Articulometodos 1

RESE ARCH | R E S E A R C H A R T I C L E S
fundamental minimum required thicknesses. IMMUNOLOGY

The complete process is summarized and some
additional discussion is provided in supple- Past history of obesity triggers persistent
mentary text S14 and S15, respectively. As
illustrated above, systems with optimized de- epigenetic changes in innate immunity and
signs already approach these limits within
some small factor (e.g., a factor of 3 or less). exacerbates neuroinflammation
Flat optical systems offer many interesting
possibilities for reducing the thickness of Masayuki Hata1,2, Elisabeth M. M. A. Andriessen3, Maki Hata1, Roberto Diaz-Marin2, Frédérik Fournier2,
systems—e.g., by using metasurfaces to elimi- Sergio Crespo-Garcia1,2, Guillaume Blot1,2, Rachel Juneau1, Frédérique Pilon1, Agnieszka Dejda1,
nate most of the thickness of lenses or other Vera Guber1, Emilie Heckel4, Caroline Daneault5, Virginie Calderon6, Christine Des Rosiers5,
elements. This work shows that, especially for Heather J. Melichar7, Thomas Langmann8, Jean-Sebastien Joyal4,
applications with large ONL, although we still Ariel M. Wilson1, Przemyslaw Sapieha1,2*
need thickness overall, much of this thickness
just needs to transport optical channels later- Age-related macular degeneration is a prevalent neuroinflammatory condition and a major cause of
ally; it may only need to be empty, uniform, blindness driven by genetic and environmental factors such as obesity. In diseases of aging, modifiable
or relatively simple waveguiding space, which factors can be compounded over the life span. We report that diet-induced obesity earlier in life
would simplify overall system design. triggers persistent reprogramming of the innate immune system, lasting long after normalization of
metabolic abnormalities. Stearic acid, acting through Toll-like receptor 4 (TLR4), is sufficient to remodel
RE FE RENCES AND N OT ES
chromatin landscapes and selectively enhance accessibility at binding sites for activator protein-1
1. C. Guo, H. Wang, S. Fan, Optica 7, 1133–1138 (2020).
2. O. Reshef et al., Nat. Commun. 12, 3512 (2021). (AP-1). Myeloid cells show less oxidative phosphorylation and shift to glycolysis, ultimately leading to
3. J. T. R. Pagé, O. Reshef, R. W. Boyd, J. S. Lundeen, proinflammatory cytokine transcription, aggravation of pathological retinal angiogenesis, and neuronal
Opt. Express 30, 2197–2205 (2022). degeneration associated with loss of visual function. Thus, a past history of obesity reprograms
4. K. Shastri, O. Reshef, R. W. Boyd, J. S. Lundeen, F. Monticone,
Optica 9, 738–745 (2022). mononuclear phagocytes and predisposes to neuroinflammation.
5. A. Chen, F. Monticone, ACS Photonics 8, 1439–1447 (2021).
A
6. A. Silva et al., Science 343, 160–163 (2014).
7. H. Wang et al., ACS Photonics 9, 1358–1365 (2022).
8. A. Overvig, A. Alù, Laser Photonics Rev. 16, 2100633 (2022).
ge-related macular degeneration (AMD) immunogenic modified lipoproteins, comple-
9. D. A. B. Miller, J. Opt. Soc. Am. B 24, A1–A18 (2007). is a multifactorial neuroinflammatory ment proteins, vitronectin, amyloid proteins,
10. D. A. B. Miller, Phys. Rev. Lett. 99, 203903 (2007). disease of the aging eye that is caused by and immunoglobulins (8), which attract and
11. M. Gerken, D. A. B. Miller, Appl. Opt. 44, 3349–3357 (2005).
12. Z. Kuang, L. Zhang, O. D. Miller, Optica 7, 1746–1757 (2020).
genetic and environmental risk factors activate cells of the innate immune system.
13. M. I. Abdelrahman, F. Monticone, How Thin and Efficient Can (1, 2). It is the leading cause of irreversible These cells include resident retinal microg-
a Metasurface Reflector Be? Universal Bounds on Reflection blindness in the world (3). Investigation of the lia and recruited monocytes or macrophages
for Any Direction and Polarization. arXiv:2208.05533
genetic predisposition to AMD identified var- (9, 10), which attempt to achieve tissue repair
[physics.optics] (2022).
14. P. Chao, B. Strekha, R. Kuate Defo, S. Molesky, A. W. Rodriguez, iants in susceptibility gene loci but not causa- through low-grade inflammation. However,
Nat. Rev. Phys. 4, 543–559 (2022). tive gene mutations. Therefore, environmental similarly to other neurodegenerative diseases,
15. D. A. B. Miller, Opt. Express 20, 23985–23993 (2012). factors likely contribute to the pathogenesis of early AMD is triggered by chronic parainflam-
16. D. A. B. Miller, Adv. Opt. Photonics 11, 679–825 (2019).
17. N. K. Fontaine et al., Nat. Commun. 10, 1865 (2019). AMD. This notion is supported by monozygotic mation, overly activated phagocytes, and dys-
18. S. Pai et al., IEEE J. Sel. Top. Quantum Electron. 26, 1–13 (2020). twin studies demonstrating that, along with regulation of the complement system (11, 12).
19. B. J. Shastri et al., Nat. Photonics 15, 102–114 (2021). age, environmental factors determine disease Moreover, inadequate clearance of drusen by
20. F. Ashtiani, A. J. Geers, F. Aflatouni, Nature 606, 501–506 (2022).
21. X. Lin et al., Science 361, 1004–1008 (2018). susceptibility (4). phagocytes precipitates further pathologic in-
22. W. Bogaerts et al., Nature 586, 207–216 (2020). In the early stages of AMD, insoluble extra- flammatory and angiogenic responses that lead
23. G. Wetzstein et al., Nature 588, 39–47 (2020). cellular deposits called drusen, containing lip- to tissue damage that predisposes to late AMD
24. D. A. B. Miller, Photon. Res. 1, 1–15 (2013).
25. D. A. B. Miller, J. Lightwave Technol. 35, 346–396 (2017).
ids, proteins, hydroxyapatite, and trace metals, (13). Late AMD is classified into two forms, neo-
26. V. Blahnik, O. Schindelbeck, Adv. Opt. Technol. 10, 145–232 (2021). form in the subretinal pigment epithelium vascular AMD (nAMD) and non-neovascular
27. J. W. Goodman, Introduction to Fourier Optics (Macmillan, (RPE) space (5–7). Components of drusen also AMD (geographic atrophy). nAMD causes >80%
2017).
28. D. A. B. Miller, An introduction to functional analysis for science
include various inflammatory factors such as of vision loss in patients with AMD (14) and is
and engineering. arXiv:1904.02539 [math.FA] (2019). characterized by the growth of abnormal blood
1
29. B. Fornberg, Math. Comput. 51, 699–706 (1988). Department of Ophthalmology, Maisonneuve-Rosemont vessels from the choroid under the macula,
Hospital Research Centre, University of Montreal, Montreal, known as choroidal neovascularization (CNV).
ACKN OW LEDG MEN TS Quebec H1T 2M4, Canada. 2Department of Biochemistry and
The author acknowledges stimulating conversations with S. Fan. Molecular Medicine, Maisonneuve-Rosemont Hospital CNV progression leads to retinal edema or
Funding: D.A.B.M. received funding for this work from the Research Centre, University of Montreal, Montreal, Quebec hemorrhage, culminating in photoreceptor
MURI program supported by AFOSR grant no. FA9550-21-1-0312. H1T 2M4, Canada. 3Department of Biomedical Sciences, death. Abnormal neovascularization ultimate-
Competing interests: The author declares that he has no Maisonneuve-Rosemont Hospital Research Centre, University
of Montreal, Montreal, Quebec H1T 2M4, Canada. 4Departments
ly causes fibrovascular scarring and can lead
competing interests. Data and materials availability: All data
are available in the main text or the supplementary materials. License of Pediatrics, Ophthalmology, and Pharmacology, Centre to permanent loss of central vision (15). In
information: Copyright © 2023 the authors, some rights reserved; Hospitalier Universitaire Ste-Justine Research Center, Montreal, addition to a local immune response, progres-
exclusive licensee American Association for the Advancement of Quebec H3T 1C5, Canada. 5Department of Nutrition, University
of Montreal, Montreal, Quebec, Plateforme métabolomique
sion of AMD is also influenced by heightened
Science. No claim to original US government works. https://www.
science.org/about/science-licenses-journal-article-reuse de l’Institut de Cardiologie de Montréal, Montreal, Quebec H3C systemic inflammation (16). How distal in-
3J7, Canada. 6Bioinformatics & Molecular Biology Core Facility, flammation influences AMD remains incom-
SUPPLEMENTARY MATERIALS Institut de Recherches Cliniques de Montréal, Montreal, Quebec
pletely understood.
H2W 1R7, Canada. 7Department of Medicine, Maisonneuve-
science.org/doi/10.1126/science.ade3395 Among modifiable factors, obesity is the
Rosemont Hospital Research Centre, University of Montreal,
Supplementary Text
Montreal, Quebec H1T 2M4, Canada. 8Laboratory for Experimental second most important risk factor after smok-
Figs. S1 to S11
Immunology of the Eye, Department of Ophthalmology, ing for late AMD (17). Patients with nAMD
References (30–40)
Faculty of Medicine and University Hospital Cologne,
Submitted 9 August 2022; accepted 31 October 2022 University of Cologne, 50931 Cologne, Germany. have significantly larger volumes of visceral
10.1126/science.ade3395 *Corresponding author. Email: [email protected] adipose tissue (18), and each increase of 0.1 in
SCIENCE science.org 6 JANUARY 2023 • VOL 379 ISSUE 6627 45

RES EARCH | R E S E A R C H A R T I C L E S
Fig. 1. DIO triggers long-term changes in eWAT that exacerbate pathologi- subjected to laser-induced CNV. Two weeks after CNV induction, mice were
cal angiogenesis. (A) Experimental schematic in which mice started a HFD euthanized and eyes collected. (B and C) ITT on RD-RD and HFD-RD mice at
at 8 weeks of age and were then switched back to a RD at 19 weeks, which is time point 1 (RD-RD, n = 5; HFD-RD, n = 6) (B) and time point 2 (RD-RD, n = 10;
time point 1; this group is HFD-RD. Control mice were fed a RD throughout and HFD-RD, n = 11) (C). (D and E) GTT on RD-RD and HFD-RD mice at time
are referred to as RD-RD. At 28 weeks, which is time point 2, mice were point 1 (RD-RD, n = 20; HFD-RD, n = 21) (D) and time point 2 (RD-RD, n = 18;
46 6 JANUARY 2023 ¥ VOL 379 ISSUE 6627 science.org SCIENCE

