|Home | About | Journals | Submit | Contact Us | Français|
Microglial activation and overproduction of inflammatory mediators in the central nervous system (CNS) have been implicated in Alzheimer's disease (AD). Elevated levels of the pro-inflammatory cytokine Tumor Necrosis Factor (TNF) have been reported in serum and post-mortem brains of patients with AD, but its role in progression of AD is unclear. Using novel engineered dominant negative TNF inhibitors (DN-TNFs) selective for soluble TNF (solTNF), we investigated whether blocking TNF signaling with chronic infusion of the recombinant DN-TNF XENP345 or a single injection of a lentivirus encoding DN-TNF prevented the acceleration of AD-like pathology induced by chronic systemic inflammation in 3xTgAD mice. We found that chronic inhibition of solTNF signaling with either approach decreased the LPS-induced accumulation of 6E10-immunoreactive protein in hippocampus, cortex, and amygdala. Immunohistological and biochemical approaches using a C-terminal APP antibody indicated that a major fraction of the accumulated protein was likely to be C-terminal APP fragments (β-CTF) while a minor fraction consisted of Aβ 40 and 42. Genetic inactivation of TNFR1-mediated TNF signaling in 3xTgAD mice yielded similar results. Taken together, our studies indicate that soluble TNF is a critical mediator of the effects of neuroinflammation on early (pre-plaque) pathology in 3xTgAD mice. Targeted inhibition of solTNF in the CNS may slow the appearance of amyloid-associated pathology, cognitive deficits, and potentially the progressive loss of neurons in AD.
Although the etiology of idiopathic Alzheimer's disease (AD) in humans is unknown, mutations in Amyloid Precursor Protein (APP) or Presenilin-1, a component of the γ-secretase complex, result in overproduction of Aβ40 and 42 peptides and are sufficient to cause disease (Cai et al., 1993; Citron et al., 1992; Goate et al., 1991; Lemere et al., 1996; Lopera et al., 1997; Nee et al., 1983; Whalen et al., 2005). Cleavage of APP by α- and β-secretases also results in the retention of intracellular fragments which can be further cleaved by γ-secretase. Specifically, cleavage by α-secretase results in an 83 amino acid fragment, known as C83 or α-C-terminal fragment (α-CTF (Estus et al., 1992) which is further processed by γ-secretase to produce the non-amyloidogenic fragment p3 (Haass et al., 1993). Cleavage of APP by β-secretase produces a 99 amino acid fragment known as C99 or β-CTF (Sinha and Lieberburg, 1999) which is further cleaved by γ-secretase to produce Aβ peptides. β-CTF alters ionic homeostasis (Fraser et al., 1997) and exerts neurotoxicity in differentiated PC12 cells (Yankner et al., 1989) and in transgenic mice (Oster-Granite et al., 1996). Neuronal cells bearing familial AD (FAD) mutations accumulate β-CTF intracellularly (McPhie et al., 1997), implicating its involvement in AD pathogenesis.
Experimental and clinical evidence suggests a close association between neuroinflammation and AD pathogenesis (Akiyama et al., 2000). Overproduction of inflammatory mediators in the brain occurs when microglia, which are often found in close physical association with amyloid plaques in AD brains, become chronically activated. It has been proposed that elevated levels of pro-inflammatory cytokines, including Tumor Necrosis Factor (TNF), may inhibit phagocytosis of Aβ in AD brains thereby hindering efficient plaque removal by resident microglia. Clinically, TNF-driven processes have been strongly implicated in AD pathology and may contribute to cognitive dysfunction and accelerated progression of AD (Alvarez et al., 2007; Collins et al., 2000; Fillit et al., 1991; Ma et al., 2004; Paganelli et al., 2002; Tan et al., 2007; Tobinick et al., 2006). In support of the idea that chronic systemic inflammatory disease may accelerate AD pathogenesis, the bacterial endotoxin lipopolysaccharide (LPS), a potent trigger of inflammation that elicits production of TNF and other cytokines (Flick and Gifford, 1986), can accelerate the appearance and severity of AD pathology in several animal models of AD. These include the APPV717F mouse (Qiao et al., 2001), the APPSwe Tg2576 mouse (Sheng et al., 2003), and the 3xTgAD mouse (Kitazawa et al., 2005) in which up-regulation of TNF mRNA precedes the appearance of amyloid-associated pathology (Janelsins et al., 2005) and correlates with cognitive deficits (Billings et al., 2005). Taken together, these observations provide compelling rationale to investigate new therapeutic approaches that selectively target the TNF pathway in pre-clinical models of AD. The main goal of our studies was to investigate the feasibility and efficacy of novel anti-TNF therapeutics in blocking early development of amyloid-associated pathology in 3xTgAD mice (Oddo et al., 2003a; Oddo et al., 2003b) exposed to chronic peripheral inflammation.
Unless otherwise stated, all chemical and tissue culture reagents were obtained from Sigma-Aldrich Corp (St. Louis, MO). Tissue culture plastic was obtained from Corning (Corning, NY). Recombinant mouse TNF was obtained from R&D Systems (Minneapolis, MN). The dominant negative XENP345 was provided by Xencor, Inc., DN-TNF neutralizes solTNF signaling by forming dominant negative heterotrimers that have significantly attenuated binding affinity for TNF receptors (Steed et al., 2003; Zalevsky et al., 2007) thereby lowering the effective concentration of solTNF without affecting overall production levels of TNF or the biological activity of transmembrane TNF (tmTNF), including its role in maintaining innate immunity (Zalevsky et al., 2007).
3xTgAD mice are of a mixed 129/C57BL6 genetic background (Oddo et al., 2003b) and were generated by Dr. Frank M. LaFerla (University of California at Irvine). TNFR1-deficient mice on a C57BL6 genetic background were obtained from Jackson Laboratories (Bar Harbor, ME). Animals were housed in pathogen-free climate controlled facilities and allowed to have food and water ad libitum at the Animal Resources Center at The University of Texas Southwestern Medical Center. All animal studies were approved by the Institutional Animal Care and Use Committee at The University of Texas Southwestern Medical Center at Dallas.
