Predict your next investment

SOFTWARE (NON-INTERNET/MOBILE) | HR & Workforce Management Software
searchinco.co

See what CB Insights has to offer

Stage

Incubator/Accelerator | Alive

About Searchin

Searchin is an AI powered talent sourcer discovering and analyzing candidates by skills, personality and company culture within minutes.

Searchin Headquarter Location

Tel Aviv,

Israel

Latest Searchin News

Systematic elucidation of neuron-astrocyte interaction in models of amyotrophic lateral sclerosis using multi-modal integrated b...

Nov 4, 2020

1 Altmetric Abstract Cell-to-cell communications are critical determinants of pathophysiological phenotypes, but methodologies for their systematic elucidation are lacking. Herein, we propose an approach for the Systematic Elucidation and Assessment of Regulatory Cell-to-cell Interaction Networks (SEARCHIN) to identify ligand-mediated interactions between distinct cellular compartments. To test this approach, we selected a model of amyotrophic lateral sclerosis (ALS), in which astrocytes expressing mutant superoxide dismutase-1 (mutSOD1) kill wild-type motor neurons (MNs) by an unknown mechanism. Our integrative analysis that combines proteomics and regulatory network analysis infers the interaction between astrocyte-released amyloid precursor protein (APP) and death receptor-6 (DR6) on MNs as the top predicted ligand-receptor pair. The inferred deleterious role of APP and DR6 is confirmed in vitro in models of ALS. Moreover, the DR6 knockdown in MNs of transgenic mutSOD1 mice attenuates the ALS-like phenotype. Our results support the usefulness of integrative, systems biology approach to gain insights into complex neurobiological disease processes as in ALS and posit that the proposed methodology is not restricted to this biological context and could be used in a variety of other non-cell-autonomous communication mechanisms. Introduction The rationale for the elucidation of mechanisms that mediate cell-to-cell communication processes is that ligands released by one cell type may induce specific changes in gene-product activity in one or more additional cell types, ultimately resulting in an observable molecular phenotype. The latter can range from the activation of specific immune response pathways to the aberrant reprogramming of cell state and even cell death. Classical hypothesis-driven mechanistic elucidation of these intercellular signaling pathways requires complex, time-consuming, and laborious work to gradually restrict the number of molecular players that may mediate the cell-to-cell interaction of interest. As a result, high-throughput methodologies for the systematic prioritization of process-specific ligand–receptor interactions are critically needed, yet, still largely elusive. By combining recent results on modeling transcriptional, signal transduction and other context-specific molecular interaction networks with proteomics, we propose an approach for the Systematic Elucidation and Assessment of Regulatory Cell-to-cell Interaction Networks (SEARCHIN) to prioritize cross-compartment ligand–receptor interactions that may mediate specific cellular phenotypes. Specifically, SEARCHIN implements a workflow that integrates several algorithms, such as ARACNe 1 , VIPER 2 , MINDy 3 , and PrePPI 4 , that were not originally designed to study the cell-cell communication process. As a proof-of-concept for its utilization, we elect to study amyotrophic lateral sclerosis (ALS), a common fatal, paralytic disorder that provides an especially relevant context. Indeed, although ALS is characterized by preferential death of motor neurons (MNs) 5 , the contribution of aberrant cell-to-cell interactions involving non-neuronal cells such as glia to the neurodegenerative process is increasingly acknowledged 6 . Specifically, several investigators have exploited the fact that mutations in superoxide dismutase-1 (mutSOD1) gene not only cause a familial form of ALS 7 , but also provide an ideal molecular tool to study non-cell autonomous mechanisms given their cell-ubiquitous expression and toxic gain-of-function properties. In keeping with this, we and others have reported that astrocytes expressing mutSOD1 selectively kill wild-type (WT) mouse primary spinal MNs and embryonic stem cell-derived MNs (ES-MNs) 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 . This spontaneous neurodegenerative phenotype was observed either when MNs were cultured in the presence of mutSOD1-expressing astrocytes or when they were exposed to medium conditioned by mutant astrocytes 8 , 9 , 20 . Similar to these findings from mouse mutSOD1-expressing astrocytes, we showed that astrocytes derived from postmortem CNS samples from human sporadic ALS (sALS) patients were also associated with a loss of human ES-MNs 9 . Furthermore, it has been reported that mutSOD1-expressing glial-restricted precursor cells grafted onto spinal cords (SCs) of WT rats were also associated with MN loss in living animals 21 and that selective reduction of mutSOD1 levels in astrocytes prolonged survival in transgenic (Tg) SOD1G37R mice 22 . Taken together, these observations suggest that astrocyte-mediated MN degeneration is a general phenomenon in ALS and is not restricted to in vitro systems, mouse cells, or mutSOD1-linked ALS. Since our initial observation 20 , we gained significant insights into the MN-intrinsic molecular cascade that drives neurodegeneration in vitro, both in murine mutSOD1 and in human sALS cells 8 , 9 . For instance, our published data support NF-κB1 as a likely apical master regulator (MR), whose translocation to the nucleus of cultured MNs mediates mutSOD1 astrocyte-induced neurodegeneration 8 . As a result, it is reasonable to assume the existence of a ligand released by mutSOD1-expressing astrocytes, capable of activating a receptor-mediated signal in MNs, ultimately leading to NF-κB1 activation. However, the precise mechanisms promoting astrocyte-induced demise of neighboring MNs in ALS remain elusive. In this paper, we developed the SEARCHIN pipeline to integrate information on proteins enriched in medium conditioned by mutSOD1 astrocytes with regulatory network-based analysis of their cognate receptor-mediated NF-κB1 activation in MNs, a previously identified MR of MN demise. The analysis identified amyloid precursor protein (APP) as the most likely toxic factor released by the astrocytes. In addition, SEARCHIN identified the Tumor Necrosis Factor Receptor Superfamily member 21 (TNFRSF21)—also known as death receptor-6 (DR6)—as the APP cognate receptor most likely to be responsible for transducing mutSOD1 astrocyte-induced death in MNs. We confirmed these predictions experimentally by showing that MN toxicity depends on the expression of APP in astrocytes and DR6 in MNs, in both mouse mutSOD1 and human sALS astrocytes. In vivo, RNA interference (RNAi)-mediated knockdown of DR6 in MNs-attenuated neurodegeneration. In light of these data and of our previous studies 8 , 9 , we propose a pathogenic model of neurodegeneration in ALS in which astrocyte-specific release of a soluble fragment of APP, and possibly of other ligands prioritized by our analysis, activates DR6 at the surface of MNs, thus triggering a death signal that culminates in the demise of spinal MNs via NF-κB1-dependent pathway. In addition, these data support the use of SEARCHIN as a pipeline to systematically prioritize candidate ligand–receptor interactions that may mediate a variety of pathophysiological processes, across diverse cellular compartments. Results Mouse mutSOD1 astrocytes mediate MN death by a toxic proteinaceous factor(s) Based on our previous data, we have observed significant reduction in MN number after exposure to either mouse mutSOD1 and human sALS astrocytes or their corresponding astrocyte-conditioned media (ACMs) 8 , 9 , 20 , resulting in a ~50% reduction of healthy MNs by day 7 in vitro (DIV 7). To determine whether the deleterious effect of ALS astrocytes is mediated by either a progressive loss-of-supportive function or a buildup-of-toxic factor(s), three sets of experiments were performed. First, both mouse mutSOD1 and non-transgenic (NTg) control ACMs were subjected to ion-exchange liquid chromatography. We found that the Q-column (i.e., positively charged quaternary amines-coated column) eluates