14/70 (20%) patients, respectively, corresponding to 4/50 (8%) and 8/69 families (12%) (Fig 1). We were not able to establish the molecular diagnosis in 27 of 60 (45%) patients analyzed for all three known genes, corresponding to 25/49 (51%) families. For 16 patients we were unable to analyze all three genes (Table I). Screening of PRF1 identified nine different mutations; two non-sense, five missense and two in-frame deletions. Four different mutations were identified in UNC13D; one non-sense, two missense and one splice mutation, and three different mutations were identified in STX11; one non-sense mutation, a single nucleotide deletion and one out-of-frame mutation were found. The mutations identified are detailed in Table II.
Table II. Mutation spectrum in PRF1, UNC13D and STX11 in the patients studied.
The clinical data collected from each patient were reviewed in detail and correlated to the genotype (Table I). The patients were classified into four distinct subgroups based on whether they carried biallelic mutations in PRF1 (n = 13), UNC13D (n = 6), STX11 (n = 14), or no mutation in any of these three genes (n = 27). The remaining 16 patients were not included in any subgroup because it was not possible to analyze all three genes in these individuals. To evaluate the presence of phenotypic distinctions, each genotype group was compared to the other genotype groups. In addition, to investigate whether ethnicity correlated to any of the genotypes, patients were classified into subgroups based on their ethnic origin: the Nordic countries, Turkey and the Middle East. To provide a more valid estimate of the prevalence of the different genotypes, the correlation of ethnicity to genotype was performed based on the 65 families rather than the 76 patients (Table III). Regarding ethnicity, each genotype group was compared to all other patients analyzed for their respective genotype.
Table III. Correlation between genotype and ethnicity in the 65 unrelated families studied.
PRF1 (n = 63)
UNC13D (n = 50)
STX11 (n = 59)
No mutation (n = 49)
Values in parentheses are expressed as percentages.
Nordics (n = 18)
Turkey (n = 33)
Middle East (n = 14)
Ethnicity in correlation to genotype
History of familial disease and consanguinity was found to have a strong association with ethnicity, being rare in the Nordic patients. We observed a significantly higher prevalence of PRF1 mutations in the Middle East families compared to the families originating from the Nordic countries [6/13 vs. 1/18 (P = 0·012)] (Table III). In addition, a higher incidence of STX11 mutations was noted among Turkish families compared to the Nordic group [7/28 vs. 0/17 (P = 0·034)]. In the Nordic group the frequency of families with no identified mutation was significantly higher compared with both families from Turkey [13/14 vs. 10/24 (P < 0·002)] and families from the Middle East [13/14 vs. 2/11 (P < 0·001)].
Clinical symptoms and treatment response in relation to genotype
Each genotype group was compared to the other individual groups to evaluate the presence of phenotypic distinctions. We observed significant differences regarding the history of familial disease (P = 0·027), consanguinity (P < 0·001) and ethnical origin (P < 0·001). We also observed significant differences in the clinical presentation regarding age < three months at onset (P = 0·040), age <6 months at onset (P = 0·055), jaundice (P = 0·030), and pathological CSF (P = 0·031). No differences were detected with regard to the presence of symptoms included in the diagnostic criteria, hepatomegaly, oedema, skin-rash, or ferritin at the time of diagnosis. In addition, we were not able to identify any differences in response to initial therapy (measured as if the patients were dead or alive with active or inactive disease at 2 months after start of therapy) between the four genotype groups.
