For understanding the elusive physiological role of APP we studied the interactome of APP's intracellular domain (termed AICD) using an in-vitro reconstitution system that we have recently established [27]. In this system AICD was coupled to the surface of liposomes to mimic the organisation of its intracellular domain in the native APP together with a membrane context [28]. These so-called 'proteo-liposomes' were then used to recruit interaction partners followed by isolation and identification by mass spectrometry [27–29].

Combining AICD proteo-liposomes with label-free mass spectrometry [30] we established the interactome of APP's intracellular domain and, surprisingly, found all three subunits of the mammalian PIKfyve complex, namely PIKfyve (NCBI: NM_015040), Vac14 (NCBI: NM_018052) and Fig 4 (NCBI: NM_014845) to associate with APP's intracellular domain (Fig 1). Given that inactivation of the PIKfyve complex leads to neurodegeneration and that APP plays a key role in Alzheimer's disease this finding prompted us to investigate this interaction in more detail.

We confirmed the interaction between the intracellular domain of APP and Vac14 and the PIKfyve complex by analysing the protein recruitments using Western blotting, showing that Vac14 of the PIKfyve complex binds specifically to AICD-presenting proteo-liposomes while no significant amounts of Vac14 were found in any of three different, negative controls (Fig 1).

When analysing the steady state localisation of APP and Vac14 in HeLa cells by co-expressing APP-CFP and Vac14-mCIT we found them to display strong colocalisation on vesicles (Additional line scans are presented in S1 Fig). Some of these were positive for EEA1 indicating early endosomes while some were LampI positive late endosomes (Fig 1). The dynamic association of APP with Vac14 was also observed in live-cell imaging where APP co-localises and co-migrates with Vac14 on tubules and vesicles moving through the cytoplasm (Fig 1). Additionally, we detected significant colocalisation on mobile vesicles when expressing only the intracellular domain of APP, AICD, fused to mCherry together with Vac14 (Fig 1). This is interesting given that AICD, lacking a transmembrane domain, is expected to be cytosolic. However, these data show that a portion of AICD is clearly membrane associated and colocalises with Vac14. These data are consistent with the APP/Vac14 interaction established in Fig 1 and suggest that APP associates with the PIKfyve complex in cells via its intracellular domain.

What is the function of the protein-protein interaction between APP and the PIKfyve complex in-vivo? Currently it is not clear how PIKfyve function is regulated in multicellular organisms. The fact that APP binds the PIKfyve activator Vac14 raised the possibility that APP may control PIKfyve activity. For testing this hypothesis we used the genetic model organism Caenorhabditis elegans. The significant advantage of C. elegans is that only one orthologue of APP exists (named apl-1, NCBI: NM_076469). In contrast, in mammals the APP-like gene family consists of APP, APLP-1 and APLP-2. Mouse genetics clearly demonstrated a high level of functional redundancy between these three genes, considerably complicating the analysis of APP gene function in mammals [31].

Both APL-1 and PIKfyve (PPK-3 in C. elegans, NCBI: NM_077754) are essential for worm development and viability [32, 33]. The most prominent phenotype of apl-1 null alleles is developmental arrest during larval stages caused by a loose cuticle. This lethal phenotype is rescued by the expression of the soluble, extracellular domain of APL-1 [32]. Whether APL-1 has additional functions that are masked by the severity of the cuticle phenotype remains unclear.

The PIKfyve orthologue PPK-3 is also essential and required for embryogenesis [33]. Additionally, PPK-3 plays an important role in endosomal homeostasis. Partial loss-of-function alleles (hypomorphs) lead to a drop in PI(3,5)P₂ levels and cause the formation of large vacuoles in hypodermal and intestinal cells, mirroring closely the defects induced by loss of mammalian PIKfyve [33]. The vacl-14 gene (NCBI: NM_059814), the C. elegans orthologue of mammalian Vac14, has not been characterised in C. elegans to date.

First, we wanted to test whether in C. elegans the putative Vac14 orthologue VACL-14 plays a similar role as Vac14 in the mammalian complex and is indeed required for the activation of PPK-3. We tested this by creating the ppk-3(n2835); vacl-14(ok1877) double mutant which showed synthetic lethality, confirming that in C. elegans VACL-14 is necessary for full PPK-3 function and suggesting that the C. elegans complex functions similarly as the human complex.