from mutSOD1 ACMs were associated with greater MN loss at 7 DIV, compared with unfiltered mutSOD1 ACMs after 5 days incubation (Fig. 1a ). Meanwhile, Q column eluates from NTg ACMs were not associated with any significant reduction of MN numbers, compared with their unfiltered counterparts (Fig. 1a ). In contrast, there was no significant difference in MN numbers at 7 DIV between cultures incubated for 5 days with the S column (i.e., negatively charged sulfate derivatives-coated column) eluates from mutSOD1 or NTg ACMs (Fig. 1a ). These results suggest the presence of a negatively charged toxic factor in mouse mutSOD1 ACM that can be enriched using Q column ion-exchange liquid chromatography. Fig. 1: Characterization of the deleterious effects of ACM from Tg mutSOD1 mice. a Representative images of MN cultures exposed to NTg (blue) and mutSOD1 (red) ACM stained with the neuronal marker SMI-32. Scale bar: 50 μm. ACM from NTg (n = 3) or mutSOD1 (n = 3) astrocytes were passed through anion exchange Q column or cation exchange S column. Elutes were collected and then applied to mouse primary MNs. GFP+MNs were counted manually using epifluorescent microscope. Data are means ± SEM of independent experiments (n) analyzed by two-way ANOVA (Interaction F(2,12) = 17.61; P = 0.0003) followed by Sidak post hoc test: ***P = 0.0001 Control NTg vs mutSOD1 (CI 22.61–57.39%; d = 4.73); ****P ≤ 0.0001 Q eluate NTg vs mutSOD1 (CI: 47.61–82.39%; d = 28.47). b ACM from NTg (blue; n = 5) or mutSOD1 (red; n = 5) were passed through different molecular weight cutoff filters. The retentates were resuspended in fresh media and then applied to mouse primary MN culture and counted manually. Data are means ± SEM of independent experiments (n) analyzed by two-way ANOVA (Interaction F(5,24) = 19.14; P < 0.0001) followed by Sidak post hoc test: ****P ≤ 0.0001 Control NTg vs mutSOD1 (CI: 34.26–70.94%; d = 6.35) and 5 K NTg vs mutSOD1 (CI: 31.06–67.74%; d = 5.76); **P = 0.0015 10 K NTg vs mutSOD1 (CI: 9.06–45.74%; d = 1.55). c To examine molecular property of the deleterious effect in MNs, ACM from NTg (blue) or mutSOD1 (red) mice was either untreated (n = 10), heat-inactivated using water bath (n = 8 HI 15 min and n = 7 HI 30 min), treated with pepsin (n = 8), charcoal (n = 6) or chloroform (n = 7) and then applied to mouse primary MNs and counted manually. Data are mean ± SEM of independent experiments (n) analyzed by two-way ANOVA (Interaction F(5,40) = 21.22; P < 0.0001) followed by Sidak post hoc test: ****P ≤ 0.0001 Control NTg vs mutSOD1 (CI: 33.61–53.99%; d = 5.86), HI 15 min NTg vs mutSOD1 (CI: 11.98–34.77%; d = 1.63), charcoal NTg vs mutSOD1 (CI: 20.68–46.99%; d = 3.24), and chloroform NTg vs mutSOD1 (CI: 26.82–51.18%; d = 6.10). In Fig. 1 , all primary spinal MN cultures were from Tg HB9::EGFP mice. See also Supplementary Fig. 1 . Source data provided as source data file. In light of the above findings, we asked whether, akin to MMP9 ref. 46 , deletion of DR6 might confer a generalized resistance of MNs to death-inducing agents. Accordingly, we set out to examine MN susceptibility to domoic acid and Fas activation 47 , two compounds known to cause MN degeneration by mechanisms distinct from those involved in astrocyte-induced MN death 20 . For this experiment, we went back to our primary MN culture system and, by following the protocol of Raoul and collaborators 47 , we demonstrated that both domoic acid and agonistic anti-Fas antibody killed WT and DR6−/− MNs to a comparable degree (Fig. 6c ), indicating that DR6 deficient MNs are not resistant to any insults. Thus, our findings support the notion that DR6—expressed at the surface of MNs—not only contribute to the demise of these cells in response to ALS astrocytes, but may also modulate MN vulnerability to this astrocyte-derived insult. Discussion Molecular interactions between distinct cellular niches have emerged as a key determinant in an increasing number of pathophysiologic phenotypes. Yet, their systematic elucidation is lagging and still represents an unmet challenge in molecular biology. We have introduced SEARCHIN, as an analytic pipeline supporting the integration of diverse regulatory and proteomic evidence sources to prioritize the ligand–receptor interactions that most likely mediate these pathophysiologic phenotypes. Critically, our bioinformatics approach depends on the availability of regulatory networks, which can be inferred from gene expression data, on gene expression signatures that recapitulate the effect of the interaction, and on the availability of proteomics data identifying candidate ligands. As a result, SEARCHIN should be generalizable to a variety of additional contexts. Furthermore, recent results from the application of network-based approaches, such as metaVIPER to single cells, suggest that SEARCHIN may be effectively applied to elucidate mediators of cell-to-cell interactions at the single cell level 48 . However, analytical results are limited by the existing repertoire of available PPIs, in this case based on the PrePPI algorithm, which, albeit constantly expanding, is still not fully comprehensive. Herein, we sought to apply our proposed integrative approaches to a cell-to-cell interaction setting provided by a MN death phenotype induced, non-cell autonomously, by neighboring human sALS and mouse mutSOD1 astrocytes 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 . As a preamble to our computational investigations, we performed a series of experiments aimed at better defining the biological framework of our selected cell-to-cell disease model. We began by clarifying whether MN degeneration is due to a lack of beneficial property or to a gain of rather toxic properties by ALS astrocytes. Relevant to this question is the work of Tyzack, et al. 49 showing that the neuroprotective EphB1-ephrin-B1 cross-talk between MNs and astrocytes is disrupted in human ES-derived astrocytes and mutSOD1 mouse models of ALS. Although this study suggests that part of the non-cell autonomous arm of ALS pathogenesis is related to a loss of function, our present data argue that mouse mutSOD1 and human sALS astrocytes kill MNs by a gain of toxic function. Indeed, through a combination of biochemical methods, we provide evidence that mouse mutSOD1 astrocytes release a soluble toxic protein with a molecular mass between 5–30 kDa (Fig. 1b ). This fact allowed us to rule out small molecules (≤1kDa) including oxygen and nitrogen reactive species and neurotoxic amino acids such as glutamate, which have often been proposed as causal in the ALS-related MN degeneration 50 as well as lipids and prostanoids, since chloroform extraction fails to alter mutSOD1 astrocyte-induced MN death (Fig. 1c ). In addition, using immunodepletion and supplementation experiments, we also excluded SOD1 itself, which has been reported as causing MN degeneration in vitro 13 , 23 . Given the proteinaceous nature of the putative toxic factor, we then performed an unbiased proteomics analysis of the ACMs, and, first, examined these data for particular astrocyte inflammatory phenotypes such as A1 and A2 reactive signature 26 . Through a meta-analysis, we found marginal overlap between our proteomics and the genomic data of Liddelow, et al. 26 , allowing us to infer that mutSOD1 astrocytes are unlikely to kill MNs via an A1 astrocyte mechanism 26 . The corollary of this conclusion is that the A1 neurotoxic concept may be more context-specific than initially anticipated and raises the possibility that astrocytes contribute to neuron death by mechanisms that call for greater molecular nuances than the proposed A1/A2 dichotomy 26 . All of the aforementioned investigations, not only confirmed the usual poor yield of candidate-based strategies aimed at elucidating the nature of the cell-to-cell communication, but also provided a solid rationale to try using an unbiased approach that combined ACM proteomics with MN functional genomics in a multistep bioinformatics process to address this challenging question. As such, we thought to utilize SEARCHIN as a proof-of-principle to show that meaningful biology can be identified via an integrative approach, rather than provide a comprehensive benchmark of the methodologies that include testing a variety of parameters, algorithms or assessing the impact