HFD-RD, n = 21) (E). (F and G) Plasma insulin levels in RD-RD and HFD-RD in the backs of the recipient mice. Note the blood vessels that sprouted
mice at time point 1 (RD-RD, n = 4; HFD-RD, n = 5) (F) and time point 2 (RD-RD, and are now supplying the grafts. (O) Compilation of representative compressed
n = 6; HFD-RD, n = 5) (G) during GTT. (H) Representative confocal images Z-stack confocal images showing IB4-stained laser burns with FITC-dextran–
of IB4-stained laser burns with FITC-dextran–labeled CNV and IBA1-stained labeled CNV and IBA1-stained MNPs in sham-operated mice and RD-RD ATT
MNPs in RD-RD and HFD-RD mice. Scale bar, 50 μm. (I to K) Quantification of and HFD-RD ATT recipient mice. Scale bar, 50 μm. (P to R) Quantification of the
the FITC-dextran–labeled CNV area (I), the IB4-stained laser impact area (J), FITC-dextran–labeled CNV area (P), the IB4-stained laser impact area (Q),
and the ratio of FITC/IB4 per laser burn (K) relative to RD-RD mice at D14; and the ratio of FITC/IB4 per laser burn (R) relative to sham at day 14; n = 30
n = 46 burns for RD-RD; n = 54 burns for HFD-RD. (L) Number of IBA1+ MNPs burns for sham, n = 31 burns for RD-RD ATT, and n = 32 burns for HFD-RD
around the laser impact area at D14; n = 46 burns for RD-RD; n = 56 burns ATT. (S) Number of IBA1+ MNPs around the laser impact area at day 14 ;
for HFD-RD. (M) Experimental schematic of the ATT in which recipient mice were n = 29 burns for sham, n = 35 burns for RD-RD ATT, and n = 34 burns for HFD-
transplanted with eWAT fat pads at 8 weeks of age from either RD-RD or RD ATT. Data information: Comparisons between groups were analyzed
HFD-RD donor mice. Sham surgeries of controls were performed similarly but using two-way ANOVA with Sidak’s multiple-comparisons test [(B) to (G)],
without ATT. Mice were subjected to laser burns at 11 weeks and euthanized Student’s unpaired t test [(I) to (L)], or one-way ANOVA with Tukey’s multiple-
at week 13. (N) Adipose tissue grafts 3 weeks after ATT. The grafts (white comparisons test [(P) to (S)]. *P < 0.05, **P < 0.01, ***P < 0.001, ****P <
arrowheads) were dissected from the epididymis of the donor mice and inserted 0.0001. Error bars represent mean ± SEM.
the waist/hip ratio, a measure of abdominal ilar to RD-RD mice within 6 weeks and remain- Given that adipose tissue undergoes mor-
obesity, is associated with a 75% increase in the ing similar throughout the course of study (fig. phological and functional changes during obe-
odds of late AMD (17). Longitudinal anal- S1A, time point 2). The systemic metabolic con- sity and is one of the largest immunologically
ysis from the Age-Related Eye Disease Study sequences of DIO in HFD mice were normal- active organs, we tested whether adipose tis-
(AREDS) showed that a higher body-mass in- ized, with subsequent weight loss upon return sue from formerly obese mice displayed any
dex (BMI > 30) was associated with progres- to RD, as demonstrated by a comparable re- residual properties that might contribute to
sion to late AMD and more years of disease (19). sponse to insulin challenge [insulin tolerance disease. We conducted adipose tissue trans-
The mechanisms through which obesity pre- test (ITT); Fig. 1, B and C, and fig. S1, B and C] plantations (ATTs) in which equal amounts
disposes to late AMD remain poorly defined. and glucose challenge [glucose tolerance test (500 mg) of epididymal visceral white adipose
Compounded effects of obesity throughout (GTT); Fig. 1, D and E]. Plasma insulin concen- tissue (eWAT) fat pads from RD-RD mice or
life have been heavily investigated, revealing trations during GTT (Fig. 1, F and G) at time HFD-RD mice were transplanted into C57BL/
that weight loss in obese patients reduced point 2 were similar between HFD-RD and 6J recipient mice (28, 29) (Fig. 1M). All mice
adipose tissue inflammation and reinstated RD-RD control mice. progressively gained weight from 1 week after
glycemic control (20–23). However, the long- After 20 weeks, RD-RD mice and HFD-RD ATT, and no significant differences in weight
term impact of prior obesity on the immune mice were subjected to laser-induced photoco- were observed among groups (fig. S2A). Surgery-
response in later life remains unknown. We agulation in the eye to trigger CNV (27) (Fig. 1A). induced inflammation had subsided in recip-
sought to determine the long-term conse- Two weeks after treatment, CNV lesions and ient mice within 7 days, as demonstrated by
quences of a past history of obesity on neuro- laser-burned areas were quantified with high the normalization of plasma tumor necrosis
inflammatory complications of the retina that molecular weight fluorescein isothiocyanate factor (TNF), as well as circulating MNPs and
lead to CNV and retinal degeneration and (FITC)–dextran and isolectin B4 (IB4) staining neutrophils (fig. S2, B to I). To verify successful
whether weight loss restores immune homeo- in choroid flatmounts. FITC-dextran perfusion ATT, we confirmed that grafts were healthy
stasis in the aging eye. permits visualization of neovessels with lu- and vascularized without necrosis (Fig. 1N).
men, and IB4 stains endothelial cells in neo- Three weeks after transplantation, recipient
RESULTS vascularization and choriocapillaris beneath mice were subjected to laser-induced CNV.
Past history of obesity triggers persistent the laser-burned Bruch’s membrane. Quantifi- Mice receiving HFD-RD ATT developed more
changes in visceral adipose tissue cation of FITC-dextran–perfused newly formed CNV than RD-RD ATT recipient mice or sham-
and predisposes mice to pathological vessels revealed a 40% increase in CNV in HFD- operated mice (Fig. 1, O to R). The number of
angiogenesis in the retina RD mice compared with that of RD-RD mice MNPs recruited to sites of injury also did not
Diet-induced obesity (DIO) promotes systemic (Fig. 1, H to K). The average size of IB4-labeled differ significantly (P = 0.55) between groups
inflammation (24). Persistent obesity exacer- lesion areas did not differ between groups, (Fig. 1S). Collectively, these data demonstrate
bates CNV through activation of systemic in- suggesting that the observed effect is directly that despite normalization of body weight
nate immunity (25, 26). However, whether on the nascent vasculature (Fig. 1, J and K). The and correction of metabolic abnormalities, adi-
weight loss after obesity can reverse this ef- numbers of mononuclear phagocytes [(MNPs), pose tissue from formerly obese mice retains
fect is poorly understood. We therefore set up which are labeled with ionized calcium-binding properties that promote pathological neovas-
weight gain–weight loss (WGWL) experiments adaptor molecule 1 (IBA1)] recruited to sites of cularization after experimental injury in distal
in which we placed male C57BL/6J mice on injury in RPE-choroid-sclera complexes was tissues.
a high-fat diet (HFD: 60% lipid content) for comparable between RD-RD mice and HFD-RD
11 weeks to provoke DIO and then switched mice (Fig. 1, H and L). Flow cytometric analysis Adipose tissue macrophages are primed for
them to a regular diet (RD: 10% lipid content, further confirmed that the numbers of MNPs cytokine production by past obesity and
with lipid source as in the HFD) for 9 weeks to and, more specifically, CX3C motif chemokine maintain proinflammatory profiles after weight loss
induce weight loss (HFD-RD mice). Control receptor 1–positive (CX3CR1+) microglia, were To investigate the source of angiogenic mem-
groups were fed a RD throughout the study comparable between RD-RD mice and HFD- ory in the visceral adipose tissue of HFD-RD
(20 weeks; RD-RD mice) (Fig. 1A). After 11 weeks, RD mice (fig. S1, D to H). Thus, a past history mice, we first conducted a gross histological
mice on HFD gained three times more weight of obesity leads to persistent changes in later assessment with hematoxylin & eosin (H&E)
than mice on RD (fig. S1A, time point 1). The life that enhance neovascularization in retinal staining (Fig. 2A). The weight of adipose tissue
weight of HFD-RD mice gradually decreased tissue without affecting the absolute numbers in HFD-RD mice returned to levels similar to
after return to RD, returning to weights sim- of lesion-associated MNPs. that in RD-RD mice, whereas that of mice fed


Fig. 2. ATMs are primed by prior obesity and maintain a proinflammatory and YFP+ MNPs in retinas from sham-operated and LysMCre/+ ATT and LysMCre/+:
profile after weight loss. (A) H&E staining of eWAT sections from HFD mice Ai3EYFP/+ ATT recipient mice. (L) Percentage of YFP+ cells of viable MNPs in
(HFD feeding for 11 weeks), RD-RD mice, and HFD-RD mice. Scale bar, 50 μm. sham (n = 9) and LysMCre/+ ATT (n = 10) and LysMCre/+:Ai3EYFP/+ ATT
(B) Fat pad weight per body weight (g/g) from eWAT of HFD (n = 8), RD-RD (n = 10) recipient mice. (M) mRNA expression of Il1b, Il6, Tnf, and Nfkb in
(n = 6), and HFD-RD (n = 6) mice. (C and D) Adipocyte sizes in eWAT sections of monocytes isolated from BM of mice 72 hours after laser burn relative to RD-RD;
HFD, RD-RD, and HFD-RD mice (n = 6). Area under the curve (AUC) of the n = 4 RD-RD, n = 3 HFD-RD. (N) Experimental schematic of BMT in which
percent total of large [> 7000 μm2 (C)] and small [≤ 7000 μm2 (D)] adipocytes. lethally irradiated B6.SJL (CD45.1) recipient mice were reconstituted with BM
(E) Quantification of crown-like structures in adipose tissue sections of HFD cells at 8 weeks of age from RD-RD or HFD-RD C57BL/6J (CD45.2) mice.
(n = 6), RD-RD (n = 5), and HFD-RD (n = 6) mice. (F to I) Flow cytometry Blood was collected from recipient mice for flow cytometry analysis 8 weeks
analysis of eWAT from RD-RD and HFD-RD mice. (F) Representative flow plots of after BMT, and then recipient mice were subjected to laser-induced CNV.
ATMs in eWAT of RD-RD and HFD-RD mice. (G) Quantification of the number (O) Quantification of CD45.1+ and CD45.2+ circulating monocytes (Ly6G–/CD11b+/
of macrophages (Ly6G–/CD45.2+/CD11b+/CD11clo/int/ F4/80+) per gram fat pad F4/80+/Ly6C+) of a CD45.1+ control mouse and an irradiated CD45.1+ mouse
in eWAT of RD-RD (n = 15) and HFD-RD (n = 14) mice. (H) Representative having received CD45.2+ RD-RD or HFD-RD BM. (P) Representative confocal
flow plots of CD38+ and CD206+ ATMs in eWAT of RD-RD and HFD-RD mice. images showing IB4-stained laser burns with FITC-dextran–labeled CNV
(I) Quantification of CD38+ macrophages (Ly6G–/CD45.2+/CD11b+/CD11clo/int/ and IBA1-stained MNPs in RD-RD and HFD-RD BMT mice. Scale bar, 50 μm.
F4/80+/ CD38+) in RD-RD mice and HFD-RD mice (n = 4). (J) Experimental (Q) Quantification of the ratio of FITC/IB4 per laser burn relative to RD-RD
schematic of ATT in which recipient mice were transplanted with eWAT fat mice; n = 32 burns for RD-RD BMT, n = 33 burns for HFD-RD BMT. Data information:
pads at 8 weeks of age from either LysMCre/+:Ai3EYFP/+ or LysMCre/+ donor mice. Comparisons between groups were analyzed using one-way ANOVA with Tukey’s
Sham surgeries of controls were performed similarly but without ATT. Mice multiple-comparisons test [(B) to (E), (L), and (O)] or Student’s unpaired t test [(G),
were subjected to laser burns at 11 weeks and euthanized 3 days later. (I), (M), and (Q)]. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Error bars
(K) Representative flow plots of MNPs (Ly6G–/CD11b+/F4/80+/CX3CR1+/CD45.2+) represent mean ± SEM.
HFD throughout the course of experimenta- labeled ATMs. We used lysozyme 2 (LysM)Cre/+ (RT-qPCR) and found that levels of interleukin
tion (HFD mice) remained significantly greater mice, which express Cre in the myeloid cell 1b (Il1b) and Il6 rose significantly in HFD-RD
(Fig. 2B). The mean of frequency of adipocyte lineage, and bred them with Ai3EYFP/+ mice, mice compared with RD-RD controls (Fig. 2M).
size did not differ between RD-RD and HFD- which harbor a targeted mutation of the To test whether the BM from formerly obese
RD mice (fig. S3, A and B); we however found Gt(ROSA)26Sor (R26) locus with the loxP-flanked mice might contribute to the development of
a similar increase in the number of large adi- STOP cassette, preventing the transcription of CNV, we performed BM transfers (BMTs) from
pocytes (>7000 mm) in HFD and HFD-RD mice enhanced yellow fluorescent protein (EYFP). RD-RD or HFD-RD C57BL/6J (CD45.2) mice into
compared to RD-RD mice and a larger number The resulting mice had fluorescent monocytes, lethally irradiated B6.SJL (CD45.1) recipient
of small adipocytes (≤ 7000 mm) in RD-RD mice mature macrophages, and granulocytes. We mice (Fig. 2N and fig. S4, E and F). Eight weeks
(Fig. 2, C and D). These results suggest that transplanted eWAT fat pads from LysMCre/+: after BMT, the circulating population of mono-
obesity-induced changes in the size and distribu- Ai3EYFP/+ mice or LysMCre/+ mice into C57BL/ cytes expressed the donor mouse allelic variant
tion of adipocytes persist long after weight loss. 6J recipient mice (Fig. 2J). CNV was induced of the pan-hematopoietic cell marker CD45,
Several immune cell subsets in adipose tis- 3 weeks after transplantation, and once surgi- attesting to successful transfer (Fig. 2O and fig.
sue have important roles in modulating obesity- cal inflammation had subsided (fig. S2, B to I), S4G). Most ATMs in recipient mice were de-
associated inflammation (30–32). Although we assessed whether myeloid cells within trans- rived from donor mice (fig. S4H), indicating
prior exposure to HFD feeding did not grossly planted fat pads could migrate and contribute that ATMs were replaced by BM-derived cells
affect T cell numbers (fig. S3, C to G) or T cell– to AMD pathology locally. Three days after in irradiated recipient mice. When chimeric
mediated cytokine production (fig. S3, H to CNV induction, we detected EYFP+ MNPs in re- mice were subjected to laser-induced CNV
J), a substantial difference in adipose tissue tinas of mice that received eWAT fat pads from 8 weeks after BMT, those that had received
macrophage (ATM) phenotype was observed. LysMCre/+:Ai3EYFP/+ donor mice. Transplants BMT from HFD-RD mice developed more
Crown-like structures, which form a syncytium from LysMCre/+ mice were used as controls to CNV than recipients of RD-RD BMT (Fig. 2,
of macrophages surrounding dead adipocytes determine background autofluorescence. Thus, P and Q, and fig. S4, I and J). This occurred
in obese mice (33), remained a prominent fea- it appears possible that myeloid cells from despite a lack of difference in the number of
ture of visceral adipose tissue in HFD-RD mice adipose tissue directly infiltrate CNV lesions MNPs recruited to the sites of injury (fig. S4K).
after weight loss (Fig. 2, A and E). Further, and locally contribute to disease progression Together, these data demonstrate that BM-
although the number of ATMs per gram of (Fig. 2, K and L). derived myeloid cells transferred from formerly
fat pad was similar between RD-RD mice and obese mice eventually repopulate adipose tis-
HFD-RD mice (Fig. 2, F and G, and fig. S3, C Bone marrow transfer from mice with past sue and retain a phenotype that drives CNV.
and K), flow cytometric analysis of the tissue obesity aggravates CNV in lean mice
revealed a threefold increase in proinflam- The myeloid compartment of the bone mar- Past obesity induces persistent epigenomic
matory ATMs [F4/80+ or EGF-like module- row (BM) is affected by obesity, so it is possible reprogramming of ATMs toward enhanced
containing mucin-like hormone receptor-like 1 that BM-derived myeloid cells from HFD mice angiogenic and inflammatory responses
(EMR1), CD11b+, CD38+] (34) in HFD-RD mice might also affect CNV. We observed a signif- We next characterized potential epigenetic
compared with RD-RD mice 3 days after icantly higher frequency of myeloid-biased changes in ATMs induced by HFD-driven obe-
laser burn (Fig. 2, H and I, and fig. S3L). multipotent progenitor cells (MPP3) in HFD sity. Transplanted adipose tissue or BM-derived
Even before laser injury, a greater number mice, a trend that was similar in HFD-RD mice myeloid cells retained the effects of HFD expo-
of ATMs were CD38+ in both HFD mice and compared with RD-RD controls (fig. S4, A to sure and exacerbated pathological neovascular-
HFD-RD mice than in RD-RD controls (fig. D). We assessed transcript levels of innate ization (Figs. 1 and 2), and ATMs from HFD
S3, M and N). immunity–related genes in monocytes iso- mice persistently remained in a proinflamma-
To further investigate how ATMs influence lated from the BM of RD-RD mice and HFD- tory state (F4/80+, CD11b+, CD38+) after animals
CNV, we performed ATT with fluorescently RD mice using real-time quantitative PCR were returned to a RD (Fig. 2). We therefore