Young adult (4.5 month old) 3xTgAD mice were anesthetized continuously with 2.5% halothane (Halocarbon Laboratories, River Edge, NJ) and placed in a stereotaxic frame (David Kopf Instruments, Tujunga, CA). We inserted a cannula (gauge 28; Plastics One, Roanoke, VA) connected via polyethylene tubing to a subcutaneously implanted osmotic minipump (model 2004, Alzet, Cupertino, CA) preloaded with vehicle (sterile PBS with 10% glycerol) or the treatment agent XENP345 (0.01 mg/kg/day) at the coordinates for hippocampus CA1 in the right hemisphere (anterioposterior (AP): −2.0 mm from bregma, mediolateral (ML): −2.0 mm and dorsoventral (DV): −1.6mm below dura), per the mouse brain atlas (Paxinos, 2001). The recombinant dominant-negative human TNF variant XENP345 (Steed et al., 2003; Zalevsky et al., 2007) was bacterially produced and formulated by Xencor, Inc. (Monrovia, CA) to contain <0.1 endotoxin units (E.U.)/mg. Cannulae were secured to the skull with surgical glue and left in position for 4 weeks. The mice were injected with either 0.25 mg/kg (7.5 × 105 endotoxin units E.U./kg LPS (from Escherichia coli O111:B4; 3.0 × 106 E.U./mg, Sigma-Aldrich Corp., St. Louis, MO) or an equivalent volume of sterile saline (B. Braun Medical Inc., Bethlehem, PA). intraperitoneally (i.p.) twice weekly for 4 weeks.
The human pro-DN-TNF sequence, provided to us by David E. Szymkowski (Xencor, Inc., Monrovia, CA), included a signal peptide sequence and was that of the TNF variant A145R/I97T. The DN-TNF sequence was subcloned into a constitutive self-inactivating lentiviral vector based on the plasmid pLV (Pfeifer et al., 2002) 5' of an internal ribosome entry site (IRES) followed by the GFP coding sequence. The GFP expressing lentiviral plasmid has been described previously (Pfeifer et al., 2002; Taylor et al., 2006). DN-TNF or GFP expression was driven by the CMV enhancer/chicken ß-actin hybrid promoter (CAG).
Lentivirus stocks were produced and purified according to a previously published protocol (Taylor et al., 2006). The final titer was 125 μg/mL P24 and 1.6 × 109 infectious units/mL for the negative control lentivirus-GFP and 980μg/mL p24 and 8 × 108 IU/mL for lentivirus-DN-TNF.
Neuro2a cells were infected with 10 M.O.I. lentivirus-DN-TNF. Media was collected and replaced every 24 hours up to 72 hours post-infection. The DN-TNF produced by the cells was measured by quantitative ELISA specific for human TNF and non-crossreactive with mouse TNF (Biosource/Invitrogen, Carlsbad, CA).
Young adult (3 month-old) 3xTgAD mice were anesthetized continuously with 2.5 % halothane and placed in a stereotaxic frame. A 30 gauge needle (Hamilton Company, Reno, NV) was inserted to reach the coordinates for the 3rd ventricle (AP: −0.82 mm from bregma, ML: −0.47 mm from midline, DV: −2.5 mm from dura) per the mouse brain atlas (Paxinos, 2001). Due to the risk of damaging blood vessels by injecting so close to midline, we used an angled approach to reach the coordinates above (12° angle, AP: −0.82 from bregma, ML: −1.0 mm from midline, DV: −2.55 mm from dura). An injection of 1 μL of lenti-DN-TNF or lenti-GPF at 100 μg p24/mL was administered via an automated injector (Stoelting Co., Wood Da1e, IL) using a 5 μL Hamilton microsyringe. Mice were allowed to express the lentivirus for 6 weeks and were then injected with 0.25 mg/kg (7.5 × 105 E.U./kg) LPS i.p. twice weekly for 6 weeks.
Primary microglial cells were isolated from postnatal day 1 (P1) 3xTgAD mouse pups and cultured as previously described (Saura et al., 2003). Quantification of activated microglia was performed by counting the number of primary microglia which stained brightly positive for CD45 in 10 fields per well in 3 separate wells for each treatment condition. Total number of microglia was quantified by counting all cells stained positive for CD45 (strongly and weakly); these counts matched the total number of nuclei counterstained with Hoechst 33258).
The murine astrocyte cell line SS01 (a gift from Dr. Robert Bachoo, University of Texas Southwestern Medical Center, Dallas, TX) was maintained in DMEM supplemented with 10 % heat-inactivated FBS.
Immunohistochemistry was performed on brain sections as previously described (McCoy et al., 2006). 10 μm cryosections were incubated with mouse anti-human Aβ clone 6E10 (Chemicon Temecula, CA, 1:3000). Immunoreactivity was visualized using an Alexa fluor-conjugated secondary antibody, goat anti-mouse 488 (Molecular Probes/Invitrogen, Carlsbad, CA, 1:1000). Sections were counterstained with the nuclear dye Hoechst 33258 (bisbenzimide, 1:20,000). Immunocytochemistry was performed as previously described (McCoy et al., 2006). Cells were incubated with rabbit anti-NFκB p65 Re1a (Santa Cruz Biotechnology, Santa Cruz, CA, 1:200) or rat-anti-CD45 to visualize microglia (Serotec, 1:500). Immunoreactivity was visualized using Alexa fluor-conjugated secondary antibodies (goat anti-rabbit 594 or goat anti-rat 594, 1:1000) followed by nuclear counterstain as above.
Brain sections on slides were stained using a previously published protocol (Frank et al., 2003). Epitope unmasking required for anti-Aβ (6E10),Aβ40- or 42-specific, and N-terminal APP staining consisted of a brief incubation in 88 % formic acid (Fisher Scientific, Fair Lawn, NJ). 30 μm cryosections were incubatef with mouse anti-human Aβ clone 6E10 (1:1000), rabbit anti-human Aβ40 (Chemicon, 1:100), mouse anti-Aβ40 (Mab 13.1.1, 1:500, (Das et al., 2003; Kim et al., 2007), rabbit anti-human Aβ42 (Invitrogen, 1:100), mouse anti-Aβ42 (Mab 2.1.3, 1:500, (Das et al., 2003; Kim et al., 2007), rabbit anti-APP (C9, 1:500, (Kimberly et al., 2005), chicken anti-GFP (Chemicon, 1:500), rat anti-mouse CD68 (Serotec, 1:500), or rat anti-mouse CD45 (Serotec, 1:500), mouse anti-APP (22C11, 1:100, Chemicon). Sections were then incubated with biotinylated secondary antibodies (horse anti-mouse, goat anti-rabbit, goat anti-chicken, or rabbit anti-goat, Vector Laboratories, Burlingame, CA, 1:1000) followed by incubation with neutravidin-HRP (Pierce Biotechnology, Rockford, IL). Immunoreactivity was visualized by reaction with diaminobenzidine (DAB) with nickel sulfate (to produce a purple-blue reaction product) or without ammonium nickel sulfate (to produce a brown reaction product).
