Deletion of the gene Pip4k2c, a novelphosphatidylinositol kinase, results in hyperactivationof the immune systemHyeseok Shima,b,c,1, Chuan Wud, Shivan Ramsamooja,b, Kaitlyn N. Boscha,b, Zuojia Chend, Brooke M. Emerlinga,b,Jihye Yuna,b, Hui Liue, Rayman Choo-Winga,b, Zhiwei Yanga,b, Gerburg M. Wulfe, Vijay Kumar Kuchrood,and Lewis C. Cantleya,b,2

aMeyer Cancer Center, Weill Cornell Medical College, New York, NY 10065; bDepartment of Medicine, Weill Cornell Medical College, New York, NY 10065;cBiological and Biomedical Sciences Graduate Program, Harvard Medical School, Boston, MA 02115; dEvergrande Center for Immunologic Diseases, HarvardMedical School and Brigham and Women’s Hospital, Boston, MA 02115; and eDivision of Hematology/Oncology, Beth Israel Deaconess Medical Center,Boston, MA 02115

Contributed by Lewis C. Cantley, May 6, 2016 (sent for review January 28, 2016; reviewed by Richard A. Flavell and Robin F. Irvine)

Type 2 phosphatidylinositol-5-phosphate 4-kinase (PI5P4K) convertsphosphatidylinositol-5-phosphate to phosphatidylinositol-4,5-bisphos-phate. Mammals have three enzymes PI5P4Kα, PI5P4Kβ, and PI5P4Kγ,and these enzymes have been implicated in metabolic control,growth control, and a variety of stress responses. Here, we showthat mice with germline deletion of type 2 phosphatidylinositol-5-phosphate 4-kinase gamma (Pip4k2c), the gene encoding PI5P4Kγ,appear normal in regard to growth and viability but have increasedinflammation and T-cell activation as they age. Immune cell infiltratesincreased in Pip4k2c−/− mouse tissues. Also, there was an increase inproinflammatory cytokines, including IFNγ, interleukin 12, and inter-leukin 2 in plasma of Pip4k2c−/− mice. Pip4k2c−/− mice had an in-crease in T-helper-cell populations and a decrease in regulatory T-cellpopulations with increased proliferation of T cells. Interestingly,mammalian target of rapamycin complex 1 (mTORC1) signalingwas hyperactivated in several tissues from Pip4k2c−/− mice andtreating Pip4k2c−/− mice with rapamycin reduced the inflamma-tory phenotype, resulting in a decrease in mTORC1 signaling intissues and a decrease in proinflammatory cytokines in plasma.These results indicate that PI5P4Kγ plays a role in the regulationof the immune system via mTORC1 signaling.

PIP4K2C | mTORC1 | autoimmunity | PI5P4K | inflammation

Type 2 phosphatidylinositol-5-phosphate 4-kinase (PI5P4K) con-verts phosphatidylinositol-5-phosphate to phosphatidylinositol-

4,5-bisphosphate. Mammals have three genes, PIP4K2A, PIP4K2B,and PIP4K2C that encode the enzymes PI5P4Kα, PI5P4Kβ, andPI5P4Kγ, respectively.All three isoforms are highly expressed in brain, whereas their

relative expressions in other tissues vary. PI5P4Kα is highlyexpressed in spleen and the peripheral blood. PI5P4Kβ is highlyexpressed in muscle, whereas PI5P4Kγ is highly expressed inkidney (1). In kidney, PI5P4Kγ is mostly detected in the cortexand the outer medulla (1). All tissues examined appear to expressat least one isoform of PI5P4K. At the cellular level, PI5P4Ks arefound in several organelles, including plasma membrane, cytosol,nucleus, and vesicular compartments (2, 3). It is not simple todefine unique compartmentalization of the individual enzymesbecause all three isoforms can homodimerize or heterodimerizewith each other. At the sequence level, PI5P4Kα and PI5P4Kβare more homologous to each other than either is to PI5P4Kγ.Also, PI5P4Kγ is only about 1% as active as the other isoforms,raising the possibility that its major role may be to localize orregulate the activities of PI5P4Kα and PI5P4Kβ.Germline deletion of both alleles of Pip4k2a or Pip4k2b in

