Essential protective role of tumor necrosis factor receptor 2 in neurodegeneration

Significance

TNF is known to play an important role in various neurodegenerative diseases. However, anti-TNF therapeutics failed in clinical trials of neurodegenerative diseases. This failure is most likely due to antithetic effects of the TNF receptors in the central nervous system, whereby TNFR1 promotes inflammatory degeneration and TNFR2 neuroprotection. Here we show that novel TNFR-selective therapeutics, i.e., a TNFR1 antagonist and a TNFR2 agonist, block neuroinflammation and promote neuronal survival in a mouse model of neurodegeneration related to Alzheimer disease as well as other neurodegenerative diseases. Most important, neuroprotection mediated by the TNFR1 antagonist is abrogated by simultaneous blockade of TNFR2 activation, revealing that neuroprotection requires TNFR2 signaling and uncover why anti-TNF drugs failed in treatment of neurodegenerative diseases.

Abstract

Despite the recognized role of tumor necrosis factor (TNF) in inflammation and neuronal degeneration, anti-TNF therapeutics failed to treat neurodegenerative diseases. Animal disease models had revealed the antithetic effects of the two TNF receptors (TNFR) in the central nervous system, whereby TNFR1 has been associated with inflammatory degeneration and TNFR2 with neuroprotection. We here show the therapeutic potential of selective inhibition of TNFR1 and activation of TNFR2 by ATROSAB, a TNFR1-selective antagonistic antibody, and EHD2-scTNF_R2, an agonistic TNFR2-selective TNF, respectively, in a mouse model of NMDA-induced acute neurodegeneration. Coadministration of either ATROSAB or EHD2-scTNF_R2 into the magnocellular nucleus basalis significantly protected cholinergic neurons and their cortical projections against cell death, and reverted the neurodegeneration-associated memory impairment in a passive avoidance paradigm. Simultaneous blocking of TNFR1 and TNFR2 signaling, however, abrogated the therapeutic effect. Our results uncover an essential role of TNFR2 in neuroprotection. Accordingly, the therapeutic activity of ATROSAB is mediated by shifting the balance of the antithetic activity of endogenous TNF toward TNFR2, which appears essential for neuroprotection. Our data also explain earlier results showing that complete blocking of TNF activity by anti-TNF drugs was detrimental rather than protective and argue for the use of next-generation TNFR-selective TNF therapeutics as an effective approach in treating neurodegenerative diseases.

Tumor necrosis factor (TNF) is a master proinflammatory cytokine that plays an important role in the initiation and orchestration of immunity and inflammation (1, 2). Elevated TNF levels have been associated with different autoimmune diseases, and deregulation of TNF expression and signaling can lead to chronic inflammation and tissue damage (3–6). Therefore, several anti-TNF therapeutics are clinically approved and successfully used to treat autoimmune diseases, such as rheumatoid arthritis, psoriasis, or inflammatory bowel disease. Up-regulated TNF expression has also been associated with different neurodegenerative diseases, e.g., Alzheimer’s disease, Parkinson’s disease, stroke, and multiple sclerosis (MS) (7). However, an anti-TNF therapeutic failed in a phase II randomized, multicenter, placebo-controlled study for the treatment of relapsing remitting MS (Lenercept study) because symptoms of Lenercept-treated patients were significantly increased compared with patients receiving placebo, and neurologic deficits tended to be more severe in the Lenercept treatment groups (8). Next to induction or aggravation of demyelinating diseases, all anti-TNF therapeutics may induce severe side effects such as serious infections, including reactivation of tuberculosis and invasive fungal and other opportunistic infections. An increased susceptibility to develop additional autoimmune diseases and lymphomas has also been reported (3).

The negative results of the Lenercept study and the observed severe side effects of anti-TNF therapeutics in approved indications might be explained by the pleiotropic actions of TNF, including both pro- and antiinflammatory functions and other immune regulatory as well as regenerative activities. Blocking all effects of TNF therefore might be counterproductive. Because most of the proinflammatory actions of TNF are mediated by TNFR1, a more effective therapeutic approach could be the selective blocking of TNFR1 signaling; this would spare TNFR2 signaling, which has been implicated in various protective and regenerative responses, particularly in the central nervous system: TNFR2 signaling was shown to promote neuronal survival and oligodendrocyte regeneration in in vivo models of ischemic and neurotoxic insults (9, 10), respectively. Specific activation of TNFR2 rescues neurons (11) and oligodendrocytes (12) from oxidative stress and promotes oligodendrocyte differentiation and myelination (13, 14). In addition, TNFR2 signaling protects neurons from glutamate-induced excitotoxicity in vitro (15, 16).

Glutamate, a key neurotransmitter, can interact with the ionotropic AMPA and NMDA receptors. Exacerbated activation of these glutamate receptors may lead to progressive neuronal cell death, which is a hallmark of acute and chronic neurological diseases (17). Consequently, NMDA receptor antagonists such as dizocilpine (MK-801) or memantine were developed and revealed neuroprotective effects against ischemia-induced neuronal death in vitro and ischemic brain damage in vivo (18, 19). However, inhibitors of NMDA receptors largely failed in clinical studies (20).

Because TNFR2 signaling protects neurons from glutamate-induced excitatory cell death in vitro, ligands promoting TNFR2 signaling might be superior to NMDA antagonists because they do not completely inhibit glutamate-induced signal transmission, but buffer excitotoxicity likely by acquisition of a resistant state of affected cells. Because of the antithetic action of TNF via its two receptors, we reasoned that both an inhibition of TNFR1 signaling and a selective activation of TNFR2 signaling, respectively, could shift the balance of endogenous TNF activity toward an overall neuroprotective/regenerative response. We here report on the therapeutic activity of the human TNFR1-selective antagonist ATROSAB (21, 22) and a human TNFR2 selective agonist (EHD2-scTNF_R2) in in vitro and in vivo models of excitotoxic brain damage (23, 24) using humanized TNFR1 and TNFR2 knock-in mice.

Results

Generation and Characterization of a TNFR2 Selective Agonist.

Efficient activation of TNFR2 requires receptor oligomerization by the membrane form of TNF (tmTNF) or oligomerized soluble forms of tmTNF mimicking receptor activation via tmTNF (11, 25) such as a TNFR2-selective TNF mutein, fused to the trimerization domain tenascin C (TNC–scTNF_R2) recently developed by the authors (11). We here describe a TNFR2-selective TNF mutein, which consists of a covalently stabilized human TNFR2-selective (D143N/A145R) (25) single-chain TNF (scTNF_R2) fused to the dimerization domain EHD2 derived from the heavy chain domain CH2 of IgE (26), constituting a disulfide bonded dimer that is, with respect to TNF domains, hexameric (EHD2–scTNF_R2; Fig. 1A). The purity of the recombinant protein was confirmed by SDS/PAGE and Coomassie staining (Fig. 1B). Under reducing conditions, the TNF variant exhibited an apparent molecular mass of ∼70 kDa, matching the calculated molecular mass of 68.35 kDa. Under nonreducing conditions the expected dimer of ∼130 kDa was observed (Fig. 1B). The oligomerization state of EHD2–scTNF_R2 was further characterized by size exclusion chromatography (SEC; Fig. 1C). EHD2–scTNF_R2 eluted as a single major peak, indicating high purity.

Fig. 1.

Genetic engineering and bioactivity of EHD2–scTNF_R2. (A) Schematic representation of EHD2–scTNF_R2. (B) Coomassie staining and immunoblot of EHD2–scTNF_R2 under reducing (column 1) and nonreducing (column 2) conditions. (C) HPLC-SEC analysis of EHD2–scTNF_R2 using a BioSep-Sec-2000 column. Peak positions of relevant standard proteins are indicated. (D and E) Binding of soluble huTNF (□) and membrane-TNF mimetic EHD2–scTNF_R2 (●) to huTNFR1 (huTNFR1–Fc) or huTNFR2 (Etanercept; huTNFR2–Fc) was determined by ELISA (n = 3; ±SEM). (F) L929 cells were cultivated with titrations of recombinant human TNF (huTNF) or EHD2–scTNF_R2. Cell viability was determined by crystal violet staining (n = 4; mean ± SEM). (G) Kym-1 cells were stimulated with (open symbols) or without (closed symbol) the TNFR2 cross-linking mAb 80M2 for 30 min at 37 °C. Then titrations of scTNF_R2 (squares) or EHD2–scTNF_R2 (circles) were added and cells were incubated for 24 h. Cell viability was determined by crystal violet staining (n = 4–6; mean ± SEM).