of alternative evidence integration methodologies. Incidentally, in developing SEARCHIN, we realized that effective gold-standard databases for astrocyte-MN-specific interactions are lacking and, to our knowledge, there are no alternative methodologies to prioritize the cell-to-cell interactions that mediate a specific phenotype. As a result, any effort to compare the rankings generated by different methodologies would have had to rely on direct experimental validation, which is overly complex and time-consuming in this context. Within this framework, we found that SEARCHIN was able to distill a list of 19,370 potential interactions between 65 over-abundant proteins in the mutSOD1 ACM and 298 candidate surface and nuclear receptors in MNs to only three statistically significant ligand–receptor pairs (Nme1-Ptprn; Cdh2-Ryk; and App-Dr6), which represents a major implosion in the number of hypotheses that may need to be experimentally validated. Thus, although additional efforts to further validate the pipeline are still ongoing, it should be clear that SEARCHIN addresses an important, yet poorly addressed problem in molecular biology. Prior to discussing our experiment validation, potential users must be aware that for this study, we have used well-validated algorithms like VIPER, CINDy, and PrePPI to test a methodology to generate reasonable and testable hypotheses for experimental validation. Yet, we do not assert that our proposed approach is the sole or even the optimal strategy to gain insights into the nature of non-cell autonomous pathogenic mechanisms, but merely a demonstration that will encourage investigators to develop and benchmark suitable variants—or even different implementations—as well as to generate appropriate gold-standard sets for systematic validation. It is also important to emphasize that for each individual step, we used the default parameters and thresholds, based on extensive benchmarks performed in the original manuscripts in which the various algorithms were introduced 2 , 3 , 51 , 52 . For instance, since PrePPI uses a likelihood rather than p value-based model, p = 0.5 represents the natural threshold at which an interaction is more likely to be true than false. In the corresponding manuscript 51 , we showed that >80% of the predictions made by PrePPI using this analysis were confirmed experimentally. In future studies, it will thus be interesting to determine whether different choice of parameters might produce even deeper biological insights into the cell-to-cell communication in models of human diseases, a possibility that cannot be assessed at this stage, owing to experimental complexity, lack of alternative methodologies, and lack of effective gold-standard datasets. At last, it should be remembered that we used an evidence integration approach, as our ultimate goal was to achieve a global score that represents the balanced integration of all the available evidences, rather than being dominated by a specific evidence source (e.g., PrePPI or VIPER). For this same reason, we favored a rank-based integration method rather than a p value-based one, which may bias some of the algorithms. As a result, in our study—as also typically done when using evidence integration approaches based on multiple weak clues—it is appropriate to use more relaxed statistical threshold for each individual evidence source contributing to the ultimate outcome. As mentioned above, of the three significant ligand–receptor pairs identified by SEARCHIN, we began our experimental validation by studying the role of APP and DR6 in our ALS models, as APP–DR6 interaction has been previously validated 32 , 33 , 34 and proposed in a few neurobiology-relevant phenotypes. For instance, APP-DR6 emerged as instrumental to axonal pruning during development 32 , 34 and both DR6 and APP are critical for axonal plasticity in adult mice 53 , 54 . In addition, it was reported that APP-DR6 has a role in the pathophysiology of Alzheimer’s disease 32 , a view that could not be confirmed 33 . In contrast, for the two other statistically significant ligand–receptor pairs identified by SEARCHIN, we found no experimentally proven interaction between Nme1 and Ptprn while the interaction between Cdh2 and Ryk has been confirmed only by affinity capture-mass spectometry 55 . Yet, all of these four proteins have neurobiological activities relevant to brain development, axon guidance, synapse adhesion as well as pathological situations such as brain tumor, traumatic brain injury, and stroke 56 , 57 , 58 , 59 , which deserve further investigations in the context of neurodegeneration. In support of the role of APP-DR6 in ALS, we found that DR6 is highly expressed in MNs (Fig. 5a ) and that its ablation in these neurons was protective against the toxicity of both mouse mutSOD1 and human sALS astrocytes (Fig. 4 ), confirming that the identified non-cell autonomous toxic mechanism is not specific to mutSOD1 or mice. Furthermore, the importance of DR6 in MN degeneration was not restricted to our in vitro models as its reduction by viral gene silencing attenuated MN loss in Tg mutSOD1 mice (Fig. 5b , c ). We also provide evidence that DR6 expression in MNs parallels that of MMP9 (Fig. 6a , b ). As MMP9 is expressed in selectively vulnerable MNs 46 , our findings led us to posit that MNs with the highest levels of DR6 may be more prone to degeneration. However, more investigations are needed to elucidate the mechanism and pathogenic significance of the DR6/MMP9 correlation. As for our validations of APP, our results suggest that MN toxicity involves an astrocyte-derived protein whose sequence is within the ectodomain of APP upstream to the β-secretase major cleavage site 60 . Relevant to this interpretation are the demonstrations that both ablation of APP 61 and antibody inhibition of β-secretase 62 mitigate the ALS-like phenotype in Tg mutSOD1 mice. Of note, unlike Rabinovich-Toidman, et al. 62 , we did not find, as reported in the results section, that reduction of β-secretase function using viral gene therapy altered SOD1 expression nor did we find, that silencing mutSOD1 altered APP expression. Our idea that astrocyte toxicity is mediated by a soluble fragment of APP is also consistent with previous crystallography studies 63 and cell-based binding assays 34 , 63 showing that a fragment of sAPP-β can bind to DR6. Aside from APP, our results also raise the possibility for a role played by APLP1 in mutSOD1 astrocyte-induced toxicity. Indeed, we found that silencing APLP1, but not APLP2, in astrocytes consistently phenocopied the knockdown of APP (Supplementary Fig. 3 ). This finding suggests that both APP and APLP1 might function in the same molecular pathway that contribute to the death of MNs in response to mutSOD1 astrocytes. In light of our results, we propose that APP-DR6 contribute to neurodegeneration via a β-secretase-dependent non-cell autonomous scenario, which is in contrast with the β-secretase-independent cell autonomous scenario put forward in the axonal homeostatic situations 32 , 34 , 53 , 54 . Given the role of DR6 in axonal pruning during development, surprisingly but importantly, DR6 silencing in our in vivo model of ALS protected against the loss of MN cell bodies but was less effective in mitigating motor axon damage (Fig. 5 ). Thus, our findings are critical since they argue for a distinct role of APP–DR6 in pathologic vs. normal neurobiological situations. Of note, dissociation between cell body and axon protection in Tg mutSOD1 mice is reminiscent of that reported for Bax deletion in these animals 64 , raising the possibility that APP/DR6 engage a molecular pathway of MN death that involves Bax, which would be consistent with our in vitro data 9 , 20 . In summary, the β-secretase-mediated role played by the APP/DR6 interaction in ALS is both novel and based on a solid foundation of prior results. Moreover, our demonstration that targeting DR6 attenuates MN death opens a unique opportunity to test therapeutic strategies aimed at blocking/slowing nerve terminal degeneration, a combination of effects that is required to ultimately extend muscle function and thus greatly improve the quality of life for patients with ALS. In conclusion, our presented combination of computational multi-modal analysis followed by experimental