To the Editor,

Hemophagocytic lymphohistiocytosis is a rare syndrome characterized by systemic inflammation, hypercytokinemia and multiorgan failure [1]. Primary HLH typically occurs in early childhood and is caused by pathogenic variants in 12 known HLH genes [1, 2]. Allogeneic hematopoietic stem cell transplantation (HSCT) remains the only definitive curative therapy for pHLH [3, 4]. According to the HLH-2004 protocol, variants in HLH genes serve as independent diagnostic criteria for pHLH and a guide to help select treatment options [5]. However, in nearly ten percent of patients clinically strongly suggested to have pHLH, no pathogenic variants in known HLH genes are present [6, 7].

In this study, we performed WES (n = 12) or WGS (n = 1) in 13 ppHLH parent-child trios (strongly suspected pHLH cases despite lacking a confirmed genetic diagnosis). A total of 6,717,749 variants were successfully called across the 13 ppHLH patients (Fig. 1a). After filtering, we identified 58 genotypes that may contribute to HLH (Additional file 1: Methods and Fig. S1, Additional file 2: Table S1). The majority of the genes annotated by these variants appeared to be patient specific, except for TMEM236 and NBAS. In PPI network analysis, three genes (RAB9B, KLC3 and AP3D1) showed molecular relationships with known HLH genes (Additional file 1: Fig. S2). Finally, only the recurrently mutated gene NBAS and two genes (RAB9B and KLC3) from the PPI network remained after Sanger confirmation and pedigree segregation analysis (Fig. 1b and Additional file 1: Fig. S3).

Fig. 1
figure 1

Identification and bioinformatic characterization of NBAS biallelic variants. a Schematic representation of gene prioritization and validation strategies applied in this study. NBAS genotypes of two ppHLH families in the discovery stage (b) and three families in the replication stage (c). Closed symbols indicate affected patients, and open symbols indicate unaffected family members. d Schematic diagrams of the genomic location of NBAS. e Distribution of NBAS variants identified in this study (top) and the evolutionary conservation of mutated amino acids in the NBAS protein among different species (bottom). All 52 exons of the NBAS gene (reference sequence NM_015909) and two known protein domains of the NBAS protein are represented. ppHLH, presumed primary HLH; WES, whole-exome sequencing; WGS, whole-genome sequencing

Full size image

To examine whether variants in the three candidate genes are present in other HLH patients, we extended our genetic study to a validation cohort of 224 pediatric HLH patients (Additional file 2: Tables S2 and S3). Variants in only NBAS were found in three additional HLH children, with confirmation by Sanger sequencing in a family setting (Fig. 1c). Collectively, a total of five patients in our study carried NBAS biallelic variants (Fig. 1d–e and Additional file 2: Table S4). The estimated frequency of NBAS variants among the pediatric HLH patients was 2.11%, which is lower than that of PRF1 but higher than that of the other 11 known pHLH genes (Additional file 2: Tables S5 and S6). The main clinical characteristics at baseline of all pHLH patients, including the 5 with NBAS variants, are summarized in Additional file 2: Table S7.

To explore the functional relevance of NBAS variants in HLH, we focused on patient P007, who presented with recurrent HLH as soon as therapy was discontinued after complete remission (Fig. 2a). Prompted by a suspicion of pHLH, functional investigations were performed after her second relapse. For both NK cells and CTLs from P007, cytotoxic function and degranulation were defective compared with those of her healthy parents (heterozygous carriers of NBAS); however, progressive recovery in these functions occurred after she received haploidentical HSCT from her father (Fig. 2b). This is consistent with the notion that HLH gene variants reported to date are generally loss-of-function. Therefore, we performed RNA interference of NBAS by using two different shRNAs in the NK-cell line IMC-1, and the NBAS shRNA-targeted (shNBAS) cells showed impaired cytotoxicity and degranulation (Fig. 2c–f and Additional file 1: Figs. S4 and S5), consistent with the abnormalities in the cells from patients.

Fig. 2
figure 2

NBAS is required for cytotoxic granulation in the NK-cell line. a Plot showing the HLH time course, therapeutic approaches and response status in patient P007 with presumed primary HLH. b Functional investigations of cytotoxic lymphocytes. NK-cell cytotoxicity (upper row) and NK (middle row) and T-cell degranulation (bottom row) were defective in P007 compared with her healthy parents but gradually recovered after HSCT. NBAS mRNA (c) and protein (d) levels in an NK-cell line (IMC-1) after NBAS knockdown (n = 3). e Histograms showing the cytotoxic activity of scramble or shNBAS-targeting IMC-1 cells analyzed by FACS after coculture with K562-GFP target cells (n = 3). K562-GFP was used as the negative control, and K562-GFP cocultured with wide-type IMC-1 (without transfection of any shRNA) was used as the positive control. f Histograms show surface CD107a expression, indicating the degranulation ability of scramble or shNBAS-targeting IMC-1 cells in the presence of K562-GFP target cells. g Representative electron microscopic images of sorted scramble or shNBAS IMC-1 cells stimulated with K562-GFP cells for 4 h. LG, lytic (cytotoxic) granule; N, nucleus; M, mitochondria; GA, Golgi apparatus. Scale bars = 500 nm. h The relative number of lytic granules per field was quantified (n = 3). i Representative images showing decreased expression of AP3B1 (magenta; indicated by white arrows) and CD107a (red; indicated by yellow arrows) stimulated by K562-GFP (green) cells for 4 h. Nuclei were stained with DAPI (blue). Scale bar: 10 μm. n = 3. j Schematic depicting the process by which cytotoxic cells kill target cells through the granule-mediated degranulation pathway (left panel). The pop-up panel shows a cartoon of NBAS, along with RINT and ZW10, as part of the syntaxin 18 complex between the endoplasmic reticulum (ER) and Golgi. Data were presented as means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant

Full size image

Considering that NBAS is known to be essential for NK-cell cytolytic function and Golgi-to-ER retrograde transport [8,9,10,11], NBAS may function upstream of the lymphocyte degranulation process. As shown in Fig. 2g–h, knockdown of NBAS decreased the number of cytotoxic vesicles, particularly near the Golgi apparatus. Furthermore, shNBAS IMC-1 cells exhibited significantly decreased expression of AP3B1, an important protein upstream of the known degranulation pathway involved in the transport of cytotoxic vesicles from the Golgi (Fig. 2i and Additional file 1: Fig. S6). Taken together, these findings suggest that NBAS defects disrupt the transport and recycling of proteins or vesicles between the ER and Golgi apparatus and then impact the downstream cytotoxic vesicle transport and degranulation cascades involved in HLH (Fig. 2j).

Biallelic mutations in NBAS have been related to a wide spectrum of symptoms, whereas mutations occurring at Sec39 domain and C-terminus mainly associated with liver failure and multisystemic features, respectively [12]. NBAS mutated HLH in this study was not associated with these clinical manifestations, and 90% of mutations identified clustered within the latter region of the Sec39 and C-terminal domains. It remains to be determined whether differences in mutation spectrum contribute to differential phenotypic manifestations. Collectively, our data provide compelling evidence that the recurrent mutated gene NBAS, known for being involved in transport between the Golgi and ER, is an HLH-predisposing gene that may play a role upstream of the known degranulation pathway in NK cells.