TNF receptor selectivity of EHD2–scTNF_R2 was analyzed by binding studies with immobilized huTNFR1–Fc and huTNFR2–Fc fusion proteins. Whereas EHD2–scTNF_R2 did not interact with huTNFR1, the fusion protein efficiently bound to huTNFR2 (Fig. 1D). In contrast, soluble human TNF (huTNF) efficiently bound to huTNFR1, whereas it just weakly interacted with huTNFR2 (Fig. 1D). Furthermore, EHD2–scTNF_R2 did not activate TNFR1-dependent cell death in L929 (Fig. 1E), verifying that EHD2–scTNF_R2 had lost affinity for TNFR1 due to the mutations D143N/A145R. In contrast, EHD2-scTNF_R2 efficiently induced cell death in Kym-1 cells, which endogenously express both TNF receptors and are highly sensitive to endogenous TNF-induced TNFR1 mediated cytotoxicity (27) (Fig. 1G). Interestingly, induction of cell death was not enhanced by addition of the cross-linking TNFR2-specific monoclonal antibody 80M2, which in concert with soluble TNF is able to mimic the action of tmTNF (11, 28). Differently, the bioactivity of scTNF_R2, a covalently stabilized huTNFR2-selective TNF trimer, was strongly enhanced in the presence of 80M2 (Fig. 1G), indicating that dimerization of scTNF_R2 via EHD2 is necessary and sufficient to fully activate huTNFR2.

ATROSAB and EHD2–scTNF_R2 Are Neuroprotective in Vivo.

To assess both EHD2–scTNF_R2 and ATROSAB in an in vivo model of neurodegeneration, hu/mTNFR-knock-in mice (Figs. S1–S4) were generated and used in the nucleus basalis lesion model. The hu/mTNFR knock-in (k/i) mice express a chimeric TNFR, where the extracellular part of the human receptor is fused to the transmembrane- and intracellular region of the mouse TNFR. This chimeric hu/mTNFR was introduced into the germline of C57BL/6 mice by knock-in technology, replacing the endogenous mouse TNFR, thereby using the regulatory elements from the wild-type mouse TNFR; this should lead to an expression pattern of the hu/mTNFR comparable to the wild-type mouse TNFR. The expression and functionality of the chimeric hu/mTNFR in hu/mTNFR-k/i mice was investigated using primary cells isolated from different tissues and in vivo experiments. Using primary immune cells, neurons, and mouse embryonic fibroblasts (MEFs), we demonstrated that expression of chimeric hu/mTNFR1 and hu/mTNFR2 resemble the expression pattern of the wild-type TNFRs (Figs. S5 and S6). Furthermore, we demonstrated the functionality of the chimeric hu/mTNFRs both using in vitro experiments with primary MEFs and thymocytes as well as in vivo experiments showing TNF sensitivity of hu/mTNFR1-k/i mice (Fig. S7).

Fig. S1.

Humanized hu/mTNFR1-k/i mouse models. Schematic representation of the homologous recombination in the TNFR1 gene locus. Shown is the wild-type genomic locus (A), the targeted locus, including the neo cassette (B), and the genomic locus of chimeric mice (Cre’d locus; C).

Fig. S4.

Southern blot analysis of mouse tissue. DNA from lung, heart, liver, brain, thymus, kidney, and skin tissue from homozygous hu/mTNFR1-k/i (A). (B) hu/mTNFR2-k/i mice were digested by BamHI, blotted on nylon membrane, and probed with the endogenous probes enP. Genomic DNA from tail biopsies of wild-type mice was used as a control. The DNA size marker phage Lambda DNA/StyI marker (Bioron) was used to identify size of products. In all samples, the neo cassette is removed as evident by the 5.9-kb (A; hu/mTNFR1-k/i) or 8.2-kb (B; hu/TNFR2-k/i) product.

Fig. S5.

Expression of transgenic hu/mTNFR1-k/i and hu/mTNFR2-k/i in homozygous mice. Splenocytes or thymocytes were isolated from C57BL/6J wild-type (A and B), hu/mTNFR1-k/i (A), or hu/mTNFR2-k/i (B) mice. Then mouse TNFR1 (HP8002) and human TNFR1 (HP9002) (A) or mouse TNFR2 (HP8003) and human TNFR2 (HP9003) (B) expression was analyzed by flow cytometry using subgates for CD45R⁺ B cells, CD68⁺ macrophages (splenocytes), and CD8⁺ thymocytes. Gray histograms show the isotype controls; black lines indicate signals measured for TNFR1 or TNFR2 expression. Data are presented as normalized to unit area.

Fig. S6.

Expression of transgenic hu/mTNFR1-k/i and hu/mTNFR2-k/i in primary cells isolated from homozygous mice. Primary MEFs were isolated from wild-type (A and B), hu/mTNFR1-k/i (A), or hu/mTNFR2-k/i (B) C57BL/6 mice. Expression of mouse TNFR1 (HP8002) and human TNFR1 (HP9002) (A) or mouse TNFR2 (HP8003) and human TNFR2 (HP9003) (B) was analyzed by flow cytometry. Data are presented as normalized to unit area. (C and D) Brain tissue isolated from wild-type (C and D), hu/mTNFR1-k/I (C), or hu/mTNFR2-k/i (D) C57BL/6 mice. Tissue was lyzed and expression of huTNFR1 (wild-type, hu/mTNFR1-k/i, H398) or huTNFR2 (wild-type, hu/mTNFR2-k/i, MR2-1) was analyzed by Western blot.

Fig. S7.

Functionality of transgenic hu/mTNFR1-k/i and hu/mTNFR2-k/i mice. (A) Primary MEFs were isolated from hu/mTNFR1-k/i C57BL/6 mice. Cells were incubated with PBS or ATROSAB (20 µg/mL) for 60 min, followed by addition of rhTNF (1 ng/mL) for up to 40 min. Then, cells were fixed and localization of NF-κB p65 was analyzed by indirect immunofluorescence (n = 4; ±SD; >500 cells per condition; *P < 0.05; **P < 0.001). (B) MEF from hu/mTNFR1-k/i C57BL/6 mice were incubated with ATROSAB (20 µg/mL) for 60 min, followed by addition of rmTNF for 24 h. Then, supernatants were harvested and analyzed for presence of IL-6 by ELISA (n = 4; mean ± SEM). (C) MEF from hu/mTNFR1-k/i C57BL/6 mice were incubated with ATROSAB (20 µg/mL) for 60 min, followed by addition of rmTNF or rhTNF (each 1 ng/mL) for 2 h. Then, RNA was isolated converted into cDNA and expression of CXCL2 was quantified by qPCR (n = 2; mean ± SD). (D) hu/mTNFR1-k/i C57BL/6 mice were treated with mouse TNF (30 µg) or saline (i.v.). Animals were monitored over a period of 24 h, and body temperature and weight was documented (n = 4–6 mice per group, ±SEM). (E) MEFs isolated from hu/mTNFR2-k/i mice were stimulated with or without EHD2–scTNF_R2 (100 ng/mL) for 2 h. Then cells were lyzed, RNA was isolated and transcribed into cDNA, and expression of CXCL2 and IL-6 was quantified by qPCR (n = 2 ± SD). (F) Thymocytes were isolated from hu/mTNFR2-k/i mice and cultivated together with immobilized/plate-bound anti-CD3 and EHD2–scTNF_R2 (100 ng/mL) for 4 d. Amount of living cells was determined by MTT assay (n = 3; ±SEM).

Fig. S2.

Humanized hu/mTNFR2-k/i mouse models. Schematic representation of the homologous recombination in the TNFR2 gene locus. Shown is the wild-type genomic locus (A), the targeted locus, including the neo cassette (B), and the genomic locus of chimeric mice (Cre’d locus; C).

Fig. S3.

Southern blot analysis of ES cells. (A) Genomic integration of hu/mTNFR1 is shown for nine embryonic stem cells. DNA was digested with EcoRV and probed with a 5′ probe. The expected sizes are wt 11.9 kb and targeted 7.7 kb (Left). DNA was digested with NheI. Probing with the 3′ probe resulted in products of 11.1 kb (wt) and 7.5 kb (targeted) (Right). (B) Genomic integration of hu/mTNFR2 is shown for nine embryonic stem cells. DNA was digested with PshAI and probed with the 5′ probe. The expected sizes are wt 26.7 kb and targeted 13.5 kb (Left). Digestion with EcoRV and probing with the 3′ probe resulted in products of 32.4 kb (wt) and 17.8 kb (targeted) (Right). Size marker: Lambda DNA/HindIII (Thermo Fisher Scientific).

By stereotactic NMDA injection, lesions in the nucleus basalis magnocellularis (NBM) were generated. Lesioning of the NBM resulted in a decrease of body weight (Fig. S8), reduction of cholinergic fibers of the parietal cortex (Fig. 2 B and F), and macrophage/microglial activation within the NBM (Fig. 2 D and H and Fig. S9). These symptoms can be assessed to determine the lesion size. Both EHD2–scTNF_R2 and ATROSAB show a strong protective effect in this in vivo model when simultaneously injected with NMDA into the NBM (Fig. 2 B, D, F, and H and Fig. S9). Injection of a control IgG (anti-huEGFR), however, did not significantly alter NMDA-mediated neurodegeneration (Fig. 2 F and H).

