JAK inhibition alleviates the cellular senescence-associated secretory phenotype and frailty inold ageMing Xua, Tamara Tchkoniaa, Husheng Dinga, Mikolaj Ogrodnika,b, Ellen R. Lubbersa, Tamar Pirtskhalavaa,Thomas A. Whitea, Kurt O. Johnsona, Michael B. Stouta, Vojtech Mezeraa, Nino Giorgadzea, Michael D. Jensena,Nathan K. LeBrasseura, and James L. Kirklanda,1

aRobert and Arlene Kogod Center on Aging, Mayo Clinic, Rochester, MN 55905; and bNewcastle University Institute for Ageing, Newcastle University,Newcastle Upon Tyne, NE4 5PL, United Kingdom

Edited by Yinon Ben-Neriah, Hebrew University, Jerusalem, Israel, and accepted by the Editorial Board September 16, 2015 (received for review August 3, 2015)

Chronic, low grade, sterile inflammation frequently accompaniesaging and age-related diseases. Cellular senescence is associatedwith the production of proinflammatory chemokines, cytokines,and extracellular matrix (ECM) remodeling proteases, which com-prise the senescence-associated secretory phenotype (SASP). Wefound a higher burden of senescent cells in adipose tissue withaging. Senescent human primary preadipocytes as well as humanumbilical vein endothelial cells (HUVECs) developed a SASP thatcould be suppressed by targeting the JAK pathway using RNAi orJAK inhibitors. Conditioned medium (CM) from senescent humanpreadipocytes induced macrophage migration in vitro and inflam-mation in healthy adipose tissue and preadipocytes. When thesenescent cells from which CM was derived had been treated withJAK inhibitors, the resulting CM was much less proinflammatory. Theadministration of JAK inhibitor to agedmice for 10wk alleviated bothadipose tissue and systemic inflammation and enhanced physicalfunction. Our findings are consistent with a possible contribution ofsenescent cells and the SASP to age-related inflammation and frailty.We speculate that SASP inhibition by JAK inhibitors may contributeto alleviating frailty. Targeting the JAK pathway holds promise fortreating age-related dysfunction.

JAK/STAT pathway | cellular senescence | ruxolitinib | interleukin-6 | frailty

Ahallmark of aging is chronic, low-grade, “sterile” inflam-mation (1–3). Elevated proinflammatory cytokines and

chemokines are closely associated with mortality (4, 5) and with avariety of age-related diseases, including atherosclerosis (6), de-pression (7), cancers (8), diabetes (9), and neurodegenerativediseases (10, 11). Inflammation also is associated with frailty, a ge-riatric syndrome characterized by decreased strength and incapacityto respond to stress (2).The underlying mechanisms of age-related chronic inflam-

mation, tissue dysfunction, and frailty remain elusive. Cellularsenescence, stable arrest of cell growth in replication-competentcells, is a plausible contributor. Senescence can be induced by anumber of stimuli and stresses, including telomere dysfunction,genomic instability, oncogenic and metabolic insults, and epi-genetic changes (12). Senescent cells accumulate with aging inthe skin (13, 14), liver (15, 16), kidney (17), the cardiovascularsystem (18), and other tissues in various species (16). The se-nescence-associated secretory phenotype (SASP), largely com-prised of proinflammatory cytokines and chemokines (19, 20),links senescent cells to age-related inflammation and diseases.We found that elimination of senescent cells delayed the onsetof age-related phenotypes and enhanced healthspan (21, 22).Therefore, senescent cells and the SASP could play a role inage-related pathologies, particularly those that involve systemicinflammation.The JAK/STAT pathway plays an important role in regulating

cytokine production (23, 24). We hypothesized that it may di-rectly affect the SASP. The JAK family has four members: JAK1,

JAK2, JAK3, and tyrosine kinase 2 (TYK2). JAK1 and 2 are in-volved in inflammatory signaling and in the action of growth hor-mone and other endocrine and paracrine signals (25). JAK3 isexpressed primarily in blood cells and mediates erythropoietin sig-naling and immune cell generation (26–29). TYK2 plays an im-portant role in white blood cell function and host defenses (30, 31).A number of inhibitors that target different JAKs have been ap-proved or are in phase I–III studies for treating myelofibrosis (23, 32–34), acute myeloid leukemia (23), lymphoma (23), and rheuma-toid arthritis (23, 35, 36). JAK inhibitors recently have beenfound to reprogram the SASP in senescent tumor cells, con-tributing to improved antitumor response (37). Therefore, weinvestigated the role of the JAK pathway in age-related inflam-mation and dysfunction.We demonstrate here that senescent preadipocytes, fat cell pro-

genitors, accumulate in adipose tissue with aging and can contributeto adipose tissue inflammation. We found that JAK inhibitors de-crease the SASP in preadipocytes and human umbilical vein en-dothelial cells (HUVECs). They also decrease age-related adiposetissue and systemic inflammation as well as frailty. Our findingsprovide insights into the possible contribution of senescent cellsto age-related inflammation and, in turn, to age-related pathologies,as well as potential therapeutic targets to alleviate age-relateddysfunction.

Significance

A hallmark of aging is chronic sterile inflammation, which is closelyassociated with frailty and age-related diseases. We found thatsenescent fat progenitor cells accumulate in adipose tissue withaging and acquire a senescence-associated secretory phenotype(SASP), with increased production of proinflammatory cytokinescompared with nonsenescent cells. These cells provoked in-flammation in adipose tissue and induced macrophage migration.The JAK pathway is activated in adipose tissue with aging, and theSASP can be suppressed by inhibiting the JAK pathway in senes-cent cells. JAK1/2 inhibitors reduced inflammation and alleviatedfrailty in aged mice. One possible mechanism contributing to re-duced frailty is SASP inhibition. Our study points to the JAK path-way as a potential target for countering age-related dysfunction.

Author contributions: M.X., T.T., and J.L.K. designed research; M.X., T.T., H.D., M.O., E.R.L.,T.P., T.A.W., K.O.J., M.B.S., V.M., N.G., and M.D.J. performed research; M.X., T.T., M.O., andN.K.L. analyzed data; and M.X., T.T., and J.L.K. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. Y.B.-N. is a guest editor invited by the EditorialBoard.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. Email: kirkland.james@mayo.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1515386112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1515386112 PNAS | Published online November 2, 2015 | E6301–E6310

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

ResultsSenescent Cells Accumulate in Adipose Tissue with Aging. Preadipocytes,also termed “adipose-derived stem cells” or “fat cell progenitors”(38), constitute 15–50% of the cells in adipose tissue and arguablyare the most abundant progenitor cell type in humans (39). Wefirst tested whether the number of senescent preadipocytesincreases in adipose tissue with aging. We isolated primarypreadipocytes from the stromal-vascular fraction of abdominal s.c. fattissue from healthy old and young human subjects. We stainedthese cells with histone γH2AX and the cell proliferation markerKi-67. Cells with γH2AX+ foci and Ki-67− nuclei were con-sidered to be senescent cells. We found that more than 10%of preadipocytes in adipose tissue from old human subjects but lessthan 4% of preadipocytes in adipose tissue from young subjectswere senescent (Fig. 1). Similar results were found when these cellswere assayed for cellular senescence-associated β-galactosidase(SABG) (Fig. S1A). Consistent with our data in humans, we foundthere are more SABG+ cells in adipose tissue from 30-mo-old ratsthan in adipose tissue from 3-mo-old rats (Fig. S2A). The fre-quency of γH2AX+ (a marker of DNA damage) cells also washigher in adipose tissue from 30-mo-old rats (Fig. S2C). Levels ofp21Cip1 protein, a marker of cellular senescence, were higher inthese adipose depots in the old animals (Fig. S2B). Both activated(phosphorylated) JAK1 (Tyr1022/1023) and JAK2 (Tyr1007/1008)were increased in adipose tissue from old rats as well (Fig. S2B).We observed more senescent preadipocytes isolated from theadipose tissue of 30-mo-old rats than from the adipose tissue of3-mo-old rats (Fig. S1 B and C). In addition, we recently dem-onstrated that aging is associated with an increased abundance ofsenescent cells in the adipose tissue of mice (40). Therefore, se-nescent cells accumulate in fat tissue with aging across severalmammalian species.

