Microglial derived tumor necrosis factor-α drives Alzheimer's disease-related neuronal cell cycle events

Abstract

Massive neuronal loss is a key pathological hallmark of Alzheimer's disease (AD). However, the mechanisms are still unclear. Here we demonstrate that neuroinflammation, cell autonomous to microglia, is capable of inducing neuronal cell cycle events (CCEs), which are toxic for terminally differentiated neurons. First, oligomeric amyloid-beta peptide (AβO)-mediated microglial activation induced neuronal CCEs via the tumor-necrosis factor-α (TNFα) and the c-Jun Kinase (JNK) signaling pathway. Second, adoptive transfer of CD11b+ microglia from AD transgenic mice (R1.40) induced neuronal cyclin D1 expression via TNFα signaling pathway. Third, genetic deficiency of TNFα in R1.40 mice (R1.40-Tnfα(-/-)) failed to induce neuronal CCEs. Finally, the mitotically active neurons spatially co-exist with F4/80+ activated microglia in the human AD brain and that a portion of these neurons are apoptotic. Together our data suggest a cell-autonomous role of microglia, and identify TNFα as the responsible cytokine, in promoting neuronal CCEs in the pathogenesis of AD.

Keywords: AD; Adoptive transfer; Alzheimer's disease; AβF; AβO; BrdU; Bromodeoxyuridine; CCE; CM; COX2; GFP; IDV; IKKα/β; IL-1β; IL-6; INFγ; IκB kinase α/β; JNK; LPS; MAP2; Microglia; NOS; NSAID; Neuroinflammation; Neuronal cell cycle; PAS; PBS; PCNA; PFA; PI3K; Phosphatidylinositol 3-kinase; ROI; RT; STAT3; Signal transducer and activator of transcription 3; TNF receptor; TNFR; TNFα; TUNEL; Terminal deoxynucleotidyl transferase dUTP nick end labeling; Tumor Necrosis Factor-α; Tumor necrosis factor-α (TNFα); c-Jun Kinase; c-Jun Kinase (JNK); cell cycle events; conditioned media; cyclooxygenase-2; fibrillar amyloid-beta peptide; green fluorescent protein; integrated density value; interferon-γ; interleukin-1β; interleukin-6; lipopolysaccharide; microtubule associated protein-2; nitric oxide synthase; non-steroidal anti-inflammatory drug; oligomeric amyloid-beta peptide; p38 MAPK; p38 mitogen activated protein kinase; paraformaldehyde; phosphate buffered saline; proliferating cell nuclear antigen; protein-A-sepharose; region of interest; room temperature.

Fig. 1. Soluble factor(s) from AβO-activated microglia induce neuronal DNA replication

(A–D) Treatment of primary microglia with different concentrations AβO (0.2, 2 and 20 µg/ml), but not vehicle (Veh), for 24 h alters microglial morphology from resting (rod-like) to an activated (macrophage-like) phenotype. Scale bar 20 µm. (E) Schematic showing the gradation of microglial activation with increasing concentrations of AβO treatment. (F–H) Confocal image with orthogonal view (in H) showing AβO (4.0 µg/ml for 24 h)-activated, Iba1 positive microglia (green) containing AβO inclusions immunoreactive for A11 oligomer-specific antibody (red; merge appears yellow in H). (I–N) AβO-activated microglia (green) contains Aβ inclusions that are immunoreactive for an AβO-specific monoclonal antibody NU1 (red in J) or Aβ-specific antibody 6E10 (red in M). Merged images are shown in K and N, DAPI is shown in blue in K. In I–K and L–N, AβO concentration was 4 µg/ml and 0.2 µg/ml respectively. Scale bars 10 µm.

Fig. 2. Soluble factor(s) secreted from AβO-activated microglia induce neuronal DNA replication

(A) Schematic showing primary microglia derived from postnatal day 3 (P3) pups were incubated with 4 µg/ml AβO for 24 h, prior to collecting CM for neuronal treatment. Primary cortical neurons were incubated with 25% of the AβO-activated microglial CM containing 10 µM BrdU for 24 h prior to immunofluorescence analysis. (B–J) Double immunofluorescence revealing no DNA replication (BrdU incorporation) in the MAP2 + neurons treated with vehicle (Veh)-activated microglial CM (B, C, D). AβO-activated microglial CM induces neuronal DNA replication (E, F, G). Scale bar 30 µm. (H–K) Treatment of primary neurons with AβO-immunoprecipitated (IP’d) microglial CM showing numerous BrdU/MAP2 double positive neurons (H, I, J). In (K), AβO from microglial CM was IP’d utilizing 6E10 antibody prior to neuronal treatment. Lane 1: AβO-treated CM prior to IP; lane 2: vehicle treated microglial CM prior to IP; lane 3 and 4: AβO-treated CM after IP with 6E10; lane 5: Protein A Sepharose (PAS) beads containing 6E10-AβO complex after IP; lane 6: recombinant Aβ showing different AβOs and monomeric Aβ (AβM). (L) Quantification of double positive (BrdU/MAP2) neurons revealing a statistically significant, dose-dependent increase (vehicle versus 6.25% CM or 25% CM treatment with or without IP of AβO; n=3 independent cultures; run in triplicates/culture; ***p<0.001; two-way ANOVA followed by a Bonferroni post hoc test; mean ± SEM) in the percentage of BrdU + neurons following microglial CM treatment.

