PD-1 immune checkpoint blockade reduces pathology and improves ...

2 downloads 0 Views 987KB Size Report
Jan 18, 2016 - Lesokhin, A.M., Callahan, M.K., Postow, M.A. & Wolchok, J.D. Sci. Transl. Med. 7, ..... A derivation of MARS-seq22, developed for single-cell ...
B r i e f c o m m u n i c at i o n s

© 2016 Nature America, Inc. All rights reserved.

PD-1 immune checkpoint blockade reduces pathology and improves memory in mouse models of Alzheimer’s disease Kuti Baruch1, Aleksandra Deczkowska1, Neta Rosenzweig1, Afroditi Tsitsou-Kampeli1, Alaa Mohammad Sharif1, Orit Matcovitch-Natan1,2, Alexander Kertser1, Eyal David2, Ido Amit2 & Michal Schwartz1 Systemic immune suppression may curtail the ability to mount the protective, cell-mediated immune responses that are needed for brain repair. By using mouse models of Alzheimer’s disease (AD), we show that immune checkpoint blockade directed against the programmed death-1 (PD-1) pathway evokes an interferon (IFN)-g–dependent systemic immune response, which is followed by the recruitment of monocytederived macrophages to the brain. When induced in mice with established pathology, this immunological response leads to clearance of cerebral amyloid-b (Ab) plaques and improved cognitive performance. Repeated treatment sessions were required to maintain a long-lasting beneficial effect on disease pathology. These findings suggest that immune checkpoints may be targeted therapeutically in AD. Chronic neuroinflammation is common to nearly all neurodegenerative diseases, and it contributes to their pathophysiology1. Nevertheless, although anti-inflammatory and immunosuppressive therapies have demonstrated some efficacy in neurodegenerative disease models, these treatments have largely failed in the clinic2,3. In mouse models of AD, the trafficking of blood-borne myeloid cells (monocyte-derived macrophages) to the central nervous system (CNS) was shown to be neuroprotective. Yet, spontaneous recruitment of these cells seems to be insufficient4. By using the five familial AD mutations (5XFAD) mouse model of AD5, we recently showed that transient depletion of forkhead box P3 (FOXP3)+ regulatory T (Treg) cells induces an IFN-γ–associated systemic immune response and the activation of the brain’s choroid plexus6, which is a selective gateway for leukocyte trafficking to the CNS7,8. This response was followed by the accumulation of monocyte-derived macrophages and Treg cells at sites of CNS pathology and by Aβ plaque clearance and a reversal of cognitive decline6. We therefore suggested that in

chronic neurodegenerative conditions, systemic immunity should be boosted, rather than suppressed, to drive an immune-dependent cascade needed for brain repair4. Immune checkpoints are regulatory pathways for main­taining systemic immune homeostasis and tolerance9. Selective blockade of immune checkpoints, such as the PD-1 pathway, enhances anti-tumor immunity by mobilizing the immune system 10. The IFN-γ–dependent activity induced by PD-1 blockade in cancer immunotherapy11, in addition to our observations that leukocyte trafficking to the CNS for repair involves an IFN-γ–dependent response7,12, prompted us to explore the therapeutic potential of PD-1 immune checkpoint blockade in AD. 5XFAD mice aged 10 months—an age of advanced cerebral pathology—received two intraperitoneal (i.p.) injections (at 3-d intervals) of either a blocking antibody directed at PD-1 (anti–PD-1) or an IgG control, and were examined 7 d after the first injection. PD-1 blockade increased splenocyte frequencies of IFN-γ– producing CD4+ T cells (Supplementary Fig. 1a,b), and genomewide RNA-sequencing of the choroid plexus (Supplementary Table 1) revealed an expression profile associated with an IFN-γ– response (Fig. 1a and Supplementary Table 2). Real-time quantitative PCR (RT-qPCR) showed elevated IFN-γ (Ifng) mRNA levels at the choroid plexus (Fig. 1b). These findings pointed to a systemic IFN-γ immune response in 5XFAD mice following PD-1 blockade, particularly at the choroid plexus. We next examined whether the effect of PD-1 blockade on systemic immunity involves CNS recruitment of monocyte-derived macrophages. We analyzed myeloid cell populations in the brains of 5XFAD mice at 7 d and 14 d after the first injection of anti–PD-1 (two i.p. injections at 3-d intervals) by separately sorting CD45lowCD11b+ microglia and CD45highCD11b+ cells, which represent mostly infiltrating myeloid cells13. We observed higher frequencies of CD45highCD11b+ cells in the brains of 5XFAD mice following PD-1 blockade, relative to IgG-treated 5XFAD and wild-type (WT) controls (Fig. 1c). Genome-wide transcriptome analysis (Supplementary Table 3) of the myeloid cell populations, sorted from 5XFAD brains after PD-1 blockade, indicated that the CD45highCD11b+ cells expressed a distinct mRNA profile relative to that expressed by the CD45lowCD11b+ cells. The CD45highCD11b+ expression profile included features of infiltrating myeloid cells (characterized by high expression of lymphocyte antigen 6c (Ly6C)) (Fig. 1c), and expression of the chemokine receptor CCR2 (Supplementary Fig. 2a), which is associated with myeloid cell neuroprotection in AD14. These myeloid cells were characterized at the mRNA (Supplementary Fig. 2a,b) and protein (Supplementary Fig. 2c) levels by the expression of scavenger receptor A (SRA1), which is an Aβ-binding scavenger receptor associated with Aβ-plaque clearance15. To determine whether enhanced monocyte-derived macrophage trafficking seen after PD-1 blockade was dependent on IFN-γ, we

