Introduction
Conventional wisdom has held that the chronic neuroinflammation associated with LOAD may be a secondary or even protective event that occurs in response to Aβ deposition and may occur only in late stages of AD. However, recent genetic and genomic approaches, as well as computational strategies, have converged on immune-inflammatory pathways as risk factors and as key events in the pathogenesis of late-onset sporadic Alzheimer’s disease (LOAD) [
19]. Moreover, correlation between inflammatory genes and clinical presentation of previously asymptomatic cerebral amyloidosis (ACA) indicates a role for inflammation and microglia in the progression from ACA to the earliest stages of mild cognitive impairment (MCI) and/or mild clinical AD. Among the genes implicated by the largest available genome-wide association studies [
43], one-third is either unique to, or enriched in, microglia. Recently identified mutations and variants in genes encoding important immune receptors including
CD33,
CR3 (Complement Receptor 3), and
TREM2 (Triggering Receptor Expressed On Myeloid Cells 2), have been genetically linked to LOAD risk, highlighting the potential role of a dysregulated immune response in an early, and perhaps causative role in AD pathogenesis. Unlike autosomal dominant familial Alzheimer’s mutations that promote elevation of the Aβ42:40 ratio or of other variant hyperaggregatable Aβ species, these AD risk factors specify some of the cell surface signal transduction pathways that modulate the phagocytic machinery of microglia.
TYROBP (TYROsine kinase Binding Protein) (also known as DAP12), is a microglial transmembrane signaling polypeptide that contains an immunoreceptor phosphotyrosine-based activation motif (ITAM) in its cytoplasmic domain and is a direct partner/adaptor for immune receptors, including TREM2, CR3, and SIRPβ1 (Signal Regulatory Protein β1) all of which are independently linked to, or associated with, LOAD [
5,
7,
23,
55,
86]. Interaction of TYROBP with its partners forms phagocytosis “active zones” (known as phagocytic synapses) on the surface of microglia. In preparation for phagocytosis, there is a respiratory burst that generates reactive oxygen species (ROS) and appears to involve an interaction between TYROBP and CR3, which in turn interacts with complement component C3 associated with nearby neurites. Mice lacking the complement receptor CR3 or expressing defective TYROBP show reduced ROS production and apoptosis [
77]. A recent report demonstrates that the complement pathway can mediate the toxic effects of soluble Aβ on synapses, and that overactivation of this pathway in AD leads to excessive synapse pruning and early synapse loss [
25]. Since the discovery of a link between mutations of
TREM2 and AD, several studies have emerged regarding the role of a loss of function of TREM2 in AD. While these studies have some conflicting results, the most consistent observation is that either
Trem2 deficiency or
Tyrobp deficiency can cause reduced recruitment of microglial cells around Aβ plaques. The impact of this reduction in microglia per plaque was interpreted as deleterious in
Trem2 haploinsufficient and
Trem2 deficient mice.
Through a multi-scale integrated computational approach, we and two other independent groups [
12,
48,
86] have previously reported
TYROBP as a network hub or driver gene in LOAD. Additionally, missense mutations in
TYROBP have been recently reported as risk factors for AD [
61]. Evidence associating
TYROBP to LOAD notwithstanding, it is important to recognize that most
TYROBP mutations (as well as
TREM2 mutations) represent loss-of-function mutations that result not in AD but in an osteopathy/encephalopathy known as Nasu–Hakola disease (NHD) [
59]. One formulation of these data is that the pathogenic mechanism(s) of loss-of-function (nonsense) mutations in
TYROBP associated with NHD may cause molecular events that differ from those associated with missense polymorphisms that increase the risk for AD.