Fig. 3. Prior obesity induces epigenomic reprogramming of ATMs toward versus HFD ATMs as found by ATAC-seq. (D) Venn diagram showing DARs
proinflammatory and proangiogenic phenotypes. (A) Experimental schematic identified in the comparisons RD-RD versus HFD-RD ATMs, RD-RD versus HFD
of the ATAC-seq of ATMs. ATMs derived from RD-RD, HFD-RD, and HFD mice ATMs, and HFD-RD versus HFD ATMs. (E) Heatmap of 2022 DARs (z-score
(HFD feeding for 11 weeks) were sorted by FACS and subjected to ATAC-seq. of normalized count) identified in the comparison RD-RD versus HFD ATMs,
(B) Principal component analysis of ATAC-seq normalized read counts in peaks which were shared in the comparison RD-RD versus HFD-RD ATMs but not in the
from HFD, RD-RD, and HFD-RD ATMs. (C) Volcano plots of accessible regions comparison HFD-RD versus HFD ATMs. (F) GO circle plot displaying gene
with DARs, defined by an adjusted P < 0.05 and a |log2(fold change)| > 1.0, annotation enrichment analysis in the comparison of RD-RD versus HFD-RD
between RD-RD versus HFD-RD ATMs, RD-RD versus HFD ATMs, and HFD-RD ATMs. The inner ring is a bar plot in which the height of the bar indicates the

significance of the term (−log10 adjusted P value) and the color corresponds to the in eWAT 48 hours after laser burn relative to RD-RD of Il1b (I), n = 18 for RD-RD,
z-score. The outer ring displays scatterplots of the expression levels (log fold n = 23 for HFD-RD; Il6 (J), n = 12 for RD-RD, n = 17 for HFD-RD; Il10 (K), n = 5
change) for the genes in each term. (G and H) Heatmaps of DARs (z-score of for RD-RD, n = 11 for HFD-RD; Tnf (L), n = 14 for RD-RD, n = 17 for HFD-RD; Cxcl1 (M),
normalized count) identified in the comparison of RD-RD versus HFD-RD ATMs with n = 11 for RD-RD, n = 12 for HFD-RD; Ccl5 (N), n = 14 for RD-RD, n = 17 for HFD-RD; and
their nearest genes in the gene sets GO angiogenesis (G) and GO inflammatory Ifng (O), n = 11 for RD-RD, n = 15 for HFD-RD Socs3 (P), n = 6. Data information:
response (H). AP-1 target genes are highlighted. If multiple DARs correspond to the Comparisons between groups were analyzed using a StudentÕs unpaired t test [(I) to
same gene, then the number is indicated behind the gene. (I to P) mRNA expression (P)]; *P < 0.05, **P < 0.01, ***P < 0.001; error bars represent mean ± SEM.
evaluated global chromatin accessibility using nuclear factor of kappa light polypeptide gene phospholipids of mice fed a HFD (time point 1,
an assay for transposase-accessible chromatin enhancer in B cells 1 (Nfkb1), Tnf alpha-induced Fig. 4A and fig. S7, A and B). Additionally, we
sequencing (ATAC-seq). We performed ATAC- protein 3 (Tnfaip3), vascular endothelial growth observed higher amounts of proinflammatory
seq on nuclei extracted from sorted ATMs factor A (Vegfa), angiopoietin 1 (Angpt1), platelet- arachidonic acid (C20:4) also bound to phos-
from HFD, RD-RD, and HFD-RD mice before derived growth factor receptor (Pdgfr), and pholipids from mice fed a HFD. By contrast, we
laser-induced CNV (Fig. 3A). Principal com- beta polypeptide (Pdgfrb) were more accessible, did not observe enrichment in plasma concen-
ponent analysis of open chromatin regions whereas Il10 was less accessible in HFD-RD trations of SA or any other analyzed FA in
showed that samples segregated in accor- ATMs (Fig. 3, G and H, and fig. S5, A and B). HFD-RD mice at time point 2 after return to
dance to their experimental group, with RD-RD Together, these data demonstrate that most RD and metabolic normalization (Fig. 4B and
being the most distinct and having the highest changes in the chromatin landscape induced fig. S7C), indicating that SA only accumulated
variability (Fig. 3B). by HFD feeding are maintained as open chro- in the phospholipid fraction during HFD feed-
Hence, we analyzed specific and common matin positions for at least 9 weeks, suggesting ing. Palmitic acid (C16:0) did not differ between
epigenetic changes between groups. Compar- that HFD feeding leads to long-term changes in the two groups in our experiments (Fig. 4A).
ison of the groups identified a total of 5898 ATMs and may render them prone to proangio- Saturated FFAs, including SA and palmitic
differentially accessible regions (DARs) defined genic and proinflammatory responses. acid, induce inflammation through Toll-like
using adjusted P < 0.05 and |log2(fold change)| > To determine whether the modifications receptor (TLR) 4-JNK or nuclear factor kB
1.0 (Fig. 3, C and D). Intergroup comparisons in the chromatin landscapes described above (NF-kB) signaling (36, 39). We therefore dif-
revealed the highest diversity, and thus the potentiated phenotypic changes in cytokine ferentiated BM cells from 8-week-old male
greatest number of DARs between ATMs from production, we next assessed transcript abun- C57BL/6J mice into mature macrophages and
RD-RD and HFD-RD mice (4989 DARs), and dance of innate immunity–related genes by stimulated them with SA or palmitic acid with
considerably fewer DARs were identified be- RT-qPCR. At 48 hours after laser injury, visceral or without TAK-242 (an inhibitor of TLR4 sig-
tween ATMs from HFD-RD versus HFD mice adipose tissue levels of Il1b, Tnf, chemokine naling) (fig. S8, A to F). After acute stimulation
(403 DARs) (Fig. 3D). A total of 2022 DARs were (C-X-C motif) ligand 1 (Cxcl1), chemokine (C-C for 6 hours, SA induced the expression of pro-
shared by both the RD-RD versus HFD mice, motif) ligand 5 (Ccl5), and interferon gamma inflammatory and proangiogenic genes such as
and the RD-RD versus HFD-RD mice, but (Ifng) rose significantly in HFD-RD mice com- Tnf, Tnfaip3, Il6, chemokine (C-C motif) ligand
not by HFD-RD versus HFD mice, indicating pared with RD-RD controls, supporting the idea 2 (Ccl2), Cxcl1, Nfkb, and Vegf in BM-derived
that these regions of chromatin in ATMs from that epigenetic changes in HFD-RD mice may macrophages (BMDMs) (fig. S8B). Pretreatment
HFD-RD mice remained in a conformation lead to biased expression of proinflammatory with TAK-242 abolished the induction of the
similar to the HFD state even after return genes (Fig. 3, I to P). Consistent with this, expres- Il1b, Tnf, Tnfaip3, Il6, Ccl5, and Cxcl1 genes (fig.
to RD (Fig. 3, D and E). Among these 2022 sion of Suppressor of Cytokine Signaling-3 (Socs3), S8B). Similarly, palmitic acid induced the ex-
DARs, 1535 (75.9%) were more accessible in which is a negative regulator of cytokine signal- pression of proinflammatory and proangiogenic
the HFD ATMs (11 weeks) and HFD-RD ATMs, ing, was down-regulated. Furthermore, serum genes in a TLR4-dependent manner (fig. S8E).
whereas 487 (24.1%) were more accessible in concentrations of proinflammatory cytokines Given that SA was the most abundant
the RD-RD ATMs. By contrast, only 50 DARs such as IL1b, IL2, IL12, TNF-a, and IFNg were phospholipid-bound FA during the obese phase
were commonly identified in RD-RD versus increased in HFD-RD. Anti-inflammatory IL10 of the WGWL experiments (Fig. 4A), we inves-
HFD, and HFD-RD versus HFD mice, but not also increased (fig. S6, A to G). Collectively, these tigated its potential to affect macrophage mem-
in RD-RD versus HFD-RD mice. This indi- data suggest that obesity leads to persistent pro- ory. We treated BMDMs with SA for 24 hours
cates that fewer regions of open chromatin inflammatory changes in visceral adipose tissue and then stimulated them with lipopolysac-
in HFD mice reverted back to a state seen in and predisposes to increased systemic inflam- charide (LPS) 5 days after the removal of SA.
RD once mice were returned to RD; most re- mation after experimentally induced injury. Compared with control macrophages pretreated
mained in a conformation similar to the HFD with bovine serum albumin (BSA), SA-pretreated
state. These data suggest that sustained Stearic acid potentiates macrophage memory macrophages had increased expression of
changes in the chromatin accessibility land- through activation of TLR4 signaling mRNAs encoding Tnf, Tnfaip3, Il6, and Nfkb
scape of ATMs from HFD mice persist long During obesity, several effectors such as free in response to secondary stimulation (Fig. 4,
after return to RD. fatty acids (FFAs), triglycerides, ceramides, gut- C and D). We observed heightened expression
Association of DARs with the nearest gene derived endotoxin, and damage-associated mo- of Il1b, Il6, and Tnf in SA-pretreated macro-
and gene ontology (GO) enrichment analysis lecular patterns (DAMPs) such as S100A8-A9 and phages in response to secondary stimulation
revealed considerable pathway enrichment in High Mobility Group Box1 (HMGB1) are thought after a wash period of 10 days (fig. S8C). The
the HFD-RD group compared with the RD-RD to activate ATMs and other cells of the innate elevated expression of Il6 persisted for at least
group (Fig. 3F). GO-listed genes involved in immune system (26, 35–38). To elucidate po- another 10 days, and Il1b and Tnf showed the
pathways coding for angiogenesis, cytokine tential lipids that might drive epigenetic changes same trends (fig. S8C). Pretreatment with TAK-
production, and inflammatory response were caused by HFD, we used quantitative fatty acid 242 decreased the effect of SA on macrophage
enriched in the HFD-RD group. In gene sets (FA) profiling by gas chromatography–mass memory (Fig. 4D and fig. S8C). Consistent with
from the GO terms “angiogenesis” and “inflam- spectrometry (GC-MS). Stearic acid (SA) (C18:0) a role for TLR4 in this process, SA-pretreated
matory response,” AP-1 target genes such as Il1b, was the most abundant FA bound to plasma macrophages subsequently stimulated with

Fig. 4. SA potentiates macrophages through activation of TLR4 signaling FAs in plasma phospholipids in RD-RD (n = 6 each time point) and HFD-RD
and induces sustained metabolic rewiring. (A and B) Concentration of (n = 6 each time point) mice at 19 weeks (time point 1, 11 weeks HFD