Fig. 3. TNFoc mediates microglia-mediated neuronal cell cycle events in vitro

(A) A representative image showing a four-plex mouse cytokine infrared (IR) array spotted with antibodies against IL-6, TNFα, IL-1β and INFγ (spots are identified in the schematic on the right) display increasing intensity in the IR signal when hybridized with increasing concentrations of a mixture of these four recombinant cytokines (top panel). A representative image of an IR array probed with microglial CM that are treated with either vehicle (Veh) or different concentrations of AβO. Note a concentration dependent increase IR signal for IL-6 and TNFα, but not for IL-1β or INFγ (bottom panel). (B) Quantification of IR signal intensities for each of the four cytokines in microglial CM reveal a dose-dependent increase in the levels of secreted TNFα, to less extent for IL-6, but not for IL-1β or INFγ following treatment with different concentrations of AβO. (C–N) Double immunofluorescence of 21DIV primary cortical neurons for MAP2 and BrdU reveal that 30 min pre-treatment of neurons with anti-TNFα antibody (F-H), but not mouse IgG (C–E), prior to 24 h of incubation with 25% of AβO-activated microglial CM (AβO-CM) reduced the number of double (BrdU and MAP2) positive neurons. Direct incubation of 21DIV primary cortical neurons with recombinant mouse TNFα (250 pg/ml; 24 h) induced the number of double (BrdU and MAP2) positive neurons (L-N) compared to those treated with mouse IgG alone (I–K). Scale bar 30 µm. (O–Q) Western blot analysis of 21 DIV primary cortical neurons revealing significantly elevated (*p<0.05 with unpaired t test; graphs shown in P and Q) levels of cyclin D1 and PCNA following rTNFα, but not vehicle, treatment. (R) Quantification of double (BrdU and MAP2) positive cells from different conditions described in C–N reveal statistically significant (mean ± SEM; *p<0.05 versus Veh – clear white bar) increase in double positive cells following AβO-CM treatment, a significant decrease (mean ± SEM; *p<0.05) following pre-incubation of neurons with anti-TNFα antibody, but not by mouse IgG (mean ± SEM; **p<0.01). Treatment with mouse IgG alone (without AβO-CM) is similar to vehicle treatment with no difference in the percentage of BrdU+ neurons. Direct treatment of neurons with recombinant 1 ng of TNFα (rTNFα) significantly (mean ± SEM; ***p<0.001 versus Veh or mouse IgG alone) increased the number of double positive neurons. For the cytokine array experiments; samples were ran in quadruplicates per treatment; all other data sets derived from triplicates for each treatment and from at least two independent neuron and microglial cultures. One-way ANOVA followed by a Newman-Keuls multiple comparison test were used.

Fig. 4. Purified microglia from R1.40 mice induces neuronal cell cycle events in vivo

(A–B) Schematic showing the magnetic-based isolation of CD11b+ microglia and intracerebral injections of purified microglia into recipient mouse brain. Microglia purified from R/R-Cx₃cr1^gfp/+ mouse brain are viable and show robust GFP expression in vitro prior to adoptive transfer (in B; Scale bar 10 µm). (C) Five serial sections (30 µm thick), two anterior (N+1 and N+2) and two posterior (N−1 and N−2) were analyzed for quantification and the 1–2 mm area marked in the hashed circle defined as ROI for quantification. (D–E) Representative images captured under low power (D) and high power (E) of a medial section showing GFP + microglia from R/R-Cx₃cr1^gfp/+ donor within recipient mouse brain, showing activated phenotype after 48 h (inset in E). Scale bars (30 µm in E; 10 µm for inset in E). (F–Q) 6-month-old R/R microglia, but not WT, was capable of inducing cyclin D1 expression (red in G, J, M, P) with NeuN+ neurons (green in F, I, L, O) in the ipsilateral, but not contralateral, cortex of 6-month-old WT recipient mouse brain. Recipient mice receiving vehicle (RPMI) did not exhibit neuronal cyclin D1 expression (F-H). Merged images are in H, K, N and Q. Scale bar 30 µm. (R) Quantification of cyclin D1 and NeuN + neurons revealed a statistically significant (n=3 recipients per treatment; five sections per mouse; six random fields per section were scored; mean ± SEM; *p<0.05; one-way ANOVA with a Tukey post hoc test) increase in the percentage of double + (cyclin D1 and NeuN) neurons in the ipsilateral cortex of recipient mouse brain that received microglia from 6 month old R/R donors compared to other recipients receiving either microglia from WT or vehicle.

