Genotype characteristics and immunological indicator evaluation of 311 hemophagocytic lymphohistiocytosis cases in China

Background Primary hemophagocytic lymphohistiocytosis (pHLH) is a genetic disorder that is classically diagnosed by genetic testing. Secondary HLH (sHLH) is usually caused by infections, malignancies, or autoimmune disorders, but may display some mutations or polymorphisms. Rapid immunological assays examining natural killer (NK) cell activity, degranulation function (CD107a), and protein expression related to genetic deficiencies have been recommended for early pHLH identification. Methods A retrospective analysis of 311 HLH patients from a Chinese population was performed to evaluate the potential correlations between genetic testing and rapid immunological assays; genotyping characteristics, age of onset, and etiology were examined. Results Among the 128 (128/311) patients who were positive in the genetic screening, the most frequently detected mutant gene was UNC13D (29%), followed by LYST (21%), PRF1 (17%), and STXBP2 (10%). Among pHLH patients (n = 39), the majority (67%) had PRF1 and UNC13D defects. FHL-2 was predominant (12/27, 44%) in patients aged under 18, while FHL-3 was the most common (6/12, 50%) in adults. Differences in genetic variant types and etiological components were noted in HLH patients based on the age of onset. NK cell activity and CD107a were observed to show a consistent trend (Ptrend < 0.001) when grouping patients according to the severity of the genetic variant type. Moreover, NK cell activity was generally consistent within a certain range of ΔCD107a values (Ptrend < 0.001). The PPV for bi-allelic degranulation gene mutations in patients with CD107a < 5% was 38.9% (7/18), while the PPV in patients with CD107a ≤10% was 16.7% (13/78). The PPV for pHLH was 41.4% (29/70) with NK cell activity ≤13%. To further evaluate the diagnostic efficacy of NK cell activity assay in pHLH, a receiver operating characteristic (ROC) curve was generated and showed an area under the curve (AUC) of 0.872, and the optimal cutoff value was determined to be 13.425% with a sensitivity of 84.21% and specificity of 80.67% when the corresponding Youden index was maximized. Flow cytometry screening for deficient proteins, including perforin, SAP, and XIAP, showed a relatively high sensitivity (83.33–93.33%). The positive predictive values (PPVs) of perforin and XIAP were relatively low (20.83–26.92%), but the negative predictive values (NPVs) for all three were excellent (all > 98%). Conclusions Various immunological indicators have different clinical prediction and application values for the diagnosis of pHLH. The degree of reduction of immunological indicators also needs attention, and choosing appropriate cutoff value may be of important significance in guiding clinical judgment for pHLH.