Fig. S8.

Body weight changes after treatment in both mouse lines. (A) EHD2–scTNF_R2 significantly prevented the reduction of body weight induced by NMDA injection into the NBM in hu/mTNFR2-k/i mice at 2 and 4 d after injection. (B) ATROSAB prevented the significant reduction of body weight induced by NMDA injection into the NBM in hu/mTNFR1-k/i mice at 1 d after injection, but this was not statistically significant. (C) Body weight effects of the TNFR2 antagonistic antibody MAB426, when administered together with NMDA alone, or NMDA and ATROSAB together.

Fig. 2.

EHD2–scTNF_R2 and ATROSAB prevent NMDA-induced NBM lesion and macrophage/microglia activation. (A–D) hu/mTNFR2-k/i and (E–H) hu/mTNFR1-k/i mice were used. (A and E) Representative images show choline acetyltransferase (ChAT)-positive cholinergic innervations in the somatosensory cortex. NMDA injected into the NBM induced an extensive cholinergic fiber loss in the layer V of somatosensory cortex compared with the control group. However, EHD2–scTNF_R2 or ATROSAB treatment attenuated fiber loss. Parallel bars in indicate the layer V of the somatosensory cortex in which quantitative measurements were performed. (B and F) Quantification of cholinergic fiber density in layer V of the somatosensory cortex. Fiber density was measured in eight sections per mouse, n = 7 mice per group. All data in bar charts represent means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.0001, one-way ANOVA with post hoc comparisons Tukey. (C and G) Representative images show CD11B-positive activated macrophage/microglia in magnocellular nucleus basalis. NMDA injected into the NBM induced a massive volume of macrophage/microglial activation, compared with those in both control groups. However, EHD2–scTNF_R2 or ATRSOAB treatment significantly reduced macrophage/microglial activation induced by NMDA. (D and H) Quantification of total extent of activated macrophage/microglia around the injections. Macrophage/microglial activation was measured in a series of sections with macrophage/microglial activation, n = 7 mice per group. All data in bar charts represent means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.0001, one-way ANOVA with post hoc comparisons Tukey.

Fig. S9.

EHD2–scTNF_R2 and ATROSAB prevent NMDA-induced macrophage/microglia activation. Representative images show IBA1-positive activated macrophage/microglia in magnocellular nucleus basalis. (A) Treatment of hu/mTNFR1-k/i mice with NMDA and/or ATROSAB and MAB426 is shown. (B) Treatment of hu/mTNFR2-k/i mice with NMDA and/or EHD2–scTNF_R2 is shown. NMDA injected into the NBM induced a massive volume of macrophage/microglial activation, compared with those in both control groups. However, EHD2–scTNF_R2 or ATROSAB treatment significantly reduced macrophage/microglial activation induced by NMDA.

ATROSAB and EHD2–scTNF_R2 Improve Memory Performance in the NBM Lesion Model.

Memory was evaluated as parameter for the protective effects of EHD2–scTNF_R2 in the hu/mTNFR2-k/i mice and ATROSAB in the hu/mTNFR1-k/i mice against NMDA-induced neurotoxicity in the nucleus basalis and subsequent cholinergic fiber loss in parietal cortex. The sequence of memory tests performed in hu/mTNFR2-k/i and hu/mTNFR1-k/i mice are depicted in Fig. 3A and in Figs. S10 and S11. The NMDA-induced NBM lesion and the compounds tested had no significant effect on short-term memory performance in the spontaneous alternation task (Figs. S10B and S11B) or number of entries recorded (Figs. S10C and S11C). Anxiety-like behavior assessed by the elevated plus maze showed no significant changes in the total number of entries (Figs. S10D and S11D), entries into the open arms (Figs. S10E and S11E), and time spent in the dark and light arms and center of the maze (Figs. S10F and S11F) in all of the mice tested. Results from the passive avoidance paradigm showed that NMDA, EHD2–scTNF_R2, and ATROSAB did not result in differences of the preshock latency to enter the dark compartment in the hu/mTNFR2-k/i mice (Fig. 3B) and hu/mTNFR1-k/i mice (Fig. 3D). However, NBM injection of NMDA caused a significant impairment in the postshock latency in all of the mice tested (Fig. 3 C and E). Both EHD2–scTNF_R2 and ATROSAB obliterated NMDA-mediated memory impairment (Fig. 3 C and E). EHD2–scTNF_R2 and ATROSAB when given without NMDA had no effect on postshock latency (Fig. 3 C and E). These results indicate the lack of neurotoxicity of EHD2–scTNF_R2 and ATROSAB and accentuate their protective effects against NMDA toxicity.

Fig. 3.

EHD2–scTNF_R2 and ATROSAB attenuate NMDA-induced memory impairment. (A) A graphical description for the sequence of behavioral and memory assessments in hu/mTNFR1-k/i (B and C) and hu/mTNFR2-k/i (D and E) mice. Assessment of long-term memory with the passive avoidance paradigm showed no significant differences in (B and D) preshock latency between the experimental groups. NBM injection of NMDA caused a significant impairment of long-term memory, measured as postshock latency, which was obliterated by cotreatment of (C) EHD2 B and C scTNF_R2 or (D) ATROSAB. Error bars indicate ±SEM. *P < 0.05, **P < 0.01, one-way ANOVA with post hoc comparisons Tukey.

Fig. S10.

EHD2–scTNF_R2 attenuate NMDA-induced memory impairment. (A) A graphical description for the sequence of behavioral and memory assessments in hu/mTNFR2-k/i mice. NBM injection of NMDA and/or EHD2–scTNF_R2 had no effect on (B) short-term memory, presented as percentage of alternations and (C) number of total entries that were measured with spontaneous alternation. For anxiety-like behavior assessed by elevated plus maze, no significant effects in the (D) total number of entries, (E) entries into the open arms, and (F) time spent in the different arms of the maze was observed between the experimental groups. Error bars indicate ± SEM.

Fig. S11.

ATROSAB attenuates NMDA-mediated memory impairment. (A) An illustration for the series and time intervals in which behavior and memory were assessed in hu/mTNFR1-k/i mice. No significant differences were found between the experimental groups in the spontaneous alternation test for (B) the percentage of alterations and (C) the total number of entries into the different arms of the maze. Evaluation of anxiety-like behavior with the elevated plus maze showed no significant differences between the groups in the (D) number of entries into the different arms, (E) entries into the open arms, and (F) time spent in the different arms of the maze. Error bars indicate ± SEM.

Emergence of Neuroprotective TNFR2 Signaling in Vitro and in Vivo upon ATROSAB-Mediated Inhibition of TNFR1.

Previously we have shown that the PI3K–PKB/Akt pathway mediates TNFR2-promoted neuroprotection from excitotoxic cell death (10, 15, 16). To investigate the molecular pathways underlying the neuroprotective effects of ATROSAB, we isolated primary neurons from hu/mTNFR1-k/i mice and investigated TNF-induced phosphorylation of PKB/Akt (Ser473). As expected, stimulation with a nonreceptor- selective wild-type mouse variant of EHD2–scTNF_R2, with tmTNF-mimetic activity (EHD2-sc–mTNF), induced phosphorylation of PKB/Akt. Interestingly, phosphorylation was enhanced in the presence of the TNFR1 antagonist ATROSAB (Fig. 4A), suggesting that PKB/Akt activation occurs via TNFR2, and concomitant TNFR1 signaling interferes with this pathway. In addition, TNF was shown to protect primary neurons from glutamate-induced cell death in a TNFR2-dependent manner (15). Similarly, we could show that the mouse TNFR2-specific variant (D135N/A137R; 13) of the human TNFR2-selective EHD2–scTNF_R2 (EHD2-sc–mTNF_R2) protects primary hu/mTNFR1-transgenic neurons from excitotoxic cell death (Fig. 4B). In accordance with ATROSAB-mediated increase of PKB/Akt phosphorylation, ATROSAB enhanced the neuroprotective effect of EHD-sc–mTNF at lower concentrations (10 ng/mL; Fig. 4B) in this in vitro cell model, too.

Fig. 4.