1Department

of Neurobiology, Weizmann Institute of Science, Rehovot, Israel. 2Department of Immunology, Weizmann Institute of Science, Rehovot, Israel. Correspondence should be addressed to K.B. ([email protected]) or M.S. ([email protected]). Received 12 October 2015; accepted 4 December 2015; published online 18 January 2016; doi:10.1038/nm.4022

nature medicine



B r i e f c o m m u n i c at i o n s a

b

c

Percentage

Forward scatter

Side scatter

e

© 2016 Nature America, Inc. All rights reserved.

gave 5XFAD mice an IFN-γ–blocking antibody before administering PD-1 blockade. IFN-γ neutralization reduced monocyte-derived macrophages recruitment to the CNS (Fig. 1d) and interfered with mRNA expression of intercellular adhesion molecule 1 (Icam1) and chemokine (C-C motif) ligand 2 (Ccl2) by the choroid plexus, induced by PD-1 blockade (Fig. 1e); these leukocyte-trafficking determinates were previously associated with myeloid cell entry into the CNS via the choroid plexus–cerebrospinal fluid pathway6,7. To examine the potential impact of PD-1 blockade on AD pathology, we first treated 10-month-old 5XFAD mice with either anti–PD-1 antibody or IgG control, and evaluated the effect of the treatment on spatial learning and memory by using the radial arm water maze (RAWM) task. 5XFAD mice that received PD-1 blockade (two i.p. injections at 3-d intervals) were analyzed 1 month later, at which point they exhibited reduced cognitive deficits relative to IgG-treated or untreated age-matched controls (Fig. 2a). 5XFAD mice that received two sessions of PD-1 blockade, with a 1-month interval between sessions, were tested 2 months after the first session, and they exhibited improved cognitive performance relative to IgG-treated or untreated 5XFAD control mice, reaching performance levels comparable to those of WT mice (Fig. 2b). Notably, when 5XFAD mice that had received a single session of PD-1 blockade were examined 2 months after the treatment, only a marginal improvement in memory was observed when compared to IgG-treated mice (Fig. 2b), which suggests that repeated treatment sessions are needed to maintain the beneficial effects on cognition and memory. After behavioral testing, 2 months following treatment initiation, we examined the brains of 5XFAD mice that had received either one or two sessions of PD-1 blockade. Cerebral Aβ plaque load was reduced in the hippocampus (specifically, in the dentate gyrus) (Fig. 2c,d) and in the cerebral cortex (layer V) (Fig. 2c,e), which are the main anatomical regions with robust Aβ-plaque pathology in 5XFAD mice5. Aβ clearance was more pronounced after two sessions of PD-1 blockade than after a single session, and both mouse groups had reduced plaque load relative to untreated or IgG-treated 5XFAD mice. Astrogliosis, as assessed by glial fibrillary acid protein (GFAP) immunoreactivity, was reduced in the hippocampus of 5XFAD mice