Background
Hemophagocytic lymphohistiocytosis (HLH) is a lifethreatening hyperinflammatory syndrome caused by excessive macrophage and lymphocyte activation and can be attributed to various causes. The disorder is usually divided into two categories, primary/hereditary (pHLH), which is autosomal and/or X-linked recessive, and secondary/acquired (sHLH), which is attributed to underlying diseases (e.g., infections, malignancies, or autoimmune disorders). In pHLH, pathogenesis is mainly attributed to excessive immune system activation due to a reduction or absence of natural killer (NK) cell, and cytotoxic T lymphocytes (CTL) functions caused by genetic defects [1]. To diagnose pHLH, the Histiocyte Society recommends utilizing genetic testing, with at least 12 pHLH-related genes (including PRF1, UNC13D, STX11, STXBP2, RAB27A, LYST, AP3B1, SH2D1A, BIRC4, ITK, CD27, and MAGT1) recognized currently [2]. In addition, more primary immunodeficiency disorder (PID)-associated genes have been reported [3][4][5] with an increasingly in-depth understanding of PID-associated HLH. Furthermore, HLH can be considered a threshold disease, with genetic factors and other multiple endogenous and exogenous components interplaying until a critical point is reached, with sHLH also considered to have a certain degree of genetic underpinning [6].
For the early identification and prediction of pHLH, various rapid immunological assays, such as NK cell activity assay, CD107a degranulation assay, and screening for deficiency in protein expression of relevant genes, have been recommended due to their fast turnaround time. Among them, a reduction or absence of NK cell activity is considered an important HLH indicator that reflects immunodeficiency. To assess NK cell activity, various approaches, such as Cr 51 radioimmunoassay [7][8][9], lactate dehydrogenase (LDH) release assay [10,11], and various types of flow cytometry [12][13][14][15][16][17] have been adopted under different conditions and parameters, with different degrees of sensitivity and specificity. Thus far, no uniform measurement standard has been adopted. CD107a, also called lysosomal-associated membrane protein-1 (LAMP-1), is a major component of vascular membrane proteins. Utilizing flow cytometry analysis to quantify degranulation in cytotoxic cells can help rapidly and clearly distinguish conditions with granulocytosis dysfunction [e.g., familial hemophagocytic lymphohistiocytosis types 3-5 (FHL- [3][4][5], Griscelli syndrome type 2 (GS-2), Chediak-Higashi syndrome (CHS), or Hermansky-Pudlak syndrome type 2 (HPS2)] from conditions without granulocytosis dysfunction [e.g., familial hemophagocytic lymphohistiocytosis type 2 (FHL-2), X-linked lymphoproliferative disease (XLP) with deficiency in SLAM associated protein (SAP) or X-linked inhibitor of apoptosis protein (XIAP), or sHLH] [18]. Thus, the CD107a assay is being considered for inclusion in the diagnostic criteria for HLH. In addition, screening for perforin, SAP, XIAP, Munc13-4, synthaxin-11, and Munc18-2 expression via flow cytometry or western blot can also provide rapid identification of relevant defective genes. Once immunological indicators suggest the presence of a genetic basis for HLH, subsequent genetic identification should be performed. The two approaches need to be combined and mutually validated to guide the treatment strategies.
This study analyzed the results of genetic testing and rapid immunological assays in 311 patients with HLH in China to evaluate the accuracy and diagnostic efficacy of these immunological assays and to explore the characteristics of genotyping, age of onset, and etiology.  [19]. Correlations and accuracy between genetic testing results and rapid immunological indicators, including NK cell activity, CD107a, and protein expression (perforin, SAP, and XIAP), were examined. Some patients had multiple results for a given immunological assay, but only the first valid result was selected.

Genetic testing and variant analysis
High-throughput sequencing (targeted gene sequencing panels and WES) and Sanger sequencing were utilized to evaluate the 311 HLH patients. Mutations associated with pHLH, including PRF1, UNC13D, STX11, STXBP2, LYST, RAB27A, ADTB3A, SH2D1A, BIRC4, ITK, CD27, and MAGT1, were examined and interpreted using the Human Gene Mutation Database (HGMD; http://www. hgmd.cf.ac.uk/), the Single Nucleotide Polymorphism database (dbSNP; http://www.ncbi.nlm.nih.gov/snp/), and Ensemble (http://asia.ensembl.org/index.html). The 1000 Genomes Project, Exome Aggregation Consortium (ExAC), and NHLBI Exome Sequencing Project (ESP6500) were referenced for variant frequencies; while Sorting Intolerant From Tolerant (SIFT) and PolyPhen-2 were used for SNP predictions. Variant pathogenicity classifications were determined in accordance with the standards and guidelines recommended by the American College of Medical Genetics and Genomics (ACMG) [20]. The genetic variants selected in this study were those identified by pathogenicity analysis as pathogenic, likely pathogenic, or uncertain significance. These variants also met at least one of the following requirements: (1) the mutation in the population has a minor allele frequency ≤ 0.01 based on the annotation from the ALlele FREquency Database (AL-FRED); (2) the mutation was determined to be harmful/ possibly harmful by at least one of the two prediction algorithms (SIFT or Polyphen-2); (3) the mutation had been reported to be pathogenic in the literature; or (4) the mutation was a nonsynonymous mutation that had not been previously reported in the literature and had no allele frequency annotation or pathogenicity prediction.

