1. Biochemistry and Chemical Biology
  2. Neuroscience
Download icon

Somatostatin binds to the human amyloid β peptide and favors the formation of distinct oligomers

  1. Hansen Wang
  2. Lisa D Muiznieks
  3. Punam Ghosh
  4. Declan Williams
  5. Michael Solarski
  6. Andrew Fang
  7. Alejandro Ruiz-Riquelme
  8. Régis Pomès
  9. Joel C Watts
  10. Avi Chakrabartty
  11. Holger Wille
  12. Simon Sharpe
  13. Gerold Schmitt-Ulms  Is a corresponding author
  1. University of Toronto, Canada
  2. The Hospital for Sick Children, Canada
  3. University of Alberta, Canada
Research Article
  • Cited 5
  • Views 1,741
  • Annotations
Cite this article as: eLife 2017;6:e28401 doi: 10.7554/eLife.28401

Abstract

The amyloid β peptide (Aβ) is a key player in the etiology of Alzheimer disease (AD), yet a systematic investigation of its molecular interactions has not been reported. Here we identified by quantitative mass spectrometry proteins in human brain extract that bind to oligomeric Aβ1-42 (oAβ1-42) and/or monomeric Aβ1-42 (mAβ1-42) baits. Remarkably, the cyclic neuroendocrine peptide somatostatin-14 (SST14) was observed to be the most selectively enriched oAβ1-42 binder. The binding interface comprises a central tryptophan within SST14 and the N-terminus of Aβ1-42. The presence of SST14 inhibited Aβ aggregation and masked the ability of several antibodies to detect Aβ. Notably, Aβ1-42, but not Aβ1-40, formed in the presence of SST14 oligomeric assemblies of 50 to 60 kDa that were visualized by gel electrophoresis, nanoparticle tracking analysis and electron microscopy. These findings may be relevant for Aβ-directed diagnostics and may signify a role of SST14 in the etiology of AD.

https://doi.org/10.7554/eLife.28401.001

eLife digest

Treating Alzheimer’s disease and related dementias is one of the major challenges currently facing healthcare providers worldwide. A hallmark of the disease is the formation of large deposits of a specific molecule, known as amyloid beta (Aβ), in the brain. However, more and more research suggests that smaller and particularly toxic amyloid beta clumps – often referred to as oligomeric Aβ – appear as an early sign of Alzheimer’s disease.

To understand how the formation of these smaller amyloid beta clumps triggers other aspects of the disease, it is important to identify molecules in the human brain that oligomeric Aβ binds to. To this end, Wang et al. attached amyloid beta or oligomeric Aβ molecules to microscopically small beads. The beads were then exposed to human brain extracts in a test tube, which allowed molecules in the extracts to bind to the amyloid beta or oligomeric Aβ. The samples were then spun at high speed, meaning that the beads and any other molecules bound to them sunk and formed pellets at the bottom of the tubes. Each pellet was then analyzed to see which molecules it contained.

The experiments identified more than a hundred human brain proteins that can bind to amyloid beta. One of them, known as somatostatin, selectively binds to oligomeric Aβ. Wang et al. were able to determine the structural features of somatostatin that control this binding.

Finally, in further experiments performed in test tubes, Wang et al. noticed that smaller oligomeric Aβ clumps were more likely to form than larger amyloid beta deposits when somatostatin was present. This could signify a previously unrecognized role of somatostatin in the development of Alzheimer’s disease.

Further studies are now needed to confirm whether the presence of somatostatin in the brain favors the formation of smaller, toxic oligomeric Aβ clumps over large innocuous amyloid beta deposits. If so, new treatments could be developed that aim to reduce oligomeric Aβ levels in the brain by preventing somatostatin from interacting with amyloid beta molecules. Wang et al. also suggest that somatostatin could be used in diagnostic tests to detect abnormal levels of oligomeric Aβ in the brain or body fluids of people who have Alzheimer’s disease.

https://doi.org/10.7554/eLife.28401.002

Introduction

The amyloid beta (Aβ) peptide represents a 37 to 49 amino acids endoproteolytic fragment of the amyloid precursor protein (APP), a single-span transmembrane protein, which in humans is coded on the long arm of Chromosome 21 within cytogenetic band q21.3 (Kang et al., 1987). The cellular biology that governs the formation and clearance of Aβ has been understood to play a critical role in Alzheimer disease (AD) since it was shown that fibrillary aggregates of Aβ represent the main constituents of extracellular amyloid plaques that accumulate in the brains of individuals afflicted with the disease (Glenner and Wong, 1984). The endoproteinases responsible for the release of Aβ are broadly referred to as secretases, with β- and γ-secretases being responsible for the critical cleavage reactions that cause the N- and C-terminal release of the Aβ peptide from its APP precursor, respectively. Mutations in the human APP gene (Goate et al., 1991) or the genes coding for presenilin 1 or 2, catalytic subunits of γ-secretases (Rogaev et al., 1995; Sherrington et al., 1995), which cause an increase in Aβ levels or shift the balance of Aβ peptides of different lengths in favor of the production of Aβ1-42, remain the only known causes for early-onset familial manifestations of AD.

Although the extracellular amyloid deposits themselves are increasingly being viewed as innocuous sinks for misfolded Aβ, they may continue to play a role as reservoirs from which monomeric or oligomeric Aβ, hereafter referred to as mAβ or oAβ (see also (Yang et al., 2017)), can emanate (Selkoe, 2001). These soluble forms of Aβ may interact with other molecules they meet in the extracellular space (Narayan et al., 2011), or may bind to molecules embedded in the plasma membrane (Laurén et al., 2009), which may be a precursor to their endocytosis (Jin et al., 2016). Although it is not known how precisely Aβ, which has been taken up into the cell by endocytosis, can overcome the bilayer that surrounds the endolysosomal vesiculo-tubular network, it is well established that the peptide can also reach the cytoplasm and associates with other intracellular compartments, including mitochondria (Lustbader et al., 2004).

The ability of Aβ to overcome compartmental boundaries allows the peptide to encounter a wide variety of proteins that may essentially exist anywhere in the brain. Within the extracellular space, Aβ has, for instance, been shown to interact with apolipoproteins E (Strittmatter et al., 1993) and J (the latter is also known as clusterin) (Narayan et al., 2011; Ghiso et al., 1993). These interactions with apolipoproteins are relevant in the AD context, as certain polymorphisms in the genes coding for them have been shown to bestow an increased risk to develop late-onset AD (Lambert et al., 2009; Harold et al., 2009; Corder et al., 1993). When neurons are exposed to soluble oAβ, a cascade of events is thought to unfold, which leads to the intracellular deposition of hyperphosphorylated tau in the form of so-called neurofibrillary tangles (NFT) (Grundke-Iqbal et al., 1986a, 1986b). The relationship between these two pathobiological features of the disease has remained enigmatic and, although there is broad agreement that signaling downstream of oAβ is toxic for cells, a bewildering number of hypotheses exist regarding the mechanism by which oAβ-dependent toxicity manifests. Receptor candidates proposed to mediate oAβ toxicity include RAGE receptors (Origlia et al., 2008), insulin receptor-sensitive Aβ-binding protein (De Felice et al., 2009), P/Q calcium channels (Nimmrich et al., 2008), sphingomyelinase (Grimm et al., 2005), the Aβ parent molecule APP itself (Lorenzo et al., 2000; Shaked et al., 2006; Sola Vigo et al., 2009), amylin receptors (Fu et al., 2012), a subset of integrins (Wright et al., 2007), the prion protein (PrP) (Laurén et al., 2009) and the metabotropic glutamate receptor 5 (which may act as a co-receptor for PrP (Um et al., 2013; Hu et al., 2014)). Within cells, Aβ has been reported to associate with more than a dozen additional proteins, including glyceraldehyde dehydrogenase (GAPDH) (Verdier et al., 2005), the mitochondrial ATP synthase complex (Schmidt et al., 2008), and a 17-β-hydroxysteroid dehydrogenase X (HSD10) (Yan et al., 1997), also known as 3-hydroxyacyl-CoA dehydrogenase type-2 or Aβ-binding alcohol dehydrogenase (ABAD).

Countless isolated experimental paradigms, differences in Aβ preparations and a broad spectrum of methods underlie the discoveries of the aforementioned Aβ binding candidates, a dissatisfactory status quo, also lamented by others (Mucke and Selkoe, 2012). We therefore set out to locate pertinent literature studies that made use of a more systematic mass spectrometry (MS)-based discovery approach for the identification of Aβ interactors. Surprisingly, whereas several studies are available that interrogated the APP interactome by affinity purification MS (Kohli et al., 2012; Bai et al., 2008), a similar investigation has, to our knowledge, not yet been reported for the Aβ peptide.

To address this shortcoming, we have now undertaken an in-depth search for proteins that can bind to Aβ using biotinylated Aβ peptides as baits and human frontal lobe extract as the biological source for the capture of Aβ binding proteins. We made use of isobaric tagging for the relative quantitation of proteins and capitalized on recent advances in mass spectrometry instrumentation. In addition to confirming many of the previously known Aβ interactors and revealing more than a hundred novel Aβ binding candidates, the study uncovered a surprising and selective interaction between oligomeric Aβ and the small cyclic peptide somatostatin (SST). We demonstrate that SST (i) preferentially binds to aggregated Aβ, (ii) influences Aβ aggregation, (iii) traps a proportion of Aβ in oligomeric assemblies of 50 to 60 kDa, (iv) masks the ability of several widely-used antibodies to detect Aβ, and (v) may form a complex with monomeric Aβ that can induce tau hyperphosphorylation in primary hippocampal neurons.

Results

Workflow of Aβ interactome analyses

The primary objective of this study was to generate an in-depth inventory of human brain proteins that oligomeric preparations of Aβ1-42 can bind to using an unbiased in vitro discovery approach. Synthetic Aβ1-42 peptides and brain extracts generated from adult human frontal lobe tissue served in these studies as baits and biological source materials, respectively. oAβ1-42 was prepared by aggregating the peptide at 4°C for 24 hr, using previously described procedures known to generate amyloid-β-derived diffusible ligands (ADDLs) (De Felice et al., 2009; Krafft and Klein, 2010). ADDLs are understood to be composed of a heterogenous mixture of oligomeric and prefibrillar Aβ aggregates. This heterogeneity ensured that the analysis was not limited to particular oligomeric Aβ assemblies, which were observed to predominate with some alternative preparation protocols (Barghorn et al., 2005; Ahmed et al., 2010). Because the interaction with a given binding partner may involve a binding epitope that comprises N- or C-terminal residues of Aβ1-42, initially two separate experiments (I and II) were conducted, which differed in the orientation designated for tethering the oAβ1-42 bait to the affinity matrix (Figure 1A). To facilitate meaningful comparisons across experiments, the method of Aβ1-42 capture was not based on immunoaffinity reagents. Instead, alternative Aβ1-42 baits were equipped with biotin moieties attached to the N- or C-terminus by a 6-carbon linker chain, enabling their consistent affinity-capture on streptavidin agarose matrices. Large aggregates were removed prior to the bait capture step by centrifugal sedimentation. Biotin-saturated streptavidin agarose matrices served as negative controls and three biological replicates of samples and controls were generated for each interactome dataset by reproducing the affinity-capture step side-by-side on three separate streptavidin agarose affinity matrices that had been saturated with the biotinylated baits. To identify differences in protein-protein interactions of monomeric versus oligomeric Aβ1-42, a third interactome experiment was conducted in which oAβ1-42-biotin or mAβ1-42-biotin served as baits. Digitonin-solubilized brain extracts, which are known to primarily comprise extracellular and cellular proteins (except for nuclear proteins) served as biological starting materials, consistent with the main subcellular areas previously reported to harbor Aβ. Following extensive washes of affinity matrices in their protein-bound state, binders to the bait peptides were eluted by rapid acidification, fully denatured in 9 M urea, and trypsinized. To avoid notorious confounders related to variances in the subsequent handling and analysis of samples, individual peptide mixtures were labeled with distinct isobaric tandem mass tags (TMT) in a six-plex format, then combined and concomitantly subjected to ZipTip-based pre-analysis clean-up by strong cation exchange (SCX) and reversed phase (RP) separation. Four-hour split-free reversed phase nanospray separations were online coupled to an Orbitrap Fusion Tribrid mass spectrometer, which was configured to run an MS3 analysis method. Tandem MS spectra were matched to peptide sequences by interrogating the human international protein index (IPI) using Sequest and Mascot algorithms. The relative levels of individual peptides in the six samples could be determined by comparing the intensity ratios of the corresponding TMT signature ions in the low mass range of MS3 fragment spectra.

Summary of Aβ1-42 interactome analyses.

(A) Workflow of interactome studies designed to capture binders to oligomeric Aβ1-42 tethered to the streptavidin matrix by N-terminal (Experiment I) or C-terminal (Experiment II) biotin groups, or comparing binders to oligomeric versus monomeric Aβ1-42 (Experiment III). (B) Representative chart from interactome dataset generated in Experiment I, depicting the false discovery rates of peptide-to-spectrum matches and benchmarks of the analysis depth. (C) ‘Cellular Component’ Gene Ontology analysis of top 200 proteins that exhibited the most pronounced oAβ1-42 co-enrichment in Experiment I on the basis of their isobaric signature ion distribution. Asterisks indicate TMT labels that were omitted in a subset of quantitative mass spectrometry experiments.

https://doi.org/10.7554/eLife.28401.003

The Aβ1-42 interactome

The three comparative Aβ1-42 interactome analyses (Experiments I-III) conducted in this study led to mass spectrometry datasets, which were characterized by similar benchmarks of data quality and enabled confident assignments of several thousand mass spectra to human peptides. For example, Experiment I generated 9661 spectra, which passed confidence criteria applied. These filtered spectra could be matched to 4352 unique peptides, which in turn formed the basis for the identification of 1074 protein groups. The designation ‘protein groups’, as opposed to ‘proteins’, reflects a reality of a subset of tryptic peptide sequences not being uniquely associated with a specific protein. Whenever encountered, only ambiguous assignments are possible. For this specific dataset, the 1074 protein groups were annotated to comprise 1619 unique proteins. No attempt was made to resolve this residual source of ambiguity at the individual peptide level. Instead, a majority of uncertain identifications were removed by requiring protein identifications to be based on the confident assignment of at least three unique peptides with a minimum length of six amino acids (Figure 1B).

To begin to characterize the Aβ1-42 interactome, the top-listed 200 proteins, whose levels were most pronouncedly co-enriched with biotin-oAβ1-42, were subjected to a gene ontology (GO) analysis. This analysis revealed amongst the biotin-oAβ1-42 binding candidates a highly significant overrepresentation (p<1.0 E-50) of ‘Cellular component’ GO classifiers that indicate an association with the ‘extracellular vesicular exosome’, ‘cytosol’, ‘vesicle’ and ‘extracellular region part’ (Figure 1C). A more detailed analysis of the list of biotin-oAβ1-42 (B-oAβ1-42) candidate interactors revealed that all top-listed 50 candidates were identified on the basis of more than ten peptide-to-spectrum matches (PSMs) and with sequence coverages exceeding 50% of their primary structures (Table 1). Moreover, indicative of high selectivity of the affinity capture procedures applied, the abundance levels of these 50 candidate interactors in the B-oAβ1-42-specific samples exceeded their levels in the negative control sample by more than four-fold. Not surprisingly, the Aβ peptide itself was amongst the proteins identified whose enrichment levels in specific versus unspecific affinity capture samples were most pronounced. In contrast, assignments of APP peptides that map to regions outside of Aβ were low scoring (that is, did not pass a 95% significance threshold) and therefore not credible. The highest levels of B-oAβ1-42 co-enrichment were exhibited by peroxiredoxin-5, rab-1A, lactate dehydrogenase, hypoxanthine-guanine phosphoribosyltransferase, ATP synthase subunits beta and gamma, and acyl CoA thioester hydrolase. Several of the Aβ1-42-biotin candidate interactors revealed by this analysis were previously known to interact with Aβ peptides, including ATP synthase (Verdier et al., 2005; Schmidt et al., 2008) and glyceraldehyde-3-phosphate dehydrogenase (Verdier et al., 2005, 2008), and/or had been shown to be present at altered abundance levels in cells exposed to Aβ (Lovell et al., 2005) (Table 1). Observations by others, which preceded this study, have provoked a hypothesis that a free N-terminus of Aβ1-42 might be critical for inducing neurofibrillary degeneration (Jin et al., 2011) and synapse loss (Shankar et al., 2008). We therefore explored how the orientation of Aβ-tethering might affect its protein-protein interactions. When the C-terminally biotinylated oAβ1-42-B bait was employed for affinity capture, a lower number of proteins was consistently observed to co-enrich with the bait and their ratios of co-enrichment (relative to biotin-only negative controls) were generally lower than those observed with the N-terminally tagged B-oAβ1-42 bait (see also Figure 1—source data 1).