Herein, we report the effects of a constitutively null mutation in
Tyrobp on the phenotype of an
APP/PSEN1 mouse model of AD. In the
Tyrobp-null mouse, there is a deletion of exons 3 and 4 resulting in loss of function of the TYROBP protein by deletion of the transmembrane region and part of the cytoplasmic region including the first tyrosine of the ITAM motif [
2]. The
APP/PSEN1 mouse model [
29] expresses
APP
KM670/671NL
/PSEN1
Δexon9
in neurons and accumulates in the interstitial spaces of the brain fibrillar amyloid that goes on to form typical amyloid plaques accompanied by neuritic dystrophy, age-dependent synaptic loss without neuronal loss, and abnormalities in spatial memory [
25,
29,
40,
41]. Since
TYROBP expression is increased in the LOAD brain [
86], we hypothesized that the
APP/PSEN1 phenotype may be improved in the presence of reduced TYROBP levels. Since
Tyrobp is not expressed in neurons, our observations in this report describe non-cell autonomous effects wherein signals arising from microglia perturb the homeostasis of nearby neurons or nerve terminals or the pathophysiology of evolving structural intraneuronal or extracellular Alzheimer’s pathology.
Methods
Mouse husbandry
The experimental procedures were conducted in accordance with NIH guidelines for animal research and were approved by the Institutional Animal Care and Use Committee (IACUC) at Icahn School of Medicine at Mount Sinai. APP
KM670/671NL
/PSEN1
Δexon9
(APP/PSEN1) and Tyrobp knockout (KO) mice were obtained from Jackson Laboratories and Taconic/Merck Laboratory, respectively. APP/PSEN1 mice were crossed with Tyrobp KO mice to obtain APP/PSEN1 mice heterozygous or KO for Tyrobp. Four-month-old male and female mice were killed by decapitation. One hemisphere was collected for immunohistochemical analysis. The second hemisphere was collected for transcriptomic and biochemical analyses.
Immunohistochemical and biochemical analyses
Immunohistochemical and biochemical characterization were performed as previously described [
40,
41,
44,
76]. For biochemical analysis, hemibrains were processed via differential detergent solubilization to produce TBS-soluble, Triton-X-soluble, and formic-acid soluble Aβ fractions. For analysis of native oligomeric Aβ peptides, 2 μl protein samples from the TBS-soluble fraction were spotted onto activated/pre-wetted PVDF membrane (0.22 μm; Millipore, Billerica, MA). Membranes were incubated with rabbit pAb A11 (anti-prefibrillar oligomers, 0.5 μg/ml), rabbit pAb OC (anti-fibrillar oligomers and fibrils; 0.25 μg/ml), and mouse mAb Nu-4 (anti-oligomers; 1 μg/ml) [
44,
76]. Normalization to total APP/Aβ signal was achieved by detection of human APP transgene metabolites with the mouse pAb 6E10 antibody (1:1000; Covance, Princeton, NJ). To quantify total Aβ levels, human/rat Aβ 1–40/1–42 ELISA kits (Wako) were used according to the manufacturer’s instructions.
For immunohistochemistry, 30 µm thick sagittal sections were incubated with the following antibodies: rabbit anti-Iba1 (1:500; Wako, Richmond, VA), mouse anti-6E10 (1:1000; Covance, Princeton, NJ), and rat anti-CD68 (1:200, mca1957, AbD Serotec BioRad). Sections were then incubated with the appropriate secondary antibody: anti-rabbit Alexa Fluor 488 or Alexa Fluor 568 (1:400; Invitrogen, Carlsbad, CA), anti-mouse Alexa Fluor 568 (1:400; Invitrogen, Carlsbad, CA), and anti-rat Alexa Fluor 488 (1:400; Invitrogen, Carlsbad, CA) antibodies. ThioflavinS (Sigma-Aldrich, T1892, 1% w/v stock solution) was used for labeling amyloid deposits.
For measuring microglia number, Iba1-immunolabeled sections were thresholded and particles analyzed with Fiji (v2.0.0). Sizes of 6E10 immunoreactive plaques and fluorescent intensities were analyzed with Fiji (v2.0.0). The regions of interest were determined by manual tracing. Thioflavin S fluorescence intensity and circularity were analyzed as described [
85].
