The BEACH is hot: A LYST of emerging roles for BEACH-domain containing proteins in human disease

Abstract

BEACH (named after ‘Beige and Chediak-Higashi’) is a conserved ~280 residue domain, present in nine human BEACH domain containing proteins (BDCPs). Most BDCPs are large, containing a PH-like domain for membrane association preceding their BEACH domain, and containing WD40 and other domains for ligand binding. Recent studies found that mutations in individual BDCPs cause several human diseases. BDCP alterations affect lysosome size (LYST and NSMAF), apoptosis (NSMAF), autophagy (LYST, WDFY3, LRBA), granule size (LYST, NBEAL2, NBEA), or synapse formation (NBEA). However, the roles of each BDCP in these membrane events remain controversial. After reviewing studies on individual BDCPs, we propose a unifying hypothesis that BDCPs act as scaffolding proteins that facilitate membrane events, including both fission and fusion, determined by their binding partners. BDCPs may also bind each other, enabling fusion or fission of vesicles that are not necessarily of the same type. Such mechanisms explain why different BDCPs may have roles in autophagy; each BDCP is specific for the cell type or the cargo, but not necessarily specific for attaching to the autophagosome. Further elucidation of these mechanisms, preferably carrying out the same experiment on multiple BDCPs, and possibly using patients’ cells, may identify potential targets for therapy.

1. INTRODUCTION

Protein domains are distinct structural and/or functional units within proteins. They are typically responsible for specific interactions or functions. Analysis of a protein’s domains often precedes predictions of its function. Identifying larger collections of proteins sharing the same domain(s) can help formulate hypotheses about the functional roles of such domains (1). Here, we review the expanding eukaryotic family of BEACH domain containing proteins (BDCPs), and aim to identify common mechanisms that aid in elucidating their cellular functions.

The ‘BEACH’ domain was originally identified as a conserved region within the lysosomal trafficking regulator (LYST) protein, mutated in Chediak-Higashi syndrome (CHS) (2). The name BEACH is extracted from ‘Beige and Chediak-Higashi’, where beige is the name of the CHS mouse model (2–4). BEACH domains have subsequently been identified in eight other human proteins (Figure 1): neurobeachin (NBEA), neurobeachin-like 1 (NBEAL1), neurobeachin-like 2 (NBEAL2), lipopolysaccharide-responsive, beige-like anchor protein (LRBA), WD and FYVE zinc finger domain containing protein 3 (WDFY3), WD and FYVE zinc finger domain containing protein 4 (WDFY4), neutral sphingomyelinase activation-associated factor (NSMAF), and WD repeat domain 81 (WDR81) (Table 1). BEACH domains are sometimes preceded by a concanavalin Alike lectin binding domain and/or a PH-like domain and followed and/or preceded by WD repeats (Figure 2).

Figure 1. Amino acid alignment of the BEACH domains of the human BDCPs.

Completely conserved amino acids are highlighted in cyan, semi-conserved in grey, and non-conserved are given in lower case letters with no highlight. Amino acid position numbers are given on the right and represent the last amino acid on each line. These numbers are based on translation of the longest mRNA transcript of each BDCP (shown in Table 1). Note the 8–15 amino acid insertion in LYST compared to other BDCPs. Red boxes represent missense mutations that have been identified in the BEACH domain. Arrows and numbers call attention to the columns with the red boxes. 1 = LRBA - p.R2224W (UDP data); 2 = NBEAL2 - p.P2100L (10); 3 = NBEAL1 - p.R2077Q (UDP data); 4 = NBEAL2 - p.H2263Y (9); 5 = NBEAL2 - p.S2269L (10); 6 = WDR81 - p.G538R (UDP data); 7 = WDR81 - p.A546G (UDP data); and 8 = LYST - p.N3376S (16).

Table 1.

Molecular features of human BDCPs¹

Protein	Synonyms	GenBank mRNA	Locus	Exons	Total amino acids	BEACH domain2 (aa #)	WD40 domains2,3 (aa #)	Other relevant domains2 (aa #)
LYST	CHS, CHS1	NM_000081	1q42.1-q42.2	53	3801	3132..3422	662–700 (WD1) 1582–1626 (WD2) 3563–3778 (WD3-WD7)	1390–1691(ConA-like lectin domain)4 3023–3103 (PH-like domain*)5
NBEA	BCL8B, LYST2	NM_015678	13q13.3	59	2946	2286..25638	1326–1368 (WD1) 2718–2941 (WD2-WD5)	255–402 (ConA-like lectin domain) 1966–2133 (DUF1088)6 2150–2257 (PH-like domain)
NBEAL1	ALS2CR16, ALS2CR17	NM_001114132	2q33.2	55	2694	2004..2284	2439–2531 (WD1-WD2)	577–716(ConA-like lectin domain) 1895–1978 (PH-like domain)
NBEAL2	BDPLT4, GPS	NM_015175	3p21.31	54	2754	2065..2345	1515–1555 (WD1) 2457..2724 (WD2-WD5)	598–728 (Con A-like lectin domain) 1918–2003 (PH-like domain)
LRBA	BGL, CDC4L, CVID8, LBA, LAB300	NM_001199282	4q31.3	59	2851	2201..2478	1301–1343 (WD1) 2580–2846 (WD2-WD6)	205–376 (ConA-like lectin domain) 672–690 (ARM) 1880–2048 (DUF1088) 2065–2172 (PH-like domain)
WDFY3	ALFY, ZFYVE25	NM_014991	4q21.23	68	3526	2695–2976	3077–3447 (WD1-WD5)	1191–1250 (ConA-like lectin domain) 2611–2654 (PH-like domain*) 3449..3515 (FYVE domain)
WDFY4	C10orf64, KIAA1607	NM_020945	10q11.23	62	3184	2540..2821	2930–3184 (WD1-WD5)	2463–2507 (PH-like domain*) no FYVE domain7
NSMAF	FAN	NM_001144772	8q12 -q13	32	948	333..606	645–947 (WD1-WD6)	220–278 (GRAM domain)
WDR81	CAMRQ2	NM_001163809	17p13.3	12	1941	351..5919	1646..1941 (WD1-WD5)

¹Features of the longest mRNA transcripts and translated proteins are listed.

²Domains are extracted from Entrez and the Conserved Domain Database (CDD; 81) by rps-BLAST with default parameters, except for the * marked PH-like domains.

³For WD domains there are multiple position-specific score matrices, leading to slightly different boundary assignments.

⁴The ConA-like lectin domain in LYST was detected by Burgess et al. (20), but is not detectable using CDD.

⁵PH-like domains are hard to recognize by sequence analysis (14, 24). We relaxed the E-value threshold by one (LYST, WDFY3) or two (WDFY4) orders to find the instances marked by * and these PH-like domains may be fragments. This technique did not identify a PH-like domain in NSMAF or WDR81. In the case of NSMAF, the membrane-associated PH-like domain was replaced with a membrane-associated GRAM domain.

⁶The fact that DUF1088 is present in NBEA and LRBA is one indication that these two proteins are more similar to each other.

⁷Despite its name, WDFY4 does not contain a FYVE domain, since WDFY4 is 342 amino acids shorter than WDFY3, lacking the C-terminal segment in which the FYVE domain of WDFY3 is located.

⁸When it was crystallized, the NBEA BEACH domain was identified as spanning positions 2264–2553 (14), but the current Entrez annotation labels the BEACH domain in positions 2286–2563, although the NBEA protein sequence is unchanged.

⁹WDR81 has three shorter isoforms lacking the BEACH domain; in all other BDCPs with multiple isoforms, the BEACH domains are identical in all isoforms.

Abbreviations: ARM domain = ARMadillo/beta-catenin-like repeats; DUF1088 = Domain of Unknown Function 1088; FYVE domain = FYVE zinc finger domain found in Fab1, YOTB, Vac1, and EEA1 (as well as many other proteins); GRAM domain = (found in) Glucosyltransferases, Rab-like GTPase Activators and Myotubularins.

Figure 2. Schematic diagram displaying recognized protein domains in human BDCPs.