mice results in healthy mice that live a normal life span. ThePip4k2b−/− mice have increased insulin sensitivity and are pro-tected from obesity, insulin resistance, and type 2 diabetes whenplaced on a high-fat diet (4). Also, crossing the Pip4k2b−/− mice

with Trp53−/− mice results in early embryonic lethality for thesubset of embryos that are Pip4k2b−/−, Trp53−/−, indicating asynthetic lethality relationship between these genes. In contrast,Pip4k2a−/− mice do not exhibit any of the phenotypes observed inthe Pip4k2b−/− mice (5). They are not protected from obesity orinsulin resistance, they do not exhibit a synthetic lethality re-lationship with Trp53, and in all ways examined, they resemblewild-type mice (5). However, germline deletion of one allele ofPip4k2a in the context of germline Pip4k2b−/− causes enhance-ment of the phenotypes of the Pip4k2b−/− mice. Deletion of bothalleles of Pip4k2a and Pip4k2b did not have any observable effecton embryonic development up until the time of birth, but resultedin perinatal lethality of all pups. These results indicate that thesegenes do not play a major role in embryonic development andthat Pip4k2a provides a backup for Pip4k2b that becomes criticalat the time of birth.

Significance

The mammalian target of rapamycin complex 1 (mTORC1) sig-naling pathway is an important facet of the immune system,including that it regulates T-cell differentiation and activation.Here, we report that type 2 phosphatidylinositol-5-phosphate4-kinase gamma (protein, PI5P4Kγ; gene, PIP4K2C) plays a rolein the regulation of the immune system by manipulatingmTORC1 signaling. These results suggest that the SNP at thePIP4K2C locus (rs1678542) in human patients with autoimmunitymight cause a decrease in PI5P4Kγ expression and thereby anincrease in mTORC1 signaling. In addition, these results implythat inhibition of mTORC1 would be beneficial to these patients.These studies also suggest that agents that inhibit PIP4K2Cfunction could be useful to enhance cancer immunotherapy.

Author contributions: H.S., C.W., and L.C.C. designed research; H.S., C.W., S.R., K.N.B., Z.C.,J.Y., H.L., R.C.-W., and Z.Y. performed research; B.M.E., G.M.W., and V.K.K. contributednew reagents/analytic tools; H.S. and C.W. analyzed data; and H.S. and L.C.C. wrotethe paper.

Reviewers: R.A.F., Yale School of Medicine, Howard Hughes Medical Institute; and R.F.I.,Cambridge University.

Conflict of interest statement: L.C.C. owns equity in, receives compensation from, andserves on the board of directors and scientific advisory board of Agios Pharmaceuticals.Agios Pharmaceuticals is identifying metabolic pathways of cancer cells and developingdrugs to inhibit such enzymes to disrupt tumor cell growth and survival. L.C.C. ownsequity in, receives compensation from, and serves on the scientific advisory board of PetraPharmaceuticals, a company that develops targeted therapies for cancer treatment. Inaddition, Petra Pharmaceuticals will be providing funds to support research in the L.C.C.laboratory, although the research described in this paper predated the existence of thiscollaborative agreement.1Present address: Petra Pharmaceuticals, New York, NY 10016.2To whom correspondence should be addressed. Email: lcantley@med.cornell.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1600934113/-/DCSupplemental.

7596–7601 | PNAS | July 5, 2016 | vol. 113 | no. 27 www.pnas.org/cgi/doi/10.1073/pnas.1600934113

