Written informed consent was obtained from all participants and the study was approved by the institutional review board (University of Freiburg ethics committee’s protocol numbers 143/12 and 40/08).

Whole exome sequencing was performed with samples from patients VI and VIII (Table 1). Three μg DNA was fragmented using sonication technology (Diagenode, Seraing, Belgium). The fragments were end repaired and adaptor ligated. Library enrichment process was done by using the Nimblegen SeqCap EZ Human Exome Library v2.0 (Roche, Madison,WI) enrichment kit and sequenced with the Illumina HiSeq2000 instrument (Illumina, San Diego, CA). Data analysis of reads passing filter was done with Burrows-Wheeler Aligner-short in combination with Genome Analysis Toolkit and SAMtools as implemented in Varbank (Cologne Center for Genomics). In-house scripts were applied for filtering against dbSNP, the 1000 Genomes Project, the gnomAD database and an internal database of exome variants. Rare (<1% MAF) missense, nonsense, frameshift, and splice site mutations were taken into account assuming autosomal recessive inheritance.

NK or T cell degranulation was performed from fresh PBMC or cultured PHA/IL-2 blasts as described previously (35). For analysis, CD3−CD56+ NK cells were gated and assessed for surface expression of CD107a (BD Pharmingen). The difference between the percentage of NK cells expressing surface CD107a after K562 stimulation and after incubation with medium alone (ΔCD107a). For analysis of CTL degranulation, CD3+CD8+ T cells were stimulated with anti-CD3/CD28 activation beads (dynabeads) and difference in the mean fluorescence intensity (MFI) of CD107a expression was compared with the unstimulated control sample. Cells were analyzed with flow cytometry (Navios flow cytometer, Beckman Coulter) and FlowJo software.

Peripheral blood mononuclear cells (PBMCs) or lymphocytes from tonsil sample (patient IX) were stimulated with PHA (1.25 μg/mL) and ∼100 U/mL human IL-2 (produced from IL-2 expressing X63 myeloma cells) in the presence of irradiated (30 Gy for 5 minutes) allogeneic PBMCs as feeder cells. Cells were cultured in Iscove modified Dulbecco medium (Gibco, ThermoFisher, UK) with 10% human serum (Sigma-Aldrich, USA) and 1% penicillin and streptomycin (Gibco, ThermoFisher, UK). Every 14 to 18 days, T cells were re-stimulated with fresh IL-2 and feeder cells.

Virus was produced after transfection of HEK293T cells with 6 μg pHR-SIN-RAB27A-GFP, 4 μg pMDG VSV-G (envelope plasmid) and 4 μg pCMV delta 8.9 (packaging plasmid) using TransIT transfection reagents (Mirrus). Virus containing supernatant was harvested after 48 hours and concentrated with LentiX concentration reagent (Clontech, Takara). CTLs were transduced with concentrated viral particles in the presence of 6 μg protamine sulfate (P4020-1G, MilliporeSigma).

Western blot was performed from CTLs as described previously (14). In brief: 2 × 10⁷ cells/ml were lysed in lysis buffer (50 mM Tris–HCl (pH7.4), 150 mM NaCl, 1% Triton X-100, 1% NP-40, 2 mM EDTA, 1x Halt™ protease inhibitors (ThermoFisher)) for 30 min on ice before cell debris was pelleted. Cell lysate was mixed with 4x NuPAGE™ LDS reducing sample buffer (Tris base (141 mM), Tris HCl (106 mM), LDS (2%), EDTA (0.51 mM), 50mM DTT) and heated at 95°C for 5 min prior to separation on a 4–12% NuPage™ Bis-Tris gel in MES running buffer (MES (50 mM), Tris base (50 mM), SDS (0.1%), EDTA (1 mM), pH 7.3 (all ThermoFisher)). To identify protein Plus Protein Kaleidoscope Prestained Protein Standards (Biorad) was used. Protein transfer onto nitrocellulose membranes (Trans-Blot^® Turbo™ Mini Nitrocellulose Transfer Packs (BioRad) was performed with a semi-dry Trans-Blot Turbo Transfer System (BioRad). Membranes were blocked in TBS, 5% non-fat dried milk and 0.05% Tween-20 (Sigma-Aldrich). Incubation of membrane with primary antibodies rabbit anti-RAB27A (Abcam) at 1:100 in blocking buffer at 4°C overnight or 1:5000 rabbit anti-actin (Sigma) for 1h at room temperature. Secondary antibody was 1:10,000 diluted goat anti-rabbit (H+L) HRP labelled (ThermoFisher). Membranes were developed using ECL Prime Western Blotting solution (Amersham) and imaged using a BioRad Chemidoc.

