Biogenesis of lysosome-related organelles complex-2 is an evolutionarily ancient proto-coatomer complex

Hermansky-Pudlak syndrome (HPS) is an inherited disorder of intracellular vesicle trafﬁcking affecting the function of lysosome-related organelles (LROs). At least 11 genes underlie the disease, encoding four protein complexes, of which biogenesis of lysosome-related organelles complex-2 (BLOC-2) is the last whose molecular action is unknown. We ﬁnd that the unicellular eukaryote Dictyostelium unexpectedly contains a complete BLOC-2, comprising orthologs of the mammalian subunits HPS3, -5, and -6, and a fourth subunit, an ortholog of the Drosophila LRO-biogenesis gene, Claret. Lysosomes from Dictyostelium BLOC-2 mutants fail to mature, similar to LROs from HPS patients, but for all endolysosomes rather than a specialized subset. They also strongly resemble lysosomes from WASH mutants. Dictyostelium BLOC-2 localizes to the same compartments as WASH, and in BLOC-2 mutants, WASH is inefﬁciently recruited, accounting for their impaired lysosomal maturation. BLOC-2 is recruited to endolysosomes via its HPS3 subunit. Structural modeling suggests that all four subunits are proto-coatomer proteins, with important implications for BLOC-2’s molecular function. The discovery of Dictyostelium BLOC-2 permits identiﬁcation of orthologs throughout eukaryotes. BLOC-2 and lysosome-related organelles, therefore, pre-date the evolution of Met-azoa and have broader and more conserved functions than previously thought


In brief
Thomason et al. discover that biogenesis of lysosome-related organelles complex-2 (BLOC-2) is an ancient protocoatomer complex widely conserved throughout eukaryotes.In the unicellular Dictyostelium, it functions in a central endolysosomal vesicle maturation pathway, perhaps reflective of its role in the last eukaryotic common ancestor.

INTRODUCTION
Hermansky-Pudlak syndrome (HPS) (MIM #203300) (https://www.hpsnetwork.org) is an autosomal recessive disease affecting the biogenesis and function of lysosome-related organelles (LROs).Its prevalence is 0.5-1 3 10 À6 (https://www.orpha.net),although this is much higher in a region of Puerto Rico, 1 where incidence is 1/1,800 and the carrier rate is 1/21. 2,35][6][7] General features include oculocutaneous albinism, visual impairment, and bleeding disorders; the more severe subtypes include immune deficiency or life-threatening lung fibrosis. 4,8,9][12] LRO dysfunction underlies all types of HPS. 10,11These organelles store and commonly secrete, by exocytosis, cargoes acting as physiological mediators, [13][14][15][16] and include platelet-dense granules, melanosomes, and lung alveolar lamellar bodies.Through these well-documented roles, LROs are often viewed as being specialized cellular compartments having tissue-specific functions. 12However, because all LROs are derived at least partly from endosomes 17 and can acquire contents from multiple trafficking origins (macropinosomes, sorting and recycling endosomes, lysosomes, and biosynthetic secretory pathway), they can be considered more fundamentally as elaborations of core vesicle-trafficking pathways, and the name endolysosome-related organelles (ELROs) has been proposed to reflect this. 17Indeed, compartments similar to animal cell LROs have been recognized in unicellular organisms (reviewed in Bowman et al. 11 and Delevoye et al. 17 ), but, significantly, much of the known biosynthetic machinery has seemed to be lacking outside of Metazoa.
A molecular function of BLOC-2 remains elusive. 7,11Mammalian BLOC-2, isolated from cells, is thought to be a trimer of HPS3, -5, and -6 subunits, [38][39][40] all of which are sizable proteins (1,004, 1,129, and 775 aa, respectively, for the main isoforms).They have no conserved domains or similarity to other proteins, save for a putative WD-repeat domain at the N terminus of HPS5. 38This lack of recognizable properties has restricted progress in understanding BLOC-2 function.Studies in melanocytes again provide the best evidence: BLOC-2 localizes to endosomal tubules 41 and is required to stabilize-possibly by tethering-connections between BLOC-1-dependent, VAMP7-containing tubules and maturing melanosomes. 42Documented BLOC-2 binding partners include BLOC-1, 41 a number of Rab GTPases, [43][44][45][46] clathrin, 47 and dynactin, 48 all of which have roles in trafficking pathways.
By searching databases with available BLOC-1, -2, and -3 sequences, several subunits were identified in non-metazoan organisms. 49This gave rise to the suggestion that the BLOCs arose early in evolution, but this conclusion was based almost entirely on three BLOC-1 subunit genes found in several of the surveyed organisms.It was subsequently realized that these subunits are not specific to BLOC-1 but are shared with another complex, the then-unknown BORC. 50Once this is accounted for, only one organism lends support for the ''early origin'' hypothesis: Dictyostelium.
We and others have previously studied vesicle trafficking in Dictyostelium, [51][52][53] a tractable system with highly conserved pathways.5][56][57] Using a genetic screen for disrupted trafficking, we have now identified a complete and functional BLOC-2 in Dictyostelium and characterized its composition and activity.We explore specific roles of each subunit and how the complex is recruited to vesicle membranes.Using these novel sequences, we discover candidate genes encoding BLOC-2 subunits in a wide array of unicellular eukaryotes, greatly extending the range of organisms in which the genes are conserved.We use structural modeling to explore the properties of BLOC-2 subunits, finding that they all belong to the proto-coatomer family, many members of which have roles in vesicle trafficking.Our findings have implications for the cellular role of BLOC-2 and of LROs themselves.