Fig. 5. Microglia-derived TNFα induces neuronal cyclin D1 expression in vivo

(A–F) Double immunofluorescence for NeuN (green), cyclin D1 (red) show reduced number of double (cyclin D1 and NeuN) positive cells (arrows) in the layer VI of the cortex when purified R/R microglia was transferred with anti-TNFα antibody, but not with mouse IgG, into two-month-old non-transgenic (WT) recipient mouse brain. (G) Percentage of cyclin D1 + neurons are significantly (****p<0.0001; unpaired t test; n=3 recipient mouse brains) reduced in the recipient mouse brain that received purified R/R microglia with anti-TNFα antibody than compared to those that received R/R microglia with mouse IgG. (H) Western blot analysis of detergent soluble cortical lysates revealed an increase in the levels of TNFα in the six-month-old R/R mouse brain compared to age-matched WT. (I) Quantification of Western blots for TNFα revealed a statistically significant increase (mean ± SEM; **p<0.01; n=4 per group; unpaired t test) in the integrated density value (IDV) ratio for TNFα/GAPDH in the six-month-old R/R compared to age-matched WT. (J–R) Layer II/III of six month old R/R showing significantly higher number of NeuN + neurons expressing cyclin D1 (arrows) compared to age-matched non-transgenics (WT) and R/R-Tnfα^−/− mice in the identical brain region. Scale bar 20 µm.. (S) Percentage of cyclin D + neurons are significantly (mean ± SEM; ***p<0.001; one-way ANOVA with Tukey post hoc test; n=4 animals per group) higher in layer II/III of the cortex in 6 month old R/R mice compared to age-matched WT or R/R-Tnfα^−/− mice.

Fig. 6. Activation of TNFα induces neuronal DNA replication and apoptosis

(A–B) Double immunofluorescence for MAP2 (green) and BrdU (red) reveal reduction in the double (MAP2 and BrdU) positive neurons following 30 min pre-incubation of 21DIV neurons with SP600125 (a JNK inhibitor), but not vehicle, prior to 24 h incubation with 250 pg/ml rTNFα. Scale bar 20 µm. (C) Quantification reveal statistically significant (mean ± SEM; ***p<0.001; unpaired t test; n=3 replicates in two independent cultures) decrease in the percentage of BrdU + neurons following SP600125 pretreatment. (D–K) Triple immunofluorescence of 21DIV primary cortical neurons for MAP2 (blue), BrdU (green), TUNEL (red) following 24 h treatment with 250 pg of rTNFα reveal a neuron positive for both BrdU and TUNEL Scale bar 10 µm. (L) Quantification of the triple positive (MAP2+BrdU+TUNEL) neurons reveal that about 20% (*p<0.05; unpaired t test; mean ± SEM; n=3) of MAP2 positive neurons are also positive for both BrdU and TUNEL in response to rTNFα

Fig. 7. Relationship between neuroinflammation, neuronal cell-cycle events and neurodegeneration in the R1.40 transgenic mouse model and in human AD brain

(A–B) Immunohistochemical analysis revealing the presence of CD45 immunoreactivity in layer II and III of the cortex of 6 month old R/R mouse brain (B) but not in age-matched non-transgenics (A). (C–D) CD45 immunoreactivity is significantly enhanced in 20-month-old R/R mouse brain (D) compared to age-matched non-transgenics (C). (E–H) Triple immunofluorescence labeling reveals cyclin D1 expression (red in F) within NeuN + neurons (purple in E; arrows in H) in the temporal cortex of a human AD brain where numerous F4/80 + (green in G) microglia/macrophage also co-exist (merged image in H). I–L) Triple immunofluorescence labeling reveals that a portion of NeuN + neurons (purple in I) expressing cyclin D1 (green in K) is also Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) positive (red in J; arrows in L), merged image in L). Scale bars 20 µm. (M–P) As an unaffected regional control, deeper layers of the temporal gyri show no cyclin D1 expression and are also devoid of TUNEL reactivity in NeuN positive neurons. Scale bar 20 µm. (Q) Schematic showing microglial activation by AβO results in the release of TNFα that binds to TNFR present on susceptible neurons. Interaction between TNFα and TNFR recruits several adaptor proteins including Act1 and TNF Receptor Associated Factor (TRAF) and induces activation of JNK directly or via Mitogen Activated Protein Kinase Kinase (MKK). Activation of JNK induces the expression of cyclin D1, and thus renders “at risk” neurons to acquire the identity of a mitotic cell.