validation provides an effective proof-of-principle for a generalizable methodology directed at elucidating cell-to-cell communication mechanisms. Indeed—based on its ability to reduce the number of candidate hypotheses to a manageable number—use of proteomics data to prioritize candidates from one compartment may be effectively replaced by their gene expression, wherever possible. Although this will increase the number of candidate hypotheses, it will also allow direct application to a number of studies for which data are already available. Methods Humans and animals All of the studies with human postmortem tissues were approved by Columbia IRB Committee protocol AAAA8153. Procedures related to in vitro experimentation with cells produced or derived from mice B6SJL were approved by Columbia IACUC protocols AAAD8107, AAAL2502, and AAAN2050. All animal related procedures and euthanasia were approved by Columbia University’s Institutional Animal Care and Usage Committee. The entire mouse study was performed using B6.Cg-Tg(SOD1*G93A)1Gur/J mice (Cat# 004435; Jackson Laboratory) and NTg controls B6SJLF1/J mice (Cat# 100012; Jackson Laboratory). Primary astrocyte cultures AMLs from mice were prepared and genotyped as described 20 . In brief, postnatal day 1 pups from WT or mutSOD1 with B6SJL background were killed using approved IACUC protocol. The cortices were dissected out from the brain after carefully removing the meninges and deeper cortical structures. The cortex from each animal was dissociated by passing through 18 G needle (305196, BD PrecisionGlide) and plated on 75-cm2 flasks containing Dulbecco’s Modified Eagle Medium (DMEM, 12430-054, Gibco), 10% fetal bovine serum (FBS, A31604-02, Gibco) and 1% Pennicillin/Streptomycin (15140-122, Gibco). After 2 weeks, the cells were agitated (200 rpm, 6 h) to remove residual microgila. The cells were trypsinized and plated on 75-cm2 flask in astrocyte media described at a density of 100,000 cells/mL and incubated for 7 days. The ACMs were prepared from astrocyte cultures of different genotypes, collected, frozen in several aliquots, and then supplemented the day of the experiments 20 . Frozen aliquots from three individual human non-neurological disease control and three sALS astrocytes obtained from Re, et al. 9 were used for these investigations. In brief, fresh autopsied tissues were received from Columbia Medical Center Morgue or National Disease Research Interchange (NDRI) in (1:1) media containing DMEM (12430, Gibco) and F10 (11550, Gibco) and 1% gentamicin (G1272, Sigma-Aldrich). After removal of blood vessels and meninges, the tissue was cut into smaller sections. This crude suspension was incubated in Hanks' Balanced Salt Solution (14025, Thermofisher Scientific) containing 0.25% typsin (2500056, Thermofisher Scientific), 0.2 mg/mL ethylenediaminetetraacetic acid (EDTA) (ED2SS, Sigma-Aldrich), 1 mg/mL glucose (G8769, Sigma-Aldrich) and 0.1 mg/mL bovine pancreatic DNaseI (LK003172, Worthington) for 20 min at 37 °C. The cell suspension was then triturated with DMEM/F10 media containing 10% FBS (SH30070.02, Fisher Scientific) and 1% penicilliln/streptomycin (15140-163, Thermofisher Scientific). These cultures were passed through 0.22 μm filter and then plated on 75-cm2 flasks for 2 h at 37 °C to allow monocytes and macrophages to adhere. The supernatant from these cultures are transferred to another 75-cm2 flask coated with Poly-l-Lysine (15 μg/mL, Sigma-Aldrich) and media was changed after 48 h. The cells were allowed to grow until confluence and then dissociated and plated for experiments or frozed for additional experiments. As described previously 9 these human cell cultures expressed the astrocytic markers GFAP, vimentin and CD44 while they were immunonegative for microglial protein ionized calcium binding adaptor protein-1, oligodendritic marker 2’,3’-Cyclic nucleotide 3’-phosphodiesterase, and neuronal marker microtubule-associated protein 2. Primary MN cultures Primary SC neuronal cultures were prepared following the protocol described in ref. 20 using embryos from Tg mice expressing GFP driven by the mouse HB9 promoter 65 , and mice homozygote for a null mutation in DR6 (Stock # 2994, B6.129X.tm1Sjk/J from the Jackson Labs., MI, USA) kindly gifted by Genentech or for a null muation in p75NTR (Stock #2213, B6.129S4-Ngfrtm1Jae/J from the Jackson Labs. MI, USA). In brief, SCs dissected from embryos were incubated using 0.025% trypsin (15090-046, Gibco) and 1 mg/mL DNase (LK003172, Worthington) for 10 mins at 37 °C. The cells were then triturated and gently placed on 4% bovine serum albumin (BSA, A9418, Gibco) cushion and centrifuged for 300 × g for 5 min. The pellet was finally resuspended in Neurobasal medium (ME120079L2; Gibco), 2% B27 supplement (17504-044; Gibco), 2% HS (26050070; Gibco), 0.5 mM l-glutamine (25030-081; Gibco), 25 μM β-mercaptoethanol (ES-007-E; Millipore), and 1% P/S (15140-122; Gibco), supplemented with 1 ng/ml brain-derived neurotrophic factor (BDNF) (450-02-10UG; PeproTech), 0.5 ng/ml glial cell line-derived neurotrophic factor (GDNF) (450-10-10UG; PeproTech), 10 ng/ml ciliary neurotrophic factor (CNTF) (450-13-20UG; PeproTech), counted and plated at desired cell density as described below. Mouse ES-derived neuron cultures ES cells, derived from Tg mice expressing GFP driven by the mouse HB9 promoter 65 were differentiated into MNs using well-established differentiation protocol 20 . In short, mouse ES cell line Hb9::eGFP (a gift from Dr. Hynek Wicheterle) were plated on gelatinized 25 cm2 flask in DMEM media (SLM220B, Millipore Sigma) containing 15% ES-FBS (ES-009-B, Millipore Sigma), 1% penicillin/streptomycin (15140-122, Gibco), 1% glutamine (25030-081, Gibco), 1% non-essential amino acids (TMS-001-C, Millipore Sigma), 1% nucleosides (ES-008-D, Millipore Sigma), 1% β-mercaptoethanol (ES-007-E, Millipore Sigma), 1% Sodium pyruvate (S8636, Sigma-Aldrich) and 0.1% LIF (ESG1107, Millipore Sigma). After 48 h, the cells were trypsinized and plated at 1–2 million cells in 10 cm2 culture dish (353003, BD Falcon) containing αDFNK media containing 1:1 ratio of Advanced DMEM (12634, Gibco) and Neurobasal A (12349015, Thermofisher Scientific), 10% Knock Out Serum Replacement (ES-007-E, Millipore Sigma), 1% penicillin/streptomycin (15140-122, Gibco), 1% glutamine (25030-081, Gibco) and 1% β-mercaptoethanol (ES-007-E, Millipore Sigma for allowing non ES cells to adhere. The supernatant containing embryoid bodies was removed and transferred to Ultra low attachment dish (326, Corning). Next day, differentiating factors 1 µM retinoic acid (R2625, Sigma-Aldrich) and 0.25 µM Smoothened Agonist (566660, Calbiochem) were added for 72 h. Differentiated EBs were then transferred to fresh media for 24 h and dissociated using 0.5% trypsin EDTA (15400-054, Gibco) for 5 min at 37 °C. After halting the trypsinization with Horse serum (26050070, Thermofisher Scientific) the EBs were gently triturated using 1 mL tip and transferred to 4% BSA cushion, spun at 300 × g for 5 mins. The pellet was resuspended in ES media (similar to Primary Neuron media recipe described above except 10 ng/mL BDNF and 10 ng/mL GDNF). The cells were plated at density of 3000 GFP+ve cells/well in a 96-well plate. Gene silencing in astrocytes and MNs Primary astrocytes All shRNAs and EVs used in this study were from Sigma MISSION. For astrocytes, a suspension of 100,000 cells/mL was treated with 8 μg/mL of hexadimethrine bromide (107689; Sigma-Aldrich, St. Louis, MO), exposed to lentiviral particles at an MOI of 15 or 50, centrifuged (800 × g, 30 min, room temperature), and plated (20,000 cells/well on a 96-well plate or 80,000 cells/well in 24-well plates). After 48 h, cells were selected by adding 1 μg/ml puromycin (A11138); Invitrogen, Carlsbad, CA) for 4.5 days and placed in normal medium for 7 days before being co-cultured with MNs. For primary and ES-derived MN transduction, hexadimethrine bromide and puromycin were omitted. Primary neuronal cultures and co-cultures All shRNAs and EVs used in this study were constructed in pLKO.1-puro plasmids (Sigma MISSION®). This plasmid contains a puromycin resistance cassette inserted behind a human phosphoglycerate kinase eukaryotic promoter ( http://www.sigmaaldrich.com/life-science/functional-genomics-and-rnai/shrna/library-information/vector-map.html#pLKO ). For human SOD1 silencing, the clones TRCN0000009869, TRCN0000018344, TRCN0000039808, and TRCN0000039812 were used. For APP silencing, the clone TRCN0000054877 was used. For APLP1 silencing, clone TRCN0000106646 and for APLP2 silencing clone TRCN0000080052 were used. As controls, SHC001V (empty vector, EV) was used in all of the experiments, and SHC002H, a pLKO.1-puro non-mammalian shRNA control or SHC012V, a pLKO.1-puro-CMV-TagRFP™ positive control was used back to back with EV in several of these. For co-culture experiments, astrocytes were processed as described above and neurons were added on top of the astrocytes (added cell density: 12,000 in 96-well plates and 35,000 mouse primary MNs for 24-well plates). In each experiment, some of the wells did not receive MNs so as to serve as control wells to verify the efficiency of silencing by mRNA or protein extraction. Quantitative real-time PCR RNA samples were obtained from 24- or 96-well cell extraction (AM1931, RNAqueous Micro-Kit, Invitrogen, Carlsbad, CA). RNA (50–100 ng) was first reverse transcribed to cDNA (500 ng/μL) using Superscript II three-step reverse transcription PCR (11904-018, Invitrogen, Carlsbad, CA). A two-step qRT-PCR was carried out with Realplex 4 Mastercycler PCR System (Eppendorf, NY) using FAM-conjugated gene specific TaqMan® primers (Applied Biosystems, Invitrogen, Carlsbad, CA) in triplicate for each sample. Primers for GAPDH (Applied Biosystems, Invitrogen, Carlsbad, CA) were used as housekeeping control gene to normalize all target gene qRT-PCR results. qRT-PCR results were quantified by the Ct method following published recommendations (Schmittgen & Livak, 2008). Target genes for which Ct ≥35 were considered as unamplified and were thus referred to as non-detected. Pharmacological and recombinant treatments Treatment with SOD1 and E1/E2 recombinants, secretase inhibitors, domoic acid and agonistic anti-Fas antibody was done similar to our previously published article 9 . SOD1 recombinants (gift from J.-P. Julien) were added at the indicated concentarions and  E1 and E2 recombinant protein were added to ES-derived MNs at 3 µm on DIV2 monitored daily until DIV 7. β-secretase inhibitor AZ29 (gift from Genentech), α-secretase inhibitor TAPI (579051, Millipore Sigma), γ-secretase inhibitor (565771, Millipore Sigma), domoic acid (0269, Tocris), agonistic anti-Fas antibody (AB16982, Millipore Sigma), and anti-P75NTR neutrilizing antibody (Millipore clone ME20.4, cat number 05-446) were added at DIV 1 and DIV4. Cell survival and death were evaluated at the end of the DIV 7 by counting GFP-positive MNs. Immunodepletion experiments For immunodepletion experiments, anti-SOD1 antibody (Abcam 13498) was first incubated with magnetic agarose beads for 15 min at room temperature on rotation. After removing all of the unbound supernatant, beads were resupsended in antibody binding and washing buffer. Next, the beads were incubated with ACM from NTg and mutSOD1 astrocytes for 15 min at room temperature while rotating. After incubation with ACM, the antibody-bound beads were then separated using magnets to pull down the bound antibody complexes. The cleared CM was then applied to WT primary MNs and GFP+MNs remaining after DIV 7 were then counted as described above. Immunocytochemistry Cells were processed for immunocytochemistry as previously described 20 . Primary antibodies used here were: rabbit polyclonal anti-GFP (1:1000; A11122; Molecular Probes, Eugene, OR), rabbit polyclonal anti-GFP (1:1000; A11120, Molecular Probes, Eugene, OR), chicken polyclonal anti-MAP-2 (1:5000; ab5392; Abcam, Cambridge, England), mouse anti-SMI-32 for non-phosphorylated neurofilament heavy chain (1:500; SMI-32R; Covance, Princeton, NJ), mouse monoclonal anti-vimentin (1:500; V2258 Sigma-Aldrich, St. Louis, MO), and rabbit monoclonal anti-GFAP (1:500; a gift from Dr. James Goldman at Columbia University) antibodies. After overnight incubation (4 °C), coverslips were washed twice with Dulbecco’s phosphate-buffered saline (DPBS) (14190; Invitrogen, Carlsbad, CA) and incubated (1 h, room temperature) with the appropriate fluorescent-conjugated secondary antibodies (1:400; Molecular Probes, Eugene, OR). Coverslips were then washed and incubated with DAPI (D9564, Sigma-Aldrich, St. Louis, MO) at 0.1 μg/mL in phosphate-buffered saline (PBS) at room temperature for 5 min. After two DPBS washes, coverslips were mounted in Dako Cytomation Fluorescent mounting medium (S302380-2; Dako, Glostrup, Denmark). Neuromuscular junction analysis Mice at P90 and P125 were killed as per our IACUC approved protocol. The mice were perfused with 4% paraformaldehyde and the tibialis anterior muscle was dissected out. The muscle was stored in PBS overnight and then transferred to 30% sucrose for 24 h. After washing the muscle with PBS, the muscle was frozen on dry ice using clear OCT compound (Fisher Healthcare #4585). The blocks of muscle were then cryo-sectioned longitudinally into 20 µm slide-mounted sections. For staining, the slides were washed with PBS for 5 min and then blocked with 4% BSA+1% Triton-X 100 in PBS for 1 h at room temperature. The slides were then treated with Tetramethylrhodamine α-bungarotoxin (1:200; ThermoFisher Scientific T1175) and rabbit anti-vesicular acetylcholine transporter antibody (1:16,000; Covance CV1721) in 0.4% BSA+0.1% Triton-X 100 overnight at 4 °C in a humid chamber. The slides were washed and treated with donkey anti-rabbit Alexa 488 (1:400) and DAPI for 1.5 h at room temperature in a humid chamber. Finally, the slides were washed three times with PBS containing 0.2% Triton-X 100 and mounted with coverslips then and imaged on a Leica confocal microscope. At least 50 NMJs were counted for each muscle. Immunohistochemistry Mice were perfused using 4% paraformaldehyde. SC from P125 mice was dissected out post-fixed in 4% PFA overnight. SC was then transferred to 10% sucrose for at least 48 h at 4 °C. The SC was embedded in 7.5% gelatin and 10% sucrose at 37 °C and frozen on dry ice. 15 µm sections were cut on a cryostat. The sections were blocked with 10% Donkey serum and 0.2% Triton-X 100 in PBS. Primary antibody donkey anti-ChAT (1:125; Millipore AB144P) was added to the sections overnight in a humid chamber. The sections were washed with 0.2% Triton-X 100 in PBS followed by secondary antibody anti-goat Alexa 594 (1:250; Life Technologies A-11058) and 4′,6-diamidino-2-phenylindole (DAPI) for 3 h at room temperature. The slides were cover-slipped then mounted and imaged on a Leica Confocal microscope to count surviving MN numbers. ChAT-positive neuron on each hemi-section were counted. RNAscope assay L4/L5 section of SC tissue from WT mice was dissected out and frozen on dry ice in aluminum foil with minimum delay. The frozen tissue was embedded in OCT right away and stored in −80 °C for 24 h. Then, 20 μm thick sections from L4/L5 were cut with a cryostat and mounted on Frost-Free glass slides. The sections were stored at −20 °C for at least 24 h before processing them for RNAscope assay. RNAscope Fluorescent Multiplex Reagent kit (320513 for fresh frozen tissue and 320293 for Multiplex Reagent kit, Advanced Cell Diagnostics) was used for detecting ChAT, MMP9 and DR6 mRNA. The assay was performed exactly as per manufacturer’s recommended protocol. DAPI was used to label cell nuclei. The slides were allowed to dry overnight and imaged using Leica Confocal microscope. Behavior tests Loaded grid test: The loaded grid test described by Barneoud et al. 66 was used to assess limb muscle strength. One week before starting the test, the mice were pre-trained to hold the grid. Mice were weighed every time the assay was performed. Starting at P70, grids of increasing weight (10, 20, 30, and 40 g in ascending order) were given to the mice to hold for 30 s. Three trials were given for each weight with 15 s rest between each similar weight. Before moving to a different weight, 30 s rest period was given. The mice were suspended by their tails and given the weighted grid between their forelimbs and hindlimbs for 30 s. The best of three trials was used for statistics. Following equation was used to compute final score for each mice. $$Loaded\,grid\,score = \frac{{10\left( {t1} \right) + 20\left( {t2} \right) + 30\left( {t3} \right) + 