Senescent Preadipocytes Develop a SASP. To test if human primarypreadipocytes exhibit a SASP upon becoming senescent, we ir-radiated preadipocytes from healthy human kidney transplantdonors with 10 Gy to induce senescence. After 20 d, more than70% of preadipocytes became SABG+ (Fig. 2A), with increasedp16Ink4a and p21Cip1 mRNA abundance (Fig. 2B). We also in-duced cellular senescence by serial passage. By passage 27–30, mostpreadipocytes had stopped proliferating, and more than 80% wereSABG+ (Fig. 2A) with increased p16Ink4a and p21Cip1 transcriptabundance (Fig. 2B). To test cytokine secretion by these cells, wecollected conditioned medium (CM) from senescent and controlcells for 24 h. In an addressable laser bead-based multiplex assay,most of the cytokines and chemokines measured were signifi-cantly higher in CM from irradiation-induced senescent than from

control preadipocytes (Table S1). The most highly induced cy-tokines and chemokines were IL-6 (19-fold), GM-CSF (28-fold), granulocyte-colony stimulating factor (G-CSF) (10-fold),chemokine (C-X-C motif) ligand 1 (CXCL-1) (27-fold), macro-phage inflammatory protein-1α (MIP-1α) (30-fold), IL-8 (46-fold),monocyte chemotactic protein-1 (MCP-1) (19-fold), regulated onactivation, normal T-cell–expressed and –secreted (RANTES) (15-fold), and MCP-3 (17-fold) (see Fig. 4B; full results are in TableS1). Similar results were found for IL-6 and MCP-1 proteins in CMby ELISA (Fig. S3A). In serial passage-induced senescent cells, wealso detected the induction of secreted cytokines including IL-6,IL-8, MCP-1, and plasminogen activator inhibitor-1 (PAI-1) usingthe multiplex assay (Fig. S3B). Consistent with the increased se-creted cytokines, we found increased mRNA levels for several keySASP components, including IL-1α, IL-1β, IL-6, IL-8, MCP-1,matrix metalloproteinase (MMP)-3, MMP-12, PAI-1, and TNF-αin both irradiation- and serial passage-induced senescent pre-adipocytes compared with nonsenescent controls (Fig. 2C).

Senescent Preadipocytes Induce Inflammation in Adjacent Cells. Next,we tested whether senescent preadipocytes confer inflammation toadjacent healthy cells. If so, this effect would suggest that theSASP can amplify adipose tissue inflammation. Exposure ofnonsenescent preadipocytes or adipose tissue from healthy donorsto CM from irradiation-induced senescent preadipocytes for 24 hinduced the expression of inflammatory markers (Fig. 3 A and C,respectively) in a dose-dependent manner (Fig. S4), comparedwith exposure to CM from nonsenescent preadipocytes. Infiltra-tion of macrophages into adipose tissue is associated with adiposetissue dysfunction and inflammation (41). To determine whethersenescent cells could drive this event, we conducted cell migrationassays using THP1 monocytes. Relative to CM from nonsenescentpreadipocytes, CM derived from senescent cells caused a morethan fourfold increase in THP1 monocyte migration within 6 h(Fig. 3B).

Inhibition of JAK Suppresses the SASP in Senescent Preadipocytes.Given the role of the JAK/STAT pathway in inflammation, weexamined whether inhibiting the JAK/STAT pathway blunts theSASP. We treated irradiation-induced senescent preadipocytesfor 72 h with various concentrations (0–1 μM) of JAK inhibitor 1,which inhibits all four JAKs. We found that at a concentration of0.6 μM JAK inhibitor 1 suppressed the mRNA levels of keySASP components, including IL-6, IL-8, and MCP-1, by up to60%. Increasing the concentration of JAK inhibitor 1 to 1 μMdid not further suppress the SASP (Fig. 4A). JAK inhibitor 1 hadlittle effect on the same components in nonsenescent controlcells (Fig. S5). In a multiplex assay for cytokines and chemokines,

B

0

4

8

12

Human

Sene

scen

t cel

ls %

Young

Old

*

Young

γ-H2A.X Ki67 merged

Old

γ-H2A.X Ki67 merged

A Human

200μm

Fig. 1. Senescent preadipocytes accumulate in adipose tissue with aging. Primary preadipocytes were isolated from femoral s.c. fat obtained by needlebiopsy from four young (31 ± 5 y) and four old (71 ± 2 y) healthy male volunteers and then were stained with γH2AX and Ki-67. (A) Representative images ofhuman cells. (B) The percentages of γH2AX+ and Ki-67− preadipocytes were counted. Results are expressed as mean ± SEM; *P < 0.05; n = 4.

E6302 | www.pnas.org/cgi/doi/10.1073/pnas.1515386112 Xu et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

we observed that 16 of 32 proteins were significantly lower in theCM from senescent preadipocytes treated with JAK inhibitor 1(0.6 μM for 72 h; CM collected for the last 24 h) than in the CMfrom vehicle-treated senescent preadipocytes (Table S1). IL-6(58%), GM-CSF (68%), G-CSF (60%), IFN-γ–induced protein10 (IP-10) (92%), CXCL-1 (48%), MIP-1α (90%), IL-8 (62%),MCP-1 (59%), RANTES (52%), and MCP-3 (71%) were amongthe SASP components most prominently suppressed by JAKinhibitor 1 (Fig. 4B; full results are in Table S1). MMP3 andMMP12 mRNA levels were decreased by 80% when senescenthuman preadipocytes were treated with JAK inhibitor 1 (Fig. 4C).We also assessed the extent to which JAK inhibitor 1 affected therelease of key SASP components in serial passage-induced senescentpreadipocytes. Similar to our observations in irradiation-inducedsenescent cells, CM from passage-induced senescent cells treatedwith JAK inhibitor 1 contained less IL-6, IL-8, and MCP-1 than CMfrom vehicle-treated cells (Fig. S3B). Expression levels of several keySASP components also were reduced in passage-induced senescentcells treated with JAK inhibitor 1 (Fig. 4D). Endothelial cells are amajor population of cells in adipose tissue and elsewhere, and wefound that senescent endothelial cells accumulate in adipose tissue(42). JAK inhibitor 1 suppressed the SASP in irradiation-inducedHUVECs (Fig. S6), suggesting that inhibition of the JAK/STATpathway could suppress the SASP in multiple senescent cell types.

To exclude the possibility of nonspecific effects of JAK in-hibitor 1, we treated irradiation-induced senescent preadipocyteswith two other JAK inhibitors, momelotinib (CYT387) andINCB18424, both of which specifically inhibit JAK1 and JAK2.These two specific inhibitors blunted the SASP similarly to JAKinhibitor 1 (Fig. 4E). STAT3 is a STAT family member that inducesand maintains an inflammatory microenvironment by regulating arange of inflammatory cytokines and mediators, including the SASPcomponents IL-6, IL-8, PAI-1, MCP-1, and GM-CSF (24, 43). Weobserved that phosphorylated STAT3 (Tyr705) was elevated in se-nescent preadipocytes and was substantially decreased upon treat-ment with JAK inhibitor 1 (Fig. 4F), consistent with a role forJAK1/2 activation in the SASP.All three JAK inhibitors target both JAK1 and JAK2. To

distinguish the role of JAK1 from that of JAK2, we used siRNAto knock down either JAK1 or JAK2 expression in senescentpreadipocytes. We used two different siRNAs against the samegene transcripts to exclude off-target effects of these siRNAs.We were able to knock down JAK1 and JAK2 mRNA levels byabout 85% and 80%, respectively. Reduced expression of eitherJAK1 or JAK2 led to SASP inhibition (Fig. 5). Together, thesefindings indicate that both JAK1 and JAK2 have roles in regu-lating the SASP.

