Erythropoietin signaling in peripheral macrophages is required for systemic β-amyloid clearance

Abstract

Impaired clearance of beta-amyloid (Aβ) is a primary cause of sporadic Alzheimer's disease (AD). Aβ clearance in the periphery contributes to reducing brain Aβ levels and preventing Alzheimer's disease pathogenesis. We show here that erythropoietin (EPO) increases phagocytic activity, levels of Aβ-degrading enzymes, and Aβ clearance in peripheral macrophages via PPARγ. Erythropoietin is also shown to suppress Aβ-induced inflammatory responses. Deletion of EPO receptor in peripheral macrophages leads to increased peripheral and brain Aβ levels and exacerbates Alzheimer's-associated brain pathologies and behavioral deficits in AD-model mice. Moreover, erythropoietin signaling is impaired in peripheral macrophages of old AD-model mice. Exogenous erythropoietin normalizes impaired EPO signaling and dysregulated functions of peripheral macrophages in old AD-model mice, promotes systemic Aβ clearance, and alleviates disease progression. Erythropoietin treatment may represent a potential therapeutic approach for Alzheimer's disease.

Keywords: Alzheimer's disease; erythropoietin signaling; peripheral macrophages; systemic Aβ clearance.

Figure 1. Aβ₄₂ promotes its self‐clearance through macrophage EPO signaling

1. A, B

   Peritoneal macrophages (PMs) were treated with Aβ₄₂ (0.5–5 μM) for 12 h, phagocytosis of Aβ₄₂ (A) and Aβ₄₀ (B) was measured by flow cytometry (n = 4). Vehicle (cold) was macrophage treated with Aβ resolution incubated at 4°C. It was used as a negative control to exclude Aβ bound with macrophage from Aβ engulfment. Vehicle was macrophage treated with Aβ resolution incubated at 37°C as control.

2. C, D

   WT mice (8 weeks) were treated with Aβ₄₂ (0.2 μmol/kg/day, i.v., twice per day) for 24 h, splenic macrophages were purified for phagocytosis assay of Aβ₄₂ (C) and Aβ₄₀ (D) (n = 4).

3. E

   Splenic macrophages from young APP/PS1‐21 mice (6‐week‐old) were isolated for phagocytosis assay of Aβ₄₂ (n = 4).

4. F–H

   Macrophages from (A–E) were detected by immunoblotting with the indicated antibodies (n = 3).

5. I

   PMs from Lyz2‐Cre ^+/+ /Epor ^loxp/loxp mice (EPOR‐MKO) or Lyz2‐Cre ^+/+ /Epor ^+/+ mice (EPOR‐C) were treated with Aβ₄₂ (5 μM) or vehicle for 12 h. Phagocytosis of Aβ₄₂ was measured by flow cytometry (n = 4).

6. J

   EPOR‐MKO mice (8 weeks) were treated with Aβ₄₂ (0.2 μmol/kg/day, i.v.) for 24 h, splenic macrophages were used for phagocytosis assay of Aβ₄₂ (n = 4).

Data information: For multiple groups, one‐way ANOVA was used. Unpaired t‐tests were used for the comparison of two groups. Results are presented as mean ± s.d., *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; ns denotes no statistical significance. Aβ, amyloid‐β; EPO, erythropoietin. EPOR‐KO macrophages were PMs isolated from EPOR‐MKO mice. Source data are available online for this figure.

Figure EV1. Aβ42 promotes itself phagocytosis by macrophages

1. A

   A series of figures show the gating strategy of flow cytometry analysis used to identify PMs that engulfed HiLyte‐488‐labeled Aβ₄₂ in vitro.

2. B, C

   PMs were treated with Aβ₄₂ of indicated concentration for 12 h, and then incubated with HiLyte‐488‐labeled Aβ₄₂ (B) or HiLyte‐555‐labeled Aβ₄₂ (C) for 1 h after being washed for three times. Phagocytosis of Aβ₄₂ was detected by confocal microscopy (n = 6).