6-month old 3xTg-AD mice were sacrificed by Euthasol overdose and cardiac perfusion with 0.01 M phosphate buffered saline (PBS). Brains were rapidly removed, postfixed in 4 % paraformaldehyde (pH-7.4) for 48 hr, and cut as 50 Pm thick coronal sections using a Vibratome. To examine intraneuronal Aβ, sections were double-labeled with the monoclonal antibody 6E10 (Covance, Emeryville CA) and a polyclonal antibody directed against the C-terminus of amyloid precursor protein (APP) (Invitrogen). Primary antibodies were applied overnight at 4°C and following rinses were detected with anti-mouse Alexa 555 (Invitrogen, 1:200, red) and anti-rabbit Alexa 488 (Invitrogen, 1:200, green). After rinsing, sections were mounted on slides and coverslipped using Fluoromount-G (Southern Biotech). Specificity of all primary antibodies was confirmed by Western blot and also by omission of primary antibody which demonstrated no staining (data not shown). 6E10 recognizes amino acids 1–16 within the Aβ sequence and thus can also recognize full-length APP or the β-CTF of APP. Double-label confocal microscopy was therefore utilized to identify interneuronal Aβ (red only) versus full-length APP or β-CTF immunoreactivity (overlap of red and green), or α-CTF (green only). Immunofluorescent sections were visualized using a Bio-Rad 2100 confocal imaging system equipped with Argon, HeNe, and Red Diode lasers (Bio-Rad Laboratories, Hercules, CA). To avoid non-specific bleed-through, each laser line was excited and detected independently using lambda-strobing mode. All images represent either single confocal Z-slices or Z-stacks. Kruskal-Wallis analysis was performed to assess significance between experimental conditions.
Stereological analyses to estimate the number of amyloid-positive cells were performed using the optical fractionator module of Stereoinvestigator software (MicroBrightField, Williston, VT). Contours were traced around the hippocampus, cortex, and amygdala under the magnification of a 2X objective as delineated by the mouse brain atlas (Paxinos, 2001). Aβ positivve (ommunoreactive for 6E10) cells were counted under a 60X oil immersion objective using random and systematic counting frames (size: 50 μm × 50 μm) with an 18 um optical dissector, 2 μm upper and lower guard zones in a 430 μ × 280 μm grid. All reported values had a Gundersen coefficient of error (CE) values of less than 0.10.
Aβ from mouse brains was extracted and measured as oreviously described (Levites et al, 2006). Briefly, each frozen hemi brain was sequentially extracted in a two-step extraction involving, extraction in RIPA buffer followed by 70 % formic acid. The following antibodies against Abeta were used in sandwich capture ELISA for measuring Aβ levels as described before (Levites et al., 2006). For brain Aβ40, Ab9 (Aβ1–16) was used for capture and Ab40.1-HRP for detection. For brain Aβ42, Ab42.2 was used for capture and Ab9-HRP for detection.
Flash-frozen mouse hemispheres were homogenized in ice-cold RIPA buffer with protease inhibitor cocktail (Roche, Indianapolis, IN). Protein samples were diluted in 2X Laemmli sample buffer (62.5 mM Tris-HCl, pH 6.8, 20% glycerol, 2% SDS, 0.1% bromophenol blue, 5% 2-mercaptoethanol), boiled for 5 min, and run on a 4–20% Tris-Glycine gel(for PS1 and BACE1) or a 10–20% gel (APP and CTFs) for 90 min at 125V in 1X Tris-Glycine-SDS running buffer or 90 min at 130V in 1X Tricine-SDS running buffer. The gels were transferred onto a 0.2 μm nitrocellulose membrane (Biorad) at 30V for 1 hr 30 min. The membranes were blocked in 5% milk in TBST (1X TBS with 0.1% Tween-20) for 1 hour and incubated in primary antibodies (anti-APP C9, 1:500; anti-PS1, 1:2500; anti-BACE1, 1:1000; anti-alpha-tubulin, 1:5000) overnight at 4°C. The membranes were then washed three times 10 min in TBST, incubated in secondary antibodies (goat-anti-mouse-HRP, 1:2500, or goat-anti-rabbit-HRP, 1:5000) for 1 hour at room temperature, then washed five times 10 min in TBST. The membrane was then incubated for 1 min at room temperature in WestDura chemiluminescent substrate (Thermo Scientific, Rockford, IL) and imaged using the ChemiImager 5500 (Alpha Innotech, San Leandro, CA).
Quantitative real-time PCR was performed as previously described (Valasek and Repa, 2005). Briefly, hippocampi were isolated from 3xTgAD mice given chronic systemic i.p. injections of saline or low-dose (7.5 × 105 E.U./kg) LPS for 6 weeks. RNA was extracted using Tri-Reagent (Molecular Research Center, Inc., Cincinnati, OH), treated with DNAse I (Invitrogen), and reverse transcribed using Superscript II RNAse H- reverse transcriptase (Invitrogen). Quantitative real-time PCR was performed using an ABI Prism 7900HT Fast Detection System (Applied Biosystems, Forester City, CA). Each reaction was performed in a volume of 20 μL that contained 50 ng cDNA, 10 μL SYBR Green PCR Master Mix (Applied Biosystems), and 150 nM of each (forward and reverse) PCR primer. All reactions were performed in triplicate. Levels of the various mRNAs were normalized to those of the mouse housekeeping gene cyclophilin. The following forward and reverse primers were used:
The differences between means of stereological estimates of Aβ positive cells in lentivirus-infected animals and between means of nuclear counts within the hilus and amyloid beta-positive cells were subjected to a one-tailed Student's t test. The differences between means of stereological estimates of 6E10-positive cells in the hippocampus of animals that received XENP 345 infusion via indwelling cannulae were subjected to a one-way ANOVA followed by Tukey's test for post-hoc analysis. Differences between means of stereological estimates of 6E10-positive cells in cortex, hippocampus and amygdala were subjected to a one-tailed t test. Differences in numbers of activated primary microglia were subjected to a one-way ANOVA followed by Tukey's test for post-hoc analysis. Differences in Aβ40 and Aβ42 levels were subjected to one-factor ANOVA. For immunofluorescence confocal microscopy, Kruskal-Wallis analysis was performed to assess significance between experimental conditions. Values expressed are the means ± S.E.M., * denotes p< 0.05.