The BEACH domain is aligned for all nine proteins and the drawing is to scale, where the scale bar represents 250 amino acids. Note the similarity in number and positions of the WD40 repeat domains following, and the PH domains preceding the BEACH domain in 7 out of the 9 proteins. Other recognized domains include ConA-like lectin, DUF1088, ARM, FYVE and GRAM domains.

The exact functions of BDCPs also remain largely unknown. BDCPs are generally large, and act in seemingly diverse cellular mechanisms, including vesicular transport, apoptosis, membrane dynamics and receptor signaling. This protein family is of emerging clinical importance since several discoveries in the past three years associated genetic variations in their encoding genes with human disorders (Table 2). LYST mutations cause CHS (2–4), upregulation of LRBA may facilitate cancer growth (5) and LRBA mutations cause a generalized autoimmunity syndrome (6), NBEA is a candidate gene for autism (7), NBEAL1 is upregulated in glioma (8), NBEAL2 mutations cause gray platelet syndrome (GPS) (9–11), a polymorphism in WDFY4 is associated with systemic lupus erythematosus (SLE) (12), and a mutation in WDR81 causes a syndrome of quadrupdal locomotion, mental retardation, and cerebro-cerebellar hypoplasia (13). The Mendelian disorders due to mutations in LYST, LRBA, NBEAL2, and WDR81 all share the property that the inheritance is autosomal recessive and heterozygous mutation carriers have no obvious phenotype.

Table 2.

Human disorders and clinical features associated with defective BDCPs and some animal models

Protein	Disorder or condition	Species	Disorder Association	Neurological features	Immune features	Other features	Literature
LYST	Chediak-Higashi Syndrome/beige	Human/Mouse	AR Mendelian mutations/murine Lyst gene disruptions	Yes	Yes	Hematologic Hypopigmentation	(2–4, 16, 37)
NBEA	Body size	Mouse	heterozygous Nbea disruption	No	No		(42)
	Autism	Human	heterozygous NBEA disruption	Yes	No		(7, 45)
	Platelet formation	Human	heterozygous NBEA disruption	No	No		(41, 45)
	Obesity	Mouse	heterozygous Nbea disruption	Yes	No		(46)
	Obesity	Human	NBEA SNPs association	-	No		(46)
	Multiple Myeloma	Human	Somatic deletions, variable NBEA expression	No	Yes	Cancer	(47, 48)
NBEAL1	Glioma	Human	overexpressed NBEAL1	Yes	No		(49)
NBEAL2	Gray Platelet Syndrome	Human	AR Mendelian mutations	No	No	Hematologic Myelofibrosis Splenomegaly	(9–11)
LRBA	Sutoimmune syndrome/CVID	Human	AR Mendelian mutations	No/Yes1	Yes		(6, 53, 54)
	Breast cancer	Human	overexpressed LRBA	No	No	Cancer	(5)
	Prostate cancer	Human	overexpressed LRBA	No	No	Cancer	(5)
WDFY3	Neurodegeneration	Fly		Yes	No		(61)
WDFY4	Systemic Lupus Erythematosus (SLE)	Human	WDFY4 SNPs association	No	Yes		(12, 33, 64)
NSMAF	Subtle immune deficiency	Mouse	Nsmaf knock-out mice	No	Mild	Defective wound healing	(67, 68)
WDR81	CAMRQ	Human	AR Mendelian mutations	Yes	No		(9)

¹Secondary neurological features in 2 patients. AR = Autosomal Recessive.

We describe the BEACH domain organization (Section 2) and provide detailed information about the nine known human BDCPs and their associated human disorders (Section 3). Furthermore, we report occurrence of genetic variants in BDCPs encoding genes extracted from literature and from a large set of exome sequencing data (Section 4). Finally, we propose possible cellular functions of BEACH domains (Section 5).

2. BDCP STRUCTURE AND EVOLUTION

A. BEACH domains

Since the designation of the BEACH domain in the LYST protein, another eight human BDCPs have been identified (Table 1). Crystallization studies revealed that BEACH domains have a backbone fold that was novel at the time of publication (14). The BEACH domain structure is unusual because it contains several hydrophobic segments that do not meet standard criteria to be called beta-strands since they are not sufficiently extended.

By aligning the BEACH domains in the 9 known BDCPs (Figure 1), we can see that BEACH domains consist of ~280 amino acids and are highly conserved, except for WDR81. Comparing the amino acid positions in Figure 1 and the protein lengths in Table 1, one can see that the BEACH domain is usually located at the C-terminal half of the protein (except for NSMAF and WDR81), but not completely at the end (Figure 2). Relative to the other family members, LYST has one large insertion in the middle of the BEACH domain (Figure 1). In one in vitro experiment, a C-terminal fragment of murine Nbea (including BEACH and WD domains) could not produce the phenotype of enlarged lysosomes, whereas a corresponding construct of the C-terminal fragment of Lyst (including BEACH and WD domains) did produce enlarged lysosomes (15). Thus, the insertion in the BEACH domain of Lyst may be functionally important. Alignments of BEACH domains are useful for evaluating sequence substitutions seen in patients. To date, there has been one such published mutation in LYST (16) and there have been three in NBEAL2 (9, 10) (Figure 1).

There have been at least two evolutionary expansions of BDCPs, one in lower eukaryotes and another in vertebrates. Yeast contains only one BDCP, called bph1 (beige protein homolog 1); it is a non-essential protein, but deletion of the gene leads to a mild defect in sorting vacuolar proteins (17). In contrast, Dictyostelium contains six BDCPs (18) and this number stays the same in Caenorhabditis elegans and Drosophila. Humans have nine BDCPs. A phylogenetic tree of human BEACH domains (Figure S1) illustrates the close distances of the pairs NBEA/LRBA and NBEAL1/NBEAL2, which were subject to duplications in the transition from invertebrates to vertebrates. Further evidence of the NBEA/LRBA duplication is that they have orthologous nested genes, MAB21L1 and MAB21L2 inside one intron (19; Table 3).

Table 3.

Nested genes within introns of human BDCPs encoding genes

Gene	Intronic Genes	Gene ID	Strand	Name	Comments
LYST	RNU5E-2P	100873827	same	RNA, U5E small nuclear 2, pseudogene	Pseudogene
	MIR1537	100302139	same	MicroRNA 1537	MicroRNA
	LDHAP2	100190799	opposite	Lactate dehydrogenase A pseudogene 2	Pseudogene
NBEA	PHBP13	100418843	same	Prohibitin pseudogene 13	Pseudogene
	MAB21L1	4081	opposite	Mab-21-like 1 (C. elegans)	Protein coding. Role in eye and cerebellum development
NBEAL1	RPL7P14	100129743	same	Ribosomal protein L7 pseudogene 14	Pseudogene
	LOC100419679	100419679	opposite	Zinc finger protein 670 pseudogene	Pseudogene
	RPL23AP36	100271296	opposite	Ribosomal protein L23a pseudogene 36	Pseudogene
	RPL12P16	100130447	same	Ribosomal protein L12 pseudogene 16	Pseudogene
NBEAL2	-	-	-	-	-
LRBA	ZBTB8OSP1	729566	opposite	Zinc finger and BTB domain containing 8 opposite strand pseudogene 1	Pseudogene
	MAB21L2	10586	opposite	Mab-21-like 2 (C. elegans)	Protein coding Role in neural development
	LOC729558	729558	opposite	Uncharacterized LOC729558	?
	LOC100419161	100419161	same	Nei endonuclease VIII-like 2 (E. coli) pseudogene	Pseudogene
WDFY3	WDFY3-AS2*	404201	opposite	WDFY3 antisense RNA 2*	Noncoding RNA, precedes WDFY3 gene locus
WDFY4	RPL13AP19	100132924	opposite	Ribosomal protein L13a pseudogene 19	Pseudogene
	LRRC18	474354	opposite	Leucine rich repeat containing 18	Protein coding
NSMAF	RPS26P7	619443	same	Ribosomal protein S26 pseudogene 7	Pseudogene
WDR81	-	-	-	-	-

^*Immediately precedes the WDFY3 gene locus on the opposite strand.