CD45

Icam1 mRNA relative expression

-γ An I + ti– gG An P ti– D-1 PD -1

N

IF

i–

An t

i–

An t

An t

i–

IF

IF

N

N

-γ An I + ti– gG an P ti– D-1 PD -1

-γ Ant Ig + i–P G an D ti– -1 PD -1

Percentage of CD45highCD11b+

Ccl2 mRNA relative expression

d

Percentage

5X F

AD

+ 5 5X an XF FA ti–P AD D D + -1 Ig G

Ifng mRNA relative expression

Figure 1  PD-1 blockade promotes myeloid cell 8 ** * recruitment to the CNS via IFN-γ. (a) Gene Cellular response to IFN-γ 6 Cellular response to cytokine stimulus Ontology (GO) annotation terms enriched in 4 Immune response 69% the choroid plexus of 10-month-old 5XFAD mice Response to cytokine Single 2 cells treated with anti–PD-1 (n = 5) and examined on Response to IFN-γ 0 day 10 after the first injection, when compared to 0 5 10 15 Forward scatter Pulse height CD11b IgG-treated (n = 5) and untreated (n = 4) 5XFAD 4.2% 5.65 13.67 WT 5XFAD 12.2% controls (based on Supplementary Table 2; IgG GO enrichment score 86% color scale corresponds to negative log10 of –log10 (P value) P value). (b) mRNA expression levels of Ifng 92.4% 82.3% * * (encoding IFN-γ) in the choroid plexus of anti– 2.0 4 20 ** * *** PD-1–treated (n = 5), IgG-treated (n = 5) and 5XFAD 23.5% 5XFAD 17.9% 1.5 3 15 anti-PD-1 anti-PD-1 untreated (n = 3) 5XFAD mice (one-way analysis d14 d7 1.0 10 2 0% of variance (ANOVA) and Bonferroni post-test; 5 0.5 1 74.2% 71.2% data are representative of three independent 0 0 0 experiments). (c) 5- to 6-month-old 5XFAD CD11b Ly6c mice (n = 3 per group) were i.p. injected on days 1 and 4 with either anti–PD-1 or IgG, and examined at days 7 (d7) and 14 (d14). Flow cytometry sorting gating strategy and quantitative analysis of brain CD45lowCD11b+ (indicated by blue gates and bar fills) and CD45highCD11b+ (indicated in orange) myeloid cells. Myeloid cell populations showed distinct differential expression of Ly6c. (d) 6-month-old 5XFAD mice were injected with IFN-γ–neutralizing antibodies 1 d before PD-1–specific antibody injections and were then examined on day 7. Flow cytometry analysis of CD45 highCD11b+ cell frequencies in the brains of IgG-treated (n = 4) and anti–PD-1–treated (with (n = 5) or without (n = 6) anti−IFN-γ) 5XFAD mice. (e) mRNA expression levels of Ccl2 and Icam1 in the choroid plexus of the same mice (one-way ANOVA and Bonferroni post-test). Error bars represent mean ± s.e.m.; *P < 0.05; **P < 0.01; ***P < 0.001.

treated with either one or two sessions of PD-1 blockade, relative to that in IgG-treated controls (Fig. 2f). We also examined the effect of PD-1 blockade in another AD model, APP/PS1 mice16, which develop Aβ-plaque pathology at a more advanced age than do 5XFAD mice. APP/PS1 mice were tested at two stages of disease progression (8 and 15 months). PD-1 blockade reduced hippocampal Aβ plaque load in PD-1–treated APP/PS1 mice when compared to IgG-treated controls (Fig. 2g,h). Our findings show that in the context of neurodegenerative disease, PD-1 blockade evokes a systemic IFN-γ–dependent immune response that enables the mobilization of monocyte-derived macrophages to the brain. This process is reminiscent of tissue-specific immune surveillance induced by immune checkpoint blockade in cancer therapy10,11,17. PD-1 blockade treatment reduced the cerebral Aβ plaque load in two mouse models of AD in advanced stages of the disease. In particular, repeated treatment sessions were required for maintaining a long-lasting beneficial effect on disease pathology. Given that immune checkpoint blockade releases self-reactive T cells from immune tolerance mechanisms18, these findings support a neuroprotective role for CNS-specific cell-mediated immunity19. Notably, immune checkpoint blockade is not meant to target a single disease-causing etiologic factor in AD; rather, this approach is meant to augment the overall ability of the immune system to clear brain pathology. In cancer immunotherapy, anti–PD-1 and anti–PD-ligand antibodies were shown to be relatively safe and well tolerated20. Taken together, our findings identify immune checkpoint blockade as a novel therapeutic strategy for AD and, potentially, for other neurodegenerative diseases. Methods Methods and any associated references are available in the online version of the paper. Note: Any Supplementary Information and Source Data files are available in the online version of the paper. Acknowledgments We thank I. Slutsky (Tel Aviv University, Tel Aviv, Israel) for APP/PS1 mice, S. Schwarzbaum for proofreading the manuscript, M. Azulai for animal handling and the Krenter Institute for equipment grant support. This work was supported by