ATROSAB is neuroprotective via enhanced TNFR2 signaling. (A) Primary neurons, isolated from hu/mTNFR1-k/i mice, were stimulated with or without ATROSAB (100 µg/mL) for 30 min followed by addition of wild-type EHD2-sc–mTNF (10 ng/mL). Then cells were incubated for 24 h and lyzed, and phosphorylation of PKB/Akt was quantified by Western blot (n = 4, ±SEM). (B) Primary neurons, isolated from hu/mTNFR1-k/i mice, were stimulated with or without ATROSAB (100 µg/mL) for 30 min followed by addition of non–TNFR-selective EHD2-sc–mTNF or mouse TNFR2-selective EHD2-sc–mTNF_R2. After 24 h, glutamate (5 µM) was added and cells were incubated for an additional hour. Then medium was exchanged to remove glutamate and cells were incubated for 23 h. Cell viability was determined by MTT assay. Data are shown as percentage of MTT signal of untreated control cells (n = 3; ±SEM). Representative images show (C) ChAT-positive cholinergic innervations in the somatosensory cortex or (E) CD11B-positive activated macrophage/microglia in magnocellular nucleus basalis. NMDA injected into the NBM induced an extensive cholinergic fiber loss in the layer V of somatosensory cortex and a massive volume of macrophage/microglial activation compared with the control group. However, ATROSAB treatment attenuated fiber loss and significantly reduced macrophage/microglial activation induced by NMDA. ATROSAB neuroprotection against fiber loss was prevented by TNFR2 antagonistic MAB426. However, MAB426 alone did not significantly alter NMDA-induced NBM lesion. Parallel bars indicated the layer V of the somatosensory cortex in which quantitative measurements were performed. (D) Quantification of cholinergic fiber density in layer V of the somatosensory cortex. Fiber density was measured in eight sections per mouse. (F) Quantification of total extent of activated macrophage/microglia around the injections. Macrophage/microglial activation was measured in a series of sections with macrophage/microglial activation. n = 7 mice/group. All data in bar charts represent means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.0001, one-way ANOVA with post hoc comparisons Tukey.

To prove the essential role of TNFR2 signaling in the in vivo NBM lesion model, we simultaneously injected NMDA, ATROSAB, and the mouse TNFR2-specific antagonist MAB426 (Fig. 4 C–F) into hu/mTNFR1-k/i mice. Under these conditions, the protective effect of ATROSAB on both cholinergic innervation and macrophage/microglial activation was completely reverted.

Discussion

TNF can exert opposite effects via its two receptors, TNFR1 and TNFR2. This dual role of TNF signaling becomes particularly obvious in the central nervous system (5, 6). Using the cuprizone-induced mouse model of demyelination, it was shown that TNFR2 is critical for oligodendrocyte regeneration, whereas TNF signaling via TNFR1 promoted nerve demyelination (9). Similar, investigations in a mouse model of retinal ischemia revealed that TNFR1-deficient animals were protected from ischemic lesions. In contrast, TNFR2^−/− mice showed enhanced neuronal loss and a more severe pathology compared with wild-type animals, indicating an antagonistic function of the two TNFRs. In the retinal ischemia model, TNFR2’s neuroprotective activity in vivo was associated with PI3K-PKB/Akt activation, which was counterbalanced by the neurodegenerative action of TNFR1 (10). The same principle applied to cortical neurons, where TNFR2 mediated in vitro protection from glutamate-induced cell death in a PI3K-PKB/Akt-dependent manner (15, 16). We here show in an in vivo model of NMDA-induced cellular degeneration and loss of neuronal functions that both a TNFR2 selective agonist (EHD2–scTNF_R2) and a TNFR1 antagonistic antibody (ATROSAB) protect from loss of cholinergic fibers and associated neurologic deficits.

Exacerbated activation of glutamate receptor-coupled calcium channels and subsequent increase in intracellular calcium concentrations ([Ca²⁺]_i), followed by sustained disturbances in the [Ca²⁺]_i homeostasis, are established hallmarks of neuronal cell death in acute and chronic neurological diseases (29, 30). Interestingly, activation of potassium intermediate/small conductance calcium-activated channel K_Ca2 prevented [Ca²⁺]_i deregulation and reduced neuronal death following glutamate toxicity and cerebral ischemia (31). We previously demonstrated that the neuroprotective effect of TNFR2 against a glutamate challenge was associated with an increased expression of K_Ca2.2 channels (16), outlining a potential molecular mechanism of TNFR2-mediated protection of neurons from death during exposure of a priori excitotoxic stimuli.

Studies on lovastatin actions further support TNFR2 involvement in neuroprotection. Lovastatin is a cholesterol-lowering drug with reported neuroprotective properties that can reduce the incidence of stroke and progression of Alzheimer's disease. Lovastatin increased the expression of TNFR2 in cortical neurons in vitro (32) and was neuroprotective in TNFR1^−/−neurons, whereas lovastatin’s protection was lost in neurons from TNFR2^−/− mice (32). Furthermore, lovastatin-mediated neuroprotection led to an increase in PI3K-dependent PKB/Akt phosphorylation, whereas inhibition of PKB/Akt activation entirely abolished lovastatin-induced neuroprotection. This finding is in line with previous findings that TNFR2-mediated neuroprotection is dependent on the PI3K-PKB/Akt pathway (15), and suggests that lovastatin-induced neuroprotection is dependent on TNFR2 signaling. Administration of lovastatin protected cholinergic neurons and their cortical projections against NMDA-induced excitotoxic damage in vivo (23). Furthermore, treatment with the PI3K inhibitor LY294002 to block activation of PKB/Akt resulted in a strong reduction of lovastatin-mediated neuroprotection (23).

In the present report, we evaluated the therapeutic potential of the TNFR2 agonist EHD2–scTNF_R2 and the TNFR1 antagonistic antibody ATROSAB in this mouse model of NMDA-induced neurodegeneration. Similar to lovastatin, both EHD2–scTNF_R2 and ATROSAB reduced the extent of areas of macrophage/microglia activation at the site of the lesion and protected cholinergic neurons and the neocortical innervations against NMDA-induced excitotoxic damage. Functional consequences of NMDA-induced lesions and the therapeutic activity of TNFR targeting reagents became phenotypically apparent in behavioral performance studies: damage to the NBM selectively affected neocortical cholinergic denervation and its memory functions, while leaving particular hippocampal innervation and functions unaffected. Treatment of such animals with either EHD2–scTNF_R2 or ATROSAB fully restored the affected cholinergic memory function.

ATROSAB-induced protection from excitotoxicity induced neurodegeneration was found to be linked to an enhancement of TNFR2 signaling leading to PKB/Akt activation; this is evident from abrogation of neuroprotection in vivo upon cotreatment with TNFR2 blockers and from in vitro studies showing increased phosphorylation of PKB/Akt as well as enhanced resistance of primary cortical neurons toward excitotoxic insult by a tmTNF mimetic TNF in the presence of a TNFR1 blockade. We propose that the neuroprotective activity of ATROSAB is accomplished by simultaneous action on two different cellular targets: (i) as a consequence of the excitotoxic insult, activated macrophage/microglia promote, via soluble TNF production, the expansion of the degenerative tissue response in an autocrine TNFR1-dependent way; and (ii) inhibition of TNFR1 limits this process both at the level of microglia/macrophage activation and at the level of neurons of TNFR1-mediated degenerative signaling. Further, for neuronal cells, it was shown previously that proper TNFR2 activation induces protection from excitotoxicity in vitro (15, 16); in the in vivo model analyzed here, in neurons in the vicinity of the acute insult and not immediately succumbing to excitotoxic death, competition of TNFRs for limiting amounts of endogenously produced membrane form of TNF exists, which results in suboptimal activation of TNFR2 of this cell population under a nontreatment condition. Moreover, at the level of intracellular signaling, it is known that signal cross-talk between TNFR1 and 2 exists in terms of competition for common signal transducers such as TRAF2 (33–35), with TNFR1 outperforming TNFR2 pathways in the case of abundance of soluble TNF, which triggers exclusively TNFR1 (3, 6). In the presence of ATROSAB, neuronal response is shifted toward neuroprotective TNFR2 signaling because of blocked neuronal TNFR1, allowing the induction of a resistant state by endogenous tmTNF (Fig. 5). The therapeutic activity of TNFR2 activation by treatment with EHD2–scTNF_R2 without interfering with TNFR1 signaling supports the view of limited TNFR2 activation by endogenous TNF. Accordingly, TNFR2’s role as an important neuroprotective pathway emerges when TNFR1 ligand binding is stalled and/or when TNFR2 activation is optimized by application of exogenous TNFR2-selective ligands.

Fig. 5.

Mechanism of TNFR1 antagonist ATROSAB and TNFR2 agonist EHD2–scTNF_R2-mediated neuroprotection. (A) TNFR1 signaling induces neurodegeneration and inflammation, whereas TNFR2 signaling promotes neuroprotection. (B) ATROSAB, an antagonistic TNFR1-selective antibody, can block TNFR1 signaling, leading to prevention of neurodegeneration and inflammation. More importantly, in the presence of ATROSAB, tmTNF stays free to activate neuroprotective signaling via TNFR2. Our data indicate that, by blocking TNFR1, ATROSAB shifts the endogenous tmTNF signaling toward neuroprotective TNFR2 signaling. Similarly, the TNFR2 agonist EHD2–scTNF_R2 can directly activate TNFR2 and thereby induce neuroprotection.