for HFD-RD mice) (A) and 28 weeks (time point 2, 9 weeks back on RD for collected. (H) Representative confocal images of IB4-stained laser burns with
HFD-RD mice) (B). (C) Experimental schematic of the in vitro SA-induced FITC-dextran–labeled CNV and IBA1-stained MNPs in RD-RD and HFD-RD Tlr4
immune memory model. BMDMs from C57BL/6J mice were pretreated with knockout mice. Scale bar, 50 μm. (I and J) Quantification of the FITC-dextran–
DMSO (control) or a TLR4-specific inhibitor (TAK-242) for 1 hour. The BMDMs labeled CNV area (I) and the ratio of FITC/IB4 per laser burn (J) relative to
were then stimulated with BSA (control) or SA. After 24 hours of stimulation, RD-RD Tlr4 knockout mice at week 30 (day 14 ); n = 24 burns for RD-RD,
BMDMs were cultured in basal medium for another 5 days before a secondary n = 23 burns for HFD-RD. (K) Number of IBA1+ MNPs around the laser impact
stimulation with LPS for 4 hours (memory phase). Total RNA was extracted and area relative to RD-RD Tlr4 knockout mice at day 14; n = 24 burns for RD-RD,
analyzed for gene expression by qPCR. Seahorse assay was performed on n = 23 burns for HFD-RD. (L) OCR of BMDMs from BSA- and SA-pretreated
BMDMs without TAK-242 pretreatment before secondary stimulation with LPS mice with or without LPS restimulation as indicated in (C); n = 11 (BSA + PBS
[(L) to (Q)]. (D) BMDM mRNA expression of Il1b, n = 6; Il6, n = 6; Tnf, n = 7; in vitro), n = 11 (SA + PBS in vitro), n = 9 (BSA + LPS in vitro), n = 12 (SA + LPS
Vegfa, n = 6; Tnfaip3, n = 8; Ccl2, n = 6; Ccl5, n = 6; Cxcl1, n = 8; and Nfkb, n = 10 in vitro). (M and N) Basal respiration (M) and maximal respiration (N) of
(BSA + LPS) and n = 9 (SA + LPS and TAK-242 + SA + LPS) as indicated in BMDMs from BSA- and SA-pretreated mice with or without LPS restimulation;
(C). (E) Experimental schematic of the in vitro SA-induced immune memory model n = 11 BSA + PBS in vitro, n = 11 SA + PBS in vitro, n = 9 BSA + LPS in vitro,
in Tlr4 knockout (Tlr4−/−) mice. BMDMs from Tlr4−/− mice were stimulated and n = 12 SA + LPS in vitro. (O and P) Glycolysis (O) and glycolytic capacity (P)
with BSA (control) or SA for 24 hours. After 5 days of washout, BMDMs were of BMDMs from BSA- and SA-pretreated mice with or without LPS restimulation;
restimulated with or without LPS for 4 hours before gene expression analysis by n = 11 BSA + PBS in vitro, n = 11 SA + PBS in vitro, n = 9 BSA + LPS in vitro,
qPCR. (F) BMDM mRNA expression of Il1b, Il6, Tnf, Tnfaip3, and Nfkb (n = 3 and n = 12 SA + LPS in vitro. (Q) Energy map of the four conditions tested charging
per condition) as indicated in (E). (G) Experimental schematic in which half ATP-linked respiration versus glycolytic capacity. Data information: Comparisons
of the Tlr4−/− mice were started on a HFD at 8 weeks of age and switched between groups were analyzed using a Student’s unpaired t test [(A), (B),
back to a RD at 19 weeks; this group is referred to as HFD-RD. Control mice were (I to K), (M), (N), (O), (P)], a multiple t test (F), or a one-way ANOVA with
fed a RD throughout and are referred to as RD-RD. Mice were subjected to Tukey’s multiple-comparisons test (D). *P < 0.05, **P < 0.01, ***P < 0.001,
laser-induced CNV at 28 weeks and euthanized at week 30, when eyes were ****P < 0.0001; error bars represent mean ± SEM.
HMGB1 (a TLR4 ligand that provokes a lesser innate immune memory by biasing metabo- Specifically, more accessible DARs in HFD-
inflammatory response than LPS) showed greater lism. We assessed mitochondrial respiration by RD ATMs were enriched for regions contain-
cytokine production than control BSA-pretreated measuring the oxygen consumption rate (OCR) ing consensus-binding sites for the AP-1 family
macrophages (fig. S8, G and H). By contrast, in BSA (control) or SA-pretreated BMDMs (at of transcription factors, which comprises sev-
palmitic acid–pretreated macrophages showed memory phase 5 days after treatment; Fig. 4C). eral members, including c-JUN, c-FOS, and
a lower-magnitude regulation of gene expres- Small but statistically significant decreases were ATF. The top eight highest-ranked motifs cor-
sion in response to subsequent stimulation noted in basal respiration and maximal respi- respond to binding of AP-1 family members
with LPS (fig. S8F). ration in SA-pretreated BMDMs both before and based on a Hypergeometric Optimization of
Consistent with the results of pharmaco- after restimulation with LPS (Fig. 4, L to N). Motif EnRichment (HOMER) motif search
logic inhibition of TLR4, SA-induced effects We also measured the extracellular acidifi- (Fig. 5, B and C). AP-1 binding itself leads to
on cytokine and NF-kΒ mRNA were abolished cation rate (ECAR) to assess glycolysis in BSA chromatin remodeling by recruiting histone-
in BMDMs derived from mice with a coding (control) or SA-pretreated BMDMs. Consis- modifying enzymes that trigger proinflam-
sequence deletion of Tlr4 causing loss of func- tent with a primed state for inflammatory re- matory genes (42, 43). With this in mind, we
tion (Tlr4Lps–del; hereafter referred to as Tlr4−/−) sponse (Fig. 4D), SA-treated BMDMs showed next tested whether c-JUN, the major compo-
(Fig. 4, E and F). Tlr4−/− BMDMs retained re- small but significant shifts in glycolysis and nent of AP-1 family, and histone acetyltrans-
sponse to LPS or SA, likely through other TLRs glycolytic capacity (Fig. 4, O and P) (41). Sub- ferase (HAT) are recruited to the promoter of
(fig. S9, A and B). Tlr4−/− mice fed RD-RD or HFD- sequent treatment with LPS caused a further the Tnf gene in cells exposed to SA (Fig. 5D). A
RD showed the same body weight change as increase in glycolysis and glycolytic capacity in 1-hour stimulation of BMDMs with SA in-
observed in WT C57BL/6J mice (fig. S1A; Fig. both groups (Fig. 4, O and P). We suggest that creased the phosphorylation of c-JUN (Fig. 5,
4G; and fig. S9, C and D). The systemic meta- the memory state of myeloid cells is asso- E to G). Phosphorylation of c-JUN promotes
bolic consequences (as shown by GTT and ITT) ciated with metabolic reprogramming and the expression of target genes by facilitat-
of DIO in Tlr4−/− mice were also similar to those that exposure to SA may shift myeloid cell ing its physical interaction with the histone
in C57BL/6J mice (fig. S9, E to H). However, metabolism toward glycolysis with less reli- acetyltransferase p300 (EP300) (44, 45). Chro-
Tlr4−/− mice did not retain a memory pheno- ance on oxidative metabolism (Fig. 4Q), an matin immunoprecipitation (ChIP)–qPCR assays
type, as described above (Figs. 1 and 2), and effect that persists long after weight loss. of SA-stimulated BMDMs revealed recruit-
had similar magnitudes of CNV- and FITC- ment of c-JUN to the promoter region of the
perfused areas as RD-RD and HFD-RD mice Obesity-driven reprogramming of macrophages Tnf gene in cells treated with SA (Fig. 5H). This
(Fig. 4, G to K, and fig. S9I). Thus, SA, the most correlates with chromatin remodeling was accompanied by recruitment of EP300
enriched FA in plasma phospholipids of HFD at AP-1Ðbinding sites to the promoter region of the Tnf gene and
mice, potentiates macrophages for future cyto- ATAC-seq from nuclei extracted from FACS- significant increases in acetylation of histone
kine production through TLR4 and leads to sorted ATMs of RD-RD or HFD-RD mice re- H3 on lysine 27 (H3K27ac), leading to higher
innate memory in DIO that aggravates path- vealed a landscape of open chromatin regions activation of transcription (46) (Fig. 5, I and J).
ological angiogenesis in response to experi- that are enriched in proximity of genes related However, despite a trend, statistically signifi-
mental injury. to the mitogen-activated protein kinase (MAPK)/ cant changes were not detected in the recruit-
c-Jun N-terminal kinase (JNK) or extracellular ment of c-JUN and EP300 to the promoter
Prior exposure to SA shifts macrophage signal–regulated kinase (ERK) signaling pathway region of the Il6 gene after stimulation with
metabolism toward glycolysis in HFD-RD ATMs (Fig. 5A). We thus hypoth- SA (Fig. 5, K to M). These data show chromatin
Given that cellular energy metabolism affects esize that genes encoding effectors of the MAPK remodeling at sites that exhibited AP-1 binding
myeloid cell phenotype and function (40), we signaling pathway might be epigenetically with SA as a potential mechanism of macro-
investigated whether exposure to SA provokes modulated in ATMs of formerly obese mice. phage reprogramming.

Fig. 5. Essential role of AP-1 in chromatin remodeling during obesity- signalingÐrelated pathways enriched in ATAC-seq data from DARs between
driven reprogramming of macrophages. (A) GO enrichment analysis of MAPK ATMs from RD-RD versus HFD-RD mice. (B) Top 10 enriched transcription factor

recognition sequences in ATAC-seq peaks of ATMs from the HFD-RD group in BMDMs stimulated with BSA or SA. (F and G) Quantification of phospho–c-
on the basis of HOMER. Transcription factors indicated with asterisks are JUN expression in BMDMs stimulated with BSA or SA (n = 3). (H to M) ChIP-
members of the AP-1 family. (C) Heatmaps showing density of mapped ATAC- qPCR showing the enrichment of cJun [(H) and (K)], EP300 [(I) and (L)],
seq reads from single biological replicates 1 kb up- and downstream of and H3K27ac [(J) and (M)] on the chromatin of proximal promoter regions of Tnf
transcriptional start sites (TSS) of the AP-1 motif. (D) Experimental schematic and Il6. Data are shown as the percentage of input DNA versus IgG control
in which BMDMs were stimulated with BSA or SA for 1 or 4 hours and harvested antibodies. Data information: Comparisons between groups were analyzed using
for immunoblotting or ChIP-qPCR analysis of the Tnf promoter, respectively. a Student’s unpaired t test [(F) to (M)]. *P < 0.05, **P < 0.01; error bars
(E) Representative immunoblots showing c-JUN and phospho–c-JUN expression represent mean ± SEM.
CNV is influenced by obesity-induced mice (fig. S11B). Fourteen days after laser in- of the innate immune response and influence
reprograming of ATMs and retinal myeloid cells jury, CNV was more pronounced in HFD-RD experimentally induced pathological events
Given that cells of myeloid origin retain changes LysMCre mice than in RD-RD mice. This pheno- in CNS tissue such as the retina.
associated with previous obesity and that a past type was not detected in LysMCre/+:R26iDTR/+
history of obesity influences CNV in the retina mice, suggesting a contribution of tissue-resident Loss of retinal function associated with light-
after injury, we sought to determine whether cells of myeloid origin such as ATMs in the pro- induced retinal degeneration in previously obese
acquired innate immune memory is found in longed effects of DIO (fig. S11, C to G). mice is prevented by depletion of myeloid cells
systemic myeloid cells, local retinal microglia, To confirm whether ATMs contribute to To determine whether obesity-induced myeloid
or both. We selectively eliminated distinct im- CNV after laser injury, we transplanted eWAT memory could influence visual function asso-
mune cell populations through diphtheria fat pads from RD-RD or HFD-RD LysMCre/+: ciated with AMD, we used a non-neovascular
toxin (DT)–mediated apoptosis. We assessed R26iDTR/+ or LysMCre/+ mice into C57BL/6J remodel to evaluate photoreceptor damage and
the efficiency and specificity of two different cipient mice. Mice were then subjected to laser- retinal function in mice (48) (fig. S12A). Ex-
myeloid-specific promoters, Lysm and Cx3cr1, induced CNV with targeted ablation of ATMs cessive light exposure in the experimental
in transgenic mice expressing fluorescent pro- by DT injections (Fig. 6A). Fourteen days after model is associated with accumulation of mye-
teins. ATMs were investigated as tissue-resident laser injury, CNV was more pronounced in loid cells in the subretina and choroid (fig.
cells of myeloid origin that can influence recipients of fat pads from HFD-RD LysMCre S12, B to E) and photoreceptor damage (fig.
obesity-driven phenotypes. The Lysm pro- control mice than in recipients of fat pads S12, F to H), leading to retinal dysfunction
moter drove expression preferentially in ATMs from RD-RD LysMCre mice. Selective elim- (fig. S12, I to K).
compared with the Cx3cr1 promoter (78.9 versus ination of myeloid cells from adipose tissue After 20 weeks of diet (time point 2), RD-RD
30.4% of total ATMs; fig. S10, A and B). By con- after ATT from LysMCre/+:R26iDTR/+ mice and mice and HFD-RD mice were subjected to
trast, in naïve retina and RPE-choroid-sclera treatment with DT abrogated the increased blue light-emitting diode (LED) light exposure
complexes, the Cx3cr1 promoter drove expres- CNV in recipients of fat from HFD-RD mice (1500 lux for 5 days) (Fig. 6I). In vivo live im-
sion in 83.8% of retinal MNPs versus 65.9% (Fig. 6, B to D, and fig. S11, H and I). We sug- aging by optical coherence tomography (OCT)
with Lysm; 3 days after laser injury, this shifted gest that ATMs may retain immune memory after the blue light exposure showed gradual
to 69.4 versus 37.5%, respectively (fig. S10, C of past obesity and contribute to experimen- thinning of the outer retina (photoreceptor
and D). Lysm and Cx3cr1 promoters drove ex- tal CNV development. layer) in both RD-RD and HFD-RD mice. Five
pression at similar levels in blood monocytes We then evaluated the contribution of local days after light exposure, mice in the HFD-RD
(69.7 versus 75.1%, respectively; fig. S10, E to G). myeloid cells to the prolonged effects of obesity- group had significantly thinner photoreceptor
To deplete myeloid cells in vivo, we gener- induced immunity. We generated compound layers at each point of measurement (day 10)
ated compound heterozygous mice carrying heterozygous mice carrying the Cx3cr1CreER (Fig. 6, J and K, and fig. S12, L and M). Consis-
the LysmCre allele and the R26iDTR allele allele and the R26iDTR allele (Cx3cr1CreER/+: tent with our data from the laser-induced CNV
(LysMCre/+:R26iDTR/+). The resulting mice ex- R26iDTR/+). After intravitreal DT administration, model (Fig. 1L), the number of accumulated
pressed DT receptors specifically in myeloid ablation of local myeloid cells was verified by IBA1+ cells in the retina and choroid did not
cells, rendering them susceptible to targeted flow cytometry. We confirmed that MNPs, in differ between the two groups (Fig. 6, L and M,
elimination. We investigated depletion of ATMs particular CX3CR1+ cells, were decreased in and fig. S12, N to Q). Amounts of transcripts of
for proof of concept of targeted elimination of the retina and RPE-choroid-sclera complexes proinflammatory genes such as Il1b and Tnf
tissue-resident myeloid cells. Successful ablation of Cx3cr1CreER/+:R26iDTR/+ mice (fig. S11, J to L). showed a small but significant rise in retina-
of ATMs by DT was verified by flow cytometry. RD-RD and HFD-RD mice were intraperito- RPE-choroid-sclera complexes in HFD-RD mice
We observed reduction of ATMs within visceral neally injected with tamoxifen for 3 consecutive compared with control RD-RD mice (Fig. 6N).
adipose tissue in LysMCre/+:R26iDTR/+ mice (fig. days, and DT was administered intravitreally Consistent with these results, we observed sig-
S10, H and I). ATMs in control LysMCre mice 4 and 5 weeks afterwards (Fig. 6E). Fourteen nificantly greater retinal dysfunction (decreased
devoid of the R26iDTR allele were unaffected. days after laser injury, Cx3cr1CreER/+ control amplitude of a-wave and b-wave) in HFD-RD mice
No significant reduction was noted in circu- mice fed HFD-RD showed increased size of compared with RD-RD mice (Fig. 6, O to Q).
lating monocytes of LysMCre/+:R26iDTR/+ mice CNV area compared with RD-RD mice. This To determine whether the observed accen-
3 days after DT injections, suggesting this popu- was in contrast to Cx3cr1CreER/+:R26iDTR/+ mice, tuated retinal dysfunction was provoked by
lation was replenished at the time of analysis in which no change was observed (Fig. 6, F to H, sustained effects of past HFD on MNPs, we
(fig. S10, J and K) (47). and fig. S11, M and N). Overall, these data reveal depleted myeloid cells in LysMCre/+:R26iDTR/+
Both LysMCre/+:R26iDTR/+ and control LysMCre that myeloid cells in both tissues distal to the mice. Mice were put on the RD-RD or HFD-
mice were put on RD-RD or HFD-RD and then retina (i.e., ATMs) and local retinal myeloid cells RD regime and subjected to blue light expo-
subjected to laser-induced CNV (fig. S11A). (i.e., tissue-resident microglia and infiltrating sure with intraperitoneal injections of DT (Fig.
Forty-eight hours after laser injury, we did not monocytes and macrophages) appear to contrib- 6R). Retinal dysfunction was more pronounced
observe any changes in the serum concentra- ute to DIO-mediated changes in immunity. More in HFD-RD LysMCre/+ mice, but this phenotype
tion of proinflammatory cytokines between broadly, these data suggest that a past history of was completely abrogated in LysMCre/+:R26iDTR/+
either diet paradigm in LysMCre/+:R26iDTR/+ obesity can reprogram local and distal effectors mice (Fig. 6, S to U). We propose that myeloid

