Diagnostic evaluation of patients with suspected haemophagocytic lymphohistiocytosis

Authors

  • Kai Lehmberg,

    1. Department of Paediatric Haematology and Oncology, University Medical Centre, Hamburg, Germany
    Search for more papers by this author
  • Stephan Ehl

    Corresponding author
    1. Centre of Chronic Immunodeficiency, University Hospital Freiburg, Freiburg, Germany
    • Department of Paediatric Haematology and Oncology, University Medical Centre, Hamburg, Germany
    Search for more papers by this author

Correspondence: Stephan Ehl, Centre of Chronic Immunodeficiency, Breisacher Str. 117, 79106 Freiburg, Germany.

E-mail: stephan.ehl@uniklinik-freiburg.de

Summary

Haemophagocytic lymphohistiocytosis (HLH) is a life-threatening hyperinflammatory syndrome characterized by severely disturbed immune homeostasis. It can affect all age groups. Diagnostic evaluation of the patient with suspected HLH has to address three main questions: (i) does the patient have HLH? There is no simple diagnostic test, but a number of clinical and laboratory criteria define this clinical syndrome. (ii) Can a trigger be identified? A variety of infections, malignant or autoimmune diseases can contribute to the disturbed immune homeostasis with important consequences for treatment. (iii) Does the patient suffer from a genetic disease predisposing to HLH? Recent advances in the understanding of the genetic and pathophysiological basis of HLH have enabled a better and more rapid answer to this question, which is relevant for prognosis and the decision to perform haematopoietic stem cell transplantation. This review summarizes the current diagnostic approach to the patient with HLH.

Haemophagocytic lymphohistiocytosis (HLH) is a life-threatening disorder of immune regulation (Janka, 2012). It is characterized by highly activated lymphocytes and macrophages that infiltrate tissues and produce large amounts of proinflammatory cytokines (Henter et al, 1991a; Osugi et al, 1997). A set of clinical, histopathological, and laboratory features define the acute syndrome, including unremitting fever, hepatosplenomegaly, haemophagocytosis (the engulfment of blood cells by activated macrophages) in different tissues and organs, cytopenia, hypofibrinogenaemia, elevated ferritin, triglycerides, soluble CD25 (sCD25, sIL2RA,), and liver enzymes (Bode et al, 2012). Several genetic diseases predispose to HLH (Table 1). They include familial haemophagocytic syndromes (FHL), immunodeficiencies with albinism, and various primary immunodeficiencies, in particular diseases associated with an impaired or aberrant immune response to infection with Epstein-Barr virus (EBV). The pathophysiological basis of the molecularly defined forms of FHL and of the albinism syndromes is an impairment of lymphocyte cytotoxicity (Pachlopnik Schmid et al, 2010). Mutations in the gene encoding the pore-forming protein perforin (PRF1) cause FHL-2 (Stepp et al, 1999), while genes involved in the transport and exocytosis of perforin-containing lytic granules including UNC13D (MUNC13-4) (Feldmann et al, 2003), STX11 (zur Stadt et al, 2005), and STXBP2 (MUNC18-2) (Cote et al, 2009; zur Stadt et al, 2009) cause FHL3-5. The genes causing albinism and immunodeficiency include RAB27A (Griscelli syndrome type II - GSII), LYST (Chediak-Higashi syndrome - CHS), and AP3B1 (Hermansky-Pudlak syndrome type II - HPSII) (Johnson et al, 2010). Their association with HLH, but also with platelet dysfunction, impaired pigmentation, and infection susceptibility reflects their role in lysosomal trafficking in a number of different organ systems (Stinchcombe et al, 2004).

Table 1. Classification of HLH. HLH may develop on the grounds of a genetic predisposition or as acquired form in association with a variety of disorders
 Genetic defects
Familial HLHPRF1 (perforin) – FHL2
UNC13D (MUNC13-4) – FHL3
STX11 (syntaxin 11) – FHL4
STXBP2 (MUNC18-2) – FHL5
Albinism syndromesRAB27A – Griscelli syndrome 2
LYST – Chediak-Higashi syndrome
AP3B1 – Hermansky-Pudlak syndrome type 2
OthersSH2D1A (SAP) - X-linked lymphoproliferative disease (XLP) 1
XIAP (BIRC4) – X-linked lymphoproliferative disease 2
ITK (IL2-inducible T-cell kinase)
CD27
Conditions associated with acquired HLHInfections
Malignancies
Autoinflammatory disease
Autoimmune disease
Immunosuppression
Inborn errors of metabolism

Defects in natural killer (NK) cell- and cytotoxic T cell (CTL)-mediated cytotoxicity may impair elimination and control of intracellular pathogens, leading to prolonged and enhanced stimulation of immune cells and subsequent immunopathology (Kagi et al, 1996). Notably, the cytotoxic effector response not only targets infected cells, but also antigen-presenting cells (APC). Elimination of APC in the context of an effective immune response is an important negative feedback to curtail the immune response. Impaired cytotoxicity disrupts this negative feedback (Fischer et al, 2007) and leads to continued stimulation of activated NK cells and CTL secreting large amounts of cytokines, in particular γ-interferon (IFNγ) (Binder et al, 1998; Jordan et al, 2004), which is a potent macrophage-activating stimulus.