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

0

Whereas less is known about PI5P4Kγ, it has been linked tothe mammalian target of rapamycin signaling complex (mTORC)and to cellular immunity. Mackey et al. argued that PI5P4Kγ wasnegatively regulated by mTORC1 through direct phosphorylationat serine 324 and serine 328 (6). On the other hand, it wasreported that knocking out the only PI5P4K isoform in Drosophilaresulted in lower body weight and that this correlated with de-creased mTORC1 signaling (7). Of particular interest, multiplestudies have shown a link between a SNP (rs1678542) in thehuman PIP4K2C locus and familial autoimmunity (8, 9). Theimmune system is a complex network that evolved to protectorganisms from invasion of various microbes. Whereas an activeimmune system protects from microbes, overactivation of theimmune system can result in autoimmunity. An understanding ofthe molecular underpinnings of an overactive or underactiveimmune system is important for developing therapies for immune-related diseases from immunodeficiency to autoimmunity andeven cancers. mTORC1, a central regulator of cell survival,growth, and metabolism, also plays a critical role in regulation ofimmune cells. mTORC1 responds to intra- and extracellularsignals such as growth factors, oxygen levels, energy status, andamino acid levels (10). The mTOR pathway is activated duringvarious cellular processes, including T-cell activation, insulinresistance, and tumor formation. The immunosuppressive drugrapamycin that has been used clinically to prevent transplantrejection, directly binds to mTORC1 to suppress immune responses(11). One of the identified roles of mTORC1 in the immune systemis to direct T-cell-fate decisions. mTORC1 positively regulatesdifferentiation of the Th1 and Th17 subset of Th cells (12). Onthe other hand, mTORC1 is a negative regulator of Treg dif-ferentiation (13) and at the same time is required to maintainTreg function (14).Here, we present the first characterization to our knowledge of

mice with germline deletion of Pip4k2c. Surprisingly, Pip4k2c−/−

mice exhibit a phenotype that is quite different from Pip4k2a−/− orPip4k2b−/− mice. These mice develop normally and are not pro-tected from obesity, insulin resistance, or diabetes, but ratherdevelop enhanced immune responses, resulting in autoimmunity.

In addition, they exhibit hyperactivation of mTORC1 signaling inmultiple tissues, suggesting that Pip4k2c negatively regulatesmTORC1. These results, along with a recent observation that theenzyme encoded by this gene is a substrate of mTORC1 (6),suggest a close relationship between mTORC1 and Pip4k2c.Moreover, the hyperimmune phenotype of the Pip4k2c−/− micecould be partially ameliorated by treatment with the mTORC1inhibitor, rapamycin. Importantly, a SNP (rs1678542) located nearthe PIP4K2C locus has been correlated with familial autoim-munity (8) and the results presented here suggest that loss ofPI5P4Kγ function could explain this disease.

ResultsGeneration of Pip4k2c−/− Mice. To investigate the role of PI5P4Kγin mammals, PI5P4Kγ knockout (Pip4k2c−/−) mice were gener-ated using Pip4k2c-targeted ES cell clones obtained from theknockout mouse project (KOMP) repository (www.komp.org).In the targeting construct, exons 3 and 4 of Pip4k2c werebracketed by loxP sequences, so deletion of these exons couldremove much of the kinase domain and place the protein out ofthe correct reading frame (Fig. 1A). ES cell clones were kar-yotyped and selected for microinjection. Because the targetedclones were C57BL/6N (agouti)-derived JM8A3.N1, the cellswere injected into C57BL/6J (black) blastocysts. Three chimericmice were born and they were backcrossed with C57BL/6 micefor three generations. Then whole body knockout mice wereobtained by crossing the transgenic mice with B6.C-Tg(CMV-cre)1Cgn/J. The Pip4k2c+/− mice were crossed with Pip4k2c+/−

mice to get Pip4k2c+/+ (WT) mice and Pip4k2c−/− (KO) mice touse for the ensuing experiments. The wild-type allele and knock-out allele of each mouse was confirmed by genomic DNA PCR(Fig. 1B) and Western blot for protein expression (Fig. 1C).

Increased Inflammation in Pip4k2c−/− Mice. Pip4k2c−/− mice brednormally and grew into adulthood, displaying no obvious growthor behavioral abnormalities (Fig. S1 A–E). Unlike Pip4k2b−/−

mice, these mice did not have enhanced insulin sensitivity andwere not protected from obesity on a high-fat diet (Fig. S1 A–E).