Fig. 6. Depletion of adipose tissue or retinal myeloid cells reverses a were transplanted with eWAT fat pads at 8 weeks of age from either LysMCre/+
proinflammatory and proangiogenic phenotype in formerly obese mice or LysMCre/+:R26iDTR/+ donor mice fed either RD-RD or HFD-RD (as shown
and restores vision loss associated with retinal degeneration after light in fig. S13A). DT was administered intraperitoneally for 3 consecutive days (at
exposure. (A) Experimental schematic of ATT in which recipient C57BL/6J mice week 10), followed by one additional injection at week 12. At 11 weeks of age,

mice were subjected to laser-induced CNV and then euthanized at week 13 for light exposure (baseline), just after light exposure (day 5), and 5 days after
analysis of eyes. (B) Representative confocal images showing IB4-stained light exposure (day 10). Mice were euthanized and eyes were collected for flatmount
laser burns with FITC-dextran–labeled CNV and IBA1-stained MNPs in LysMCre/+ analysis 5 days after light exposure (day 10). (J) Representative SD-OCT
ATT and LysMCre/+:R26iDTR/+ ATT mice after either RD-RD or HFD-RD. Scale bar, images of RD-RD and HFD-RD mice before (baseline) and after blue LED light
50 μm. (C and D) Quantification of the FITC-dextran–labeled CNV area (C) exposure (days 5 and 10). (K) Outer retinal thickness (ORT) at increasing distances
and the ratio of FITC/IB4 per laser burn (D) relative to RD-RD mice at day 14; (–1750 μm: nasal, +1750 mm: temporal) from the center of the optic disk (0 μm)
n = 27 burns for RD-RD LysMCre/+, n = 30 burns for HFD-RD LysMCre/+, in 250-μm steps in RD-RD and HFD-RD mice 5 days after exposure (day 10);
n = 25 burns for RD-RD LysMCre/+:R26iDTR/+, and n = 31 burns for HFD-RD n = 10 in each group. (L) Representative images of IBA1-stained choroidal
LysMCre/+:R26iDTR/+. (E) Experimental schematic in which half of the flatmounts at day 10. Scale bar, 500 μm. (M) Quantification of subretinal and
Cx3cr1CreER/+ and Cx3cr1CreER/+:R26iDTR/+ mice started a HFD at 8 weeks of choroidal IBA1+ cells/mm2 of RD-RD (n = 7) and HFD-RD (n = 8) mice. (N) mRNA
age and were then switched back to a RD at 19 weeks. This experimental group expression in retina-RPE-choroid-sclera complexes at day 10 relative to RD-RD of
is referred to as HFD-RD. Control mice were fed a RD throughout and are Il1b, n = 11 for RD-RD and n = 9 for HFD-RD; Il6, n = 11 for RD-RD and n = 10 for
referred to as RD-RD. Tamoxifen (TAM) was administered intraperitoneally for HFD-RD; and Tnf, n = 11 for RD-RD and n = 10 for HFD-RD. (O to Q) Scotopic
3 consecutive days starting at week 24, and DT was administered intravitreally at electroretinography before (baseline) and 3 days after light exposure (day 8).
weeks 28 and 29. At 28 weeks of age, mice were subjected to laser-induced Representative images are shown in (O). (P and Q) Amplitudes of the a-waves
CNV and were then euthanized at week 30 and eyes collected. (F) Representative (P) and b-waves (Q) in RD-RD and HFD-RD mice; n = 12 per each group.
confocal images showing IB4-stained laser burns with FITC-dextran–labeled (R) Experimental schematic of LysMCre/+ and LysMCre/+:R26iDTR/+ mice fed
CNV and IBA1-stained MNPs in Cx3cr1CreER/+ and Cx3cr1CreER/+:R26iDTR/+ mice either RD-RD or HFD-RD and subjected to blue LED light exposure (1500 lux for
after either RD-RD or HFD-RD feeding. Scale bar, 50 μm. (G and H) Quan- 5 days) at 28 weeks of age. DT was administered intraperitoneally for 3
tification of area of FITC-dextran–labeled CNV (G) and the ratio of FITC/IB4 per consecutive days at week 27. ERG was performed 3 days after (day 8) light
laser burn (H) relative to RD-RD Cx3cr1CreER/+ mice at day 14; n = 23 burns exposure. (S to U) Scotopic electroretinography of LysMCre/+ or LysMCre/+:
for RD-RD Cx3cr1CreER/+, n = 19 burns for HFD-RD Cx3cr1CreER/+, n = 27 burns R26iDTR/+ mice after receiving either RD-RD or HFD-RD. Representative images
for RD-RD Cx3cr1CreER/+:R26iDTR/+, and n = 23 burns for HFD-RD Cx3cr1CreER/+: are shown in (S). Amplitudes of the a-waves (T) and b-waves (U); n = 8 for
R26iDTR/+. (I) Experimental schematic of the blue LED light exposure AMD RD-RD LysMCre/+, n = 7 for HFD-RD LysMCre/+, n = 8 for RD-RD LysMCre/+:
model. After 20 weeks of diet either RD-RD or HFD-RD, C57BL/6J mice were R26iDTR/+, and n = 7 for HFD-RD LysMCre/+:R26iDTR/+). Data information:
subjected to blue LED light exposure (1500 lux for 5 days) after dark adaptation Comparisons between groups were analyzed using a Student’s unpaired t test
overnight at 28 weeks of age. ERG was performed 3 days before (day –3) [(C), (D), (G), (H), (M), (N), (P), (Q), (T), and (U)] or a multiple t test (K); *P <
and 3 days after (day 8) light exposure. SD-OCT was performed 1 day before 0.05, **P < 0.01, ****P < 0.0001; error bars represent mean ± SEM.
cells retain the effects of prior DIO and can and induce innate immune memory in the BM SA led to a persistent hyperresponsiveness state
exacerbate the neuroinflammation and ret- (51). Although we found absolute numbers of in macrophages. LPS (the microbial ligand for
inal damage associated with vision loss in mod- immune cells unaffected in formerly obese mice, TLR4) also has the ability to trigger a long-
els of AMD. we saw sustained proinflammatory phenotypes lasting hyperresponsive state in myeloid cells
in ATMs. ATAC-seq analysis revealed preferen- (55), suggesting a conserved mechanism through
DISCUSSION tial chromatin decondensation at binding sites TLR4 for initiating innate memory. Persistent
Although polymorphisms in several genes re- for AP-1, with enrichment in accessibility re- epigenetic reprogramming of long-lived cells of
lated to the immune and inflammatory path- gions containing consensus binding sites for myeloid lineage by past events associated with
ways are associated with the progression of several AP-1 family members such as c-JUN, lipid overload identifies potential nodes that
AMD (49), genetics can only partially explain c-FOS, and ATF. AP-1 binding is reported to can be targeted to mitigate neuroinflammation
disease incidence. We propose that in obesity, lead to chromatin remodeling by recruitment in later life.
sustained systemic exposure to increased con- of histone modification enzyme to a promoter In summary, we show in mice that a past
centrations of lipids such as SA may lead to of proinflammatory genes (43). In agreement history of obesity has the propensity to induce
long-term changes in innate immunity and with this, we observed HAT EP300 bound to the long-term chromatin remodeling in tissue-
render distinct cells of myeloid lineage, such promoter region of the Tnf gene in SA-stimulated resident myeloid cells such as ATMs and that
as ATMs and retinal microglia, susceptible to macrophages. Moreover, histone acetylation at this ultimately influences neuroinflammation
triggering heightened proangiogenic and pro- lysine 27 of histone 3 (H3K27ac) was enriched after experimental injury in distal tissues such
inflammatory responses. These cells may thus in the promoter region of the Tnf gene, and as the retina. These sustained changes in the
be primed to exacerbate CNV, even once nor- AP-1 target genes such as Vegfa, Il1b, Tnfaip2, chromatin landscape of innate immune cells
mal weight and metabolic status have been and Cxcl1 were in the open conformation, exacerbate pathological angiogenesis and neu-
reestablished. Targeted elimination of either which is consistent with epigenetic modifica- ronal degeneration in experimental models,
cell population (ATM or retinal microglia) at- tion leading to a primed innate immune system. suggesting a link between systemic innate
tenuated experimentally induced CNV in pre- The observed changes in the innate immune immune training and retinal disease. In ad-
viously obese mice, consistent with a contribution cells of formerly obese mice can be recapitulated dition, future therapeutic avenues to influence
of both local and peripheral reprogrammed with exposure to selected FAs. Elevated circulat- epigenetic reprogramming of the innate im-
myeloid cells to neovascular AMD. ing saturated FAs can provoke inflammation mune system or to eliminate subpopulations
The concept of immune memory, conven- in obesity after HFD (52). SA and palmitic acid of reprogrammed innate immune cells may
tionally thought to occur uniquely with adap- activate TLR4 signaling pathways and induce delay or prevent the onset of both neovascular
tive immunity, has been extended to include the production of inflammatory mediators such AMD and non-neovascular AMD.
innate immunity with the demonstration that as TNF-a, IL6, NF-kB, and COX2 in macro-
monocytes and macrophages undergo stable phages (36, 39, 53, 54). We found that both of MATERIALS AND METHODS
epigenetic and metabolic reprogramming (50). these saturated FAs increased the expression Reagents
Consistent with our study, recent work suggests of proinflammatory and proangiogenic genes See table S3 for a detailed description of all
that Western diets trigger sterile inflammation in a TLR4-dependent manner; however, only reagents.
SCIENCE science.org 6 JANUARY 2023 ¥ VOL 379 ISSUE 6627 57

Animal studies neal injections with DT for 3 consecutive days 30 min. The RPE-choroid-sclera complex was
Mice were housed in the Hospital Maisonneuve- at 27 weeks of age. dissected and separated from the neuroretina.
Rosemont Research Center animal facility in a Tissue was incubated for 1 hour in a blocking
12-hour light-dark cycle and with ad libitum Primary BM-derived macrophages solution (3% BSA + 0.3% Triton X-100), fol-
access to food and water unless otherwise in- Donor mice (8-week-old male C57BL/6J or Tlr4−/−) lowed by an overnight incubation with anti-
dicated. All animals used in this study were were euthanized and leg bones were dissected. bodies. Blood vessels were investigated using
males. The experimental procedures were ini- Femur and tibia cavities were flushed with rhodamine-labeled Griffonia (bandeiraea)
tiated when mice were 8 weeks of age unless phosphate-buffered saline (PBS) supplemented Simplicifolia Isolectin I (1:100), and MNPs
otherwise indicated. with 10% fetal bovine serum (FBS) using a sy- were detected with IBA1 (1:350). After several
C57BL/6J and B6.B10ScN-Tlr4lps–del/JthJ ringe, resuspended, and passed through a 70-mm washes with PBS, RPE-choroid-sclera complexes
(referred to as Tlr4−/−) mice were purchased strainer. Red blood cells (RBCs) were removed were incubated with secondary fluorochrome-
from The Jackson Laboratory (Bar Harbor, ME, using RBC lysis buffer (catalog no. 00-4333-57; conjugated species-appropriate antibodies for
USA) and bred at the Hospital Maisonneuve- eBioscience). After centrifugation, BM cells were 1 hour. The tissue was then mounted onto a glass
Rosemont Research Center animal facility. recovered and cultivated in Dulbecco’s mod- slide and imaged using an Olympus FluoView
Homozygous B6.129P2(C)-Cx3cr1tm2.1(cre/ERT2)Jung/ ified Eagle’s medium (DMEM) supplemented FV1000 laser scanning confocal inverted mi-
J (referred to as Cx3cr1CreER) mice and homo- with 10% FBS and Penicilin-Streptomycin 1% croscope (Olympus Canada, Richmond Hill,
zygous B6.129P2-Lyz2tm1(cre)Ifo/J (referred to as (100 U/ml). Macrophage colony–stimulating fac- ON). For analysis, the Z-stacks were compressed
LysMCre) mice were crossed in-house with homo- tor (M-CSF) (20 ng/ml, catalog no. PMC2044; into one image. The area of neovascularization
zygous C57BL/6-Gt(ROSA)26Sortm1(HBEGF)Awai/J Invitrogen) was used to generate in vitro BMDMs (FITC-dextran+) and the burn area (isolectin+),
(referred to as R26iDTR) mice to obtain hetero- from BM progenitor cells. After 3 days of incu- as well as the number of MNPs (IBA1+) were
zygous Cx3cr1CreER/+:R26 iDTR/+ and LysMCre/+: bation at 37°C with 5% CO2, medium containing quantified using ImageJ software (version 1.0;
R26 iDTR/+ mice, respectively. Homozygous B6.129P- M-CSF was refreshed. Cells were allowed to National Institutes of Health, Bethesda, MD,
Cx3cr1tm1Litt/J (referred to as Cx3cr1GFP) mice differentiate for a total of 6 days before their USA). The quantification of the number of IBA1+
were crossed in-house with C57BL/6J mice to medium was replaced by complete medium cells was performed blinded to diet group.
obtain heterozygous Cx3cr1GFP/+ mice. Homo- without M-CSF. Macrophage purity was ~99%
zygous B6.Cg-Gt(ROSA)26Sortm3(CAG–EYFP)Hze/J as evaluated by flow cytometry. Blue LED light exposure AMD model
(referred to as Ai3EYFP) mice were crossed in- Mice were dark adapted overnight, and then
house with homozygous LysMCre mice to ob- METHOD DETAILS pupils were dilated using atropine sulfate
tain heterozygus LysMCre/+:Ai3EYFP/+ mice. Diet paradigm ophthalmic ointment before exposure to light.
All animal studies were performed in com- The fat component of the HFD, lard 245 g and Mice were exposed to blue light from an LED
pliance with the Animal Research: Reporting soybean oil USP 25 g/773.85 g, was the same as (wavelength 450 nm) at a light intensity of
of In Vivo Experiments (ARRIVE) guidelines that for the RD, lard 20 g and soybean oil USP 1500 lux for 5 days and then returned to regular
and the Association for Research in Vision 25 g/1055.5 g. Eight-week-old mice were fed conditions under a standard 12-hour light-dark
and Ophthalmology (ARVO) Statement for with either HFD (60% kcal fat) or RD (10% cycle until sacrifice on day 5 after illumination.
the Use of Animals in Ophthalmic and Vision kcal fat) to study weight gain experimentally,
Research and were approved by the Animal as outlined in the text, and mouse weight was Spectral domainÐOCT image acquisition
Care Committee of the Hospital Maisonneuve- monitored regularly. The time course of the diet and measurement of retinal thickness
Rosemont Research Center in agreement with feeding is indicated when appropriate. Upon Spectral domain–OCT (SD-OCT) examinations
the guidelines established by the Canadian sacrifice, the eWAT fat pads were weighed. using Multiline OCT (based on a Spectralis OCT,
Council on Animal Care. Heidelberg Engineering) were performed in
Laser-induced CNV blue LED light exposure AMD model at base-
Myeloid cell depletion Animals were anesthetized using a mixture of line, just after exposure (day 5), and 5 days
Microglia depletion was performed using 10% ketamine and 4% xylazine (10 ml/g of body after exposure (day 10). Volume scans were
Cx3cr1CreER/+:R26 iDTR/+ or LysMCre/+:R26 iDTR/+ weight) injected intraperitoneally, and pupils performed with the optic nerve head centered,
mice. In Cx3cr1CreER/+:R26 iDTR/+ mice, the ac- were dilated with mydriacyl 0.5%. Using an and the horizontal scan images, the center of
tivation of Cre recombinase (under the control argon laser, we ruptured their Bruch’s mem- which were adjusted to the center of the optic
of the Cx3cr1 promoter) can be induced by brane to induce CNV as described previously nerve head, were used for analysis. The soft-
tamoxifen treatment and leads to surface ex- (27). Each eye received four distinct laser burns ware used for measurement of retinal thick-
pression of DTR on CX3CR1-expressing cells. (400 mW, 100 mm, 0.05 s) distributed equidis- ness was based on the built-in Spectralis OCT
At 24 weeks of age, mice were subjected to daily tantly and following the optic nerve head as and provided by Heidelberg Engineering soft-
intraperitoneal injections with tamoxifen di- a central reference. Disruption of the Bruch’s ware to facilitate manual assessment of the
luted in corn oil (4 mg per mouse per day, stock membrane was verified through observation B-scan images. The distance between the in-
solution at 20 mg/ml) for 3 consecutive days. of a visible heat bubble at the site of injury. ner border of the outer plexiform layer and
To deplete CX3CR1+ cells in Cx3cr1CreER/+: the inner border of the retinal pigment epi-
R26 iDTR/+ mice, DT was administered intra- CNV evaluation thelium was calculated as outer retinal thick-
vitreally (25 ng/ml saline per eye) at 28 and Fourteen days after CNV induction, mice were ness (ORT). ORT and total retinal thickness
29 weeks of age. Depletion of LysM+ cells in anesthetized with isoflurane gas and intracar- were assessed at distances of 750, 1000, 1250,
LysMCre/+:R26 iDTR/+ mice was achieved by intra- dially perfused with 0.5 ml of FITC-dextran 1500, and 1750 mm from the optic nerve head
peritoneal injections with DT (100 ng per mouse (average molecular weight, 2,000,000). FITC- (nasally and temporally).
per day) for 3 consecutive days at 27 weeks dextran was left in circulation for 5 min while
of age, followed by one additional injection the animals were under anesthesia; subse- MNP quantification in choroidal and retinal
at 29 weeks of age. For the blue LED light quently, the mice were sacrificed and the eye flatmounts of mice exposed to blue LED light
model, depletion of LysM+ cells in LysMCre/+: globes were enucleated and fixed in paraform- Eyes were enucleated and fixed in 4% PFA
R26 iDTR/+ mice was achieved by intraperito- aldehyde (PFA) 4% at room temperature for at room temperature for 60 min. Dissected
58 6 JANUARY 2023 • VOL 379 ISSUE 6627 science.org SCIENCE