nature medicine

B r i e f c o m m u n i c at i o n s

RAWM 12 #

6

### ###

*** ***

4 2 0 2

3

4

5

Day 1 - training

Trials

6

7

8

### ###

Anti–PD-1 (1 session) Anti–PD-1 (2 sessions) IgG (2 sessions) 1

2

3

4

Anti–PD-1 (1 session)

30 20 10 0

###

***

IgG (2 sessions)

Cortex DG

5 6 7 8 9 10 Trials Day 2 - testing

Anti–PD-1 (2 sessions)

g

IgG (2 sessions)

f 6 4 2 0

## ** ###

***

8 6 4 2 0

## ## **

GFAP-normalized immunoreactivity (a.u.)

0

* ### ***

40

Aβ-plaque cortex (layer V) area (%)

2

#

Mean DG Aβ-plaque number

8

Anti–PD-1 ( 2 sessions)

Day 1 - training

e 4

Untreated 5XFAD

2

Day 2 - testing

Untreated 5XFAD

6

WT

* ***

4

9 10

d Aβ-plaque DG area (%)

#

6

Anti–PD-1 (1 session)

Layer V

12

0 1

Time (months) 8

1.5 1.0 0.5 0

IgG # # ** **

h

Anti–PD-1 APP/PS1

HC

9

Time (months) 15

Anti–PD-1 CA1

Anti–PD-1 APP/PS1 IgG HC

16 Anti–PD-1

CA1

Advanced European Research Council (ERC) grants (no. 232835 to M.S. and no. 309788 to I.A.), by the EU Seventh Framework Program HEALTH-2011 (grant no. 279017 to M.S.), by an Israeli Science Foundation grant (no. 1782/11 to I.A.) and by the Weizmann-Tanz collaboration for research in Alzheimer’s disease (to M.S.). M.S. holds the Maurice and Ilse Katz Professorial Chair in Neuroimmunology. Author Contributions K.B. and M.S. conceived and designed the study. K.B., A.D., N.R., A.T.-K., A.M.S. and A.K. performed experiments and analyzed and interpreted the data. O.M.-N. and E.D., under the supervision of I.A., performed RNA-seq analysis. K.B. and A.D. prepared the data for presentation. The manuscript was written by K.B. and M.S. COMPETING FINANCIAL INTERESTS The authors declare competing financial interests: details are available in the online version of the paper. Reprints and permissions information is available online at http://www.nature.com/ reprints/index.html. 1. Heneka, M.T., Kummer, M.P. & Latz, E. Nat. Rev. Immunol. 14, 463–477 (2014).

nature medicine

Mean HC Aβplaque number

Aβ-plaque HC area (%)