A B

Ex3

Ex4

En2SA

IRES gal

act ::neo

pAEx5

Ex2

Cassette inserted locus

Knockout locus

Wild type locus

En2 SA

IRES gal

pAEx5

Ex2

Ex3

Ex4

Ex5

Ex2

X CMV-Cre

pwtF pwtR

pkoF pkoR

: FRT: loxP

pA

WT(pwtF,pwtR)

KO (pkoF,pkoR)

+/+ +/- -/- Mouse tail DNA

PI5P4K

Mouse kidney protein+/+ -/- +/+ +/+ -/- -/-

C

-actin

E

D

Imm

une

infil

trate

sto

tal a

rea

0

0.5

1

1.5

2

WT KO

*

CD3B220 MAC-2

KO

1 W

T

(%) WT KO

KO

1K

O2

Fig. 1. Generation of Pip4k2c−/− mice. (A) Schematic representation of the Pip4k2c locus before and after deletion of the critical exons (exons 3 and 4). Thedeletion of the critical exons was induced by Cre-lox recombination. β-gal, β-galactosidase; En2SA, engrailed 2 gene splice acceptor sequence; Ex, exon; neo,neomycin; pA, polyA signal. (B) PCR analysis of genomic DNA prepared from mouse tail. Primers used for genotyping by PCR are pwtF, pwtR, pkoF, and pkoR.(C) Western blotting of proteins prepared from mouse kidney. (D) Immune cell infiltration in the liver of Pip4k2c−/− mice. H&E staining readily identifiesimmune infiltrates in the liver tissue from Pip4k2c−/− mice. (Scale bars, 100 μm.) The area of immune infiltrates of H&E-stained liver tissues was quantifiedusing ImageJ software. Twenty-five Pip4k2c−/− mice and 12 wild-type mice (four images per mouse) were examined. The results are presented as means ± SEof the means. (E) Immune infiltrates in liver tissue of Pip4k2c−/− mice are mostly T cells and B cells. The Pip4k2c−/− liver sections of two animals (KO1 and KO2)were incubated with anti-CD3, anti-B220, and anti-Mac2 antibodies, respectively.

Shim et al. PNAS | July 5, 2016 | vol. 113 | no. 27 | 7597

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

0

Because a SNP (rs1678542) near the human PIP4K2C locus hasbeen linked to autoimmunity, we examined whether Pip4k2c−/−

mice exhibit any inflammatory phenotype. We carried out acomplete necropsy of Pip4k2c−/− mice at different ages. We foundthat immune cells formed clusters in the organs of maturePip4k2c−/− mice (8–14 mo of age). The immune cell infiltrationwas observed in the liver, kidney, salivary glands, lungs, and in-testine of the mice (Fig. 1D and Fig. S1F). To measure thesurface area of infiltrating immune cells in the liver, we para-formaldehyde fixed and paraffin embedded liver tissues andperformed H&E staining. The ratio of immune infiltrates pertotal area was significantly increased in the Pip4k2c−/− mice (Fig.1D), which indicated that the Pip4k2c−/− mice developed chronicinflammation without a specific trigger such as infection or in-juries. To identify which type of immune cells infiltrated the or-gans, we stained the liver tissue sections with anti-CD3, anti-B220,and anti-Mac2 antibodies. The infiltrating immune cells in thePip4k2c−/− livers consisted of mostly CD3+ T cells and B220+ Bcells (Fig. 1E).Moreover, the plasma levels of various proinflammatory cy-

tokines increased in the Pip4k2c−/− mice, including the Th1-typecytokines IFNγ, IL-12, and IL-2 (Fig. 2A). Additionally, IL-17and IFNγ secreted by CD4+ T cells isolated from spleen werehigher in the Pip4k2c−/− mice (Fig. 2B). As T-cell cytokines canaffect Ig class switching, we performed Ig isotyping. It is knownthat IL-17 drives B cells to undergo preferential isotype classswitching to IgG3 and IgG2a (15). We found that the level of

IgG3 increased in the plasma of Pip4k2c−/− mice, which agreedwith the increase in IL-17 levels (Fig. 2C).