So little is known about the ninth human BDCP, WDR81, that this protein was omitted from previous lists of human BDCPs. Three of the four human isoforms of WDR81 are shorter and lack the BEACH domain. In some vertebrates (e.g., cow) all protein isoforms of WDR81 currently in Entrez, lack the BEACH domain, but further investigation of mRNA and protein sequences are needed to confirm this.

B. PH domains

Most BDCPs also contain a pleckstrin homology (PH) domain, a weakly conserved stretch of ~100 amino acids, directly preceding the BEACH domain region (Table 1, Figure 2) (20). PH domains are present in a wide range of proteins involved in intracellular signaling or as constituents of the cytoskeleton. PH domains have a common core fold, which can associate with biological membranes through phospholipid binding (in particular phosphoinositides) and some PH domains bind other ligands including proteins (21, 22). PH domains may assist in targeting their host protein to the cytosolic face of membranes, thus localizing them to appropriate cellular compartments.

In some proteins, PH domains may be difficult to recognize due to minimal sequence homology, however, their three-dimensional structures are remarkably conserved (23). Similarly, in some BDCPs the PH domains are very difficult to recognize by in silico methods, such that they are not annotated as PH domains in Entrez (for LYST, WDFY3, WDFY4; annotated in Table 1). These PH domains may be fragments rather than full domains.

Crystallization studies of amino acid segments of NBEA (14) or LRBA (24) containing the PH-like and the BEACH domains identified that the PH-like domains had a similar fold to canonical PH domains, as well as a strong interaction with their BEACH domains. These studies reasoned, based on weak sequence alignments, that NSMAF also has a PH-like domain preceding the BEACH domain and suggested that the two domains may function as a single unit (24). However, our current use of better databases and more modern domain-finding tools, shows that NSMAF contains a different membrane-associated domain called GRAM (see below and Table 1) rather than a PH-like domain. It was hypothesized that the structural groove/interface (estimated at 1100 square Angstroms) between the PH-like and BEACH domains of BDCPs may be involved in binding a partner (possibly a peptide segment) (14). Unlike canonical PH domains, some of the PH-like domains of BDCPs may have, at best, weak affinity to bind phospholipids, because the binding site seems to be structurally blocked. Furthermore, their surfaces do not have the clustering of positively charged side chains that may be important for binding the highly negatively charged phospholipids (14, 21, 24). Biochemical analysis with a panel of phospholipids confirmed absence of binding to the PH-BEACH domains of LRBA, NBEA (24), but NSMAF does bind phospholipids (25).

C. WD domains

A WD40 repeat (also known as a WD or beta-transducin repeat) is a short structural motif of approximately 40 amino acids, often terminating in a tryptophan and aspartic acid (W-D) dipeptide (26). Several of these repeats are combined to form a WD protein domain. WD-containing proteins have 4 to 16 repeating units, which are thought to form a circularized beta-propeller structure, although WD domains with fewer than 7 WD40 repeats (as in most BDCPs, Figure 2) may be less stable (27, 28). WD-repeat proteins are implicated in a variety of cellular functions ranging from cell cycle control, transcription regulation, signal transduction, to autophagy and apoptosis. The most relevant among the known functions of WD-repeat proteins is vesicle trafficking (27, 28). These proteins are thought to coordinate multi-protein complex assemblies, where WD domains serve as a scaffold for heteromeric protein interactions. PH domains and WD domains co-occur frequently either in the same protein or as binding partners (23). The specificity of the WD-repeat proteins are determined by the sequences outside the repeats themselves.

D. Concanavalin A-like lectin binding domain

Six of the nine BDCPs contain a Concanavalin A (ConA)-like lectin domain, although the instance in LYST is hard to recognize due to a large (WD40 domain) insertion (Figure 2) (20). Lectins bind carbohydrates and the ConA-like lectin domain occurs in a diverse set of proteins with little sequence similarity and diverse functions (20). Burgess et al. (20) proposed that the ConA-like lectin domain in BDCPs could be involved in oligosaccharide binding associated with protein traffic and sorting along the secretory pathway, especially in relation with components of the vesicle fusion machinery.

E. Other protein domains in BDCPs

There are three other domains present in at least one human BDCP, which are FYVE, GRAM, and DUF1088 (Table 1). WDFY3 is the only BDCP that contains a FYVE domain. This domain occurs in several other human proteins and has the function of binding the PtdIns(3)P form of phosphorylated inositide (29). Since several BDCPs also contain PH-like domains, it is interesting to note that many PH domains bind other phosphorylated inositides, but usually not PtdIns(3)P (29). PtdIns(3)P is a marker for endosomal membranes and plays various roles in protein sorting and trafficking across membrane boundaries (29). At the amino acid level, binding is dependent on a perfectly conserved tetramer HHCR motif along with several surrounding residues that are less well conserved (29, 30). The two consecutive H(istidine) residues create a pH dependency, so that the FYVE domain binds much better when the cytosolic pH is low (30). FYVE domain activity is enhanced by binding to acidic lipids and by multimerization (30).

NSMAF is the only BDCP that also contains a GRAM domain (short for (found in) Glucosyltransferases, Rab-like GTPase Activators and Myotubularins)) (31). The GRAM domain is much more widespread than BEACH since it can be found in bacteria and fungi as well as plants and animals (31). GRAM-domain containing proteins have diverse functions, but a common theme is that they function at or near membranes of cells or organelles (32). In NSMAF, the GRAM domain at positions 220–278 immediately precedes the BEACH domain and occupies part of the interval 183–298 that previous studies (14, 24) claimed was a PH-like domain.

The DUF1088 domain (short for domain of unknown function 1088) is shared only by NBEA and LRBA and may indicate a specific function of these two BDCPs not shared by others.

F. Nested genes within BDCPs

Most of the genes encoding human BDCPs have the unusual characteristic of having genes and/or pseudogenes nested within their intronic regions (Table 3). As exemplified by the case of LRRC18 nested in WDFY4 (33), the gene nesting can substantially complicate genetic and genomic studies. If a variant in a gene is associated with a phenotype (as is the case for WDFY4 and lupus, see below), it may be difficult to determine whether the association is truly with the BDCP-encoding gene or the nested gene. If deletions of the BDCP gene are observed (as is the case for NBEA in myeloma, see below), it is unclear whether the deletion of the nested gene contributes to the phenotype. It has not been studied whether expression of BDCP genes and their nested genes are co-regulated.

3. HUMAN BDCPs AND ASSOCIATED HUMAN DISORDERS

A. LYST/CHS1

Lysosomal-trafficking regulator (LYST) is the most studied and first identified BDCP in humans. Although the mutant gene was identified more than 15 years ago, its sequence was relatively unrevealing as there are few motifs with assigned functions. LYST has been proposed to potentially act as a scaffold protein in the mediation of fusion/fission events of vesicles (34). Like most other BDCPs, LYST contains a C-terminal PH-like domain, followed by a BEACH domain and WD repeats (2). The presence of a putative perilipin-like domain (15) and of HEAT/Armadillo-like repeats (2) were proposed, but their precise localization in LYST was not defined (therefore, these domains were omitted from Table 1 and Figure 2).

LYST is the gene mutated in Chediak-Higashi syndrome (CHS), which is a rare, autosomal recessive disorder that can cause severe immunodeficiency, hypopigmentation of the eyes and skin, prolonged bleeding, and neurological symptoms in humans (16, 35). CHS is distinguishable from other hypopigmentation with immunodeficiency syndromes (e.g., Hermansky-Pudlak Syndrome type 2) because in CHS, the granulocytes and some other leukocytes have enlarged lysosomes with abnormal morphology. The large majority of the more than a hundred patients reported to date have a severe, childhood-onset form with all of these symptoms (36). The only effective treatment for these patients is hematopoietic stem cell transplantation (HSCT), and this can cure the susceptibility to infections and bleeding, but not the neurodegeneration or the hypopigmentation.