3xTgAD mice display age-dependent AD-like pathology in the hippocampus and entorhinal cortex which can be accelerated by chronic systemic inflammation induced by bacterial endotoxin lipopolysaccharide (LPS), a potent trigger of inflammation that elicits production of TNF and other cytokines (Flick and Gifford, 1986). In order to investigate the extent to which soluble TNF mediates the LPS-induced accumulation of 6E10-immunoreactive protein reported in 3xTgAD mice (Kitazawa et al., 2005), we fitted 4.5 month old mice with osmotic pumps to administer 0.1 mg/kg/day of the DN-TNF variant XENP345 or an equivalent volume of vehicle into the hippocampus for 28 days. Our experimental design consisted of placing a cannula in the CA1 region of the hippocampus in the right hemisphere of the brain and, concurrent with the XENP345 or saline infusion, administering 0.25 mg/kg LPS (7.5 × 105 E.U./kg) intraperitoneally (i.p.) twice a week for one month (Figure 1A). The hippocampus was chosen as a brain region of interest based on the report that 3xTgAD mice display AD-like pathology in the hippocampus by 6 months of age (Oddo et al., 2003a). Moreover, we chose the chronic systemic low-dose LPS regimen because it has been shown to accelerate the appearance of AD-like pathology in several other mouse models of AD (Kitazawa et al., 2005; Qiao et al., 2001; Sheng et al., 2003). Immunohistological analysis of brain sections from 3xTgAD animals fitted with pumps dispensing XENP345 and exposed to chronic systemic inflammation showed a marked qualitative reduction in neurons accumulating 6E10-immunoreactive material in the hilus compared to animals fitted with pumps dispensing saline and exposed to chronic systemic inflammation (Figure 1B). Non-transgenic mice of the same genetic background did not display accumulation of 6E10-IR protein in response to chronic systemic inflammation in hippocampus (Figure 1B) or any other brain region. Counts in the hilar region of the hippocampus revealed that administration of XENP345 significantly reduced the appearance of 6E10-IR protein in the hilus of the LPS-treated 3xTgAD mice compared to the saline-infused animals (p<0.05 by one-tailed Student's t-test) (Figure 1C). This effect was also observed in parietal, temporal and entorhinal cortex as well as amygdala of LPS-treated mice (data not shown). Of note, the XENP345 treatment did not result in a reduction in the number of total cells (as measured by bisbenzimide-positive nuclei) in the regions where reduction in 6E10-IR was observed (Figure 1D). To rule out the possibility that increased expression of 6E10-immunoreactive protein was a result of increased transcription of the human APP transgene, we performed real-time quantitative PCR analyses on freshly dissected hippocampal tissue to measure mRNA levels of the human APP transgene as well as mouse APP, BACE1 and ApoE. In agreement with previous findings (Kitazawa et al., 2005), we found that the chronic systemic inflammation regimen did not induce robust changes in expression of the transgene or the aforementioned murine genes (Supplementary Figure S1). In a second experimental paradigm, 4.5 month old 3xTgAD mice were fitted with osmotic pumps for delivery of either saline or XENP345 for 1 month. During this time, the mice were injected i.p. twice a week with either 7.5 × 105 E.U./kg of LPS or an equivalent volume of saline (Figure 2A). The brains of these mice were stained with the 6E10 antibody and subjected to stereological counts of 6E10-positive cells throughout the hippocampus ipsilateral and contralateral to the cannula. A gross comparison of brain regions stained with 6E10 revealed a qualitative increase in the extent of intraneuronal accumulation of 6E10-IR species in the total hippocampus (Figure 2B) and specifically the hippocampal CA1 region (Figure 2C) of animals infused with saline and exposed to chronic systemic LPS which did not occur in animals infused with XENP345 in the hippocampus and exposed to chronic systemic inflammation. Quantitative stereological analysis revealed that intrahippocampal administration of XENP345 inhibited the appearance of 6E10-positive cells (p < 0.05 by one-factor ANOVA and Tukey's posthoc test) in LPS-injected mice (Figure 2D) on both ipsilateral and contralateral hemispheres, possibly due to circulation of the inhibitor to the contralateral side. On the basis of these results we concluded that short term inhibition of solTNF in the hippocampus of young adult 3xTgAD mice blocks the accumulation of a 6E10-IR amyloid-associated protein(s) induced by chronic systemic inflammation.
Although chronic infusion of DN-TNF enabled us to block soluble TNF signaling in the hippocampus for a period of one month, a different approach was needed to obtain longer term inhibition of TNF activity and reach other parts of the brain affected by amyloid accumulation, such as the entorhinal cortex. To achieve this end, we chose to use a lentivirus-based approach because lentivirii infect non-mitotic cells and because expression of the desired gene can be maintained for periods of at least one year (Blomer et al., 1997). The sequence of the human TNF variant A145R/I97T (Steed et al., 2003), including the TACE cleavage site that converts the transmembrane form of the protein into a soluble form (Aggarwal, 2000), was subcloned into a self-inactivating lentiviral vector downstream from the CMV enhancer/chicken β-actin (CAG) promoter (Supplementary Figure S2A). The lentiviral vector also contains an IRES-driven green fluorescence protein (GFP) sequence which allowed us to monitor in vitro and in vivo expression using endogenous GFP fluorescence or anti-GFP immunohistochemistry. The control vector consisted of the GFP sequence driven by the CAG promoter.