3. D

   A series of representative images from a live‐cell video microscopy experiment were showed. Phagocytosis of fluorescently labeled Aβ₄₂ by primary macrophage was depicted using phase contrast images overlaid with fluorescence microscopy images. Yellow arrowheads indicate Aβ₄₂ (n = 4).

4. E

   PMs were treated with Aβ₄₂ (5 μM) for 12 h, then with LA (200 nM) (phagocytosis inhibitor) or Dynasore (0.1 μM) (endocytosis inhibitor) for another hour and were applied for Aβ phagocytosis assay by flow cytometry (n = 4).

5. F

   Number and viability of macrophages from Fig 1A were analyzed by CCK8 assay (n = 7).

6. G, H

   PMs were treated with Aβ₄₀ (5 μM) for 12 h and then phagocytosis of HiLyte‐Aβ₄₂ (G) or HiLyte‐Aβ₄₀ (H) was measured by flow cytometry (n = 4).

7. I

   Bone marrow‐derived macrophages (BMDMs) were treated with Aβ₄₂ (0.5–5 μM) for 12 h, and applied for HiLyte‐Aβ₄₂ phagocytosis assay (n = 4).

8. J

   Gating of splenic macrophages was set, and macrophages were confirmed as F4/80⁺CD11b⁺.

9. K, L

   Splenic macrophages from WT mice (8 weeks) treated with Aβ₄₀ (0.2 μmol/kg/day, i.v. for 1 day) were purified for phagocytosis assays of Aβ₄₂ (K) and Aβ₄₀ (L) (n = 4).

10. M, N

    Spleen levels of Aβ₄₂ were measured by ELISA for following groups: WT mice (8 weeks) treated with Aβ₄₀ (0.2 μmol/kg/day, i.v.) or vehicle for 24 h (M) (n = 4); APP/PS1‐21 mice or WT mice of 6‐week‐old age (N) (n = 7).

11. O

    Co‐immunoprecipitation of EPOR and Aβ shows no direct binding of Aβ to EPOR. IgG was used as negative control for this assay (n = 4).

Data information: For multiple groups, one‐way ANOVA was used. Unpaired t‐tests were used for the comparison between two groups. Results are presented as mean ± s.d., *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; ns denotes no statistical significance; LA, Latrunculin A. Source data are available online for this figure.

Figure 2. EPO promotes macrophage clearance of Aβ and suppresses Aβ‐induced inflammatory response

1. A

   PMs treated with EPO of indicated concentration were measured by flow cytometry (n = 4). Vehicle (cold) was macrophage treated with PBS incubated at 4°C. It was used as a negative control to exclude Aβ bound with macrophage from Aβ engulfment. Vehicle was macrophage treated with PBS incubated at 37°C as control.

2. B

   PMs from Lyz2‐Cre ^+/+ /Epor ^loxp/loxp mice (EPOR‐MKO) or Lyz2‐Cre ^+/+ /Epor ^+/+ mice (EPOR‐C) were treated with PBS (Vehicle) or rhEPO for 12 h and then applied for Aβ phagocytosis assay.

3. C–F

   F4/80⁺CD11b⁺ splenic macrophages from PBS (Vehicle) or rhEPO (10,000 IU/kg/day, i.p., twice per day for 1 day) treated EPOR‐C mice (C), or EPOR‐MKO mice (D), or from rhEPO (5,000 IU/kg, i.p., every second day for 5 days) treated APP/PS1‐21 mice (5 months) (E), or from APP/PS1‐21+EPOR‐MKO mice (APP/PS‐1‐21 ^+/+ /Lyz2‐Cre ^+/+ /Epor ^loxp/loxp , 5 months) (F) were purified for Aβ₄₂ phagocytosis assay (n = 4).

4. G, H

   PMs from (A, B) were treated with PBS (Vehicle) or rhEPO for 12 h, then with Aβ₄₂ (5 μM) for another 12 h and applied for detection of inflammatory factor expression by QPCR (n = 4).

5. I–K

   Macrophages from (C–F) were applied for detection of inflammatory factor expression by QPCR (n = 4). Vehicle (H, I) is Aβ resolution and Vehicle (J) is PBS.