This disturbance of immune homeostasis can also be caused by infections, autoimmune or neoplastic disease in the absence of a genetic defect of lymphocyte cytotoxicity. An excellent example is acquired EBV-induced HLH (Ishii et al, 2007). Here, EBV-infected B cells serve as APC and as a result of their rapid expansion in the context of the viral infection, the balance between activating APC and their elimination by virus-specific CTL is disturbed (Hislop et al, 2007). The result is a prolonged antigenic stimulation leading to a dramatic expansion of EBV-specific CTL, which initiate the pathophysiological cascade of events leading to the clinical picture of HLH. However, because the cytotoxic effector mechanisms are intact, the APC are eventually eliminated restoring immune homeostasis. This pathway of EBV-triggered HLH is favoured in several hereditary immunodeficiencies associated with an impaired or dysregulated response to EBV infection, particularly SAP (SHD2D1A) (Coffey et al, 1998) and XIAP deficiency (x-linked lymphoproliferative disease (XLP) 1 and 2) (Rigaud et al, 2006; Yang et al, 2012), but also CD27 (van Montfrans et al, 2012) and interleukin 2-inducible T-cell kinase (ITK) deficiency (Huck et al, 2009). The pathophysiological basis of the association of severe, autoinflammatory diseases such as systemic juvenile idiopathic arthritis (sJIA) and autoimmune conditions with HLH is less well understood (Grom & Mellins, 2010).

Diagnosis of HLH

Diagnosis of HLH is critical for appropriate therapy. A key prerequisite for diagnosis is a high level of suspicion and familiarity with the particular clinical features of the disease. Three main diagnostic questions have to be addressed in the patient presenting with suspected HLH: (i) does this patient have HLH?, (ii) is there an acute trigger for the disease?, (iii) does the patient have genetic disease predisposing to HLH? Recent advances in the understanding of the genetic and immunological basis of the disease have a significant impact on the current diagnostic approach to patients with HLH.

Does the patient have HLH?

The diagnosis of the acute episode of HLH is a challenge. Each single feature of this syndrome is unspecific and only the composite picture of a variety of clinical and laboratory characteristics results in the diagnosis of HLH. It represents the extreme upper end on the continous gradient of adequate inflammatory response to excessive hyperinflammation and, frequently, the diagnosis can only be made during the course of the disease, if the progression of parameters allows classification of the condition as HLH. Occasionally, episodes of incomplete HLH resolving spontaneously may precede the full picture and are only retrospectively recognized as such.

To establish a common definition of HLH, the FHL study group of the Histiocyte Society established a diagnostic guideline in 1991 (Henter et al, 1991b) which included the following criteria: (i) fever of >7 d duration, (ii) splenomegaly, (iii) cytopenia of at least 2 cell lines (haemoglobin <90 g/l and <100 g/l in neonates, platelet count <100 × 109/l, neutrophil count <1 × 109/l), (iv) hypofibrinogenaemia (<1·5 g/l) and/or fasting hypertriglyceridaemia (>2 mmol/l), and (v) the finding of haemophagocytosis in the bone marrow (Fig 1A), spleen, or lymph node. Each of these features was considered mandatory for diagnosis. The fulfilment of these criteria was used as inclusion requirement for the HLH-94 study of the Histiocyte Society.

Figure 1.

Morphological findings in HLH. (A) Macrophage phagocytosing erythrocytes and erythroblast (bone marrow). (B) Extracellular amastigotes in visceral leishmaniasis (bone marrow). (C), (D) Pigment clumps in hair shafts of patients with Griscelli syndrome type II (C) and Chediak-Higashi syndrome (D). (E) Giant granule in monocyte (bone marrow) and (F) small granules in neutrophil (peripheral blood) in Chédiak Higashi syndrome. Smears 600x, hair 100x. Modifed form Bode, S.F., Lehmberg, K., Maul-Pavicic, A., Vraetz, T., Janka, G., Stadt, U.Z. & Ehl, S. (2012) Recent advances in the diagnosis and treatment of hemophagocytic lymphohistiocytosis. Arthritis Res Ther, 14, 213.

Increasing knowledge on the features of HLH led to revision of the criteria for the following study HLH-2004 (Table 2) (Henter et al, 2007): (vi) Defective NK-cell activity, (vii) hyperferritinaemia (>500 μg/l), and (viii) sIL2RA of >2400 μ/ml were included as criteria and the cut-off of triglycerides was increased to 3 mmol/l. HLH was now diagnosed if at least 5 out of 8 criteria were fulfilled. In addition, the finding of a genetic defect consistent with HLH alone was considered sufficient for the diagnosis of HLH. However, it must be emphasized that the latter feature does not prove the acute clinical syndrome of HLH but a predisposition to develop the condition. These revised criteria are widely used to define HLH for scientific purposes and are useful in clinical practice. However, there are several additional characteristics that may underscore the clinical suspicion: moderate lymph node enlargement, jaundice, oedema and pleural or pericardial effusions, skin rash, hypoproteinaemia and hyponatriemia (Henter et al, 1991b), elevation of liver transaminases and lactate dehydrogenase (LDH).

Table 2. Diagnostic Criteria for HLH. According to the revised diagnostic criteria guideline of the HLH-2004 protocol, HLH is assumed if either a genetic diagnosis consistent with HLH or ≥5 out of 8 criteria are fulfilled (Henter et al, 2007). Of note, the finding of a molecular defect is not proof of acute hyperinflammatory HLH, but only a marker of predisposition to the syndrome
  1. Supportive evidence is lymph node enlargement, rash, oedema, central nervous system symptoms, hypoproteinaemia, hyponatriaemia, elevation of liver transaminases, lactate dehydrogenase, and cell count and protein of cerebrospinal fluid.