The CHS immunodeficiency has at least three aspects: defective innate immunity due to protein missorting in neutrophil granulocytes, loss of natural killer (NK) cell function, and defective adaptive immunity due to defective exocytosis, which damages antigen presentation (34, 35). The bleeding diathesis is due to defective platelet delta granule biogenesis. The hypopigmentation is due to defective melanosome biogenesis in melanocytes. A unifying feature of all of these intracellular functions is the involvement of protein sorting and fusion/fission of vesicles.

For understanding which parts of the LYST protein are functionally important, it may be useful to study the 10–20% of CHS patients that have a later onset and fewer symptoms, although the disease is still lethal in adulthood because of the neurodegeneration (37). Karim et al. (37) further subdivided these patients into two categories. The medium patients have onset in adolescence and they may have some immune symptoms, but these tend to diminish with age. The mild patients have onset in adulthood and only neurological symptoms. Karim et al. observed that missense mutations were seen only in the two later-onset forms (37). They reported four such mutations, two of which are located in the ConA-like lectin domain. Westbroek et al. (16) reported the first amino acid substitution mutation in the BEACH domain (N3376S), and they showed that the genotype-phenotype correlation could be seen at the cellular level too. The pigment cells of the N3376S patient have melanosomes and lysosomes that are not as enlarged as those of a childhood-onset CHS patient.

There are animal models of CHS in mouse, mink, cats, and cattle. The mouse phenotype is called beige and the gene was identified at approximately the same time as human LYST (4). It seems that the mouse gene is highly mutable, as several distinct mutations leading to an indistinguishable beige phenotype have arisen in laboratory stocks in the past century (4).

B. NBEA

Neurobeachin (NBEA) is the second-most studied of the BDCPs, after LYST. The NBEA protein has the same trio of domains as LYST, a PH-like domain followed by a BEACH domain followed by WD repeats. NBEA also contains a ConA-like lectin domain and a Domain of Unknown Function 1088 (DUF88), also present in LRBA (Table 1 and Figure 2).

In an early functional study of this protein, NBEA was called BCL8 because of sequence similarity to the important B-cell protein BCL6 (38). The hypothesis that a gene with resemblance to BCL6 would have a B-cell function turned out to be indirectly correct, but the B-cell function is associated with LRBA, the paralog of NBEA (described below, 6, 39).

In 2000, NBEA studies turned away from lymphocytes and towards the brain because Wang et al. (40) showed that expression of NBEA is high in brain; subsequent studies showed additional high expression in endocrine cells and platelets. Wang et al. (40) also showed that NBEA localized near the Golgi apparatus, consistent with the hypothesis that NBEA, as all human BDCPs, has a role in trafficking of proteins and vesicles. NBEA binds to an important signaling complex protein kinase A, but this property may not be shared by other BDCPs, since they do not necessarily share the PKA binding site (40). Reduced expression of NBEA increases vesicle secretion (41).

Various studies, summarized below, consider either humans or mice with disrupted NBEA. Whether the nested MAB21L1 gene is also disrupted and whether that might have additional phenotypic consequences has not been considered. NBEA has been studied in relation to the phenotypes of body length (mice), synaptic spine patterns of neurons (mice), autism (humans), platelet development (humans), obesity (mice and humans), and multiple myeloma (humans).

Studies on near-knockout Nbea mice (with slight residual expression) revealed that heterozygous mice were much shorter/smaller than wild-type mice, and homozygous mice were paralyzed and died perinatally due to lack of synaptic transmissions (42, 43). This may explain why no human patients with biallelic mutations of NBEA have been reported; such a genotype may be lethal in utero or perinatally. In cultured neurons of homozygous Nbea knockout mice, the localization of actin filaments was altered and the number of synapses with spines was reduced, but not the structure of the spines (44). Changes in synaptic spine patterns have been observed in some mental illnesses.

In humans, NBEA was first implicated in autism an autistic patient with a translocation disrupting one allele of NBEA was described (7). Other case studies have been reported in which autistic patients had heterozygous deletions eliminating NBEA, but possibly also adjacent genes (45). Platelets of the autistic patients with disrupted NBEA have dense granules with an abnormal morphology (41). This is interesting both because of the roles of NBEAL2 or LYST in platelet alpha or delta granule formation respectively, and because abnormal platelet dense granules have been observed in approximately 30% of patients with autism (45).

Heterozygous mice for a gene trap disruption of Nbea (46) were, surprisingly, not smaller than normal as described for the heterozygous mice generated by Su et al. (42), and they showed no behavioral phenotype, as was suggested by the human autism studies. Instead, the heterozygous Nbea gene trap mice develop mild obesity due to increase in adipose tissue associated with increased eating (46). The obesity and overeating was responsive to leptin and naltrexone, suggesting that this form of obesity involved neurological signals (46). A targeted genotyping analysis showed that two SNPs in human NBEA are associated with weight and body-mass index (BMI), though this locus was not identified by genome-wide association studies (46).

NBEA is located in a fragile site on human chromosome 13 and is a recurrent target of somatic deletions in multiple myeloma (47). However, expression studies showed variable expression of NBEA, including high expression in some multiple myeloma patients with deletions. Low expression of NBEA would be a surprise because the paralog LRBA is overexpressed in several tumor types (5). In several multiple myeloma cell lines in which NBEA is disrupted by translocation and gene fusion, the gene partner PVT1 on chromosome 8 is the more critical gene, rather than NBEA on chromosome 13 (48). NBEA may be a recurrent fusion partner of PVT1 because the NBEA gene is in a fragile site rather than because of its protein function.

C. NBEAL1

Neurobeachin-like 1 (NBEAL1) and Neurobeachin-like 2 (NBEAL2) are two of the least studied and understood of the nine human BDCPs. NBEAL1 is a typical BDCP, a large protein containing a ConA-like lectin domain, a PH-like domain, BEACH domain and WD domains (Table 1 and Figure 2). In addition, NBEAL1 appears to contain a vacuolar-targeting peptide motif ILPK, which suggests that the protein might be located within the lysosome, however, this has not been confirmed by cellular localization studies (49).

Cloning of NBEAL1 was reported by Chen et al. (49) in 2004, three years after initial sequencing of the human genome. The protein described by Chen et al. contained only 1001 amino acids from the C-terminal portion of the now (in 2012) assumed full length NBEAL1 protein of 2646 amino acids. This partial initial cloning of NBEAL1 is reminiscent of what happened in the gradual identification of LRBA (39, 50). The difficulty in identifying the full length protein indicates that the large size of most BDCPs has been a fundamental limitation in sequencing and biology studies of these genes and proteins.

NBEAL1 is widely, but not ubiquitously expressed: high expression occurs in brain, kidney, prostate, and testis (49). NBEAL1 was found to be expressed at higher levels in gliomas (49), analogous to LRBA overexpression in some other types of cancers (5).

D. NBEAL2

Neurobeachin-like 2 (NBEAL2), like its homolog NBEAL1, contains a ConA-like lectin domain, a PH-like domain, BEACH domain and WD domains (Table 1 and Figure 2). NBEAL2 has no other known functional domains that give clues to protein function. No reports on NBEAL2 gene structure or protein function were available, until recently, when NBEAL2 mutations were found to be the long sought genetic defect of gray platelet syndrome (GPS) (9–11). NBEAL2 is located in a region of chromosome 3 to which the disease gene had been previously mapped (51).

GPS patients have excessive bleeding related to the lack of alpha granules in their platelets and their megakaryocytes (9). The severity of the bleeding varies between patients, even among relatives homozygous for the same mutation (9). Other symptoms of GPS include myelofibrosis (fibrotic bone marrow) and splenomegaly (9). Among the GPS mutations are the amino acid substitutions, P2100L, H2263Y and S2269L located in the BEACH domain of NBEAL2 (9, 10). Although asymptomatic, carriers of one NBEAL2 mutated allele were found to have platelet macrocytosis and significant reduction of platelet alpha-granule content.