Data information: For multiple groups, one‐way ANOVA was used. Unpaired t‐tests were used for the comparison of two groups. Results are presented as mean ± s.d., *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; ns denotes no statistical significance. QPCR, quantitative polymerase chain reaction. EPOR‐KO macrophages were PMs isolated from EPOR‐MKO mice.

Figure EV2. Activation of macrophage EPO signaling promotes Aβ phagocytosis and degradation

1. A

   PMs were treated with rhEPO (10–100 IU) for 12 h, and applied for Aβ₄₀ phagocytosis assay (n = 4).

2. B

   Number and viability of macrophages from (A) were analyzed by CCK8 assay (n = 7).

3. C

   BMDM was treated with rhEPO (10–100 IU) for 12 h and applied for Aβ₄₂ phagocytosis assay (n = 4).

4. D–F

   The mRNA levels of IDE and NEP of indicated macrophages from Fig 2H–K were measured by QPCR (n = 4).

5. G

   Supernatants of macrophages from Fig 2G were collected. Levels of NO were normalized to supernatants of cell cultures treated with vehicle. Levels of inflammatory factors were measured by ELISA (n = 4).

6. H

   Macrophage lysate from Fig 1F were detected by immunoblotting with the antibodies against PPARγ or SRA (n = 3).

7. I

   Cell lysates from Fig 1G were detected by immunoblotting with the antibodies against PPARγ or SRA (n = 3).

8. J

   Cell lysates from Fig 1H were detected by immunoblotting with the antibodies against PPARγ or SRA (n = 3).

Data information: For multiple groups, one‐way ANOVA was used. Unpaired t‐tests were used for the comparison between two groups. Results are presented as mean ± s.d., *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; ns denotes no statistical significance. Source data are available online for this figure.

Figure 3. EPO promotes Aβ clearance via PPARγ in macrophages

1. A–F

   Macrophages from Fig 2A–F were detected by immunoblotting with indicated antibodies (n = 3).

2. G

   PMs from Lyz2‐Cre ^+/+ /Epor ^+/+ mice (EPOR‐C) were pretreated with PPARγ antagonist GW9662 (10 μM, 12 h), then with rhEPO (100 IU, 12 h), and then applied for phagocytosis assay of Aβ₄₂ (n = 3).

3. H

   PMs from Lyz2‐Cre ^+/+ /Epor ^loxp/loxp mice (EPOR‐MKO) or from EPOR‐C mice were pretreated with PPARγ agonist RSG (50 μM, 12 h), and then applied for phagocytosis of Aβ₄₂ by flow cytometry (n = 4).

4. I–L

   Splenic macrophages from EPOR‐C mice treated with GW9662 (1 mg/kg/day, i.p.) or/and rhEPO (10,000 IU/kg/day, i.p.) twice per day for 1 day (I), from EPOR‐MKO mice treated with RSG (0.1 mg/kg/day, i.p.) for 24 h (J), from APP/PS1‐21 mice treated with GW9662 (1 mg/kg, i.p., every second day) and/or EPO (10,000 IU/kg, i.p., every second day) for 5 days (K), or from APP/PS1‐21 ^+/+ /Lyz2‐Cre ^+/+ /Epor ^loxp/loxp mice (APP/PS1‐21+EPOR‐MKO) treated with RSG (0.1 mg/kg, i.p., every second day) for 5 day (L), were purified for Aβ₄₂ phagocytosis assay (n = 4).

5. M–O

   Macrophages from (G–L) were applied for detection of inflammatory factor expression by QPCR (n = 4).

Data information: All the vehicles of Fig 3 were PBS. For multiple groups, one‐way ANOVA was used. Unpaired t‐tests were used for the comparison of two groups. Results are presented as mean ± s.d., *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; ns denotes no statistical significance. GW, GW9662; RSG, Rosiglitazone. EPOR‐KO macrophages were PMs isolated from EPOR‐MKO mice. Source data are available online for this figure.