To establish that transduction of cells with lenti-DN-TNF resulted in production of measurable DN-TNF protein, the level of DN-TNF production in the conditioned media of cells transduced with the lentivirus was measured using a quantitative human TNF ELISA to specifically detect the human DN-TNF protein. The DN-TNF protein is not detected by ELISA specific for mouse TNF, nor is mouse TNF detected by ELISA specific for human TNF (data not shown). We found that Neuro2a murine neuroblastoma cells infected with lenti-DN-TNF produced daily amounts of DN-TNF into the conditioned media in the 10 ng/mL range over the 3-day time course assayed (Supplementary Figure S2B), an amount that should be sufficient to neutralize up to 10 ng/mL solTNF in vitro (Steed et al., 2003; Zalevsky et al., 2007). To confirm that the lentiviral-derived DN-TNF was biologically active, we used several different in vitro functional assays. First, we assessed TNF pathway activation status by measuring cytoplasmic-to-nuclear translocation of the NFκB subunit p65 RelA (Hohmann et al., 1990) immunocytochemically in a murine astrocyte cell line. NFκB translocation in response to 2 ng/mL TNF was detectable in mock-transduced and lenti-GFP-transduced cells (Figure 3A). In contrast, transduction with lenti-DN-TNF blocked TNF-induced cytoplasmic-to nuclear translocation of the p65 subunit of NFkB (Figure 3A), indicating that transduction of cells with a lentivirus encoding DN-TNF resulted in proper expression of functionally active protein capable of antagonizing solTNF signaling. Next, we investigated the extent to which primary microglia transduced with lenti-DN-TNF responded to a TNF challenge. Primary microglial cultures from 3xTgAD mice were transduced with lenti-DN-TNF, lenti-GFP, or mock-transduced. We quantified microglial activation in response to 10 ng/mL TNF using immunocytochemistry for the microglial activation marker CD45. We used CD45 as a pan-microglia marker, as it stained all microglia in the dish, and used its upregulation especially in the nucleus as a measure of activation. The numbers of CD45-positive cells in each treatment group were used to normalize the numbers of activated microglia. Mock- and lenti-GFP transduced microglia responded to TNF treatment by upregulating CD45 expresssion (Figure 3B). Mock-transduced cells, cells transduced with lenti-DN-TNF, or cells transduced with the negative control lenti-GFP were challenged with 10 ng/mL solTNF in the presence of excess dominant negative TNF protein inhibitor XENP345 (200 ng/mL) as a positive control for suppression of TNF signaling. Cultures that were mock-transduced or transduced with lenti-GFP displayed a significant increase in the percent of activated microglia in response to solTNF treatment, whereas lenti-DN-TNF-transduced cells displayed baseline levels of microglia activation, equivalent to the reduction in signaling achieved using XENP345 (Figure 3C). On the basis of these studies, we hypothesized that transduction of mouse brains with lenti-DN-TNF should also result in functional antagonism of TNF-dependent processes in vivo.
To investigate the extent to which lentiviral-derived DN-TNF could block the LPS-induced accumulation of intraneuronal amyloid-related species accumulation, we selected 3 month-old 3xTgAD mice to coincide with the time of TNF mRNA upregulation (Janelsins et al., 2005) and the earliest appearance of intraneuronal 6E10-immunoreactive protein in the entorhinal cortex (Oddo et al., 2003a). Intracerebroventricular (ICV) injections of lenti-DNTNF or lenti-GFP were performed to deliver the lentivirus bilaterally and maximize virus spread. Lentivirus-based DNTNF (or GFP) expression was allowed to proceed for 6 weeks prior to the start of the chronic systemic inflammatory stimulus, which consisted of intraperitoneal injections of 7.5 × 105 E.U./kg of LPS twice a week for another 6 weeks, an experimental paradigm previously reported to hasten the appearance of AD-like pathology in 3xTgAD mice (Kitazawa et al., 2005) (Figure 4A). The efficiency of lentiviral transduction of cells in the brains of mice injected with lenti-DN-TNF ICV was assessed by immunohistochemical analysis with an anti-GFP antibody to detect expression from the lentiviral vector. GFP-immunoreactivity was readily and specifically detected in the brain parenchyma of mice injected with lenti-DN-TNF, confirming that transduced brain cells were indeed expressing the proteins encoded by the lentivirus (Figure 4B). In support of our hypothesis that TNF is a key mediator of the enhanced amyloid-associated pathology induced by chronic systemic inflammation, the brains of mice injected with lenti-DN-TNF displayed a drastic reduction in intraneuronal APP-derived 6E10-immunoreactive species in parietal cortex (Figure 5A, black brackets), in the CA1 region of the hippocampus (Figure 5A, white brackets), in amygdala (Figure 5B) and in entorhinal and temporal cortex (not shown) compared to mice injected with the control lenti-GFP. We performed unbiased stereology to estimate the number of 6E10-immunoreactive cells in the hippocampus, cortex and amygdala using the optical fractionator method, and found that lenti-DN-TNF-transduced brains displayed a significant (~ 60 %) reduction in accumulation of intraneuronal APP-derived 6E10-immunoreactive protein compared to lenti-GFP transduced brains (p < 0.05 by one-tailed Student's t-test) (Figure 5C). Therefore, we concluded that in vivo inhibition of TNF signaling via infusion of XENP345 or transduction with lenti-DN-TNF results in significant attenuation of 6E10-immunoreactive protein.