For immunoblotting, membranes were incubated with either anti-CD68 (1:1000, mca1957, AbD Serotec BioRad), anti-phospho-Tau pSer202/Thr205 (1:1000; MN1020, Thermo Fisher Scientific, Waltham, MA), anti-Tau (1:1000; MN1000, Thermo Fisher Scientific, Waltham, MA), anti-Synaptophysin (1:200; ab16659, Abcam, Cambridge, MA), anti-Lamp1 (1:200; ab24170, Abcam, Cambridge, MA), anti-C3 (1:50; ab11862, Abcam, Cambridge, MA), and anti-GAPDH (1:5000; sc32233, Santa Cruz, Dallas, TX) antibodies. Integrated density of immunoreactive bands was measured using MultiGauge Software (FujiFilm). At least two independent western blot analyses were performed and normalized using APP/PSEN1 female mice as controls.
Behavior analysis
The Barnes Maze test was performed using a standard apparatus [
3,
74]. Four-month-old mice were transported from their cage to the center of the platform via a closed starting chamber where they remained for 10 s prior to exploring the maze for 3 min. Mice failing to enter the escape box within 3 min were guided to the escape box by the experimenter, and the latency was recorded as 180 s. Mice were allowed to remain in the escape box for 1 min before the next trial. Two trials per day during 4 consecutive days were performed. The platform and the escape box were wiped with 70% ethanol after each trial to eliminate the use of olfactory cues to locate the target hole. All trials were recorded by video camera and analyzed with ANY-maze video tracking software (Stoelting Co, Wood Dale, USA).
Field electrophysiology
Coronal brain slices containing the hippocampal formation were prepared as previously described [
17]. Following anesthesia with isoflurane, brains were rapidly removed and cut into 400 µm thick coronal sections using a vibratome VT1000S (Leica Microsystems, Germany). Brain slices were incubated at room temperature for ≥3 h in a physiologic ACSF containing 120 mM NaCl, 3.3 mM KCl, 1.2 mM Na
2HPO
4, 26 mM NaHCO
3, 1.3 mM MgSO
4, 1.8 mM CaCl
2, 11 mM Glucose (pH 7.4) and then transferred to a recording chamber perfused with ACSF at a flow rate of ~2 mL/min; experiments were performed at 28.0 ± 0.1 °C. Recordings were acquired with a GeneClamp 500B amplifier (Axon Instruments, Union City, CA) and Digidata 1440A (Molecular Devices, Sunnyvale, CA). All signals were low-pass filtered at 2 kHz and digitized at 10 kHz. For extracellular field recordings, a patch-type pipette was filled with ACSF and placed in the middle third of stratum radiatum in area CA1. Field excitatory postsynaptic potentials (fEPSPs) were evoked by activating Shaffer Collaterals with a Concentric Bipolar Electrode stimulator (FHC, St Bowdoin, ME) placed in the middle third of stratum radiatum 150–200 µm away from the recording pipette. Square-wave current pulses (60 ms pulse width) were delivered through a stimulus isolator (Isoflex, AMPI). Input–output curves were generated by a series of stimuli in 0.1 mA steps. Paired-pulse facilitation was measured by delivering two stimuli at 20, 50, and 100 ms inter-stimulus intervals. Each inter-stimulus interval was repeated three times and the resulting potentials were averaged. The paired-pulse ratio was calculated by dividing the slope of the second EPSP by the slope of the first EPSP. All results were analyzed by ANOVAs followed by Tukey post hoc tests. Baseline recordings (stable for 20 min) were made every 30 s using stimuli that yielded a response equal to 50% of spike threshold. LTD was induced using a 1-Hz train of 900 bursts, each burst containing three stimuli delivered at 20 Hz, using stimulus strength just superthreshold for evoking a population spike during baseline.
Molecular biological analyses
RNA isolation, library preparation, differential expression analysis and gene set enrichment analyses were performed as described [
6,
26,
27,
64,
66,
67].