Age at onset in relation to genotype
In our cohort of patients the median age at diagnosis was 2·3 months, 6·2 months and 14·4 months for patients with PRF1, UNC13D and STX11 mutations, respectively, and 4·9 months for the patients without any identified gene defect. The patients from the Middle East were younger at diagnosis (age at diagnosis <6 months) compared to those from Turkey [10/15 vs. 14/41 (P = 0·029)]. Logistic regression analysis was used to evaluate the influence of ethnicity or genotype on age at diagnosis and determine which of those two factors it was most dependent of. The dependent variable studied was age at diagnosis less than six months. Patients carrying PRF1 mutations were shown to have a significantly increased risk of an early onset of the disease compared with patients carrying mutations in STX11. The association remained after adjusting for ethnicity as a potential co-founding factor [odds ratio 8·23 (95% confidence interval [CI] = 1·20–56·40), P = 0·032] (Table IV). For patients with PRF1 mutations, the mean age at diagnosis in patients carrying non-sense mutations (n = 4) was 5 months (range 2·1–10·7 months) compared to 21·4 months (range 0·9–61·5 months) in the patients with missense mutations (n = 3).
Table IV. Frequency of early age (<6 months) at onset of disease in relation to ethnic origin and mutated gene (multivariate analyses, n = 59).
OR, odds ratio; 95% CI, 95% confidence interval.
CNS disease at diagnosis in relation to genotype
Of the 33 patients with known biallelic mutations in any of the genes, neurological symptoms at onset were reported in nine. Among patients with PRF1 mutations, five of 13 had neurological symptoms; two with symptoms not specified, one with encephalopathy, one with seizures and one with weakness of the left leg and balance disturbances. Among those with UNC13D mutations, two of five (nd = 1) patients were reported with neurological symptoms at onset; one with microencephaly and mental retardation, and one with seizures. In patients with STX11 mutations, only two of 14 patients were reported to have neurological symptoms at onset; one with signs of development delay and one with seizures. In the group with no biallelic mutation found, 11 of 26 (nd = 1) had neurological symptoms. The most common symptoms in this group were cranial nerve palsies, seizures and irritability.
Regarding the clinical presentation at diagnosis, there was a trend to more pathological CSF in patients from the Middle East compared to patients from Turkey [9/12 vs. 13/28 (P < 0·096)].
To investigate a possible association between CNS disease and genotype logistic regression analysis was performed for the four genetic subgroups. Pathological CSF at diagnosis was chosen as the dependent variable (Table V). The unadjusted odds ratio showed an increased risk of pathological CSF at diagnosis for both patients with PRF11 mutations and those with no identified mutation if compared to the patients with STX11 mutations. After adjusting for ethnicity as a potential co-founding factor this association remained for patients without identified mutations compared to the patients with STX11 mutations [odds ratio 26·37 (95% CI = 1·90–366·82), P = 0·015].
Table V. Frequency of pathological CSF at onset of disease in relation to ethnic origin and mutated gene (multivariate analyses, n = 46).
OR, odds ratio; 95% CI, 95% confidence interval.
Familial hemophagocytic lymphohistiocytosis is characterized by a clinical homogeneity in spite of the genetical heterogeneity. We performed a genotype-phenotype study comparing four genetically different groups of FHL patients: three groups with patients harboring biallelic mutations in the three genes PRF1, UNC13D and STX11, and a fourth group comprising patients where biallelic mutations in none of these genes had been detected. The study identified significant differences for ethnicity and clinical presentation with regard to age at diagnosis and pathological CSF between these four subgroups.
This study established a relationship between ethnicity and genotype. PRF1 mutations were found to be more frequent in patients originating from the Middle East compared to patients originating from the Nordic countries and Turkey. In addition, STX11 mutations were shown to be more common in patients from Turkey in contrast to the Nordic countries and the Middle East. For the majority of patients originating from the Nordic countries no mutation was identified in any of the known genes. This is consistent with previous studies that have shown that PRF1 mutations are an uncommon cause for FHL in patients originating from Nordic countries, as well as in Germanic patients (Goransdotter Ericson et al, 2001; zur Stadt et al, 2006). Biallelic STX11 mutations have so far only been reported in patients originating from Turkey and Lebanon (zur Stadt et al, 2005; Rudd et al, 2006; Bryceson et al, 2007).