Etiological analysis
Differences in etiological components of 311 HLH patients, such as pHLH, infection, malignancy, autoimmune disorders, and other unknown causes, in different age groups, including ≤2 years old, > 2 to < 18 years old, and ≥ 18 years old, were analyzed. Additionally, gene mutations in various underlying diseases associated with sHLH were also examined.

Rapid immunologic assays NK cytotoxicity assay
To detect NK cytotoxicity, a new flow cytometry assay (China National Invention Patent, No. ZL201410005008.7) was used [21]. For this assay, K562 cells, as the standard target for NK cells, were engineered to stably express enhanced green fluorescent protein (EGFP) via lentiviral transfection. Then the EGFP-K562 cells (target cells) were co-incubated with the peripheral blood mononuclear cells (PBMCs, effector cells) of HLH patients based on the specific effector: target cell ratio. After the co-culture of the target and effector cells, the percentage of apoptotic EGFP-K562 cells reflected NK cytotoxicity. To determine the proportion of apoptotic target cells, samples were labeled with Annexin V-PE, and 7-amino-actinomycin D (7-AAD) (eBioscience, San Diego, CA, USA) and flow cytometry were performed.

Perforin, SAP, and XIAP expression in NK cells
PBMCs were isolated and labeled with anti-CD3-FITC, anti-CD8-PerCP, and anti-CD56-APC before cell fixation and cell membrane rupture. The intracellular proteins were further labeled with anti-perforin, anti-SAP, and anti-XIAP. Additionally, these proteins were quantified in NK cells using flow cytometry.

Statistical analysis
Analyses were performed using the IBM SPSS Statistics 19.0 software (IBM SPSS Inc., Chicago, IL). Independentsamples t-test was used for quantitative data with a normal distribution. Differences in etiology were analyzed using a non-parametric Chi-square test. A multi-sample distribution trend comparison of non-normally distributed data was performed using a Jonckheere-Terpstra test. Receiver operating characteristic (ROC) curves were generated and used to determine the optimal threshold (cutoff values) of NK cell cytotoxicity that would identify pHLH patients with maximum sensitivity and specificity (Youden index). For the flow cytometry analyses, the sensitivity, specificity, positive predictive values (PPVs), and negative predictive values (NPVs) were determined for predicting genetic mutations by perforin, SAP, or XIAP expression based on the laboratory-generated normal ranges for each test. Statistical significance is defined as P < 0.05, and the significance threshold was corrected for pairwise comparisons between multiple groups.

General information
The examined HLH patients (n = 311) had a male-tofemale ratio of 167:144, with an age range from 2 months to 74 years old and a median age of 19 years.
Genetic testing was performed on all of the patients, with 128 patients, 67 males and 61 females, found to have potential disease-related variants (Supplementary  Table S1), with an age range of 2 months to 70 years and a median age of 14.5 years (≤ 2 years of age, n = 26; > 2 to < 18 years old, n = 46; ≥ 18 years of age, n = 56). The type of variants included frameshift/non-frameshift, nonsense, missense, and splice-site (upstream and downstream of 5′ splice sites) variants. Furthermore, 183 cases lacked genetic findings (Fig. 1).

Differences in NK cell activity among different mutation types
Of the 311 patients, 307 had their NK cell activity evaluated. When grouping these patients based on the severity of the genetic mutation type, the NK cytotoxicity showed a consistent trend as the mutation type became less severe (P trend < 0.001). When performing a pairwise comparison of different groups, patients with homozygous/hemizygous mutations and compound heterozygous mutations (n = 38) had a lower NK cell activity relative to patients with bigenic/polygenic heterozygous mutations (n = 12; P = 0.014), heterozygous monogenic mutations (n = 75; P < 0.001), or those who had negative findings during genetic screening (n = 182; P < 0.001). Moreover, in patients with bigenic/polygenic heterozygous mutations, NK cell activity was lower than in patients that tested negative for mutations (P = 0.023). However, when comparing NK cell activity between patients with heterozygous monogenic mutations and bigenic/polygenic heterozygous mutations (P = 0.275) or patients that tested negative for mutations (P = 0.132; Fig. 5), no significant difference was noted.