RESULTS
Trafficking mutants identify a tetrameric Dictyostelium BLOC-2 To extend our understanding of the genetic control of vesicle trafficking, we employed a reporter assay.Cells ingest fluorescent dextran by macropinocytosis, the compartments sequentially becoming endosomes, lysosomes, and post-lysosomes (PLs).9][60] Mutants with defects in this pathway have delayed or blocked exocytosis and increased fluorescence compared with wild-type (WT) cells.Using this assay, we previously identified Dictyostelium MROH1, a regulator of lysosomal maturation. 53In the current screen, one strongly affected clone was in a gene (Dictybase: DDB_G0275889; 1,469 aa) previously described as a putative homolog of HPS3. 49Direct BLASTP alignment showed contiguous similarity with human HPS3 over 320 aa, supporting this identification.As human HPS3 is a component of BLOC-2, 39,40 we searched the Dictyostelium genome for the other expected subunits: we identified a potential HPS5 ortholog (as done previously 49 ), but no HPS6 could initially be found.
To identify interacting proteins, we expressed Dictyostelium HPS3-GFP and performed GFP-nanotrap pull-downs from cells.Samples were analyzed by mass spectrometry: three potential specific interactors were found (Table S1).One (Dictybase: DDB_G0272600; 1,536 aa) was the putative HPS5 protein.
The two others (Dictybase: DDB_G0279509, 1,198 aa; and Dictybase: DDB_G0283369, 2,083 aa) were previously undescribed.Sequence searches of the larger gene product returned alignments dominated by proteins containing an RCC1-like domain. 61his occupies only a small proportion ($350 aa) of the protein, and searches of the remainder of the sequence showed it to be an ortholog of a family that includes Drosophila Claret 62,63 and C. elegans GLO-4 (also known as X-linked retinitis pigmentosa GTPase regulator homolog). 64Both have been linked to LRO function in these organisms.The smaller protein was more cryptic, returning no similarity when its amino acid sequence was searched against Metazoa using BLASTP 65 or HMMER. 66However, a search against the human proteome by the structure-based predictor HHPRED 67 yielded an unambiguous match: HPS6.This indicates that Dictyostelium HPS6 has diverged considerably from its metazoan counterparts yet is structurally homologous.It also explains why no HPS6 was found in previous searches. 49llowing band identifications: C (DDB_G0283369, i.e., Claret), 5 (DDB_G0272600, i.e., HPS5), 3 (DDB_G0275889, i.e., HPS3), and 6 (DDB_G0279509, i.e., HPS6).The bands marked * are the GFP-fusion protein.Lanes 1-3 are from one experiment, lane 4 from another.MW, molecular weight (kDa).(B) TRITC-dextran exocytosis assays.Left graph: WT strains plus hps3 KO and hps5 KO and rescues; middle graph: hps6 KO and claret KO and rescues.All mutants had a trafficking t 1/2 > 5 h compared with a WT t 1/2 of 1 h.Graphs show averages and SD, n R 4 for each strain.Right graph: area under the curve and statistical analyses (one-way ANOVA with Tukey's correction for multiple comparisons) for all mutants vs. their respective parents and rescues: adjusted p < 0.0001).Each point on the graph represents one experiment for that strain.(C) Delayed vesicle neutralization in mutants.Cells incubated in FITC/TRITC dextran.In WT cells, neutral vesicles (white) were abundant by 65 min ($half-time of dextran trafficking) but not seen in mutants until R3 h.All mutants behaved similarly, as did wash KO , though hps5 KO   We cloned the three genes and expressed them as GFP-fusions, followed by nanotrap pull-downs.All identified each other, and HPS3, as specific binding partners (Figures 1A and S1A; Table S1).To test whether the four proteins form a stably interacting complex, we prepared cytosol from HPS6-GFP cells and performed size-exclusion chromatography.Column fractions were subjected to GFP-trap followed by SDS-PAGE and silver stain or western blot (for GFP).All four proteins precisely co-eluted stoichiometrically (Figure S1B).Additionally, due to its over-expression, there was a second peak of HPS6-GFP alone (Figures S1B and S1C).From their sequences, the estimated molecular mass of the complex is expected to be 700 kDa, but it eluted at an unexpectedly large estimated hydrodynamic radius of 180 A ˚(the excess HPS6-GFP had a largerthan-expected estimated radius of 69 A ˚).This suggests that the complex could be in an oligomeric form (multiple of each subunit) or that its structure is not globular.
We disrupted each gene by homologous recombination in the laboratory strain Ax3 68 (as used for our library of exocytosis mutants) or, for HPS5, the very similar strain Ax2 (the hps5 locus being duplicated in Ax3).All four mutants had severely impaired exocytosis of fluorescent dextran, with a trafficking half-time > 5 h compared with $1 h for WT (Figure 1B).Expression of GFP-tagged versions of the proteins (as used in the pulldowns) restored normal trafficking.
Trafficking of dextran could be disrupted at several different steps of the endolysosomal pathway.In order to understand the nature of the fluorescent compartments in the mutants, we used a mixture of pH-insensitive tetramethylrhodamine isothiocyanate (TRITC)-dextran (Figure 1C, magenta) and fluorescein isothiocyanate (FITC)-dextran (Figure 1C, green), which is quenched at low pH 51,69 : acidic lysosomes therefore appear magenta whereas neutral PLs are white. 51,69We found that all four mutants had delayed vesicle neutralization (Figures 1C  and 1D).In WT cells, PLs were prominent at 1 h (roughly the half-time of trafficking) but absent from mutants; instead, they appeared slowly over several hours.PLs were not seen in wash KO , assayed in parallel.The proportion of PLs was calculated for each strain and plotted (Figure 1D), highlighting that this is the primary defect in BLOC-2 mutants.

BLOC-2 is an ancient and widespread proto-coatomer
The four proteins appear to form a tetrameric Dictyostelium BLOC-2, comprising orthologs of HPS3, HPS5, HPS6, and Claret.That the HPS6 sequence is so diverged from known orthologs presented an opportunity to find others that might not have been previously recognized.We used an iterative method to search widely in eukaryotes: first HMMER, to identify putative HPS6-related sequences, then HHPRED (against human and Dictyostelium) to confirm or reject these; positive hits were then used as new HMMER queries.This way, we identified orthologs in many families of organisms where there was no prior evidence of genes for BLOC-2 proteins, including choanoflagellates, protists, and oomycetes (Figure 2A).Interestingly, we identified HPS6 orthologs in arthropods but not in insects, including Drosophila, Apis, and Bombyx (using HHPRED), suggesting that it has been lost.
We carried out similar searches for HPS3, HPS5, and Claret, and extended the distribution of these genes across eukaryotes (Figure 2A).Most organisms that possess any BLOC-2 subunit genes contain all four; exceptions include those mentioned for HPS6, and for Claret, which, though found in animals, appears absent from vertebrates.All four genes are unambiguously ancient and, given their widespread distribution, were likely present in the last eukaryotic common ancestor. 70,72,73This further suggests that BLOC-2's molecular role, as well as compartments with properties of LROs, are evolutionarily ancient.
Structural modeling with AlphaFold 71 suggests that all four proteins are homologous, having an N-terminal b-propeller followed by a-solenoid occupying the rest of the protein (Figure 2B).Claret is predicted to have this arrangement in tandem.[76] BLOC-2 localizes to endolysosomes via its HPS3 subunit In mammals, BLOC-2 is required for maturation of developing LROs, many of which are exocytic, 14,15,19 and in Dictyostelium, BLOC-2 is involved in the maturation of exocytic lysosomes.Where does BLOC-2 act?To probe this, we examined GFP-fusions of HPS3, HPS5, HPS6, and Claret.All four localized to compartments resembling lysosomes (Figure 3A).When fluorescent dextran was added, this labeled the same vesicles, showing they derive from endocytosis.
To address whether all four proteins were present on the same vesicles-an expectation of proteins forming a complex-we co-expressed pairs tagged with EGFP and mRFPmars.In all cases, they were co-localized on the same compartments (Figure 3B).From these co-expressing cells, we immunoprecipitated each GFP-tagged protein and probed for the mRFP-tagged protein, confirming in all cases that the subunits co-precipitated (Figure S2 related to Figure 3).
None of the constituent members of mammalian BLOC-2 (HPS3/5/6) have any specifically ascribed roles.We investigated the subunits in Dictyostelium by examining their individual behavior in each mutant.In hps3 KO the other subunits were entirely cytosolic (Figure 4A, top row): HPS3 is thus essential for their localization.Conversely, HPS3-GFP still localized-more strongly than normal-in all the other mutants (Figures 4A, left column; Fig- ure 4C).The vesicles so labeled were larger than those in WT cells ($50% increase in diameter) (Figure 4D), with an estimated 3-fold higher surface content of HPS3-GFP.fungi.HPS6 is present in arthropods but has not been identified in insects.Claret is present in all animal phyla examined, including Chordates (e.g., Branchiostoma), but is not found in vertebrates.The tree is based on Burki et al., 70  Disruption of hps5 did not prevent localization of any of the remaining subunits (Figure 4A, second row).Knockout of hps6 and claret had intermediate effects that were similar to each other but subtly different (Figure 4A, 3 rd and 4 th rows).In both, HPS5-GFP was again completely de-localized.In claret KO , HPS6-GFP was cytosolic, whereas in hps6 KO , Claret-GFP showed some residual vesicular localization (but less than in WT cells).The contrasting behavior of HPS3 and HPS5 was confirmed by co-expressing them in hps6 KO and claret KO : HPS3 was strongly localized to vesicles but, in the same cells, HPS5 was cytosolic (Figure 4B).
To complement these localization studies, in each mutant, we co-expressed pairs of tagged (EGFP/mRFPmars) proteins and tested their co-immunoprecipitation (Figures 4E and S3).All gave reliable coIP, except the following, where it was lost or greatly reduced: in hps6 KO , HPS3/HPS5 and HPS5/Claret; in claret KO , HPS3/HPS5 and HPS3/HPS6.These findings closely