(A) Genetic defect consistent with HLH
(B) ≥5 out of 8 clinical and laboratory criteria fulfilledFever
Splenomegaly
Cytopenia ≥2 cell lines
haemoglobin <90 g/l, in neonates <100 g/l
platelet count <100 × 109/l
neutrophil count <1 × 109/l
Hypertriglyceridaemia (≥3 mmol/l) or hypofibrinogenaemia (<1·5 g/l)
Hyperferritinaemia (≥500 μg/l)
Soluble CD25 > 2400 U/ml
Haemophagocytosis in bone marrow, spleen, or lymph nodes
Low or absent NK-cell cytotoxicity

Given that each feature in itself is unspecific, the magnitude (unusually long-lasting fever, pronounced splenomegaly, excessive hyperferritinaemia, and profound hypofibrinogenaemia and cytopenia) may arouse suspicion of HLH and prompt further work-up. Of note, a patient may well have substantial ‘pre-HLH’ immune dysregulation without formally fulfilling the above-mentioned criteria.

Assessment of parameters

The clinical value of the parameters varies. In some cohorts, the prevalence of fever (resulting from hypercytokinaemia) and splenomegaly (caused by infiltration of lymphocytes and macrophages) is approximately 90–100% beyond the neonatal age (Janka, 1983, 2009). However, in neonates both symptoms are frequently absent, particularly if born preterm (Suzuki et al, 2009).

Generally, the sensitivity of laboratory parameter cut-off points is more readily available than specificity, as proper studies with well-defined appropriate disease control cohorts have been performed only for few parameters. At diagnosis, the bicytopenia criterion (attributed to hypercytokinaemia and haemophagocytosis) is fulfilled in most cases (Janka, 2009). Neutrophils are usually the last cell line to drop below the cut-off point (Janka, 1983), which can be considered a red flag feature of disease severity.

In HLH, the excess of ferritin reflects macrophage activation. The sensitivity of the 500 μg/l cut-off was 84% in a cohort of 34 patients in the HLH-94 study (Henter et al, 2007). Higher levels (>3000 or >10 000 μg/l) with presumed increased specificity have been proposed (Jordan et al, 2011). In a retrospective single centre study of 10 HLH patients and a large control group with different conditions, a sensitivity and specificity of >90% was found if a cut-off point of 10 000 μg/l was applied (Allen et al, 2008). The diagnostic accuracy of a low fraction of glycosylated ferritin (<25%) was superior to the absolute level of ferritin in a study in adults with HLH as compared to controls (Fardet et al, 2008).

Hypofibrinogenaemia <1·5 g/l is present at diagnosis in approximately 2/3 of patients (Janka, 1983), while other clotting factors are usually in the normal range (McClure et al, 1974). Of note, fibrinogen as an acute phase protein is usually increased in febrile illness. Thus, a level of fibrinogen in the lower range of normal in a child with fever for several days is suspicious of incipient immune dysregulation, potentially leading to the full picture of HLH. In our experience, profound hypofibrogenaemia in a febrile child is a quite specific feature, provided there is no severe liver failure or disseminated intravascular coagulation of other origin with consecutive general deficiency of coagulation factors.

High triglyceride levels have been attributed to a decreased level of lipoprotein lipase (Henter et al, 1991c). At diagnosis, this criterion (<3·0 mmol/l) is fulfilled in 70% of cases (Janka, 2012). For clinical purposes it is the least practical parameter as it is hardly possible to obtain true fasting levels of triglycerides in infants. Parenteral application of lipids during the course of disease may hamper its use as follow-up parameter.

sCD25 was initially described as a parameter that distinguished HLH from infectious mononucleosis and pseudomonas infection (Komp et al, 1989). The cut-off of 2400 U/ml was chosen on the grounds of a comparative analysis with septicaemia, juvenile myelomonocytic leukaemia, and Langerhans-Cell histiocytosis, where a sensitivity of 93% and specificity of 100% was documented M Schneider, Universitaetsklinikum Ulm, Ulm, Germany, personal communication). However, other relevant disease controls were not examined. For example, most patients with autoimmune lymphoproliferative syndrome due to FAS mutations have high levels of sCD25.

Liver transaminases (alanine and aspartate aminotransferase) are increased >100 U/l in approximately 1/3 of patients (Janka, 2009). Kidney function may be reduced in HLH-related multi-organ failure and should be determined in order to adjust drug doses (including HLH-directed therapy) accordingly.

A bone marrow puncture is mandatory, not only to look for haemophagocytosis, but also to exclude leukaemia as trigger of HLH. Haemophagocytosis in the bone marrow is neither very sensitive nor specific (Gupta et al, 2008). As the term haemophagocytic lymphohistocytosis seems to imply that haemophagocytosis is a mandatory feature, physicians less familiar with the condition may consider a negative bone marrow aspirate as incompatible with HLH. Quite the contrary, a negative initial specimen must not delay diagnosis and start of therapy, if the other parameters are conclusive. Liver, spleen, and lymph node biopsies merely for demonstration of haemophagocytosis should be avoided. However, occasionally they must be performed to exclude hepatic diseases or malignancies. If a liver biopsy is performed, a prominent finding apart from haemophagocytosis is lymphocytic infiltration of the portal tracts (Ost et al, 1998).