Relationship between CD107a and variants of degranulation-related genes
A total of 260 patients completed the degranulation function test. After grouping patients according to the severity of the genetic variant type in degranulationrelated genes, it was found that the NK-ΔCD107a results also showed a consistent trend (P trend < 0.001). Additionally, pairwise comparison of the NK-ΔCD107a results was assessed and significant differences were identified between biallelic variants in degranulation-related genes (FHL-3,5; GS-2, n = 13) and heterozygous monogenic variants involving one degranulation-related gene (n = 55, P < 0.001), between "potential digenic mode"-heterozygous variants involving two degranulation-related genes (n = 9) and variants not involving degranulation-related genes (n = 32, P = 0.011), and between heterozygous monogenic variants involving one degranulation-related gene and negative findings following genetic screening (n = 151, P = 0.028). However,  no significant differences were noted when comparing biallelic variants in degranulation-related genes vs. digenic variants in degranulation-related genes (P = 0.150); digenic variants in degranulation-related genes vs. heterozygous monogenic variants involving one degranulation-related gene (P = 0.218); or heterozygous monogenic variants involving one degranulation-related gene vs. variants not involving degranulation-related genes (P = 0.127). Moreover, there were no significant differences between patients who had variants in genes not related to degranulation and patients that tested negative for variants (P = 1.000; Fig. 6).

Consistency between NK cell activity and CD107a
Of the 311 examined patients, 256 patients had both NK cell activity and CD107a detection assays performed.