Human NBEAL2 is predicted to produce 15 different mRNA transcripts, of which 7 would be protein coding (Ensembl). A unique combination of 4 NBEAL2 transcripts is solely expressed in megakaryocytes and platelets, which may help explain the clinical phenotype of GPS (9).

NBEAL2 is predicted by in silico methods to interact with two other BDCPs, LYST and WDFY3, but this interaction has not been proven in vitro (9). The putative interaction between NBEAL2 and LYST is intriguing since each protein is deficient in a human disorder of platelet granule formation. Platelet delta granules are deficient in CHS (LYST deficiency). Platelet alpha granule deficiency occurs due to mutations in NBEAL2, PLAU, VPS16B, or VPS33B (52). The functional relationship between NBEAL2 and the other three proteins is unknown, but it is speculated that the known interaction of VPS16B with VPS33B establishes the vesicles containing alpha granules, and then NBEAL2 is needed for further maturation of the alpha granules (52).

E. LRBA

Lipopolysaccharide-responsive, beige-like anchor protein (LRBA) was partially cloned in 1992 (50) and fully cloned in mouse and human in 2001 (39). LRBA mRNA is expressed in almost all cell types, with elevated expression in immune cells (6, 39). The LRBA protein is a typical BDCP with PH-like, BEACH, WD and ConA-like domains.

Functional studies identified that LRBA expression is responsive to lipopolysaccharides (LPS), which are carbohydrate-linked lipids at the outer surface of gram-negative bacteria and could trigger an immune response in an immunocompetent human host (39). Wang et al. showed that LRBA is overexpressed in breast and prostate cancer (5). These studies suggested the hypotheses that LRBA could be functionally important for host defenses against infections and for cell proliferation.

Confirming the first hypothesis, Lopez-Herrera et al. identified five patients from four families with biallelic, functionally null mutations of LRBA who came to clinical attention because of a combination of hypogammaglobulinemia (low IgA and IgG leading to chronic infections) and autoimmunity (6). Functional studies showed that the LRBA-deficient patients have inadequate proliferation of B-cells, thus confirming the second hypothesis as well (6). Subsequent studies identified six additional LRBA-deficient patients from two families, also having homozygous null mutations (53, 54). The clinical presentation of LRBA deficiency is variable. All 12 reported patients have severe autoimmunity, but the forms of autoimmunity vary among diarrhea associated with inflammatory bowel disease, idiopathic thrombocytopenia purpura (ITP), arthritis, autoimmune hemolytic anemia (AIHA), and myasthenia gravis (6, 53, 54).

Most, but not all patients present in childhood with hypogammaglobulinemia and recurrent infections. These hypogammaglobulinemic patients were assigned a diagnosis of common variable immunodeficiency (CVID). LRBA-deficient patients with hypogammaglobulinemia represent the most extreme form of CVID that has early onset and autoimmunity (55). The clinical variability is best represented in the family presented in (53) with two nuclear families of cousins sharing the same mutation, such that one branch has hypogammaglobulinemia and the other does not. This variability is analogous to that in GPS, in which affected relatives homozygous for the same NBEAL2 mutation have different disease severity (9).

Since all six germline LRBA mutations reported to date are functionally null, there has been no opportunity to study genotype-phenotype correlation, as has been done for LYST mutations in reference to CHS (16, 37). Whether mild amino acid substitutions lead to any phenotype and if so, how that phenotype will be noticed clinically are pending questions. As for LYST and NBEAL2, heterozygous carriers of LRBA mutations have no phenotype (6).

The PH and BEACH domain segments of NBEA and LRBA were crystallized and structures computed (14, 24). The LRBA structure has very similar characteristics to the NBEA structure, though the LRBA structure has to be rotated by four degrees to achieve the best superposition on NBEA (24). We note that the 414 amino acid sequence used in (24) matches positions 2065–2478 of LRBA (NM_001199282, Table 1), except for one amino acid substitution Y2168F, and the LRBA BEACH domain is annotated in Entrez as occupying positions 2201–2478.

Wang et al. (39) showed in vitro that LRBA co-localized with lysosomes, the trans-Golgi network, the ER, the perinuclear ER, and endocytic vacuoles. These localizations are consistent with a role for LRBA in autophagy. Based on this evidence and the fact that several other BDCPs have a proposed role in autophagy and that the Drosophila ortholog of NBEA and LRBA has a role in autophagy (56), Lopez-Herrera et al. (6) conducted several functional experiments to prove that LRBA-deficient patients have deficient autophagy in their lymphocytes. Their experiments suggest that under conditions where lymphocytes in healthy individuals undergo autophagy, lymphocytes in LRBA-deficient individuals die by apoptosis (6). The excessive lymphocyte apoptosis may be the trigger for the autoimmune phenotypes (57). Moreover, this general mechanism for the onset of autoimmunity suggests that stochastic, environmental factors (rather than genetics) determine in which cell types the autoimmunity manifests in different LRBA-deficient patients.

F. WDFY3

WD and FYVE zinc finger domain containing protein 3 (WDFY3) is one of four human WDFY proteins whose canonical example, WDFY1, was characterized by having repeated WD domains involved in diverse functions including signal transduction (58) and a FYVE zinc finger domain, which has been found in a variety of proteins with roles in endocytosis (59). Two of these four, WDFY3 and WDFY4, are large proteins that also have PH and BEACH domains (Table 1) and hence belong to the BDCP family. The two others, WDFY1 and WDFY2, are medium-size proteins lacking BEACH domains and are therefore, not considered further. The WDFY designation is misleading for WDFY4 because it lacks a FYVE domain (Table 1).

Before mammalian WDFY3 was characterized, Finley et al. (60) discovered the Drosophila ortholog, called blue cheese (bchs). Knockout of bchs in either the homozygous or compound heterozygous states leads to neurodegeneration, protein aggregation, and significantly shorter lifespan in flies (60). This finding influenced later studies to look for a role of WDFY3 in protein clearance, especially of neurologically relevant protein complexes.

WDFY3 is ubiquitously expressed, and its protein product may be found either in the nucleus or in the cytoplasm, but it is not a transmembrane protein (59). Rather, WDFY3 can localize near organelle membranes, where it can bind to phospholipids such as PtdINs(3)P, and the BEACH domain is essential for this binding (25, 59). More specifically, WDFY3 localizes to autophagosomes, implicating this protein (like LYST and LRBA) in autophagy (59). The initial study suggested a role for WDFY3 in macroautophagy, under starvation conditions (59), but more recent experiments identified a role for WDFY3 in more unusual selective autophagy (61). Due to the autophagy function and the FYVE domain, WDFY3 is usually called ALFY standing for “Autophagy-Linked FYve-domain-containing protein”.

WDFY3 can interact with the ubiquitin binding adaptor protein sequestosome 1 (SQSTM1, also called p62), which is required for the delivery of several ubiquitinated cargos to the autophagosome. SQSTM1 can recruit WDFY3 to move out of the nucleus, into the cytoplasm and to the proximity of organelle membranes by binding to the PH and BEACH domains of WDFY3 (62). The WDFY3/SQSTM1 heterodimer then forms a scaffold linking cargo to the macroautophagy complex, including ATG5 (61). Mutagenesis experiments showed that the colocalization of WDFY3 and ATG5 depends only on the WD domains that are on the C-terminal end of WDFY3, beyond the BEACH domain. Although ATG5 is part of the macroautophagy complex, the role of WDFY3 seems to be in a selective autophagy of protein aggregates, named “aggrephagy” (62), rather than general macroautophagy, induced by cell starvation (61).

The canonical example of a neurodegenerative protein aggregate is the expanded polyglumatine tract of mutant huntingtin, which is implicated in Huntington’s disease (61). Since a variety of neurological diseases (sometimes called “triplet repeat diseases”) are associated with expanded polyglumatine tracts in different proteins, it is worth considering the possibility that overexpression of WDFY3 would mitigate the neurological effects by more rapidly clearing the polygutamine aggregates (62).