T Cells Are Hyperactivated in Pip4k2c−/− Mice. Flow cytometry wasused to determine the activity of T cells by counting CD44+ andCD62L+ T cells. We found that Pip4k2c−/− mice have moreCD44+-activated T cells and fewer CD62L+ naïve T cells com-pared with the wild-type mice (Fig. 2D). These data suggested thatT cells were more activated in Pip4k2c−/− mice. Also, Pip4k2c−/−

mice showed increased CD4+ and CD8+ populations, indicatingan increase of Th cells and cytotoxic T cells (Fig. S1G).To investigate proliferation of T cells, we isolated CD4+ T

cells from the spleens from Pip4k2c−/− mice and wild-type mice.The cells were seeded in 96-well plates coated with anti-CD3 andanti-CD28. After 48 h of culture, 3H-thymidine was added foranother 16 h before measuring the 3H-thymidine incorporation(Fig. 2E). The Pip4k2c−/− T cells showed significantly higherproliferation rates than T cells from wild-type mice. In accor-dance with increased T-cell activation in Pip4k2c−/− mice, thenumber of Foxp3+CD4+ Treg cells decreased in Pip4k2c−/− mice(Fig. 2F). Therefore, these results indicate that Pip4k2c−/− micedeveloped inflammation with an activation of T cells.

Increased mTORC1 Signaling in Pip4k2c−/− Mice. Because the im-mune system is hyperactivated in Pip4k2c−/− mice, as indicatedby an increase in inflammation concomitantly with an increase inT-cell activation, an increase in Th-cell population and a de-crease in Treg-cell population, we next investigated if mTORC1

0.1

1

10

100

1000

10000

IFN

IL-1

2 (p

40)

IL-1

2 (p

70)

IL-2

IL

-9

IL-1

7 R

AN

TES

IL

-3

MIP

-1B

M

CP

-1

IL-5

IL

-10

GM

-CS

F IL

-4

VE

GF

MIP

-1a

IL

-6

M-C

SF

IL-1

3

IL-1

B

TNFa

K

C

MIP

-2

IP-1

0 LI

X

G-C

SF

IL-1

a M

IG

LIF

Eot

axin

IL

-7

IL-1

5

WT KO

pg/m

l (lo

g sc

ale)

IL-17

IFN

WT KO

IL-1

7 (n

g/m

l)

IFN

(ng

/ml)

WT KO WT KO

Abs

orba

nce

at 4

50nm

0

0.5

1

1.5

2

2.5

IgG1

IgG2a

IgA IgM

WTKO

*

CD44

CD

62L

IgG2b

IgG

3

CD4

Foxp

3

WT KO

0 10 20 30 40 50

WT KO

CD4+Foxp3+ (%)

10 3 1 0.3 0.10

20000

40000

60000

80000*

**

3 H U

ptak

e (c

.p.m

) WTKO

Anti-CD3

**

* *

**

0

20

40

60

80

WT KO

CD4+ CD44+ CD62L- (%) 44.2 20.7

A

B

D E F

C

Fig. 2. Proinflammatory cytokines are increased in Pip4k2c−/− mice. (A) Plasma cytokines were detected using multiplex cytokine ELISA. The experimentswere performed on >10 mice (12–14 mo old) per group, with two measurements per mouse. The y axis is in logarithmic scale. The results are presented asmeans ± SE of the means. *P < 0.05. (B) T-cell–derived IFNγ and IL-17 levels are elevated in Pip4k2c−/− mice. Flow cytometry of IL-17 and IFNγ secretion by CD4+

T cells isolated from spleen-indicated groups. The data are representative of three independent experiments with n > 3 mice (12–14 mo old) per group. *P <0.05 (Student’s t test, error bars represent SD). (C) Plasma IgG3 levels are elevated in Pip4k2c−/− mice. Plasma Ig levels were measured using ELISA. ThreePip4k2c−/− mice (12 mo old) and four age-matched wild-type mice were examined. The results are presented as means ± SE of the means. *P < 0.05.(D) Pip4k2c−/− mice exhibit an increase in CD44+ active T cells and a decrease in CD62L+ naïve T cells. Flow cytometry is shown of CD44+ and CD62L+ T cellsisolated from spleen-indicated groups. The most representative data from three independent experiments are given, with n > 3mice (12–14mo old) from each group.(E ) T cells from Pip4k2c−/− mouse have enhanced growth rates. CD4+ T cells were isolated from spleens of 12-mo-old mice and cultured in media con-taining 3H-thymidine. The level of radioactivity was measured by liquid scintillation. The data are presented as mean 3H-thymidine incorporation (cpm ± SEM,performed in triplicate). *P < 0.05. (F) Regulatory T cells are suppressed in Pip4k2c−/− mice. Flow cytometry of Foxp3+ and CD4+ T cells isolated from spleen-indicated groups. The most representative data from three independent experiments are given, with n > 3 mice (12–14 mo old) from each group.