0

40

80

120

IL6 IL8 IL1α IL1β MCP1 MMP3 MMP120

6

12

18

PAI1 TNFα

CON

IRA

SP

A

200μm

CON IRA SP

SABG

DAPI

C

Rela

�ve

mRN

A le

vel

0

1

2

3

4

p16 p21

CON

IRA

SP

B

Rela

�ve

mRN

A le

vel

*

**

*

*

*

** *

*

*

**

*

**

*

*

*

*

**

Fig. 2. Senescent preadipocytes develop a SASP. (A) Preadipocytes were isolated from healthy human subjects. Control (CON, Left), irradiated (IRA, Center),and serially passaged (SP, Right) preadipocytes were assayed for cellular SABG activity and stained with DAPI. (B) The relative mRNA abundance of p16 andp21 in control (CON) and irradiation-induced (IRA) and serial passage-induced (SP) senescent preadipocytes is shown. Results are expressed as mean ± SEM;*P < 0.05; n = 7. (C) The relative mRNA abundance of key SASP components in control (CON), irradiation-induced (IRA), and serial passage-induced (SP)senescent preadipocytes is shown. Results are expressed as mean ± SEM; *P < 0.05 compared with nonsenescent controls; n = 5.

Xu et al. PNAS | Published online November 2, 2015 | E6303

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

JAK Inhibition Decreases Inflammation Induced by Senescent Cells.Because JAK inhibitors suppress the SASP, we tested if SASPsuppression reduces the inflammation and chemoattraction causedby senescent cell CM. Nonsenescent preadipocytes expressed sig-nificantly less IL-6, MCP-1, and PAI-1 when incubated with CMgenerated from senescent cells treated with JAK inhibitor thanwhen incubated with CM generated from vehicle-treated senescentcells. Interestingly, IL-1α, IL-1β, and IL-8 expression did not changewith JAK inhibitor treatment (Fig. 6A), suggesting that the SASP issegmentally suppressed by JAK inhibitors. Macrophage migrationwas reduced in response to CM from senescent preadipocytestreated with a JAK inhibitor compared with CM from vehicle-treated senescent cells (Fig. 6B). Because the JAK inhibitor waspresent in CM from the senescent cells treated with JAK inhibitor,we tested whether the addition of a JAK inhibitor to CM fromsenescent preadipocytes not treated with a JAK inhibitor couldprevent SASP-induced inflammation. We treated nonsenescent pre-adipocytes with CM from irradiation-induced senescent preadipocytesfor 24 h in the presence of vehicle or 0.6 μM JAK inhibitor 1 (bothvehicle and JAK inhibitor 1 had been placed in a 37 °C incubatorfor 24 h to mimic the CM collection process). JAK inhibition hadlittle effect on reducing target cell inflammation induced by CMfrom senescent preadipocytes (Fig. 6C). Therefore, JAK inhibitors

alleviated inflammation mainly by suppressing the SASP in se-nescent cells rather than by acting on nonsenescent cells to pre-vent the inflammation caused by the action of SASP components.

Inhibition of JAK/STAT Pathways Decreases Systemic and AdiposeTissue Inflammation in Old Mice. Aging is associated with chroniclow-grade sterile inflammation (1–3). JAK inhibition partiallyrescues the chronic low-grade sterile inflammation caused bysenescent preadipocytes in adipose tissue in vitro. Therefore, weexamined whether inhibition of the JAK pathway reduced sys-temic inflammation in aged mice. We treated 24-mo-old C57BL/6male mice with ruxolitinib (INCB18424), a selective JAK1/2 in-hibitor approved by the Food and Drug Administration, or vehicle(DMSO) for 10 wk. We measured the levels of an array of cyto-kines and chemokines in sera from these old mice and from eightyoung (6-mo-old) mice as controls (Table S2). Seven targetswere increased significantly with age: IL-6, G-CSF, IP-10, CXCL1,MIP-1β, IL-17, and chemokine (C-C motif) ligand-11 (CCL-11).JAK inhibitor treatment significantly decreased five of these tar-gets as well as GM-CSF, CXCL9, and IL-15, which were notsignificantly increased with age (Fig. 7A; full results are in TableS2). Notably, IL-6, GM-CSF, G-CSF, IP-10, and CXCL-1 levelswere suppressed by JAK inhibitor both in CM from senescent

Rela

�ve

mRN

A le

vel

0

2

4

6

8

IL1a IL1b IL6 IL8 MCP1 PAI-1

CON

SEN

C

* *

* *

**

B

0%

100%

200%

300%

400%

500%

600%

CON

SEN

Perc

enta

ge c

hang

e

*

0

4

8

12

16

20

IL1a IL1b IL6 IL8 MCP1 PAI-1

CON

SEN

A

Rela

�ve

mRN

A le

vel

*

*

*

*

*

*

THP1 cell migra�on

Fig. 3. Senescent preadipocytes induce preadipocyte inflammation, macrophage migration, and adipose tissue inflammation. (A) Preadipocytes were iso-lated from healthy human subjects and treated with CM collected from control (CON) or senescent (SEN) preadipocytes for 24 h. Then RNA was collected, andreal-time PCR was performed. (B) THP-1 cells were placed in inserts in the top chamber, and CM from control (CON) or senescent (SEN) preadipocytes wasintroduced into the lower chamber. Cell migration was tested following 6 h of exposure using the CyQUANT cell proliferation assay. Results are shown as thefold-change of cell number in each treatment vs. the control blank CM group. (C) Subcutaneous adipose tissue was isolated from healthy human subjects andtreated with CM collected from control (CON) or senescent (SEN) preadipocytes for 24 h. Then RNA was collected, and real-time PCR was performed. Theresults were obtained from six sets of CM from the preadipocytes of six subjects and are expressed as mean ± SEM. *P < 0.05.

E6304 | www.pnas.org/cgi/doi/10.1073/pnas.1515386112 Xu et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

preadipocytes (Fig. 4B) and in serum from aged mice (Fig. 7A).IL-6, G-CSF, and CXCL-1 levels were induced by both cellularsenescence (Fig. 4B) and aging (Fig. 7A). To test whether thedecreased levels of cytokines were related to reduced numbers ofwhite blood cells, we performed complete blood cell counts inthese mice. We observed that 10 wk of treatment with INCB18424did not alter number of peripheral blood immune cells in agedmice (Table S3). Adipose tissue is a major source of the cytokinesin aged animals (44, 45). We found that the expression of severalinflammatory markers and matrix metalloproteinases in adiposetissue, including IL-6, TNF-α, MMP-3, MMP-12, and EGF-likemodule-containing mucin-like hormone receptor-like 1 (Emr1 orF4/80), a macrophage marker, was decreased in the INCB18424-treated group, (Fig. 7B). This decrease in Emr-1 is consistent withthe reduction in senescent CM-induced macrophage diapedesiswe found following JAK inhibition in vitro (Fig. 6B). In addition,we found increased phosphorylated JAK2 (Tyr1007/1008) andSTAT3 (Tyr705) in aged adipose tissue along with p21Cip1. Onlyphosphorylated STAT3 was significantly reduced by JAK inhibitortreatment (Fig. 7 C and D). This finding suggests that STAT3plays a role in age-related adipose tissue inflammation.