Table 1

Top-listed 50 interactors of biotin-Aβ1-42 observed in Experiment I. Except for the Aβ1-42 bait, which is shown in first position, proteins are listed by their enrichment (relative to the biotin-only saturated negative control matrix) observed in Experiment I. Note the extensive amino acid sequence coverage exceeding 44% for all proteins listed. Whenever the same proteins were also observed in Experiments II and III, their corresponding enrichment ratios and counts of peptides quantified are shown in additional columns on the right. (see also Figure 1—source data 1 and Supplementary file 1).

https://doi.org/10.7554/eLife.28401.005
PeptidesExperiment I
B-oAβ1-42/B
Experiment II oAβ1-42-B/B*Experiment III oAβ1-42-B/ mAβ1-42-B
AccessionDescriptionCoverageUniqueTotalRatioCountRatioCountRatioCountAAs
 IPI00006608.1APP770 of Amyloid beta A4 protein (Fragment)62.34%2488.33185.77132.8037770
 IPI00024915.3Peroxiredoxin-5, mitochondrial86.45%81812.2332.1535.9023214
 IPI00005719.1Ras-related protein Rab-1A80.49%41810.523

3.286205
 IPI00219217.3L-lactate dehydrogenase B chain74.55%6269.879

2.463334
 IPI00218493.7Hypoxanthine-guanine phosphoribosyltransferase64.22%7229.0012

2.458218
 IPI00303476.1ATP synthase subunit beta, mitochondrial83.55%14488.593

2.6121529
 IPI00395769.2ATP synthase subunit gamma, mitochondrial59.60%6258.573

3.2919297
 IPI00219452.1Cytosolic acyl CoA thioester hydrolase76.60%12318.3831

4.0331329
 IPI00027223.2Isocitrate dehydrogenase [NADP] cytoplasmic62.56%5308.204


414
 IPI00743713.4Pyruvate kinase64.77%3417.9551.8974.508599
 IPI00029469.1Centractin, beta76.06%3267.877

3.2539376
 IPI00003482.12,4-dienoyl-CoA reductase, mitochondrial88.06%12377.7421

4.3228335
 IPI00000949.1Mu-crystallin homolog80.25%11227.6511

2.4117314
 IPI00028520.2NADH dehydrogenase flavoprotein 177.59%12317.52323.4844.0820464
 IPI00018352.1Ubiquitin carboxyl-terminal hydrolase L163.68%9177.49381.8721.3919223
 IPI00478231.2Transforming protein RhoA51.30%5117.487
5.748193
 IPI00028091.3Actin-related protein 354.78%6247.459

1.814418
 IPI00029468.1Centractin, alpha72.34%7367.34153.9433.1138376
 IPI00335509.3Dihydropyrimidinase-related protein 577.13%8427.056

3.798564
 IPI00479186.7Pyruvate kinase isozymes M1/M293.03%3707.041381.681004.33123531
 IPI00219018.7Glyceraldehyde-3-phosphate dehydrogenase98.21%35506.961602.361212.62251335
 IPI00873027.2Carbohydrate kinase domain-containing protein95.68%15316.95182.1731.7232347
 IPI00179187.4DnaJ homolog subfamily A member 374.17%6316.933

9.578453
 IPI00216456.5Histone H2A type 1 C92.31%3216.83152.03100.5954130
 IPI00016801.1Glutamate dehydrogenase 1, mitochondrial85.13%13466.8216

2.9922558
 IPI00218448.4Histone H2A.Z83.59%5146.7841.8850.5130128
 IPI00220301.5Peroxiredoxin-680.80%12186.6971.4221.5422224
 IPI00472047.1NAD-dependent deacetylase sirtuin-278.32%7286.684

2.5815369
 IPI00169383.3Phosphoglycerate kinase 192.09%18516.61481.38131.7321417
 IPI00895801.1Medium-chain specific acyl-CoA dehydrogenase69.41%8396.393

1.8212425
 IPI00026268.3Guanine nucleotide-binding protein beta-183.82%7356.384

2.5830340
 IPI01015522.1Actin, cytoplasmic 196.83%8376.25332.07264.71129347
 IPI00909207.1Peroxiredoxin-292.35%9236.00131.51142.4013183
 IPI00970967.1GSTT2 protein51.64%5146.007


244
 IPI00298547.3DJ-195.24%11225.87171.4241.8430189
 IPI00796462.1GTP-binding nuclear protein Ran90.60%13245.82264.82144.4246234
 IPI00784154.1Heat shock protein, 60 kDa, mitochondrial94.76%35715.68692.27174.2173573
 IPI00003362.378 kDa glucose-regulated protein58.72%6445.65201.3912.427654
 IPI00011229.1Cathepsin D84.95%11355.5215

2.1026412
 IPI00013683.2Tubulin beta-3 chain84.44%10435.3951.1812.0485450
 IPI00032406.1DnaJ homolog subfamily A member 290.29%10385.285

6.1019412
 IPI00419802.43-hydroxyisobutyryl-CoA hydrolase64.25%5225.277


386
 IPI00219037.5Histone H2A.x92.31%5235.2331.6990.6748143
 IPI00556376.2Dihydropyrimidinase-related protein 189.21%21605.21571.39273.7167686
 IPI00000792.1Quinone oxidoreductase68.69%3185.184

329
 IPI00295400.1Tryptophanyl-tRNA synthetase, cytoplasmic66.67%6335.179

3.972471
 IPI00440493.2ATP synthase subunit alpha, mitochondrial99.46%38895.17622.16324.88131553
 IPI00029111.3Dihydropyrimidinase-related protein 380.99%4495.13151.42103.4019684
 IPI00019171.1Endophilin-A178.41%10375.0371.2811.0740352
 IPI00005198.2Interleukin enhancer-binding factor 244.36%3134.984

1.416390
 IPI00030363.1Acetyl-CoA acetyltransferase, mitochondrial86.65%19504.81332.64102.4543427
 IPI00413344.3Cofilin-290.96%6224.79251.81136.9311166
 IPI00007068.1Actin-related protein 3B50.96%4214.7711


418
 IPI00012011.6Cofilin-193.98%15294.67561.75375.3456166
  1. *Note that SST14 is not included in this list because this protein only came to the fore as the most selective oAβ1-42-B interactor when the ≥6 amino acids and ≥3 peptides per protein requirements were waived.

To determine the extent to which the binding of individual Aβ interactors was governed by Aβ aggregation, we next compared directly the interactome of proteins that bind to monomeric versus oligomeric preparations of the Aβ1-42-biotin bait. This experiment revealed a preference amongst the most highly enriched Aβ1-42 candidate interactors for binding to pre-aggregated bait peptides (Table 1). Exceptions represented the robust and preferred binding of histones H2, H3 and H4 to monomeric Aβ1-42 bait matrices (please see Supplementary Table 1 in Supplementary file 1). Moreover, a majority of proteins, which exhibited lower levels of Aβ1-42 co-enrichment in Experiments I and II, were observed to bind preferentially to monomeric Aβ bait peptides (Figure 1—source data 1).

Pre-aggregated Aβ binds to SST14 but not to its preprosomatostatin precursor through a binding epitope in the N-terminal half of Aβ1-42

When we waived the requirement that peptides had to be at least six amino acids long to be considered for protein identification, a tandem mass spectrum assigned to the five amino acid sequence ‘NFFWK’ came to the fore, which exhibited pronounced preference for binding to C-terminally tethered oAβ1-42-B. A query of human genome databases revealed that this peptide, owing to its unusual composition, and despite its short length, could only have originated from the well-known paralogs preprosomatostatin or preprocortistatin (Figure 2A). This conclusion was strengthened by the fact that this peptide is naturally preceded by a tryptic cleavage site and the exquisite match between observed and in silico predicted fragment ions (Figure 2B). In light of the high intensity ion counts of fragments observed for this peptide and its robust co-enrichment with oAβ1-42-B in three biological replicates (Figure 2C), it first seemed puzzling that the identification of this protein group was not corroborated by other spectra that map to sequences outside of the ‘NFFWK’ sequence stretch. Searching for a plausible explanation, it became apparent that preprosomatostatin and preprocortistatin give rise to cyclic neuropeptide hormones through a series of posttranslational trimming steps, and the ‘NFFWK’ peptide is the only tryptic peptide derived from the mature hormone that is of sufficient length and distinct sequence to be readily identifiable by MS (Figure 2A). To further explore if binding had occurred to the mature neuropeptide, as opposed to the precursor, and to characterize the Aβ binding requirements, streptavidin-based affinity matrices were next side-by-side pre-saturated with biotin, N- or C-terminally biotinylated Aβ1-42 or an N-terminally truncated Aβ17-42-biotin bait. The subsequent application of an identical interactome analysis workflow as outlined above (but replacing TMT with iTRAQ isobaric labeling) (Figure 1A) corroborated the propensity of ‘NFFWK’ to bind to oAβ1-42-B but also established that the free N-terminus is indeed essential for this peptide-peptide interaction (Figure 2D). Although neither preprosomatostatin nor preprocortistatin had passed stringent filtering criteria required for inclusion in the Aβ interactome data tables (because their identification could not be based on at least three unique peptides with a minimum length of six amino acids), close inspection of Experiment III data under omission of these filters suggested that mature somatostatin was not only present in the dataset but represented the protein, whose levels were most selectively enriched in oAβ1-42-B affinity capture eluates. This conclusion was further corroborated by two additional PSMs that mapped to regions outside the SST14 neuropeptide sequence and one additional five amino acid peptide of the sequence ‘TFTSC’ that could be assigned (albeit not unambiguously) to the mature SST14 neuropeptide sequence itself. More specifically, consistent with the notion that Aβ1-42 interacts separately with the SST14 neuropeptide and preprosomatostatin precursor, the distributions of TMT signature ions derived from tryptic ‘NFFWK’ and ‘TFTSC’ peptides derived from the mature SST14 neuropeptide exhibited identical TMT signature ion profiles patterns that differed fundamentally from the respective TMT profiles of preprosomatostatin peptides which mapped to sequences upstream of the SST14 sequence domain (Figure 2E). In agreement with the interpretation that the ‘NFFWK’ had probably originated from SST14, not CST17, no peptides were observed in this or other experiments that could be uniquely assigned to preprocortistatin. Finally, these experiments repeatedly established that the ‘NFFWK’ peptide binds preferentially to oligomeric (pre-aggregated) but not monomeric Aβ1-42-biotin (Figure 2E).

Discovery of somatostatin as a candidate interactor of oligomeric Aβ1-42.

(A) Sequence alignment of preprocortistatin and preprosomatostatin. The signal sequence and the boundaries of the bioactive cortistatin and somatostatin peptides are indicated by horizontal bars. Identical residues are highlighted by black background shading, and peptide sequences observed by mass spectrometry are shown in colored fonts. (B) Example tandem MS spectrum supporting the identification of a peptide with the amino acid sequence ‘NFFWK’. Fragment masses attributed to B- and Y- ion series are shown in red and blue colors, respectively. (C) Expanded view of MS3 spectrum derived from ‘NFFWK’ parent spectrum in interactome study based on oAβ1-42-biotin bait and biotin only negative control (Experiment II). In this view, the relative intensities of signature ions reflect the relative abundances of the ‘NFFWK’ peptide in the six side-by-side generated affinity purification eluate fractions. (D) SST/CST in human frontal lobe extracts binds to oAβ1-42-biotin but not to N-terminal biotinylated or truncated Aβ baits. iTRAQ signature ion intensity distribution in experiment probing the relative ability of four different biotin baits to capture SST/CST from human brain extract. The exclusive presence of a high intensity 116 ion indicates that the ‘NFFWK’ fragment spectrum, which gave rise to this peak distribution, was dependent on SST/CST exclusively associating with oAβ1-42-biotin. (E) Preferential binding of SST to pre-aggregated oAβ1-42. TMT signature ion intensity distributions of four MS3 spectra assigned to preprosomatostatin based on oligomeric or monomeric Aβ1-42-biotin baits (Experiment III). PSMs derived from SST-14 (‘TFTSC’ and ‘NFFWK’ peptides) had in common signature ion intensity distributions characterized by high intensity even-numbered TMT fragments. In contrast, signature ion intensity distributions of preprosomatostatin-derived tryptic peptides outside of the SST-14 coding region were relatively evenly distributed.

https://doi.org/10.7554/eLife.28401.006

Validation of SST14 binding to oAβ1-42

To further characterize the binding of SST14 to oAβ1-42, we next undertook a series of biochemical binding experiments. In a first reductionist approach, we explored if binding of SST14 requires co-factors or occurs directly and, therefore, can be observed with synthetic peptides. For this experiment, oAβ1-42 preparations (generated by incubating mAβ1-42 purified by gel filtration at 37°C for 2 hr with shaking at the speed of 700 rpm) gave rise to a pronounced signal of high molecular mass (HMM), not seen with mAβ1-42, when analyzed by denaturing SDS-PAGE and immunoblotting with the 6E10 antibody directed against an N-terminal Aβ1-42 epitope (Figure 3A). Next, streptavidin agarose matrices were pre-saturated with N-terminally biotinylated SST14 and exposed to unlabeled monomeric or oligomeric Aβ1-42 preparations. Finally, affinity capture matrices were boiled in Laemmli sample buffer and again immunoblotted with the 6E10 antibody (Figure 3B). The analysis revealed in samples which had been exposed to oAβ1-42 robust 6E10 signals both in the HMM and low mass region of the blot, the latter migrating at the same level as mAβ1-42. Because very little low mass signal was observed in samples, which had been incubated with mAβ1-42, we concluded that the low mass signal in oAβ1-42 exposed samples most likely reflected a release of a fraction of Aβ1-42 molecules from SST14-bound oAβ1-42 aggregates under the harsh denaturing analysis conditions. Taken together, this experiment validated that SST14 binding only occurs when Aβ1-42 is available in oligomeric form and established a direct mode of interaction.

Validation of SST binding to oAβ1-42 but not to mAβ1-42.

(A) Western blot analysis of synthetic mAβ1-42 and oAβ1-42. The asterisk designates a signal caused by the partial release of mAβ1-42 from high molecular mass (HMM) oAβ1-42 in the presence of SDS. (B) Biotin-SST affinity capture of oAβ1-42 but not mAβ1-42. (C) Evidence for fluorescence energy transfer between (FRET) between Edans-SST14 donor and Aβ1-42-TMR acceptor. Samples containing donor and acceptor peptides at 20 μM concentrations were incubated overnight at physiological pH. Note the profound quenching of the donor signal and increase in acceptor fluorescence relative to negative control preparations that contained only donor or acceptor peptides but were otherwise treated identically. (D) Competition FRET analysis based on configuration shown in panel ‘D’ but with unlabeled SST14 or the negative control AVP peptide being added at varying concentrations to the assay mix. Note the rescue of donor fluorescence in the presence of unlabeled SST14 but not AVP.

https://doi.org/10.7554/eLife.28401.007

To further validate the SST14-oAβ1-42 interaction and derive a binding constant we employed fluorescence resonance energy transfer (FRET) methodology based on a pairing of an SST14 donor and Aβ1-42 acceptor labeled with Edans and TMR fluorophores, respectively. To enable the formation of Aβ1-42 oligomers and aggregates, fluorescence spectra were recorded following a 24 hr incubation. Following excitation at a wavelength of 335 nm, the recording of independent fluorescence spectra of donor or acceptor peptides over a window of wavelengths spanning 350 nm to 650 nm gave rise to the expected Edans and TMR fluorescence intensity maxima at 493 nm and 596 nm, respectively (Figure 3C). Mixing of the two peptides caused profound quenching of the donor signal, concomitant with an increase in acceptor fluorescence, indicating energy transfer. The acceptor sensitization at 596 nm discouraged an interpretation whereby the donor quenching merely reflected an unspecific amorphous co-precipitation of SST14 with Aβ. Next, this experimental setup was exploited to derive a first indication of the binding constant that governs the interaction between SST14 donor and Aβ1-42 acceptor. To this end, various concentrations of unlabeled SST14 were added to the reaction mixture in a FRET competition experiment to determine the concentration (EC50) for restoring donor fluorescence (Figure 3D). A half-maximal effective concentration of 12.7 μM SST14 could be deduced from the recorded graph. In contrast, addition of [Arg8]-vasopressin (AVP), a cyclic neuropeptide of similar size and biochemical characteristics as SST14, did not restore donor quenching in a control experiment, further strengthening the conclusion that the energy transfer reflected a specific interaction between the SST14 donor and Aβ1-42 acceptor.