B (legend on next page)
parallel the localization results, and from them we propose roles for the different subunits: HPS3 is absolutely required for the BLOC-2 complex and any of its subunits to localize to endolysosomes.HPS5 will only bind when all other subunits are present in a complex; it is thus likely to be the most distal subunit from the vesicle (and from HPS3).HPS6 and Claret appear to perform an intermediate role between HPS3 and HPS5.
BLOC-2 arrival and departure BLOC-2 localizes to lysosomes and is important for their neutralization and subsequent exocytosis.When does it arrive and when does it leave?To answer this, we used HPS3-GFP as a marker and did time-lapse imaging during pulse-chase experiments.
At the population level, HPS3-GFP vesicular delivery began about 20 min after dextran addition, increasing to a plateau by 45-50 min, at which point all dextran-containing vesicles were HPS3-positive (Figures 5A and 5B).To examine HPS3 arrival at individual vesicles, we used rapid z stack imaging at intervals of 6-8 s.It was only possible to image cells for a few minutes due to the harmful effects of the illumination.Nevertheless, we observed vesicles transitioning from unlabeled to fully labeled.The HPS3-GFP signal first appeared as puncta, and over the ensuing 5 min the vesicle became decorated with a smooth continuous outline (Figure 5C; Video S1).Therefore, although the cell's population of vesicles takes about 25 min to fully transition to being HPS3-positive, individual vesicles convert quite rapidly (Figure 5D).
To determine the cycle of BLOC-2 localization, we monitored lysosomes throughout their lifetimes.HPS3 was present until shortly before exocytosis: $30 s prior to this event, HPS3-GFP rapidly and smoothly dissipated from the vesicle surface.For the last 20 s of its residence in the cell, when it was docked with the plasma membrane, the compartment was devoid of HPS3 (Figures 5E and 5F; Video S2).We conclude that once it has been recruited to lysosomes, BLOC-2 is present on them for the rest of their lifetime and is recycled into the cytosol just prior to vesicle exocytosis.