JAK Inhibition Enhances Physical Activity in Aged, Frail Mice. Theprevalence of frailty increases in old age. Phenotypic characteristics

of frailty include reduced muscle strength, poor endurance, and lowphysical activity (46). Increased serum IL-6 is associated with de-creased physical activity and frailty in older people (47–50). Wetested whether the decreased abundance of IL-6 and other in-flammatory cytokines in response to JAK1/2 inhibitors is associatedwith enhanced physical function in aged mice. Before treatment,there were no differences in baseline measures of physical activity,hanging endurance, or grip strength between groups of mice ran-domized to treatment with either vehicle or the JAK inhibitor. After10 wk of treatment, administration of the JAK1/2 inhibitor enhancedphysical activity (Fig. 8A), including rearing [reduced rearing is amarker of frailty (51)] (Fig. 8B) and ambulation counts (Fig. 8C).The administration of the JAK1/2 inhibitor also improved hangingendurance (Fig. 8D) and grip strength (Fig. 8E) in these aged mice.Moreover, inhibition of the JAK pathway partially restored the age-related decline in coordination (Fig. 8F). Body weight and food in-take remained similar between the vehicle-treated mice and thosetreated with the JAK inhibitor. Thus, JAK inhibitor treatment en-hanced overall physical function in frail, aged mice.

DiscussionSystemic inflammation increases with advancing age (52). Theexact mechanism for this increase is still not completely un-derstood. Potential mechanisms include a deregulated immune

0

0.4

0.8

1.2

MMP3 MMP12

SEN SEN+JAKi

* *

0

0.4

0.8

1.2

IL6 IL8 MCP1

SP SP+JAKi

Rela

�ve

mRN

A le

vel

D

***

0

0.4

0.8

1.2Veh 0.15μM 0.3μM 0.6μM 1μM

ARe

la�v

e m

RNA

leve

l

*

* **

*

**

**

*

**

*

0

0.4

0.8

1.2

IL6 IL8 MCP1 PAI-1

Veh Jak1 CYT INCB

ERe

la�v

e m

RNA

leve

l

**

***

**

*

**

CON SEN SEN+JAKi

P-STAT3

T-STAT3

GAPDH

F

IL6 IL8 MCP1

Rela

�ve

mRN

A le

vel

C

0

20

40

60

80

100

120

IL-6 GM-CSF G-CSF IP-10 CXCL-1 MIP-1α IL-8 MCP-1 RANTES MCP-3

Cyto

kine

leve

ls (

% o

f SEN

)

CON SEN SEN+JAKi

B

*

*

*

*

*

**

*

*

* *

*

*

*

*

*

*

**

Fig. 4. JAK inhibitor suppresses the SASP in senescent preadipocytes. (A) Irradiation-induced senescent preadipocytes were treated with different con-centrations of JAK inhibitor 1 for 72 h. Then RNA was collected, and real-time PCR was performed. Results are expressed as mean ± SEM; *P < 0.05 comparedwith the vehicle (Veh) group; n = 3. (B) CM was collected from control adipocytes (CON), senescent adipocytes treated with DMSO (SEN), and senescentpreadipocytes treated with 0.6 μM JAK inhibitor 1 (SEN+JAKi). Cytokine protein levels in CM were measured by multiplex assay and were normalized by cellnumbers. Selected cytokines levels are shown as the percentage of SEN for primary preadipocytes from each human subject. Results are expressed as mean ±SEM; *P < 0.05 compared with SEN; n = 6. (C) Irradiation-induced senescent preadipocytes were treated with 0.6 μM JAK inhibitor 1 for 72 h. Then RNA wascollected, and real-time PCR was performed. Results are expressed as mean ± SEM; *P < 0.05; n = 6. (D) Serial passage-induced senescent preadipocytes weretreated with 0.6 μM JAK inhibitor 1 for 72 h. Then RNA was collected, and real-time PCR was performed. Results are expressed as mean ± SEM; *P < 0.05; n = 6.(E) Irradiation-induced senescent preadipocytes were treated with different JAK inhibitors, including JAK inhibitor 1 (Jak1), momelotinib (CYT), and rux-olitinib (INCB), each at 1 μM for 72 h. Then RNA was collected, and real-time PCR was performed. Results are expressed as mean ± SEM; *P < 0.05 comparedwith the vehicle (Veh) group; n = 5. (F) Radiation-induced senescent preadipocytes were treated with 0.6 μM JAK inhibitor 1 for 72 h. Then cell lysates wereprepared. Phosphorylated STAT3 (P-STAT3) (Tyr705), total STAT3, and GAPDH protein abundance are shown. n = 4; representative images are depicted.

Xu et al. PNAS | Published online November 2, 2015 | E6305

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

system (53) or altered redox homeostasis (54). Increased num-bers of T cells and inflammatory macrophages also are observedin aged adipose tissue (44). Our findings support the possibilitythat senescent cells contribute to age-related systemic inflam-mation. Evidence for this notion includes the following points.(i) Senescent cells accumulate in aged adipose and other tissuesin the mouse (40), rat, and human. (ii) Senescent preadipocytesdevelop a proinflammatory SASP that can induce inflammationin adjacent healthy cells and whole adipose tissue. (iii) p16Ink4a-positive senescent cells were the main cause of increased IL-6expression in the stromal vascular fraction of adipose tissue digestsfrom older progeroid mice (21). The stromal vascular fraction is,in turn, the principal site of cytokine release by adipose tissue (44).(iv) Inhibiting the JAK-STAT pathway suppresses the SASP inpreadipocytes and endothelial cells as well as SASP-induced adi-pose tissue inflammation in vitro. We found that JAK inhibitionalso attenuates age-related adipose tissue and systemic inflamma-tion together with frailty in vivo. Collectively, these findings suggestthat targeting senescent cells could be a promising way to alleviateage-related chronic inflammation of adipose tissue.

In this study, we mainly used human primary preadipocytes tostudy cellular senescence and the SASP because fat tissue isa major source of circulating inflammatory cytokines in agedpopulations (44, 45) and because fat tissue is the largest organ inmost humans and preadipocytes thus are perhaps the mostabundant progenitor cell type in humans (39). It is likely that se-nescent cells in other tissues contribute to age-related inflamma-tion as well. Consistent with this possibility, we found that JAKinhibitors decreased the SASP in both preadipocytes (Fig. 6) andendothelial cells (Fig. S6), cell types from two distinct lineages.This finding is consistent with the observation that JAK inhibitorscan reprogram the SASP in senescent tumor cells, leading toimproved antitumor immune responses (37).There are three potential approaches for targeting senescent

cells. One is to prevent them from arising by disabling p16- orp53-related processes or other upstream mechanisms that drivethe generation of cellular senescence. This approach, however, islikely to induce cancer (55). The second is to eliminate senescentcells that already have formed. The third is to blunt the proin-flammatory nature of the SASP. Mounting evidence suggests thatsenescent cells can have both harmful and beneficial effects (56).

0

0.4

0.8

1.2

siCON

siJAK1-a

siJAK1-b

0

0.4

0.8

1.2

siCON

siJAK2-a

siJAK2-b

Rela

�ve

mRN

A le

vel

Rela

�ve

mRN

A le

vel

**

***

*

*

*

*

*

**

*

*

*

*

*

*

*

*

A

B

T-JAK2

T-JAK1

GAPDH

GAPDH

siCON siJAK2-a siJAK2-b

siCON siJAK1-a siJAK1-b

IL6 IL8 MCP1 PAI-1 JAK2

IL6 IL8 MCP1 PAI-1 JAK1

Fig. 5. Knockdown of either JAK1 or JAK2 suppresses the SASP in senescent preadipocytes. (A, Left) Irradiation-induced senescent preadipocytes weretransduced with nontargeting siRNA (siCON) or either of two JAK1 siRNAs (siJAK1-a and siJAK1-b). (B, Left) Irradiation-induced senescent preadipocytes weretransduced with nontargeting siRNA (siCON) or one of two JAK2 siRNAs (siJAK2-a and siJAK2-b). RNA was collected 48 h after transduction, and real-time PCRwas performed to detect mRNA of key SASP components. Results are expressed as mean ± SEM; *P < 0.05 compared with the siCON group; n = 4 for eachgroup. Total JAK1, total JAK2, and GAPDH protein levels are shown also. (A and B, Right) Representative images are shown.