Given that immunoreactivity for 6E10 can represent APP, C-terminal APP fragments, and/or Aβ peptides, we performed brightfield immunohistology to investigate if the intraneuronal APP-derived 6E10-immunoreactive species in brains of young adult 3xTgAD mice exposed to chronic systemic inflammation was composed of mainly Aβ peptides. We staind intervening brain sections with Aβ 40- and Aβ42-specific antibodies from two independent sources to compare the staining patterns with that of the 6E10. Brightfield immunohistological analysis of hippocampal brain regions with Aβ 40 and Aβ-42 antibodies revealed no detectable immunoreactivity in neuronal populations which were clearly positive for 6E10 in young-adult 3xTgAD mice that received chronic systemic LPS injections and either saline infusion via osmotic pump in the hippocampus CA1 region (Supplemental Figure S3A) or lenti-GFP (Supplemental Figure S3B) as negative controls for XENP345 and lenti-DNTNF, respectively. As a positive control for the Aβ 40 and Aβ-42 antibodies, hippocampal brain regions from an untreated APPswe/PS1Δ9 transgenic mouse were analyzed in the same way and found to display identical staining patterns for 6E10, Aβ 40 and Aβ-42 in and around amyloid plaque deposits (Supplemental Figure S3C). We considered that our inability to detect Aβ 40 and Aβ-42 peptides using brightfield immunohistology could be due to the reduced sensitivity of the technique. To test this possibility, we performed double immunofluorescence labeling of hippocampal sections with the 6E10 antibody (red) and a C-terminal-specific APP antibody (green) to generate confocal images and construct Z-stack series in order to determine the relative abundance of Aβ peptides (6E10 signal minus the C-terminal-specific signal subtracted) versus β-CTF, α-CTF, and APP in our 3xTgAD mice (See Materials and Methods). The levels of APP/ β-CTF species (regions positive for both 6E10 and C-terminal immunoreactivity, yellow overlay signal) were detectably higher than those of Aβ (shown as Aβ subtracted signal) in young adult 3xTgAD mice transduced with lenti-DNTNF and exposed to chronic systemic inflammation (Figure 6A). Consistent with previous observations that chronic systemic LPS did not significantly increase Aβ levels measured by ELISA (Kitazawa et al., 2005), this confocal subtraction analysis indicated that chronic exposure to systemic LPS did not significantly increase the levels of Aβ peptides and inhibition of TNF signaling with the DN-TNF inhibitor XENP345 (Figure 6B, p = 0.34 by Kruskal-Wallis analysis) or with lenti-DN-TNF (Figure 6C, p = 0.15 by Kruskal-Wallis analysis) had no effect on their levels either. To confirm these observations biochemically, we performed quantitative high-sensitivity ELISAs for soluble and insoluble Aβ40 and 42 in lysates of microdissected cortices and hippocampi from 3xTgAD mice injected with lenti-DN-TNF (or lenti-GFP) and exposed to chronic systemic inflammation. In agreement with our confocal analyses of hippocampal brain sections, we found that levels of Aβ40 and Aβ42 were low and not significantly changed by TNF signaling inhibition (Figure 6D, one-way ANOVA, n = 4 for lenti-DN-TNF, n = 3 for lenti-GFP, Aβ40 p = 0.12; Aβ42 p = 0.10). Taken together, results from these multiple independent approaches led us to conclude that the bulk of the 6E10-IR species that accumulated intraneuronally as a result of the chronic systemic inflammatory stimulus and was significantly attenuated when solTNF signaling was inhibited was likely to be APP or β-CTF.
To confirm that the major fraction of 6E10-IR protein was APP or β-CTF, we used the C-terminal APP antibody C9 (Kimberly et al., 2005) to stain brain sections of treated 3xTgAD mice. We compared brain sections stained with C9 from animals fitted with osmotic pumps delivering DN-TNF and treated with LPS to those from animals fitted with osmotic pumps delivering saline and treated with LPS. We also compared hippocampal CA1 brain sections stained with C9 from 3xTgAD mice that received lenti-GFP injections followed by systemic LPS with those of mice that received lenti-DN-TNF followed by systemic LPS. In agreement with 6E10-IR results, accumulation of intraneuronal APP C9-immunoreactive material in the hippocampus CA1 region was evident in 3xTgAD mice that were fitted with a saline pump in the CA1 and injected with low-dose LPS (i.p.) and this increase was inhibited by infusion of XENP345 (Figure 7A). Similarly, lenti-DN-TNF/LPS-treated mice displayed reduced C9-immunoreactivity in the parietal cortex and hippocampus CA1 region compared to lenti-GFP/LPS-treated mice (Figure 7B), suggesting that TNF signaling is a critical step in this process. Given the overlapping immunohistological pattern between 6E10 and C9 and the position of the epitopes for 6E10 and C9 on the APP protein, our data suggests that the bulk of the accumulated 6E10-immunoreactive protein species is likely to be composed of full-length APP and/or β-CTF and a small amount of Aβ peptides. To directly investigate whether full length was one of the 6E10-immunoreactive species accumulating intraneuronally in LPS-treated mice, we stained brain sections from our experimental animals as well as some APPswe/PS1Δ9 mice using two separate antibodies against the N-terminus of APP. An antibody obtained from Sigma (cat. #A8967) failed to stain N-terminal APP in either 3xTgAD or APPswe/PS1Δ9 brain sections. However, immunohistological analysis with 22C11 anti-N-terminal APP antibody yielded a punctate pattern in neurons in brain sections from APPswePS1Δ9 mice (positive control) but failed to detect robust intraneuronal staining in the brains of 3xTgAD mice treated with LPS systemically (data not shown). Taken together with results from confocal analyses these data led us to conclude that the 6E10-immunoreactive species is not full-length APP but rather mostly β-CTF, (recognized by 6E10 and C9) and some α-CTF (recognized only by C9). Next, we aimed to extend and confirm these findings using a biochemical approach. Densitometric analyses of immunoblots of whole-hippocampus and cortex detergent lysates from lenti-GFP and lenti-DN-TNF-injected animals revealed no significant differences in protein levels of full-length APP, the various APP-derived species, BACE1, or presenilin-1 (PS1), the catalytic subunit of γ-secretase between the mice exposed to chronic systemic LPS and injected with lenti-GFP versus lenti-DN-TNF (Supplemental Figure S4). However, given that immunohistological analyses 6E10 and C9 indicate that only a subpopulation of cells displays robust accumulation of the APP-derived species, it is not surprising that the effects of TNF signaling inhibition on this sub-population cannot be appreciated in the background of many other cells which are not affected. On the basis of these results, we concluded that inhibition of TNF signaling by gene transfer of DN-TNF via lentivirus prevents LPS-induced intraneuronal accumulation of β-CTF the same extent as that achieved with chronic delivery of the DN-TNF biologic XENP345. Taken together, these findings suggest that persistent inactivation of TNF signaling during the early stages of AD-like amyloid neuropathology may be an effective way to block the cascade of downstream events (such as accumulation of toxic APP-derived fragments and/or overproduction of Aβ oligomers) that contribute to cognitive decline.