Mean HC Aβplaque number

Figure 2  PD-1 blockade reduces AD pathology and improves memory in 5XFAD IgG IgG and APP/PS1 mice. Male 5XFAD mice (average cohorts aged 10 months) were Anti–PD-1 Anti–PD-1 treated with either PD-1–specific antibody or IgG control. Experimental design 2.0 *** 40 6 60 ** ** ** is presented. Black arrows indicate time points of treatment, and illustrations indicate 1.5 30 4 40 1.0 20 time points of cognitive testing or Aβ plaque–burden assessment. (a) RAWM performance 2 20 0.5 10 of anti–PD-1–treated mice (n = 9), of IgG-treated (n = 6) 5XFAD mice and of untreated 0 0 0 0 5XFAD (n = 9) and wild-type (WT) (n = 9) controls; two-way repeated-measures ANOVA and Bonferroni post-test). (b) RAWM performance, comparing one anti–PD-1 treatment session (n = 9) to two sessions with a 1-month interval (n = 6), and untreated aged-matched 5XFAD (n = 7) and IgG-treated (n = 9) controls, and WT (n = 9) controls (combined data from separate experiments which included treated and control groups; two-way repeated-measures ANOVA and Bonferroni post-test). (c–f) Representative immunofluorescence images (c), and quantitative analysis (d–f) of Aβ and astrogliosis, assessed 2 months after the first treatment, in the brains of anti–PD-1–treated 5XFAD mice (after either one session (n = 4) or two sessions (n = 6)) and of controls (untreated (n = 7) and IgG-treated (n = 6) 5XFAD mice). Brains were immunostained for Aβ (in red), GFAP (in green) and Hoechst nuclear staining. Scale bars, 50 µm. Mean plaque area and numbers were quantified (in 6-µm brain slices) in the dentate gyrus (DG) and in the cerebral cortex (layer V), and GFAP immunoreactivity was measured in the hippocampus (one-way ANOVA and Bonferroni post-test). (g,h) APP/PS1 mice were treated with either PD-1–specific antibody or IgG control and examined 1 month later. Brains were immunostained for Aβ (in red) and Hoechst nuclear staining. Scale bars, 250 µm. Mean Aβ plaque area and numbers were quantified in the hippocampus (HC) (in 6-µm brain slices; Student’s t test). Representative immunofluorescence images and quantitative analysis of 8-month-old male mice (n = 4 per group) (g) and 15-month-old female mice (n = 4 per group) (h). CA1, region I of hippocampus proper. Error bars represent mean ± s.e.m.; *, anti–PD-1–treated versus IgG-treated controls; #, anti–PD-1–treated versus untreated controls; *,#P < 0.05; **,##P < 0.01; ***,###P < 0.001. Aβ-plaque HC area (%)

© 2016 Nature America, Inc. All rights reserved.

11

Mean cortex (layer V) Aβ-plaque number

RAWM number of errors

8

Untreated 5XFAD

RAWM

Aβ Hoechst

11

5XFAD Time 10 (months) 8

c

Plaque load

Anti–PD-1

Aβ Hoechst

5XFAD Time 10 (months)

Anti–PD-1

Aβ GFAP Hoechst

b

Anti–PD-1

RAWM number of errors

a

2. Wyss-Coray, T. & Rogers, J. Cold Spring Harb. Perspect. Med. 2, a006346 (2012). 3. Arvanitakis, Z. et al. Neurology 70, 2219–2225 (2008). 4. Schwartz, M. & Baruch, K. EMBO J. 33, 7–22 (2014). 5. Oakley, H. et al. J. Neurosci. 26, 10129–10140 (2006). 6. Baruch, K. et al. Nat. Commun. 6, 7967 (2015). 7. Kunis, G. et al. Brain 136, 3427–3440 (2013). 8. Shechter, R. et al. Immunity 38, 555–569 (2013). 9. Pardoll, D.M. Nat. Rev. Cancer 12, 252–264 (2012). 10. Lesokhin, A.M., Callahan, M.K., Postow, M.A. & Wolchok, J.D. Sci. Transl. Med. 7, 280sr1 (2015). 11. Peng, W. et al. Cancer Res. 72, 5209–5218 (2012). 12. Baruch, K. et al. Science 346, 89–93 (2014). 13. Gate, D., Rezai-Zadeh, K., Jodry, D., Rentsendorj, A. & Town, T. J. Neural Transm. 117, 961–970 (2010). 14. Naert, G. & Rivest, S. J. Mol. Cell Biol. 5, 284–293 (2013). 15. El Khoury, J. et al. Nature 382, 716–719 (1996). 16. Jankowsky, J.L. et al. Hum. Mol. Genet. 13, 159–170 (2004). 17. Okazaki, T. et al. Nat. Immunol. 14, 1212–1218 (2013). 18. Kong, Y.C. & Flynn, J.C. Front. Immunol. 5, 206 (2014). 19. Moalem, G. et al. Nat. Med. 5, 49–55 (1999). 20. Chinai, J.M. et al. Trends Pharmacol. Sci. 36, 587–595 (2015).