CTLs were used at day10 of culture and were starved overnight without IL-2. CTLs were stimulated with plate-bound anti-CD3e (1μg/mL, UCHT1, BD Biosciences, UK) or with anti-CD3e (1μg/mL) coated P815 cells. Anti-CD107a (2μg/mL, Clone H4A3, eBioscience, UK) was added for the 3hrs stimulation time and cytotoxic T cells were stained with anti-CD8, Clone: MEM-31 (Abcam, UK) or 53-6.7 (BioLegend). Cells were analyzed with flow cytometry (Attune NxT, ThermoFisher, UK) and FlowJo software.

Cytotoxicity was measured as described before (36). In brief, cultured CTLs were stimulated with 1 μg/mL anti-CD3e antibody (clone UCHT1, BD Biosciences – Pharmingen) coated P815 cells, which express a NucLight Red dye. The used CTL-to-target ratio was 10:1. The killing was measured by the reduction of red fluorescence intensity, indicating target cell death. The assay was measured every 30 minutes for 15 hrs using the IncuCyte S3 live-cell analysis system (Essen Bioscience).

CTLs were washed and allowed to adhere for 12 minutes onto hydrophobic mutliwell slides (Hendles-Essex) in FCS-free IMDM. For conjugation assays, blue P815 target cells were coated with anti-CD3 (UCHT1), plated on multiwall slides in serum free media and CTLs were added for 10-15min. Cells were fixed with 4% paraformaldehyde (15710-S, Electron Microscopy Systems, USA) and incubated in quenching solution (50 mM ammonium chloride in 1X PBS) (15 min at RT) and rinsed twice in PBS. Permeabilization was done with 0.2% Saponin in PBS (5 min, RT) followed by washing and blocking for 30min in blocking buffer (1xPB, 1%BSA, 0.2% saponin). Cells were stained with anti-LAMP1 (H4A3, hybridoma supernatant), anti-RAB27A (abcam, UK) for 1 hour, followed by fluorophore-conjugated secondary antibodies (donkey anti-mouse 488 and goat anti-rabbit 647, Invitrogen, UK) (1 hour at RT) together with phalloidin 568 (Invitrogen, UK). Nuclei were stained for 5 minutes at RT with Hoechst (Thermo Fisher, UK), and samples were mounted in ProLong Diamond Antifade Reagent (Thermo Fisher, UK).

The cDNA encoding mouse RAB27A(Asp7Tyr) was prepared by the standard molecular biology techniques using the following mutagenic oligonucleotide with a BamHI linker: 5’-CTAGGATCCATGTCGGATGGAGATTACTATTAC-3’. The RAB27A(Asp7Tyr) cDNA was subcloned into the pEF-FLAG tag expression vector (37) and pEGFP-C1 vector (Takara Bio Inc., Shiga Japan). The RAB27A(Asp7Tyr/Gln78Leu/Cys219Ala/Cys221Ala) cDNA was similarly prepared by PCR using pGBD-C1-Rab27A(Gln78Leu/Cys219Ala/Cys221Ala) (Gln78Leu: constitutively active mutation (38); as a template and subcloned into the pGBD-C1 vector (39). The mouse MLPH/SLAC2-A-SHD (Slp homology domain) (amino acids 1–153) and SLP2-A-SHD (amino acids 1–87) were subcloned into the pAct2 expression vector (Takara Bio Inc.). pEF-T7-MUNC13-4 and pEF-FLAG-RAB27A were prepared as described previously (37, 40).