40(t4)}}{{weight\,of\,mouse}}$$ where t1, t2, t3, and t4 represent the seconds the mouse held on to weights 10, 20, 30, and 40 g, respectively. Bioinformatics analysis The SEARCHIN pipeline implements a systematic and unbiased genome-wide approach to the identification of cross-compartment ligand–receptor pairs responsible for transducing signals from one cellular compartment to activate or inhibit MR proteins in the other. Specifically, the proof-of-concept SEARCHIN implementation described in this manuscript is aimed at identifying the ligand–receptor interactions that mediate transduction of death signals that determine MN demise when exposed to culture media from mutSOD1 astrocytes. The full source code used to run the experiment discussed in the manuscript, and a comprehensive tutorial on how to use the methodology and how the datasets were built. The source code is available under the GNU GPLv3 license at the following URL: http://www.github.com/Califano-lab/SEARCHIN . The tutorial is also available at the same link. The pipeline is conceptually summarized as follows. Ligands binding partner predictions To prioritize ligand–receptor binding interactions, we leveraged a large collection of PPIs inferred by the PrePPI algorithm 4 , 51 . PrePPI integrates 3D protein structural information as well as other sources of information, such as Gene Ontology and experimentally determined interactions. The current version of the PrePPI database contains 1,545,710 pairs of interacting proteins that are based on the human proteome, considered as high confidence predictions (probability > 0.5). To generate a null model, we used the full collection of potential PPI pairs that PrePPI may infer (i.e., 203,931,823 based on 20,259 proteins). Specifically, we generated a data driven empirical null model for each candidate interaction, by integrating the PrePPI scores from (a) binding partners of a ligand identified by MS, and (b) proteins predicted to interact with a differentially active receptor inferred by VIPER. In brief, for each possible interaction between two proteins P1 and P2, we built an empirical null model using the collection of all the PrePPI scores between P1 and all its putative binding partners present in PrePPI database (i.e., 20259 protein coding genes); the same was done for P2. The p value was generated with the empirical p value method that calculates p values from a set of observed test statistics (see ref. 67 ). This approach enabled helped us correct any bias owing to the number and specificity of individual PPIs, thus providing a more robust biological rationale for the empirical null model. We used an open source implementation of a graph database (i.e., the Neo4J Open Source Community Edition 3.2.5, http://neo4j.org/ ) to build a PPIs interactome from PrePPI inferences and the Cypher Query Language to access it programmatically through Python 3.6 scripts. An empirical p value was computed using the package q value-2.8.0 in the R environment version 3.4.1. This step of the SEARCHIN pipeline generated a list of 1478 candidate PPIs (Supplementary Table  2 ). To combine MS data (i.e., proteins from rat ACM) with PrePPI (i.e., human proteome), we used the programmatically accessible web services provided through BioMart 68 for biological identifier conversion (e.g., to translate orthologous genes between organisms or gene identifiers and their protein products). Membrane receptors inferred activity We used the ARACNe algorithm 1 to reverse-engineer a membrane protein-specific regulatory network from 437 mouse brain expression profiles (GSE10415). As regulators with fewer than 25 targets were shown to reduce inference accuracy, as comprehensively benchmarked in ref. 2 , only plasma membrane proteins, annotated in the Gene Ontology 69 database as “signal transduction activity” (GO:0005886 and GO:0007165), with ≥25 target genes (among 16,589 possible ones) were included in the network, for a total of 873 murine receptor and 376,672 transcriptional interactions. The plasma membrane interactome is available through FigShare at the following permanent URL: https://doi.org/10.6084/m9.figshare.6387671 . The VIPER algorithm 2 was used to infer changes in plasma membrane signaling protein activity based on the mutSOD1 gene expression signature and mouse brain plasma membrane regulatory model. Because the inferred activity is based on the transcriptional programs induced by the treatment, the algorithm will identify the specific receptors as well as other regulatory proteins associated with similar transcriptional programs. We used the list of differentially active receptors, based on the rank-ordered p values (from the most to least significant) as another source of evidence to further prioritize the candidate paracrine interactions. The list of candidate receptors is shown in Supplementary Table  3 . NF-κB1 modulators inference The CINDy algorithm 3 was used to predict the most likely modulators of NF-κB1 activity among the membrane receptor proteins. We used gene expression profiles from 437 mouse brain samples (GSE10415) as input to CINDy. The null model was generated using the number of triplets (i.e., receptor-TF-target) identified by the algorithm for NF-κB1. Specifically, only the triplets that had a Conditional Mutual Information (CMI) greater than or equal to the average CMI of the data set were considered. The list of candidate modulators is shown in Supplementary Table  4 . Scores integration To combine the scores from the different sources of evidence, we used the Robust Rank Aggregation method for meta-analysis 27 . From the R package RobustRankAggreg version 1.1, we used the function “aggregateRanks” with method “RRA”. The final list of ranked ligand–receptor pairs is shown in Table  1 . For validation, we selected the first pair, i.e., APP/TNFRSF21, with the highest “experimental score” as reported in PrePPI DB, whose receptor was also shown to be differentially activated. SEARCHIN code and implementation Although the present manuscript provides the outline of the algorithm, its full source code is available at the following URL: https://github.com/califano-lab/SEARCHIN . Treatment of ACM ACM were passed through Q or S columns (Pierce Strong Anion Exchange Spin column, Thermo Scientific #90011; Cation Exchange Spin column, Thermo Scientific #90009). Sample flow-through from the column after passing the sample was discarded. Molecules bound to the column were eluted by increasing concentrations of NaCl (0.5, 1, and 2 M respectively). The eluate was desalted using spin columns (Amicon Ultra-15 Centrifugal Filter Unit with Ultracel-3 membrane, Millipore, Cat. # UFC900396). The elute was brought up to original volume and supplemented with HS, B27, P/S, l-glutamine, and β-mercaptoethanol and then applied to primary SC MN cultures. In another set of experiments, ACM media from NTg and mutSOD1 were passed through cutoff filters of 5, 10, 30, 50, and 100 kDa to separate potential toxic moieties based on molecular weight. The components of the ACM retained on the filters were then resuspended in fresh MN culture media before being applied to primary SC MN cultures to examine their effect on viability. For heat inactivation, ACM from NTg and mutSOD1 astrocytes were incubated in a water bath at 95 °C for 15 or 30 min, cooled down to 37 °C and then resupplemented with horse serum (2%) and trophic factors as described in ref. 20 before being applied on mouse primary SC MN cultures. To eliminate proteins, ACM from NTg and mutSOD1 were first adjusted to a pH 2.5 with 1 N HCl and then incubated with pepsin at 37 °C for 2 hours. At the end of the incubation the pH was readjusted to 7.4 with 1 N NaOH and ACM were re-supplemented as described above. Control incubations identical to those above were performed except that pH was not adjusted and ACM were kept at room temperature to prevent pepsin activity. For charcoal treatment, ACM were incubated for 5 min at RT with 50 mg/mL charcoal (Norit: activated and neutralized; Sigma 97876), then centrifuged at 11,500 × g for 3 min. The supernatants were then filtered and re-supplemented as described above before treatment. For lipid depletion, ACM were mixed with chloroform (v/v), vortexed for 30 s and centrifuged for 5 min at 750 × g. The chloroform phase was then removed