SST14 and CST17 delay Aβ1-42 aggregation in a Thioflavin T fluorescence assay

To explore the influence of SST14 on Aβ1-42 more rigorously, a Thioflavin T (ThT) fluorescence assay was applied that incorporated previously reported methodology advancements (Hellstrand et al., 2010), including the removal of residual Aβ1-42 aggregates by size-exclusion chromatography immediately prior to the recording of ThT fluorescence spectra (Figure 4A). The application of this method revealed that the presence of 15 μM SST14 (sourced from two independent vendors) in the reaction mix reproducibly extended the lag phase of Aβ1-42 aggregation by more than two hours and lowered the signal amplitude during the subsequent stationary phase (Figure 4B). The presence of an intact disulfide bridge within SST14 was not required for this delay to occur. No delay was observed when SST14 was replaced by the AVP negative control peptide. In further variations to this experimental setup it was established that pre-incubation of SST14, at a concentration which causes this peptide to pre-aggregate (see also Figure 4—figure supplement 1), did not affect its ability to delay ThT incorporation into Aβ1-42 aggregates (Figure 4C). Importantly, the SST14-dependent lag phase extension correlated directly with the concentration of SST14 in the reaction mix (Figure 4D). Moreover, a similar lag phase extension and/or reduction in ThT fluorescence intensity was not observed when the Aβ1-42 peptide was replaced in the same assay with amylin, another well-known amyloidogenic peptide of similar size (37 amino acids). More specifically, the presence of SST had no influence on the ThT fluorescence curve collected with a reaction mix containing 10 μM amylin. This result was observed even when SST was added to the reaction well at a considerably higher concentration of 100 μM (Figure 4—figure supplement 2). Finally, an experiment in which SST14 was replaced by CST17 revealed that the latter peptide is even more potent in its ability to interfere with the Aβ1-42 aggregation-dependent ThT incorporation. Thus, whereas the addition of 15 μM SST14 delayed Aβ1-42 aggregation by a few hours, addition of CST at the same concentration completely abrogated ThT incorporation (Figure 4E).

Figure 4 with 2 supplements see all
SST14 and CST17 delay Aβ1-42 aggregation in ThT fluorescence assay.

(A) Workflow of ThT-based aggregation assay. (B) Representative ThT fluorescence charts using synthetic Aβ1-42 alone or in combination with SST14 or negative control AVP peptides. Note that SST14 alone does not contribute to ThT fluorescence in this assay at these relatively low concentrations. (C) Pre-aggregation of SST14 did not alter its effect on Aβ1-42 aggregation. (D) Evidence that the SST14-dependent delay in Aβ1-42 aggregation is SST14 concentration dependent. (E) Like SST, CST17 causes a concentration-dependent inhibition of Aβ1-42-dependent ThT fluorescence. Note that at 15 μM concentrations, CST17 appears more potent in this regard than SST14. Please see legend for experimental conditions.

https://doi.org/10.7554/eLife.28401.008

At concentrations above 3% w/w (~20 mM) the SST14 peptide had been observed to acquire amyloid characteristics, including Congo Red birefringence (van Grondelle et al., 2007). In our hands, even considerably lower concentrations of SST14 exceeding 250 μM reproducibly caused SST14 to incorporate ThT following a 2 hr lag phase (Figure 4—figure supplement 1). We used this property to determine if the presence of monomeric or oligomeric 1 μM Aβ1-42 influences SST aggregation. Note that the presence of 1 μM Aβ1-42 in this experiment does not contribute to ThT incorporation in this reversed assay configuration. More specifically, whereas the presence of 15 μM merely causes a profound lag phase extension (Figure 4D) the addition of 250 μM SST14 abolishes the ability of 1 μM Aβ1-42 to form aggregates that can incorporate ThT (not shown). Also note that in contrast to the first assay configuration, when low micromolar concentrations of SST14 were added to 1 μM Aβ1-42, in this second configuration of the ThT assay, substoichiometric quantities of 1 μM Aβ1-42 were added to a large excess of 250 μM SST. Strikingly, this analysis revealed that the presence of 1 μM Aβ1-42 caused a profound delay in SST14 aggregation-dependent ThT incorporation but only when Aβ1-42 had been allowed to pre-aggregate (Figure 4—figure supplement 1B). Taken together, these experiments corroborated the notion that the interaction of SST14 with Aβ1-42 relies on the presence of pre-aggregated oAβ1-42.

Binding of SST14 or CST17 precludes detection of Aβ1-42 with commonly used antibodies

To determine if the presence of SST14 or CST17 simply prevented Aβ1-42 aggregation, or had caused the formation of alternative heteromeric complexes that failed to incorporate ThT, samples prepared as described in the previous section (but under omission of ThT) were analyzed by Western blotting using antibodies targeting N-terminal, central or C-terminal epitopes within Aβ1-42 (Figure 5A). Initially, we were interested in observing the influence of incubation time on sample composition (Figure 5B). As expected, the incubation of monomeric Aβ1-42 alone over a span of 18 hr caused its signals to gradually shift to higher molecular mass bands, reflecting the appearance of Aβ1-42 aggregates. In the presence of SST14 or CST17, 6E10-reactive signals that migrated at the level of monomeric Aβ1-42 also disappeared over the course of 18 hr albeit at a slower rate than observed with Aβ1-42 alone, most likely reflecting the delayed aggregation (Figure 4D and E). Strikingly, however, in the presence of SST14 or CST17 only very weak HMM signals of lower apparent molecular weight (50-70 kDa) than those seen in preparations of Aβ1-42 alone (55-200 kDa) could be seen following 18 hr incubation (Figure 5B, lanes 8 and 10). To further explore the apparent disappearance of 6E10-reactive bands during this time-course experiment, we next repeated the 18 hr incubation of Aβ1-42 in the presence of varying concentrations of SST14 or CST17. This experiment documented again an inverse correlation of the signal intensity of SST14 and CST17 concentrations and 6E10-reactive blot signals (Figure 5C). This effect was not limited to detection by the 6E10 antibody but was similarly observed with antibodies targeting a central Aβ1-42 epitope (4G8) or the very C-terminus of Aβ1-42 (12F4) (not shown). The disappearance of signals could not be attributed to some unexpected disappearance of the Aβ1-42 peptide in these samples, because silver-stain analyses of the same fractions revealed relatively stronger Aβ1-42 signals in CST17 and SST14 containing samples. Rather, all available data suggest that the interaction of SST14 or CST17 masked the ability of these antibodies to bind to Aβ1-42.

Figure 5 with 1 supplement see all
Binding of SST14 or CST17 precludes detection of Aβ1-42 with commonly used antibodies.

(A) Schematic highlighting Aβ1-42 binding epitopes of antibodies used to generate this and the subsequent figure. (B) Western blot-based time-course analysis of Aβ1-42 aggregation in the presence or absence of SST14 or CST17. (C) SST14- and CST17-dependent masking of Aβ1-42 binding epitopes. Note the relatively more intense silver-stained bands of monomeric Aβ1-42 (red arrowhead) in samples containing the highest concentrations of SST14 or CST17 (lanes 6, 7 and 9), yet virtual absence of Aβ-specific immunoblot signals in the corresponding lanes. Note also the faster migrating band (green arrowhead) representing monomeric CST or SST observed in samples, which contained the highest concentrations of CST or SST (lanes 7 and 9).

https://doi.org/10.7554/eLife.28401.011

Aβ1-42 forms a distinct 50-60 kDa SDS-stable complex in the presence of SST14

We next sought to compare the effect of SST14 (or CST17) on the aggregation of Aβ1-42 and Aβ1-40. To learn more about how the presence of SST14 (or CST17) affects Aβ aggregation, ThT assay fractions were in this experiment collected and further analyzed by Western blotting. To compensate for the abovementioned epitope masking effect of SST14 or CST17 (Figure 5C), these experiments were undertaken with 5 to 10 times higher peptide concentrations than those described above (Figures 4 and 5). As expected, at these higher concentrations, Aβ1-42 and Aβ1-40 exhibited shorter lag phases in the ThT fluorescence assay of 30 min and 45 min, respectively (Figure 6A). Consistent with the ThT fluorescence data collected at lower peptide concentrations (Figure 4), the presence of CST17 (or SST14, not shown), but not AVP, extended the lag phase of Aβ aggregation considerably (by approximately one hour). However, except for slightly longer lag phases observed in fractions containing Aβ1-40, there was no difference in Aβ1-42 and Aβ1-40 kinetics revealed by this assay. Next, Western blot analyses of ThT assay fractions were undertaken without or with prior boiling in SDS. Interestingly, Aβ1-42-containing samples subjected to the milder treatment, which had been incubated for 18 hr in the presence of SST14 or CST17, gave rise to 6E10-reactive bands that migrated at 50-60 kDa, not seen in samples incubated either without these peptides or with AVP (Figure 6B, lanes 2 and 3). These intermediate-sized 6E10-reactive bands were even more pronounced when the same samples had been boiled in the presence of SDS (Figure 6B, lanes 10 and 11) but escaped detection with an antibody directed against the C-terminus of Aβ1-42 (Figure 6B, lanes 6 and 7). Interestingly, similar bands were not observed when Aβ1-40 served as the substrate for aggregation (Figure 6B, lanes 22 and 23), suggesting that the two most C-terminal residues of Aβ1-42 confer properties that are essential for the formation of the 50-60 kDa oligomeric assemblies. Only when SDS was present in the sample buffer, low mass bands appeared in the Western blot that migrated at a level expected for the monomeric Aβ peptide, consistent with the interpretation that these peptides were released from larger oligomeric or fibrillar aggregates under denaturing conditions (see also Figure 5). Strikingly, in samples that contained SST14 (or CST17) a fraction of these low mass bands, less than what would be needed for their detection by silver-staining, were occasionally (in well-resolved gels) observed to migrate at levels expected for heterodimers of Aβ1-42 and SST14 (or CST17) (Figure 6B, lanes 5, 6, 10 and 11). Consistent with the 3-amino acid smaller size of SST14, relative to CST17, the respective heterodimers containing SST14 migrated slightly faster than those containing CST17. A five-fold increase in Aβ1-42 levels in the reaction mix further emphasized the appearance of these low mass heterodimers but also revealed strong signals migrating at the expected size of Aβ1-42 dimers (Figure 6B, lanes 13 to 16). Notably, whereas evidence for heterotrimers composed of two Aβ1-42 molecules and one SST14 (or CST17) molecule was never obtained, we occasionally observed signals that migrated at molecular masses expected for SDS-stable complexes consisting of three Aβ1-42 peptides linked to SST14 (or CST17) (Figure 6B, lanes 14 and 15). Corroborating this interpretation was the fact that bands matching this size were never observed in fractions lacking SST14 (or CST17) or containing AVP as a negative control peptide.

Aβ1-42 forms SDS-stable oligomeric complexes of 50-55 kDa in the presence of SST14 (or CST17).

(A) In the presence of CST17 (or SST14, not shown) the ThT fluorescence curve or Aβ1-42 or Aβ1-40 is characterized by an extension of the lag phase and a reduction in ThT fluorescence. (B) CST17 (or SST14) co-assemble with Aβ1-42 into oligomers of 50-55 kDa that withstand boiling (lanes 2 and 3) but partially disintegrate in the presence of SDS. Immunoblot analyses with antibodies directed against the C-terminus (x-42) (lanes 6 and 7) or an N-terminal epitope (6E10) (lanes 10 and 11) revealed bands of 5-6 kDa, consistent with the existence of SDS-resistant heterodimeric complexes of mAβ1-42 and SST14 (or CST17). Note the well-defined oligomeric bands of 50 and 55 kDa (lanes 10 and 11) that were observed in samples derived from the co-incubation of SST14 (or CST17) with Aβ1-42 (lanes 10, 11, 18, 19). Note also that signals interpreted to represent trimeric Aβ1-42 (t) (lanes 13-16), but not dimeric Aβ1-42 (d), can be seen to migrate slower in the presence of SST14 (or CST17) (lanes 14 and 15). Finally, intensity levels of homodimeric Aβ1-42 bands are reduced in the presence of SST14 (or CST17) (compare lanes 17 and 20 with lanes 18 and 19). Black arrowhead labeled with ‘m’, ‘d’, and ‘t’ designate bands interpreted to consist of monomeric, dimeric and trimeric Aβ1-42. Green and red arrowheads were used to label bands interpreted to represent SDS-stable heteromeric building blocks consisting of SST14 (or CST17) bound to monomeric and trimeric Aβ1-42, respectively.

https://doi.org/10.7554/eLife.28401.013

These biochemical analyses indicated the formation of SDS-stable oligomeric complexes of Aβ1-42 and SST14 (or CST17) made of building blocks comprising one or three Aβ1-42 peptides.

Tryptophan-8 is critical for SST14-dependent perturbation of Aβ1-42 seeding and fibril growth

To delineate key residues and minimal components required for SST to interact with Aβ1-42, we next had a series of SST-derived peptides custom-synthesized that differed in one or two amino acids from the wild-type SST sequence or were truncated, thereby missing the internal disulfide bridge. Based on prior high-resolution NMR models available for the cyclic peptide and structure-activity relationship data derived from SST-receptor docking studies, we hypothesized that a striking hydrophobic ‘belt’ composed of three phenylalanines and a tryptophan, which is flanked by two lysine side-chains in the 3D rendering of SST (Figure 7A,B and C), might also be critical for binding to Aβ1-42. With regard to the juxtaposed phenylalanines, we were in particular interested to learn if the well-known ability of their side-chains to engage in pi stacking plays a role in the SST-Aβ1-42 interaction. Interestingly, the replacement of one or two of the phenylalanines to leucines did not rescue the lag phase extension phenotype in the ThT fluorescence assay, indicating that the aromatic nature of these residues is not essential for the interaction. However, when we replaced the single tryptophan present in position 8 both the lag phase extension and the reduction in total ThT fluorescence observed in the presence of wild-type SST were rescued (Figure 7D). This outcome was documented when tryptophan was replaced with alanine, proline, histidine or tyrosine. The fact that even the replacement with tyrosine rescued the lag phase extension phenotype indicates that tryptophan does not merely act in this context by providing an aromatic side-chain but suggests that other structural features present in tryptophan provide the necessary fit for binding to Aβ1-42 in this PBS-based system.

Figure 7 with 2 supplements see all
Tryptophan-8 in SST14 sequence is essential for lag phase extension of Aß1-42 in ThT incorporation assay.

(A–C) NMR-structure of SST14 in 5% D-mannitol (RCSF PDB structure ID: 2MI1, Model 1), adapted from Anoop et al. (2014). Renderings were generated in NGL 3D viewer (powered by MMTF). (A) Backbone of SST14 emphasizing secondary structure. (B) and (C) Stick and surface models of SST14 with coloring emphasizing relative hydrophobicity. (D) and (E) Thioflavin T absorbance assay data based on SST point mutants and deletion constructs, respectively. Please see legend for sample compositions.

https://doi.org/10.7554/eLife.28401.014

Analogous experiments with truncated SST peptides revealed that a peptide encompassing SST residues 5-11 can replace wild-type SST with regard to its effect on Aβ1-42 (Figure 7E), thereby upholding the conclusion that the ability of SST to form its internal disulfide bridge is not essential for this effect (Figure 4B). However, further stripping of one amino acid from each end of this peptide resulted in a SST6-10 peptide, which lacked the ability to influence Aβ1-42 aggregation.

We next compared Aβ aggregate size populations in the presence or absence of SST by nanoparticle tracking analysis (based on Nanosight technology [Filipe et al., 2010]). This orthogonal method documented that Aβ aggregates present in ThT aggregation assay fractions after 75 min of incubation were profoundly shifted toward smaller particle sizes when samples were co-incubated with SST (Figure 7—figure supplement 1A and B). Importantly, the replacement of SST with an SST-W8P negative control peptide did not result in this particle size shift toward smaller aggregates (Figure 7—figure supplement 1C), indicating that the altered particle size distribution is caused by the SST-Aβ interaction, rather than by some unspecific effect of the peptide.

Negative stain electron microscopy corroborated the reduction in aggregate size when Aβ1-42 was incubated with SST versus Aβ1-42 alone (Figure 7—figure supplement 2). Incubating 50 μM of Aβ1-42 under aggregation conditions resulted in the appearance of characteristic amyloid fibrils (Figure 7—figure supplement 2A), while in the presence of equimolar concentrations of SST only oligomeric structures were observed (Figure 7—figure supplement 2B).

Taken together, these results demonstrated exquisite specificity of the SST-Aβ1-42 interaction and corroborated the conclusion that Aβ assemblies in the presence of SST are smaller than those observed following Aβ-only aggregation.