Importantly, initial studies with anti-TNF therapeutics showed that TNF blocking drugs cannot be used to treat neurodegenerative diseases such as MS. In support of this finding, we here provide compelling evidence in an in vivo model of NMDA-induced acute neuronal lesions that abrogation of complete TNF signaling by blocking both TNFRs is not protective, because mounting of a neuroprotective TNFR2-dependent response is prevented. However, preclinical studies in the experimental autoimmune encephalomyelitis (EAE) model (36–38) or a mouse model of spinal cord injury (39) revealed that selective neutralization of sTNF/TNFR1 signaling is neuroprotective. Of note, in the EAE model, systemic application of TNFR1-blocking antibodies proved to be therapeutically effective (40). Accordingly, TNFR1 antagonists such as ATROSAB should be superior to conventional anti-TNF drugs in the treatment of neurodegenerative diseases, as they spare TNFR2 and even enhance TNFR2 signaling but still block detrimental signals transmitted via TNFR1. In addition to the potential therapeutic use of TNFR1-specific antagonistic antibodies, TNFR2-selective agonists seem to be particularly suitable to treat inflammatory, demyelinating diseases, because next to the direct neuroprotective effects shown in this report, data from different laboratories outline that TNFR2 is also involved in immune suppression via expansion and stabilization of regulatory T cells (41–47), and induces remyelination (9, 13, 14). Thus, like TNFR1 antagonists, TNFR2 agonists might promote therapeutic effects via multiple cellular targets. Limiting for the treatment of neurodegenerative diseases is the blood–brain–barrier (BBB), which restricts the transport of therapeutics into the CNS. Under physiologic conditions, antibodies, e.g., ATROSAB, do not cross the BBB. However, of relevance, the BBB undergoes rapid changes in permeability in response to CNS insults, e.g., after spinal cord injury or during MS, potentially enabling therapeutics to reach the CNS parenchyma without specific transport mechanisms. Indeed, the reported therapeutic activity of ATROSAB in EAE models (40) supports this reasoning. Whether alterations of the BBB permeability in different neurodegenerative diseases are sufficient to reach therapeutic effective concentrations in the CNS is presently unclear. Therefore, strategies are now available to overcome limits of BBB passage. Thus, antibodies engineered to be actively transported via BBB have been developed that were shown to be beneficial for the treatment of Alzheimer’s disease (48, 49). Cytokines including TNF can cross the BBB (50). By analogy, we assume that the engineered TNF mutein EHD2–scTNF_R2 could be able to penetrate the BBB, too.

In summary, in an acute neurodegenerative disease model, we provide proof of concept that both the TNFR2-selective TNF variant EHD2–scTNF_R2 and the antagonistic TNFR1-selective antibody ATROSAB protect from neurological deficits due to excitotoxic cell death induced by excessive glutamate exposure in vivo. Our data provide a rational base for previous failure of clinical studies with anti-TNF drugs in neurodegenerative diseases and highlight the essential protective role of TNFR2 in the central nervous system. Further investigations in additional neurodegenerative diseases models on the therapeutic potential of the TNFR2 agonist EHD2–scTNF_R2 and the TNFR1 antagonist ATROSAB are warranted and will reveal whether and for which indications these or functionally similar proteins can be successfully applied.

Materials and Methods

Detailed methods can be found in SI Materials and Methods. The TNFR2 agonist EHD2–scTNF_R2 and the tmTNF-mimetic TNF mutein EDH2–scTNF are described in this work. The transgenic hu/mTNFR-k/i mice were generated as contracted by Ozgene Pty Ltd. Animal care and treatment were carried out in accordance with Committee Groningen (Dierexperimenten Comissie Van de Rijksuniversiteit Groningen) and Committee Stuttgart (Regierungspräsidium Stuttgart) guidelines on the use of experimental animals at the University of Groningen (permit no. DEC6523) and the University of Stuttgart (permit no. 35-9815.81-0350), respectively.

SI Materials and Methods

Materials.

Primers for gene expression studies were obtained from Biomers and for genotyping of humanized TNFR from Life Technologies. The TNFR2-specific antibody 80M2 (28) and the TNFR1-selective antagonistic antibody ATROSAB (21, 22) have been described. Antibodies against NF-κB p65, Akt, and Akt (Ser473) were from Cell Signaling Technology. The anti-TNF and anti-TNFR2 (MAB426) antibody was from R&D Systems. Antibodies against mouse (HP8002, HP8003) and human TNFRs (HM2020, HP9002 and HM2023, HP9003) were from Hycult Biotec. Antibodies against CD45R, CD68, and CD8 were from Miltenyi Biotech. Secondary antibodies coupled to Alexa Fluor 488 were from Life Technologies, and HRP-labeled antibodies were purchased by Jackson ImmunoResearch Laboratories. Primary antibodies used for immunohistochemistry were goat anti-ChAT IgG (Millipore), rabbit anti-mouse IBA-1 (WAKO Chemicals), and mouse anti-rat integrin αM (CD11b) monoclonal antibody (Chemicon International). Secondary antibodies were rabbit anti-goat IgG (Sigma) and horse anti-mouse IgG (Vector; Brunschwig Chemie), respectively. Stainings were developed by use of the Vectastain Elite ABC Kit (Vector Laboratories) followed by incubation with 3,3′-diaminobenzidine (DAB; Sigma). Actinomycin D and 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) were from Sigma-Aldrich. All other chemicals were of analytical grade.

Production and Purification of TNF Muteins.

HEK293-6E cells (51), grown in F17 medium (Life Technologies), were transiently transfected with expression constructs of TNF muteins using polyethyleneimine (Sigma). The day after, Tryptone N1 (Organotechnie; TekniScience) was added to the cell culture, and cells were cultivated for additional 4 d. Then, supernatant was collected and recombinant proteins were purified by immobilized metal ion chromatography (IMAC). For this purpose, supernatant was loaded onto a column containing Ni-NTA agarose (Macherey-Nagel) and unbound proteins were washed away using IMAC wash buffer (50 mM sodium phosphate buffer). Bound proteins were eluted with IMAC elution buffer (50 mM sodium phosphate buffer, 250 mM imidazole) and dialyzed (cutoff 4–6 kDa; Roth) against PBS overnight at 4 °C. Finally, eluted proteins were purified by SEC. Protein concentration was determined by measuring the absorbance at 280 nm.

SDS/PAGE and Coomassie Staining.

A total of 2 µg of the purified TNF muteins were denatured in Laemmli buffer and resolved by 8% (mass/mass) SDS/PAGE (100 V; 90 min). For Coomassie staining of proteins, the gel was incubated in staining solution for 60 min at room temperature (RT) and destained.

Immunoblot Primary Neurons.

Cells were lysed in homogenization buffer [10 mM Hepes (pH 7.5), 1.5 mM MgCl₂, 1.5 mM KCl, 1% Nonidet P-40, 0.2 mM PMSF, 20 mM β-glycerophosphate, and 100 μM Na₃VO₄] at 4 °C for 30 min. Lysates were centrifuged (2 min at 9,600 × g) and protein concentration of supernatants were determined using the BCA method (Pierce). Twenty micrograms total protein were denatured in Laemmli buffer and resolved by 12% (mass/mass) SDS/PAGE (100 V; 90 min). Then, proteins were transferred onto nitrocellulose membranes (semidry blot; 1.5 mA/cm² gel for 90 min) and nonspecific protein binding was blocked with 5% (mass/mass) skim milk powder solution in PBS/0.1% Tween 20 for 30 min at RT; the membrane was incubated overnight at 4 °C using specific antibodies. After incubation with HRP-conjugated secondary antibodies for 90 min at RT, the signals were detected by enhanced chemiluminescence (Super Signal; Pierce).

Immunoblot Whole-Brain Samples.

The whole-brain protein samples were prepared as described previously (15). Protein concentration from brain tissues were quantified and adjusted to 1 mg/mL. Samples were incubated on ice for 20 min, 25 μL of 0.1% Nonidet P-40 was added for 2 min, and the lysates were centrifuged at 8,000 × g for 10 min at 4 °C. The lysates were boiled for 5 min in Laemmli’s sample buffer [2% (mass/mass) SDS, 5% (mass/mass) DTT], and proteins was separated by SDS/PAGE. After transfer to PVDF membranes (Millipore), membranes were blocked for 1 h with 1% I-blocker (Tropix) in TBS containing 0.0625% Tween 20 and subsequently incubated overnight with primary antibody at 4 °C [anti-human TNFR1 was purchased from Hycult Biotech (H398); anti-human TNFR2 was from Abcam (MR2-1)]. Afterward, the membranes were washed with TBS containing 0.0625% Tween 20 and incubated with the appropriate HPR-conjugated secondary antibody diluted to 1:5,000 with TBS containing 0.0625% Tween 20 for 1 h. Proteins were detected using enhanced chemiluminescence (Pierce Biotechnology). Hypoxanthine guanine phosphoryltransferase served as internal standard protein.