Melan-ash cells, an immortal mouse melanocyte cell line derived from a RAB27A-deficient mouse (ashen mouse)(obtained from the Wellcome Trust Functional Genomics Cell Bank at St George’s, University of London), and were cultured as described previously (41). Transfection of pEGFP-C1 plasmids into melan-ash cells was achieved by using Lipofectamine^® 2000 (Thermo Fisher Scientific, Waltham, MA) according to the manufacturer’s instructions. Two days after transfection, the cells were fixed in 4% paraformaldehyde and examined for EGFP fluorescence with a confocal fluorescence microscope through an objective lens (×60 magnification, N.A. 1.40) (FluoView 1000-D, Evident/Olympus, Tokyo, Japan). Melanosome distribution assays were performed as described previously (42). The protein expression level of EGFP-tagged proteins in melan-ash cells was confirmed by immunoblotting with anti-GFP rabbit polyclonal antibody (MBL, Nagoya, Japan) and anti-β-actin-HRP mouse monoclonal antibody (Proteintech, Rosemont, IL).

Co-immunoprecipitation assays in COS-7 cells and immunoblotting were performed as described previously (14). Anti-FLAG M2 magnetic beads (Sigma-Aldrich, St. Louis, MO), horseradish peroxidase (HRP)-conjugated anti-FLAG mouse monoclonal antibody (Sigma-Aldrich), and HRP-conjugated anti-T7 mouse monoclonal antibody (Merck Biosciences Novagen, Darmstadt, Germany) were obtained commercially. Yeast two-hybrid assays were also performed as described previously (38, 39). In brief, yeast cells expressing pGBD-C1-RAB27A (Gln78Leu/Cys219Ala/Cys221Ala) (WT or Asp7Tyr) plasmids and pAct2-MLPH/SLAC2-A-SHD or pACT2-SLP2-A-SHD plasmids were streaked on a growth medium (a synthetic complete medium lacking leucine and tryptophan: SC-LW) and on a selection medium (a synthetic complete medium lacking adenine, histidine, leucine, and tryptophan: SC-AHLW) and incubated two days and four days, respectively.

The affected siblings of consanguineous parents presented with susceptibility to severe EBV infections and EBV-driven development of lymphoma (2/7 children). Patient IV was diagnosed with EBV-associated Hodgkin lymphoma with CD30+ cells (Figure 1A). At the time of diagnosis, EBV was detected by PCR in a bronchioalveolar lavage (16469 copies/μg) and blood (604 copies/μg). The patient received stem cell transplantation from his HLA-matched brother (VIII) after his fifth relapse of lymphoma. At the time of transplantation, genetic analysis of patient VIII had not been performed. The patient (IV) died of post-transplant lymphoproliferative EBV-driven disorder (PTLD) at 21 years of age. Patient VI developed CD20+ NLPHL (nodular lymphocyte predominant Hodgkin lymphoma) at 17 years of age, which in contrast to his brother’s lymphoma, was negative for CD30 and EBV antigens (Figure 1B). The patient is in remission after radiotherapy and surgical resection (R0). The youngest sister IX developed EBV-positive chronic-hyperplastic tonsillitis, which required a tonsillectomy due to sleep apnea. There was no sign of malignancy observed in the histological analysis. Her lymph nodes showed presence of abundant neutrophils. Normal T and B cell distribution was seen in the tonsils via flow cytometry. Brother V was EBV seropositive without any clinical manifestation. Patient VIII was PCR negative and serological investigation did not show a history of EBV infection until the age of 13.