SST14 and CST17 can potentiate tau hyperphosphorylation when added to the cell culture medium together with monomeric Aβ1-42

A 2009 milestone report revealed that signaling downstream of SST positively modulates transcription of the neprilysin gene (Saito et al., 2005). Consistent with the known role of neprilysin as one of only a few proteases known to contribute to Aβ1-42 degradation, the authors reported an inverse relationship between SST and Aβ1-42 levels in mice deficient for the somatostatin gene. To begin to assess the physiological significance of the interaction between Aβ1-42 and SST14 observed in this study, an experimental paradigm was needed that is both amenable to manipulation and can be provoked to exhibit a phenotype considered relevant in the context of AD research. The primary hippocampal neuron assay, first described in detail by Jin et al. (2011), fits this description (Figure 8A) and was favored over an analysis based on the technically challenging enrichment of the relatively small number of GABAergic interneurons known to release SST naturally. In agreement with prior data by others, robust tau hyperphosphorylation was observed when mouse hippocampal neurons, which had been cultured for 18 days, were exposed to oAβ1-42 but not when mAβ was added to the cell culture dish (Figure 8B). For the purposes pursued here, we restricted the tau phospho-epitope characterization to monitoring occupancy of pThr231 (AT180) and pSer202/pThr205 (AT8) phospho-acceptor sites, two of several known sites that undergo robust hyperphosphorylation also in AD brains, because we had noted that these phospho-epitopes were most responsive to oAβ exposure in the original report (Jin et al., 2011). We then modified the assay by the concomitant addition of synthetic SST14, other cyclic control peptides, or bombesin, which does not carry an internal disulfide bridge but was observed to share aggregation characteristics with SST (Maji et al., 2009). In a series of follow-up experiments, we added a pre-aggregation step for the peptide and/or Aβ1-42, varied the pH or duration under which this pre-aggregation was conducted, or changed the duration of cellular exposure to the peptide mixture as described in detail in the figure legend (Figure 8A,C,D and E). Because SST is naturally released from cells through the regulated secretory pathway when densely packed granules fuse with the plasma membrane (Maji et al., 2009), their content does not dissolve instantly when it encounters the extracellular milieu but dissolves with a time-scale of hours (Anoop et al., 2014). We therefore reasoned that low micromolar SST concentrations, including residual SST aggregates of varying size, might also be locally encountered by neurons in the brain in proximity to SST release sites.

Exposure of primary hippocampal neurons to SST14 and Aβ1-42 can potentiate phosphorylation of tau at sites known to undergo hyperphosphorylation in AD.

(A) Workflow of primary hippocampal neuron assay. Please see methods section for details on peptide preparation. (B) Addition of preaggregated (24 hr) but not monomeric (0 hr) Aβ1-42 to primary hippocampal neurons causes within 24 hr an increase in tau phospho-occupancy at AT8 and AT180 phosphorylation sites, in the absence of an effect on total tau levels (K9JA). (C) Addition of SST or CST bioactive peptides in the absence of Aβ1-42, leads to a reduction in tau phospho-occupancy. (D) Addition of SST14 or CST17, but not negative control cyclic peptides, can potentiate tau phospho-occupancy in the presence of monomeric Aβ1-42. (E) Tau hyperphosphorylation observed following addition of SST14 and mAβ1-42 to the cell culture appears to depend on conducive SST14 pre-incubation conditions. In all panels red and green font designates parameters (pH and pre-aggregation periods) applied to the preparation of Aβ1-42 and other neuropeptides, respectively. Peptides were added separately to the cell culture medium except for samples 3-11, Panel E, when peptides were mixed and pre-incubated for an additional 1 or 3 hr. Exposures of cells to peptides were for 24 hr.

https://doi.org/10.7554/eLife.28401.017

When primary hippocampal neurons were exposed to SST14 alone, levels of tau phospho-occupancy at pSer202/pThr205 (AT8) were slightly reduced relative to vehicle-treated control cells. A similar reduction in phospho-occupancy was seen when SST14 was replaced with SST28 (Figure 8C, lanes 2-5), the longer physiologically relevant and bioactive version of this peptide, or the dominant preprocortistatin-derived regulatory peptides, CST17 or CST29. The addition of these peptides together with oAβ1-42 did not prevent the anticipated tau hyperphosphosphorylation-inducing effect of oAβ1-42 exposure (Figure 8C, lanes 8-11). Interestingly, the SST- dependent reduction in tau phosphorylation at the AT8 site, which was also observed when SST was subjected to a pre-aggregation step before its addition to primary hippocampal neurons (Figure 8D, lane 2), was not seen in the presence of other identically treated control neuropeptides, including AVP, bombesin and oxytocin (Figure 8D, lanes 3-5). Most remarkably, however, both SST14 and CST17, but not the aforementioned control peptides, were observed to potentiate tau hyperphosphorylation under the latter conditions when Aβ1-42 was concomitantly added in monomeric form to the cell culture medium (Figure 8D, lanes 7 and 8). This observation was notable on several grounds: (i) it indicated that SST14 and CST17 might be equivalent also in this experimental paradigm; and (ii) it revealed a novel way in which tau hyperphosphorylation can be stimulated that, remarkably, was reliant on the co-addition of monomeric, not oligomeric, Aβ1-42 to the cell culture medium.

The objective of a series of subsequent experiments was to better define the experimental conditions that needed to be met for this SST-dependent tau hyperphosphorylation to occur. This line of investigation was pursued because, in contrast to the oAβ1-42-induced tau hyperphosphorylation phenotype, which was robustly observed, SST-dependent tau hyperphosphorylation following co-administration of monomeric Aβ1-42 was not observed every time. Several plausible explanations can be invoked to explain such an experimental behavior, including reliance on experimental paradigms that are governed by threshold-effects of critical parameters or involve a stochastic component. Our current results argue against a stochastic limitation and are instead consistent with the view that once conducive conditions are met, the phenotype can be reliably observed. This view is, for instance, corroborated by one series of experiments, in which SST14 pre-aggregation was allowed to proceed for 24 or 48 hr. This experiment led to a tau hyperphosphorylation phenotype in three independent biological replicates only when the 48 hr pre-aggregated SST14 was added to the cell culture dish (Figure 8E, lanes 5, 8 and 11). As before, this molecular phenotype depended on the co-addition of monomeric Aβ1-42 and was again not observed when monomeric Aβ1-42 was added to the cell culture medium alone (Figure 8E, lanes 3, 6 and 9).

Discussion

The current study was conducted with the intent to produce an in-depth inventory of proteins within the human brain that oAβ1-42 binds to. The study made use of a hypothesis-free discovery approach that capitalized on advanced workflows for the high-pressure nanoflow reversed-phase separation and relative quantitation of peptides, as well as recent improvements to mass spectrometry instrumentation. Taken together, these advances afforded an unprecedented depth of analysis of oAβ1-42 interactors in three independent interactome datasets, each undertaken with three biological replicates of samples and controls. More specifically, these analyses facilitated the direct comparisons of binders to oAβ1-42-baits that were tethered to affinity capture matrices through N- or C-terminal biotin moieties, or had been captured in oligomeric versus monomeric form. From the large amount of data the study generated, several observations stand out: (1) despite its small size, oAβ1-42 baits were observed to bind reproducibly to more than one hundred proteins in all analyses undertaken; (2) in particular, when oAβ1-42 was conjugated to the affinity matrix through an N-terminal biotin moiety, a large number of candidate interactors co-enriched with the bait, corroborating a scenario whereby the hydrophobic C-terminal amino acid sequence stretch within Aβ1-42 is probably mostly protein- or lipid-bound; (3) the small cyclic peptide somatostatin, which has previously been implicated in the etiology of AD but had not been shown to interact with Aβ, was observed to bind directly to oAβ1-42 in several orthogonal biochemical assays; (4) the presence of SST14 was observed to compromise antibody-based detection of Aβ1-42; (5) distinct assemblies of 50-60 kDa, reminiscent of the previously reported Aβ*56, were robustly generated in the presence of SST in reaction mixtures with Aβ1-42 but not with Aβ1-40; and (6) whereas Aβ alone does not induce tau hyperphosphorylation in a primary hippocampal neuron assay, unless it is added to the cell culture medium in a pre-aggregated form, the concomitant exposure of cells to monomeric Aβ and pre-aggregated SST can induce tau hyperphosphorylation at key AD-associated phosphorylation sites.

Several of the oAβ1-42 binders observed in this work, including clusterin (Narayan et al., 2011; Ghiso et al., 1993), 17-β-hydroxysteroid dehydrogenase X (HSD10) (Yan et al., 1997), the ATP synthase complex (Schmidt et al., 2008), GAPDH (Verdier et al., 2005), and the prion protein (Laurén et al., 2009), had previously been proposed to represent Aβ interactors. HSD10, also known as ERAB, 3-hydroxyacyl-CoA dehydrogenase type-2 or Aβ-binding alcohol dehydrogenase (ABAD), initially emerged from a yeast two-hybrid screen as an Aβ binding protein (Yan et al., 1997). Follow-up validation work, which preceded this study, generated evidence that this interaction may contribute to Aβ-mediated mitochondrial toxicity (Lustbader et al., 2004; Yao et al., 2011; Valaasani et al., 2014). The prion protein, which has been described as a major cell surface receptor for oligomeric Aβ (Laurén et al., 2009; Chen et al., 2010), was in this report found to co-enrich with oAβ1-42 tethered to the affinity matrix at its N-terminus but did not interact with C-terminally tethered oAβ1-42. A few candidate oAβ1-42 interactors revealed in this study, including the ubiquitin carboxy-terminal hydrolase L1 (Uch-L1) and the ras-related nuclear protein (RAN), had previously been proposed to play roles in the molecular etiology of AD but the mechanisms of their involvement have remained speculative. More specifically, both Uch-L1 and RAN were reported to be downregulated in AD (Poon et al., 2013; Choi et al., 2004; Mastroeni et al., 2013), and overexpression Uch-L1 was observed to restore synaptic function in hippocampal slices treated with oAβ (Gong et al., 2006). It will be interesting to explore the possible contribution of a physical interaction between oAβ1-42 and these proteins to these observations. Finally, this study revealed many new candidate interactors that had neither been known to interact with Aβ nor had otherwise been tied to the etiology of AD. The significance of these interactions is not known at this time. It is hoped that this inventory of interactors will expedite knowledge regarding the molecular underpinnings of the disease.

In extensive follow-up experiments, we validated the surprising interaction of Aβ1-42 with SST14, also known as somatotropin release-inhibiting factor (SRIF). A conclusion whether CST17 was also present in the initial oAβ1-42 interactome dataset could not be made because, in contrast to the situation for SST14, no peptide was observed that uniquely identified CST17. However, the possibility that binding of oAβ1-42 may extend to CST17 is supported by our observation that, although not yielding qualitatively different results in any of the binding assays, this peptide exerted even more pronounced effects on Aβ aggregation than SST14.

Not only did the oAβ-SST14 interaction stand out as a rare example of an interaction between two relatively small peptides but it also was remarkable in its dependence on a freely accessible N-terminus of Aβ1-42. Interest in binders to the N-terminus is heightened by the recognition that the C-terminus of Aβ1-42 is rarely accessible in vivo due to its hydrophobic nature. Although initially captured at pH 8.0, a widely-used pH for interactome analyses, owed to its wide-spread use in radioimmunoprecipitation assay (RIPA) buffer formulations, the oAβ-SST14 interaction was in subsequent validation experiments observed to influence Aβ-dependent aggregation characteristics in a buffer intended to mimic the physiological environment, and also withstood the presence of a lower pH during FRET analyses. To our knowledge, the smallest protein Aβ had previously been known to bind to is MT-RNR2, also known as humanin, a 21-24 amino acid peptide coded within the mitochondrial 16S ribosomal RNA gene (Hashimoto et al., 2001). Humanin has no resemblance to bioactive SST14 at the amino acid level, has been shown to acquire a three-turn α-helical fold, and interacts with an Aβ binding epitope encompassing residues 17-28 (Maftei et al., 2012), a segment of Aβ not sufficient for binding to SST (Figure 2D). Several previous reports have shown that synthetic macrocycles can assemble into oligomeric assemblies (Levin and Nowick, 2007) characterized by tetrameric assemblies in which β-sheet dimers pair through complementary dry interfaces (Levin and Nowick, 2007; Liu et al., 2011; Pham et al., 2014) and have the ability to serve as building blocks of larger fibril-like supramolecular assemblies (Pham et al., 2014). In the context of their shared predominant release into the extracellular synaptic clefts in the brain, the robust influence of SST14 on Aβ aggregation observed in this work recommends SST14 as a natural Aβ binding partner that could contribute to alternative Aβ conformers and the existence of Aβ strains.

An intriguing data point in this work has been the observation that Aβ1-42, but not Aβ1-40, assembles into distinct complexes of 50-60 kDa in the presence of SST (or CST). Aβ complexes of similar size, most often referred to as Aβ*56, have been observed in transgenic APP mouse models (Lesné et al., 2006) and human AD brains (Lesné et al., 2013), and their levels have been proposed to correlate with cognitive impairment and toxicity. Consistent failures to generate Aβ*56 in vitro have led to the proposition that its formation may require an unknown co-factor (Larson and Lesné, 2012). While the relationship of Aβ*56 to SST-dependent oAβ complexes of similar mass remains unresolved, the ability to generate Aβ assemblies of this size in vitro should aid their structural characterization and facilitate the raising of antibodies with the intent to evaluate their presence in patient samples. Moreover, the ability of SST to prevent the fibrillization of Aβ1-42 in vitro may have implications for the generation of toxic Aβ1-42 oligomers in AD patients. We further documented that the presence of SST profoundly diminished the detection of Aβ by some of the most frequently used antibodies. More work is needed to understand the molecular basis for this finding and gauge the extent to which its existence may have undermined Aβ detection in studies that preceded this work.

SST, originally identified as a factor that inhibits the secretion of pituitary growth hormone (Brazeau et al., 1973), is not a neophyte in the field of AD; a PubMed query conducted with the search terms ‘somatostatin’ and ‘Alzheimer’ returns more than 300 entries, including 60 review articles (e.g., (Epelbaum et al., 2009; Burgos-Ramos et al., 2008). Several observations stand out amongst the connections of SST to AD and should be considered in future efforts to understand the possible significance of the direct interaction between SST and Aβ1-42: (1) Already in 1980, it was observed that SST levels are conspicuously reduced in AD-afflicted brains (Davies et al., 1980), a finding which has been repeatedly confirmed by others (Beal et al., 1985; Gahete et al., 2010; Gabriel et al., 1993). More recently, it was shown that SST gene transcripts are amongst a relatively small number of transcripts whose abundance levels decline persistently throughout adult life (Lu et al., 2004) ; (2) A 1985 study reported a striking co-localization of SST-filled neurons with sites of amyloid plaque formation in AD (Morrison et al., 1985); (3) a manuscript published back-to-back documented that the expression of SST characterized a subclass of neurons in the brain, which were most vulnerable to forming neurofibrillary tangles and undergoing destruction in AD (Roberts et al., 1985); (4) two studies undertaken with Finnish and Chinese patient cohorts reached the conclusion that an SST gene polymorphism may increase the risk for AD (Xue et al., 2009; Vepsäläinen et al., 2007) and, (5) independent investigations revealed that SST can regulate Aβ catabolism by modulating the expression of neprilysin (Saito et al., 2005) and insulin-degrading enzyme (Tundo et al., 2012), two critical players in the best-characterized Aβ clearance pathways. More specifically, Aβ levels were reported to be approximately 1.5-fold higher in SST-deficient mice (Saito et al., 2005), an increase that did not lead to spontaneous plaque formation, thereby leaving the possible involvement of SST in this process currently untested. Once AD models become available in which SST has been eliminated, it will be informative to assess how their Aβ-related pathobiology is altered. It will also be interesting to investigate the constraints of the SST-oAβ interaction in more detail. Our current data suggest that this interaction depends on the presence of tryptophan 8 within SST in the context of flanking residues 5-11 and as such strongly overlaps the epitope within SST known to mediate binding to its cell surface receptors (Veber et al., 1979). Our results further suggest that binding does not occur between monomers of both peptides but may require at least one of the two binding partners to be present in oligomeric (pre-aggregated) form. This limitation is likely to pose technical difficulties for its detailed study but also offers the tantalizing prospect that future insights into the binding interface can be used to design compounds, which may selectively bind to the respective oligomeric forms of Aβ. Such compounds might find application in disease diagnosis, might be used to prevent the SST-Aβ interaction from forming, or could possibly be derivatized in ways that target the respective conformers, and possibly amyloid plaques, for destruction.

Materials and methods

Peptides

Synthetic human Aβ1-42 (catalog number AS-24224), Aβ1-40 (catalog number AS-24236), SST14 (catalog number AS-24277), [Arg8]-Vasopressin (catalog number AS-24289), Bombesin (catalog number AS-20665), Oxytocin (catalog number AS-24275), SST28 (catalog number AS-22902) and amylin1-37 (catalog number AS-60804) were purchased from Anaspec, Inc. (Fremont, CA, USA). The alternative synthetic human SST14 peptide (catalog number H-1490) sourced from a different vendor (see Figure 4B), CST17 (catalog number H-5536) and CST29 (catalog number H-6458) were obtained from Bachem Americas, Inc. (Torrance, CA, USA). The biotinylated Aβ peptides, including biotin-Aβ1-42, Aβ1-42-biotin, Aβ17-42-biotin, biotin-SST14 and truncated or mutant SST peptides were synthesized by LifeTein LLC (Hillsborough, NJ, USA). The Edans-SST donor and Aβ1-42-TMR acceptor peptides were in-house synthesized as described previously (Bateman and Chakrabartty, 2011).