HPLC.

The oligomerization state of the TNF variants under native conditions was analyzed by HPLC-SEC. Approximately 20 µg protein was applied to a BioSep-SEC-s2000 column (Phenomenex) equilibrated with PBS and eluted at a flow rate of 0.5 mL/min. For determining the size of recombinant proteins, standard proteins were run under the same conditions.

Binding Studies with TNF Muteins.

ELISA plates (Greiner) were coated with huTNFR1–Fc or huTNFR2–Fc (Etanercept) fusion proteins at 1 µg/mL in PBS and incubated at 4 °C overnight. Residual binding sites were blocked with 2% (mass/mass) skim milk powder in PBS at RT for 2 h. TNF muteins were diluted in 2% (mass/mass) skim milk powder in PBS and incubated for 1 h at RT. Bound proteins were detected with mouse monoclonal antibodies to TNF (clone F6C5; 1 µg/mL; incubation for 1 h at RT) and HRP-conjugated anti-mouse IgG antibodies (diluted 1:10,000; incubation for 1 h at RT), followed by incubation with 3,3′,5,5′-Tetramethylbenzidine (TMB) substrate solution. Reaction was stopped by addition of 1 M H₂SO₄, and the absorbance at 450 nm was determined with an absorbance reader (Multiskan FC; Thermo Scientific). Data were analyzed using the software Microsoft Excel and GraphPad Prism 4 (GraphPad). Between each step, nonbound proteins were removed by washing four times with 0.005% Tween-20 in PBS.

Crystal Violet Staining.

L929 or Kym-1 cells (1.5 × 10⁴ cells/well) were grown in 96-well flat-bottom cell culture plates overnight. L929 cells were treated with actinomycin D (1 µg/mL) for 30 min and Kym-1 cells with or without 80M2 (1 μg/mL) for 30 min before addition of TNF variants. Then cells were incubated with different concentrations of TNF muteins for 24 h at 37 °C. Cells were washed with PBS and incubated with crystal violet [20% (vol/vol) methanol; 0.5% crystal violet] for 20 min to stain viable cells. The dye was washed away under rinsing water and cells were air-dried. Crystal violet was resolved with methanol and the optical density at 550 nm was determined. Each sample was analyzed in triplicates and data were analyzed using the software Microsoft Excel and GraphPad Prism 4 (GraphPad).

MTT Assay.

Cells were incubated with MTT (0.5 mg/mL) for 2 h at 37 °C. Then lysis buffer [10% (mass/mass) SDS, 20 nM HCl] was added, cells were lyzed overnight, and optical density at 550 nm was determined. Each sample was analyzed in triplicates, and data were analyzed using the software Microsoft Excel and GraphPad Prism 4.

Immunofluorescence.

Primary fibroblasts from human/mouse tumor necrosis factor receptor one knock in (hu/mTNFR1-k/i) mice were cultivated on poly-d-lysine (10 µg/mL)-coated coverslips. After incubation with TNF derivatives, the cells were fixed with PBS/4% (mass/mass) paraformaldehyde (PFA) for 30 min on ice at point in time 0. Then cells were permeabilized with PBS/0.1% Triton X-100 for 10 min at RT. Unspecific binding sites were blocked with PBS/4% (mass/mass) BSA for 30 min and cells were subsequently incubated with primary antibodies against NF-κB p65 for 60 min followed by the incubation with appropriate fluorescence-labeled secondary antibodies for 45 min in PBS/2% (mass/mass) BSA. After staining the nuclei with DAPI, the cells were mounted with Fluoromount-G (Southern Biotech). The labeled cells were analyzed by wide-field fluorescence microscopy (CellObserver with Colibri LED modules; Carl Zeiss). Isotype or secondary antibody control stainings were used to determine background fluorescence and to adjust the intensity of specific stainings.

To quantify the nuclear translocation of NF-κB p65, mean fluorescence intensities of nuclei and cytoplasm within a cell were determined automatically using Cell Profiler (52). Ratios of nuclear to cytoplasmic fluorescence were calculated from at least 500 cells for every condition.

ELISA.

Primary embryonic fibroblasts were stimulated as indicated, and supernatants were collected at the indicated time intervals and analyzed by an ELISA specific for IL-6 according to the instructions of the manufacturer (BioLegend). The absorbance at 450 nm was determined, and the amount of released IL-6 was determined with the provided standard and calculated using the software GraphPad Prism 4.

RNA Isolation, cDNA Synthesis, and Quantitative Real-Time PCR.

Total RNA was extracted using the peqGOLD Total RNA Kit (Peqlab) and then transcribed into cDNA with the Affinity Script Multiple Temperature cDNA Synthesis Kit and oligo(dT) primers (Agilent). The obtained cDNA was used to determine the gene expression by quantitative real-time PCR (qPCR; CFX96; BioRad) using specific primers for CXCL2, IL-6, and GAPDH (Table S1) and the KAPA SYBR FAST qPCR Mastermix (Peqlab). To determine the expression of distinct genes, ΔΔCt values were determined by correlating the obtained values to the housekeeping gene GAPDH. Data are presented relative to the unstimulated controls as normalized fold expression.

Table S1.

Oligonucleotide sequences used for PCR

Gene	Forward	Reverse
CXCL2	5′-AGTGAACTGCGCTGTCAATG-3′	5′-TTCAGGGTCAAGGCAAACTT-3′
IL-6	5′-CCGGAGAGGAGACTTCACAG-3′	5′-CTCTGAAGGACTCTGGCTTTG-3′
GAPDH	5′-GTGGCAAAGTGGAGATTGTTGCC-3′	5′-GATGATGACCCGTTTGGCTCC-3′
huTNFR1	5′-TGTCCCACCAAAACACACAC-3′	5′-CTGGCTGTTGTCCCTAGCAT-3′
mTNFR1	5′-CGGCTTCTTTTGCTTGTTTC-3′	5′-ACCTTTCCGACATGTCTTGC-3′
huTNFR2	5′-CTGGACTTTGTGGGGACAGT-3′	5′-GACAGCTGGAAGCCAAAGAG-3′
mTNFR2	5′-AAGGACCAGAGGTCTCAGCA-3′	5′-GCAGGAACAGAGGAGACGAG-3′

Humanized TNFR Knock-In Mice.

The humanized Tnfrsf1a and Tnfrsf1b knock-in mouse lines (hu/mTNFR1-k/i and hu/mTNFR2-k/i) were generated as contracted by Ozgene Pty Ltd. For targeting TNFR1, the human sequence from chromosome 12 position 6330763–6332522 was inserted in place of mouse chromosome 6 positions 125306866–125310626. For TNFR2, the human sequence from chromosome 1 position 12188610–12193082 was inserted in place of mouse chromosome 4 positions 144814170–144819194. The targeting constructs were electroporated into C57BL/6 ES cell line, Bruce4 (53). Homologous recombinant ES cell clones were identified by Southern hybridization and injected into BALB/c-albino C57BL/6 blastocysts. Male chimeric mice were obtained and crossed to albino C57BL/6 females to establish heterozygous germline offspring on C57BL/6 background. The germline mice were crossed to a ubiquitous Cre mouse line to remove the loxP-flanked selectable marker cassette.

To initially distinguish between hetero- and homozygous mice, the genotype of hu/mTNFR transgenic mice was tested by using mouse TNFR-specific forward primers (mTNFR1 OZ fwd and mTNFR2 OZ fwd; Table S2) in combination with mouse- or human-specific reverse primers (mTNFR1 OZ rev, huTNFR1 OZ rev, mTNFR2 OZ rev, and huTNFR2 OZ rev; Table S2). The mouse/mouse TNFR1-specific primers for wild-type TNFR1 lead to a product of 1,217 bp, whereas the mouse/human TNFR1-specific primers for the chimeric TNFR1 lead to a product of 702 bp. Similarly, the mouse/mouse TNFR2-specific primers for wild-type TNFR2 lead to a product of 1,169 bp, whereas the mouse/human TNFR2-specific primers for the chimeric TNFR2 lead to a product of 771 bp. After homozygous chimeric mice were bred for more than three generations, genotyping was carried out by PCR with primers specific for huTNFR1 (Table S1), leading to a 313bp product for the human TNFR1 allele, and mTNFR1 (Table S1), inducing a 270-bp fragment for the wild-type allele. Homozygous hu/mTNFR2-k/i animals were genotyped by PCR, using primers specific for huTNFR2 (Table S1) to produce a 300-bp fragment for human TNFR2 allele and using primers for mTNFR2 (Table S1), which results in 380-bp fragments for the wild-type allele.