WDFY3 interacts with SQSTM1 in osteoclasts (63). This location is interesting because heterozygous mutations in SQSTM1 cause Paget’s disease. However, under in vitro starvation conditions to induce autophagy, both wild type SQSTM1 and a construct encoding the most frequent Paget’s disease mutation show the same interaction with WDFY3 in moving from the nucleus to the cytoplasm to form large aggregates, leaving involvement of WDFY3 in Paget’s disease elusive (63). WDFY3 is also predicted to interact with NBEAL2 (9), but no specific role for WDFY3 in the formation of platelet alpha-granules has been established.

G. WDFY4

WD and FYVE zinc finger domain containing protein 4 (WDFY4) has multiple transcripts. The longest transcript encodes a 3184 amino acid protein. Similar to its closest paralog, WDFY3, WDFY4 contains WD40 domains and a BEACH domain. Unlike WDFY3, WDFY4 does not contain a ConA-like lectin domain, and its FYVE zinc finger domain is truncated. WDFY4 is predicted to be a transmembrane protein. WDFY4 is most strongly expressed in immune tissues such as lymph node, spleen, thymus and tonsil, thus it may be important in autoimmune disease (64). The narrow expression pattern of WDFY4 distinguishes it from WDFY3, which is expressed in a wide variety of tissues and organs (64).

WDFY4 came to the attention of medical geneticists when the SNP rs1913517 in WDFY4 was associated with systemic lupus erythematosus (SLE) as part of a genome-wide association study (GWAS) of lupus in Chinese populations (33). Connecting the initial finding to one gene was difficult because rs1913517 is located within LRCC18, which is nested within an intron of WDFY4 (Table 3). This gene ambiguity was resolved by a second, independent GWAS in which 20 SNPs in WDFY4, including rs1913517 were associated with SLE in another collection of Chinese and Thai cohorts (64). Besides the 20 associated SNPs on the genotyping chip, Yang et al. also genotyped a coding SNP in WDFY4, R1816Q, and found that it strongly associated with SLE. However, they did not make suggestions for a mechanism by which the R1816Q variant might be causal.

Instead, Zhao et al. (12) examined genetic effects of several of the intronic, SLE-associated SNPs in WDFY4. The SNP rs877819 best explained the association (12). This SNP is within an intron of WDFY4, but not within LRCC18. WDFY4 expression was lower in SLE patients, while LRRC18 expression was not significantly affected in patients, thus confirming that WDFY4 is the disease-associated gene. In vitro assays confirmed lower transcriptional activity of a WDFY4 constructs that included the minor A allele as compared to the major G allele of the SNP (12). This SNP was identified as a binding site for the transcription factor Yingyang1 (YY1), binding of which is stronger to the major G allele than to the minor A allele (12).

To date, the association of WDFY4 and SLE has not been replicated in non-Asian populations and no rare coding variants have been reported in SLE patients. More research is needed to establish the causal link between WDFY4 and susceptibility to lupus or other autoimmune diseases.

H. NSMAF

The neutral sphingomyelinase activation-associated factor (NSMAF) protein can be viewed as the ‘exception that proves the rule’ in several aspects of characterizing BDCPs. NSMAF differs from all other BDCPs in that it is not exceptionally large, with a size of 948 amino acids (longest isoform). It does contain the BEACH and WD domains, but has a minimal N-terminal segment preceding the BEACH domain and lacks a ConA-like lectin domain. NSMAF is the only BDCP that contains a membrane-associated GRAM domain, seemingly replacing the membrane-associated PH-like domain seen in several other BDCPs.

NSMAF has at least two transcripts that are ubiquitously expressed. NSMAF was cloned in 1996 (65), the same year in which LYST was identified as the gene mutated in CHS. The initial motivation for cloning NSMAF was that it interacts with the tumor necrosis factor (TNF) pathway; specifically, NSMAF interacts with the NSD (N-Smase activating Domain) of the 55kDa form of the TNF receptor (65). Although NSMAF is the current official name, the original and still mostly used name is FAN (Factor Associated with N-Smase activation) (65). Möhlig et al. showed that NSMAF-deficient cells have enlarged lysosomes, like LYST-deficient cells, but not as extreme (58).

Because NSMAF transmits signals from TNF and TNF delivers many immune signals, it was expected that NSMAF would have a role in the immune system. Some specific early evidence for an immune role is that CD40, a signaling molecule important in lymphocyte proliferation, binds to NSMAF (66). To evaluate the possible roles of NSMAF in immunity, NSMAF-deficient mice were generated (67). NSMAF and NBEA are the only two BDCPs for which knockout mice have been generated, but no humans with biallelic mutations have been described.

When a gene has an essential role in immunity, it is typical that knockout of that gene leads to either an immunodeficiency (e.g., LYST deficiency) or autoimmunity (as may be the case for WDFY4) or both (e.g., LRBA deficiency). Therefore, it was surprising that initial phenotypic characterization of the NSMAF-deficient mice showed only a defect in wound healing, but no immune phenotype (67). A later microarray analysis did show that lack of NSMAF leads to a reduction in expression of interleukin 6 (IL6) and various chemokines important in immunoregulation (68). This in vitro finding led to a more exhaustive battery of immune, in vivo challenges of the NSMAF-deficient mice to see if they had a selective or subtle immunodeficiency. Indeed, it turned out that NSMAF-deficient mice have a poor response to the BSA antigen (68), but this does not imply any immunodeficiency or inadequate vaccine response at the level of the whole organism. In particular, NSMAF is not sensitive to stimulation by LPS (68), which was the defining property of LRBA (39).

In its TNF signaling role, NSMAF is pro-apoptotic, enabling TNF to deliver signals to the downstream molecule caspase 8 and other caspases by generating ceramide (66). This was proved by replacing wild type NSMAF with a dominant negative construct in various types of cells including rat cardiomyocytes (69). One effect of the pro-apoptotic signal via NSMAF is to make lysosomes permeable, and this permeability is blocked when a dominant negative form of NSMAF is expressed (70). Pro-apoptotic function of NSMAF is in direct contrast to LRBA whose expression promotes autophagy instead of apoptosis in lymphocytes (6) and cardiomyocytes (71) and WDFY3, which has a central role in autophagy (62).

The crystal structures of the NBEA and LRBA PH and BEACH domains (14, 24) enabled pinpoint investigations on the functions of PH, GRAM, and BEACH domains. The TNF signal through NSMAF stimulates filopodia formation and actin polymerization (25). Two necessary conditions for the signal to be delivered are, first, that NSMAF has to bind phospholipids such as PtdIns(4,5)P and second, NSMAF has to localize to the plasma membrane (25). For the first requirement, mutagenesis experiments showed that the membrane-associated domain (which was thought to be PH, but is GRAM) is necessary and the BEACH domain is dispensable. Mutation of residues K199 and H212 in a basic segment just preceding the GRAM domain is sufficient to eliminate binding to PtdIns(4,5)P (25). However, for the second requirement, both the GRAM domain and the BEACH domain are needed (25), confirming the hypothesis of (14) that the two adjacent domains function as one unit.

I. WDR81

A recent homozygosity mapping study identified a novel BDCP, WD repeat domain 81 (WDR81), to be mutated in one family with a recessive syndrome of quadrupedal walking, mental retardation, and cerebellar hypoplasia (CAMRQ syndrome) (13). Other forms of this syndrome are caused by mutations in VLDLR or CA8 (13), but these are not BDCPs.

WDR81 shares the characteristic of having a BEACH domain followed by WD repeats, but unlike most other BDCPs, WDR81 does not have a recognizable PH domain or ConA-like lectin domain preceding the BEACH domain (13) (Table 1 and Figure 2). Only the long isoform (NP_001157281) of WDR81 contains the BEACH domain; three shorter isoforms (NP_689561, NP_00157145, and NP001157283) do not. Either the unique domain architecture or the variant isoforms may explain why WDR81 was missed in previous reports summarizing the BDCPs. Also in contrast to other BDCPs, WDR81 is predicted to be a transmembrane protein with six transmembrane regions. WDR81 is ubiquitously expressed. WDR81 shares with LYST and possibly NBEA the property of being mutated in a human disease with neurological manifestations. The patients with the WDR81 mutation are not reported to have any immune or bleeding symptoms, though their granulocytes and platelets were not examined (13).