Antibodies

Primary antibodies used in this study were the anti-Aβ1-16 antibody (6E10) (catalog number 803015), the anti-Aβ17-24 antibody (4G8) (catalog number 800701), and the anti-Aβ1-42 antibody (12F4) (catalog number 805501), which were all sourced from BioLegend (San Diego, CA, USA) (note that epitope ranges embedded in vendor names for these antibodies are imprecise, and validated epitopes for the above antibodies are depicted in Figure 5A). The polyclonal antibody recognizing total Tau (K9JA) (catalog number A0024) was a Dako product (Agilent Technologies Canada, Mississauga, ON, Canada). The monoclonal antibodies recognizing Tau phosphorylated at Ser202/ Thr205 (AT8) (catalog number MN1020) or at Thr231 (AT180) (catalog number MN1040) were from Thermo Fisher Scientific (Burlington, ON, Canada).

Western blot and silver staining

Request a detailed protocol

All Western blot reagents were purchased from Thermo Fisher Scientific (Burlington, ON, Canada). For denaturing gels (Figures 3, 5 and 6), peptide samples were mixed with Bolt LDS Sample Buffer (catalog number B0007) in the presence of 2.5% 2-mercaptoethanol and boiled at 70 ˚C for 10 min before loading. The samples were separated on Bolt 12% Bis-Tris Plus gels (catalog number NW00125BOX) in MES SDS Running Buffer (catalog number NP0002) at 100 to 120 V for 1.5 to 2 hr. For native gels (Figure 6B), the peptide samples were mixed with Novex Tris-Glycine Native Sample Buffer (catalog number LC2673) without boiling before loading. The samples were separated on SDS-free 4-20% Tris-Glycine Mini Gels (catalog number XP04202BOX) in Novex Tris-Glycine Native Running Buffer (catalog number LC2672) at 150 V for 1.5 hr. 0.3 to 0.5 µg of Aβ peptide were loaded to each lane for both denaturing and native gels. For the analysis of hippocampal neuron lysates, samples were prepared and separated in the same way as described for denaturing gels except that 4% to 12% Bis-Tris gradient gels were used and 20 to 30 µg of total protein were loaded to each lane. For all immunoblot analyses, peptides were transferred to polyvinylidene difluoride (PVDF) membranes at 50 V in Tris-Glycine buffer containing 10-20% methanol for 1.5 to 2 hr. Membranes were blocked for 2 hr in conventional tris-buffered saline and 0.1% Tween 20 (TBST) containing 5% fat-free milk and probed overnight with the respective primary antibodies. After at least three washes with TBST, membranes were incubated for 2 hr with 1: 2000 to 1: 5000 diluted anti-mouse or anti-rabbit horseradish peroxidase-conjugated secondary antibodies (Bio-Rad Laboratories, Inc., Hercules, CA, USA). The band signals were visualized using enhanced chemiluminescence reagents (catalog number 4500875; GE Health Care Canada, Inc., Mississauga, ON, Canada) and X-ray films. Silver staining of gels was done with reagents from Pierce Silver Stain Kit (catalog number 24612) (Thermo Fisher Scientific, Burlington, ON, Canada) following the protocol provided by the manufacturer.

Preparation of monomeric versus pre-aggregated Aβ, SST and other neuropeptides

Request a detailed protocol

Aβ peptides were dissolved in 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP) at a concentration of 1 mg peptide per mL for one hour at room temperature, then dried in a centrifugal evaporator and stored at −80°C until use. Monomeric Aβ peptides were prepared by dissolving peptides in dimethyl sulfoxide (DMSO) at a concentration of 2 mM and further diluting them in phosphate buffered saline (PBS, pH 7.4) to 100 μM, followed by centrifugation (14,000 g, 20 min) to remove traces of insoluble aggregates. Oligomeric Aβ was created by incubating the monomeric preparation at 4°C for 24 hr. The Aβ oligomers were then purified by centrifugation (14,000 g, 20 min). The supernatant containing Aβ oligomers was further diluted as required and indicated in figure legends describing specific experiments. When high-purity monomeric Aβ peptides at concentrations below 10 μM were required for interactome and ThT assay analyses, the solubilized Aβ preparations were passed through a size-exclusion column (see below for details). Pre-aggregated SST was generated by incubating 1 mM SST in PBS at 37°C for 1 hr with shaking at the speed of 700 rpm. For SST and other neuropeptides prepared for the hippocampal neuron culture assay, the peptides were dissolved in PBS (at pH 4.5, 6.5 or 7.4) at 400 µM without incubation (Figure 8C) or were incubated at room temperature for 24 or 48 hr (Figure 8D and E). 50 μM Aβ1-42 and 40 μM SST/other neuropeptides were mixed (Figure 8D) and incubated for 1 or 3 hr (Figure 8E) before they were added to cell culture medium. The final concentration of Aβ1-42 and SST/other neuropeptides in the medium were 5 and 4 µM, respectively.

Affinity capture of Aβ1-42 and its binding proteins in human frontal lobe extracts

Request a detailed protocol

The biotinylated Aβ oligomers or monomers were captured on Streptavidin UltraLink Resin beads (Thermo Fisher Scientific, Burlington, ON, Canada) by overnight incubation in PBS at 4°C and continuous agitation on a slow-moving turning wheel. Additional negative control samples were generated by saturation of Streptavidin UltraLink Resin (catalog number 53113, Thermo Fisher Scientific, Inc.) with biotin. Subsequently, the bait peptide- or biotin-saturated beads were washed with Lysis Buffer (0.15% digitonin, 150 mM NaCl, 100 mM Tris, pH 8.0). Human frontal lobe tissue samples from individuals (two males and two females) who had died in their early 70s of non-dementia causes served as the biological source material. These samples were adopted from a former Canadian Brain Tissue Bank at the Toronto Western Hospital and are held in −80°C freezers in the biobank of the Tanz Centre for Research in Neurodegenerative Diseases. 1 g pieces each of these brain tissue samples were combined and homogenized in Lysis Buffer supplemented with Complete protease inhibitor cocktail (Roche, Mississauga, ON, Canada). Following the removal of insoluble debris by centrifugation for 30 min at 14,000 g, the protein concentration was adjusted to 2 mg/mL before the brain homogenates were added to the pre-saturated affinity capture beads for overnight incubation at 4°C. Following the affinity capture step, the affinity capture beads (100 μL per biological replicate) were extensively washed in three consecutive wash steps with a total of 150 mL of Lysis Buffer. Subsequently, the beads were additionally washed with 50 mL of 20 mM Hepes, pH 7.0, and transferred to Pierce Spin columns (catalog number 69705, Thermo Fisher Scientific, Inc.) to remove primary amines stemming from the Tris buffer and to prepare the samples for elution. Captured proteins were finally eluted by rapid acidification mediated by a solution comprising 0.2% trifluoroacetic acid and 20% acetonitrile in deionized water (pH 1.9).

Affinity capture of synthetic Aβ1-42 on biotin-SST-saturated capture beads

Request a detailed protocol

N-terminally biotinylated SST14 was captured on Streptavidin UltraLink Resin beads (Thermo Fisher Scientific, Burlington, ON, Canada) at a concentration of 50 µM by overnight incubation in PBS at 4°C with continuous agitation on a turning wheel. Next, the biotin-SST14 saturated streptavidin agarose beads were incubated at room temperature for 4 hr with monomeric or oligomeric Aβ1-42 preparations (2.5 µM), prepared as described in section ‘Preparation of monomeric versus oligomeric Aβ or SST’. Before and after the incubation, the beads were extensively washed in Lysis Buffer (0.15% digitonin, 150 mM NaCl, 100 mM Tris, pH 8.0). Finally, bound peptides were eluted from the affinity capture beads by 15 min boiling in Laemmli sample buffer and analyzed by Western blotting using NuPAGE gels as described in section ‘Western blot and silver staining’.

Size-exclusion chromatography of Aβ1-42

Request a detailed protocol

200 µg of monomeric Aβ1-42 were generated as described in the section ‘Preparation of monomeric versus oligomeric Aβ or SST/other neuropeptides’. To further remove residual traces of Aβ1-42 oligomers, the soluble centrifugation supernatant was subsequently subjected to size-exclusion chromatography on a Superdex 75 10/300 GL column (GE Healthcare Life Sciences, Mississauga, ON, Canada) in ThT assay buffer (20 mM sodium phosphate, pH 8.0, 200 μM EDTA and 0.02% NaN3) at the flow rate of 0.5 mL/min. Late eluting fractions, containing Aβ1-42 monomers, were collected on ice and the Aβ1-42 concentration in these fractions was determined by Western blot analysis, using known amounts of synthetic Aβ1-42 as signal intensity calibrants.

Sample preparation for interactome analyses

Request a detailed protocol

Affinity-capture eluates were essentially processed as described before (Gunawardana et al., 2015; Mehrabian et al., 2014; Jeon and Schmitt-Ulms, 2012). Briefly, sample tubes were moved to a centrifugal evaporator to remove the organic solvent. Additional acidity of the sample was removed following the addition of water and continuous evaporation. Subsequently, protein solutions were denatured by the addition of 9 M urea (to achieve a final concentration of 6 M urea) and 10 min incubation at room temperature. Next, the pH was raised by the addition of 100 mM HEPES, pH 8.0, and proteins were reduced for 30 min at 60°C in the presence of 5 mM tris (2-carboxyethyl) phosphine (TCEP), and alkylated for 1 hr at room temperature in the presence of 10 mM 4-vinylpyiridine (4-VP). To ensure that the residual urea concentration did not exceed 1.5 M, protein mixtures were diluted with 50 mM tetraethylammonium bromide (TEAB), pH 8.0, to a total volume of 100 μL. Samples were then digested with side-chain-modified porcine trypsin (Thermo Fisher Scientific, Burlington, ON, Canada) overnight at 37°C. The covalent modifications of primary amines with isobaric labels provided in the form of tandem mass tag (TMT) reagents (Thermo Fisher Scientific, Inc.) or isobaric tagging for relative and absolute quantitation (iTRAQ) reagents (Applied Biosystems, Foster City, CA, USA) followed instructions provided by the manufacturers. Equal amounts of the labeled digests were pooled into a master mixture and purified with C18 (catalog number A5700310) or SCX (catalog number A5700410) Bond Elut OMIX tips (Agilent Technologies, Inc., Mississauga, ON, Canada) using manufacturer instructions. Peptide mixtures were finally reconstituted in 0.1% formic acid and analyzed by tandem mass spectrometry analysis on a Tribrid Orbitrap Fusion instrument. Instrument parameters during the data acquisition were as described in detail before (Gunawardana et al., 2015).

Post-acquisition data analyses

Request a detailed protocol

The post-acquisition data analyses of interactome data sets was conducted against the human international protein index (IPI) database (Version 3.87) which was queried with Mascot (Version 2.4; Matrix Science Ltd, London, UK) and Sequest HT search engines within Proteome Discoverer software (Version 1.4; Thermo Fisher Scientific, Burlington, ON, Canada). Spectra exceeding a stringent false discovery rate (FDR) target of ∆Cn of 0.05 for input data and a FDR of 0.01 for the decoy database search were detected and removed by the Percolator algorithm (Käll et al., 2007) as described before (Gunawardana et al., 2015). PEAKS Studio software (Version 6.0; Bioinformatics Solutions Inc., Waterloo, ON, Canada) was used to assess the reproducibility of nano-HPLC separations. A maximum of two missed tryptic cleavages and naturally occurring variable phosphorylations of serines, theonines and tyrosines were considered. Other posttranslational modifications considered were carbamylations, oxidation of methionines and deamidation of glutamines or asparagines. Mass spectrometry data sets have been deposited to the ProteomeXchange Consortium (Vizcaíno et al., 2014) via the PRIDE partner repository (Vizcaíno et al., 2013) with the dataset identifier PXD004867 and have been made fully accessible.

FRET measurements

Request a detailed protocol

All fluorescence spectra were acquired with a Photon Technology International fluorescence spectrophotometer (QuantaMaster, HORIBA Scientific, London, ON, Canada) using a 1 cm quartz cuvette with an excitation wavelength of 335 nm and a 2 nm slit width at room temperature. Solutions containing 20 μM concentrations of Edans-SST14, Aβ1-42-TMR or a mixture of both fluorescently labeled peptides at pH 8.5 were rapidly adjusted to pH 5.2 and the resulting reaction mixtures incubated overnight. Fluorescence emission spectra were recorded in a continuous window spanning 350 nm to 650 nm.

FRET competition assay

Request a detailed protocol

FRET competition assays were performed using the fluorescence spectrophotometer described above. A solution of 2 μM Aβ1-42-TMR was mixed with increasing concentrations of SST14 or AVP (0-50 μM) at pH 8.5 followed by dropping the pH to 5.2 and incubating the resulting solutions overnight. To compete for binding to Aβ1-42-TMR, a solution of 2 μM Edans-SST was then added to each of the solutions at a final volume of 150 μL at continuous incubation of peptide mixtures at pH 5.2 overnight. Fluorescence spectra were recorded and binding curves were plotted against the concentration of unlabeled peptides. The curves were fitted with assistance of OriginPro 8.5 software (OriginLab Corporation, Northampton, MA) using nonlinear least squares fitting given by the equation:

y=(ymaxymin/(1+(x/EC50)n)+ymin

with y = observed donor fluorescence, x = concentration, and n = Hill coefficient.

Kinetic aggregation assay

Request a detailed protocol

Assay procedures were largely based on a protocol described by Sarah Linse’s group (Hellstrand et al., 2010). Briefly, Aβ1-42, Aβ1-40 and/or SST14 were prepared in assay buffer (20 mM sodium phosphate, pH 8.0, 200 μM EDTA and 0.02% NaN3) or in PBS, pH 7.4, at concentrations specified in the individual figures. Human amylin1-37 was initially dissolved in deionized water but then assayed under identical conditions as Aβ1-42. 100 μL of peptide solutions were supplemented with 25 μM thioflavin T (ThT) (catalog number T3516, Sigma-Aldrich Canada, Oakville, ON, Canada) and loaded into 96-Well Half-Area Microplates (catalog number 675096, Greiner Bio One International, Kremsmünster, Austria). The subsequent plate incubation proceeded at 37°C with shaking at 700 rpm for 4 of every 5 min in a microplate reader (CLARIOstar, BMG Labtech, Guelph, ON, Canada) for overall durations specified in individual figures. ThT fluorescence was measured every 5 min at excitation and emission wavelengths of 444 nm and 485 nm, respectively.

Particle sizing

Request a detailed protocol

Particle size was analyzed using a NanoSight NS300 (Malvern Instruments, Southborough, MA, USA). Samples were vortexed for 40 s before being loaded into a luer lock syringe and infused at a constant rate through the sample chamber using an attached syringe pump (Malvern). Particle movement was captured for 60 s x 5 acquisitions at a rate of 25 frames per second, at room temperature. Data were analyzed by nanoparticle tracking analysis (NTA 3.2 software, Malvern) with detection threshold held constant between replicates. Polystyrene latex microsphere beads of 100 nm (Malvern) were used as a size standard. Data are shown as mean diameter ± standard deviation.

Negative stain electron microscopy

Request a detailed protocol

Five-microliter samples were adsorbed for 1 min onto freshly glow-discharged Formvar-carbon-coated 400-mesh copper grids. The grids were washed with 0.1 M and 0.01 M ammonium acetate buffer (pH 7.4), stained with freshly filtered 1% uranyl acetate solution, and excess stain removed with filter paper. The grids were allowed to dry overnight and then viewed with a Tecnai F20 electron microscope (FEI Company / Thermo Fisher Scientific, Eindhoven, The Netherlands) at an acceleration voltage of 200 kV. Electron micrographs were recorded on an Eagle 4K charge-coupled device camera (FEI Company / Thermo Fisher Scientific, Eindhoven, The Netherlands).

Primary hippocampal neuron assay

Request a detailed protocol

Primary hippocampal cultures were generated from E16 ~18 C57BL/6 mouse embryos. The hippocampi were dissected while immersed in Hank’s Balanced Salt Solution buffered with 10 mM HEPES and cells were dissociated with 0.125% trypsin (Invitrogen Canada, Inc., Burlington, ON, Canada) for 10 ~ 15 min at 37°C, followed by trituration. Dissociated cells were plated at a density of 1.0 ~ 1.5 × 105 cells/cm2 in 6-well plates pre-coated with poly-D lysine and grown in Neurobasal medium with B-27 supplement and Glutamax (Invitrogen). Half the medium was refreshed every 3 days. The treatment of the primary hippocampal neurons with Aβ and other peptides was started on day 18. After treatment, cells were lysed in 0.5% SDS, 1% Nonidet P-40, 2 mM EDTA, 100 mM NaCl, 100 mM Tris, pH 7.6, supplemented with protease inhibitor and phosphatase inhibitor cocktails (Roche, Mississauga, ON, Canada) for Western blot analyses.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
    Microtubule-associated protein tau. A component of alzheimer paired helical filaments
    1. I Grundke-Iqbal
    2. K Iqbal
    3. M Quinlan
    4. YC Tung
    5. MS Zaidi
    6. HM Wisniewski
    (1986b)
    The Journal of Biological Chemistry 261:6084–6089.
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
    Alzheimer's disease: genes, proteins, and therapy
    1. DJ Selkoe
    (2001)
    Physiological Reviews 81:741–766.
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
  80. 80
  81. 81
  82. 82
  83. 83
  84. 84

Decision letter

  1. Randy Schekman
    Reviewing Editor; Howard Hughes Medical Institute, University of California, Berkeley, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

[Editors’ note: a previous version of this study was rejected after peer review and an appeal, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.]