Computational screen of TYROBP regulating compounds
Drug-induced gene expression fold change was obtained from the Connectivity Map database [
42], which consists of 6100 individual experiments, representing 1309 unique compounds. The 6100 individual expression profiles were merged into a single representative signature for the 1309 unique compounds, according to the prototype-ranked list method [
28]. Each compound was scored according to the rank of
Tyrobp expression fold change within its signature. Compounds were ranked in descending order of
Tyrobp expression fold change and used for a secondary enrichment analysis of drug-target associations. For each compound in the drug signature library, referenced drug–target associations [
45,
83] and predicted off-targeting [
36,
37] were collected. For each of these features, we calculated a running sum enrichment score, reflecting whether that feature was over-represented among the compounds at the top (associated with
Tyrobp upregulation) or at the bottom (associated with
Tyrobp down-regulation). Two-tailed
p values were based on comparison with 10,000 permuted null scores, generated from randomized drug target sets that contain an equivalent number of compounds to the true set under evaluation, and adjusted using the Benjamini–Hochberg method [
6]. Computational screening and chemogenomic enrichment analysis were performed using the R project for statistical computing version 3.2.5 [
62].
Data and software availability
Gene expression data generated contributing to the described study will be deposited electronically to the Synapse Web Portal (
https://www.synapse.org) in accordance with data sharing policies established by the NIH Accelerating Medicine Partnership (AMP) AD consortium. Specific software will also be made available upon request.
Discussion
Association of
TYROBP with LOAD arose via a multiscale computational network approach [
86]. The physical interaction between TREM2 and TYROBP as well as with other LOAD risk factors such as CR3, and SIRPβ1 [
5,
7,
23,
55,
86] provided an important lead for our experimental strategy aiming to validate the important role of
TYROBP in the pathogenesis of LOAD. We have previously defined via a multiscale computational network approach
TYROBP as a strong candidate for playing the role of a key “hub” or “driver” gene in LOAD [
86]. It is worth noting that two independent groups have also identified
TYROBP as a driver of LOAD despite having followed different and highly idiosyncratic computational strategies [
12,
48].
CD33 is a known AD risk gene and a component of the
TYROBP network. The regulation of
Cd33 by TYROBP reported herein as well as the regulation of
TREM2 by CD33 reported by de Jager and colleagues [
12] provide compelling evidence in support of the role of
TYROBP as a “driver” gene in LOAD. Capping off the evidence associating
TYROBP with LOAD is the recent discovery that missense mutations in the coding region of the
TYROBP gene are associated with AD risk [
61]. Interestingly, in the same study, in vitro overexpression of the candidate pathogenic p.D50_L51ins14
TYROBP variant led to a strong reduction of TREM2 expression [
61]. We have previously shown that
TYROBP expression is elevated in AD brain and mouse models [
64,
86], but it was not immediately apparent whether that elevation represented a pre-existing, predisposing factor or was a secondary reaction to LOAD pathology. Based on the data presented above, a
Tyrobp null mutation appears to exert effects that would be characterized as beneficial with respect to both the normal physiology of neurons and the proteinopathy of LOAD.
The effects of
Tyrobp knockdown or knockout on Aβ levels and Aβ oligomer conformers as defined by epitope content were limited to the reduction of the level of NU-4 and A11 type oligomers in TYROBP deficient
APP/PSEN1 mice. There were no consistent statistically significant differences on levels of total Aβ, Aβ40, Aβ42, or on levels of OC type Aβ oligomers. The relatively minor effect size notwithstanding, it is worth noting that the converging evidence from several laboratories (including our own) is that the NU-4 epitope is the signature of the Aβ oligomer strain that is most consistently neurotoxic [
40,
73]. A11 and OC oligomer strains are not consistently neurotoxic. As reported above and in one of our previous studies [
41], we noted sex differences in Aβ and oligomer levels suggesting an earlier progression of the disease in female than male
APP/PSEN1 mice.