A recent large single centre study compared cytokine levels of HLH with other febrile conditions. A cutoff of >75 pg/ml IFNγ and >60 pg/ml Interleukin (IL)10 detected HLH with a sensitivity and specificity of >90%. Elevation of IL6 was moderate in HLH (median 50 pg/ml) and pronounced in bacterial septicaemia (median 325 pg/ml) (Xu et al, 2012). Expression of the haemoglobin–haptoglobin scavenger receptor CD163 is restricted to monocytes and macrophages. Soluble CD163 has been described as marker of macrophage activity in adults with episodes of acquired haemophagocytic syndromes (Schaer et al, 2005) and MAS in sJIA (Bleesing et al, 2007).

Several parameters are predictive of increased pre-transplant mortality, including hyperbilirubinaemia (>50 μmol/l), hyperferritinaemia (>2000 μg/l), and cerebrospinal fluid (CSF) pleocytosis (>100 × 106/l) at diagnosis, as well as thrombocytopenia (<40 × 109/l) and hyperferritinaemia (>2000 μg/l) after two weeks of treatment according to HLH94 or HLH2004 (Trottestam et al, 2011a). The latter finding is corroborated by a study that demonstrated increased mortality if the relative reduction of ferritin during treatment was low (Lin et al, 2011). The simple parameter, platelet count, rapidly reflects decrease or increase of disease activity in most cases. In our experience, excessive ferritin levels tend to drop substantially during the first days of treatment if the patient responds adequately to treatment. However, complete normalization of ferritin levels is a matter of several weeks or months and may be delayed by transfusion of erythrocytes. If in doubt about the cause of persisting or progressive cytopenia during treatment, a bone marrow aspirate may help distinguishing between continuing disease activity and myelotoxicity of therapy. In addition, a thorough microbiological work-up should be performed to identify infections exacerbating cytopenia.

Central nervous system involvement

Central nervous system (CNS) involvement constitutes the main cause of long-term morbidity in survivors of HLH (Trottestam et al, 2011b). Diagnosis of disease activity in the CNS is based on (i) clinical examination, (ii) assessment of CSF, and (iii) magnetic resonance imaging (MRI). Neurological signs and symptoms are apparent in 37% of patients, most frequently irritability, seizures, meningism, and reduced level of consciousness. Elevated CSF protein or cell count (>5 cells/μl) are currently defined as CSF involvement and are found in approximately one half of patients (Horne et al, 2008; Deiva et al, 2012). Typical findings on cytological investigation of the CSF are lymphocytic pleocytosis, activated monocytes, and haemophagocytosis, which can support the diagnosis. In a cohort of mostly acquired HLH, 39% of patients displayed abnormalities on computed tomography (CT) or MRI scan (Yang et al, 2010). In other studies the prevalence of neuroradiological pathology could not be determined because the investigations were performed predominantly in symptomatic patients. Virtually all areas of the brain can be affected. Cerebral atrophy, diffuse white matter abnormalities, and multiple focal lesions are the most common findings (Horne et al, 2008; Yang et al, 2010; Deiva et al, 2012; Rego et al, 2012). Cerebral atrophy after start of treatment, however, may be therapy-associated. Of note, the clinical, neuroradiological, and CSF findings are not necessarily coherent (Horne et al, 2008; Deiva et al, 2012). Thus, all three means of evaluation should be performed, provided there are no contraindications against a lumbar puncture (e.g. severe coagulopathy) or MRI (e.g. patient not fit for general anaesthesia).

The HLH-2004 protocol recommends intrathecal treatment only in patients with CNS involvement persisting after two weeks of systemic treatment. To this end, a follow-up CSF is recommended on day 14. Several authors have reported CNS-HLH without clear-cut systemic features of HLH (Gurgey et al, 2008). Thus, this differential diagnosis should also be considered in patients with unspecified CNS disease.

Is there an acute trigger?

Infectious triggers

Infectious triggers are frequently found both in hereditary and acquired HLH. Therefore, the identification of an infectious agent does not allow differentiation between these different forms of HLH. Viruses constitute by far the largest group of triggering infectious agents, in particular herpesviridae. However, many other viruses (and vaccinations), protozoa, fungi, bacteria, and mycobacteria have been reported as HLH triggers. Interestingly, virus-associated HLH was first reported in a cohort predominantly consisting of solid organ transplant recipients during immunosuppressive treatment (Risdall et al, 1979).

EBV is the most frequent and potent trigger of HLH (Ishii et al, 2007; Beutel et al, 2009). In particular, full-blown acquired HLH beyond infancy is commonly triggered by EBV. Patients with XLP 1 are extremely susceptible to overwhelming HLH at primary infection of EBV (Booth et al, 2011). Cytomegalovirus (CMV) and herpes simplex virus (HSV) are frequent triggers in neonates and infants (Imashuku et al, 2005; Suzuki et al, 2009). Viral stimulation of lymphocytes may be found in the peripheral blood smear and bone marrow. Polymerase chain reaction (PCR) is recommended for viral investigations as the diagnosis may initially be missed by serology. For screening, we recommend to test for EBV, CMV, HSV, parvovirus B19, human herpes virus 6, adenovirus, and varicella zoster virus. Further viruses should be included according to local prevalence and clinical presentation.

Interestingly, intracellular pathogens seem to have a propensity to set off HLH. Visceral leishmaniasis is one of the most frequent non-viral infectious triggers of HLH. The risk of missing the diagnosis of leishmania infection is high for several reasons: First, against general belief, leishmania infection may occur in temperate zones. Second, direct proof of amastigotes in the bone marrow smear (Fig 1B) is successful only in a minority of cases. Third, serological testing may be negative, particularly in the initial stages. (Gagnaire et al, 2000) Thus, PCR from a bone marrow specimen should be performed. Diagnostic work-up for mycobacteria and salmonella species should be performed in countries with high prevalence or if clinically suspected.