Thank you for submitting your work entitled "The human Aβ interactome: binding to somatostatin favors formation of distinct oligomers and impairs detection of Aβ1-42" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by a Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Colin Masters (Reviewer #2).

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

Although the manuscript provides potentially useful information about Aβ interactors, after a thorough discussion among the editors and reviewers, we are particularly concerned that the physiologic basis of an effect of somatostatin14 on the aggregation of Aβ(1-42) has not been addressed in cells or in vivo. Further, the evidence for a role of somatistatin14 in the aggregation of Aβ(1-42) is not compelling.

Summarizing other issues that you will see in the full reviews appended below:

One implication of this work is that Aβ(1-42) can be found in the cytoplasm of nerve cells. What evidence is there that this is so? Are its termini free? What was the difference here between monomer and oligomer? Were there any fibrils? Did the authors use EM and aSEC? How do these structures relate to the recent papers by Wälti et al. and Colvin et al.?

The focus on somatostatin is explained, but this section is far too long. The authors must also discuss the effects of corticostatin on Aβ aggregation, given that this hormone was not identified as a positive interactor.

The issue of n numbers must be addressed. The number must be at least three, meaning extracts from three different brains. Much more detail must be given in relation to tissue preparation.

The authors must explain why the capture reaction was done at pH 8.0. How did this pH influence the interactors detected?

The denaturation and trypsinization steps do not seem to figure in Materials. Quite often, more details are needed, so that one could attempt to replicate these findings. The authors must go through the manuscript with this in mind. For instance, in subsection “Binding of SST14 or CST17 precludes detection of Aβ1-42 with commonly used antibodies”, much more detail is required about Western blotting using Aβ antibodies. At times, the authors refer to TMR-Aβ, at other times to Aβ-TMR. Does this mean the TMR was at varying ends?

Reviewer #1:

This is a comprehensive proteomic study. It provides useful background information, even though some of the interpretations might be open to question. The authors identified proteins within the human brain that monomeric and oligomeric Aβ(1-42) bind to. This work implies that oligomeric Aβ(1-42) is found in the cytoplasm. What is the evidence for this? How do the structures of monomeric and oligomeric Aβ(1-42) differ and how does this relate to the structures reported by Colvin et al. and Wälti et al.? What do the monomers and oligomers used here look like by EM? How do the authors know that there were no fibrils? Somatostatin was found to bind to the N-terminal region of Aβ(1-42) oligomers, but it is not listed in Table 1. The discussion relating to somatostatin was much too long. Overall, this manuscript is sophisticated in some parts, but it is rather unsophisticated in others. I could not understand what the n numbers were. Did the authors only use one frontal cortex (age of patient, post mortem delay, control or AD?). If so, the n number was one (which would be insufficient). Is it known if the post mortem delay could influence these interactions?

Reviewer #2:

This is a reasonable proteomic approach but the results are difficult interpret because of the experimental conditions used. Some would argue that monomeric Aβ does not occur in vivo, and that the c-terminus is rarely available for protein-protein interactions given its hydrophobic nature. Nevertheless the data are of use in future studies when other putative interactions are being evaluated. The focus on SST14 is a little surprising given that it did not come up in the primary screen.

Reviewer #3:

This is an interesting manuscript that describes a quantitative mass spec approach to investigate the Aβ interactome. Orthogonal approaches using N-terminal and C-terminal tags, and aggregated versus monomeric Aβ are appreciated and confirm several proteins previously implicated as Aβ binders. The data in Figures 1 and 2 and associated tables will be of interest to many involved in Aβ research. By relaxing the requirement that peptides had to be at least six amino acids long to be considered for protein identification, the sequence "NFFWK" was identified. The fact that this sequence resides not simply in the preprosomatostatin and preprocortistatin, but also the mature hormones (SST14 and CST17) together with evidence of other SST14 sequence (TFTSC) justifies their focus on SST14. In Figure 3B the authors present evidence that SST14 can bind Aβ aggregates in vitro, and in Figure 4 they demonstrate that SST14 and CST17 can inhibit Aβ aggregation. However, CST17 is a much better inhibitor of Aβ aggregation, but given that this hormone was not identified as a positive interactor the relevance of this finding is unclear. The approach (SDS-PAGE and WB) used in Figures 5 and 6 is weak and the conclusions drawn are shaky.

In terms of the various subsections of the manuscript, the Introduction was a joy to read, but the Results section was somewhat contorted and the Materials and methods lacked many details. Indeed, certain parts of the Materials and methods are incomplete rendering it impossible for independent replication. The Discussion contained some strong points, but was wandering and highly speculative. In sum, this manuscript contains several interesting and innovative points, but it is not suitable for publication in its current form.

A list of specific points documenting errors/inconsistencies and omissions is provided.

1) There are no details on how the brain extract was prepared. This is an important issue as the method used will influence the proteome in the extract. Similarly, there are no details regarding the donor(s) – i.e. cause of death, gender, age, etc. While the authors refer to 3 biological replicates, it is unclear if these means 3 extracts from the same brain, or extracts from 3 different brains. Clearly a proteomic analysis based on extracts from a single human brain would need validate using other human brains.

2) While the FRET experiments in Figure 3 are consistent with other data that support SST binding to Aβ aggregates, the experimental design of the experiments in Figure 3 preclude interpretation since they were done at pH close to the isoelectric point of Aβ. The apparent FRET and competition could result due to co-precipitation and formation of amorphous aggregates.

3) It is not clear why the capture reaction was done at pH 8.0, nor is there consideration of how this mildly alkaline pH may influence the interactors detected.

4) Subsection “Affinity capture of Aβ1-42 and its binding proteins in human frontal lobe extracts”: "the beads were additionally washed with 50 mL of 20 mM Hepes, pH 7.0, and transferred to Pierce Spin columns (catalog number 69705, Thermo Fisher Scientific, Inc.) to remove primary amines and prepare the samples for elution" – how were primary amines removed? Does this refer to depletion of the Tris in the lysis buffer?

5) Subsection “Workflow of Aβ interactome analyses”: "binders to the bait peptides were eluted by rapid acidification, fully denatured in 9 M urea, and trypsinized" – there are no details of the denaturation and trypsinization steps in the Results or Materials and methods.

6) Subsection “The Aβ1-42 interactome”: "Not surprisingly, the Aβ peptide itself was amongst the proteins identified whose enrichment levels in specific versus unspecific affinity capture samples were most pronounced" and Supplementary Table 1. Were the tryptic fragments detected restricted to Aβ, or as suggested on Supplementary Table 1, did they also include other regions of APP?

7) Subsection “Validation of SST14 binding to oAβ1-42”: "For this experiment, oAβ1-42 preparations (generated by incubating mAβ1-42 purified by gel filtration at 37 ºC for 2 h with shaking at the speed of 700 rpm)" – further details are required. Moreover, the material used should be characterized by EM and aSEC as it seems likely that the material used are not oligomers, but short fibrils. SDS-PAGE (Figure 3A) is not a suitable means to determine native assembly size.

8) Subsection “FRET measurements”: "Solutions containing 20 μM concentrations of Edans-SST14, TMR-Aβ1-42 or a mixture of both fluorescently labeled peptides at pH 8.5 were rapidly adjusted to pH 5.2 and the resulting reaction mixtures incubated overnight". pH 5.2 is essentially the isoelectric point for Aβ, making interpretation of the presented result uncertain. TMR is hydrophobic and has a tendency to aggregate in aqueous solutions under conditions where the labeling density is sufficient to permit dye-dye interactions. A consequence of these interactions is fluorescence self-quenching.

9) At times the authors refer to TMR- Aβ, and at other times Aβ -TMR, they should clarify whether the TMR was on the N or C-terminus.

10) Subsection “Binding of SST14 or CST17 precludes detection of Aβ1-42 with commonly used antibodies”: "samples prepared as described in the previous section (but under omission of ThT) were analyzed by Western blotting using antibodies targeting N-terminal, central or C-terminal epitopes within Aβ1-42" – more details are necessary.

11) Figure 5 – the Ag stain data also show a decrease in the amount of peptide detected in the presence of SST14 or CST17. Moreover, truncation of the Ag stained gels at the 6 kDa marker makes it difficult to compare the Ag stained and WB samples.

12) Subsection “Aβ1-42 forms a distinct 50-60 kDa SDS-stable complex in the presence of SST14”: "To compensate for the abovementioned epitope masking effect of SST14 or CST17 (Figure 5C), these experiments were undertaken with 5 to 10 times higher peptide concentrations than those described above (Figure 4 and 5)." Why distort the experiment by using such high concentrations, why not simply use Ag staining since the authors claim that Ag stain was not influenced by the presence of SST14 or CST17.

13) Figure 6A, lanes 1-4 are labeled "no SDS" – is this correct?

[Editors’ note: what now follows is the decision letter after the authors’ appeal.]

I apologize for the delay at all stages in the review of your work and I take responsibility for the decision to decline publication of this work. Because of our editorial process and the volume of submissions we receive (over 8000 last year), I do not become involved until the very end and then only in cases where the manuscript appears to be moving toward acceptance.

In the case of your submission, you are correct that the reviews contained suggestions that would have been possible for you to address, but on reading the comments of the referees in their individual critiques an in the ensuing consultation session, I was struck by the lack of any biological or even functional biochemical validation of your conclusion concerning the interaction of somatostatin and the Aβ peptide. I then conferred with two other eLife neuroscience Senior Editors who agreed that your results were simply too preliminary for one to conclude that the interaction has meaning in relation to Alzheimer's Disease. And in further consultation with the reviewers, two expressed concerns with the strength of your data regarding on influence of somatostatin on the aggregation of Aβ. However, they also were not in favor of demanding animal studies, which may be of dubious value given the inadequate animal models of Aβ aggregation and the development of dementia. (e.g. reviewer #3 - In my opinion, studies to show that the interaction between Aβ and SST14 has physiological or pathophysiologic significance is beyond the scope of this paper. Rather it is more important to understand if the experimental conditions in Figures 3 and 4 really justify the conclusions. With regard to SST14's effect on Aβ aggregation, this is the least compelling part of the manuscript. I consider Figures 5 & 6 to be weak.)

So we are left with an unsatisfying conclusion that the solid proteomic data you have developed may or may not have any bearing on the physiologic role of somatostatin in the fate of extracellular Aβ peptide or oligomers. As a result, I felt we had no choice but to decline this work. However, if as you suggest it may be possible to evaluate the effect of somatostatin secreted by cells on the activity of Aβ monomer or oligomer in the phosphorylation of tau, then this at least could lead to a more plausible if not conclusive connection to disease. If you are willing to go one more round with us, I suggest you develop such an analysis and send it back to us for review by the same individuals who read the current version of your paper.

Sincerely, Randy Schekman

Editor-in-Chief

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for submitting your article "The human amyloid β peptide interactome: binding to somatostatin favors formation of distinct oligomers" for consideration by eLife. Your resubmission has been evaluated by a Senior Editor in consultation with the original reviewers. The editors would be happy to consider your work further once you have responded to new concerns raised by one of the original reviewers.

We remain somewhat conflicted about this submission. It has some strengths and the new data are a plus, but the physiological or more importantly pathophysiological role of the Aβ -SST interaction is still unclear.

The new data don't directly address the request I made in our exchange in your appeal letter:

"However, if as you suggest it may be possible to evaluate the effect of somatostatin secreted by cells on the activity of Aβ monomer or oligomer in the phosphorylation of tau, then this at least could lead to a more plausible if not conclusive connection to disease."

Instead, you used synthetic SST and synthetic Aβ. However, the new data in Figure 8 do indicate that co-addition of SST with Aβ monomer to primary neurons leads to increased phosphorylation of tau. So the authors have gone some way towards the request. Please comment on your use of synthetic peptides rather than those interacting in the medium of cultured cells.

The new data in Figure 7 demonstrate that Trp8 in SST is essential for the inhibition of Aβ aggregation, and that SST causes production of relatively small aggregates, this certainly bolsters the specificity and effect of SST, but this was considered a minor point.

https://doi.org/10.7554/eLife.28401.021

Author response

[Editors’ note: the author responses to the first round of peer review follow.]

Summarizing other issues that you will see in the full reviews appended below:

One implication of this work is that Aβ(1-42) can be found in the cytoplasm of nerve cells. What evidence is there that this is so?

While there are in our view compelling prior reports by others that documented the presence of Aβ in mitochondria and the cytosol, including by super‐resolution imaging work [2], this is not something we emphasized in our report. Our choice to work with digitonin‐solubilized brain extracts during the discovery phase allowed us to keep an open mind in this regard, as these would include extracellular and most cellular proteins (except for nuclear proteins), consistent with the main subcellular areas previously reported to harbor Aβ. We then confirmed that Aβ can interact with a subset of previously reported intracellular Aβ interactors when they are offered in a brain extract (e.g., ATP synthase or ABAD). Importantly, for somatostatin, the most striking interactor revealed by our analysis, cytosolic uptake is not required for its Aβ interaction to occur, since both somatostatin and Aβ are secreted at the synapse.

Are its termini free? What was the difference here between monomer and oligomer?

We had tethered Aβ to affinity capture matrices using N‐terminal or C‐terminal biotin tags, which were coupled through a short linker (Experiments 1 and 2). We thereby generated separate interactome datasets for Aβ with free C‐ or N‐termini, which we reported in detail in the data tables. How oligomers were prepared was also described in our original submission.

Were there any fibrils? Did the authors use EM and aSEC? How do these structures relate to the recent papers by Wälti et al. and Colvin et al.?

Soluble oligomeric Aβ used for affinity capture were prepared using a previously published method (Goate et al., 1991 and Rogaev et al., 1985) known to generate amyloid‐β‐derived diffusible ligands (ADDLs). EM images characterizing these preparations have been published by others before but could also be included with a resubmission of our paper. The reviewer is correct in asserting that these included both oligomeric and prefibrillar Aβ. Larger aggregates would have sedimented during a 14,000 g centrifugation step, which we had applied prior to the capture of the Aβ bait.

The focus on somatostatin is explained, but this section is far too long.

This could easily be truncated.

The authors must also discuss the effects of corticostatin on Aβ aggregation, given that this hormone was not identified as a positive interactor.

One peptide we observed (NFFWK) could have originated from both cyclic hormones. A conclusion whether cortistatin was missing in the interactome dataset cannot be made. This is because, in contrast to somatostatin, no peptide was observed that uniquely identifies cortistatin. This observation could reflect one of several things, including the fact that the expression of cortistatin has been reported to be less abundant and more restricted in the brain, or could merely represent inadvertent ionization properties of the one tryptic peptide that is longer than 3 amino acids, which could uniquely identify cortistatin. Given that one peptide (TFTSC) uniquely identified somatostatin, it made sense to focus more on somatostatin during validation and in our discussion. The reviewer is correct in asserting that the effect of cortistatin on aggregation was even more pronounced than that observed for somatostatin. However, we would like to note that cortistatin did not behave in a qualitatively different manner from somatostatin in any of the assays.

Since November we have generated additional somatostatin‐Aβ interface mapping data, which inform about the minimum epitope required for binding and this could be included with a revised submission.

The issue of n numbers must be addressed. The number must be at least three, meaning extracts from three different brains. Much more detail must be given in relation to tissue preparation.

We are ourselves routinely frustrated by papers which report on interactome data that do not pass stringent conventions in the field. Our work is not one of them and describes three independent datasets, which together corroborated the conclusion that somatostatin was the most selective binder of oligomeric Aβ based on several peptide‐to‐spectrum matches and several biological and technical replicates (Experiments 1‐3). We did not anticipate that the robustness of the interactome data would be contentious and therefore did not mention that, in addition, we had previously generated a completely separate but almost identical Aβ interactome dataset. Samples in both studies were independently sourced, mass spectrometry analyses were undertaken on fundamentally different instruments (orbitrap versus triple time‐ of‐flight), one study was based on iTRAQ labeling, the other on TMT labels. Each time, brain extracts represented 1:1:1 (w/w/w) mixtures of frontal lobes from three elderly individuals that had died of non‐dementia causes. Each time, the NFFWK peptide derived from cortistatin/somatostatin was identified as the tryptic peptide that most selectively co‐purified with oligomeric Aβ but not monomeric Aβ. The datasets we included in the paper were produced on the cutting‐edge Orbitrap Fusion Tribrid mass spectrometer. If it was deemed helpful, we would be happy to also present the first complete dataset produced earlier on an AB Sciex 5600 machine when resubmitting our paper.

The authors must explain why the capture reaction was done at pH 8.0. How did this pH influence the interactors detected?