Table S2.

Oligonucleotide sequences used for genotyping

Primer	Sequence
mTNFR1 OZ fwd	5′-CTAAACATTCCTTGACCGGC-3′
mTNFR1 OZ rev	5′-TTCCCACACAAATCTTGACG-3′
huTNFR1 OZ rev	5′-ATGCTAGGGACAACAGCCAG-3′
mTNFR2 OZ fwd	5′-GGTCCAAACCTTCTAAGCCC-3′
mTNFR2 OZ rev	5′-ACATCAATATAGGCCAGCCG-3′
huTNFR2 OZ rev	5′-GCGTAGGGTGTAAATGCCAC-3′

Male homozygous hu/mTNFR-k/i C57BL/6J mice (12 wk, 24–30 g) were used for experiments. All animals were individually housed with a free access to food and tap water and kept in an air-conditioned room (21 ± 2 °C) with a 12/12 h light/dark circle (lights on 7:00 AM).

Southern Blot Analysis ES Cell DNA.

ES cell DNA was isolated, digested with restriction enzymes, and analyzed on agarose gels. Specificity of digestion products was tested by Southern blotting: DNA was transferred to a nylon membrane and hybridized with the 5′ or 3′ external probe. Membranes were washed and the pattern of hybridization was visualized on a X-ray film by autoradiography.

Southern Blot Analysis Mouse Tissue.

Genomic DNA from lung, heart, liver, brain, thymus, kidney, and skin tissue of both homozygous hu/mTNFR1-k/i and hu/mTNFR2-k/i mice, respectively, was isolated, digested with BamHI, and blotted on nylon membrane and probed with the endogenous probes (enP) for TNFR1:

• 5′CACATCTACCTCTTCCTGACACTGCCTGATCTGTTGGTTTGGCTTCAGGTTCTCTGATGGGGTTGGAAGTACCACTGACCTTAGGTGCTCCAAGCATTTCTTCTTCGGGGAAAGGAACCACACTTTCATGATTGGGAAGTTCTTATCATAACTAACCCTTCCTGTCACCCTGGAAGCCTCTGTGTGTCGTGAGGGGTGGGGGTGCACCCCTGCCTGAGAGATTGCTGGTGTGCTTTCTGTGTGTGGCTTCTTGGGTCTATGGCTGAGGCAAGGGGCTTCCTGGCCCGTGCAGCTGCTGTGCCGAGGAGGTAGCACTTCTAGTAACAGCAGCTGACAGCAGGGTGCAAGCTGCCAGCCTCTTTCCAAACGGAGCTTTGGGGGTTGCAGAGCCCCCAAAGGCAGCTGTGAGTCTAGGTGTTAGGTCTCTCCTGAATGTGATCTGATTGGTCAGTTGCTTCTGCATCTGTCTTGAAGACCTCCGCTATCTTGACGTAACAACGCTATGCGAGGGGGGGAGGGGAGATGAAAGAAGGAAGGCTGAGGGAAACCCAAAACCGGGAGAGATCAAAGAGCAAGGTCTACATTGAAACTAAGAGGGCTTTGCGCTCTCCATACTGAAGCTGTTTACATACAAGAAAGCCAGTTGGTTGCTATGAGCTTTGCCAAGATTATAAAGCCAGCCCTCTAGAATGCCCTGCTTTGGGACTCCTGGCACATCCTCAGCTCCTGAATTGAGGGTCATCTTGAATCAGATCACCAATCTCTGGTCAACTCTAGTACATGACACATTTGTCTTTGGTCTTACTTAATCCTTGTCGCGCCTGCATGCTCGGAGTTCTTCCTTTTGTTTCTGTCCCGCAATG3′

and for TNFR2:

• 5′TCATCTGGGGTTACTGGGTGAGCATGAAGACAAGGCAGGGGGCTGGCTGTATTGGGTTGCTTATCCAGGGCACGGTGTGCACAGTCGCCTGAAGAGGGAAGCACTCTGTGTGGACTTTTGGCCAAAGGCGCAAGAGGCAGCTAGGGTACTGTGGCAGGTCCAGAAGAGCCAAAGACTGTCCACAGCATAGAACCCCAGTCCCTTGAGTCATGGCTTCTGGTGGCTTTCAGCTGTTATCACTGAGTGGCTGAGCAAAGAGAAACATGATAGGCTCAGAGGTCACTCAGTGCACAGCACTAGGCAAGGTAGGGTGATGCTCTCCAGACACAGCTGGGGATGGGAAGGGGAAAGACAGAAAGCAGAGCCACTGCCCACAGATCATGGGAGTTTGTGCAGTCTATACAGACCAAATGAGCCTAATAGTAAGGTCGGCATCTTCTCTGATTGGCTTGTTTGCAGTGGCTTTTCAGAGACTGATCCATGCCTTTGCTGTGCTAATTGTTGGATATTTTATGTATCTCTCTTGCTCCCATAGAATTGAGTCAGGAACCGATGTCCAGAGAGATAGGGACACTTGATTGACTATGACCATACAGCTTCTGGATGGAAGAATAAGAAAAGAAGACCACATCTCTCTAACTCTCCACCTTTATCTGAGGAGTCAGGGGATTCATTTGGGTTACTGGTTTAATAAACATTTATTAAGCTCCTGCTGACCATATGACGGCAGCTGCTGGTACAATCAAGAGTGCGGAGCAGCCTTCCTCTTTAACCTAGAACCACCCCTTGGTCATTTCCAGGTGACATCAGCAGAGGCTGCAAGGGTGTTGAGTGTGTGTTCACACCCTCTTTGCTCTTGTAG3′.

DNA from tail biopsies of C57BL/6 wild-type mice was used as a control.

Primary Mouse Embryonic Fibroblasts.

Brain and dark red organs were removed from embryonic (E14–16) hu/mTNFR1-k/i or hu/mTNFR2-k/i mice. Remaining tissue was minced and digested using trypsin-EDTA (Gibco) and 1% DNase I (Sigma) for 30 min at 37 °C. Digestion was stopped by addition of FCS, and homogenates were centrifuged (5 min, 200 × g). Digested tissue was triturated in culture medium [DMEM, 10% (vol/vol) FCS, l-glutamine, 1% (vol/vol) penicillin/streptomycin (P/S)] and plated in cell culture dishes. After 1–2 d, cells were frozen or used for experiments.

Isolation of Primary Thymocytes.

Thymus of hu/mTNFR2-k/i mice was isolated and mashed through a 40-µm cell strainer (Flacon). Cells were centrifuged (300 × g, 5 min) and washed once with culture medium [RPMI 1640, 10% (vol/vol) FCS, 50 µM β-mercaptoethanol, P/S]. Then, 1.5 × 10⁵ cells were plated into anti-CD3 (clone 17A2; BioLegend)-coated (6 h at 4 °C, 1 µg/mL) 96-well (U) plates and cultivated for 4 d in the presence of EHD2–scTNF_R2. Amount of cells was then determined by MTT assay.

Isolation of Thymocytes and Splenocytes for Flow Cytometry.

Thymus and spleen of hu/mTNFR1-k/i or hu/mTNFR2-k/i mice were isolated and mashed through a 40-µm cell strainer (Flacon). Cells were centrifuged (300 × g, 5 min) and washed once with PBA (PBS, 0.5% BSA, 0.02% NaN₃). Then, 0.5 × 10⁶ cells were incubated with antibodies against mouse TNFR1 (HP8002; Hycult Biotec), human TNFR1 (HP9002; Hycult Biotec), mouse TNFR2 (HP8003; Hycult Biotec), and human TNFR2 (HP9003; Hycult Biotec) for 45 min on ice. Cells were washed twice with PBA and incubated with fluorescence-labeled secondary antibodies for 30 min on ice. Then cells were washed again twice with PBA, and fluorescence was analyzed by flow cytometry (MACSQuant Analyzer 10; Miltenyi). Normalized data are presented as unit area vs. fluorescence intensity.

Primary Cortical Neuron Culture.

Primary cortical neurons were prepared from embryonic brains (E14–16) of hu/mTNFR1-k/i mice. The meninges were removed and the cortical neurons were separated by mechanical dissociation. Cells were plated in a density of 12 × 10⁴ cells/well (96-well plates) and 2 × 10⁶ cells/well (six-well plates) on poly-d-lysine precoated plates. Neurobasal medium (Gibco) with 2% (vol/vol) B27-supplement, 0.5 mM glutamine, P/S was used as culture medium. After 48 h, neurons were treated with 10 µM cytosine arabinoside (Sigma) for another 66 h to inhibit nonneuronal cell growth. Subsequently, the medium was completely exchanged and after 7 d of in vitro culture, the neuronal cell cultures (<1% contaminating glial cells) were used for experiments.