To test whether APL-1 and the PPK-3 complex share a common function we analysed the phenotypes of apl-1, ppk-3 and vacl-14 mutant alleles. For APL-1 we chose to study the yn5 deletion in which the 3'-end of the gene encoding the transmembrane domain and intracellular domain are deleted [32]. The yn5 allele thus represents a powerful tool for characterising the role of the cytoplasmic domain of APL-1 without additional, deleterious effects caused by lack of the extracellular domain [32]. This approach allowed us to characterise phenotypes that are masked by the lethality of apl-1 null mutants. According to our working hypothesis APL-1 binds the PPK-3 complex via its intracellular domain and controls its activity. Loss of the intracellular domain of APL-1 will prevent it from binding to the PPK-3 complex and disrupt PPK-3 regulation. By consequence, APL-1 mutants should have a similar phenotype as PPK-3 loss-of-function mutations.

Reduced PPK-3 activity by partial loss of function mutations in ppk-3 or deletion of its activator VACL-14 led to the formation of large vacuoles that are most easily visible in hypodermal cells ([33] and Fig 2). Interestingly, when analysing the apl-1(yn5) mutant we observed hypodermal vacuolation similar to that of ppk-3 and vacl-14 mutants. This showed that the apl-1(yn5) mutant phenocopies the partial loss of PPK-3 function (Fig 2 and [32]).

While this observation suggests that APL-1 and PPK-3 may share a common function it is not clear whether they act in the same pathway. To test this we analysed the epistasis of the apl-1 and ppk-3 genes. For this we combined the apl-1(yn5) mutation with hypomorphic alleles of ppk-3. For reducing PPK-3 activity two hypomorphic ppk-3 alleles were used. ppk-3(n2668), introducing a point mutation in the catalytic loop, significantly reduced PPK-3 activity [33]. A weaker allele, ppk-3(n2835), leads to the formation of a truncated PPK-3 lacking the C-terminal eleven amino acids [33]. Additionally, we utilised the vacl-14(ok1877) deletion mutant. Using these mutants we created double mutants in the apl-1, ppk-3 and vacl-14 genes and compared their phenotypes. In all three double mutants [apl-1(yn5); vacl-14(ok1877), apl-1(yn5) ppk-3(n2835) and apl-1(yn5) ppk-3(n2668)] the accumulation of strongly swollen vacuoles was evident (Fig 2 and S2 Fig). These proved to be significantly more numerous compared to single mutants or control animals (Fig 2), showing that truncation of APL-1 considerably enhanced the defects caused by reduced PPK-3 activity. This was particularly evident in the apl-1(yn5) ppk-3(n2668) double mutant containing the stronger ppk-3 mutant allele. These animals were filled with vacuoles along the entire length of the animal in both the hypoderm and intestine (Fig 2A and S2 Fig). The strength of the phenotypic effect of this double mutant precluded us from counting individual vacuoles, as large portions of the animals were filled with vacuoles. Moreover, these animals were unable to produce any viable offspring and the strain was non-viable. Thus the truncation mutation in apl-1(yn5) combined with the ppk-3(n2668) point mutation are synthetically lethal, demonstrating that APL-1 is required for PPK-3 function for suppressing vacuoles and maintaining viability.

For independent confirmation of the results obtained using the apl-1(yn5) mutant we used RNAi suppression of APL-1 and showed that knock-down of APL-1 led to the accumulation of huge vacuoles in the hypoderm and intestine of ppk-3 mutant animals compared to wildtype controls (S3 Fig), mirroring the effect of truncating APL-1. These data show that it is indeed the loss of APL-1 that enhances loss of PPK-3 function.

If loss of APL-1 exacerbates the effects of PPK-3 impairment it is reasonable to assume that overexpression of APL-1 may rescue the PPK-3 deficiency phenotype. We tested this by expressing APL-1::GFP (characterised in [32]) in both hypomorphic ppk-3 mutant strains and measured the number of vacuoles. In Fig 3 it is shown that overexpression of APL-1::GFP significantly reduced the number of vacuoles in both ppk-3 mutants. These data are compatible with APL-1 functioning within the PPK-3 pathway.

While APL-1 and PPK-3 appear to act in the same pathway it was not clear whether APL-1 acts upstream or downstream of PPK-3. Furthermore we have not fully excluded the possibility that APL-1 and PPK-3 share a common function but act in parallel pathways.

Overexpression of APL-1 ameliorated the partial loss of PPK-3 function induced by hypomorphic ppk-3 alleles. If APL-1 acted downstream of PPK-3, overexpression of APL-1 should also ameliorate the phenotype caused by complete loss of PPK-3 function in a ppk-3 null mutant. Similarly, if APL-1 functions in a pathway parallel to PIKfyve, overexpression of APL-1 should also rescue the phenotype of a ppk-3 null allele. In contrast, overexpression of APL-1 should not result in a rescue if APL-1 functions upstream of PPK-3. We tested the hypothesis that APL-1 functions upstream of PPK-3 by overexpressing APL-1::GFP in the homozygous ppk-3(mc46) null mutant [33]. ppk-3 null animals from heterozygous hermaphrodite mothers grow up to the adult stage but fail to produce any viable offspring. Overexpression of APL-1::GFP in hermaphrodites did not rescue lethality of their offspring. Additionally, we studied vacuolation in the anterior tip of young apl-1::GFP; ppk-3(mc46) adults by measuring the area of the anterior tip covered by vacuoles. This was necessary as the vacuolation phenotype in these animals was so pronounced that it was rarely possible to discriminate individual vacuoles because of their fused appearance (Fig 4). An average 17% of the anterior tip was covered by vacuoles in both the ppk-3(mc46) and the apl-1::GFP; ppk-3(mc46) animals, showing that APL-1 overexpression did not rescue the ppk-3 null phenotype.