It is particularly important not to miss the diagnosis of infections for which treatment by anti-infectious therapy alone (leishmania by liposomal amphotericin B) or in conjunction with an HLH-directed therapy can be considered, e.g. anti-viral drugs for CMV (Wolschke et al, 2010), HSV (Suzuki et al, 2009), influenza A (Henter et al, 2010), and adenovirus, and rituximab for EBV (Beutel et al, 2009)).

Autoimmune and autoinflammatory disease

By convention, HLH in association with autoimmune or autoinflammatory disease is termed macrophage activation syndrome (MAS) (Stephan et al, 1993). It is commonly associated with sJIA (Stephan et al, 2001), but has also been described in systemic lupus erythematosus (Parodi et al, 2009), inflammatory bowel disease (James et al, 2006), and Kawasaki disease (Titze et al, 2009).

The diagnosis of MAS in sJIA is difficult to make, as patients with autoinflammatory disease including sJIA usually display high levels of white blood cells, neutrophils, platelets and fibrinogen, as a feature of disease activity. A drop of these parameters from elevated to normal values in an ongoing disease flare may indicate MAS (Sawhney et al, 2001), which will not be interpreted correctly when using the HLH-2004 criteria. To address this issue, a preliminary diagnostic guideline for MAS has been suggested, based on the analysis of a cohort of patients with MAS compared to a group of patients with an acute flare of sJIA (Ravelli et al, 2005) that includes the following criteria: platelet count <262 × 109/l, aspartate amino transferase >59 U/l, white blood count <4 × 10 × 109/l, fibrinogen <2·5 g/l, hepatomegaly, neurological symptoms, haemorrhage. As a second challenge, MAS frequently manifests during the first episode of the autoinflammatory disease before the diagnosis of sJIA is made. As unequivocal arthritis may be initially absent, this condition can be confounded with other forms of HLH. C-reactive protein >90 mg/l, neutrophil count >1·8 × 109/l and sCD25 < 7900 U/ml indicate MAS in sJIA rather than FHL or virus-associated HLH (Lehmberg et al, 2011). In our own experience, granulopoiesis tends to be predominant in the bone marrow of MAS patients whereas erythropoiesis is more prominent in FHL and acquired virus-associated HLH: The median ratio of myeloid (excluding eosinophils and basophils) and erythroid compartment (M/E) is 12 in MAS (n = 15), 0·66 in FHL (n = 50), and 1·1 in acquired virus-associated HLH (n = 30), constituting a statistically significant difference between each group.

Malignancies

The most frequent malignant triggers of HLH in children are T-cell lymphomas and leukaemias, in particular anaplastic large cell lymphoma (Lehmberg et al, 2012). However, cases with acute B-precursor lymphoblastic leukaemia (Kelly et al, 2011) and occasional reports for a variety of other malignancies have also appeared (Celkan et al, 2009). As simultaneous triggering infections may occur (Lehmberg et al, 2012), infectious work-up is mandatory. HLH concurring with, preceeding or following a malignancy has been reported (O'Brien et al, 2008). However, HLH during chemotherapy for any malignancy is frequently associated with infections (Lackner et al, 2008) and should in most cases be regarded as infection-associated HLH rather than malignancy-associated.

The proportion of patients with malignant disease is higher in adults (Takahashi et al, 2001) than in the paediatric cohort. In particular, in HLH patients beyond toddler age with normal degranulation and perforin, SAP, and XIAP expression, an infectious trigger or autoinflammatory and autoimmune conditions, malignant disease should be thoroughly excluded. This may sometimes require MRI, CT, or positron emission tomography. Given that XLP is associated both with HLH and lymphomas, SH2D1A mutations should be ruled out in males with lymphomas (Booth et al, 2011). In addition, malignant disease has been found in patients with biallelic STXBP2 (Pagel et al, 2012), PRF1 (Clementi et al, 2005) and STX11 (Rudd et al, 2006) mutations.

Inborn errors of metabolism

Several case reports have documented HLH in a variety of inborn errors of metabolism, (Ikeda et al, 1998; Kardas et al, 2011; Gokce et al, 2012). Whereas most are individual cases, patients with lysinuric protein intolerance appear to have a predisposition to display features of HLH and intermittently fulfil the HLH criteria (Duval et al, 1999). Patients with Gaucher disease may display hepatosplenomegaly, haemorrhages, cytopenia, hyperferritinaemia, slightly increased transaminases and LDH (Mekinian et al, 2012). Lipid-laden Gaucher cells in the bone marrow are easily differentiated from activated monocytes and macrophages in HLH. There is substantial overlap in the signs and symptoms of Wolman disease and HLH. Two reported patients were treated for suspected FHL before the diagnosis of Wolman disease (Perry et al, 2005).

Does the patient have a genetic disease predisposing to HLH?

The question, whether the patient has a genetic disorder predisposing to HLH, is of key importance for the further management and prognosis of the acute disease (Henter et al, 2007). Patients with HLH in the context of an underlying genetic disease frequently suffer from relapses during treatment or experience additional episodes of HLH with a high risk of a lethal outcome. This can only be prevented by correction of the defective immune system by haematopoetic stem cell transplantation. Rapid diagnosis of an underlying genetic disease can prompt the immediate search for an appropriate donor, the use of prophylactic drugs and a very careful follow-up until a donor is available. It also allows family investigations in order to identify additional individuals at risk for HLH.