BLOC-2 localizes to the same lysosomal compartment as WASH and is required for its efficient recruitment
The defective PL exocytosis in BLOC-2 mutants is secondary to delayed neutralization.Wiskott-Aldrich syndrome protein homolog (WASH) is the critical regulator of this step, its delivery to acidic lysosomes inducing filamentous actin (F-actin) polymerization and V-ATPase removal. 51,52BLOC-2 mutants phenotypically resemble wash KO , and we therefore sought to investigate functional links between them.Co-expression of GFP-WASH and HPS3-mRFPmars showed two populations of vesicles: the majority having WASH and HPS3, and a minority only WASH (Figure 6A).For those vesicles having both, their surface distribution differed: WASH was patchy, as previously observed, 51,52,77 whereas HPS3 was continuous.
We examined the proteins' behavior during the time window of their expected arrival onto endolysosomes (20-25 min after dextran uptake).Z stack images showed that dextran-containing vesicles that were positive for one protein were positive for both.At these early stages of lysosome maturation, the proteins were present in one or a few puncta (Figure 6B) and, although on the same vesicles were not co-localized.This did not reveal the source of the WASH-only vesicles, but as WASH also has an early phase of vesicular localization at newly ingested macropinosomes, 77 we examined each protein during an acute dextran pulse.GFP-WASH labeled vesicles beginning at 1 min (the earliest time point), reaching a maximum at 3-5 min (Figures 6C and S4A); HPS3-GFP did not label any vesicles in this assay.Therefore, only WASH, but not HPS3, associates with early endosomes, 77 while both are recruited to endolysosomes.
We could not jointly examine HPS3-mRFP/GFP-WASH arrival onto lysosomes in greater temporal detail due to their relatively low co-expression levels.Instead, we imaged HPS3-GFP with the WASH effector, V-ATPase.On lysosomes (but not early endosomes 77 ), WASH arrival catalyzes the removal of V-ATPase and, consequently, shows an inverse distribution with it. 51Accordingly, we observed reciprocal dynamic behavior of V-ATPase with respect to HPS3 (Figure S4B; Video S3), with them transitioning over a few minutes.This further supports the temporal and functional correlation of WASH and BLOC2 in lysosome maturation.Further, we found WASH and BLOC-2 on PLs until just before exocytosis (Figure S4C; Video S4).Their localization therefore spans their near-synchronous arrival and departure.
How do the two complexes influence each other's localization and function?Using hps5 KO to represent BLOC-2 deficiency, we  KO , all other subunits completely de-localized.Conversely, HPS3-GFP strongly localized in all other mutants.In hps5 KO , all other subunits still localized.hps6 KO and claret KO behaved similarly to each other: HPS3-GFP localized, whereas HPS5-GFP did not.The mutants were subtly different: HPS6-GFP did not localize in claret KO , but Claret-GFP still localized (to a small extent) in hps6 KO .Scale bars, 5 mm.(B) Localization of HPS3-mRFPmars (magenta) and absence of localization of HPS5-GFP (green) when co-expressed in hps6 KO and claret KO cells.Scale bars, 5 mm.(C) Quantitation of localization of HPS3-GFP in WT vs. mutants.Signal around vesicle was measured at the focal plane of greatest intensity and expressed relative to the cytosolic signal.HPS3-GFP localization increased by 76% in hps5 KO , 56% in hps6 KO , and 39% in claret KO vs. WT cells (and vs. hps3 KO -rescue).5 independent experiments were done (shown by different colored dots); statistical tests done on experimental means: hps5 KO vs. Ax2 (two-tailed paired t test); hps6 KO and claret KO vs. Ax3 (paired measures ANOVA with Geisser-Greenhouse correction, and Dunnett's correction for multiple comparisons).(D) Diameter of HPS3-GFP-positive vesicles in WT vs. BLOC-2 mutant strains, using data from the 5 experiments shown in (C).Vesicles were measured at their cross-section of maximum diameter from z stacks.Paired analyses were carried out using the statistical tests as in (C).Experiments indicated by green and orange symbols used 10% dextran, the others used 2% dextran.This changed vesicle size in all strains but did not affect the differences between strains.For (C) and (D): total number of vesicles analyzed: Ax2 1,537, hps5 KO   found GFP-WASH to be sparsely localized on HPS3-positive vesicles (Figure 6D), but localized normally on early endosomes (which lack HPS3).In wash KO , BLOC-2 (shown by HPS3 and HPS5) localized strongly on vesicles, with a normal smooth appearance (Figure 6D).This suggests BLOC-2 acts genetically upstream of WASH in its localization onto endolysosomes.
In BLOC-2 mutants, we tracked the behavior of GFP-WASH with respect to vesicle age (dextran pulse-chase), and found that its arrival was delayed (Figures 6E and 6F).In WT cells, GFP-WASH strongly labeled these compartments (in characteristic large patches) by 1 h.In BLOC-2 mutants, GFP-WASH arrived later and in a lower concentration, appearing as sparse puncta for >2 h (reminiscent of the early stages of its arrival in WT cells [cf. Figure 6B]).The delayed and muted recruitment of WASH likely underlies the impaired vesicle neutralization and exocytosis.Similarly, the lysosome regulator MROH1, which we have previously found to be co-localized with WASH, 53 showed delayed arrival in mutants of BLOC-2 (represented by hps3 KO ; Figure S4D).This impairment may contribute to the increased size of lysosomes in BLOC-2 mutants (see Figure 4D).
Rab7B can induce recruitment of Dictyostelium BLOC-2 via the HPS3 subunit HPS3 localizes BLOC-2.We explored possible mechanisms underlying this.In mammalian and insect cells, BLOC-2 can bind to, or be affected by, several Rab proteins, including Rab9, 46,78 Rab22A, 45 and Rab32/38. 43Our HPS3-GFP-nanotrap experiments (which were not optimized for Rab binding) also identified some Rab proteins (Rab8A/B and Rab14) as low-level hits (Table S1).We tested all these Rabs for possible interactions with BLOC-2, using in vitro and cellular reporter assays.For Rab9 and Rab22A, we used their closest Dictyostelium homologs (Rab7A/B and Rab5, respectively).
We co-expressed HPS3-GFP with each mCherry-Rab and examined their cellular distribution.Four of the 10 Rabs (Rab7A, 7B, 32B, and 32C) were present on the same compartments as HPS3, whereas the six others were not (Figures 7A and  S5A).Using putatively activated versions of the four mCherry-Rabs, we performed RFP-trap: all four were able to co-immunoprecipitate detectable HPS3-GFP, the highest levels coming from Rab7B and Rab32B (Figure S5B).
To test whether any of the ten Rabs were able to cause HPS3 recruitment in cells, each mCherry-Rab was localized to the outer mitochondrial membrane, using a C-terminal anchor, and co-expressed with HPS3-GFP.Only one Rab had any effect: Rab7B.Mito-Rab7B relocalized HPS3-GFP in essentially all co-expressing cells (Figures 7B and S5C).This ability of mito-Rab7B to relocalize HPS3 matches that of its close human homolog, Rab9, in an equivalent assay, 46 but in those experiments, no other BLOC-2 subunits were recruited.We therefore tested whether, in Dictyostelium, Rab7B's influence was restricted to HPS3 or extended to other BLOC-2 components.
Co-expression of mito-Rab7B with GFP-tagged HPS5, HPS6, or Claret did not cause any of them to relocalize (Figure 7C).However, when HPS3 was additionally expressed, it enabled recruitment of HPS6 and Claret (Figure 7D), though not of HPS5.Finally, successful recruitment of HPS5 was achieved by the combined expression of mito-Rab7B, HPS3, and HPS6 (>80% of cells expressing this combination showed recruitment of HPS5).We did not observe convincing recruitment of HPS5 by mito-Rab7B/HPS3/Claret (Figure 7E), though this was a harder combination to express.
Rab7B thus enabled the recruitment of HPS3, as well as the stepwise recruitment of HPS3/HPS6, HPS3/Claret, and HPS3/ HPS6/HPS5.These observations closely parallel our findings on the subunit-dependence of BLOC-2 localization to its endogenous site of action, the endolysosome.