retinas and RPE-choroid-sclera complexes were and triglycerides was evaluated using GC-MS 5% FA-free BSA or 200 mM SA for 6 hours.
incubated with anti-IBA1 (1:350) followed by as described previously (56, 57). In brief, total Cells were then fixed with 1% formaldehyde
secondary antibody anti-rabbit Alexa Fluor 488. lipids were extracted with a mixture of methyl- for 8 min to cross-link chromatin. Chroma-
Choroids and retinas were flatmounted and tert-butyl ether, methanol, and water. Phos- tin samples were sheared by 10 cycles of 30 s
imaged using Zen Blue edition 3.2 software on pholipids and triglycerides were eluted on an ON and 30 s OFF with the Bioruptor Pico
an AxioImager Z2 microscope (Zeiss, Jena, aminopropyl column (Bond Elut LRC-NH2, sonicator (Diagenode), and 1% of samples de-
Germany). Choroids and retina flatmounts are 500 mg; Agilent Technologies). FFAs were an- rived from 1 × 106 cells were kept for input
large and thick samples. The full images were alyzed as methyl esters after a direct trans- measurements. Chromatin was immunopre-
obtained using mosaic and stitching features esterification with acetyl chloride/methanol on cipitated using antibodies to c-Jun (catalog no.
of the software. Z-stacks with 3-mm steps were a 7890B gas chromatograph coupled to a 5977A 9165; Cell Signaling Technology), P300 (cata-
optically sectioned using Apotome.2 with three mass selective detector (Agilent Technologies) log no. MAI-16622; Invitrogen), or histone 3K27ac
phase images and processed using normaliza- equipped with a capillary column (J&W Select (H3K27ac, catalog no. ab4729; Abcam) or using
tion, local bleaching correction, and strong FAME CP7420; 100 × 250 m inner diameter; 1 mg of rabbit immunoglobulin G (IgG) from
Fourier filter features provided by the software. Agilent Technologies) and operated in the posi- the kit as a negative control. Cross-links were
The number of MNPs (IBA1+) were counted on tive chemical ionization mode using ammonia then reversed, and the purified DNA was am-
whole RPE/choroidal flatmounts and on the as the reagent gas. Samples (2 ml) were analyzed plified by qPCR using iTaq Universal SYBR
outer segment side of the retina, and their under the following conditions: injection at Green Supermix (catalog no. 1725124; Bio Rad
density was determined using ImageJ software. 270°C in a split mode (split ratio: 50:1) using Laboratories) in an Applied Biosystems Real-
high-purity helium as the carrier gas (constant Time PCR machine (Thermo Fisher Scientific,
Electroretinography flow rate: 0.44 ml/min) and the following tem- Waltham, MA, USA). Primers for qPCR were
The retinal function of the blue LED light ex- perature gradient: 190°C for 25 min, increased previously reported (58) and are listed in table
posure AMD model mice was investigated using by 1.5°C/min until 236°C. The FAs were an- S2. Transcription factor–binding sites in the Tnf
electroretinography (ERG). ERG measurements alyzed as their [M + NH3]+ ion by selective ion and Il6 regulatory region between –227 and
were performed after an overnight dark adap- monitoring, and concentrations were calcu- –11 base pairs (bp) and between –262 and –448 bp
tation. Before measurements, mice were an- lated using standard curves and isotope-labeled were selected as described previously (59, 60).
esthetized with 10% ketamine and 4% xylazine internal standards. Total FA concentrations in
(10 ml/g of body weight) injected intraperitoneally. plasma phospholipids and triglycerides were Sample preparations for flow cytometric analysis
The pupils were then dilated with cyclopentolate calculated as the sum of each FA measured. Retinas and RPE-choroid-sclera complexes
hydrochloride 1%. Proparacaine hydrochloride The percentage of phospholipid or triglyceride were cut into small pieces and homogenized
0.5% was used to anesthetize the eye. Animals FA was calculated as the concentration of each in a solution of 750 U/ml DNAse I (catalog no.
were placed on a heating pad (Harvard Appa- FA over total FAs, multiplied by 100, measured D4527; Sigma-Aldrich) and 0.5 mg/ml colla-
ratus, Holliston, MA) during the entire record- in each fraction. genase D (catalog no. 11088866001; Roche) for
ing session to maintain their body temperature 20 min at 37°C. Homogenates were filtered
at 37°C. All manipulations were performed Transcription analysis by RT-qPCR through a 70-mm cell strainer.
under dim red light. ERG was recorded in the Mouse tissue from in vivo experiments or BM- eWAT fat pads were freshly dissected, homo-
left eye by placing an electrode (DTL Plus; Di- derived macrophages from in vitro assays were genized using scissors, and incubated in a solu-
agnosys LLC) in contact with the cornea, a ref- harvested and snap frozen immediately upon tion containing 1 mg/ml collagenase type II
erence electrode was placed on the tongue, collection. RNA extraction was performed using (catalog no. C6885; Sigma-Aldrich) for 45 min
and a neutral electrode was inserted in the TRIzol reagent (catalog no. 15596026; Invitrogen) at 37°C. EDTA (10 mM) was used to stop the
tail. Mice were placed under the ERG dome and digested with DNase I (catalog no. D4527; digestion, and the samples were incubated for
(Diagnosys LLC color dome model D125), and Sigma-Aldrich) following the manufacturer’s an additional 5 min at room temperature. The
scotopic ERGs were recorded with five flashes instructions to avoid genomic DNA amplifi- cell suspension was then filtered through a
of 25 (P)cd.s/m2. Results were analyzed with cation. Reverse transcription was performed 100-mm strainer and centrifuged at 500g for
Diagnosys version 6.63 software. The ampli- using a 5X All-In-One RT MasterMix (catalog 10 min at 4°C to separate the mature adipo-
tude of the a-wave was measured from base- no. G590; Applied Biological Materials Inc.), cytes from the stromal vascular fraction (SVF).
line to trough and the b-wave from the trough of and gene expression was analyzed using Bright Supernatants were discarded, and the SVF-
the a-wave to the highest peak of the b-wave. Green2X qPCR Master Mix-Low Rox in an Ap- containing pellet was resuspended with RBC
plied Biosystems 7500 Real-Time PCR System lysis buffer (catalog no. 00-4333-57; eBioscience)
Serum cytokine profiling (Thermo Fisher Scientific, Waltham, MA, USA). to remove RBCs.
Blood was obtained by cardiac puncture, and Primer sequences used in this study are listed Blood was collected from the tail vein by
serum was collected after centrifugation and in table S1. Analysis of expression was fol- bleeding after distal cut with a 21-gauge nee-
preserved at –80°C until analysis. The concen- lowed using the DDCT method. Actb expression dle, and BM cells were obtained by flushing
trations of IL1b, IL2, IL6, IL10, IL12, IFNg, and was used as the reference housekeeping gene. both tibias and femurs. RBC lysis buffer (cata-
TNF-a were measured using the Bio-Plex Pro Statistical analysis was performed on DDCT log no. 00-4333-57; eBioscience) was used to
Mouse Cytokine Th1 Panel according to the man- values, and the data are presented as the ex- remove RBCs.
ufacturer’s instructions (catalog no. L6000004C6; pression of the target genes normalized to
Bio-Rad). Cytokine levels are expressed as total Actb (fold increase). Flow cytometry staining and analysis
pg/ml. Viability of the cells was assessed by staining
ChIP-qPCR with 7-amino-actinomycin D (20 min at room
Plasma phospholipid and triglyceride FA profiling ChIP-qPCR experiments were performed using temperature) or Zombie Aqua (15 min at room
Blood was collected by cardiac puncture, and the iDeal ChIP-seq kit for histones (catalog no. temperature), and samples were incubated
plasma was obtained after centrifugation and C01010051; Diagenode, Liège, Belgium) follow- with LEAF-purified anti-mouse CD16/32 (cat-
preserved at –80°C until analysis. Quantitative ing the manufacturer’s protocol. Specifically, 5 × alog no. 101310; BioLegend) for 10 min at 4°C
profiling of FAs bound to plasma phospholipids 106 BM-derived macrophages were treated with to block FC receptors. Cells were then incubated