The difference in Aβ levels observed between the male and female mice is of importance considering the sexual dimorphism observed in the phosphorylation status of TAU in
APP/PSEN1 background but not in WT background. Thus, the effect of a decreased
Tyrobp expression on the stoichiometry of TAU phosphorylation appeared to be different in the presence or absence of
APP/PSEN1 mutations leading to amyloid deposition. Indeed, TYROBP deficiency tends to increase the phosphorylation of TAU on a WT background, but, on a
APP/PSEN1 background, loss of TYROBP decreased the phosphorylation status of TAU in female mice in the setting of higher Aβ loads as compared with males. Although the mechanism(s) by which microglia exert their effects on neuronal tau pathology remains unclear, several reports have linked TREM2 expression and hyperphosphorylated TAU [
31,
33,
35]. These reports suggest that TREM2 deficiency could increase tauopathy in human tau-expressing models but could decrease tau pathology in AD mouse models displaying cerebral amyloidosis. Herein we report that a decreased expression of
Tyrobp can have beneficial effects on tau pathology and neuronal injury in
APP/PSEN1 mouse model of AD. In accordance with our data, Strittmatter and colleagues [
75] recently reported that mouse deficient for
Progranulin presented an overexpression of
Tyrobp network genes correlating with an increased neuronal injury and tau pathology in the absence of amyloid pathology [
75].
As mentioned above, no differences were noted in number and size of Aβ plaque depositions and the general histological impact of TYROBP deficiency on plaque morphology and microglia recruitment was identical in appearance to that reported by Colonna and colleagues in their studies of TREM2
-deficient mice [
85]. Indeed,
Tyrobp KO mice presented fewer microglia decorating each amyloid plaque without modification in the total number of microglia, and plaques exhibited less compact morphology. However, unlike the Colonna report wherein the reduced numbers of microglia per plaque were predicted to be associated with
increased severity of the phenotype [
85], we observed that this histological appearance was instead associated with
beneficial effects on neuritic dystrophy, TAU metabolism, learning behavior, and neuronal electrophysiology. Although beyond the scope of this study, it will be interesting to determine whether overexpression leads to opposite results. In addition, recent papers from Lamb and colleagues [
30], Yu and colleagues [
32,
34], and Raha-Chowdhury and colleagues [
63] raise the possibility that there could be aging-related and/or disease-stage-related changes in the effect of TYROBP. These papers focused on TREM2 and suggest that reduced TREM2 may be beneficial early in life (~4 months) while reduced TREM2 late in life (~8 months) could be detrimental. We are in the process of assessing whether a similar phenomenon occurs with TYROBP.
Electrophysiological assays revealed that the loss of TYROBP normalized some of the synaptic dysfunctions caused by the
APP/PSEN1 mutations. The strong increase in basal synaptic efficiency seen in the
Tyrobp
−/− mice is of particular interest. If observed in isolation, this phenomenon might lead to overactivation of pyramidal neurons and damage, but the same effect could prove protective in the context of LOAD-associated factors that
reduce neuronal function. The protective effect of TYROBP deficiency in an early AD context is confirmed by the improvement in the behavioral performance of
APP/PSEN1 mice deficient in TYROBP. The effect of the
Tyrobp-null background on the electrophysiological findings and gene set enrichment (synapse assembly, ion transport, and neurotransmission) of
APP/PSEN1; Tyrobp
−/− vs.
APP/PSEN1 are in keeping with the growing appreciation for the role of microglia in maintaining normal synaptic physiology [
4]. Indeed, in addition to their pro-inflammatory and phagocytic functions, microglia release cytokines including TGFβ and interleukin-1β that acutely modulate synaptic plasticity at hippocampal synapses [
11,
71].
Thus, in a comprehensive panel of transcriptomic, biochemical, electrophysiological, and behavioral paradigms, reduction or ablation of TYROBP prevented the expression of many of the corresponding APP/PSEN1 phenotypes at 4 months of age. These results would appear to argue against the possibility that early TYROBP deficiency is likely to be a predisposing factor for LOAD. Indeed, these results would indicate that a decrease in TYROBP activity could represent an important therapeutic opportunity either for treating or preventing LOAD or else for slowing or arresting the progression of MCI or early AD to full-blown clinical and pathological LOAD.