and this extraction was repeated twice. The aqueous phase was then poured in a petri dish and placed in an incubator at 45 °C for 30 min to remove all traces of chloroform. Finally, the volumes were readjusted to the original volume of the aqueous phase with sterile water, filtered, and re-supplemented as described in ref. 20 . Mass spectrometry ACM derived from rat astrocytes was analyzed by Gel LC-MS/MS as described previously 24 . In brief, ACM were resolved on NuPAGE 10% Bis-Tris gels (Invitrogen, Carlsbad, CA). Proteins were visualized by Colloidal blue (Invitrogen, Carlsbad, CA). Uniform (2 mm) slices were cut into 1 × 1 mm cubes followed by in-gel digestion with trypsin. Tryptic digests were analyzed on hybrid LTQ-Orbitrap mass spectrometer (Thermofisher Scientific, San Jose, CA) coupled with a NanoLC pump (Eksigent Technologies, Livermore, CA). Peptide and protein identification and quantification were performed using MaxQuant. The raw intensities of peptides in MS1 were transformed to log2 values, which were then used to calculate the difference in relative levels of proteins between mutSOD1 and NTg. Cell counting and statistics Results correspond to independent experiments and each experiment is the average of 3–6 coverslips or 96-wells per time point and per condition. For mouse ES-MN number with co-culture, MN number was assessed by automatic quantification using the Metamorph Imaging System software (Universal Imaging Corporation, Downingtown, PA). For primary cultures, MNs were counted manually using the Nikon Epifluorescent microscope. The investigator counting the MN numbers was always blind to the treatment conditions. Differences between means were analyzed by a two-tailed Student’s t test. Photoshop v19 was used for preparing figures. NIH Fiji ImageJ was used for measuring intensity for RNAscope analysis. Differences among means were analyzed by one- or two-way analysis of variance (ANOVA) with the different types of astrocytes, MNs or treatments as independent factors. When the ANOVA showed significant differences, pair-wise comparisons between means were tested using Sidak post hoc testing, unless stated otherwise. In all analyses, the null hypothesis was rejected at the 0.05 level. All statistical analyses were performed using SigmaPlot for Windows (version 12.0; Jandel Corp., San Rafael, CA) and software GraphPad Prism version 7 (San Diego, CA). Loaded grid test data were fitted by a non-linear iterative least squares method provided by the software GraphPad Prism. Then, the two different curves were compared by the extra sum-of-squares F test 70 . Survival statistics was performed by Kaplan–Meyer analysis also with GraphPad Prism. Whenever suitable, statistical results include 95% confidence interval and the Cohen’s d effect size. Reporting summary Further information on research design is available in the  Nature Research Reporting Summary linked to this article. Data availability The authors declare that all the main data supporting the findings of this study are available within the article and its Supplementary Information files. All the data sets that were used as input for the study and the code to reproduce the manuscript findings are available at the following citable link: https://doi.org/10.5281/zenodo.4037259 71 . The mass spectrometry proteomics data are available via ProteomeXchange at https://www.ebi.ac.uk/pride/ with the data set identifiers PXD021773. Code availability The authors declare that the SEARCHIN source code and tutorial are available under the GNU GPLv3 license at the following URL: http://www.github.com/Califano-lab/SEARCHIN 71 . References 1. Margolin, A. A. et al. ARACNE: an algorithm for the reconstruction of gene regulatory networks in a mammalian cellular context. BMC Bioinformatics 7, S7 (2006). 2. Alvarez, M. J. et al. Functional characterization of somatic mutations in cancer using network-based inference of protein activity. Nat. Genet. 48, 838–847 (2016). 3. Giorgi, F. M. et al. Inferring protein modulation from gene expression data using conditional mutual information. Plos ONE 9, e109569 (2014). 4. Zhang, Q. C., Petrey, D., Garzon, J. I., Deng, L. & Honig, B. PrePPI: a structure-informed database of protein-protein interactions. Nucleic Acids Res. 41, D828–D833 (2013). 5. Rowland, L. P., Mitsumoto, H. & Przedborski, S. In Merritt’s Neurology (eds L. P. Rowland & T. A. Pedley) Ch. 128, 802–808 (Lippincott, Williams & Wilkins, 2010). 6. Garden, G. A. & La Spada, A. R. Intercellular (mis)communication in neurodegenerative disease. Neuron 73, 886–901 (2012). 7. Renton, A. E., Chio, A. & Traynor, B. J. State of play in amyotrophic lateral sclerosis genetics. Nat. Neurosci. 17, 17–23 (2014). 8. Ikiz, B. et al. The regulatory machinery of neurodegeneration in in vitro models of amyotrophic lateral sclerosis. Cell Rep. 12, 335–345 (2015). 9. Re, D. B. et al. Necroptosis drives motor neuron death in models of both sporadic and familial ALS. Neuron 81, 1001–1008 (2014). 10. Di Giorgio, F. P., Boulting, G. L., Bobrowicz, S. & Eggan, K. C. Human embryonic stem cell-derived motor neurons are sensitive to the toxic effect of glial cells carrying an ALS-causing mutation. Cell Stem Cell 3, 637–648 (2008). 11. Di Giorgio, F. P., Carrasco, M. A., Siao, M. C., Maniatis, T. & Eggan, K. Non-cell autonomous effect of glia on motor neurons in an embryonic stem cell-based ALS model. Nat. Neurosci. 10, 608–614 (2007). 12. Meyer, K. et al. Direct conversion of patient fibroblasts demonstrates non-cell autonomous toxicity of astrocytes to motor neurons in familial and sporadic ALS. Proc. Natl Acad. Sci. USA 111, 829–832 (2014). 13. Haidet-Phillips, A. M. et al. Astrocytes from familial and sporadic ALS patients are toxic to motor neurons. Nat. Biotechnol. 29, 824–828 (2011). 14. Marchetto, M. C. et al. Non-cell-autonomous effect of human SOD1 G37R astrocytes on motor neurons derived from human embryonic stem cells. Cell Stem Cell 3, 649–657 (2008). 15. Fritz, E. et al. Mutant SOD1-expressing astrocytes release toxic factors that trigger motoneuron death by inducing hyperexcitability. J. Neurophysiol. 109, 2803–2814 (2013). 16. Cassina, P. et al. Mitochondrial dysfunction in S0D1(G93A)-bearing astrocytes promotes motor neuron degeneration: Prevention by mitochondrial-targeted antioxidants. J. Neurosci. 28, 4115–4122 (2008). 17. Pehar, M. et al. Complexity of astrocyte-motor neuron interactions in amyotrophic lateral sclerosis. Neurodegener. Dis. 2, 139–146 (2005). 18. Pehar, M. et al. Astrocytic production of nerve growth factor in motor neuron apoptosis: implications for amyotrophic lateral sclerosis. J. Neurochem. 89, 464–473 (2004). 19. Barbeito, L. H. et al. A role for astrocytes in motor neuron loss in amyotrophic lateral sclerosis. Brain Res. Brain Res. Rev. 47, 263–274 (2004). 20. Nagai, M. et al. Astrocytes expressing ALS-linked mutated SOD1 release factors selectively toxic to motor neurons. Nat. Neurosci. 10, 615–622 (2007). 21. Papadeas, S. T., Kraig, S. E., O’Banion, C., Lepore, A. C. & Maragakis, N. J. Astrocytes carrying the superoxide dismutase 1 (SOD1G93A) mutation induce wild-type motor neuron degeneration in vivo. Proc. Natl Acad. Sci. USA 108, 17803–17808 (2011). 22. Yamanaka, K. et al. Astrocytes as determinants of disease progression in inherited amyotrophic lateral sclerosis. Nat. Neurosci. 11, 251–253 (2008). 23. Urushitani, M. et al. Chromogranin-mediated secretion of mutant superoxide dismutase proteins linked to amyotrophic lateral sclerosis. Nat. Neurosci. 9, 108–118 (2006). 24. Greco, T. M., Seeholzer, S. H., Mak, A., Spruce, L. & Ischiropoulos, H. Quantitative mass spectrometry-based proteomics reveals the dynamic range of primary mouse astrocyte protein secretion. J. Proteome Res. 9, 2764–2774 (2010). 26. Liddelow, S. A. et al. Neurotoxic reactive astrocytes are induced by activated microglia. Nature 541, 481–487 (2017). 27. Kolde, R., Laur, S., Adler, P. & Vilo, J. Robust rank aggregation for gene list integration and meta-analysis. Bioinformatics 28, 573–580 (2012). 28. Vosa, U., Kolde, R., Vilo, J., Metspalu, A. & Annilo, T. Comprehensive meta-analysis of microRNA expression using a robust rank aggregation approach. Methods Mol. Biol. 1182, 361–373 (2014). Xenarios, I. et al. DIP: the database of interacting proteins. Nucleic Acids Res. 28, 289–291 (2000). 