E6306 | www.pnas.org/cgi/doi/10.1073/pnas.1515386112 Xu et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

Therefore, partial suppression of the SASP seems to be a rea-sonable option, particularly when short-term alleviation of age-related dysfunction might be indicated. Several drugs, includingglucocorticoids (57), metformin (58), and rapamycin (59), havebeen found to inhibit the SASP. Inhibition of the JAK/STATpathway appears to be a viable option for the following reasons.The JAK/STAT pathway regulates the expression of a numberof proinflammatory cytokines (24, 43). We found that phos-phorylated JAK1 and -2 levels are increased in aged adiposetissue (Fig. S2B). Furthermore, several JAK inhibitors are be-ing used to treat cancers and hematological and rheumatologicconditions in humans (33, 36). However, few studies have inves-tigated their potential benefits in old age. Here, we found thatthese inhibitors suppress particular SASP components by 40–60%within 3 d, under conditions in which they had little evident ef-fect on nonsenescent cells (Fig. S5). Importantly, JAK inhibitortreatment for 10 wk decreased systemic inflammation and im-proved physical function in frail, aged mice. Consistent with thisresult, a recent human trial using roxolitinib for myelofibrosissuggested that JAK inhibition can increase activity in humansubjects and reduce frailty-like constitutional symptoms relatedto hematological disease (33, 60). Thus, inhibition of the JAKpathway is a potentially promising strategy for alleviating age-relateddysfunction.IL-6, a major SASP component, signals through the JAK/

STAT pathway (61) and has been shown to be critical in stabilizingcellular senescence (62). Therefore, we tested whether JAK in-hibition reduces cellular senescence. However, we did not observea significant alteration in senescent cell abundance in either invitro or in vivo models. This lack of effect could be related to thepossibility that we treated cells or aged mice with JAK inhibitor

once cellular senescence was established. Because established se-nescent cells already are present in older individuals, JAK in-hibition could prove to be a way to alleviate frailty or cellularsenescence-related chronic diseases in late life, points that warrantfurther preclinical and clinical study.Roxolitinib can have side effects, including anemia and

thrombocytopenia (33, 63), in humans who have myelofibrosis.We did not observe an alteration of red blood cell numbers orhemoglobin levels in old mice treated with roxolitinib (Table S3),although platelet counts tended to be reduced (not significantly;n = 9). Roxolitinib treatment had minimal effect on the numbersof peripheral blood immune cells in aged mice (Table S3). Onereason is that roxolitinib is a selective JAK1/2 inhibitor that hasmuch less effect on JAK3 and TYK2 (64), the two JAK familymembers associated with immune cell generation and function(28–31). Notably, in humans treated with roxolitinib for 24 wkfor myelofibrosis, major infections were no more frequent thanin the placebo group: only one case of clostridial infection wasfound in the ruxolitinib group, and one case of staphylococcalinfection was found in the placebo group (33). More work isneeded to uncover potential side effects of JAK1/2 inhibition inolder populations.Possibly, the alleviation of frailty by JAK1/2 inhibition that we

observed occurs through effects on cells other than senescent cells,such as immune cells or the brain. However, we did not find al-tered peripheral white blood cell counts (Table S3), and at leastone test of brain function remained unaffected in aged micetreated with roxolitinib (Fig. S7). In addition, roxolitinib treatmentdid not alter the physical activity of 6-mo-old mice (Fig. S8).Other mechanisms for the effects of JAK1/2 inhibition on frailtymerit future investigation. However, it appears reasonable to

Rela

�ve

mRN

A le

vel

B

0%

100%

200%

300%

400%

500%

600%

CON

SEN

SEN+JAKi

Perc

enta

ge c

hang

e

* *

0

4

8

12

16

20

IL1a IL1b IL6 IL8 MCP1 PAI-1

CON

SEN

SEN+JAKi

A

Rela

�ve

mRN

A le

vel

*

*

**

*

**

**

THP1 cell migra�on0

0.5

1

1.5

2

IL6 IL8 MCP1

SEN-CM

SEN-CM+JAKi

C

Fig. 6. JAK inhibition decreases inflammation induced by senescent cells. (A) Preadipocytes were isolated from healthy human subjects and were treated for24 h with CM collected from control (CON) preadipocytes, senescent (SEN) preadipocytes, or senescent preadipocytes treated with JAK inhibitor 1 (SEN+JAKi).Then RNA was collected, and real-time PCR was performed. (B) THP-1 cells were placed in inserts in the top chamber, and CM from different groups wasintroduced into the lower chamber. Cell migration was tested following 6 h of exposure using the CyQUANT cell-proliferation assay. Results are shown as thefold-change of cell number in each treatment vs. the control CM group. (C) Preadipocytes isolated from healthy human subjects were treated for 24 h withCM collected from senescent preadipocytes in the presence of DMSO (SEN-CM) or 0.6 μM JAK inhibitor (SEN-CM+JAKi). Then RNA was collected, and real-timePCR was performed. Results are expressed as mean ± SEM; n = 4.

Xu et al. PNAS | Published online November 2, 2015 | E6307

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

hypothesize that cellular senescence is one potential mechanismassociated with aging-related adipose tissue inflammation. Fattissue inflammation contributes to elevated circulating cytokinesin old age. These increased cytokines, in turn, are associated withor can cause frailty in older humans and experimental animals.Furthermore, genetic or pharmacological clearance of senescent

cells in older mice alleviated frailty (21, 22), much as JAK in-hibitors did in this study and in humans with myeloproliferativedisorders (63).In summary, our study suggests that the JAK1/2 pathway plays

an important role in the SASP and could be a target for futureinterventions to alleviate age-related dysfunction.

0

20

40

60

80

100

Stat3 P/T Jak2 P/T p21/GAPDH

Young

Old Veh

Old INCB

p21

P-Jak2

GAPDH

T-Jak2

P-Stat3

T-Stat3

Arbi

trar

y un

its

*** *

Young Old Veh Old INCB

A

C D

0

0.4

0.8

1.2

1.6

IL6 TNFa MMP3 MMP12 PAI-1 Emr1

Veh

INCB

Rela

�ve

mRN

A le

vel

**

* *

*

B

#

*

*

*

*

*

N/A

# #

#

#

**

Fig. 7. JAK inhibition decreases systemic and adipose tissue inflammation in old mice. Twenty-four–month–old male mice were treated with vehicle (OldVeh) or ruxolitinib (Old INCB) for 10 wk. Six-month-old male mice (Young) were used as young control group. (A) Serum levels of cytokines were measured bymultiplex assay. Selected cytokines levels are shown as the fold change over the average level of Young group. Results are expressed as mean ± SEM; *P < 0.05compared with Young; #P < 0.05 compared with Old Veh; n = 8. (B) RNA of inguinal adipose tissue from old mice was collected, and real-time PCR wasperformed. Results are expressed as mean ± SEM; *P < 0.05; n = 8. (C) Cell lysates were prepared from inguinal adipose tissue. The abundance of phos-phorylated JAK2 (P-JAK2) (Tyr1007/1008), total JAK2, P-STAT3 (Tyr705), total STAT3, p21, and GAPDH protein abundance is shown. (D) These signals werequantified by densitometry using ImageJ. The ratios of phosphorylated to total STAT3 (STAT3 P/T), phosphorylated to total JAK2 (Jak2 P/T), and total p21 levelto GAPDH (P21/GAPDH) are expressed as mean ± SEM; *P < 0.05; n = 4.