Patient (VIII) showed a complex clinical phenotype including metabolic abnormalities with increased creatine kinase (827- 3942 U/L) and myopathy, abdominal pain and hypoglycemia. The patient had suffered thus far from five episodes of lactic acidosis (1.9-7.74mmol/L), abdominal pain, hypoglycemia and encephalopathy. At the age of 8 years, a muscle biopsy revealed a pathological increase of intracellular lipid droplets. The patient suffered from severe muscle pain after 2 hours of normal physical activity. Aspartate-aminotransferase (GOT) was always slightly increased, measuring 50-73 U/L. The forearm ischemia test was pathological (lactate increase from 1.9mmol/L to 4mmol/L). During catabolic episodes (e.g. infection-associated lung inflammation with respiratory insufficiency) at the ages of 9 and 13 years, the patient presented with abdominal pain, vomiting and loss of consciousness. Laboratory results showed hypoglycemia, increased lactate (7.74mmol/L) and increased pyruvate (0.22mmol/L). The episode at the age of 13 years led to hemophagocytic lymphohistiocytosis (HLH). A urine sample showed increased levels of lactate and pyruvate, as well as proline and alanine. Two genetic variants potentially associated with inborn errors of metabolism were detected: Fructose-1,6-biphosphatase (FBP1) deficiency was diagnosed based on a homozygous deletion of exon 1 of FBP1 (FBP1^m/m , NM_000507.4), which contains the translation start codon (Figure 2A and Table 1). Deletion of exon 1 and other FBP1 deletions have been described previously in cases of FBP1 deficiency (43). A diagnosis of ACAD9 deficiency was made based on the detection of a homozygous missense mutation in the acyl-CoA dehydrogenase gene ACAD9 (NM_014049.5; c.206A>G; p.(Gln69Arg), ACAD9^m/m) (Figure 2B and Table 1). This variant is exceedingly rare (allele frequency 0.000016 according to gnomAD) and was classified as deleterious or likely pathogenic by most prediction algorithms including SIFT, PolyPhen-2, REVEL and MetaLR. CADD scores categories this variant in the 0.5% most deleterious single nucleotide variants (score 23.4). Mutations in either of these genes can result in abdominal pain, hypoglycemia and lactic acidosis (44). The homozygous FBP1 deletion was also found in sister IX, whose main clinical manifestations were episodes of hypoglycemia and increased lactate, ranging from 2.7-7.8 mmol/L. Additionally, abdominal pain was reported for brother III by family members, and we suspect that the same FBP1 homozygous deletion or ACAD9 may have been present. This patient died at the age of 1.5 years, in the context of complicated otitis media and hypoglycemia. No DNA was available for genetic confirmation. The homozygous ACAD9 variant was also found in brother VI, who was reportedly suffering from catabolic episodes. A heterozygous ACAD9 variant was assumed to underlie the mildly increased creatine kinase levels and myopathy of brothers V and VII, but no genetic analysis could be performed.

The FBP1 and ACAD9 mutations alone did not explain the susceptibility to severe EBV infections. Whole exome sequencing revealed a previously unknown homozygous mutation in RAB27A (c.19G>T, p.(Asp7Tyr), RAB27A^m/m ) in patients VI, VIII and IX and heterozygosity in the remaining family members (Figure 2C and Table 1). The affected family members did not show any macroscopic signs of hair hypopigmentation or any characteristic pigment accumulations, which are usually seen in the hair shaft of GS2 patients (Figure 3A). As described above, GS2 patients often develop the life-threatening syndrome hemophagocytic lymphohistiocytosis (HLH) which is an indication for stem cell transplantation.

At 13 years of age, patient VIII suffered a flare of HLH, which could be controlled by treatment with corticosteroids, etoposide, and cyclosporine A. Six months later, he underwent stem cell transplantation from a matched (10/10) unrelated donor, after conditioning with treosulfan, fludarabin, thiotepa, and ATG. Post transplantation, the patient experienced acute GvHD IV° (including gut, liver, and skin) and chronic extensive GvHD of the skin, which was resolved by immunosuppressive treatment. Patient IX, who also harbours the RAB27A mutation and EBV infection, has not experienced any HLH symptoms up to the time of writing and the family has so far decided against stem cell transplantation.