TNF Tolerance in Vivo.

Transgenic hu/mTNFR1-k/i mice were injected (i.v.) with recombinant mouse TNF (30 µg) or saline. Body weight and temperature were controlled regularly as indicator for TNF-induced toxicity.

Nucleus Basalis Magnocellularis Lesion Model.

Injection into the NBM was performed as previously described by Luiten et al. (54). Briefly, after anesthesia with isoflurane, mouse heads were fixated in a Kopf stereotactic frame (Kopf Instruments model 900). A volume of 0.6 µL solution was slowly infused in two steps of 0.3 µL into the NBM of the right hemisphere (coordinates: 0.6 mm posterior to bregma; 2.1 mm lateral to the sagittal suture; 4.6/4.4 mm ventral to the dura) (55). After the injection (0.1 μL/min), the needle of a 1-μL Hamilton syringe was left in place for 2 min to allow for an effective drug diffusion. The right side of the NBM and the contralateral intact hemisphere served as internal control without interference. The mice were killed for analysis 8 d after the NBM injection.

Hu/mTNFR2-k/i mice were divided into four groups: (i) PBS (0.01 M, pH 7.40); (ii) PBS with 580 ng EHD2–scTNF_R2; (iii) 55 nmol NMDA (Sigma); and (iv) 55 nmol NMDA with 580 ng EHD2–scTNF_R2.

Hu/mTNFR1-k/i mice were divided into seven groups: (i) PBS; (ii) PBS with 3 µg ATROSAB; (iii) 55 nmol NMDA; (iv) 55 nmol NMDA together with 3 μg ATROSAB; (v) 55 nmol NMDA, 3µg ATROSAB, and 150 ng mouse TNFR2 antagonist (MAB426); (vi) 55 nmol NMDA together with 150 ng MAB426 only; and (vii) 55 nmol NMDA and 2 µg anti-human epidermal growth factor receptor (huEGFR) IgG as an unspecific control antibody.

Tissue Processing and Immunohistochemistry.

Under a deep anesthesia via an i.p. injection of pentobarbital sodium, mice were transcardially perfused by a short prerinse with 0.5% heparin physiological saline and subsequently perfused and fixated with 4% (mass/mass) PFA in phosphate buffer (PBS, 0.1 M, pH 7.40). Brain tissues were removed, postfixed 48 h with 4% (mass/mass) PFA in PBS, subsequently stored in PBS at 4 °C for 24 h, and cryoprotected by immersion in 30% (mass/mass) sucrose at room temperature for 24 h. Afterward, frozen coronal sections (20-µm thickness) were cut using a Leica cryostat microtome and collected in PBS (0.01 M, pH 7.40) containing 0.1% sodium azide.

Cholinergic Fiber Stainings.

After rinsing several times in PBS, free-floating tissue sections were processed for ChAT immunohistochemistry by incubation in PBS with 0.3% H₂O₂ for 45 min, rinsed several times again, and preincubated in PBS containing 5% (vol/vol) normal rabbit serum (NRS; Zymed) and 0.4% Triton X-100 at RT for 1 h. Subsequently, sections were incubated with goat anti-ChAT IgG (diluted 1:333; Millipore) in PBS containing 1% NRS, 0.5% BSA, and 0.4% Triton X-100 at 4 °C for 3 d. After incubation with the primary antibody, sections were rinsed in PBS and incubated with rabbit anti-goat IgG (diluted 1:500; Sigma) in PBS containing 1% NRS, 0.2% Triton X-100, and 0.5% BSA at RT for 4 h and at 4 °C overnight. Sections were rinsed in PBS and incubated in PBS with Vectastain Elite ABC Kit (A and B solution were both diluted to 1:500) at RT for 2 h. Finally, DAB (Sigma) reaction to peroxidase as chromogen was visualized. Visualization of DAB reaction was enhanced by ammonium nickel sulfate (BDH Chemicals Ltd.).

Macrophage/Microglial Activation.

Tissue sections were incubated in 0.3% H₂O₂ in PBS for 45 min, rinsed in PBS several times, preincubated in 5% (vol/vol) normal horse serum (NHS) and 0.4% Triton X-100 PBS. Thereafter, sections were incubated with rat anti-mouse integrin αM monoclonal antibody (CD11b, diluted 1:1,000, Chemicon International) or anti-mouse IBA-1 (WAKO Chemicals, Neuss, Germany) in PBS containing 1% NHS and 0.4% Triton X-100 at 4 °C for 3 d. Thereafter, sections were incubated in biotinylated horse anti-mouse IgG (Vector; Brunschwig Chemie) diluted 1:500 at 4 °C overnight, followed by incubation of Vectastain Elite ABC Kit (diluted to 1:500) at RT for 2 h. Finally, peroxidase was visualized by DAB as chromogen. Visualization of the DAB reaction was enhanced by ammonium nickel sulfate.

Quantitative Analysis of Cholinergic Fibers.

ChAT-positive innervation density was measured in layer V of the somatosensory cortex by using a program of the LAS Image Analysis Software (Leica Quantimet). Surface area density of cortical ChAT-positive fibers was measured in eight coronal sections (coordinates: 0.6 mm posterior to bregma), which represent a strong reaction of cholinergic innervations from the lesioned NBM subdivision (56). After ground subtraction and gray-scale threshold determination, the surface area of ChAT-positive fibers was calculated as a percentage by (the area coverage of ChAT-active cholinergic fibers)/(the total sampling area) × 100 in each coronal section. The correlated value of fiber reduction caused by the unilateral NBM lesion was calculated as the percentage difference between the surface area density of cortical fibers in the damaged side and that in the contralateral side (57).

Measurement of Macrophage/Microglial Activation.

Quantification of the reactive volume of macrophage/microglial activation has been previously described (24). Briefly, the size of the reactive region of CD11b-positive macrophage/microglial activation was measured on the surface area of the lesion site by using a program of the Quantimet 600 system (Leica). The reactive surface area of microglial activation was quantified in a series of CD11b-positive sections of each mouse brain. The total reactive volume of microglial activation in a brain was calculated as (x1+ x2 + … xn) × 100 μm³, where xn is the reactive surface area of each section (µm²) and 100 μm was the distance between two consecutive sections.

Behavioral Studies.

Spontaneous alternation task.

Short-term spatial memory performance was evaluated by recording of spontaneous alternation behavior in a Y-maze paradigm as described previously (23). The test was performed 6 d after NBM injection in a Y-maze consisting out of three symmetrical orientated transparent Plexiglas tubular arms (5 cm in diameter, 27.5 cm length). Mice were placed in the center of the maze and allowed to freely explore for 8 min. The sequence of arm entries was recorded visually. Arm entry was considered complete when a mouse entered an arm with all four paws. A successful spontaneous alternation was defined as consecutive entries into all three different arms. Percentage alternation was calculated as the number of triads containing entries into all three arms divided by the maximum possible alternations (the total number of arms entered minus 2) × 100.

Elevated plus maze.

Anxiety-related exploration was assessed in an elevated plus maze (58). The apparatus consisted of two open arms (5.5 cm width × 30 cm length) and two enclosed arms (5.5 cm width × 30 cm length × 15 cm high), which were elevated 50 cm above the ground. After 6 d of NBM injection, mice were placed the center of the maze and allowed to freely explore for 8 min. Transitions were scored as the mice moved from one arm to the other. Arm entry was considered to be complete when the mouse had entered one arm with all four paws. The time spent in the open or closed arms, the number of arm entries, and total entries were recorded.

Passive avoidance paradigm.

Long-term associative memory was evaluated with the step-through passive avoidance paradigm (59). The apparatus consists of a light and a dark compartment separated by a retractable door. On day 7 after injection into the NBM, mice were first habituated by placing them into the illuminated platform facing the closed retractable door. After 60 s the retractable door was raised and mice were allowed to explore freely for 2 min. The training interval was performed after 24 h by placing the mice in the illuminated compartment facing to the closed door and allowed to explore the light compartment for 60 s. The door was raised, and the latency to enter the dark compartment (preshock latency) was recorded. After entering the dark compartment, the door was closed and mice received an electric shock (0.3 mA, 2 s). Following a 24-h interval, memory retention was determined by the latency (postshock latency) of the animals to enter the dark compartment, which was recorded for up to a maximum of 8 min.

Statistics.

Data are presented as mean ± SD or SEM of n independent experiments. Normal distribution was analyzed by Shapiro–Wilk normality test. Statistical analyses were performed by Student’s t test or one-way ANOVA, followed by a post hoc multiple-comparison Tukey test. *P < 0.05 (**P < 0.01; ***P < 0.001) was considered significant.