for 25 min at 4°C with the following antibodies: Immunohistological evaluation of adipose tissue BMT
for retinas and RPE-choroid-sclera complexes, eWAT depots were dissected and fixed with For the generation of chimeric mice, BM cells
BV711 anti-CD11b (catalog no. 101242; BioLegend), 10% formalin overnight. Tissues were dehy- were obtained by flushing both tibias, femurs,
PE anti-CX3CR1 (catalog no. FAB5825P; R&D drated, embedded in paraffin, and posteriorly and iliac crests of 28-week-old RD-RD and HFD-
Systems), APC anti-CD45.2 (catalog no. 109814; cut in 12-mm-thick sections following stan- RD C57BL/6J (CD45.2) donor mice. Eight-week-
BioLegend), APC/Cy7 anti-Ly-6G (catalog no. dardized histological procedures. eWAT sec- old B6.SJL (CD45.1) recipient mice were ran-
127624; BioLegend), and PE/Cy7 anti-F4/80 (cat- tions were then deparaffinized, rehydrated domly assigned to the different groups inde-
alog no. 123114; BioLegend); for ATMs, BV711 with decreasing concentrations of ethanol, pendently of the origin of the donor BM cells.
anti-CD11b (catalog no. 101242; BioLegend), PE and stained with Harris H&E. Sections were Recipient mice were lethally irradiated (12 Gy)
anti-F4/80 (catalog no. 123110; BioLegend), APC subsequently dehydrated and mounted with and reconstituted with 5 × 106 BM cells. Mice
anti-CD64 (catalog no. 139305; BioLegend), FITC PERTEX (HistoLab Products AB) for the visu- were closely monitored, and peripheral blood
anti-CD38 (catalog no. 102705; BioLegend), APC/ alization of cellular components. Differential chimerism was analyzed 8 weeks after recon-
Cy7 anti-Ly-6G (catalog no. 127624; BioLegend), interference contrast images were acquired at stitution, after which laser burn–induced CNV
BV785 anti-CD11c (catalog no. 117335; BioLegend), 20× using an AxioObserver.Z1 (Live Cell Zeiss was performed.
and PE/Cy7 anti-CD206 (catalog no. 141719; Imaging System, Zeiss, Jena, Germany). Adi-
BioLegend); for T cells in adipose tissue, BV711 pocyte size was evaluated and quantified using ATM cell sorting
anti-CD11b (catalog no. 101242; Biolegend), BV785 a custom-written program in MATLAB R2019b ATMs were stained with the antibodies men-
anti-CD4 (catalog no. 100551; BioLegend), PerCP/ (9.7.0.1190202, The MathWorks, Inc). tioned above (except for CD206 and CD38
Cy5.5 anti-CD8a (catalog no. 100733; BioLegend), antibodies) and sorted from the SVF of eWAT.
PE anti–TCR-b (catalog no. 109207; BioLegend), GTT Viable Ly6G–/CD45.2+/CD11b+/CD64+/F4/80+
and APC/Cy7 anti-CD45R/B220 (catalog no. 103223; Mice were starved for 12 hours overnight. cells were sorted using a FACS Aria instrument
BioLegend); for peripheral blood after ATT, BV711 Blood glucose was measured from the tail (BD Biosciences, Mississauga, ON, Canada) and
anti-CD11b (catalog no. 101242; BioLegend), FITC vein using AlphaTrak2 test strips at baseline recovered in FBS. After sorting, ATMs were
anti-CD3e (catalog no. 100305; BioLegend), BV785 and at 15, 30, 60, 120, and 240 min after in- immediately prepared for ATAC-seq sample
anti-CD45.2 (catalog no. 109839; BioLegend), PE/ traperitoneal injection of PBS +10% D-glucose preparation.
Cy7 anti-F4/80 (catalog no. 123113; BioLegend), (2 mg/kg of weight). Plasma insulin concen-
BV421 anti–Ly-6G (catalog no. 127627; BioLegend), trations were determined by an ultrasensitive ATAC-seq sample preparation
and APC anti-CD19 (catalog no. 115511; BioLegend); mouse insulin ELISA kit (catalog no. 90080; FACS-sorted ATMs were used for ATAC-seq,
for peripheral blood to characterize cells of do- Crystal Chem) at each time point of the GTT. and nuclei were isolated as previously described
nor and host origins, BV711 anti-CD11b (catalog (61). Briefly, isolated cells were centrifuged at
no. 101242; BioLegend), APC/Cy7 anti-Ly-6G (cat- ITT 500g for 5 min at 4°C and then resuspended
alog no. 127624; BioLegend), PE anti-F4/80 (cat- Mice were starved for 6 hours during daylight in ice-cold PBS + 0.04% BSA. Cell lysis was
alog no. 123110; BioLegend), FITC anti–Ly-6C hours. Blood glucose was measured from the performed for 5 min on ice by adding 45 ml of
(catalog no. 128006; BioLegend), Pacific Blue tail vein using AlphaTrak2 test strips at baseline lysis buffer (10 mM Tris-HCl, pH 7.4, 10 mM
anti-CD45.1 (catalog no. 110722; BioLegend), and and at 30, 60, and 120 min after intraperitoneal NaCl, 3 mM MgCl2, 0.1% Tween-20, 0.1% Nonidet
Alexa Fluor 700 anti-CD45.2 (catalog no. 109822; injection of insulin (0.75 U/kg body weight). P-40, 0.001% digitonin, and 1% BSA). After lysis,
BioLegend); for BM cells, APC/Cy7 anti-CD117 50 ml of ice-cold wash buffer (10 mM Tris-HCl,
(c-kit) (catalog no. 105825; Biolegend), FITC anti– ATT pH 7.4, 10 mM NaCl, 3 mM MgCl2, 0.1% Tween
Ly-6A/E (Sca-1) (catalog no. 108105; BioLegend), Eight-week-old recipient mice were randomly 20, and 1% BSA) was added and then cen-
PE/Cy7 anti-CD34 (catalog no. 128617; BioLegend), assigned to the different groups independently trifuged at 500g for 5 min at 4°C. After an
BV421 anti-CD48 (catalog no. 103427; BioLegend), of the origin of the donor eWAT. Donor mice additional wash, the nuclei were counted and
BV711 anti-CD150 (SLAM) (catalog no. 115941; were anesthetized, sacrificed, and their eWAT used for the transposase reaction.
BioLegend), PE anti-CD135 (Flt3) (catalog no. fat pads were carefully excised and weighed. ATAC-seq was performed as previously de-
135305; BioLegend), and APC mouse lineage The recipient mice were anesthetized with iso- scribed with slight modifications (62). Briefly,
antibody cocktail or APC mouse lineage isoflurane and subjected to multiple dorsal incisions 3250 to 16,000 nuclei were directly treated
type control cocktail (catalog no. 558074; BD (four to six incisions) to allow subcutaneous with Tn5 transposase at 37°C for 30 min. After
Pharmingen). engraftment of the same amount (~500 mg) of the enzymatic reaction, the DNA was purified
For T cell cytokine production, cells were stim- donor eWAT. Sham surgeries of control ani- with Zymo Column DNA Clean-5 and enriched
ulated with phorbol myristate acetate (50 ng/ml; mals were performed in the same manner but by 13 cycles of PCR. The libraries were recov-
catalog no. P1585; Sigma-Aldrich) and ionomycin without fat pad transplantation. Mice were ered from PCR by purification followed by size
(1 mg/ml) (catalog no. I0634; Sigma-Aldrich) in closely monitored for 3 weeks before perform- selection (180 to 750 bp) with KARA Pure
the presence of BD GolgiPlug (BFA) (catalog ing laser-induced CNV. Successful engraft- Beads and were paired-end sequenced using
no. 555028; BD Bioscience) for 4 hours at 37°C. ment of the transplantation was verified by an Illumina NovaSeq6000 Flowcell SP-PE50.
Cells were permeabilized and fixed with the evaluating the transplanted tissue (e.g., blood Tn5 tagmentation, DNA purification, library
BD Cytofix/Cytoperm kit (catalog no. 554722; vessel reperfusion and lack of necrosis) 14 days preparation, and bioinformatic analysis were
BD Bioscience) and then incubated with FITC after laser-induced CNV. performed at the Genomics Platform of the
anti-IFNγ (catalog no. 11-7311-41; eBioscience) Institut de Recherches Cliniques de Montréal.
and PE/Cy7 anti–TNF-a (catalog no. 506323; Plasma TNF measurement for ATT mice Sequencing reactions were performed at the
BioLegend) for intracellular staining for 25 min Plasma was harvested as above, and the levels Centre d’Expertise et de Services Génome
at 4°C. of TNF-a were determined before and 2, 7, 14, Québec sequencing platform.
Samples were acquired using a Fortessa X-20 and 21 days after ATT. A mouse TNF-a ELISA
cell analyzer (BD Biosciences, Mississauga, ON, kit (catalog no. BMS607-3; Invitrogen) was used In vitro BMDM assays
Canada) and analyzed using FlowJo Software to detect the secretion of TNF-a in plasma ac- BMDMs were starved for 6 hours and subse-
(vVersion 10.2; FlowJo, Ashland, OR, USA). cording to the manufacturer’s instructions. quently treated with either the TLR4 inhibitor

TAK-242 (final concentration 1 mM) or di- 2.10.0 R package was used to generate the count 26. E. M. Andriessen et al., EMBO Mol. Med. 8, 1366–1379 (2016).
methyl sulfoxide (DMSO) (vehicle) for 1 hour. matrix of Tn5 insertion site numbers for each 27. V. Lambert et al., Nat. Protoc. 8, 2197–2211 (2013).
28. O. Gavrilova et al., J. Clin. Invest. 105, 271–278 (2000).
After pretreatment, FA-free BSA (5%), palmi- consensus peak (peaks that were present in at
29. S. Klebanov, C. M. Astle, O. DeSimone, V. Ablamunits,
tate (500 mM), or SA (200 mM) was added to least two samples). DARs were identified be- D. E. Harrison, J. Endocrinol. 186, 203–211 (2005).
the medium for 24 hours. The medium was tween conditions using the DESeq2 R package 30. C. N. Lumeng, J. L. Bodzin, A. R. Saltiel, J. Clin. Invest. 117,
then refreshed and changed periodically for with adjusted P < 0.05 and |fold change|>2 175–184 (2007).
31. Q. Wang, H. Wu, Front. Immunol. 9, 2509 (2018).
5 days. Before collecting the cell lysates, BMDMs (66). A positive DAR value means that the 32. H. Wu et al., Circulation 115, 1029–1038 (2007).
were treated with LPS (10 ng/ml) for 4 hours chromatin is more open and a negative DAR 33. B. F. Zamarron et al., Diabetes 66, 392–406 (2017).
or HMGB1 (500 ng/ml) for 6 hours. value means that the chromatin is more closed 34. S. A. Amici et al., Front. Immunol. 9, 1593 (2018).
35. C. Erridge, J. Leukoc. Biol. 87, 989–999 (2010).
in the HFD-RD or HFD samples compared with
Seahorse extracellular flux analyzer assays 36. H. Shi et al., J. Clin. Invest. 116, 3015–3025 (2006).
RD-RD samples. DARs were annotated with 37. B. Vandanmagsar et al., Nat. Med. 17, 179–188 (2011).
Real-time analysis ECAR and OCR were per- their closest feature, and different transcription- 38. H. Wen et al., Nat. Immunol. 12, 408–415 (2011).
formed on BMDM from BSA-pretreated and binding motifs were identified with HOMER 39. J. Y. Lee, K. H. Sohn, S. H. Rhee, D. Hwang, J. Biol. Chem. 276,
16683–16689 (2001).
SA-pretreated mice with the XFe96 Flux Ana- version 4.8.0. 40. K. Ganeshan, A. Chawla, Annu. Rev. Immunol. 32, 609–634 (2014).
lyzer (Seahorse Bioscience) as per the manu- 41. A. J. Freemerman et al., J. Biol. Chem. 289, 7884–7896 (2014).
facturer’s instructions. BMDMs were plated in Statistical analysis 42. N. Vo, R. H. Goodman, J. Biol. Chem. 276, 13505–13508
(2001).
a 96-well Seahorse plate at a seeding density Data are expressed as mean ± SEM unless 43. W. Zhao et al., J. Immunol. 186, 3173–3179 (2011).
of 1.2 × 105 cells per well the day before exper- otherwise indicated. Multiple Student’s t tests, 44. A. J. Bannister, T. Oehler, D. Wilhelm, P. Angel, T. Kouzarides,
iments. Cells were incubated in GlycoStress test two-tailed Student’s t test, or ANOVA t test Oncogene 11, 2509–2514 (1995).
45. L. N. Tsai, T. K. Ku, N. K. Salib, D. L. Crowe, Mol. Cell. Biol. 28,
assay medium (DMEM 5030, 2 mM glutamine; were used to compare two or more than two 4240–4250 (2008).
catalog no. 25030-081; Life Technologies), pH groups, respectively. Statistical significance was 46. D. Lara-Astiaso et al., Science 345, 943–949 (2014).
7.4, 1 hour before ECAR measurements. Glu- considered when P < 0.05 and is indicated as 47. I. Goren et al., Am. J. Pathol. 175, 132–147 (2009).
48. F. Sennlaub et al., EMBO Mol. Med. 5, 1775–1793 (2013).
cose (25 mM; catalog no. G7528-250G; Sigma- follows in the figures: *P < 0.05, **P < 0.01, 49. L. G. Fritsche et al., Nat. Genet. 48, 134–143 (2016).
Aldrich), oligomycin (1 mM, catalog no. O4876; ***P < 0.001, and ****P < 0.0001. All experi- 50. M. G. Netea et al., Science 352, aaf1098 (2016).
Sigma-Aldrich), and 2-deoxy-glucose (100 mM; ments were repeated at least three times ex- 51. A. Christ et al., Cell 172, 162–175.e14 (2018).
52. G. Boden, Endocrinol. Metab. Clin. North. Am. 37, 635–646 (2008).
catalog no. sc-202010A; Santa Cruz Biotechnol- cept for the ATAC-seq, which was performed
53. D. Hwang, FASEB J. 15, 2556–2564 (2001).
ogy) were added sequentially. Glycolysis and once. The number of experimental replicates 54. E. K. Anderson, A. A. Hill, A. H. Hasty, Arterioscler. Thromb.
glycolytic capacity were calculated as ECAR. is indicated for each experiment in the figure Vasc. Biol. 32, 1687–1695 (2012).
For OCR, cells were incubated in MitoStress legends. 55. M. G. Netea et al., Nat. Rev. Immunol. 20, 375–388 (2020).
56. J. Thompson Legault et al., Cell Rep. 13, 981–989 (2015).
test assay medium containingDMEM 5030 57. V. Lamantia et al., J. Nutr. 149, 57–67 (2019).
added with 25 mM glucose (catalog no. G7528- RE FERENCES AND NOTES 58. H. Zhou et al., Front. Immunol. 9, 3082 (2019).
250G; Sigma-Aldrich), 2 mM glutamine (catalog 1. H. R. Coleman, C. C. Chan, F. L. Ferris 3rd, E. Y. Chew, Lancet 59. M. A. Phillips et al., BMC Dermatol. 4, 2 (2004).
372, 1835–1845 (2008). 60. A. Baena et al., PLOS ONE 2, e621 (2007).
no. 25030-081; Life Technologies), and 1 mM 61. J. D. Buenrostro, B. Wu, H. Y. Chang, W. J. Greenleaf, Curr.
2. A. Swaroop, E. Y. Chew, C. Bowes Rickman, G. R. Abecasis,
NaPyr (catalog no. P5280-25G; Sigma-Aldrich) Annu. Rev. Genomics Hum. Genet. 10, 19–43 (2009). Protoc. Mol. Biol. 109, 21.29.1–21.29.9 (2015).
1 hour before the experiment. Oligomycin (1 mM; 3. W. L. Wong et al., Lancet Glob. Health 2, e106–e116 (2014). 62. J. D. Buenrostro, P. G. Giresi, L. C. Zaba, H. Y. Chang,
4. J. M. Seddon, R. Reynolds, Y. Yu, M. J. Daly, B. Rosner, W. J. Greenleaf, Nat. Methods 10, 1213–1218 (2013).
catalog no. O4876; Sigma-Aldrich), carbonyl 63. B. Langmead, S. L. Salzberg, Nat. Methods 9, 357–359 (2012).
Ophthalmology 118, 2203–2211 (2011).
cyanide p-trifluoromethoxyphenylhydrazone 5. J. W. Crabb et al., Proc. Natl. Acad. Sci. U.S.A. 99, 64. F. Ramírez et al., Nucleic Acids Res. 44, W160–W165 (2016).
(2 mM; catalog no. ab120081; Abcam), and 14682–14687 (2002). 65. Y. Zhang et al., Genome Biol. 9, R137 (2008).
6. C. A. Curcio, C. L. Millican, T. Bailey, H. S. Kruth, Invest. 66. S. Anders, W. Huber, Genome Biol. 11, R106 (2010).
1 mM rotenone-antimycin (catalog no. A8674;
Ophthalmol. Vis. Sci. 42, 265–274 (2001).
Sigma-Aldrich) were added sequentially. Basal 7. R. B. Thompson et al., Proc. Natl. Acad. Sci. U.S.A. 112, AC KNOWLED GME NTS
respiration and maximal respiration were cal- 1565–1570 (2015). We thank all other laboratory members for helpful discussions;
culated as OCR. Upon completion of Seahorse 8. D. H. Anderson, R. F. Mullins, G. S. Hageman, L. V. Johnson, M. Dupuis, E. Massicotte, and M. Faquette for cell sorting and
Am. J. Ophthalmol. 134, 411–431 (2002). reagent preparation; the animal care technicians for mice
analysis, cells were lysed and their protein
9. X. Guillonneau et al., Prog. Retin. Eye Res. 61, 98–128 (2017). husbandry; and O. Neyret from the Institut de Recherches
content quantified. OCR and ECAR data were 10. G. S. Hageman et al., Prog. Retin. Eye Res. 20, 705–732 Cliniques de Montréal for help with ATAC-seq. Funding: P.S. holds
normalized to protein content. (2001). the Wolfe Professorship in Translational Research, a Canada
11. H. Xu, M. Chen, J. V. Forrester, Prog. Retin. Eye Res. 28, Research Chair in Retinal Cell Biology, and is the Fonds de
QUANTIFICATION AND STATISTICAL ANALYSIS 348–368 (2009). Recherche en Ophtalmologie de l’Université de Montréal (FROUM)
12. K. Rashid, I. Akhtar-Schaefer, T. Langmann, Front. Immunol. 10, Endowed Chair. Mas.H. holds the Banting Fellowship from the CIHR
ATAC-seq analysis 1975 (2019). and Fellowship from Japan Society for the Promotion of Science
After read quality control with FASTQC ver- 13. N. Joachim, P. Mitchell, G. Burlutsky, A. Kifley, J. J. Wang, (JSPS). This work was supported by operating grants to P.S from
Ophthalmology 122, 2482–2489 (2015). the Canadian Institutes of Health Research (foundation grant
sion 0.11.8, alignment was performed using 14. R. Klein, T. Peto, A. Bird, M. R. Vannewkirk, Am. J. Ophthalmol. 353770), an Alcon Research Institute Senior Investigator Award,
Bowtie2 version 2.2.6 (63) on the GRCm38 137, 486–495 (2004). Diabetes Canada (grant DI-3-18-5444-PS), and The Heart and
mouse reference genome. Alignments were post- 15. L. S. Lim, P. Mitchell, J. M. Seddon, F. G. Holz, T. Y. Wong, Stroke Foundation of Canada (grant G-16-00014658) and the
Lancet 379, 1728–1738 (2012). BrightFocus Foundation (grant M2022015I). Additional support
processed to remove PCR duplicates (Picard tool 16. V. Behnke, A. Wolf, T. Langmann, Cell. Mol. Life Sci. 77, was provided by the FROUM and the Réseau en Recherche
version 2.4.1) and reads mapping to mitochon- 781–788 (2020). en Santé de la Vision. This work also benefited from infrastructure
drial DNA (samtools version 1.8). To represent 17. M. K. Adams et al., Am. J. Epidemiol. 173, 1246–1255 (2011). and personal support by the Canadian Foundation for Innovation
18. P. Haas, K. E. Kubista, W. Krugluger, J. Huber, S. Binder, (grant 20415) and the Montreal Heart Institute Foundation
the real Tn5 transposase-binding sites of 9 bp, Acta Ophthalmol. 93, 533–538 (2015). (C.D.R.). R.D.-M. is supported by a research scholarship from
the coordinates of the reads were shifted by 19. J. M. Seddon, R. Widjajahakim, B. Rosner, Invest. Ophthalmol. FROUM. S.C.-G. holds a Fonds de Recherche Santé du Québec
4 bp for the plus strand and by 5 bp for the Vis. Sci. 61, 32 (2020). (FRQS) scholarship. J.-S.J. was supported by the Canadian
20. J. B. Dixon et al., JAMA 299, 316–323 (2008). Institute of Health Research (CIHR grant 390615), the National
minus strand using Deeptools 3.0.1 (64). The
21. A. Golay et al., Int. J. Obes. 9, 181–191 (1985). Sciences and Engineering Research Council of Canada (NSERC
former was also used to remove ENCODE’s 22. S. D. Long et al., Diabetes Care 17, 372–375 (1994). grant 06743), and the FRQS. Author contributions: Mas.H.,
blacklisted regions (signal artifact regions) 23. J. Olefsky, G. M. Reaven, J. W. Farquhar, J. Clin. Invest. 53, E.M.M.A.A., J.-S.J., and P.S. designed the research. Mas.H.,
64–76 (1974). E.M.M.A.A., Mak.H., R.D.-M., F.F., G.B., R.J., F.P., A.D., V.G., E.H.,
and to convert Bam files to BEDPE format.
24. T. Deng, C. J. Lyon, S. Bergin, M. A. Caligiuri, W. A. Hsueh, C.D., and A.M.W. performed the research. Mas.H., E.M.M.A.A., Mak.H.,
MACS2 was used to identify significant peaks Annu. Rev. Pathol. 11, 421–449 (2016). G.B., E.H., C.D., C.D.R., J.-S.J., and P.S. analyzed the data. Mas.H.
using a q-value < 0.05 (65). The Diffbind version 25. A. Sene et al., Cell Metab. 17, 549–561 (2013). and P.S. wrote the paper with contributions from E.M.M.A.A.,