4. GENETIC VARIANTS IN HUMAN BDCPs

Since autosomal recessive mutations in LYST, NBEAL2, LRBA and WDR81 exist in human disease (2, 6, 9–11, 13), we speculated that other BDCP-encoding genes may be mutated in other, previously unrecognized human diseases. To test this hypothesis, we interrogated whole exome data from the National Institutes of Health (NIH) Undiagnosed Disease Program (UDP) for variants in the nine genes that encode BDCPs in humans. This program recruits individuals and families with rare or new diseases where no diagnosis was made before referral to the NIH (72). Molecular genetic testing for these individuals included SNP-chip analysis to reveal any gross genomic alterations, and very often, whole exome DNA sequencing (73). A total of 374 whole exomes have been sequenced for this cohort of patients, and these data were kindly made available to us.

A total number of 1635 variants were identified in the nine genes encoding BDCPs (Table 4). After filtering for variants that (1) are not found in dbSNP 137 and (2) that are protein coding or are predicted to affect splice sites, there were respectively 523 and 45 variants remaining. These 45 variants occur in all regions of the proteins. Missense mutations identified from this study along with those that have been previously reported occurring specifically in the BEACH domains are shown in Figure 1, by red boxes surrounding the altered amino acid.

Table 4.

Genetic variants in human BDCP encoding genes extracted from whole exome sequencing data obtained from the NIH Undiagnosed Diseases Program¹

Gene	TotalVariants	After Filtering		Alleles		Total
		dbSNP	Coding	Missense	Null
LYST	221	59	8	12	6	18
NBEA	216	70	3	3	0	3
NBEAL1	173	74	6	10	0	10
NBEAL2	95	33	9	19	0	19
LRBA	205	69	6	7	2	9
WDFY3	161	80	3	3	0	3
WDFY4	358	73	5	12	0	12
NSMAF	117	36	2	4	1	5
WDR81	89	29	3	5	0	5
Total	1635	523	45	75	9	84

¹The data set of 374 whole exomes acquired by the NIH Undiagnosed Diseases program (UDP) was queried for variants in genes that encode for the 9 BDCPs, and filtered for variants that (1) are not found in dbSNP 137 and (2) that are protein coding or are predicted to affect splice sites. The numbers of variant alleles for each gene are given, either non-synonymous amino acid changes (missense) or null alleles.

In the UDP data set, there were a total of 84 instances of these 45 variants. All but one case only had a single heterozygous variant in any particular gene. The exceptional individual was shown to have two heterozygous variants in LYST (undefined whether they are bi-allelic), was enrolled in the UDP at age 10, and was previously diagnosed with fatty acid hydroxylase-associated neurodegeneration due to recessive fatty acid 2-hydroxylase deficiency, explaining his phenotype (74). For this patient, compound heterozygosity of his LYST mutations will be verified, and his cells will be further analyzed in vitro for a CHS-like phenotype. Due to his young age, and possibly mild LYST mutations, this patient may need monitoring for later onset further neurological decline typical for CHS.

For the other variants, we were expecting either homozygous or compound heterozygous variants in individuals due to the autosomal recessive inheritance pattern for diseases related to mutations in LYST, LRBA and NBEAL2. However, it is possible that other genes that encode BDCPs have autosomal dominant inheritance, that mutations could have been missed because exons of these genes are not well covered by whole exome sequencing, or that we need different cohorts or more patients to be able to identify bi-allelic variants in one patient. Nonetheless, our data suggests that variant alleles are present in the population for each of the nine genes encoding human BDCPs, and that it is possible that patients could be identified in the future with defects in these genes.

5. DISCUSSION

Since the description of the LYST protein more than a decade ago (2–4), its function and that of eight other human BDCPs, remain unresolved. Except for NSMAF (65), human BDCPs are very large, contain a C-terminal common stretch of protein domains, and genetic modifications of their coding genes result in a variety of human disorders with seemingly little clinical similarity. LYST and other BDCPs are often described generically as playing a role in membrane dynamics and/or intracellular trafficking of endosome or lysosome-related proteins and vesicles. However, how these roles are exactly executed by each BDCP remain largely elusive.

Since the phenotype of CHS, caused by LYST deficiency, includes large and mis-shaped lysosomes, it is logical to predict that LYST functions in vesicle fission or fusion, since these events reshape lysosomes and lysosome-related organelles. The recent development of a Drosophila model in which the LYST ortholog mauve is mutated could make the enigma of the cellular function of LYST more tractable (75). Mauve deficient flies have pigment defects in their eyes, an immune defect in killing E. coli, and mis-shaped lysosomes, all of which model the human CHS phenotype (75). However, the cellular experiments on the mauve model could again not resolve the longstanding fusion versus fission debate (75–77). Here, we propose that this debate cannot be resolved in the way it has so far been evaluated.

After reviewing the knowledge on all human BDCPS, we propose that LYST and likely (most) other BDCPs, act in both membrane fusion and fission events as ‘facilitators’ or scaffold proteins. The scaffold protein function of BDCPs is carried out as follows (displayed in Figure 3). BDCPs attach to membranes of the endo-lysosomal system through the structural interface created by folding of their PH-BEACH units. This attachment is likely membrane-specific for each BDCP and it was shown that this may occur through distinctive phospholipid or protein binding (14, 21, 22, 24). Additional phospholipid binding domains of some BDCPs (FYVE domain in WDFY3, GRAM domain in NSMAF) may further function in specialized membrane recognition. Some BDCPs, such as WDFY4, may adopt a transmembrane position. After membrane association, the WD domains of all BDCPS create a large platform for binding partner proteins to assist in executing their specific membrane-related functions. The ConA-like lectin domain can form attachment points for glycosylated ligands. That the Nbea C-terminal fragment did not yield the same lysosomal phenotype as the Lyst C-terminal fragment when overexpressed (15) suggests that BEACH and WD domains in different BDCPs have specificity for binding partners, likely to perform different membrane-localized functions.

Figure 3. Schematic abstract diagram showing the proposed scaffold function of BDCPs.

The example shown here is for LYST and WDFY3 because they are predicted to interact in silico and because these are the only two BDCPs for which binding partners were investigated. However, this hypothesis can be translated hetero- and homo-typic combinations of other BDCPs. (1) BDCPs first recognize their specific membranes through their PH-like and BEACH domain segment, possibly with the aid of other domains more specific to some BDCPs. (2) The BDCPs then interact with each other forming a scaffold holding vesicles or membranes in place for further membrane events. (3) The BDCP scaffold acts as a platform for other proteins and lipids to be recruited and interact with the BDCPs and/or with the vesicular membranes. These binding partners required for subsequent membrane events are largely undiscovered. In the case of WDFY3 and the autophagosome membrane, p62 (SQSTM1) is a critical interacting protein, whose interaction with WDFY3 facilitates further interactions with ATG proteins with defined roles in autophagy (references 59, 61–63).

Tchernev et al. (34) found 21 binding partners for LYST. It seems unlikely that these proteins (such as calmodulin and casein kinase) bind to LYST simultaneously or for the same function. Our alternative hypothesis is that LYST and other BDCPs serve to identify endo-lysosomal membranes that need ‘attention’ for activities such as vesicle fission, vesicle fusion, granule secretion, or melanosome trafficking. We propose that whether vesicle fusion or fission or other cellular functions occur depends on which other binding partners dock with the WD and other binding domains. Since some BDCPs (e.g., NBEA, NBEAL2, WDFY4) are highly expressed in only a few tissues, the function of the BDCP may also depend on the cell type and metabolic status of the cell in which binding partners are available, abundant, and in membrane proximity. If the BDCP provides a scaffold for the proteins primarily carrying out vesicle fusion and fission, this would permit the function-specific proteins to remain positioned near the vesicle membrane, while the membrane is moving. More generally, a multi-purpose support role for BDCPs would explain why so many studies, with the exception of WDFY3 (61) in selective autophagy, have had limited success assigning particular cellular function to any BDCP.