Figure 4. Peripheral macrophage EPO signaling regulates peripheral Aβ accumulation, brain Aβ efflux and central Aβ deposition

1. A–F

   Flag‐Aβ42 was injected into the lateral ventricles of Lyz2‐Cre ^+/+ /Epor ^loxp/loxp mice (EPOR‐MKO) or Lyz2‐Cre ^+/+ /Epor ^+/+ mice (EPOR‐C) (8 weeks) mice pretreated with rhEPO (10,000 IU/kg/day, i.p.) or ARA290 (0.03 mg/kg/day, i.p.) for 2 days (twice per day) (n = 3). One day later, Flag‐Aβ₄₂ in spleens and brains was detected by immunoblotting (A, B, D, E). Aβ₄₂ levels of spleen were measured by ELISA (C, F).

2. G–I

   Proteins were extracted from spleens and brains of APP/PS1‐21 mice (5 months) treated with rhEPO (10,000 IU/kg/day) or ARA290 (0.03 mg/kg/day) (i.p., once a day, for 14 days) (n = 10), and soluble Aβ levels were measured by ELISA (n = 5) (G, H). Aβ plaques in cortex and hippocampus (n = 5) were detected by IHC (I).

3. J–M

   Proteins were extracted from spleens and brains of 5‐ or 11‐month‐old APP/PS‐1‐21 ^+/+ /Lyz2‐Cre ^+/+ /Epor ^loxp/loxp mice (APP/PS1‐21+EPOR‐MKO) or APP/PS‐1‐21 ^+/+ /Lyz2‐Cre ^+/+ /Epor ^+/+ mice (APP/PS1‐21+EPOR‐C), and soluble Aβ levels were measured by ELISA (n = 5) (J, K). Aβ plaques in the cortex and hippocampus of mice of 5‐ (L) or 11‐month‐old (M) (n = 5) were detected by IHC.

4. N

   Timeline of bone marrow transplantation experiment.

5. O

   Splenic macrophages from APP/PS1‐21 mice transplanted with EPOR‐C or EPOR‐MKO bone marrow were purified for Aβ phagocytosis assay (n = 4).

6. P, Q

   Soluble Aβ levels of spleens and brains were measured by ELISA (n = 4).

7. R

   Aβ plaques in the cortex and hippocampus (n = 4) were detected by IHC.

Data information: For multiple groups, one‐way ANOVA was used. Unpaired t‐tests were used for the comparison of two groups. Results are presented as mean ± s.d., *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; ns denotes no statistical significance. ARA, ARA290; ELISA, enzyme linked immunosorbent assay; IHC, Immunohistochemical staining. Source data are available online for this figure.

Figure EV3. Peripheral macrophage EPO signaling regulates peripheral Aβ accumulation and brain Aβ efflux

1. Timeline of Flag‐Aβ₄₂ injection (i.c.v.) of mice treated with EPO or ARA290.

2. Expression of Iba1 and Aβ in cortex and hippocampus was detected by IF staining and statistically quantified (n = 5).

3. Blood was taken from APP/PS1‐21 mice of Fig 4G and plasma Aβ levels were measured by ELISA (n = 5).

4. Genotype of APP/PS‐1‐21^+/+/Lyz2‐Cre ^+/+/Epor ^loxp/loxp mice (APP/PS1‐21+EPOR‐MKO) was identified by genotyping assay. Mice with APP/PS1‐21 (yellow arrows), EPOR‐Flox (red arrows) and Cre (blue arrows) were APP/PS1‐21+EPOR‐MKO mice.

5. Splenic macrophages were isolated from mice of Fig 4L, and protein levels of EPOR were detected by immunoblotting (n = 3).

6. The expression of EPOR in splenic macrophages from mice of Fig 4L was measured by QPCR (n = 6).

7. The percentages of macrophage among splenic cells from mice of Fig 4L were measured by flow cytometry (n = 4).

8. Expression of EPOR or Iba1 in cortex and hippocampus from mice of Fig 4L was detected by IF staining (n = 5). The EPOR⁺Iba1⁺ microglia were pointed with white arrows.