Discussion
Of the Chinese HLH patients who underwent genetic testing (n = 311), 128 cases (41%) had disease-associated mutations ( Fig. 1; Supplementary Table S1), of which 39 were pHLH patients, with FHL-2 (16/39, 41%) and FHL-3 (10/39, 26%) being the most common. In a study examining a Japanese population, FHL-2 and FHL-3 were reported to account for 55 and 32% of FHL patients, respectively [22]. Additionally, in a study examining a Korean population, UNC13D (FHL-3) was shown to be a major pathogenic gene in FHL [23]. In the present study, we found that FHL-2 (12/27, 44%) was predominant in children and adolescents (< 18 years old), while FHL-3 (6/12, 50%) was the most common in adults (≥18 years old) (Fig. 2b). The onset age of FHL-2 patients is earlier than that of other types of FHL, mainly due to the PRF1 mutation causing more severe cytotoxicity damage. Furthermore, we found that hemizygous BIRC4 mutations (XLP-2) also account for a considerable proportion (6/27, 22%) of pHLH patients less than 18 years old. It had been reported that the number and function of T cells and NK cells were decreased and impaired in varying degrees in XLP-2 patients, and the symptoms such as recurrent infection and HLH associated with chronic EBV disease occurred at an early age [24][25][26]. Moreover, in the present study, mutation distribution was also examined for different age groups. Among them, nearly 70% of pHLH (27/ 39, 69%) occurred before the age of 18. Furthermore, pHLH patients with large-fragment variants had an earlier onset age than did pHLH patients without large-fragment variants (P = 0.033).  To explore HLH etiology, different age groups (≤ 2 years, > 2 to < 18 years, and ≥ 18 years; Fig. 3) were examined, and pHLH was found to mainly occur at a young age, with 26.2% (11/42) of HLH cases being pHLH in the ≤2 year group. Morimoto et al. [27] showed that the proportion of FHL in pediatric HLH patients under 1 year of age is nearly 50% in Japan. Herein, the proportion of pHLH cases among HLH patients under one-year-old (n = 8) was as high as 75% (6/8), thus suggesting that in HLH patients under 1 year, pHLH should be highly suspected. Moreover, the proportion of pHLH was roughly inversely correlated with age, with 15.2% of > 2 to < 18 -year group (16/105) and 7.3% of ≥18 years age group (12/164). Notably, among adult HLH patients over 40 years old, the proportion of pHLH is very low, only 3.1% (1/32).
HLH can also be associated with various diseases. In the present study, infection related HLH group was diagnosed after excluding other causes such as pHLH, malignancy, and autoimmune disease. Infection was a Fig. 7 Determination of diagnostic efficacies of CD107a and NK cell activity assays for the presence of genetic abnormalities in patients with HLH. a Comparison of CD107a and NK cell activity assays with genetic testing results. b Empirical ROC curves for the detection of NK cell activity major cause of HLH in the three different age groups with no significant difference noted between groups. However, the proportion in younger age groups (< 18 years) is higher than that in adults (P = 0.031). Infection associated HLH is mostly attributable to EBV infection, which is highly prevalent among Asian populations [28,29] and accounted for more than 90% of the infection associated HLH cases identified herein. Additionally, the proportion of malignancy associated HLH gradually increased with age. In a large-scale Japanese study examining incidence rates of lymphoma associated HLH in different age groups, lymphoma associated HLH was determined in 68% of individuals > 60 years, 38% of individuals 30-59 years, 10% of individuals 15-29 years, and 0% of patients under 14 years [30]. In our study, malignancy associated HLH accounted for 0% of ≤2 years age group (0/42), 3.8% of > 2 to < 18-year group (4/105), and 11.6% of ≥18 years age group (19/164), respectively.
Moreover, this study showed a certain proportion of pHLH-related gene mutations in sHLH patients with different types of underlying diseases (Fig. 4). Recently, HLH has been proposed to be a threshold disease, with there being no clear boundary between pHLH and sHLH because both essentially have the same outcome and are caused by a combination of multiple underlying diseases and pathogenic mechanisms [6]. Notably, this study proposed that sHLH also has a certain genetic background, such as heterozygous variants or polymorphisms in pHLH-related genes, and that sHLH is induced after a "second hit" from an exogenous trigger (e.g., a viral infection).
To compare the correlation between gene mutations and cellular function indicators (i.e., NK cell activity and CD107a), the patients were grouped based on the severity of the genetic mutation type in pHLH-related genes and in degranulation-related genes. In general, cytotoxicity levels were increased as the severity decreased (P trend < 0.001; Figs. 5 and 6). The trend of CD107a, to a certain extent, support the findings presented by Zhang et al. [31] that showed that degranulation-related bigenic heterozygous mutations cause synergistic damage, and are attributed to digenic inheritance. To assess the consistency between the NK cell activity assay and CD107a results, samples were regrouped based on the defined ΔCD107a intervals (< 5%, deficient; ≥ 5% and ≤ 10%, abnormal; and > 10%, normal) following the removal of patients with a PRF1 mutation. While the NK cell activity was found to increase in a fashion that was consistent with the CD107a results (P trend < 0.001) and both indicators reflect cytotoxic function, these two assays measure different things and are not mutually replaceable. NK cell activity reflects the direct killing ability of cytotoxic cells. Impaired or defective cytotoxicity caused by various pathways can lead to a decrease in NK cell activity. Moreover, this factor is also an important indicator of the HLH-2004 diagnostic criteria. Furthermore, CD107a reflects the exocytosis of cytolytic granules and is mainly a parameter for the rapid identification of defects in genes involved in the regulation of vesicle-mediated transport.