0

10,000

20,000

30,000

40,000

50,000

0 10 WEEKS

Ac�v

ity (c

ount

s)

CON-DayINCB-DayCON-NightINCB-Night

0

3,000

6,000

9,000

12,000

15,000

0 10 WEEKS

Rear

ing

(cou

nts)

A B C

ED F

*

0

5,000

10,000

15,000

20,000

25,000

0 10 WEEKS

Ambu

la�o

n (c

ount

s)

**

0

200

400

600

800

1000

Baseline 6 weeks

Hang

ing

Tim

e* B

ody

Wei

ght (

Sec*

g)

CON

INCB

*

0

0.2

0.4

0.6

0.8

1

1.2

Baseline 7 weeks

Grip

Str

engt

h (N

)

CON

INCB

*

0

5

10

15

20

8 weeks

Spee

d (r

pm)

Young

CON

INCB

**

Fig. 8. JAK inhibition enhances physical function in aged, frail mice. Twenty-four–month–old male mice were monitored using CLAMS before and after 10 wk ofvehicle (CON) or ruxolitinib (INCB) treatment. (A–C) Total activity (A), rearing activity (B), and ambulation (C) were analyzed. Results are shown for both light (Day)and dark (Night) cycles. (D and E) Hanging test (D) and grip strength (E) tests also were performed on these animals. The RotaRod test was performed 8 wk aftertreatment in parallel with nine 6-mo-old male mice (Young). Results are expressed as mean ± SEM; *P < 0.05; n = 8 for A–C; n = 9 for D–F.

E6308 | www.pnas.org/cgi/doi/10.1073/pnas.1515386112 Xu et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

MethodsCell Culture and Reagents. Primary human preadipocytes were isolated fromhealthy lean kidney donors aged 39 ± 3.3 y with a body mass index of 26.6 ±0.9 (mean ± SEM). Primary preadipocytes also were isolated from five young(31 ± 5 y) and five old (71 ± 2 y) healthy male volunteers for SABG assay. Theprotocol was approved by the Mayo Clinic Foundation Institutional ReviewBoard for Human Research. Informed consent was obtained from all humansubjects. The preadipocyte isolation procedure and steps taken to ensureculture purity have been described previously (65). To induce senescence,preadipocytes were subjected to 10 Gy of cesium radiation. They were se-nescent by 20 d after irradiation, with little cell growth and more than 70% ofcells exhibiting SABG positivity. Preadipocytes also were induced to becomesenescent by 27–30 serial passages, at which point cell proliferation was at-tenuated and more than 80% of cells exhibited SABG positivity. Preadipocyteswere isolated from Brown Norway rats as previously described (66). HUVECswere purchased from ATCC. CYT387 (CAS 1056634-68-4) and ruxolitinib(INCB18424, CAS 941678-49-5) were purchased from ChemieTek.

SABG Assay. Cellular SABG activity was assayed as previously described (21). Inbrief, primary preadipocytes or fat tissues were washed three times with PBSand then were fixed for 5 min in PBS containing 2% (vol/vol) formaldehyde(Sigma-Aldrich) and 0.25% glutaraldehyde (Sigma-Aldrich). Following fixa-tion, cells or tissues were washed three times with PBS before being in-cubated in SABG activity solution (pH 6.0) at 37 °C for 16–18 h. Theenzymatic reaction was stopped by washing cells or tissues three to fivetimes with ice-cold PBS. Fluorescence microscopy (Nikon Eclipse Ti) was usedfor imaging. About 8–10 images were taken from random fields in eachsample. DAPI (Life Technologies) was used to stain nuclei for cell counting.

CM. CM were made by exposing cells to RMPI 1640 containing 1 mM sodiumpyruvate, 2 mM glutamine, MEM (minimum essential medium) vitamins,MEM nonessential amino acids, and antibiotic (Life Technologies). Non-proliferating healthy and senescent preadipocytes were washed three timeswith PBS and then were cultured with CM for 24 h. For JAK inhibitortreatment, senescent preadipocytes were treated with JAK inhibitor orDMSO for 48 h, washed three times with PBS, and cultured with CM con-taining JAK inhibitor or DMSO for another 24 h.

Multiplex Protein Analyses. Luminex xMAP technology was used to quantify cy-tokines, chemokines, and growth factors. The multiplexing analysis was per-formed using the Luminex 100 system (Luminex) by Eve Technologies Corp.Human and mouse multiplex kits were from Millipore. More than 80% of thetargetswerewithin thedetectable range (signals fromall samples arehigher thanthe lowest standard). Undetectable targets were excluded from Tables S1 and S2.

Real-Time PCR. RNA from tissues or cells was extracted using TRIzol (LifeTechnologies) and was reverse transcribed to cDNA using the M-MLV ReverseTranscriptase kit (Life Technologies) following the manufacturer’s instruc-tions. Real-time PCR was performed using TaqMan fast advanced master mix(Life Technologies). The probes and primers were purchased from LifeTechnologies. TATA-binding protein was used as an internal control.

siRNA Knockdown. Senescent cells were transfected with 30 nM siRNA usingLipofectamine RNAiMAX (Life Technologies) following the manufacturer’s in-structions. The cells were collected 48 h after transfection. All siRNAs wereobtained from Life Technologies. The siRNA sequences were as follows: siJAK1-a:sense 5′-GGC AUG CCG UAU CUC UCC Utt-3′; antisense 5′-AGG AGA GAU ACGGCA UGC Ctg-3′; siJAK1-b: sense 5′-GGU UCU AUU UCA CCA AUU Gtt-3′; anti-sense 5′- CAA UUG GUG AAA UAG AAC Ctc-3′; siJAK2-a: sense 5′- GGU GUA UCUUUA CCA UUC Ctt-3′; antisense 5′-GGA AUG GUA AAG AUA CAC Ctg-3′; siJAK2-b:sense 5′-GGA GUU UGU AAA AUU UGG Att-3′; antisense 5′-UCC AAA UUUUAC AAA CUC Ctg-3′. Two nontargeting negative control siRNAs (catalog nos.AM4635 and AM4615) also were purchased from Life Technologies.

Immunostaining. Cells were fixed with 4% (vol/vol) paraformaldehyde for10 min and were permeabilized using PBS containing 0.25% Triton X-100 for10 min. After being washed with PBS, cells were blocked with PBS andTween-20 (PBST) containing 1% BSA for 30 min. Cells were incubated withprimary antibodies in 1% BSA in PBST overnight at 4 °C and then with thesecondary fluorescent antibodies (Life Technologies) in the same bufferfor 1 h in the dark. DAPI (Life Technologies) was used to stain nuclei for cellcounting. Fluorescence microscopy (Nikon Eclipse Ti) was used for imaging.

Paraffin rat fat sections were deparaffinized with Histoclear (NationalDiagnostics) and hydrated by ethanol gradient. Antigen was retrieved by

incubation in 0.01 M citrate buffer (pH 6.0) at 95 °C for 10 min. Slides wereplaced in blocking buffer (normal goat serum 1:60 in 0.1% BSA in PBS) for60 min at room temperature. Samples were blocked further with Avidin/Biotin (Vector Laboratories) for 15 min each at room temperature. Primaryantibody was applied overnight at 4 °C in the blocking buffer. Then slideswere incubated for 30 min with biotinylated, anti-rabbit secondary antibody(Vector Laboratories). Finally, Fluorescein Avidin DCS (Vector Laboratories) wasapplied for 20 min. Samples were mounted in VECTRASHIELD DAPI-containingmedium (Vector Laboratories) and were imaged using a LSM 780 Zeiss con-focal microscope under the ×100 objective. γH2AX antibody was purchasedfrom Cell Signaling, and Ki-67 antibody was purchased from Abcam.

Western Immunoanalysis. Cells or tissues were homogenized in cell lysis buffer(Cell Signaling) with protease inhibitors (Sigma). The lysate then was clearedby centrifugation (16,000 × g for 10 min at 4 °C). Total protein content wasdetermined using Coomassie Plus reagents (Pierce). Proteins were loadedinto each lane on a 4–20% gradient SDS/PAGE gel and transferred toimmunoblot PVDF membranes (Bio-Rad). Signals were developed with HRP-conjugated secondary antibodies and SuperSignal West Pico Chemilumi-nescent Substrate (Pierce). The p21 antibody was purchased from SantaCruz. All other antibodies were purchased from Cell Signaling.