A positive family history, consanguinity, recurrent disease, and coulocutaneous albinism may suggest hereditary disease. Oculocutaneous albinism is present in CHS, GSII and HPSII and causes abnormally light pigmentation of the skin, hair, and eyes leading to a fair skin and light-coloured hair, often with a metallic sheen. These changes are not always visible clinically and hair microscopy is recommended in all patients to identify the characteristic pigment distribution in GSII (Fig 1C) and – less pronounced – in CHS (Fig 1D) (Berrueco et al, 2010). In addition, blood smears and bone marrow should be evaluated to identify the characteristic intracellular granules of CHS (Fig 1E, F). In FHL5 patients, the presence of enteropathy and sensorineural hearing deficit, may point towards the genotype. (Pagel et al, 2012). Platelet secretion defects, sometimes associated with abnormal bleeding (Meeths et al, 2010), are observed in patients with HLH-associated defects in granule trafficking.

Age at manifestation of HLH is a poor indicator of genetic disease. Acquired HLH in infants is not a rare event, particularly in the context of connatal virus infections such as enteroviruses, HSV, and CMV (Imashuku et al, 2005; Suzuki et al, 2009). Moreover, there is an increasing number of reports of patients with FHL presenting in late childhood or even in adulthood (Rohr et al, 2010). The oldest reported patient with perforin deficiency was healthy until the age of 62 years, when he presented with HLH (Nagafuji et al, 2007). Later presentation is more common in immunodeficiencies with albinism (Jessen et al, 2011) and in patients with immunodeficiencies with impaired or aberrant response to EBV (Pachlopnik Schmid et al, 2011). It is therefore our policy to rule out hereditary HLH by functional testing in all patients with HLH, irrespective of their age at clinical presentation.

Although molecular sequencing of course represents the gold standard to diagnose genetic diseases, the genetic heterogeneity of HLH can make this difficult and time-consuming. Moreover, the detection of a mutation cannot, in all cases, be clearly associated with a functional defect in the encoded protein (Zhang et al, 2011). Therefore, analysis of protein expression and functional immunological tests are a very helpful addition to genetic testing and may enable important management decisions to be made rapidly and before the genetic results are available.

Cytotoxicity assays

Impaired NK cell cytotoxicity is a hallmark of HLH and represents one of the diagnostic criteria for the disease (Henter et al, 2007). NK cell cytotoxicity is usually tested by incubating peripheral blood mononuclear cells (PBMC), isolated by Ficoll gradient separation, with 51Cr labelled NK sensitive target cells, e.g. K562 cells. Lysis of target cells leads to release of chromium into the supernatant, which can be quantified. This assay is relatively simple to perform and widely available, but has a number of limitations. First, there is significant variability between different individuals, leading to a wide range of normal values. While absent cytotoxicity provides a clear result, partial impairments are more difficult to interpret. Second, impaired cytotoxicity of PBMC can be caused by a reduction in NK cell number among PBMC as well as by a reduction in NK cell lytic activity. To account for this, the percentage of NK cells among PBMC can be determined by flow cytometry, allowing the calculation of an NK cell to target ratio (rather than a PMBC to target ratio) (Fig 2A). However, in a significant number of patients with active HLH, the number of circulating NK cells is too low to appropriately judge their lytic potential in this assay. Finally, NK cell cytotoxicity can be impaired in HLH patients without underlying genetic disease and be normal in children with hereditary HLH (Filipovich, 2008).

Figure 2.

NK cell cytotoxicity and intracellular protein staining. (A) Lysis of NK-sensitive K562 target cells by PBMC from a healthy day control (circles) and a patient with GSII (squares). (B) Intracellular perforin expression in NK cells from a healthy control and a patient with FHL2. (C) SAP expression in T cells from a healthy control and a patient with XLP1. (D) XIAP expression in T cells from a healthy control and a patient with XLP2. The shaded areas represent the staining with an isotype control antibody.

With these limitations in mind, testing of NK cell cytotoxicity is still a valid diagnostic tool to identify patients with genetic defects in cytotoxicity or to confirm findings in other assays. An important advantage is its ability to characterize the whole process from target cell recognition, effector cell activation, transport and exocytosis of lytic granules to the actual lytic process leading to the leakage of the marker radionucleide from the target cell. NK cell cytotoxicity on K562 target cells is normal in the other immunodeficiencies associated with HLH. However, patients with SAP deficiency show impaired lysis of some other target cells including autologous EBV-infected B cells (Hislop et al, 2010). This is not due to a general defect in the molecular machinery of cytotoxicity, but due to a defect in the interaction of T/NK cells with B cells (Palendira et al, 2011), which also explains many other features of this highly variable immunodeficiency.

Cytotoxicity can also be assessed in CTL. To overcome the major histocompatibility complex restriction of target cell recognition, a so-called anti-CD3 redirected lysis assay is usually performed. In this assay, CTL are incubated with target cells binding anti-CD3 antibodies via their Fc part. Binding of CD3 on the CTL then induces formation of a synapse and initiates the cytotoxic process. CTL do not show lytic activity ex vivo, but must be activated by in vitro stimulation, usually performed for a period of 7–9 d. This in vitro activation may lead to mobilization of residual cytotoxic potential in patients with hypomorphic mutations decreasing the sensitivity of this assay in this group of patients (Rohr et al, 2010). Observations in the mouse model indicate that CTL rather than NK cell cytotoxicity is limiting for the protection against HLH (Jordan et al, 2004; Jessen et al, 2011). In line with this concept, it was observed in a group of CHS patients that impairment of CTL cytotoxicity rather than of NK cell cytotoxicity was predictive of the time of manifestation of HLH (Jessen et al, 2011).