In terms of the diagnostic efficacy of the two assays ( Fig. 7a and b), in the present study, a ΔCD107a > 10% was not observed in patients with confirmed FHL-3,5 and GS-2; while patients with confirmed FHL-2 and XLP, but without mutations in degranulation-related genes, did not have a ΔCD107a < 5%. Additionally, only 5.9% (1/17) of patients without degranulation-related gene mutations and 4.0% (6/151) of patients that tested negative for mutations had a ΔCD107a < 5%, thus suggesting that CD107a had significant values in exclusive diagnosis. To evaluate the diagnostic efficacy of NK cell activity assays in pHLH, the detection findings of 307 patients were compared with the genetic testing results, with a ROC curve generated with an AUC = 0.872 (P < 0.001) and a maximum Youden index of 0.649. The sensitivity and specificity were determined to be 84.21 and 80.67%, respectively, and the optimal NK-cell activity cutoff value was 13.425%. While Cr 51 has been considered the gold standard for detecting NK-cell activity, it is difficult to be used universally because of its radioactive contamination and high cost. The new flow cytometry method used in the present study provides a rapid, userfriendly way for measuring NK cell cytotoxicity and has potential to serve as another testing approach in HLH patients [21]. Furthermore, it should be of note that the degree of reduction of cytotoxic function indicators also needs attention. Indeed, in the present study, the PPV for bi-allelic degranulation gene mutations in patients with CD107a < 5% was 38.9% (7/18), the PPV for biallelic degranulation gene mutations in patients with CD107a ≤10% was 16.7% (13/78), and the PPV for primary HLH was 41.4% (29/70) in patients with NK cell activity ≤13%. In clinical practice, in the presence of patients with severely decreased NK cell killing activity and/or defective degranulation function, HLH-related genes should be investigated first.
When examining flow cytometry for the screening of gene-related protein deficiencies (perforin, SAP, and XIAP expression; Fig. 8) in conjunction with associated genetic testing, a relatively high sensitivity (83.33-93.33%) and limited false-negative outcome were obtained. However, for the perforin and XIAP deficiency screenings, the PPVs were relatively low (20.83-26.92%). Therefore, a large number of patients with a positive primary screening would be included as suspected subjects for further genetic testing validation. Moreover, SAP deficiency screening showed a relatively high specificity (98.84%) and PPV (83.33%), thus suggesting that more than 80% of patients with low SAP expression may have an SH2D1A mutation. Importantly, the NPVs for perforin, SAP, and XIAP following deficiency screening were excellent (all > 98%), thus suggesting that > 98% of patients with a negative screening result can definitively be ruled out. Additionally, these protein deficiency screenings were also compared with previous findings from Rubin et al. [32] and Gifford et al. [33], some variability among centers and patients might account for variable PPVs. In addition, the degree of reduction of protein expression in the different mutation types of patients is different. Gifford et al. [33] also optimized the cutoff value of SAP and XIAP expression. Therefore, in consideration of sensitivity and specificity, choosing the appropriate cutoff value may be of great significance in guiding clinical judgment for pHLH. Of note, the protein deficiency screenings measure the quantity of expressed proteins to reveal target gene abnormalities. While most genetic defects cause a reduction in the quantity of the expressed proteins, some missense mutations may result in protein variants synthesized in a normal quantity [33]. Therefore, protein deficiency screenings have certain limitations, and genetic testing should still be performed in some patients with a typical XLP phenotype, but with normal protein expression and no other genetic evidence [33]. In view of these findings, some optimized immunological assays are worth examining as additional testing platforms. For example, researchers have proposed the use of a functional NOD2 signaling pathway assay as a simple and reliable means of identifying a suspected XIAP deficiency [34]. Moreover, given the influence of a UNC13D mutation on platelet granule exocytosis, flow cytometry detection of Munc13-4 protein expression in platelets can be used as an alternative approach for rapid FHL-3 screening [35].
With the wide application of whole exon sequencing (WES) and whole-genome sequencing (WGS), pHLHrelated candidate genes are increasing in number; thus, additional new methods for cellular functional verification need to be explored and incorporated into the rapid pHLH diagnostic system/process.

Conclusions
In summary, this large-scale retrospective study found the gene mutations, and the proportions of etiology are different between children, adolescents, and adult HLH patients in China. For the diagnosis of pHLH, the above immunological indicators had their different clinical prediction meanings and scope of applications. In clinical practice, appropriate cutoff values of immunological indicators may be essential for guiding prejudgment of pHLH.