One testable prediction from our scaffold hypothesis is that disease-associated missense mutations that leave the BDCP intact and stable should be relatively uncommon or of milder effect, as compared to other disease-associated genes/proteins. The reasoning is that a missense mutation would be detrimental only if it damages the stability of the protein or mutates a binding site for one of the functional BDCP binding partners. Several studies to date support this hypothesis: 1) missense mutations in LYST are uncommon and are associated with milder disease of later-onset (16, 37); 2) the only missense mutation found to date in LRBA lies in the critical WD domain and affects protein stability (6); missense mutations in NBEAL2 are uncommon and three of them are in the critical BEACH domain (9, 10, Figure 1). Large-scale exome sequencing (Table 4) facilitates searching for adult-onset patients that may have biallelic mutations where at least one of which is mild. Our finding of LYST mutations in one patient already diagnosed with another (seemingly) Mendelian neurological disease also raises the possibility of looking for mutations in genes encoding BDCPs or their partner proteins as modifiers of the phenotype. Pairs of CHS- or GPS-affected siblings with the same mutations and discordant severities have been reported (9, 78, 79), suggesting that modifier genes are waiting to be found, and large scale exome or genome sequencing makes this possible.

A different line of evidence for gene-gene interaction consists of in silico prediction data that various pairs of BDCPs interact (http://www.sabiosciences.com/genenetwork/genenetworkcentral.php). There may be patients in whom mutations or rare polymorphisms of two BDCP-encoding genes contribute to the phenotype. In a cellular perspective, interactions of different BDCPs could enable fusion of pairs of vesicles that are not necessarily of the same type (depicted in Figure 3). Two different BDCPs would first identify and attach to their target vesicles and then bind to each other to bring the heterotypic vesicles in proximity. After which other vesicle-fusion-specific proteins attached to the BDCPs would be in position to complete the fusion. In silico, WDFY3 appears to interact with at least 5 other BDCPs: LYST, NBEAL1, NBEAL2, LRBA and WDFY4. WDFY3 interactions are of interest since the function of WDFY3 was described as implicated in selective autopaghy, rather than general macroautophagy (61, 62). Selective autophagy targets various cargoes such as aggregated proteins and damaged organelles for degradation. Factors involved in selective autophagy include autophagy receptors and adaptor proteins, which connect the cargo to the core autophagy machinery. WDFY3 is so far the only mammalian adaptor (‘scaffold’) protein identified for selective autophagy (80).

In this context, it is notable that recent reports of mutations in human LRBA (6), and LYST in flies (75) affect autophagy. Whether LRBA and LYST are autophagy adaptor/scaffold proteins themselves, or affect autophagy through their possible affinity with WDFY3 remains to be determined. Further elucidation of these mechanisms, possibly using patients’ cells, may identify potential targets for therapy. Furthermore, it would be of interest to evaluate whether the other predicted WDFY3 interactors, NBEAL1, NBEAL2, and WDFY4 also have roles in selective autophagy, either as direct scaffold proteins or indirectly through their -yet to be confirmed interaction- with WDFY3. Such mechanisms partially explain why many different BDCPs could have a role in autophagy; the different BDCPs are specific for the cell type or the cargo, but not necessarily specific for attaching to the autophagosome. One then wonders whether the (in silico) non-WDFY3 associated BDCPs NBEA, NSMAF, and WDR81 are involved in autophagy and/or may interact with WDFY3. In fact, NSMAF was found to be pro-apoptotic (66, 69) which is in direct contrast to the anti-apoptotic function of LRBA (6).

Another aspect that is largely unknown is cellular localization of BDCPs. WDFY3 was described to be mostly localized in the nucleus under normal conditions and recruited to the cytoplasm during conditions favoring the formation of protein aggregates (59). Nuclear localization of WDFY3 might be important, as cytoplasmic WDFY3, through its autophagic interactions may affect an appropriate autophagic response. One could pose the same question for other BDCPs. It is possible that some BDCPs are continuously associated with their target membrane(s), but binding partners change as response to cellular events. And the target membrane may switch between fusion and fission events due to switch of partners binding to BDCPs. Another possible scenario is that, like WDFY3, BDCPs that are not associated with their target membranes are stored in particular compartments to be released when certain (fusion/fission) membrane events need to be initiated.

Despite the overlapping potential roles for BDCPs, most cellular studies to date have studied one BDCP at a time. A disproportionate number of studies have been done on LYST since it is mutated in CHS (2–4) and could thus be justified as of potential clinical importance. By this clinical criterion, the recent discoveries of Mendelian human diseases due to mutations in NBEAL2, LRBA, and WDR81 quadruple the justification for studying BDCPs, perhaps changing the perception of this topic in cell biology from “cold” to “hot”, as suggested by this review’s title. The statistical association of variants in a fifth BDCP, WDFY4, with SLE adds to the medical importance.

There are intriguing clinical overlaps between the BDCP-associated diseases. Four of the five diseases (CHS, GPS, LRBA deficiency, and SLE) have an immune phenotype of immunodeficiency and/or autoimmunity. Two diseases (CHS, GPS) affect platelet granules and it was shown that platelets are atypical in autistic patients with disrupted NBEA. CHS, along with autism associated with NBEA, and WDR81 deficiency all have a neurological component. The role of BDCPs in the brain is further highlighted by the function of WDFY3 in clearing aggregates of neurological-disease proteins (61).

Accumulation of ubiquitinated protein aggregates is a common feature of many neurodegenerative and other protein aggregation diseases. It is tempting to speculate that manipulation of autophagy could be therapeutic for such disorders. Indeed, overexpression of WDFY3 or its C-terminal region led to a significant reduction of protein aggregates in a primary neuronal model of Huntington’s disease and was neuroprotective in a Drosophila eye model of polyQ disease (61).

Now that 4 out of the 9 human BDCPs are known to be mutated in recessive Mendelian diseases and NBEA is associated with autism, it is appealing to predict possible roles of the other four human BDCPs in disease. Mouse studies suggest that NSMAF may be an exception in that loss of murine Nsmaf does not lead to an obvious phenotype (67). With the sparse information available on NBEAL1, we cannot speculate about a possible disease phenotype. Based on the fly blue cheese phenotype (60) and cellular studies of human WDFY3, we suggest that WDFY3 may be altered in adult-onset neurological disease with protein aggregation. Based on the genetic association of polymorphisms in WDFY4 with SLE (12, 33, 64), it is possible that either monoallelic or biallelic mutations in WDFY4 could cause a severe autoimmune disease, perhaps resembling LRBA deficiency.

In summary, we reviewed the knowledge on human BDCPs, with a focus on their domains and their recently discovered roles in human diseases. We proposed that BDCPs function as scaffold proteins to facilitate diverse membrane-related functions including both vesicle fusion and fission. This hypothesis explains most previous BDCP-related membrane findings, some of which had been considered discrepant. If our hypothesis is correct, then the BDCP function will be largely determined by its binding partners, but the identification of such binding partners has been attempted to date only for LYST (34), WDFY3 (61, 63), and to some extent NSMAF (65, 66). BDCP-related membrane fission and fusion events involve changing membrane properties to control lysosome size (LYST and NSMAF), apoptosis (NSMAF), autophagy (WDFY3 and possibly LRBA), granule size (LYST and NBEAL2 and possibly NBEA), or synapse formation (NBEA). Several BDCPs (NBEAL1, WDFY4, WDR81) have not been tested at all for involvement in the above-mentioned membrane remodeling events. Moreover, not all these events have been investigated in relation to each BDCP. We suggest that in the future, novel overlapping and contrasting functions of BDCPs may be best understood by simultaneously carrying out the same experiments on multiple BDCPs. At a minimum, experimental studies of multiple BDCPs together will be necessary to test the in silico predictions that they bind to each other. With more BDCP functional data emerging, possible targets for therapy, including modification of autophagy, can be evaluated for the increasing number of BDCP-related human diseases.