Intracellular flow cytometry

Intracellular flow cytometry is an excellent tool to quantify intracellular protein expression. In contrast to protein detection by Western Blotting, it can be performed on a small number of cells and provides information on a per cell basis. The combination of extracellular staining followed by cell permeabilization and intracellular staining allows gating on subpopulations such as NK cells and T cells, which is useful in some situations. For diagnostic evaluation of patients with HLH, reliable intracellular stains are available that measure perforin, SAP and XIAP expression (Marsh et al, 2010). Perforin (Fig 2B) is reliably expressed in all CD16 + CD56 +  NK cells. Expression is reduced in almost all patients with mutations in PRF1, also in most patients with hypomorphic mutations allowing residual expression of the protein (Trizzino et al, 2008). Vice versa, not all patients with reduced perforin expression (e.g. patients with the PRF1 A91V mutation) show a relevant defect in cytotoxicity (Trambas et al, 2005). It is therefore our policy to sequence the PRF1 gene in all patients with reduced perforin expression, but also to confirm the functional significance of the mutation by cytotoxicity testing.

SAP and XIAP are both expressed in NK cells and T cells and can be studied in a single assay using antibodies against both proteins (Fig 2C, D). In both diseases, disease-relevant hypomorphic mutations can be associated with normal or only slightly reduced protein expression (Marsh et al, 2010). Normal expression of SAP and/or XIAP therefore does not exclude the disease and in case of a strong clinical suspicion, genetic analysis must be performed. If expression is reduced, analysis of maternal cells is a useful addition. Overall, SAP- or XIAP-deficient NK cells or CTL do not have a significant developmental/survival disadvantage, leading to the presence of random X-inactivation in these cell populations in affected carrier females (Marsh et al, 2010). They can therefore be identified by the presence of two populations with different levels of SAP/XIAP expression. Assessing the functional relevance of mutations in XLP1 or 2 is not so easy. Patients with disease-causing mutations in SAP do not have NKT cells (Pasquier et al, 2005). The assessment of 2B4-stimulated cytotoxicity or of cytotoxicity against autologous B cells are more advanced assays which are not suited for a routine diagnostic setting (Palendira et al, 2011). Enhanced apoptosis in restimulation-induced cell death assays can be demonstrated in most patients with clinically relevant XIAP mutations (Rezaei et al, 2011).

Degranulation assays

The process of lymphocyte cytotoxicity requires transport of lytic granules to the membrane at the site of contact between effector and target cell (the immunological synapse). At this site, the granules dock to the membrane, fuse and release their content (including perforin) into the synaptic cleft. During this process, proteins incorporated into the lysosomal membrane, such as CD107 (also called LAMP-1) become part of the cellular membrane and are accessible to surface staining with specific antibodies. Thus, NK cells (Fig 3A) or CTL (Fig 3B) triggered to release their lytic granules (degranulation) increase their membrane content of CD107, which can be quantified by flow cytometry. The assay provides information on both the fraction of cells upregulating CD107 in response to a given stimulus as well as on the intensity of CD107 expression in positive cells. Upon short-term stimulation with K562 target cells, 10–40% of resting NK cells upregulate CD107. The fraction of responding NK cells can be significantly increased by 24-48 h prestimulation with IL2. Only a very minor fraction of CTL can be induced to express CD107 directly ex vivo, but they become responsive after at least 48 h of in vitro activation. Anti-CD3/anti-CD28 provides the best stimulus to elicit CD107 expression in CTL.

Figure 3.

NK cell and T cell degranulation assays. (A) Degranulation of resting and interleukin 2 (IL2) activated (for 48 h) NK cells as assessed by CD107a expression after stimulation with K562 cells from a healthy day control and a patient with GSII. (B) Degranulation of T cell blasts (d7 PHA/IL-2 blasts) after stimulation with anti-CD3/anti-CD28 coated beads. The shaded areas represent incubation in medium alone.

Any defect in biogenesis, transport, docking, priming or fusion of the lytic granules leads to a decrease in CD107 expression in this assay. The initial description of the CD107 assays (Betts et al, 2003) enabled degranulation to be studied in patients with genetic diseases predisposing to HLH both in CTL (Enders et al, 2006) and in NK cells (Marcenaro et al, 2006). Subsequently, the ability of this assay to identify patients with all different variants of genetic degranulation has been confirmed in several small case series (Bryceson et al, 2007). On the basis of these data, we performed a multi-centre prospective diagnostic study in a large cohort of patients referred for evaluation for suspected HLH (Bryceson et al, 2012). The assays were well standardized and their performance was evaluated in round-robin tests.

We found degranulation of resting NK cells below 5% in 88% of patients with an FHL variant associated with impaired transport of lytic granules (n = 77) and in 77% of patients with an albinism variant (n = 13). In contrast, resting NK-cell degranulation was above 5% in 29 of 30 patients (97%) with FHL2 or with one of the 2 XLP variants and in 58 of 59 patients (98%) with secondary HLH. Therefore, the assay clearly discriminated between patients with genetic disorders predisposing to HLH that are associated with impaired degranulation and those that are not. Moreover, it also appeared useful in the discrimination of hereditary degranulation defects from acquired HLH. Overall, the sensitivity of NK cell degranulation was 96% and the specificity was 88%. Sensitivity was comparable in all age groups, whereas specificity was higher in patients less than 2 years of age (97% vs. 81%), reflecting the higher proportion of patients with genetic disease manifesting in infancy (Bryceson et al, 2012).