Macrophage Migration Assay. The THP-1 cell linewas purchased from theATCC.Migration assays were performed using Transwell polycarbonate membraneinserts with a 5-μm pore diameter (Corning). Three hundred microliters of CMcollected from cells or blank CM (without exposure to cells) were introducedinto the bottom chambers of the plates in triplicate. THP-1 cells (105)resuspended in blank CM were plated into the inserts. THP-1 cells were allowedto migrate into the bottom chamber for 6 h at 37 °C. After 6 h, the number ofcells that had migrated into the lower chamber was quantified using theCyQUANT cell proliferation assay kit (Life Technologies) according to themanufacturer’s instructions. The number of cells migrating into blank CM wasconsidered as background and was subtracted from the experimental groups.

Mice and Drug Treatments. Twenty-four-month-old C57BL/6 male mice wereobtained from the National Institute on Aging and were maintained in apathogen-free facility at 24 °C under a 12 h/12 h light/dark cycle with freeaccess to food (standard mouse diet; Lab Diet 5053) and water. For drugtreatment, ruxolitinib was dissolved in DMSO and then was ground and mixedwith food. In addition to regular food, each mouse was fed a small amount offood (0.5 g) containing ruxolitinib (60 mg/kg body weight) or DMSO daily.During the 10 wk of treatment, all mice consumed the drug-containing foodcompletely every day. The experimental procedure was approved by the In-stitutional Animal Care and Use Committee at Mayo Clinic.

Measurements of Physical Function. Forelimb grip strength [N (newton)] wasdetermined using a Grip Strength Meter from Columbus Instruments follow-ing a 3-d acclimation. Results were averaged from six trials. For the hangingtest, mice were placed on a 2-mm-thick metal wire, which was placed 35 cmabove a padded surface. Mice were allowed to grab the wire only with theirforelimbs. The hanging time was normalized to the body weight as hangingduration in seconds × body weight in grams. The results were averaged fromthree trials for each mouse. Coordination was assessed using an acceleratingRotaRod system (TSE Systems). Mice were trained on RotaRod at a speed of4 rpm for 200 s for 3 d. On the test day, mice were placed onto the stationaryRotaRod. The RotaRod then was started at 4 rpm and accelerated to 40 rpmover a 5-min interval. The speed when the mouse dropped from theRotaRod was recorded. Results were averaged from three trials.

Comprehensive Laboratory Animal Monitoring System. In a subset of eightmice per group, the habitual ambulatory, rearing, and total activity, oxygenconsumption (VO2), and carbon dioxide production (VCO2) of individual micewere monitored over a 24-h period (12 h light/12 h dark) using a Compre-hensive Laboratory Animal Monitoring System (CLAMS) equipped with anOxymax Open Circuit Calorimeter System (Columbus Instruments). Ambula-tory, rearing, and total activity was summed and analyzed for light and darkperiods under fed conditions. The VO2 and VCO2 values were used to cal-culate the respiratory exchange ratio (RER) and VO2. RER values were usedto determine the basal metabolic rate (in kilocalories per kilogram per hour).

Statistical Analysis. Analyses of differences between groups were performedby Student’s t test, one-way ANOVA, or repeated-measures ANOVA whereappropriate. Values are presented as mean ± SEM with P < 0.05 consideredto be significant.

Xu et al. PNAS | Published online November 2, 2015 | E6309

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

ACKNOWLEDGMENTS. We thank Dr. Ann Oberg for statistical support andcolleagues in the Robert and Arlene Kogod Center on Aging for com-ments and discussion. This work was supported by NIH Grants DK50456 (to

M.D.J.), AG041122 (to J.L.K.), and AG013925 (to J.L.K.). M.X. received aGlenn/American Federation for Aging Research Postdoctoral Fellowshipfor Translational Research on Aging.

1. Tchkonia T, Zhu Y, van Deursen J, Campisi J, Kirkland JL (2013) Cellular senescence andthe senescent secretory phenotype: Therapeutic opportunities. J Clin Invest 123(3):966–972.

2. Michaud M, et al. (2013) Proinflammatory cytokines, aging, and age-related diseases.J Am Med Dir Assoc 14(12):877–882.

3. López-Otín C, Blasco MA, Partridge L, Serrano M, Kroemer G (2013) The hallmarks ofaging. Cell 153(6):1194–1217.

4. Harris TB, et al. (1999) Associations of elevated interleukin-6 and C-reactive proteinlevels with mortality in the elderly. Am J Med 106(5):506–512.

5. Bruunsgaard H, Andersen-Ranberg K, Hjelmborg Jv, Pedersen BK, Jeune B (2003) El-evated levels of tumor necrosis factor alpha and mortality in centenarians. Am J Med115(4):278–283.

6. Pai JK, et al. (2004) Inflammatory markers and the risk of coronary heart disease inmen and women. N Engl J Med 351(25):2599–2610.

7. Howren MB, Lamkin DM, Suls J (2009) Associations of depression with C-reactiveprotein, IL-1, and IL-6: A meta-analysis. Psychosom Med 71(2):171–186.

8. Schetter AJ, Heegaard NH, Harris CC (2010) Inflammation and cancer: InterweavingmicroRNA, free radical, cytokine and p53 pathways. Carcinogenesis 31(1):37–49.

9. Spranger J, et al. (2003) Inflammatory cytokines and the risk to develop type 2 di-abetes: Results of the prospective population-based European Prospective In-vestigation into Cancer and Nutrition (EPIC)-Potsdam Study. Diabetes 52(3):812–817.

10. McGeer PL, McGeer EG (2004) Inflammation and the degenerative diseases of aging.Ann N Y Acad Sci 1035:104–116.

11. McGeer PL, McGeer EG (2004) Inflammation and neurodegeneration in Parkinson’sdisease. Parkinsonism Relat Disord 10(Suppl 1):S3–S7.

12. Campisi J, d’Adda di Fagagna F (2007) Cellular senescence: When bad things happento good cells. Nat Rev Mol Cell Biol 8(9):729–740.

13. Dimri GP, et al. (1995) A biomarker that identifies senescent human cells in cultureand in aging skin in vivo. Proc Natl Acad Sci USA 92(20):9363–9367.

14. Jeyapalan JC, Ferreira M, Sedivy JM, Herbig U (2007) Accumulation of senescent cellsin mitotic tissue of aging primates. Mech Ageing Dev 128(1):36–44.

15. Paradis V, et al. (2001) Replicative senescence in normal liver, chronic hepatitis C, andhepatocellular carcinomas. Hum Pathol 32(3):327–332.

16. Wang C, et al. (2009) DNA damage response and cellular senescence in tissues ofaging mice. Aging Cell 8(3):311–323.

17. Melk A, et al. (2003) Cell senescence in rat kidneys in vivo increases with growth andage despite lack of telomere shortening. Kidney Int 63(6):2134–2143.

18. Erusalimsky JD, Kurz DJ (2005) Cellular senescence in vivo: Its relevance in ageing andcardiovascular disease. Exp Gerontol 40(8-9):634–642.

19. Coppé JP, et al. (2008) Senescence-associated secretory phenotypes reveal cell-non-autonomous functions of oncogenic RAS and the p53 tumor suppressor. PLoS Biol6(12):2853–2868.

20. Coppé JP, et al. (2010) A human-like senescence-associated secretory phenotype isconserved in mouse cells dependent on physiological oxygen. PLoS One 5(2):e9188.

21. Baker DJ, et al. (2011) Clearance of p16Ink4a-positive senescent cells delays ageing-associated disorders. Nature 479(7372):232–236.

22. Zhu Y, et al. (2015) The Achilles’ heel of senescent cells: From transcriptome to se-nolytic drugs. Aging Cell 14(4):644–658.

23. Meyer SC, Levine RL (2014) Molecular pathways: Molecular basis for sensitivity andresistance to JAK kinase inhibitors. Clin Cancer Res 20(8):2051–2059.

24. Yu H, Pardoll D, Jove R (2009) STATs in cancer inflammation and immunity: A leadingrole for STAT3. Nat Rev Cancer 9(11):798–809.