To further investigate whether the mutation p.(Asp7Tyr) in RAB27A affects melanosome distribution in cultured melanocytes, an enhanced green fluorescent protein (EGFP)-tagged Asp7Tyr mutant or wild-type (WT) RAB27A were transiently expressed in RAB27A deficient melan-ash cells. The melan-ash cells normally present with black mature melanosomes aggregated in the perinuclear region (Figure 3B) (40, 41). Re-expression of RAB27A (WT) in melan-ash cells completely restored peripheral melanosome distribution (Figure 3B). When RAB27A (Asp7Tyr) was expressed ~90% of the cells still exhibited a perinuclear aggregation phenotype (Figure 3B, right panels). Interestingly, ~10% of the RAB27A (Asp7Tyr)-expressing cells, presented with restored peripheral melanosome distribution (Figure 3B, red arrowheads in the right panels, and Figure 3C). This is in contrast to other RAB27A non-albinism causing mutations e.g., p.Lys22Arg, where a cytosolic localization of mutant proteins is observed (40). Under our experimental conditions, the attenuation of the perinuclear aggregation phenotype observed after transient expression of RAB27A(Asp7Tyr) was not statistically significant (Figure 3C). However, RAB27A(Asp7Tyr) protein expression was lower than that of RAB27A(WT), by at least 2-fold, which may mask the full effects of RAB27A(Asp7Tyr) re-expression (Figure 3D, lanes 2 and 3).

We performed yeast two hybrid assays in melanocytes, to investigate whether the Asp7Tyr mutation affects RAB27A’s ability to bind to melanophilin (MLPH, also known as SLAC2-A) and SLP2-A, both being known interaction partners of RAB27A (45–47) (Figure 3E, SC-AHLW panels). These assays demonstrated normal binding of RAB27A(Asp7Tyr) to SLP2-A but only a weak interaction with MLPH. Tyr-7 is a highly conserved residue in both invertebrate and vertebrate RAB27 (48), and previous structural analyses of RAB27 in a complex with its effectors indicated that Tyr-7 of RAB27 is located in one interphase of the RAB27–effector interaction sites (49, 50). Thus, the reduced MLPH binding activity of RAB27A(Asp7Tyr) could be caused by disruption of one of the RAB27–MLPH interaction sites. SLP2-A has been shown to bind to RAB27A with ten-fold higher affinity (KD = ~ 20 nM) than MLPH (KD = ~ 200 nM) (51), thus the Asp7Tyr mutation may be of little consequence on the RAB27A–SLP-2A interaction.

We investigated the interaction between RAB27A(Asp7Tyr) and MUNC13-4 - necessary for lytic granule priming and fusion - by co-immunoprecipitation assays in COS-7 cells. MUNC13-4 interaction with RAB27A(Asp7Tyr) was decreased by ~50% as compared to RAB27A(WT) (Figure 4A).