ONLINE METHODS

© 2016 Nature America, Inc. All rights reserved.

Animals. Heterozygous 5XFAD transgenic mice (on a C57/BL6-SJL background) that overexpress familial AD mutant forms of human APP (the Swedish mutation, K670N/M671L; the Florida mutation, I716V; and the London mutation, V717I) and PS1 (M146L/L286V) transgenes under the transcriptional control of the neuron-specific mouse Thy-1 promoter5 (5XFAD line Tg6799; The Jackson Laboratory). Genotyping was performed by PCR analysis of tail DNA, as previously described5. Male and female mice were bred and maintained by the animal breeding center of the Weizmann Institute of Science. AD double-transgenic B6.C3-Tg (APPswe, PSEN1dE9) 85Dbo/ Mmjax mice17 (on a C57BL/6 background) were a gift from Dr. Inna Slutsky, Tel Aviv University, Tel Aviv, Israel. All experiments detailed herein complied with the regulations formulated by the Institutional Animal Care and Use Committee (IACUC) of the Weizmann Institute of Science. Antibodies. For PD-1 blockade, PD-1–specific blocking antibody (anti–PD-1; rat isotype; clone RPM1-14; BIOXCELL) and isotype control immunoglobulin (rat IgG2a; BIOXCELL) were administered i.p. at days 1 and 4 of each treatment session at a dose of 250 µg per mouse. For IFN-γ neutralization, mice were treated with 500 µg of an IFN-γ–specific blocking antibody (anti–IFN-γ; clone XMG1.2; BIOXCELL) on the day before each anti–PD-1 injection. RNA purification, cDNA synthesis and quantitative real-time PCR analysis. Mice were transcardially perfused with phosphate-buffered saline (PBS) before tissue excision. Choroid plexus tissues were isolated under a dissecting microscope (Stemi DV4; Zeiss) from the lateral, third and fourth ventricles of the brain. Total RNA from the choroid plexus was extracted using the RNA MicroPrep Kit (Zymo Research), and mRNA (1 µg) was converted into cDNA using a High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). The expression of specific mRNAs was assayed using fluorescence-based quantitative real-time PCR (RT-qPCR) (Fast-SYBR PCR Master Mix; Applied Biosystems). Quantification reactions were performed in triplicate for each sample using the ‘delta-delta Ct’ method. Peptidylprolyl isomerase A (Ppia) was chosen as a reference (housekeeping) gene. At the end of the assay, a melting curve was constructed to verify the specificity of the reaction. To determine the expression levels of Ifng, cDNA was pre-amplified for 14 PCR cycles, according to the manufacturer’s protocol (PreAmp Master Mix Kit; Applied Biosystems), thereby increasing the sensitivity of the subsequent real-time PCR reaction. The TaqMan Assays-on-Demand probes Mm02342430_g1 (Ppia) and Mm01168134_m1 (Ifng) were used. For other genes examined, the following primers were used: Ppia forward 5′-AGCATACAGGTCCTGGCATCTTGT-3′ and reverse 5′-CAAAGACCACATGCTTGCCATCCA-3′; Icam1 forward 5′-AGATCACATTCACGGTGCTGGCTA-3′ and reverse 5′-GCTTTG GGATGGTAGCTGGAAGA-3′; Ccl2 forward 5′-CATCCACGTGTTGG CTCA-3′ and reverse 5′-GATCATCTTGCTGGTGAATGAGT-3′; RT-qPCR reactions were performed and analyzed using StepOne software V2.2.2 (Applied Biosystems). Immunohistochemistry. Mice were transcardially perfused with PBS before tissue excision and fixation. Tissues that were not adequately perfused were not further analyzed, because autofluorescence associated with blood contamination interferes with immunostaining analyses. Tissue processing and immunohistochemistry were performed on paraffin-embedded, sectioned (6 µm thick) mouse brains. The following primary antibodies were used: mouse anti-Aβ (1:300, Covance, #SIG-39320, clone 6E10) and rabbit anti-GFAP (1:200, Dako, #Z0334, #LOT 00085137). Secondary antibodies were Cy2 and Cy3 conjugated donkey anti-mouse or anti-rabbit antibodies (1:200; all from Jackson ImmunoResearch). The slides were exposed to Hoechst nuclear staining (1:4,000; Invitrogen Probes) for 1 min, before being sealed with Aquamount (Polysciences) and glass covers. Two negative controls were routinely used in immunostaining procedures, which involved staining with isotype control antibody followed by secondary antibody, or staining with secondary antibody alone. Microscopic analysis was performed using a fluorescence microscope (E800; Nikon) equipped with a digital camera (DXM 1200F; Nikon), and with either a 20× numerical aperture (NA) 0.50 or 40× NA 0.75 objective lens (Plan Fluor; Nikon). Recordings were made on postfixed tissues using acquisition software (NIS-Elements, F3; Nikon).