This pH may be the most often employed pH for interactome analyses, because it is applied in the widely used RIPA buffer formulation. Note that subsequent validation experiments were undertaken at physiological pH (ThT aggregation assays and gel based analyses) or lower pH (FRET analyses), demonstrating that the somatostatin‐Aβ interaction is tolerant of changes to pH.

The denaturation and trypsinization steps do not seem to figure in Materials. Quite often, more details are needed, so that one could attempt to replicate these findings. The authors must go through the manuscript with this in mind. For instance, in subsection “Binding of SST14 or CST17 precludes detection of Aβ1-42 with commonly used antibodies”, much more detail is required about Western blotting using Aβ antibodies.

We are cognizant of the frustrations that come with sloppily assembled Materials and methods sections. We are confident that you would have to agree that the respective section in our paper was assembled with care. Having stated this, we agree that it can be further improved and will be happy to add the requested details to clarify procedures further.

At times, the authors refer to TMR-Aβ, at other times to Aβ-TMR. Does this mean the TMR was at varying ends?

We apologize for this inaccuracy. The same reagent TMR‐derivative of Aβ was used in all experiments.

Reviewer #1:

This is a comprehensive proteomic study. It provides useful background information, even though some of the interpretations might be open to question. The authors identified proteins within the human brain that monomeric and oligomeric Aβ (1-42) bind to. This work implies that oligomeric Aβ (1-42) is found in the cytoplasm. What is the evidence for this? How do the structures of monomeric and oligomeric Aβ (1-42) differ and how does this relate to the structures reported by Colvin et al. and Wälti et al.? What do the monomers and oligomers used here look like by EM?

As mentioned above, this information could be included with a resubmission.

How do the authors know that there were no fibrils?

Only in so far as the Aβ we conjugated was soluble after 15 minute centrifugation at 14,000 g. The preparation method we used is expected to produce a highly heterogeneous ADDL mixture. This was intentional, as we did not want to bias the capture by focusing on a particular oligomer (e.g., those described by Ahmed et al., 2010 or Barghorn et al., 2005).

Somatostatin was found to bind to the N-terminal region of Aβ (1-42) oligomers, but it is not listed in Table 1.

Although somatostatin was most selectively co‐purifying with Aβ, it did not pass stringency criteria applied for proteins listed in Table 1, namely that identifications be based on a minimum of three peptides and that the lengths of those peptides exceed five amino acids. This point was included in the text but understandably can be missed and will be further emphasized by including a footnote to Table 1 in a revised document.

The discussion relating to somatostatin was much too long.

We will truncate this section.

Overall, this manuscript is sophisticated in some parts, but it is rather unsophisticated in others. I could not understand what the n numbers were. Did the authors only use one frontal cortex (age of patient, post mortem delay, control or AD?). If so, the n number was one (which would be insufficient). Is it known if the post mortem delay could influence these interactions?

See detailed response to this point above.

Reviewer #2:

This is a reasonable proteomic approach but the results are difficult interpret because of the experimental conditions used. Some would argue that monomeric Aβ does not occur in vivo, and that the c-terminus is rarely available for protein-protein interactions given its hydrophobic nature.

We agree with this comment. Note that we describe as a strict requirement for somatostatin binding to oligomeric Aβ that the latter is tethered through its C‐terminus and, therefore, has an accessible N‐terminus.

Nevertheless the data are of use in future studies when other putative interactions are being evaluated. The focus on SST14 is a little surprising given that it did not come up in the primary screen.

Not only did cortistatin/somatostatin come up in the screen but it was the most selective Aβ binder (based on the relative quantitation afforded by iTRAQ signature ions, when comparing samples and controls), once we waived the requirement that proteins had to be identified on the basis of three unique peptides of a minimum length of six amino acids. It therefore became the natural interactor to pursue further in validation work. This point will be further emphasized in a revised submission.

Reviewer #3:

This is an interesting manuscript that describes a quantitative mass spec approach to investigate the Aβ interactome. Orthogonal approaches using N-terminal and C-terminal tags, and aggregated versus monomeric Aβ are appreciated and confirm several proteins previously implicated as Aβ binders. The data in Figures 1 and 2 and associated tables will be of interest to many involved in Aβ research. By relaxing the requirement that peptides had to be at least six amino acids long to be considered for protein identification, the sequence "NFFWK" was identified. The fact that this sequence resides not simply in the preprosomatostatin and preprocortistatin, but also the mature hormones (SST14 and CST17) together with evidence of other SST14 sequence (TFTSC) justifies their focus on SST14. In Figure 3B the authors present evidence that SST14 can bind Aβ aggregates in vitro, and in Figure 4 they demonstrate that SST14 and CST17 can inhibit Aβ aggregation. However, CST17 is a much better inhibitor of Aβ aggregation, but given that this hormone was not identified as a positive interactor the relevance of this finding is unclear. The approach (SDS-PAGE and WB) used in Figures 5 and 6 is weak and the conclusions drawn are shaky.

These data were included because, in our view, in particular, the heterodimer bands shown in Figure 5 represented a powerful validation of the direct interaction between Aβ and SST. We also found it was worth sharing that oligomeric complexes observed in the presence of SST migrated at apparent molecular masses of 55‐65 kDa, which could only be revealed by SDS‐free gel‐based analyses. Finally, the gel‐based analyses documented that these somatostatin‐ dependent complexes can be observed with Aβ1‐42 but not Aβ1‐40, which would not have been apparent had we only shown the ThT assay data, in which both Aβ peptides behaved similarly

In terms of the various subsections of the manuscript, the Introduction was a joy to read, but the Results section was somewhat contorted and the Materials and methods lacked many details. Indeed, certain parts of the Materials and methods are incomplete rendering it impossible for independent replication.

We can make sure that a revised submission will not lack in this area.

The Discussion contained some strong points, but was wandering and highly speculative.

We will remove speculative points for a resubmission.

In sum, this manuscript contains several interesting and innovative points, but it is not suitable for publication in its current form.

A list of specific points documenting errors/inconsistencies and omissions is provided.

1) There are no details on how the brain extract was prepared. This is an important issue as the method used will influence the proteome in the extract. Similarly, there are no details regarding the donor(s) – i.e. cause of death, gender, age, etc. While the authors refer to 3 biological replicates, it is unclear if these means 3 extracts from the same brain, or extracts from 3 different brains. Clearly a proteomic analysis based on extracts from a single human brain would need validate using other human brains.

We apologize for this omission and regret to have left this point unclear. As mentioned in response to the editorial summary above, the study results were NOT based on an extract generated from a single human brain but represented highly robust data based on three completely separate interactome experiments (Experiments I ‐ III) undertaken with mixtures of frontal brain samples from several elderly individuals who died of non‐dementia causes. Note that due to the complexity of the human brain, these extracts were based on mixtures of myriads of different brain cells. As briefly mentioned above, the investigation of binders to oligomeric and monomeric Aβ was undertaken completely independently twice, each time with several biological replicates and controls (and we had decided to not include the first dataset undertaken as it did not add anything of substance and seemed excessive). We will improve on this aspect of the description and list the specifics of the frontal lobe brain samples used in the revised submission.

2) While the FRET experiments in Figure 3 are consistent with other data that support SST binding to Aβ aggregates, the experimental design of the experiments in Figure 3 preclude interpretation since they were done at pH close to the isoelectric point of Aβ. The apparent FRET and competition could result due to co-precipitation and formation of amorphous aggregates.

We had considered this point. We take it that the reviewer would agree that if quenching was merely due to unspecific formation of amorphous co‐precipitation of Aβ near its physiological pH (Figure 3C), we would not have observed the acceptor sensitization at 596 nm and it would have been unlikely for the addition of unlabeled somatostatin, but not unlabeled AVP, to have rescued this quenching effect (see Figure 3D).

3) It is not clear why the capture reaction was done at pH 8.0, nor is there consideration of how this mildly alkaline pH may influence the interactors detected.

Most co‐immunoprecipitation or affinity capture interactome analyses are undertaken at this pH due to the popularity of RIPA buffer in these applications.

4) Subsection “Affinity capture of Aβ1-42 and its binding proteins in human frontal lobe extracts”: "the beads were additionally washed with 50 mL of 20 mM Hepes, pH 7.0, and transferred to Pierce Spin columns (catalog number 69705, Thermo Fisher Scientific, Inc.) to remove primary amines and prepare the samples for elution" – how were primary amines removed? Does this refer to depletion of the Tris in the lysis buffer?

Yes, the reviewer is correct that the presence of Tris would preclude labeling with the isobaric labels we routinely use. We will make this clear in a revised submission.

5) Subsection “Workflow of Aβ interactome analyses”: "binders to the bait peptides were eluted by rapid acidification, fully denatured in 9 M urea, and trypsinized" – there are no details of the denaturation and trypsinization steps in the Results or Materials and methods.

We have used and published these procedures many times before (including in detailed Methods in Molecular Biology protocols) and usually refer to them by referencing prior articles. We will ensure to provide the detail requested in a resubmission.

6) Subsection “The Aβ1-42 interactome”: "Not surprisingly, the Aβ peptide itself was amongst the proteins identified whose enrichment levels in specific versus unspecific affinity capture samples were most pronounced" and Supplementary Table 1. Were the tryptic fragments detected restricted to Aβ, or as suggested on Supplementary Table 1, did they also include other regions of APP?

Although APP peptides that map to regions outside of Aβ were ‘detected’ by the algorithm, their assignments were low scoring (95% significance) and therefore not credible. In contrast, and not surprisingly, tryptic peptides derived from Aβ were identified with highest confidence and exhibited the relative TMT signature ion quantitation profiles expected from the design of the experiments.

7) Subsection “Validation of SST14 binding to oAβ1-42”: "For this experiment, oAβ1-42 preparations (generated by incubating mAβ1-42 purified by gel filtration at 37 ºC for 2 h with shaking at the speed of 700 rpm)" – further details are required. Moreover, the material used should be characterized by EM and aSEC as it seems likely that the material used are not oligomers, but short fibrils. SDS-PAGE (Figure 3A) is not a suitable means to determine native assembly size.

EM data can be provided for these preparations together with a resubmission.

8) Subsection “FRET measurements”: "Solutions containing 20 μM concentrations of Edans-SST14, TMR-Aβ1-42 or a mixture of both fluorescently labeled peptides at pH 8.5 were rapidly adjusted to pH 5.2 and the resulting reaction mixtures incubated overnight". pH 5.2 is essentially the isoelectric point for, making interpretation of the presented result uncertain. TMR is hydrophobic and has a tendency to aggregate in aqueous solutions under conditions where the labeling density is sufficient to permit dye-dye interactions. A consequence of these interactions is fluorescence self-quenching.

The reviewer is correct and we are keenly aware of this confounder. We will be able to convincingly exclude this possibility.

9) At times the authors refer to TMR-, and at other times-TMR, they should clarify whether the TMR was on the N or C-terminus.

See above. This inconsistency in the data presentation in Figure 3 versus the Materials and methods section will be corrected.

10) Subsection “Binding of SST14 or CST17 precludes detection of Aβ1-42 with commonly used antibodies”: "samples prepared as described in the previous section (but under omission of ThT) were analyzed by Western blotting using antibodies targeting N-terminal, central or C-terminal epitopes within Aβ1-42" – more details are necessary.

We will add a more detailed description of how these samples were following ThT assay analysis (but without ThT added) dissolved in sample buffer and analyzed by gel electrophoresis.

11) Figure 5 – the Ag stain data also show a decrease in the amount of peptide detected in the presence of SST14 or CST17. Moreover, truncation of the Ag stained gels at the 6 kDa marker makes it difficult to compare the Ag stained and WB samples.

The area above 6 kDa did not reveal sample‐specific bands but was characterized by signals present in all samples at the same levels that are routinely observed in highly exposed silver stains. Because these signals were not derived from somatostatin, cortistatin or Aβ (as their intensities did not correlate at all with the differences in the amounts of these peptides in the samples) they did not aid in interpreting the binding biology that was the focus of the experiment. We therefore chose to cut them from the presentation but will present the full silver‐stained blot as supplemental information with the resubmission.

12) Subsection “Aβ1-42 forms a distinct 50-60 kDa SDS-stable complex in the presence of SST14”: "To compensate for the abovementioned epitope masking effect of SST14 or CST17 (Figure 5C), these experiments were undertaken with 5 to 10 times higher peptide concentrations than those described above (Figure 4 and 5)." Why distort the experiment by using such high concentrations, why not simply use Ag staining since the authors claim that Ag stain was not influenced by the presence of SST14 or CST17.

The silver‐staining was not sensitive enough to detect intermediate‐sized bands, which we interpreted to consist of Aβ and somatostatin, and which were detectable with 6E10.

13) Figure 6A, lanes 1-4 are labeled "no SDS" – is this correct?

Yes, these samples were generated under omission of SDS. We will provide a detailed description of the running conditions of the gel when resubmitting the paper.

[1] Jin M, Shepardson N, Yang T, Chen G, Walsh D, Selkoe DJ. Soluble amyloid beta-protein dimers isolated from Alzheimer cortex directly induce Tau hyperphosphorylation and neuritic degeneration. Proc. Natl Acad Sci USA. 2011; 108(14):5819-24

[2] Kaminski Schierle GS, van de Linde S, Erdelyi M, Esbjorner EK, Klein T, Rees E, et al. In situ measurements of the formation and morphology of intracellular beta‐amyloid fibrils by super-resolution fluorescence imaging. J AM Chem. 2011; 133(33):12902-5. [3] De Felice FG, Velasco PT, Lambert MP, Viola K, Fernandez SJ, Ferreira ST, et al. Aβ oligomers induce neuronal oxidative stress through an N‐methyl‐D‐aspartate receptor‐dependent mechanism that is blocked by the Alzheimer drug memantine. J Biol Chem. 2007;282(15):11590‐601.

[4] Krafft GA, Klein WL. ADDLs and the signaling web that leads to Alzheimer's disease. Neuropharmacology. 2010;59(4‐5):230‐42.

[5] Ahmed M, Davis J, Aucoin D, Sato T, Ahuja S, Aimoto S, et al. Structural conversion of neurotoxic amyloid‐beta(1‐42) oligomers to fibrils. Nat Struct Mol Biol. 2010;17(5):561‐7.

[6] Barghorn S, Nimmrich V, Striebinger A, Krantz C, Keller P, Janson B, et al. Globular amyloid beta‐peptide oligomer ‐ a homogenous and stable neuropathological protein in Alzheimer's disease. J Neurochem. 2005;95(3):834‐47.

[Editors’ note: the authors’ response to the appeal decision letter follows]:

[…] If you are willing to go one more round with us, I suggest you develop such an analysis and send it back to us for review by the same individuals who read the current version of your paper.

Sincerely, Randy Schekman

Editor-in-Chief

Thank you again for your thoughtful comments and for having provided us with the opportunity to resubmit a revised manuscript. Over the past few weeks, we have reworked the manuscript and have added several new analyses to strengthen its conclusions.

1. New primary hippocampal neuron assay data (Figure 8), which assessed if the presence of monomeric or pre‐aggregated somatostatin (SST) affects Aβ‐dependent tau hyperphosphorylation.

You will find that this analysis uncovered a surprising effect of SST on Aβ‐dependent tau hyperphosphorylation. Normally, Aβ only triggers tau hyperphosphorylation in this assay (first described in detail by Dennis Selkoe’s group [7]) after it has been subjected to a pre‐aggregation step, i.e., tau hyperphosphorylation is never observed following addition of a monomeric preparation of Aβ to the cell culture medium. Our new data document that the concomitant addition of pre‐aggregated SST and monomeric Aβ can induce tau hyperphosphorylation. Importantly, neither monomeric nor pre‐aggregated SST alone promotes tau hyperphosphorylation, validating that the tau hyperphosphorylation effect of pre‐aggregated SST depends on monomeric Aβ being also added to the cell culture dish. Moreover, consistent with our other in vitro data, which revealed that SST (and its paralog cortistatin) interact with Aβ, but not other cyclic neuropeptides (AVP, oxytocin or bombesin), we document that the addition of pre‐aggregated preparations of these negative control peptides and monomeric Aβ to the cell culture medium does not induce tau hyperphosphorylation.

2. New epitope mapping data that define elements within SST responsible for binding to Aβ (Figure 7).

This experiment was undertaken in response to reviewer requests to strengthen SST‐Aβ interaction data presented in Figures 3 and 4. The new results demonstrate that the SST‐Aβ interaction depends on a specific SST residue. More specifically, replacement of tryptophan‐8 within SST with another amino acid abolished the lag phase extension observed in the Aβ aggregation assay in the presence of SST. Moreover, other new data, generated with truncated SST peptides, indicate that the stretch of SST residues 5‐11 surrounding the critical tryptophan‐8 is sufficient to induce the lag phase extension. In other words, consistent with data we had included in our original submission, which documented that the lag phase extension also occurred when the disulfide bridge within SST was reduced, these new results corroborate the conclusion that binding of SST to Aβ does not rely on the SST being present in its cyclic form. Notably, the SST epitope for binding to Aβ comprises the previously known minimal epitope (phenylalanine‐7‐tryptophan‐8‐lysine‐9‐threonine‐10) required for SST’s biological activity, which depends on its binding to one of five known SST receptors (SSTR1‐5)[8]. This observation is interesting as it raises the prospect that Aβ may compete with and thereby modulate SST ligand‐receptor interactions.