We then investigated the exocytosis of lytic granules and cytotoxic function of freshly isolated NK cells. Family members with homozygous RAB27A(Asp7Tyr) (VI, VIII, IX) showed reduced (<10%) lytic granule exocytosis, as measured by CD107a surface expression after stimulation with K562 target cells. NK cells from the RAB27A(Asp7Tyr) heterozygous siblings (V and VII who did not present with any clinical manifestations or genetic alterations pointing to FBP1-deficiency) and parents (I and II) showed normal to slightly reduced lytic granule exocytosis/degranulation (Figures 4B, C). After pre-activation of NK cell with PHA/IL-2, exocytosis of CD107a was restored to normal levels for affected siblings VIII and IX (Figures 4D, E), whereas the NK cells of VI still showed reduced degranulation (Figures 4D, E). NK cell killing capacity after a 4hr co-culture with K562 target cells was reduced for VIII (RAB27A^m/m, ACAD9^m/m, FBP1^m/m). NK cells from IX (RAB27A^m/m, FBP1^m/m) also showed reduced killing, whereas NK cell killing in VI (RAB27A^m/m, ACAD9^m/m) was normal. Family members heterozygous for RAB27A(Asp7Tyr) had normal NK cell killing (Figure 4F). Similar phenotypes were seen after pre-activation of CD8 T cells with PHA and IL-2: lack of exocytosis (as measured by CD107a surface expression) after stimulation in RAB27A homozygous and reduced exocytosis in heterozygous family members (Figures 5A, B). Longer culture under PHA/IL-2 conditions (7-9 days) showed that the RAB27A heterozygous CD8+ T cells from I, II, V and VII were able to kill target cells after 4hrs co-culture with target cells, as well as from patient VI with the homozygous RAB27A mutation. Remarkably, patient VIII and IX showed no CD8+ T cell cytotoxicity. Both patients were also homozygous for the FBP1 mutation, which might influence CD8 T cell serial killing capacity in short-term cultured CD8 T cells re-stimulated with target cells (Figure 5C).

To further evaluate RAB27A protein expression in CD8 T cells, we generated cytotoxic T cell (CTL) lines, by stimulation with allogeneic PBMCs, PHA and IL-2 every 3-4 weeks. To reduce the effects of IL-2 supplementation, which can increase granule exocytosis (35, 52), we starved the CTLs of IL-2 for 16 hours before performing the assays. Expression of mutant RAB27A, as measured by Western blot, was slightly reduced in the CTLs in comparison with the control sample (Figure 6A). Consistently, immunofluorescence staining demonstrated reduced levels of RAB27A (green), while distribution of lytic granules (LAMP1, red) was normal and co-localized with RAB27A (Figure 6B). Degranulation assays with cultured CTLs from homozygous RAB27A-variant family members, revealed impaired granule exocytosis as seen by presentation of CD107a/LAMP1 on the cell surface, after stimulation with plate-bound anti-CD3 (Figure 6C). Additionally, cytotoxicity in CTLs was measured by killing of red-fluorescently labelled target cells over a time period of 15 hours using an Incucyte microscope. We observed slower to normal killing by heterozygous (V) and impaired to absent killing by homozygous RAB27A-deficient patient-derived cells (IX and VIII respectively) (Figure 6D). Of note, patient VIII’s CTLs (RAB27A^m/m, ACAD9^m/m, FBP1^m/m) presented with a total lack of killing activity. To understand if lytic granules are affected in their polarization, we stained conjugates of anti-CD3-labelled P815 target cells (blue) with CTLs using anti-LAMP1 (green) for lytic granules, phalloidin for actin (grey) and anti-pericentrin for centrosome (yellow). Successful formation of an immunological synapse is judged on the level of actin clearance and localization of the centrosome at the cell-to-cell contact site. Lytic granules polarized normally in all patients regardless of RAB27A mutations, with the notable exception of patient VIII, who is the only family member with compounding homozygous FBP1 and ACAD9-deficiency (Figure 6E). Viral transduction and reconstitution of CTLs from patient VI, VIII an IX with wild type RAB27A-GFP only partially rescued the exocytosis defect of lytic granules (Figure 6F). In summary, the RAB27A(Asp7Tyr) variant seems to impair cytotoxic function of NK cells and T cells. Siblings VIII and IX showed no CD8+ T cell cytotoxicity. Both patients also had homozygous mutation in FBP1, which might influence CD8 T cell serial killing capacity in short-term cultured CD8 T cells. In long-term cultured cytotoxic T cell lines, the killing capacity of patient IX’s CTLs was reduced and comparable to CTLs from a previously described homozygous RAB27A (Val143Ala) variant (14). Highly impaired serial-killing and impaired lytic granule polarization was only observed in CTLs from patient VIII, homozygous for the RAB27A(Asp7Tyr) variant, FBP1-deletion and ACAD9 variant (Tables 1, 2).