In patients with at least one hypomorphic missense or splice-site mutation, defects in resting NK-cell degranulation can be milder or even absent. This has been particularly reported in patients with STX11 (Bryceson et al, 2012) and STXBP2 deficiency with the exon 15 splice-site mutation (Pagel et al, 2012) known to be associated with a milder course of HLH, but has also been documented in patients with UNC13D deficiency (Rohr et al, 2010). In these cases, additional investigations have proven helpful. IL2 activation may induce reconstitution of NK-cell degranulation in patients with nonsense mutations in STX11, but less well in patients with nonsense mutations in UNC13D or STXBP2 (Bryceson et al, 2007, 2012). However, a priori, it is not possible to predict the genetic defect based on the pattern of the degranulation results. In patients with ambiguous results, the interpretation of the functional assays can be difficult and should be done by an experienced reference laboratory.

Genetic diagnosis

The relative frequency of mutations in the various genes predisposing to HLH differs among ethnicities (Zur Stadt et al, 2006; Pagel et al, 2012). The above-mentioned immunological diagnostic algorithm allows for a priorization of sequencing, but in many cases of FHL, several genes still have to be sequenced to obtain a molecular diagnosis. In some patients with impaired degranulation or cytotoxicity, no mutations in the known genes can be identified, but due to recent advances (zur Stadt et al, 2005, 2009; Cote et al, 2009; Meeths et al, 2011), a molecular diagnosis can now be obtained for most patients with FHL. In most European countries, it is estimated that no more than 10% of patients with familial disease remain without a molecular diagnosis. It is important to note that genetic sequencing of the exons and exon/intron boundaries may not be sufficient to detect all disease-causing mutations as recently demonstrated for UNC13Ddeficiency (Meeths et al, 2011). Genotype-phenotype correlations have been established for many of the genes predisposing to HLH and a number of missense mutations have been characterized that lead to a later onset and milder form of the disease (Ueda et al, 2007; Trizzino et al, 2008; Jessen et al, 2011; Sieni et al, 2011; Zhang et al, 2011; Pagel et al, 2012).

In summary: A diagnostic algorithm for HLH

Based on current experience, we recommend that all patients presenting with HLH, irrespective of age, should be investigated for expression of perforin and degranulation of resting and activated NK cells (Fig 4). It is our policy to also investigate all male patients for SAP and XIAP expression. CTL degranulation is optional, but helpful in patients with very few NK cells. Perforin expression is reduced or absent in the large majority of patients with FHL2, whereas evaluation of SAP and XIAP is not as reliable for identification of XLP patients, and functional and genetic analyses are recommended in patients with a strong clinical suspicion of XLP despite normal protein expression. In patients with normal expression of the above-mentioned proteins, resting NK-cell degranulation above 10% and normal activated NK-cell degranulation, no immediate genetic investigation is indicated. In patients with resting NK-cell degranulation between 5% and 10%, the assay should be repeated, more urgently in those in whom activated NK cell or CTL assays are abnormal. In patients with resting NK-cell degranulation below 5%, genetic investigations should be initiated unless there is good evidence for an acquired HLH and activated NK-cell and CTL assays are normal. The blood film and bone marrow aspirate (CHS granules, Fig 1B, C) and hair microscopy (abnormal pigmentation, Fig 1E, F) will determine whether genes associated with albinism or those associated with FHL should be investigated. The use of this diagnostic algorithm has helped us to provide a rapid answer to the question of whether the patient has an underlying genetic disease with important consequences for further management.

Figure 4.

Proposed diagnostic algorithm for patients presenting with HLH. Normal values have to be determined in the diagnostic laboratory. Normal intracellular expression of SAP and XIAP does not reliably exclude the molecular diagnosis of x-linked lymphoproliferative disease (XLP1 or XLP2). HLH, haemophagocytic lymphohistiocytosis; FHL, familial haemophagocytic syndrome; CHS, Chediak-Higashi syndrome; GSII, Griscelli syndrome type II; NK, natural killer cell; CTL, cytotoxic T cell; AICD, activation-induced cell death. Modified from research that was originally published in Blood. Bryceson, Y.T., Pende, D., Maul-Pavicic, A., Gilmour, K.C., Ufheil, H., Vraetz, T., Chiang, S.C., Marcenaro, S., Meazza, R., Bondzio, I., Walshe, D., Janka, G., Lehmberg, K., Beutel, K., zur Stadt, U., Binder, N., Arico, M., Moretta, L., Henter, J.I. & Ehl, S. A prospective evaluation of degranulation assays in the rapid diagnosis of familial hemophagocytic syndromes. Blood, (2012) 119, 2754-2763.

© the American Society of Haematology.

Acknowledgements

The authors would like to thank Gritta Janka, Udo zur Stadt, Andrea Maul-Pavicic, Heike Ufheil, Sandra Ammann, Sebastian Bode, Thomas Vraetz, Brigitte Strahm, and Ingo Müller for the excellent clinical, diagnostic, and experimental collaboration benefitting our HLH patients. The diagnostic algorithm was developed with an equal contribution of Yenan Brycesson, Daniela Pende, and Kimberly Gilmour. This work was supported by the Bundesministerium für Bildung und Forschung (BMBF 01 EO 0803) and the European Community's Seventh Framework Program (FP7/2007-2013 under grant agreement no. 201461).

Ancillary