DISCUSSION
We have identified a tetrameric BLOC-2 in Dictyostelium.Two subunits are worth particular attention: an HPS6 that is unambiguously orthologous to mammalian HPS6, yet sufficiently diverged to not appear in sequence searches 38,49 ; and an additional subunit, Claret. 633][64] claret belongs to the ''granule group'' of Drosophila mutants defective in eye pigment granule formation, and among these mutants, the one phenotypically closest to claret is pink, which encodes the HPS5 ortholog. 79,80Insect BLOC-2 has not been fully characterized, thus it is not known whether Claret is an integral member.Via its RCC1-like domain, Claret is proposed to have guanine nucleotide exchange factor (GEF) activity toward the Drosophila Rab32 ortholog, 63 though this has yet to be demonstrated biochemically.If correct, it implies that Dictyostelium BLOC-2 has intrinsic RabGEF function by including Claret.We have found no evidence for an interaction between Dictyostelium BLOC-2 and Rab32 homologs.Instead, we found a productive interaction between Rab7B and BLOC-2, via HPS3.We cannot be certain that Rab7B recruits BLOC-2 to endolysosomes and, if it does so, whether it acts directly.However, the stepwise recruitment of BLOC-2 in our relocalization assay, via its membrane-binding HPS3 subunit, is very similar to that of the proto-coatomer complex HOPS, by its endogenous recruiter Arl8b in mammals (via its arrives with a punctate appearance (dashed arrowheads) before becoming smooth (solid arrowhead).The relevant vesicle is kept in the middle of each displayed image (other vesicles enter and exit the frame, and focal plane, during the experiment).Scale bars, 2 mm.(D) Quantitation of HPS3-GFP arrival at individual vesicles.t0 = first time point where HPS3-GFP could be detected.The graph shows the average behavior (± SD) of seven vesicles for which we were able to obtain a negative-to-positive transition.The average half-time to maximum labeling was approximately 3-4 min.(E) BLOC-2 at exocytosis: an HPS3-positive vesicle is highlighted (filled arrowhead).About 30 s prior to exocytosis, this vesicle underwent rapid loss of signal (dashed arrowhead).It remained in the cell unlabeled for $20 s and was then exocytosed (open arrowhead).Scale bars, 5 mm.(F) Quantitation of HPS3-GFP signal on exocytic vesicles during their last 3 min.The dashed arrowhead shows the start of the rapid loss of signal, as shown in (E).The graph shows the average intensity ± SD, no. of vesicles tracked = 14 to 33, depending on time point.When the vesicle loses HPS3-GFP shortly before exocytosis, its measured signal drops below the cytosolic background due to the rim of the vesicle becoming void.Once it is exocytosed, the measured region is effectively the cytosol, hence giving a value = 1.See also Videos S1 and S2.(legend continued on next page) membrane-binding VPS41 subunit). 81The enhanced localization of HPS3-GFP to endolysosomes that we observed in the absence of intact BLOC-2 presumably indicates a de-constrained interaction with its recruiter, perhaps due to a steric effect or the absence of a regulatory property associated with the intact complex.
BLOC-2 functions in a central vesicle-trafficking pathway in Dictysotelium Dictyostelium, a unicellular eukaryote, has a complete and functional BLOC-2 that acts on a central trafficking pathway in the cell.BLOC-2 mutants share the delayed lysosome neutralization phenotype seen in WASH complex and AP3 mutants, 51,82,83 suggesting that they act in closely related processes.WASH induces F-actin polymerization on acidic lysosomes, causing V-ATPase segregation into, and removal on, recycling vesicles. 51In Dictyostelium, the WASH complex is recruited in multiple ways.Its FAM21 subunit has two membrane-localizing domains, but neither is strictly required; likewise, retromer, which binds to repeat elements in the FAM21 C-terminal tail, is not essential. 52,77ere, we show that WASH localization also depends on BLOC-2 and occurs during the same brief time window as BLOC-2 arrival.Previous observations in Dictyostelium highlight a vesicular involvement in WASH trafficking: it is present on small lysosome-adjacent vesicles during V-ATPase departure (when WASH itself would be expected to be arriving), 51 and at exocytosis, WASH appears to be removed from PLs on small vesicles. 52n Dictyostelium and mammalian cells are indirect links between BLOC-2 and WASH.We did not find them to co-precipitate from Dictyostelium cells.Both interact with BLOC-1. 29,41,84WASH also interacts on endosomes with the proto-filament of dynactin, 85 and HPS6 has been reported to co-immunoprecipitate with the p150Glued dynactin subunit from HeLa cells (we did not observe this interaction in our mass spectrometry data).Pinpointing the exact role of BLOC-2 in the trafficking pathway is challenging due to its highly dynamic and cyclical nature.The most straightforward interpretation of our results-based on phenotypes and time-course localization-is that BLOC-2 acts early in the lysosome maturation process by helping to deliver or recruit other factors, such as the WASH complex and MROH1.The impaired arrival of MROH1 in BLOC-2 mutants may underlie their increased lysosomal size: Dictyostelium mroh1 mutants share this enlarged lysosome phenotype, 53 and MROH1 has recently been discovered to function as a lysosome fission factor. 86

BLOC-2 and lysosome-related organelles are present throughout eukaryotes
The evolutionary distribution of BLOC-2 indicates that it is ancient-likely present in the last eukaryotic common ancestor-although its phylogenetic patchiness implies frequent loss. 75,76,87,88Dictyostelium BLOC-2 differs from mammalian (lacking Claret) and insect (lacking HPS6?) versions in distinct ways.Does it represent a more ancestral version of the complex?Understanding BLOC-2 composition and function in other diverse organisms will be required to clarify this point.Our findings on Dictyostelium BLOC-2 and its role in endolysosome maturation reinforce the idea that what are commonly called LROs, as well as their biosynthetic machinery, are unlikely to be recent evolutionary innovations.The specialized LROs described in animal cells are likely modified endolysosomal compartments adapted for their unique cargo. 12,14,15,17In Dictyostelium, there is no apparent distinction between a lysosome (an acidic compartment whose neutralization is WASH dependent 51,52,89,90 ) and an LRO (a compartment whose maturation is dependent on BLOC-2 39,40 ).

Implications of BLOC-2 proto-coatomer family membership
A key aspect of our work is the realization that the BLOC-2 subunits belong to the proto-coatomer family, [74][75][76]88 ancient proteins having a conserved b-propeller/a-solenoid arrangement. Thesubunits are also, therefore, structurally homologous to each other, an unexpected conclusion given that they had seemed unrelated to any known proteins.Many, but not all, proto-coatomers function in vesicle trafficking, including as coats (e.g., clathrin) and tethers (e.g., HOPS and CORVET) [74][75][76][91][92][93] ; other types of unrelated tethers exist 94,95 ).HHPRED analysis suggests that HPS3, HPS5, HPS6, and Claret (outside of its RCC1-like domain) are most closely related to subunits of HOPS/CORVET, and the possibility of BLOC-2 acting as a tethering complex has long been speculated.42 The apparent lack of similarity of BLOC-2 subunits to known tethers was problematic for this hypothesis, but our findings now add strong support to the idea.
The new BLOC-2 subunit genes that we have identified in diverse eukaryotes allow, for the first time, identification of evolutionarily conserved motifs, and we can begin to consider these and pathogenic HPS variants in the context of protein structure models (Figure S6).This is especially fruitful for HPS6, as many mutations have been identified: most cluster in the N-terminal b-propeller domain.This includes the original mouse ru  (legend continued on next page) (ruby-eye) mutation (Daa187-189), 38 which lies at the end of a b-strand.Modeling mammalian BLOC-2 using AlphaFold Multimer 96 predicts a highly intertwined HPS5/HPS6 pair atop a more exposed HPS3 subunit, perhaps providing a large binding surface for membrane and protein interactions.The full demonstration of the molecular role of BLOC-2 will require biochemical assays and structural characterization.Dictyostelium provides a new opportunity to further address these challenges.

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Robert Insall (robert.insall@glasgow.ac.uk, r.insall@ucl.ac.uk)

Materials availability
All cell lines and plasmids generated in this study are available upon request.Plasmids will be deposited with Addgene, Dictyostelium strains will be deposited with the Dicty Stock Center.

Data and code availability
d Mass spectrometry data reported in this paper have been submitted to the ProteomeXchange Consortium 97 via the PRIDE partner repository, 98 and are publicly available as of the date of publication.Accession number is listed in the key resources table.Microscopy data, compiled fluorescent exocytosis measurements, protein gel images and western blots reported in this paper will be shared by the lead contact upon request.d All original code has been uploaded to github and is publicly available as of the date of publication.The URL is listed in the key resources table.d Any additional information required to reanalyse the data reported in this paper is available from the lead contact upon request.