7598 | www.pnas.org/cgi/doi/10.1073/pnas.1600934113 Shim et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

0

signaling, which is known to regulate diverse immune cell types,(including T cells, macrophages, dendritic cells, neutrophils, andmast cells) was altered in Pip4k2c−/− mice. We found thatphosphorylation of p70-S6K on threonine 389 (Thr389, a directsubstrate of mTORC1) was increased in kidney, liver, brain, andmuscle tissues from Pip4k2c−/− mice compared with wild-typemice (Fig. 3A). Phosphorylation of Thr389 of p70-S6K was alsoenhanced in spleen, the major immune system organ (Fig. 3B).Levels of mature SREBP1 have recently been shown to be adownstream reporter for mTORC1 activity (16). Consistent withthe increased p70-S6K phosphorylation, we also found that levelsof mature SREBP1 were significantly higher in various tissuesfrom the Pip4k2c−/− mice (Fig. 3A). On the other hand, otherupstream and downstream components of the mTORC1 path-way, including Akt, AMPK, PDK1, GSK3α/β, and 4E-BP1, didnot exhibit significant changes in phosphorylation sites thatregulate the activity of this pathway (Fig. S2). The failure to seesignificant changes in these other components probably reflectsrobust feedback control at each of these steps and further sup-ports the concept that Pip4k2c regulates a step quite proximal tomTORC1. These results indicate that mTORC1 signaling ishighly activated in Pip4k2c−/− mice, which suggests that increasedmTORC1 signaling could explain the hyperactive immune systemin Pip4k2c−/− mice.

Rapamycin Reduces mTORC1 Signaling and the Inflammatory Phenotypesin Pip4k2c−/− Mice. We next examined whether rapamycin, the allo-steric mTORC1 inhibitor, could reduce the inflammatory pheno-type of Pip4k2c−/− mice. Pip4k2c wild-type and knockout micewere intraperitoneally injected with either vehicle or rapamycin(3 mg·kg−1·d−1) once a day for 2 wk. Blood was withdrawn 2 wkbefore the treatment and 24 h after the first treatment. On the finalday, blood, liver, and spleen were collected. Using protein lysatesfrom the liver and spleen, we performed SDS/PAGE and Western

blot. We found that Thr389 of p70-S6K was still hyperphos-phorylated in liver and spleen tissues of Pip4k2c−/− mice treatedwith vehicle control for 2 wk. However, in the rapamycin-treatedPip4k2c−/− mice, p70-S6K phosphorylation on Thr389 was reducedto the levels seen in tissues from wild-type mice, indicating thatrapamycin had suppressed mTORC1 signaling (Fig. 4 A and B).The levels of plasma cytokines were measured in the control

and rapamycin-treated mice. In agreement with the results inFig. 2A, before rapamycin treatment the Pip4k2c−/− miceexhibited very high plasma levels of IL-12(p40) and IFNγ com-pared with levels in plasma of wild-type mice. The plasma levelof IL-12(p40) in Pip4k2c−/− mice was reduced to the level ob-served in wild-type mice at 24 h of rapamycin treatment andremained suppressed after 2 wk of treatment. Rapamycin had no

SREBP1 (mature)

p70-S6K

-actin

-actin

p-p70-S6K (T389)

Kidney Liver Brain

Muscle

+/+ -/- +/+ +/+ -/- -/-

p-p7

0-S

6K

tota

l p70

-S6K

0 0.5

1 1.5

2

WT KO

0

0.5

1

1.5

WT ki

dney

KO kidn

ey

WT liv

er

KO liver

WT br

ain

KO brain

WT m

uscle

KO mus

cle

* p = 0.038

+/+ -/- +/+ +/+ -/- -/- +/+ -/- +/+ +/+ -/- -/-

+/+ -/- +/+ +/+ -/- -/-

p70-S6K

p-p70-S6K (T389)

Spleen

-actin

TregThp= 0.07

+/+ +/+ +/+ -/- -/- -/- -/- +/+ -/- +/+

p70-S6K

p-p70-S6K (T389)