9. Blood was taken from APP/PS1‐21 mice of Fig 4J and plasma Aβ levels were measured by ELISA (n = 5).

10. Macrophages from Fig 4N were detected by immunoblotting with indicated antibodies (n = 3).

11. Expression of EPOR or Iba1 in cortex and hippocampus taken from mice of Fig 4N was detected by IF staining (n = 4).

12. Blood was taken from mice of Fig 4N and plasma Aβ levels were determined by ELISA (n = 5).

13. Spleens were taken from mice of Fig 4N and protein levels of Aβ were detected by immunoblotting with indicated antibody (n = 3).

Data information: For multiple groups, one‐way ANOVA was used. Unpaired t‐tests were used for the comparison between two groups. Results are presented as mean ± s.d., *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; ns denotes no statistical significance. Iba1, ionized calcium binding adapter molecule 1. Source data are available online for this figure.

Figure 5. Peripheral macrophage EPOR‐deficiency exacerbates brain inflammation and behavioral deficits in APP/PS1‐21 mice

1. A, B

   Cortex and hippocampus taken from Fig 4L and M (n = 5) were applied for detection of inflammatory factor expression.

2. C–E

   Non‐cognitive behavioral impairments of APP/PS1‐21 mice (5 months) in Fig 4L were analyzed by nest construction assay and social interaction assay (C) (n = 10). Both non‐cognitive and cognitive behavioral impairments of APP/PS1‐21 mice (11 months) in Fig 4M were analyzed by nest construction assay, social interaction assay, NORT, FCT and Morris water maze test (D, E) (n = 10).

3. F

   Cortex and hippocampus were taken from APP/PS1‐21 mice in Fig 4N (n = 6) and applied for detection of inflammatory factor expression.

4. G, H

   Behavioral impairments APP/PS1‐21 of mice in Fig 4N and WT‐BM^EPOR‐C mice (irradiated WT mice transplanted with EPOR‐C bone marrow cells) were analyzed by nest construction assay, social interaction assay, NORT, FCT and Morris water maze test (n = 8).

Data information: For multiple groups, one‐way ANOVA was used. Unpaired t‐tests were used for the comparison of two groups. Results are presented as mean ± s.d., *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; ns denotes no statistical significance. NORT, novel object recognition test; FCT, fear conditioning test.

Figure EV4. The deficiency of peripheral macrophage EPO signaling contributes to Aβ accumulation in the peripheral and induces more neuronal damages in hippocampus in old APP/PS1‐21 mice

1. A

   Hippocampi were taken from mice of Fig 5B. Expression of MAP‐2 (in CA1 regions) and the apoptotic neurons were detected by IF staining and statistically quantified, respectively (n = 5).

2. B

   Hippocampi were taken from mice of Fig 5F. Expression of indicated proteins were detected by IF staining and statistically quantified (n = 5).

3. C–E

   Blood (C) and spleens (D, E) were taken from mice of Fig 6F (n = 6), and concentrations of Aβ in plasma and spleens were measured by ELISA (n = 6) (C and D). Spleen Aβ levels of APP/PS1‐21 mice at different ages were also detected by immunoblotting (n = 3) (E).

4. F

   PMs were treated with Aβ₄₂ (0, 5 or 10 μM) for 12 h, and applied for detection of inflammatory factor levels by QPCR (n = 5).

Data information: For multiple groups, one‐way ANOVA was used. Unpaired t‐tests were used for the comparison between two groups. Results are presented as mean ± s.d., *P < 0.05, **P < 0.01 and ****P < 0.0001; ns denotes no statistical significance. MAP‐2, microtubule‐associated protein 2; CA, cornu ammonis; IF, Immunofluorescent; NeuN, neuronal nuclei. Source data are available online for this figure.

Figure 6. EPO and ARA290 restore EPO signaling and dysregulated functions of peripheral macrophages in old APP/PS1‐21 mice

1. A–E

   Splenic macrophages from APP/PS1‐21 (5‐ or 11‐month‐old) or matched WT mice were detected by immunoblotting with the indicated antibodies (n = 3) (A, B). Their phagocytosis of Aβ₄₂ (C, D) were measured by flow cytometry (n = 4). The mRNA levels of IDE and NEP were measured by QPCR (n = 4) (E).