Molecular biology
The screen for disrupted exocytosis used a library of REMI mutants in the Dictyostelium Ax3 background, with selection for mutants being by FACS, as before. 53All of the hps3 insertion mutants obtained were in position 2476 of the genomic locus (numbering from the ATG start site).This is a DpnII restriction site in the middle of exon 2 of the coding sequence.New disruptants were made by homologous recombination: 5' and 3' arms of the genes of interest were cloned from genomic DNA using polymerase chain reaction (using Primestar Max) followed by insertion into pCR Blunt II TOPO or pJET2.The two arms were subcloned either side of a suitable resistance cassette to give a knockout construct.This was digested with appropriate restriction enzymes (New England Biolabs) to liberate the region required for homologous recombination, then put through a DNA cleanup column.Ax3 cells were transfected with 10-20mg of DNA by electroporation (E-buffer: 10mM NaK-phosphate pH 6.5 + 50mM sucrose; 500V, single pulse, giving a time-constant of 3ms; BTX Gene Pulser) and cells were selected the following day by transferring to 96-well plates in the presence of antibiotic.Clones were analysed by PCR from genomic DNA and by their phenotypes.The Ax2 strain was used for disruption of hps5 as this locus is duplicated in Ax3.
For expression of native proteins or ones tagged with EGFP, mRFPmars or mCherry, extra-chromosomal vectors were used. 100As these give high levels of expression, we also made use of REMI vectors integrated into the Dictyostelium genome, 101 where limited expression was required, including those shown in Figure 1A (GFP-trap purifications) and Figure 3A (localisation).

Microscopy
Confocal imaging was performed on a Zeiss Axio Observer Z1 LSM880 point-scanning microscope equipped with three separate detectors (two photomultiplier tubes and a 32-channel Quasar spectral detector) and variable band-width collection.A 63x oil NA1.4 Plan-Apochromat infinity-corrected objective (420782-9900-799) was used, typically with a 1AU pinhole set in the red channel.Zen Black 2.3 software and sequential excitation was used for all acquisitions.Airyscan imaging used the same objective, with a >1.2AU pinhole.For rapid red/green imaging a dual band-pass filter (BP 420-480/495-620) was used with sequential excitation, but for critical samples where collection of emitted light from cross-excited fluorophores had to be prevented (e.g.excitation of mCherry by 488nm laser in the mitochondrial re-localisation experiments) green and red images were taken using separate emission filters (BP 420-480/495-550 and BP 570-620/LP645, respectively, changed between captures), which introduced a short delay between acquisitions.For Airyscan imaging, full SR mode and optimal pixel count was used (pixel size 0.05mm, voxel depth usually 0.4mm for the sake of imaging speed).For Airyscan Fast imaging, SR mode was again used, and the field of view often cropped to the cell of interest.As a key factor here was to avoid movement of features between different channel captures, z-stacks were captured using filter switching every frame (where necessary) rather than after every stack.
For the vesicle neutralisation assay a mixture of TRITC dextran (4mg/ml) and FITC dextran (0.4mg/ml) was used and cells were examined using fluorescence confocal microscopy as described above.Dextran pulse-chase assays used a 5-10min labeling period of cells in glass-bottom dishes, followed by washing twice in fresh medium.For acute dextran labeling assays medium was not changed.For assays using both GFP-and mRFP-fusion proteins we added cascade blue dextran (1mg/ml), where required.For imaging dynamic arrival of HPS3-GFP at vesicles we used Texas-Red dextran (1mg/ml) as a pulse-chase marker.
For discontinuous timelapses (for population analysis), z-stacks were captured every 2-3min.For continuous timelapses (for single-cell analysis) Airyscan Fast imaging was used to capture 15-30 focal planes through the whole cell, at an axial interval of 0.4mm.Z-stacks took 1-2sec to capture and were taken at 6-8 sec intervals.For all assays at least three independent experiments were performed.Cells expressing fluorescent fusion proteins from extra-chromosomal plasmids had great variability in their expression levels, but in all cases, acquisition settings were adjusted to avoid saturation (any cells still showing saturation were not analysed).For analysis, the display of images was adjusted to enable features of different strengths to be accurately scored.For presentation of data from 4-dimensional (3D+time) acquisitions in this paper, the z-slice of greatest relevance is shown at any given time point.

GFP-trap, immuno-precipitations, western blot and size exclusion chromatography
All procedures used ice-cold buffers and all steps were carried out at 4 C.For GFP-trap a confluent 10cm dish (2x10 7 cells) was lysed in 5ml TNE buffer plus 0.1% Triton X-100, supplemented with protease inhibitors (HALT (ThermoFisher) and cOmplete inhibitor (Roche)) and 2mM DTT.Lysates were centrifuged at 16,000g for 10min and the supernatant retained.GFP-trap pulldowns were performed for 2h, beads washed three times in lysis buffer.Washed beads were processed for mass spectrometry, or alternatively were used for SDS-PAGE, typically using NuPAGE 10-well Bis-Tris mini-gels with a 4-12% acrylamide gradient, run in MOPS buffer (ThermoFisher).Gels were stained with InstantBlue or Silver Stain, as appropriate, or transferred to Protran nitrocellulose (30V for 16h to ensure completion) for western blot.Membranes were blocked in 10% fat-free milk in TBS.Primary antibodies (1/1000) were incubated in 10% milk/TBST overnight.Washed membranes were incubated with fluorescent secondary antibodies (1/20,000) in 10% milk/TBST for 2h.Washed blots were scanned on a LiCor Odyssey CLx and images analysed in ImageStudio.Where cell lysates needed to be analysed directly (without IP), these were made by adding an equal volume of 2x LDS sample buffer (ThermoFisher) at 70 C to a cell suspension containing 20mM DTT.This ensured immediate lysis and denaturation, such that no proteolysis of bands occurred.
For co-IPs of the BLOC-2 subunits (as EGFP and mRFPmars pairs) GFP IPs were done as above, followed by blotting for RFP.For analysis, signal for the RFP-fusion protein in the pulldown was divided by its signal in the lysate, and the ratio was compared to the same co-IP done from WT cells.
For size exclusion chromatography cell lysates were made from 2 x 15cm dishes (approx.10 8 cells) in 20ml TNE + 0.1% Triton X-100 + 2mM DTT, plus protease inhibitors.Alternatively, detergent-free lysis (without Triton) used two 5mm filters stacked together.Cell suspension was passed through twice using a 20ml syringe; lysis was checked by microscopy.Lysate was centrifuged at 16,000g for 10min, then the supernatant was ultracentrifuged at 120,000g for 90min (Beckman TLA100.3rotor) to yield cell cytosol (supernatant).This was concentrated down to 1ml using Amicon Ultra 10k mwco spin filters (centrifuged at 2500g) and then passed through a 0.2mm filter.0.5ml was loaded onto a Superose 6 Increase 10/300 column run at 0.4ml/min.0.5ml fractions were collected and GFP-trap was performed on each, followed by SDS-PAGE and silver stain or western blot for GFP.

Mass spectrometry
Samples from GFP-trap pulldowns either were run on SDS-PAGE followed by analysis of gel slices, or used for on-beads digestion.Further sample processing, mass spectrometric data acquisition, and analysis were performed as described in Fort et al. 102 .The raw files and the MaxQuant search results files have been deposited as ''complete submission'' to the ProteomeXchange Consortium 97 via the PRIDE partner repository 98 with the dataset identifier PXD052707.