nature medicine

For the quantification of staining intensity, total cell and background fluorescence intensity was measured using ImageJ software (from the US National Institutes of Health; NIH), and the intensity of specific staining was calculated, as previously described21. Images were cropped, merged and optimized using Photoshop CS6 13.0 (Adobe), and they were arranged using Illustrator CS5 15.1 (Adobe). Flow cytometry and sorting, and sample preparation and analysis. Mice were transcardially perfused with PBS before tissue extraction. Spleens were mashed with the plunger of a syringe and treated with ammonium chloride potassium (ACK)-lysing buffer to remove erythrocytes. Brains were removed under a dissecting microscope (Stemi DV4; Zeiss) in PBS, and tissues were dissociated using the GentleMACS dissociator (Miltenyi Biotec). All samples were filtered through a 70-µm nylon mesh and blocked with anti-Fc CD16/32 (1:100; BD Biosciences) before immunostaining. For intracellular staining of IFN-γ, the cells were incubated with phorbol 12-myristate 13-acetate (PMA; 10 ng/ml; Sigma-Aldrich) and ionomycin (250 ng/ml; Sigma-Aldrich) for 6 h, and brefeldin-A (10 µg/ml; Sigma-Aldrich) was added for the last 4 h of incubation. Intracellular labeling of cytokines was performed using BD Cytofix/ Cytoperm Plus fixation/permeabilization kit (cat. no. 555028) according to the manufacturer’s protocol. The following fluorochrome-labeled monoclonal antibodies were purchased from BD Pharmingen, BioLegend, R&D Systems or eBiosciences and used according to the manufacturers’ protocols: Brilliant-violet-421 (1:200) or PerCP-Cy5.5-conjugated anti-CD45 (1:400); phycoerythrin (PE) or Alexa Fluor 450–conjugated anti-CD4 (1:200); fluorescein isothiocyanate (FITC)-conjugated anti-TCRβ (1:200); PerCP-Cy5.5–conjugated anti-CD11b (1:400); PE-conjugated anti-Ly6c (1:200); APC-conjugated anti–IFN-γ (1:50); allophycocyanin (APC)-conjugated SRAI/MSR (SRA1; 1:20). Cells were analyzed on an LSRII cytometer (BD Biosciences) using FACSdiva (BD Biosciences) and FlowJo (Tree Star, Inc.) software. In each experiment, relevant negativecontrol groups, positive controls and single-stained samples for each tissue were used to identify the populations of interest and to exclude others. In sorting experiments, 1,500–3,000 myeloid cells were collected per sample using the FACSAriaIII sorter (BD Biosciences) into 50 µl of lysis buffer. RNA was extracted from sorted cells, DNA libraries were produced and sequencing was conducted, as described below. RNA sequencing, library construction and analysis. For each library, 10 ng of RNA from each sample was used. A derivation of MARS-seq22, developed for single-cell RNA-seq, was used to produce sensitive and robust RNA expression libraries. A minimum of two replicates were used per population. An average of 4 million reads per library was obtained and aligned to the mouse reference genome (National Center for Biotechnology Information (NCBI) 37, mm9) using TopHat v2.0.10 (ref. 23) with default parameters. Expression levels were calculated and normalized using Homer24. RNA-seq analysis of the choroid plexus was focused on genes with levels of expression above the sixtieth percentile and was robust across different cutoffs. A constant value representing the sixtieth percentile was added to each data point in order to reduce variability between low-level expressed genes. Genes were ordered according to their average expression levels in anti–PD-1 injected mice, when compared to IgG-treated and untreated 5XFAD mice, and they were analyzed for Gene Ontology (GO) enrichment using Gorilla (http://cbl-gorilla.cs.technion.ac.il/). RNA-seq analysis of myeloid cells was focused on genes with levels of expression above the fiftieth percentile (to remove low expressed genes), which were filtered for nonchanging genes (maximum median of sets − minimum median of sets > 0.75), followed by K-means clustering on median columns (K = 4). Heat maps were prepared using GENE-E (http://www.broadinstitute.org/ cancer/software/GENE-E/). Radial-arm water maze. The RAWM was used to test spatial learning and memory, as previously described in detail25. Briefly, six stainless steel inserts were placed in the tank, forming six swim arms radiating from an open central area. The escape platform was located at the end of one arm (the goal arm), 1.5 cm below the water’s surface, in a pool 1.1 m in diameter. The water temperature was kept at 21–22 °C. Water was made opaque with milk powder. In the testing room, only distal visual shape and object cues were available to the mice