Taken together, APL-1::GFP expression is able to rescue the vacuolation phenotype of hypomorphic ppk-3 alleles while it appears to be unable to rescue a ppk-3 null allele, suggesting that APL-1 acts upstream of PPK-3. One caveat of this analysis is that vacuolation may not be a very sensitive measure of PPK-3 function which may not allow to detect subtle changes of PPK-3 activity. Thus we cannot fully rule out the possibility that APL-1 may act in a pathway parallel to PPK-3.

However, given that APP interacts with the PIKfyve complex in mammals (Fig 1) the most plausible explanation is that of APP/APL-1 binding to and activating PIKfyve/PPK-3.

The PIKfyve complex is known to regulate several trafficking steps during the maturation of endosomes [15, 24, 34]. As the vacuoles induced by loss of PPK-3 function were shown to be of endosomal/lysosomal origin [15, 18] our data suggest that APL-1 participates in the regulation of endosomal homeostasis through activation of the PPK-3 complex. We therefore characterised the endosomal and lysosomal markers RAB-5, RAB-7 and LMP-1. The early endosomal marker RAB-5 appeared to be largely unaffected by single and double mutations in apl-1, ppk-3 or vacl-14 (S4 Fig) in agreement with previous data, demonstrating that PIKfyve affects mainly late endosomal compartments [15, 25]. In contrast, the late endosomal and lysosomal markers RAB-7 and LMP-1 were profoundly altered in the apl-1(yn5) ppk-3(n2835) and the apl-1(yn); vacl-14(ok1877) double mutants and showed late endosomes to be clustered together (Fig 5). These data were corroborated by RNAi suppression of APL-1 in vacl-14 and ppk-3 mutants expressing GFP::RAB-7 in which loss of APL-1 by RNAi suppression enhanced late endosomal pathology caused by loss of PPK-3 function (S5 Fig). These defects were observed both in the hypoderm as well as in the intestine. When studying the expression of a promoter fusion of APL-1 with GFP Wiese et al. could not to detect APL-1 expression in the intestine using a 5'-UTR fusion presumed to contain the complete APL-1 promoter [35]. We reassessed the expression pattern of APL-1 using a fosmid from the TransgeneOme project in which a genomic copy of APL-1 was tagged with GFP [36]. We created a transgenic line (OL0186) by microinjection of the APL-1 fosmid followed by analysis of APL-1::GFP expression. Overall the expression pattern was similar to those previously described [32, 35] with APL-1 expression detectable in head muscle cells, head neurons as well as the nerve chords. However, an important difference was that APL-1::GFP expression was also detected in the intestine (S6 Fig). These data indicate that the previously utilised 5'-UTR fusions of APL-1 drive APL-1 expression in most, but not all the tissues compared to a genomic construct. These data suggest that APL-1 is expressed in the C. elegans intestine in addition to the previously established tissues. This is fully compatible with the intestinal phenotype observed in the double mutants.

In summary these data show that APL-1 regulates late endosomal morphology in concert with the PPK-3 complex.

As defective PIKfyve function has been shown to cause profound neurodegeneration in mice and in humans [21, 25] and as APP is crucially implicated in Alzheimer's disease we wanted to test whether mutations in apl-1, ppk-3 or vacl-14 can lead to impaired neuronal function in C. elegans. We characterised the overall morphology of the neuronal system of C. elegans using the synaptic marker GFP::RAB-3 by fluorescence microscopy (Fig 6). In single apl-1(yn5) and ppk-3(n2835) mutants the neuronal system appeared to be largely unaffected. However, the apl-1(yn5) ppk-3(2835) double mutant failed to accumulate GFP::RAB-3 label in the characteristic “pearls-on-a-string” fashion to the same extent compared to control animals, leading to more homogeneous RAB-3 staining along the ventral nerve chord (Fig 6). Lack of concentrating RAB-3 vesicles in synapses along the ventral nerve chord suggests that synaptic transmission may be impaired.