Trafficking assays
Quantitative fluorescent dextran exocytosis assays were performed as described previously, 53,69 with at least four independent assays performed for each strain.Cells were incubated in HL5 medium containing TRITC-dextran (70kDa, 4mg/ml final concentration) overnight, washed twice in HL5, resuspended in 13ml fresh HL5 then shaken in 100ml conical flasks at 120rpm at room temperature.1ml samples were taken at intervals, cells pelleted in a microfuge (pulsed for 8sec to reach top speed), supernatant removed, cells washed once in KK2 buffer pH 6.5, then lysed in 0.2ml TNE (50mM Tris pH 7.5, 150mM NaCl, 0.5mM EDTA) + 0.1% Triton X-100.Lysates were read in a fluorimeter with excitation 544nm and emission 574nm.

Image analysis
Most image processing was done in Fiji 99 ; Airyscan processing was done in Zen Black 2.3, using default sharpening settings.
Measurement of fluorescent proteins on vesicles was done according to the experimental question being asked.To study the arrival of proteins onto vesicles at the population level, a vesicle was scored as positive if it contained any of the protein on its surface, no matter how much.For the assessment of GFP-WASH behaviour during dextran pulse-chase time courses in WT vs BLOC-2 mutants (Figures 6E and 6F), vesicles were scored as positive when a characteristic patch (not just a punctum) of GFP-WASH was present.This was in order to judge whether the endolysosomes were maturing or delayed/arrested.For assessment of MROH1-GFP labeling of vesicles (Figure S4D), three categories of labeling were defined: punctum, arc/small patch, and large patch/coat (occupying an estimated >50% of vesicle surface in confocal section).
For assessment of vesicle neutralisation in WT vs BLOC-2 mutants (Figure 1D), fluorescent signal in FITC and TRITC channels was calculated for each vesicle: neutral vesicles were defined as those with FITC/TRITC ratio S0.5.For statistical analysis of exocytosis data shown in Figures 1B and 1D, and of GFP-WASH data in Figure 6F, original data were plotted for each individual experiment in GraphPad Prism, and the Area Under Curve function used to generate a value for each time course.These were then tabulated and compared across strains.
For HPS3-GFP analysis in WT vs BLOC-2 mutants (Figure 4C) >2000 vesicles were measured, over 5 experiments, for each strain (done in parallel).For quantifying protein levels at individual vesicles two methods were used: manual and automated.Manual analysis used a band drawn to encompass the surface signal on the vesicle (band width 0.3mm).The signal of this was measured and divided by the cytosolic signal from the same focal plane: this gave a normalised intensity measure for each feature.From these analyses vesicle diameter was also derived.The automated method was used to do more of these analyses, and used a script written in Python, as detailed below.
For statistical analysis, to compare multiple datasets we used one-way ANOVA with Holm-Sı ´da ´k multiple comparison test.For assays in which paired observations were made (HPS3-GFP intensity in Figure 4C, vesicle size in Figure 4D, pulldowns with MBP-Rab in Figure S5B) paired analyses were used.In all analyses, n = no. of repeat experiments (always at least 3, often more): exact n is reported in each Figure legend.For some experiments, the number of features analysed (e.g.vesicles, in Figure 4) is noted for context, but analyses were always carried out per experiment, not per feature.
Ring segmentation and analysis -automated method For automated analysis of HPS3-GFP decorated endolysosomes, a code was written using the Python programming language within a jupyter notebook. 106A pipeline was developed to perform batch analysis of all the images within an experiment and output the results as a.csv file.For each experiment, the image data from a single data file is loaded, and a bespoke deep-learning model applied to the image to segment out the HPS3-GFP vesicles.The segmentation model was developed by retraining the ''cyto'' model, that is available in cellpose as part of their model zoo. 107Retraining the model was performed using code that is supplied by the cellpose development team on their GitHub (https://github.com/MouseLand/cellpose/tree/main). 108The model was trained using 26 randomly chosen frames (some whole frames, some cropped regions) that had been manually segmented using QuPath. 109The parameters used to retrain the segmentation model were: 100 epochs, a learning rate of 0.1, and a weight decay of 0.001.A bespoke model to segment the cells in the image was also developed using the same training parameters, and 39 training images.
The application of the ''ring segmentation model'' to the images sometimes resulted in an over-estimation of the number of rings in the image, due to the mischaracterisation of voids in the cell as HPS3-GFP decorated endolysosomes.To remove these false positives from the analysis sequence, all masks generated by the model were cropped from the image and analysed individually.To filter the rings, in 8 positions from the edge of the cropped image to the centre, a Gaussian curve was fitted to the data using the curvefitting tool available in SciPy. 110In the case of a mask containing a HPS3-GFP vesicle, the Gaussian curve could be fitted as the GFP decoration produced a ring with a higher fluorescence intensity than the background in the image.The R2 metric was used to assess the quality of the fitting to the imaging data.If the Gaussian curve fit failed, or if the fitting did not meet the R2 threshold of 0.8, then the object related to the mask was excluded from the analysis.The R2 value used for filtering the rings was chosen as 0.8 to maximally exclude False Positives.This value was determined by comparing the results of the algorithm to a manual selection of the HPS3-GFPpositive endolysosomes with the maximum diameter.
The program also rejects double counting of vesicles that appear across multiple z planes.For this, the central point of all segmentation masks generated is calculated for each of the frames in the image.The central positions for each mask are compared: those that have a centroid within 30 pixels in x and y from one another are filtered so that the ring with the smallest mean-full-width-halfmaxima (FWHM) is kept for analysis.If we consider the endolysosome as an approximately spherical object in 3D, when imaging at half the vesicle depth, the microscope captures the least amount of curvature for that z-slice in the image.As a result, when analysing the images, we see that at the mid-point of the vesicle, where the diameter is largest, the FWHM of the fluorescent ring is at its smallest.Therefore, by measuring the FWHM of the rings at every position in the Z-stack, we can determine the frame in the 3D stack image of the cell in which the vesicle has the largest diameter.
Once all the HPS3-GFP endolysosomes in the image had been filtered to remove double counting and false positives, the analysis of the shape and fluorescent properties of the rings could be calculated.These include: ring diameter, median fluorescence intensity, FWHM and median intensity of the cell (cytosol) that the endolysosome belongs to.The jupyter notebook used for this analysis can be found at the CRUK Scotland Institute github: https://github.com/Beatson-CompBio/Thomason_2024_A.git.
was slightly less penetrant.Scale bars, 10 mm.(D) Quantitation of vesicle neutralization.Proportion of neutral vesicles was plotted over time: graph shows mean and SD; n = 3 for all strains.Area under curve calculated, one-way ANOVA with Tukey's correction for multiple comparisons.All mutants were significantly different from their parents.See also Figure S1.