Further, a significant difference in age at onset was identified in patients carrying PRF1 mutations compared to STX11 mutations, the patients in the latter group being older (>6 months) at the time of diagnosis. It is well known that the corresponding proteins of PRF1, UNC13D and STX11 are of importance for NK and CTL cytotoxicity, however, they act on different levels. The first step in the Ca2+-dependent degranulation and exocytosis processes of cytotoxic granules is membrane fusion (Hong, 2005). UNC13D and STX11 both encode proteins that are important for the vesicle-plasma membrane fusion during exocytosis of the cytotoxic granules from CTLs and NK cells (Tang et al, 1998; Feldmann et al, 2003; Bryceson et al, 2007). The exact mechanism leading to synaptic vesicle exocytosis is not fully understood, but these proteins are thought to interact with the SNARE (soluble N-ethylmaleimide-sensitive factor attachment protein receptor) complex. In the SNARE model, t-SNAREs and v-SNAREs (present on target (t) and vesicle (v) membranes, respectively) assisted by several different proteins, are thought to recognize each other specifically and to form a core complex that brings the two membranes together, ultimately leading to fusion and release of PRF protein and other components of the cytotoxic granules (Lieberman, 2003). Thus the defect in cytotoxic activity is subordinated to the defect in degranulation in patients carrying mutations in UNC13D and STX11, which was recently reported to be subdued by in vitro stimulation with IL-2 (Bryceson et al, 2007). Subsequently, one might speculate if the same is true in vivo; that IL-2 acts as an innate rescue mechanism in the early phase of the disease, which then would explain the delayed onset in at least STX11 patients. However, we were not able to identify a significant difference in age at onset between PRF1 and UNC13D, but this might be due to the limited number of patients carrying UNC13D mutations in this study.
Since neurological symptoms were not categorized by a standardised evaluating scheme, pathological CSF was chosen as measurement of neurological disease in this study. There was a significantly decreased risk for pathological CSF in patients carrying STX11 mutations compared to the group where none of the known genes had been identified. This finding combined with the relative risk to being less than six months old at diagnosis, which was also decreased in patients harbouring STX11 mutations, indicates that patients with mutations in STX11 are more prone to develop a slightly milder course of the disease. We have previously observed a similar tendency (Rudd et al, 2006; Bryceson et al, 2007). Actually, in the present study none of the patients with STX11 mutations were less than six months and also had pathological CSF at diagnosis. The corresponding results for patients with PRF1 mutations and no identified mutation was 5/11 patients and 6/20 patients, respectively (data refer to patients with information available on CSF at onset, n = 45). A remarkable finding was that CNS disease appeared to be more prominent in patients with no known gene mutation, which might indicate that the genetic defect or defects in these children comprises gene or genes important in the CNS, as well as in the immune system.
To summarize, this study presents a large cohort of FHL patients that was thoroughly investigated and evaluated regarding genetic and clinical data, respectively. It is the first genotype-phenotype study to include patients representing all known genes in FHL and compare these to each other as well as to patients not harboring mutations in either PRF1, UNC13D or STX11. In conclusion, the study provides important and significant correlations of the genotypes to age at onset and neurological disease. Our findings suggest that patients with STX11 mutations, with a low frequency of early onset and pathological CSF, may have a less severe disease. Our study also points out previously unrecognized insights into the phenotype of the patients where the genetic defect is still undetected, and this information may hopefully add useful knowledge in the future search for additional genes.
We are very grateful to all contributing colleagues as well as the patients and their families for their participation. Supported by the Children’s Cancer Foundation of Sweden (J-I.H.); the Swedish Cancer Foundation (M.N., J-I.H.), the Cancer and Allergy Foundation of Sweden (J-I.H.); the Swedish Research Council (J-I.H.); the Tobias Foundation (J-I.H.); the Märta and Gunnar V Philipson Foundation (E.R.); the Histiocytosis Association of America (J-I.H., M.N.) and the Stockholm County Council (ALF project) (J-I.H.).