S.C.-G., C.D.R., H.J.M., T.L., and A.M.W. Competing interests: P.S. exclusive licensee American Association for the Advancement of Tables S1 to S3
and Mas.H have filed for intellectual property on the concepts Science. No claim to original US government works. https://www. MDAR Reproducibility Checklist
presented in this study. The remaining authors declare no science.org/about/science-licenses-journal-article-reuse
competing interests. Data and materials availability: All ATAC-
seq data for the study have been deposited in the National Center
for Biotechnology Information Gene Expression Omnibus and are SUPPLEMENTARY MATERIALS Submitted 9 June 2021; resubmitted 13 April 2022
accessible through GEO series accession no. GSE175614. License science.org/doi/10.1126/science.abj8894 Accepted 3 November 2022
information: Copyright © 2023 the authors, some rights reserved; Figs. S1 to S12 10.1126/science.abj8894
EXPATRIATE SCHOLARS matched by the host institution and even local

governments. All awardees are also provided
Has ChinaÕs Young Thousand Talents program with fringe benefits such as housing subsidies
and are prioritized when applying for local and
been successful in recruiting and nurturing national grants.
To be eligible, YTT applicants should ideally
top-caliber scientists? (i) work in the STEM field and be 40 years of
age or younger, (ii) have a PhD from a rep-
Dongbo Shi1*, Weichen Liu2, Yanbo Wang3* utable overseas university and three or more
years of overseas research experience, (iii) have
In this study, we examined China’s Young Thousand Talents (YTT) program and evaluated its effectiveness in a full-time overseas research position, (iv) be
recruiting elite expatriate scientists and in nurturing the returnee scientists’ productivity. We find that YTT committed to full-time employment in China,
scientists are generally of high caliber in research but, as a group, fall below the top category in pre-return and (v) be a top talent in their cohort and have
productivity. We further find that YTT scientists are associated with a post-return publication gain across the potential to become a research-field leader.
journal-quality tiers. However, this gain mainly takes place in last-authored publications and for high-caliber However, these criteria are not rigid. YTT also
(albeit not top-caliber) recruits and can be explained by YTT scientists’ access to greater funding and larger welcomes freshly minted overseas PhDs and
research teams. This paper has policy implications for the mobility of scientific talent, especially as early-career expatriate researchers with Chinese PhDs to ap-
scientists face growing challenges in accessing research funding in the United States and European Union ply if they have outstanding research records.
China’s YTT program made 3576 offers be-
I
tween 2011 and 2017. Although designed to im-
mmigrants are playing an increasing role China’s talent programs have been effective prove China’s prospect of becoming a global
in US science and engineering (1–3), and in recruiting top-caliber scientists (9) and in STEM leader, the program’s effectiveness in
China particularly has been the top sender nurturing the returnees’ productivity (10). attracting top talents and nurturing their pro-
of international students to the US’s STEM Studying talent programs is important for ductivity is unclear. On one hand, a program
programs (4, 5). In recent years, China understanding the evolving landscape of global providing substantive research support could
launched an ambitious Thousand Talents Pro- knowledge production; it is also policy-relevant motivate expatriate talents to return and even
gram (TTP) to recruit elite expatriate scientists because an increasing number of govern- help grow their productivity; on the other hand,
to return to China. This program has received ments across both high-income (e.g., Canada returnees may struggle to reintegrate into
intense attention both from the US govern- and Singapore) and middle- or lower-income China’s academia (7–9) and thus experience a
ment, as reflected in the launch of the China (e.g., Brazil and India) countries are pursuing research output slowdown. YTT scientists may
Initiative, and from the academic community, means to tap expatriates and migrant networks also be incentivized to focus on publication
especially over the Federal Bureau of Inves- for domestic knowledge production and talent quantity rather than quality, because program
tigation’s arrest of Massachusetts Institute of development. Some governments have come officials have motivations to demonstrate YTT’s
Technology professor Gang Chen. to believe that expats and returnees are the impact on publication counts, even at the cost
Despite the attention, there has been little key to building globally competitive research of quality and originality.
evidence-based research on the operation, im- institutions and dynamic, knowledge-based
pact, and policy implications of China’s talent economies. Data and methods
programs. Prior research on scientific retur- We studied the YTT program’s first four co-
nees has found productivity declines among ChinaÕs Young Thousand Talents program horts, totaling 721 awardees. Our main analy-
those returning to lower-income home coun- We examined China’s Young Thousand Talents ses excluded 309 individuals because they
tries, but such research has focused on countries (YTT) program, the “youth” branch of the TTP. either returned for nonacademic jobs (27),
other than China (6). China-specific studies have Among the country’s 200-plus talent recruit- received PhDs in China (196), were not of Chi-
suggested that returnees face difficulties re- ment programs, TTP is the most prominent nese origin (34), left China within 5 years of
integrating into the country’s research envi- initiative to bring leading global scientists to returning (5), or lacked CV information (47).
ronment (7), where administrative intervention China. In principle, TTP is open to researchers This left us with 73 scientists who rejected
and personal connections hinder scientific in- of any nationality; but in practice, few non- the YTT offers to remain overseas (hereafter,
quiry (8). This raises the question of whether Chinese have availed themselves of the program. “rejectors”) and 339 returnees who received
Established in 2010, the YTT program targets PhDs abroad, accepted the offers, and spent
1 outstanding young STEM scholars and offers at least 5 years conducting research in China
School of International and Public Affairs, Shanghai Jiao
Tong University, Shanghai, China. 2School of Public Policy generous financial support to each awardee, (hereafter, “acceptors”).
and Management, Tsinghua University, Beijing, China. including a one-off tax-exempt income subsidy We implemented two sets of analyses. First,
3
Faculty of Business and Economics, The University of Hong of 500,000 yuan RMB (~$150,200 in 2010 USD we examined YTT returnees’ educational cre-
Kong, Hong Kong.
*Corresponding author. Email: [email protected] (D.S.); purchasing power parity) and start-up grants dentials and pre-return productivity. We specif-
[email protected] (Y.W.) of 1 million to 3 million yuan. This package is ically used rejector-versus-acceptor comparisons

Copyright 2023 American Association for the Advancement of Science. All rights reserved.

Articulometodos 1

Uploaded by

Copyright:

Available Formats

Articulometodos 1

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Articulometodos 1

Uploaded by

Copyright:

Available Formats

RESE ARCH | R E S E A R C H A R T I C L E S

fundamental minimum required thicknesses. IMMUNOLOGY

SCIENCE science.org 6 JANUARY 2023 • VOL 379 ISSUE 6627 45

46 6 JANUARY 2023 ¥ VOL 379 ISSUE 6627 science.org SCIENCE

SCIENCE science.org 6 JANUARY 2023 • VOL 379 ISSUE 6627 47

48 6 JANUARY 2023 ¥ VOL 379 ISSUE 6627 science.org SCIENCE

SCIENCE science.org 6 JANUARY 2023 • VOL 379 ISSUE 6627 49

50 6 JANUARY 2023 ¥ VOL 379 ISSUE 6627 science.org SCIENCE

SCIENCE science.org 6 JANUARY 2023 • VOL 379 ISSUE 6627 51

52 6 JANUARY 2023 ¥ VOL 379 ISSUE 6627 science.org SCIENCE

SCIENCE science.org 6 JANUARY 2023 • VOL 379 ISSUE 6627 53

54 6 JANUARY 2023 ¥ VOL 379 ISSUE 6627 science.org SCIENCE

SCIENCE science.org 6 JANUARY 2023 • VOL 379 ISSUE 6627 55

56 6 JANUARY 2023 ¥ VOL 379 ISSUE 6627 science.org SCIENCE

SCIENCE science.org 6 JANUARY 2023 ¥ VOL 379 ISSUE 6627 57

58 6 JANUARY 2023 • VOL 379 ISSUE 6627 science.org SCIENCE

SCIENCE science.org 6 JANUARY 2023 • VOL 379 ISSUE 6627 59

60 6 JANUARY 2023 • VOL 379 ISSUE 6627 science.org SCIENCE

SCIENCE science.org 6 JANUARY 2023 • VOL 379 ISSUE 6627 61

EXPATRIATE SCHOLARS matched by the host institution and even local

62 6 JANUARY 2023 • VOL 379 ISSUE 6627 science.org SCIENCE

You might also like