Figure 2 .
Figure 2. Phylogenetic distribution and predicted structural features of BLOC-2 proteins (A) BLOC-2 gene distribution in eukaryotic supergroups.The three canonical subunits, plus Claret, are distributed throughout the eukaryotic tree, suggesting that they were present in the last eukaryotic common ancestor but have been frequently lost.Inset: within opisthokonts, BLOC-2-encoding genes are absent from (legend continued on next page) redrawn under CC BY 4.0 license.(B) AlphaFold 71 models for Dictyostelium BLOC-2 subunits.Ribbon depictions are colored from dark blue (N terminus) to red (C terminus).Schematic secondary structure elements are shown below: blue for b-propellers, yellow for a-helix.The Dictyostelium HPS5 RING domain is shown in pink (human HPS5 does not have a predicted RING domain).Claret is predicted to contain consecutive repeats of the b-propeller/⍺-solenoid arrangement.See also Figure S6.

Figure 3 .
Figure 3. Dictyostelium BLOC-2 localizes to endolysosomes (A) GFP-fusions localized to cytosolic vesicles, with a continuous surface appearance.Fluorescent dextran (magenta) showed these to be derived from endosomes.Scale bars, 5 mm.(B) Co-localization of the BLOC-2 subunits, expressed pairwise as EGFP (green) and mRFPmars (magenta) fusions.All proteins localized to the same vesicles and had the same appearance.Scale bars, 5 mm.See also Figure S2.

Figure 4 .
Figure 4. Interdependence of BLOC-2 subunit localization and association (A) In hps3KO  , all other subunits completely de-localized.Conversely, HPS3-GFP strongly localized in all other mutants.In hps5 KO , all other subunits still localized.hps6 KO and claret KO behaved similarly to each other: HPS3-GFP localized, whereas HPS5-GFP did not.The mutants were subtly different: HPS6-GFP did not localize in claret KO , but Claret-GFP still localized (to a small extent) in hps6KO  .Scale bars, 5 mm.(B) Localization of HPS3-mRFPmars (magenta) and absence of localization of HPS5-GFP (green) when co-expressed in hps6 KO and claret KO cells.Scale bars, 5 mm.(C) Quantitation of localization of HPS3-GFP in WT vs. mutants.Signal around vesicle was measured at the focal plane of greatest intensity and expressed relative to the cytosolic signal.HPS3-GFP localization increased by 76% in hps5 KO , 56% in hps6 KO , and 39% in claret KO vs. WT cells (and vs. hps3 KO -rescue).5 independent experiments were done (shown by different colored dots); statistical tests done on experimental means: hps5 KO vs. Ax2 (two-tailed paired t test); hps6 KO and claret KO vs. Ax3 (paired measures ANOVA with Geisser-Greenhouse correction, and Dunnett's correction for multiple comparisons).(D) Diameter of HPS3-GFP-positive vesicles in WT vs. BLOC-2 mutant strains, using data from the 5 experiments shown in (C).Vesicles were measured at their cross-section of maximum diameter from z stacks.Paired analyses were carried out using the statistical tests as in (C).Experiments indicated by green and orange symbols used 10% dextran, the others used 2% dextran.This changed vesicle size in all strains but did not affect the differences between strains.For (C) and (D): total number of vesicles analyzed: Ax2 1,537, hps5KO  3,309, Ax3 2,255, hps3 KO -rescue 2,482, hps6 KO 2,959, and claret KO 3,679.(E) CoIP of GFP and mRFP pairs expressed in WT and mutants.IPs performed at least three times (each symbol is one experiment) and expressed relative to the coIP from WT cells, done in parallel (control value C = 1, by definition).All data points are shown, as well as means and SD.One-way ANOVA with Dunnett's correction for multiple comparisons test performed.Adjusted p values are shown on the figure only for those coIPs that were substantially altered compared with controls.See also Figure S3.
3,309, Ax3 2,255, hps3 KO -rescue 2,482, hps6KO  2,959, and claret KO 3,679.(E) CoIP of GFP and mRFP pairs expressed in WT and mutants.IPs performed at least three times (each symbol is one experiment) and expressed relative to the coIP from WT cells, done in parallel (control value C = 1, by definition).All data points are shown, as well as means and SD.One-way ANOVA with Dunnett's correction for multiple comparisons test performed.Adjusted p values are shown on the figure only for those coIPs that were substantially altered compared with controls.See also Figure S3.

Figure 5 .
Figure 5. BLOC-2 is recruited to maturing endolysosomes and is present on them until exocytosis (A) WT cells expressing HPS3-GFP (green) pulse-labeled with Texas red dextran (magenta) for 10 min.Confocal z stacks captured every 2-3 min.Several time points (single planes from full z stacks) are shown as examples.Scale bars, 5 mm.(B) The proportion of dextran+ HPS3+ vesicles was calculated for each time point.The experiment was performed three times; all data points are shown (different symbols for each experiment).(C) Cells pulsed for 10 min with Texas red dextran (magenta), washed, and imaged using Airyscan Fast z stacks every 6-8 s, for as long as cells would tolerate (a few minutes).A vesicle becoming decorated with HPS3-GFP (green) over $5 min is shown.It starts with little or no HPS3-GFP (open arrowhead), which first (legend continued on next page)

(
B) Early stages of arrival of HPS3-mRFPmars (magenta) and GFP-WASH (green) onto dextran vesicles (blue) in pulse-chase experiments.At t = 20-25 min, spots of both proteins began to appear on the same vesicles (five vesicles [i-v] are enlarged to show detail).Their distribution on the vesicle was distinct, but their arrival synchronous in time at this level of resolution.Three experiments are shown as examples.(C) Acute localization of GFP-WASH, but not HPS3-GFP, to recently ingested macropinosomes.Filled arrowheads, WASH patches; open arrowheads, WASH puncta.(D) Upper, localization of HPS3-mRFPmars (magenta) and GFP-WASH (green) in hps5 KO .GFP-WASH localized normally to early endosomes (open arrowhead) but had reduced localization to endolysosomes (HPS3-positive; filled arrowheads).Lower: localization of HPS3-mRFP (magenta) and HPS5-GFP (green) to endolysosomes was not disturbed in wash KO .(E) Dextran pulse-chase assay.WT cells had multiple large vesicular patches of GFP-WASH from $30 min, which remained thereafter.GFP-WASH accumulated slowly and sparsely on endolysosomes in BLOC-2 mutants.Indicated vesicles are shown enlarged to the right.(F) Quantitation of experiments from E (n = 3 for Ax2, n = 4 for others): mean number of dextran-positive vesicles with patches of GFP-WASH, plotted over time; error bars show 95% CI.The area under the curve is shown for each experiment and strain.The hps3 KO , hps6 KO , and claret KO mutants were significantly different from their Ax3 parent (p < 0.0001 in all cases; one-way ANOVA with Dunnett's multiple comparison test), and hps5 KO was significantly different from its Ax2 parent (p = 0.0028; two-tailed t test).Scale bars: 5 mm in all images.See also Figure S4 and Videos S3 and S4.