As expected, microglial activation (measured by CD45 and CD68 immunohistology) induced by chronic systemic LPS was not completely inhibited by in vivo infusion of XENP345 or lenti-DN-TNF transduction (Supplemental Figure S5). This result is not surprising given that LPS elicits production of many other inflammatory factors that can also activate microglia in vivo. In addition, these findings are consistent with our previously published work in a rat model of Parkinson's disease (McCoy et al., 2006) in which XENP345 neutralized solTNF produced in response to LPS and rescued dopaminergic neurons despite persistent microglia activation.
Genetic deletion of the TNF receptor gene Tnfrsf1a in APP 23 transgenic mice was recently shown to result in reduction of both the number of amyloid plaques and the cognitive deficits in these mice (He et al., 2007). Therefore, we took the same genetic approach in the 3xTgAD mice and inactivated solTNF signaling by crossing these mice to TNFR1-deficient mice or to WT mice of the same genetic background (C57/B6) and bred the compound transgenics to homozygosity. We then examined intraneuronal APP-derived amyloid burden in the brains of TNFR1-deficient 3xTgAD mice and 3xTgAD; WT B6 mice exposed to chronic systemic inflammation (or given saline i.p. injections) for 6 wks. Consistent with results obtained after chronic inhibition of solTNF signaling with XENP345 or lenti-DN-TNF, 3xTgAD;WT B6 mice given i.p. LPS injections displayed increased intraneuronal 6E10 and C9-immunoreactivity whereas inactivation of TNFR1 in 3xTgAD mice prevented this intraneuronal accumulation. Interestingly, 6E10 immunoreactivity was not absent from the sections; instead, it was detectable in cellular processes characteristic of microglia and may represent engulfed Aβ (Figure 8).
The accumulation of intraneuronal APP/Aβ reported in brains of AD patients and several transgenic mouse models with AD-like pathology may be early events that promote neuronal dysfunction and death (Chin et al., 2007; Gouras et al., 2000; LaFerla et al., 1997). By blocking TNF signaling transiently (4 weeks) with solTNF-selective dominant negative TNF (DN-TNF) inhibitor XENP345 or persistently with a lentivirus encoding DN-TNF, we reduced the accumulation of intraneuronal APP β-CTF in the AD-affected brain regions of LPS-treated 3xTgAD mice. Previous studies in the 3xTgAD model reported that TNF mRNA is increased in entorhinal cortex at 3 months (Janelsins et al., 2005) prior to the onset of the cognitive deficits that correlate with the appearance of intraneuronal APP-derived 6E10-IR protein species. Therefore, we hypothesized that TNF upregulation and the appearance of that intraneuronal amyloid-associated species were causally linked. We further reasoned that targeted inhibition of TNF signaling in AD-relevant brain regions may be an effective basis to halt or attenuate progression of amyloid-associated neuropathology prior to the appearance of extracellular plaques. Our results demonstrate that both chronic infusion into the hippocampal region of 3xTgAD mice of a novel dominant negative biologic selective for soluble TNF (Zalevsky et al., 2007) and persistent inactivation of TNF signaling with a lentiviral-encoded DN-TNF drastically reduced accumulation of intraneuronal β-CTF in the brains of 3xTgAD mice exposed to chronic systemic inflammation as measured by 6E10 and C9 immunohistochemistry. Lastly, genetic deletion of TNFR1 in 3xTgAD mice prevented the LPS-induced accumulation of β-CTF. β-CTF alters ionic homeostasis (Fraser et al., 1997) and is neurotoxic in differentiated PC12 cells (Yankner et al., 1989) and in transgenic mice expressing the C-terminus of APP (Oster-Granite et al., 1996). Neuronal cells bearing familial AD (FAD) mutations accumulate β-CTF intracellularly (McPhie et al., 1997), implicating its involvement in AD pathogenesis.
Previous studies measuring reduced amyloid burden after genetic ablation of TNFR1 in APP23 transgenic mice have implicated TNF or one of its downstream targets in modulation of amyloid-associated neuropathology (Ohno et al., 2007); but this is the first study that selectively inhibits soluble TNF signaling in an acute manner in adult animals with a resulting effect on intraneuronal β-CTF. By using both a selective anti-TNF biologic and a gene transfer approach, we were able to time the inhibition of TNF to a relevant and precise interval in the progressive AD-like pathology in 3xTgAD mice; i.e., when TNF is upregulated and intraneuronal 6E10-IR protein first appears (Billings et al., 2005; Janelsins et al., 2005). As such, our findings provide strong rationale to further explore practical ways to deliver highly selective soluble TNF inhibitors such as DN-TNF biologics to the CNS for potential use in neurodegenerative diseases associated with chronic neuroinflammation. We have previously investigated the efficacy of XENP345 in other models of neurodegeneration and demonstrated that inhibition of solTNF signaling rescued approximately 50 % of the rat dopaminergic neurons that would otherwise die as a result of 6-hydroxydopamine- or LPS-induced death in models of Parkinson's disease (McCoy et al., 2006).
The DN-TNF variant we used, XENP345, is a mutant form of human solTNF with disrupted receptor-ligand interfaces (Steed et al., 2003; Zalevsky et al., 2007). It has been shown specifically to neutralize solTNF signaling by forming dominant negative heterotrimers that have significantly attenuated binding affinity for TNF receptors (Steed et al., 2003; Zalevsky et al., 2007) thereby lowering the effective concentration of solTNF without affecting overall production levels of TNF or the biological activity of transmembrane TNF (tmTNF), including its role in maintaining innate immunity (Zalevsky et al., 2007). SolTNF signals primarily through TNFR1 (Grell et al., 1998) and tmTNF signals primarily through TNFR2 (Grell et al., 1995). Therefore, selectivity of DN-TNF for solTNF may be highly relevant in the context of the brain and specifically the hippocampus, considering that TNF signaling through TNFR2 is protective against glutamate excitotoxicity (Marchetti et al., 2004), reduces seizures in response to kainic acid (Balosso et al., 2005), and most likely promotes the survival of hippocampal neuroblasts after ischemic injury (Heldmann et al., 2005). Systemic administration of drugs that inhibit both tmTNF and solTNF, such as etanercept or infliximab, have been associated with serious side effects including an increased susceptibility to infection and demyelinating disease (Scheinfeld, 2004). These effects may be related to the ability of these biologics to inhibit tmTNF signaling, which plays an important role in resolving inflammation and maintaining immunity to certain pathogens (Alexopoulou et al., 2006). Alternatively, introduction of the DN-TNF sequence into a lentiviral vector may provide several advantages over chronic infusion of DN-TNF biologics. Lentiviruses allow for long-term expression of encoded proteins (Blomer et al., 1997; Jakobsson and Lundberg, 2006), and a targeted lentiviral injection will afford long-term inhibition of TNF signaling in a specific brain region without the invasiveness of a chronic infusion pump. In the future, inducible viral vectors may permit precisely timed delivery of the drug (Blesch et al., 2005) while circumventing the inherent limitations of using a chronic infusion device.