31. Breitkreutz, B. J., Stark, C. & Tyers, M. The GRID: the General Repository for Interaction Datasets. Genome Biol. 4, R23 (2003). 32. Nikolaev, A., McLaughlin, T., O’Leary, D. D. & Tessier-Lavigne, M. APP binds DR6 to trigger axon pruning and neuron death via distinct caspases. Nature 457, 981–989 (2009). 33. Kallop, D. Y. et al. A death receptor 6-amyloid precursor protein pathway regulates synapse density in the mature CNS but does not contribute to Alzheimer’s disease-related pathophysiology in murine models. J. Neurosci. 34, 6425–6437 (2014). 34. Olsen, O. et al. Genetic analysis reveals that amyloid precursor protein and death receptor 6 function in the same pathway to control axonal pruning independent of beta-secretase. J. Neurosci. 34, 6438–6447 (2014). 35. De Strooper, B., Vassar, R. & Golde, T. The secretases: enzymes with therapeutic potential in Alzheimer disease. Nat. Rev. Neurol. 6, 99–107 (2010). 36. Coburger, I. et al. Analysis of the overall structure of the multi-domain amyloid precursor protein (APP). Plos ONE 8, e81926 (2013). 37. Vassar, R., Kovacs, D. M., Yan, R. & Wong, P. C. The beta-secretase enzyme BACE in health and Alzheimer’s disease: regulation, cell biology, function, and therapeutic potential. J. Neurosci. 29, 12787–12794 (2009). 38. Li, Q. & Sudhof, T. C. Cleavage of amyloid-beta precursor protein and amyloid-beta precursor-like protein by BACE 1. J. Biol. Chem. 279, 10542–10550 (2004). 39. Zhao, H. et al. Impaired c-Jun amino terminal kinase activity and T cell differentiation in death receptor 6-deficient mice. J. Exp. Med. 194, 1441–1448 (2001). 40. Hu, Y. et al. A DR6/p75(NTR) complex is responsible for beta-amyloid-induced cortical neuron death. Cell Death Dis. 4, e579 (2013). 41. Watanabe, T. et al. The p75 receptor is associated with inflammatory thermal hypersensitivity. J. Neurosci. Res. 86, 3566–3574 (2008). 42. Simon, C. M. et al. Converging mechanisms of p53 activation drive motor neuron degeneration in spinal muscular atrophy. Cell Rep. 21, 3767–3780 (2017). 43. Fletcher, E. V. et al. Reduced sensory synaptic excitation impairs motor neuron function via Kv2.1 in spinal muscular atrophy. Nat. Neurosci. 20, 905–916 (2017). 44. Liu, J., Shinobu, L. A., Ward, C. M., Young, D. & Cleveland, D. W. Elevation of the Hsp70 chaperone does not effect toxicity in mouse models of familial amyotrophic lateral sclerosis. J. Neurochem. 93, 875–882 (2005). 45. Kanning, K. C., Kaplan, A. & Henderson, C. E. Motor neuron diversity in development and disease. Annu. Rev. Neurosci. 33, 409–440 (2010). 46. Kaplan, A. et al. Neuronal matrix metalloproteinase-9 is a determinant of selective neurodegeneration. Neuron 81, 333–348 (2014). 47. Raoul, C. et al. Motoneuron death triggered by a specific pathway downstream of Fas. Potentiation by ALS-linked SOD1 mutations. Neuron 35, 1067–1083 (2002). 48. Ding, H. et al. Quantitative assessment of protein activity in orphan tissues and single cells using the metaVIPER algorithm. Nat. Commun. 9, 1471 (2018). 49. Tyzack, G. E. et al. A neuroprotective astrocyte state is induced by neuronal signal EphB1 but fails in ALS models. Nat. Commun. 8, 1164 (2017). 50. Taylor, J. P., Brown, R. H. Jr., Cleveland, D. W. & Decoding, A. L. S. from genes to mechanism. Nature 539, 197–206 (2016). 51. Zhang, Q. C. et al. Structure-based prediction of protein-protein interactions on a genome-wide scale. Nature 490, 556–560 (2012). 52. Szklarczyk, D. et al. The STRING database in 2017: quality-controlled protein-protein association networks, made broadly accessible. Nucleic Acids Res. 45, D362–D368 (2017). 53. Marik, S. A., Olsen, O., Tessier-Lavigne, M. & Gilbert, C. D. Death receptor 6 regulates adult experience-dependent cortical plasticity. J. Neurosci. 33, 14998–15003 (2013). 54. Marik, S. A., Olsen, O., Tessier-Lavigne, M. & Gilbert, C. D. Physiological role for amyloid precursor protein in adult experience-dependent plasticity. Proc. Natl Acad. Sci. USA 113, 7912–7917 (2016). 55. Berndt, J. D. et al. Mindbomb 1, an E3 ubiquitin ligase, forms a complex with RYK to activate Wnt/beta-catenin signaling. J. Cell Biol. 194, 737–750 (2011). 56. Romani, P., Ignesti, M., Gargiulo, G., Hsu, T. & Cavaliere, V. Extracellular NME proteins: a player or a bystander? Lab. Invest. 98, 248–257 (2018). 57. Canzio, D. & Maniatis, T. The generation of a protocadherin cell-surface recognition code for neural circuit assembly. Curr. Opin. Neurobiol. 59, 213–220 (2019). 58. Yin, W. et al. Expression profile analysis identifies a novel five-gene signature to improve prognosis prediction of glioblastoma. Front. Genet. 10, 419 (2019). 59. Lu, W., Yamamoto, V., Ortega, B. & Baltimore, D. Mammalian Ryk is a Wnt coreceptor required for stimulation of neurite outgrowth. Cell 119, 97–108 (2004). 60. Esch, F. S. et al. Cleavage of amyloid beta peptide during constitutive processing of its precursor. Science 248, 1122–1124 (1990). 61. Bryson, J. B. et al. Amyloid precursor protein (APP) contributes to pathology in the SOD1(G93A) mouse model of amyotrophic lateral sclerosis. Hum. Mol. Genet. 21, 3871–3882 (2012). 62. Rabinovich-Toidman, P., Becker, M., Barbiro, B. & Solomon, B. Inhibition of amyloid precursor protein beta-secretase cleavage site affects survival and motor functions of amyotrophic lateral sclerosis transgenic mice. Neurodegener. Dis. 10, 30–33 (2012). 63. Xu, K., Olsen, O., Tzvetkova-Robev, D., Tessier-Lavigne, M. & Nikolov, D. B. The crystal structure of DR6 in complex with the amyloid precursor protein provides insight into death receptor activation. Genes Dev. 29, 785–790 (2015). 64. Gould, T. W. et al. Complete dissociation of motor neuron death from motor dysfunction by Bax deletion in a mouse model of ALS. J. Neurosci. 26, 8774–8786 (2006). 65. Wichterle, H., Lieberam, I., Porter, J. A. & Jessell, T. M. Directed differentiation of embryonic stem cells into motor neurons. Cell 110, 385–397 (2002). 66. Barneoud, P., Lolivier, J., Sanger, D. J., Scatton, B. & Moser, P. Quantitative motor assessment in FALS mice: a longitudinal study. Neuroreport 8, 2861–2865 (1997). 67. Storey, J. D. & Tibshirani, R. Statistical significance for genomewide studies. Proc. Natl Acad. Sci. USA 100, 9440–9445 (2003). 68. Smedley, D. et al. The BioMart community portal: an innovative alternative to large, centralized data repositories. Nucleic Acids Res. 43, W589–W598 (2015). 69. Ashburner, M. et al. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat. Genet. 25, 25–29 (2000). Acknowledgements We thank Marc Tessier-Lavigne for his support and critical comments on this manuscript and for providing reagents for assays on DR6, Genentech for providing us with necessary reagents and mouse lines to study DR6, Lawrence S. Shapiro for guidance with the liquid chromatography work, Dimitar Nikolov and Jean-Pierre Julien for generously providing recombinant E2 fragment and SOD1 recombinants and Radhika Pradhan, Vernice Jackson-Lewis, James Caicedo and Norma Romero for technical expertize. Drs. M. De Rosa and C. Doege (Molecular Biology/Molecular Genetics Core of the New York Nutr

Predict your next investment

The CB Insights tech market intelligence platform analyzes millions of data points on venture capital, startups, patents , partnerships and news mentions to help you see tomorrow's opportunities, today.

Expert Collections containing Searchin

Expert Collections are analyst-curated lists that highlight the companies you need to know in the most important technology spaces.

Searchin is included in 1 Expert Collection, including HR Tech.

H

HR Tech

3,793 items

HR tech startups are helping companies manage critical pain points in HR processes such as recruitment, automation, career development, compensation, and benefits management, through a mix of software and services.

Searchin Web Traffic

Rank
Page Views per User (PVPU)
Page Views per Million (PVPM)
Reach per Million (RPM)
CBI Logo

Searchin Rank

CB Insights uses Cookies

CBI websites generally use certain cookies to enable better interactions with our sites and services. Use of these cookies, which may be stored on your device, permits us to improve and customize your experience. You can read more about your cookie choices at our privacy policy here. By continuing to use this site you are consenting to these choices.