3. New nanoparticle tracking analysis that explores particle size distributions of Aβ in the presence or absence of SST (Figure 7—figure supplement 1)

This experiment was conducted in response to reviewer criticism of Figures 5 and 6 of our original submission, which documented differences in Aβ‐specific signals when ThT aggregation assay fractions, generated in the presence or absence of SST, were analyzed by Western blotting. To determine if these differences in Aβ assemblies were merely artifacts of sample handling, i.e., introduced by denaturing conditions applied during Western blot analysis, we sought to compare Aβ aggregate size populations in the presence or absence of SST by an orthogonal method that would not be expected to suffer from Western blot confounders. Nanoparticle tracking analysis (based on Nanosight technology [9]) is such a method that can be used to characterize complex populations of molecules of differing particle size. The new data document that Aβ aggregates present in ThT aggregation assay fractions after 2 hours of incubation are profoundly shifted toward smaller particle sizes when samples were co‐incubated with SST. Importantly, the replacement of SST with an SST‐W8P negative control peptide did not result in this particle size shift toward smaller aggregates, indicating that the altered particle size distribution is caused by the SST‐Aβ interaction, rather than by some unspecific effect of the peptide.

4. Inclusion of full‐view silver‐stained gels as a figure supplement (Figure 5—figure supplement 1).

One reviewer expressed concern about the fact that we had truncated the silver‐stained images depicted in Figure 5B and C. We have now included the full view of these gels. As described in our previous appeal letter, these gels document that the respective Aβ and SST (CST) bands, which we had highlighted in our original submission were the only bands visible in the low mass range of the gel. Other signals visible on the gel in the higher mass range did not follow the logic of the experiment and, therefore, were interpreted as artefacts, often seen in overexposed silver‐stains of this kind. Importantly, the levels of these high mass signals indicate no irregularities in the loading and running that by themselves could explain the differences in bands observed in the low mass region.

5. New electron microscopy data that compare Aβ assemblies in the presence or absence of SST (Figure 7—figure supplement 2).

We followed the request to include electron micrographs that depict how the presence of SST affects visible structures. As expected, Aβ1‐42 that was incubated in PBS resulted in the formation of characteristic amyloid fibrils, while Aβ1‐42 that was incubated with equimolar concentrations of SST formed small oligomeric assemblies instead. This observation confirmed the effects SST is having on the quaternary structure of Aβ.

6. Revised Materials and Methods section.

The revised paper contains a considerably expanded Materials and Methods section. Not only were original descriptions of methodologies improved but several new experimental methods are included that describe methodologies for the primary hippocampal neuron assay, the epitope mapping analyses, the nanoparticle tracking analyses and the electron microscopy work.

7. Revised Discussion and text in other sections.

We followed the suggestion of reviewers to cut down the Discussion. In the revised document, the speculative section dedicated to how data presented in this study may lead to a revised model of the role of SST in Alzheimer’s disease has been removed. However, we have strengthened the critical evaluation of our data. We also addressed reviewer comments by inserting in several places in the manuscript additional information that clarifies aspects of our work that could easily be missed or were not well explained in the first submission.

In conclusion, we again thank all reviewers for constructive comments and for taking the time out of their busy schedules to review our manuscript. We are also most appreciative of the considerable efforts invested by the editorial team of eLife to improve our submission and hope you will agree that the manuscript is now ready for publication.

References:

[7] Jin M, Shepardson N, Yang T, Chen G, Walsh D, Selkoe DJ. Soluble amyloid beta‐protein dimers isolated from Alzheimer cortex directly induce Tau hyperphosphorylation and neuritic degeneration. Proc Natl Acad Sci U S A. 2011;108(14):5819‐24.

[8] Pohl E, Heine A, Sheldrick GM, Dauter Z, Wilson KS, Kallen J, et al. Structure of octreotide, a somatostatin analogue. Acta Crystallogr D Biol Crystallogr. 1995;51(Pt 1):48‐59.

[9] Filipe V, Hawe A, Jiskoot W. Critical evaluation of Nanoparticle Tracking Analysis (NTA) by NanoSight for the measurement of nanoparticles and protein aggregates. Pharmaceutical research. 2010;27(5):796‐810.

[Editors’ note: the author responses to the re-review follow.]

We remain somewhat conflicted about this submission. It has some strengths and the new data are a plus, but the physiological or more importantly pathophysiological role of the-SST interaction is still unclear.

In our view, the pathophysiological role of the Aβ ‐SST interaction will only come into full view, once we learn if and how the presence or absence of SST (and possibly CST) affects pathological hallmarks of Alzheimer’s disease in mice and humans. The experiments in primary hippocampal neurons provide a first hint but admittedly fall short of providing conclusive answers. While we would have loved to be able to address also this point more conclusively, in our view the in‐depth Aβ interactome data we provided, together with the discovery and detailed characterization of the Aβ ‐SST interaction, represents a major contribution in itself.

The new data don't directly address the request I made in our exchange n your appeal letter:

"However, if as you suggest it may be possible to evaluate the effect of somatostatin secreted by cells on the activity of Aβ monomer or oligomer in the phosphorylation of tau, then this at least could lead to a more plausible if not conclusive connection to disease."

Instead, you used synthetic SST and synthetic Aβ.

The perceived deviation in the design of the primary hippocampal neuron assay data was caused by a misunderstanding. When we wrote the appeal letter, we did not intend our planned experiments to be based on ‘secreted’ SST (please see below for explanations) but had in mind the very analysis we included in the resubmission. More specifically, the statement we used in our original appeal letter was: “Less conclusive, yet doable experiments, in which we will investigate if the presence of somatostatin in the cell culture medium influences effects of monomeric or oligomeric Aβ preparations on tau phosphorylation in primary hippocampal neurons (analogous to published work by Jin et al., 2011), were initiated immediately following our eLife submission and, hopefully, will lead to results within the next month.”

However, the new data in Figure 8 do indicate that co-addition of SST withmonomer to primary neurons leads to increased phosphorylation of tau. So the authors have gone some way towards the request. Please comment on your use of synthetic peptides rather than those interacting in the medium of cultured cells.

There are several considerations that guided our choice to base the experiment on the addition of synthetic Aβ and synthetic SST:

1) Early in its biogenesis SST is directed toward the regulated secretory pathway. Several conditions have been shown by others to induce or modulate the regulated release of neuropeptides, including SST (e.g., depolarization of the cellular membrane followed by Ca2+ influx through voltage‐sensitive Ca2+ channels (Kang et al, 1987), or addition of growth hormone, dopamine, GABA, acetylcholine, or a subset of so‐called small‐molecule secretagogues). However, the timed release of a considerable proportion of SST stores from neurons in a way that does not perturb other aspects of the cellular homeostasis, including kinase signaling, represents a challenge. In the absence of a robust method that causes the selective release of SST, the cell culture assay would have needed to be based on steadystate levels of this peptide in the cell culture medium. For generating a negative control, this would have forced a need to inhibit SST release, to sequester SST from the medium (so it cannot interact with Aβ), or to work with cells obtained from SST knockout mice. While we have obtained an SST knockout mouse model and are currently undertaking cross‐breeding experiments with AD model mice, our breeding colony is too small for dedicating mice at this time to experiments we consider high risk (see below). Part of this risk stems from our expectation that kinase signaling pathways might be altered in cells from these mice, due to previously established influences of SST receptor activation on kinase signaling pathways (Goate et al., 1991; Rogaev et al., 1995; Sherrinton et al. 1995 and Yang, Li, Walsh and Selkoe, 2017).

2)Note that the release of SST into the cell culture medium would still be a poor mimic of the natural release of this peptide into the extracellular space within the architecture of the brain. This is because it is likely that in the cell culture dish SST would be more rapidly dispersed and its levels diluted upon its release into the relatively excessive volume of the cell culture medium. In contrast, in the brain, where cells exist in a relatively ‘dry’ environment, even modest levels of SST release may lead to localized spikes in its concentration.

3) In fact, because SST is stored in dense core granules as amyloid‐like aggregates (Selkoe, 20010, it is to be expected that in proximity to its natural release sites high concentrations of this peptide, including residual aggregates of varying size, are present in the brain. The disintegration of amyloid aggregates released in this manner has been observed to play out in time‐scales of several hours (Narayan et al., 2012). Thus, it is plausible that the low μM concentrations of synthetic SST, which we chose for our experiments, can also be encountered in the brain. Note also that there is no evidence that secreted or synthetic SST are different in their composition or post‐translational modifications.

4)In the brain, SST is predominantly released by GABAergic interneurons. While it is conceivable that one may enrich these by FACS sorting, such methods would be expected to return a dissatisfactory cell yield, below the number of cells required for carrying out the hippocampal neuron assay. Note that even in its current form, one female mouse that is pregnant with an average number of embryos allows us to test no more than 12 different assay conditions. Because interneurons represent a small fraction of the total hippocampal cell count, their FACS selection would push this assay to its limit (or beyond it), certainly not allowing a side‐by‐side comparison of the exposure of cells harvested from littermates to different peptide mixtures.

5)The cell culture assay that monitors tau hyperphosphorylation in response to the addition of synthetic Aβ was optimized and described in detail for primary hippocampal mouse neurons harvested on day 16‐18 from embryos and cultured in a specific cell culture medium for an additional 18 days. Any deviation from this protocol would have exposed experiments to criticism that the data cannot be compared to preexisting data generated with this assay. While this may seem a minor consideration, the Alzheimer’s disease research community generally dismisses cell‐based assay results as not being reflective of signaling underlying disease. The assay format we employed is, to our knowledge, the only cell culture based assay that enjoys relatively broad acceptance in the field for experiments probing Aβ ‐dependent tau hyperphosphorylation, due to data by others, which established similar tau phosphorylation site profiles observed in the disease and in analyses based on this assay.

6) The addition of synthetic peptides allowed us to generate several informative controls. Note that had we been able to overcome some of the aforementioned hurdles and work with endogenously released SST, we still would not have been able to compare the effect of SST (or CST) with those of other cyclic neuropeptides, including AVP and oxytocin, or the peptide bombesin, which is not cyclic but has been described to share with SST similar aggregation characteristics (see Table S1 in Selkoe, 2001). Finally, the assay design we selected allowed us to add SST to the cell culture medium in monomeric or pre‐aggregated form, thereby providing the insight that for the SST‐dependent tau hyperphosphorylation to be observed, cells had to be exposed to pre‐aggregated SST AND monomeric Aβ. This mixture is remarkable in that it strictly relied on monomeric Aβ also being present, yet we have never observed monomeric Aβ by itself to cause tau hyperphosphorylation.

A few sentences have been added to the manuscript to explain some of these considerations to the interested reader.

The new data in Figure 7 demonstrate that Trp8 in SST is essential for the inhibition ofaggregation, and that SST causes production of relatively small aggregates, this certainly bolsters the specificity and effect of SST, but this was considered a minor point.

Agreed.

https://doi.org/10.7554/eLife.28401.022

Article and author information

Author details

  1. Hansen Wang

    Tanz Centre for Research in Neurodegenerative Diseases, University of Toronto, Toronto, Canada
    Contribution
    HWa, Conceptualization, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—review and editing
    Competing interests
    HWa: Holds provisionary US patent on amyloid-beta binding polypeptides based on the results of this study (filing number 62/451,309).
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4414-3372
  2. Lisa D Muiznieks

    Molecular Medicine Program, Research Institute, The Hospital for Sick Children, Toronto, Canada
    Contribution
    LDM, Data curation, Validation, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared.
  3. Punam Ghosh

    Department of Medical Biophysics, University of Toronto, Toronto, Canada
    Contribution
    PG, Data curation, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared.
  4. Declan Williams

    Tanz Centre for Research in Neurodegenerative Diseases, University of Toronto, Toronto, Canada
    Contribution
    DW, Data curation, Software, Investigation, Visualization, Writing—review and editing
    Competing interests
    No competing interests declared.
  5. Michael Solarski

    1. Tanz Centre for Research in Neurodegenerative Diseases, University of Toronto, Toronto, Canada
    2. Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, Canada
    Contribution
    MS, Investigation, Visualization, Writing—review and editing
    Competing interests
    No competing interests declared.
  6. Andrew Fang

    1. Department of Biochemistry, University of Alberta, Edmonton, Canada
    2. Centre for Prions and Protein Folding Diseases, University of Alberta, Edmonton, Canada
    Contribution
    AF, Formal analysis, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared.
  7. Alejandro Ruiz-Riquelme

    Tanz Centre for Research in Neurodegenerative Diseases, University of Toronto, Toronto, Canada
    Contribution
    AR-R, Investigation, Methodology
    Competing interests
    No competing interests declared.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6581-7132
  8. Régis Pomès

    1. Molecular Medicine Program, Research Institute, The Hospital for Sick Children, Toronto, Canada
    2. Department of Biochemistry, University of Toronto, Toronto, Canada
    Contribution
    RP, Conceptualization, Resources, Software, Supervision, Investigation
    Competing interests
    No competing interests declared.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3068-9833
  9. Joel C Watts

    1. Tanz Centre for Research in Neurodegenerative Diseases, University of Toronto, Toronto, Canada
    2. Department of Biochemistry, University of Toronto, Toronto, Canada
    Contribution
    JCW, Conceptualization, Supervision, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared.
  10. Avi Chakrabartty

    1. Department of Medical Biophysics, University of Toronto, Toronto, Canada
    2. Department of Biochemistry, University of Toronto, Toronto, Canada
    Contribution
    AC, Conceptualization, Formal analysis, Supervision, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared.
  11. Holger Wille

    1. Department of Biochemistry, University of Alberta, Edmonton, Canada
    2. Centre for Prions and Protein Folding Diseases, University of Alberta, Edmonton, Canada
    Contribution
    HWi, Conceptualization, Resources, Supervision, Investigation, Visualization, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5102-8706
  12. Simon Sharpe

    1. Molecular Medicine Program, Research Institute, The Hospital for Sick Children, Toronto, Canada
    2. Department of Biochemistry, University of Toronto, Toronto, Canada
    Contribution
    SS, Conceptualization, Resources, Data curation, Supervision, Funding acquisition, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared.
  13. Gerold Schmitt-Ulms

    1. Tanz Centre for Research in Neurodegenerative Diseases, University of Toronto, Toronto, Canada
    2. Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, Canada
    Contribution
    GS-U, Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    g.schmittulms@utoronto.ca
    Competing interests
    GS-U: Holds provisionary US patent on amyloid-beta binding polypeptides based on the results of this study (filing number 62/451,309)
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6962-0919

Funding

Canadian Institutes of Health Research

  • Gerold Schmitt-Ulms

Ontario Centres of Excellence

  • Simon Sharpe
  • Gerold Schmitt-Ulms

Alberta Innovates Bio Solutions (201600028)

  • Holger Wille
  • Gerold Schmitt-Ulms

Heart and Stroke Foundation of Canada (G-15-0009148)

  • Simon Sharpe

Canada Foundation for Innovation

  • Gerold Schmitt-Ulms

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This project received funding from the Canadian Institutes of Health Research (CIHR), a medical sciences proof-of-principle (MScPoP) funding program hosted by MaRS Innovation and the Ontario Centres of Excellence Inc., the Alberta Prion Research Institute (201600028), a Grant in Aid from the Heart and Stroke Foundation of Canada (G-15-0009148), and most generous private support from the Rosiak family. The mass spectrometry component of this study was enabled by generous philanthropic support from the Irwin family and a Leading Edge Fund (LEF) infrastructure grant from the Canada Foundation for Innovation (CFI). We acknowledge the Structural and Biophysical Core Facility, Hospital for Sick Children, for access to the NanoSight instrument. The electron microscopy facility was established with support from a New Initiatives Fund (NIF) award from the CFI.

Ethics

Animal experimentation: The work was performed in strict accordance with University of Toronto animal care and biosafety recommendations. All mice were handled according to procedures approved (AUP4183.3) by the animal care committee at University Health Network overseeing work at the Krembil Discovery Centre (Toronto). The handling of samples and reagents followed biosafety procedures approved (208-S06-2) by the University of Toronto Biosafety Program.

Reviewing Editor

  1. Randy Schekman, Howard Hughes Medical Institute, University of California, Berkeley, United States

Publication history

  1. Received: November 18, 2016
  2. Accepted: June 14, 2017
  3. Accepted Manuscript published: June 26, 2017 (version 1)
  4. Version of Record published: July 11, 2017 (version 2)

Copyright

© 2017, Wang et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,741
    Page views
  • 393
    Downloads
  • 5
    Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Crossref, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Biochemistry and Chemical Biology
    2. Microbiology and Infectious Disease
    James Lee et al.
    Research Article
    1. Biochemistry and Chemical Biology
    2. Cell Biology
    Tomoki Naito et al.
    Research Article