25. Richard AJ, Stephens JM (2011) Emerging roles of JAK-STAT signaling pathways inadipocytes. Trends Endocrinol Metab 22(8):325–332.

26. Rane SG, Reddy EP (1994) JAK3: A novel JAK kinase associated with terminal differ-entiation of hematopoietic cells. Oncogene 9(8):2415–2423.

27. Ghoreschi K, Laurence A, O’Shea JJ (2009) Janus kinases in immune cell signaling.Immunol Rev 228(1):273–287.

28. Russell SM, et al. (1995) Mutation of Jak3 in a patient with SCID: Essential role of Jak3in lymphoid development. Science 270(5237):797–800.

29. Nosaka T, et al. (1995) Defective lymphoid development in mice lacking Jak3. Science270(5237):800–802.

30. Karaghiosoff M, et al. (2000) Partial impairment of cytokine responses in Tyk2-deficient mice. Immunity 13(4):549–560.

31. Shimoda K, et al. (2000) Tyk2 plays a restricted role in IFN alpha signaling, although itis required for IL-12-mediated T cell function. Immunity 13(4):561–571.

32. Harrison C, et al. (2012) JAK inhibition with ruxolitinib versus best available therapyfor myelofibrosis. N Engl J Med 366(9):787–798.

33. Verstovsek S, et al. (2012) A double-blind, placebo-controlled trial of ruxolitinib formyelofibrosis. N Engl J Med 366(9):799–807.

34. Pardanani A, et al. (2013) Safety and efficacy of CYT387, a JAK1 and JAK2 inhibitor, inmyelofibrosis. Leukemia 27(6):1322–1327.

35. Fridman JS, et al. (2010) Selective inhibition of JAK1 and JAK2 is efficacious in rodentmodels of arthritis: Preclinical characterization of INCB028050. J Immunol 184(9):5298–5307.

36. Ghoreschi K, et al. (2011) Modulation of innate and adaptive immune responses bytofacitinib (CP-690,550). J Immunol 186(7):4234–4243.

37. Toso A, et al. (2014) Enhancing chemotherapy efficacy in Pten-deficient prostate tumorsby activating the senescence-associated antitumor immunity. Cell Reports 9(1):75–89.

38. Tchkonia T, et al. (2013) Mechanisms and metabolic implications of regional differ-ences among fat depots. Cell Metab 17(5):644–656.

39. Tchkonia T, et al. (2010) Fat tissue, aging, and cellular senescence. Aging Cell 9(5):667–684.

40. Stout MB, et al. (2014) Growth hormone action predicts age-related white adipose tissuedysfunction and senescent cell burden in mice. Aging (Albany, NY) 6(7):575–586.

41. Lumeng CN, et al. (2011) Aging is associated with an increase in T cells and in-flammatory macrophages in visceral adipose tissue. J Immunol 187(12):6208–6216.

42. Villaret A, et al. (2010) Adipose tissue endothelial cells from obese human subjects:Differences among depots in angiogenic, metabolic, and inflammatory gene ex-pression and cellular senescence. Diabetes 59(11):2755–2763.

43. Dauer DJ, et al. (2005) Stat3 regulates genes common to both wound healing andcancer. Oncogene 24(21):3397–3408.

44. Wu D, et al. (2007) Aging up-regulates expression of inflammatory mediators inmouse adipose tissue. J Immunol 179(7):4829–4839.

45. Starr ME, Evers BM, Saito H (2009) Age-associated increase in cytokine productionduring systemic inflammation: Adipose tissue as a major source of IL-6. J Gerontol ABiol Sci Med Sci 64(7):723–730.

46. Fried LP, et al.; Cardiovascular Health Study Collaborative Research Group (2001) Frailty inolder adults: Evidence for a phenotype. J Gerontol A Biol Sci Med Sci 56(3):M146–M156.

47. Ferrucci L, et al. (1999) Serum IL-6 level and the development of disability in olderpersons. J Am Geriatr Soc 47(6):639–646.

48. Reuben DB, et al. (2002) Peripheral blood markers of inflammation predict mortalityand functional decline in high-functioning community-dwelling older persons. J AmGeriatr Soc 50(4):638–644.

49. Cohen HJ, Pieper CF, Harris T, Rao KM, Currie MS (1997) The association of plasma IL-6levels with functional disability in community-dwelling elderly. J Gerontol A Biol SciMed Sci 52(4):M201–M208.

50. Reuben DB, Judd-Hamilton L, Harris TB, Seeman TE; MacArthur Studies of SuccessfulAging (2003) The associations between physical activity and inflammatory markers inhigh-functioning older persons: MacArthur Studies of Successful Aging. J Am GeriatrSoc 51(8):1125–1130.

51. Whitehead JC, et al. (2014) A clinical frailty index in aging mice: Comparisons withfrailty index data in humans. J Gerontol A Biol Sci Med Sci 69(6):621–632.

52. Grunnet L, Poulsen P, Klarlund Pedersen B, Mandrup-Poulsen T, Vaag A (2006) Plasmacytokine levels in young and elderly twins: Genes versus environment and relation toin vivo insulin action. Diabetologia 49(2):343–350.

53. Shaw AC, Goldstein DR, Montgomery RR (2013) Age-dependent dysregulation ofinnate immunity. Nat Rev Immunol 13(12):875–887.

54. Chung HY, et al. (2009) Molecular inflammation: Underpinnings of aging and age-related diseases. Ageing Res Rev 8(1):18–30.

55. Hanahan D, Weinberg RA (2011) Hallmarks of cancer: The next generation. Cell144(5):646–674.

56. Coppé JP, Desprez PY, Krtolica A, Campisi J (2010) The senescence-associated secre-tory phenotype: The dark side of tumor suppression. Annu Rev Pathol 5:99–118.

57. Laberge RM, et al. (2012) Glucocorticoids suppress selected components of thesenescence-associated secretory phenotype. Aging Cell 11(4):569–578.

58. Moiseeva O, et al. (2013) Metformin inhibits the senescence-associated secretoryphenotype by interfering with IKK/NF-κB activation. Aging Cell 12(3):489–498.

59. Laberge RM, et al. (2015) MTOR regulates the pro-tumorigenic senescence-associatedsecretory phenotype by promoting IL1A translation. Nat Cell Biol 17(8):1049–1061.

60. Tefferi A, Litzow MR, Pardanani A (2011) Long-term outcome of treatment withruxolitinib in myelofibrosis. N Engl J Med 365(15):1455–1457.

61. Heinrich PC, et al. (2003) Principles of interleukin (IL)-6-type cytokine signalling and itsregulation. Biochem J 374(Pt 1):1–20.

62. Kuilman T, et al. (2008) Oncogene-induced senescence relayed by an interleukin-dependent inflammatory network. Cell 133(6):1019–1031.

63. Verstovsek S, et al. (2010) Safety and efficacy of INCB018424, a JAK1 and JAK2 in-hibitor, in myelofibrosis. N Engl J Med 363(12):1117–1127.

64. Quintás-Cardama A, et al. (2010) Preclinical characterization of the selective JAK1/2inhibitor INCB018424: Therapeutic implications for the treatment of myeloprolifer-ative neoplasms. Blood 115(15):3109–3117.

65. Tchkonia T, et al. (2005) Abundance of two human preadipocyte subtypes with dis-tinct capacities for replication, adipogenesis, and apoptosis varies among fat depots.Am J Physiol Endocrinol Metab 288(1):E267–E277.

66. Cartwright MJ, et al. (2010) Aging, depot origin, and preadipocyte gene expression.J Gerontol A Biol Sci Med Sci 65(3):242–251.

67. Lee BY, et al. (2006) Senescence-associated beta-galactosidase is lysosomal beta-galactosidase. Aging Cell 5(2):187–195.

E6310 | www.pnas.org/cgi/doi/10.1073/pnas.1515386112 Xu et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0