The association between neuroinflammation and neurodegenerative diseases, including AD, has been investigated extensively (Eikelenboom et al., 2006; Griffin, 2006; Hoozemans et al., 2006; McGeer and McGeer, 2003; Mrak and Griffin, 2005; Wyss-Coray and Mucke, 2002). Epidemiological studies suggest a link between chronic use of non-steroidal anti-inflammatory drugs (NSAIDs) and reduced risk for AD, but thus far clinical trials using systemic administration have yielded mixed or inconclusive results (Launer, 2003; McGeer and McGeer, 2007; van Gool et al., 2003), reflecting the need to identify and target the key inflammatory mediators that promote amyloid-associated neuropathology. Our in vivo findings provide strong support for TNF as a critical inflammatory modulator of APP turnover and/or modulation of early amyloid-associated pathology in hippocampal and cortical neurons. Interestingly, neuronal expression of hAPP/Aβ in the J20 transgenic model of AD has been shown to be sufficient to reduce Reelin expression in a specific population of entorhinal cortical pyramidal neurons in vivo and qualitatively similar reductions of Reelin-expressing pyramidal neurons have been reported in the entorhinal cortex of AD brains (Chin et al., 2007). Therefore, one speculative possibility raised by our findings that will need to be addressed in future studies is that neuroinflammatory processes may contribute to accelerated cognitive decline in AD by driving TNF-dependent intraneuronal accumulation of APP β-CTF in Reelin-positive interneurons and promoting abnormal Reelin processing.
Activated microglia cluster around amyloid plaques (McGeer et al., 1987), and β-secretase (BACE1) has been shown to be upregulated in neuronal populations that are in close proximity to amyloid plaques (Zhao et al., 2007). Physiologically, one of the main effector pathways downstream of TNF (Hohmann et al., 1990), the NFκB pathway, can regulate BACE1 expression after exposure to Aβ (Bourne et al., 2007). Mechanistically, genetic ablation of BACE1 in the 5xFAD APP/PS1 transgenic mouse a model of AD resulted in marked reduction of amyloid burden, neuron loss and cognitive deficits (Ohno et al., 2007). Given that TNF can affect Aβ production (Blasko et al., 1999; Blasko et al., 2000) through upregulation of BACE1 expression (Yamamoto et al., 2007) and γ-secretase activity (Liao et al., 2004) as well as expression of APP itself (Lahiri et al., 2003), our findings support an important role for TNF-dependent regulation of APP processing in vivo and underscore the importance of targeting molecular mediators that promote BACE1 activity to pathophysiological levels. In support of this idea, genetic deletion of the TNF receptor gene Tnfrsf1a in APP 23 transgenic mice resulted in reduced amyloid plaque number and cognitive deficits in these mice (He et al., 2007).
Although our biochemical and immunohistological analyses did not reveal robust accumulation of Aβ peptides in 3xTgAD mice under basal conditions or in response to chronic systemic inflammation, our studies strongly suggest that it may be feasible to selectively target solTNF therapeutically to prevent dysregulated APP turnover and/or transport induced by chronic neuroinflammation, thereby halting or attenuating disease progression and cognitive impairment. It has been proposed that in advanced stages of AD once extensive amyloid plaques have formed, elevated levels of pro-inflammatory cytokines, including TNF, may inhibit phagocytosis of toxic Aβ species and/or hinder efficient plaque removal by brain resident microglia (Koenigsknecht-Talboo and Landreth, 2005). If abnormal APP turnover/processing and/or intraneuronal accumulation of APP, APP C-terminal fragments, or Aβ can be blocked in the earliest stages of AD with solTNF-selective inhibitors, we speculate that it may also be possible to prevent or significantly delay appearance of extracellular plaques and tau pathology given that Aβ has been shown to influence the progression of tau pathology (Gotz et al., 2001; Lewis et al., 2001). In support of this idea, chronic elevation of TNF signaling in the hippocampus of 3xTgAD mice achieved via adeno-associated virus enhanced the appearance of intraneuronal amyloid and hyperphosphorylated tau ultimately leading to neuronal death (Janelsins et al., 2008). In conclusion, our findings offer support for two novel and specific approaches to inhibit the accumulation of intraneuronal amyloid-associated proteins triggered by chronic systemic inflammation, and provide further proof-of-concept that solTNF is a valid therapeutic target to modify disease progression during the early stages of AD.
The authors would like to thank D. E. Szymkowski of Xencor, Inc. for providing us with the DN-TNF biologic XENP345 and J. Zalevsky of Xencor, Inc. for the plasmids containing the DN-TNF sequence. We thank M. Lavoie and D.Selkoe of Harvard Medical School and Brigham and Women's Hospital for the anti-APP C9 antibody; R. Vassar of the Northwestern Feinberg School of Medecine for the anti-BACE1 antibody; R. Bachoo and L. Loomis for their gift of the SS01 murine astrocyte cell line; B. R. Botterman, K.E. Tansey, and C. Wang for assistance with stereotaxic surgeries; B. Pearson and T.C. Frank-Cannon for assistance with immunohistochemical techniques and statistical analyses, M.K. McCoy for assistance with TNF ELISAs, and members of the Tansey lab for insightful discussions. This work was supported by an Alzheimer's disease research grant from the American Health Assistance Foundation (MGT), the Alzheimer's Disease Center at UT Southwestern (NIA, NIH P30AG12300), and NIH pre-doctoral training grant T32 GM 008203 (FEM).