2. F

   Splenic macrophages from APP/PS1‐21 mice at different ages were purified for detection of inflammatory factor expression (n = 6).

3. G–J

   Old APP/PS1‐21 mice (11 months) and WT mice at the same age were treated with rhEPO (10,000 IU/kg/day), ARA290 (0.03 mg/kg/day) or vehicle (i.p., once a day, for 14 days). Splenic macrophages were isolated for detection by immunoblotting with the indicated antibodies (n = 3) (G). Phagocytosis of Aβ₄₂ by the splenic macrophages was measured by flow cytometry (n = 4) (H). Protein levels of IDE and NEP were detected by immunoblotting with indicated antibodies (n = 3) (I). The mRNA levels of inflammatory factors were measured by QPCR (n = 5) (J).

4. K–M

   PMs were treated with Aβ₄₂ (0, 5, 7.5 or 10 μM) for 12 h, and cell lysates were detected by immunoblotting with the indicated antibodies (n = 3) (K). Phagocytosis of Aβ₄₂ and Aβ₄₀ was measured by flow cytometry (n = 4) (L). Protein levels of IDE and NEP were detected by immunoblotting with indicated antibodies (n = 3) (M).

Data information: All the vehicles of Fig 6 were PBS. For multiple groups, one‐way ANOVA was used. Unpaired t‐tests were used for the comparison of two groups. Results are presented as mean ± s.d., *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; ns denotes no statistical significance. IDE, insulin degrading enzyme; NEP, neprilysin. Source data are available online for this figure.

Figure 7. EPO and ARA290 ameliorate peripheral and central Aβ accumulation, brain AD‐type pathologies and behavioral deficits in old APP/PS1‐21 mice

1. A

   Timeline of rhEPO (10,000 IU/kg/day) or ARA290 (0.03 mg/kg/day) treatments of old APP/PS1‐21 mice (11 months) (n = 5).

2. B

   Plasma Aβ levels were measured by ELISA (n = 5).

3. C, D

   Soluble Aβ levels in spleens (C) and brains (D) were measured by ELISA (n = 5).

4. E

   Aβ plaques in the cortex and hippocampus (n = 5) were detected by IHC.

5. F

   Cortex and hippocampus were taken, and the mRNA levels of inflammatory factors were measured by QPCR (n = 5).

6. G

   Expression of MAP‐2 or the apoptotic neurons of hippocampus were detected by IF staining and then statistically quantified (n = 5).

7. H, I

   Both non‐cognitive and cognitive behavioral impairments were analyzed by nest construction assay, social interaction assay, NORT, FCT and Morris water maze test (n = 10).

Data information: All the vehicles of Fig 7 were PBS. For multiple groups, one‐way ANOVA was used. Unpaired t‐tests were used for the comparison of two groups. Results are presented as mean ± s.d., *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; ns denotes no statistical significance. IF, Immunofluorescent; MAP‐2, microtubule‐associated protein 2; NeuN, neuronal nuclei; TUNEL, terminal dUTP nick end labeling.

Figure EV5. Treatment with rhEPO protects neuronal damages while stimulates hematopoiesis, and shows no effect on Aβ production in old APP/PS1‐21 mice

1. Levels of indicated protein from mice of Fig 7G were detected by immunoblotting and statistically quantified (n = 3).

2. Expression of MAP‐2 or NeuN in cortex from mice of Fig 7G was detected by IF staining and statistically quantified (n = 5).

3. APP/PS1‐21 (11 months) and WT mice at the same age were treated with rhEPO (10,000 IU/kg/day, i.p.) or ARA290 (0.03 mg/kg/day, i.p.) (once daily for 14 days). Blood was taken and hematopoietic function were analyzed by RBC, Hb and HCT assays (n = 8).

Data information: For multiple groups, one‐way ANOVA was used. Unpaired t‐tests were used for the comparison between two groups. Results are presented as mean ± s.d., *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; ns denotes no statistical significance. RBC, red blood cell; Hb, hemoglobin; HCT, Hematocrit. Source data are available online for this figure.