We wanted to also assess whether the interplay of APL-1 and PPK-3 is functionally important. For this we tested the ability of single and double mutants to coordinate motor function using a thrashing assay. C. elegans reacts with rapid and sustained thrashing when placed in liquid, a simple assay for testing synaptic transmission of the main motor neurotransmitters, glutamate and gamma amino butyric acid (GABA). While the single mutants apl-1(yn5), vacl-14(ok1877) and ppk-3(n2835) displayed no significant impairment of motor control, the apl-1(yn5) ppk-3(n2835) and apl-1(yn5); vacl-14(ok1877) double mutants were both significantly impaired in their ability to thrash (Fig 6). Together with the RAB-3 morphological data this demonstrates that the APL-1/PPK-3 interplay is necessary for supporting neuronal function.

The antibody directed against Vac14 was kindly provided by Prof. L. Weisman (University of Michigan). Antibodies against EEA1 were purchased from BD Biosciences and Lampl from Santa Cruz Biotechnology.

Protein recruitments and mass spectrometry were carried out as established in [27] with the following modifications: 0.5ml of mouse brain cytosol (final concentration 3mg/ml) were used supplemented with 0.21mM GTPγS and 28μM latrunculin B. The xcalibur.raw files were processed with MaxQuant and the Perseus software of this package was used for analysis as described in [30]. Briefly, LFQ intensities were logarithmised and only proteins that were identified in all replicates of at least one of the samples were retained. Missing values were imputed by a normal distribution around the detection limit and a modified t-test (SAM) was performed with a threshold value of 0.05 and a slope value of 1. A detailed description of the method is presented in [43].

Worm maintenance, genetic crosses and other C. elegans methods were performed according to standard protocols [44]. All strains were grown at 20°C. Mutations used were LGI, vacl-14(ok1877); LGX, apl-1(yn5) lon-2(e678), lon-2(e678), ppk-3(n2835), ppk-3(n2668) [33]. vacl-14(ok1877) was back-crossed three times to WT strain N2 prior to further analysis. The ok1877 deletion was followed by PCR. As negative controls either Bristol N2 wildtype or lon-2(e678) were used as the apl-1(yn5) mutation is in close proximity to the lon-2 locus which served as a marker to follow the yn5 deletion throughout crosses.

Strains carrying integrated arrays used in this study were: pwIs50 (plmp-1::LMP-1::GFP), pwIs439 (prab-5::GFP::RAB-5), pwIs399 [prab-7::GFP::RAB-7) were kindly provided by Prof. B.Grant, jsIs682 [prab-3::GFP::RAB-3) by CGC. For RNA interference (RNAi) experiments, apl-1 cDNA was prepared from EST clone provided by Prof. Kohara (National Institute of Genetics, Japan) and subcloned into RNAi feeding vector L4440 [45]. RNA interference (RNAi) was performed by the feeding method [45]. For RNAi knockdown of apl-1, L1 worms were placed onto RNAi plates, and P0 adults were scored for phenotype after 72 h.

The transgenic APL-1::GFP strain was produced as previously described [46]. Briefly, N2 animals were bleached and synchronised. Young adults were picked with an eyelash and placed on an injection pad (2% agarose in M9 buffer). Worms were coinjected with 50 ng/μl apl-1 fosmid (Construct: 10208337314574834 H09, obtained from TransgeneOme) and 120 ng/μl pRF4 roller plasmid as coinjection marker. 20–30 worms were injected and analysed for GFP positive offspring. GFP positive offspring was isolated to create line OL0186.

was detected using ImageJ by manual circling each individual vacuole present in a DIC image of the anterior tip of the worm and measuring number and size of them using the “Analyse Particle” command.

Using ImageJ the intensity profile of GFP::RAB-3 was established along the ventral nerve chord using manually created line plots, maximas and minimas detected and the average intensity difference between maxima and minima were calculated.

L4 larval stage animals were picked onto fresh OP50 seeded plates and scored for thrashing as young adults 24 hours later. Individual animals were transferred into 30μl drop of M9 buffer into a depression well slide and thrashing was scored for 1 minutes after worms were allowed to settle for 1 minute. A single thrash was defined as a complete change in the direction of the body down the midline.

Fluorescence and DIC images were obtained using a fully motorized Zeiss Axiovert 200M fluorescence microscope (Carl Zeiss) and a Hamamatsu Orca camera controlled by Volocity software (Improvision). Confocal images were obtained using a Leica SP5 upright confocal microscopy setup. To observe live worms expressing transgenes, worms were mounted on 2% agarose pads containing 0.1M tetramisole in M9 buffer.

was performed using the 'Manual Tracking' plugin in ImageJ.

Quantification was performed using ImageJ. Briefly, a line was manually drawn along the ventral nerve chords and the pixel intensity were measured in a line plot. Intensity maxima and minima were defined and the average difference between the two was measured.