doi:10.1038/nm.4022

© 2016 Nature America, Inc. All rights reserved.

to aid in finding the location of the submerged platform. On day 1, mice were trained for 15 trials with alternating visible and hidden platforms, and the last four trials with the hidden platform only. On day 2, mice were trained for 15 trials with the hidden platform. Entry into an incorrect arm, or failure to select an arm within 15 s, was scored as an error. Spatial learning and memory were measured by counting the number of arm entry errors, by a researcher blinded to the identity of the mice. Mice that displayed motor deficits in swimming performance were excluded at the beginning of the experiments from further analysis. No motor deficits were observed in relation to treatments. Data were analyzed as the mean number of errors, for training blocks of three consecutive trials. Ab plaque quantitation. From each brain, 6-µm coronal slices were collected from five different pre-determined depths, all together covering 600 µm throughout the region of interest (of the hippocampus and cerebral cortex). Slices were immunostained, and histogram-based segmentation of positively stained pixels was performed using Image-Pro Plus software (Media Cybernetics, Bethesda, Maryland, USA). The segmentation algorithm was manually applied to each image in the hippocampus, the dentate gyrus area or in cortical layer V, and the percentage of the area occupied by total Aβ immunostaining was determined. Plaque numbers were quantified from the same 6-µm coronal brain slices, and they are presented as the average number of plaques per brain region. Prior to quantification, samples were coded to mask the identity of the mice, and plaque burden was quantified by an observer blinded to the identity of the treatment groups.

doi:10.1038/nm.4022

Statistical analysis. The specific tests used to analyze each set of experiments are indicated in the figure legends. For each statistical analysis, appropriate tests were selected on the basis of whether the data was normally distributed. Data were analyzed using a two-tailed Student’s t test to compare between two groups, and one-way ANOVA was used to compare several groups, followed by the Bonferroni post-hoc procedure for pairwise comparison of groups after the null hypothesis was rejected (P < 0.05). Data from behavioral tests were analyzed using two-way repeated-measures ANOVA, and Bonferroni post-hoc procedure was used for follow-up pairwise comparison. Sample sizes were chosen with adequate statistical power on the basis of the literature and past experience, and mice were allocated to experimental groups according to age, gender and genotype. RAWM behavioral experiments were carried out in several cohorts of mice that contained all tested groups of treated mice and controls, which were examined in constitutive days, and the data were combined for analysis. Investigators were blinded to the identity of the groups during experiments and during outcome assessment. All inclusion and exclusion criteria were pre-established according to the IACUC. Results are presented as means ± s.e.m. In the graphs, y-axis error bars represent s.e.m. Statistical calculations were performed using GraphPad Prism software (GraphPad Software, San Diego, California). 21. Burgess, A. et al. Proc. Natl. Acad. Sci. USA 107, 12564–12569 (2010). 22. Jaitin, D.A. et al. Science 343, 776–779 (2014). 23. Trapnell, C., Pachter, L. & Salzberg, S.L. Bioinformatics 25, 1105–1111 (2009). 24. Heinz, S. et al. Mol. Cell 38, 576–589 (2010). 25. Alamed, J., Wilcock, D.M., Diamond, D.M., Gordon, M.N. & Morgan, D. Nat. Protoc. 1, 1671–1679 (2006).

nature medicine