SREBP1 (mature)

p70-S6K

p-p70-S6K (T389)

p-p7

0-S

6Kto

tal p

70-S

6K

A

B C

Fig. 3. Signaling downstream of mTORC1 is up-regulated in various tissuesof Pip4k2c−/− mice. (A) p70-S6K Thr389 phosphorylation, total p70-S6K, andSREBP1 (cleaved mature form) were blotted for in kidney, liver, brain, andmuscles from 12-mo-old wild-type and Pip4k2c−/− mice (three mice pergroup). The bar graph shows the ratio of p70-S6K Thr389 phosphorylationover total p70-S6K. The results are presented as means ± SE of the means.*P < 0.05. (B) p70-S6K Thr389 phosphorylation and total p70-S6K wereblotted for spleen of 12-mo-old mice (three mice per group). The bar graphshows the ratio of p70-S6K Thr389 phosphorylation over total p70-S6K. Theresults are presented as means ± SE of the means. (C) Th cells and Treg cellsisolated from the spleens of 12-mo-old mice (three mice per group). p70-S6KThr389 phosphorylation and total p70-S6K were blotted.

0 2 4 6 8

10

T veh

T rapa

KO veh

KO rapa

Immune infiltratestotal area (%)

*

p70-S6K

p-p70-S6K (T389)

Liver

actin

KO

+ + + + + +- - - - - -rapamycin+ + + + + + - - -- - -vehicle

WT

Spleen KO

+ + + + + +- - - - - -+ + + + + + - - -- - -

WT

p70-S6K

p-p70-S6K (T389)

actin

rapamycinvehicle

p-p70-60K total p70-60K

0 0.5

1 1.5

2 2.5

WTveh

WTrapa

KO veh

KO rapa

* p = 0.021

0 0.5

1 1.5

WTveh

WTrapa

KO veh

KO rapa

p = 0.502

p-p70-60K total p70-60K

0.1

1

10

100

before 1 day 14 days Rel

ativ

e le

vel (

log

scal

e) IL-12(p40)

WT, veh KO, veh WW W

T, rapa KO, rapa

0.1

1

10

100

1000

before 1 day 14 days

IFN

A

B

C D

Fig. 4. (A and B) Rapamycin reduces the activation of mTORC1 signaling inPip4k2c−/− mice. p70-S6K Thr389 phosphorylation, total p70-S6K, and actinwere blotted for in liver (A) and spleen (B) from 12- to 14-mo-old wild-typeand Pip4k2c−/− mice treated with vehicle or rapamycin. Daily i.p. injectionswere given for 2 wk (3 mg·kg−1·d−1, three mice per group). The bar graphshows the ratio of p70-S6K Thr389 phosphorylation over total p70-S6K. Theresults are presented as means ± SE of the means. *P < 0.05. (C) Changes inplasma cytokine levels after treatment with rapamycin. Plasma cytokineswere detected using multiplex cytokine ELISA. Plasma was collected beforethe treatment (“before”), 24 h after the first treatment (“1 day”), and aday after the final treatment (“14 day”) from the following four groups:wild type with vehicle, wild type with rapamycin (3 mg·kg−1·d−1), knockoutwith vehicle, and knockout with rapamycin (3 mg·kg−1·d−1) [approximatelyfour to eight mice (12–14 mo old) per group, daily i.p. injection for 2 wk].Data were normalized to the mean plasma level of IL-12(p40) and IFNγ inwild-type mice before therapy, with 1 on the y axis indicating the mean ofwild-type pretreatment (WT, vehicle, before): IL-12(p40): 46.1875 relative lu-minescence unit (RLU), IFNγ: 2.148 RLU. The results are presented as means ±SE of the means. Results for IL-2 and IL-12(p70) are shown in Fig. S1H. (D)Changes in immune cell infiltration after treatment with rapamycin. Mouselivers were collected after the final treatment (daily i.p. injections for 2 wk)from the following groups: wild type with vehicle, wild type with rapa-mycin (3 mg·kg−1·d−1), knockout with vehicle, and knockout with rapa-mycin (3 mg·kg−1·d−1), approximately four to eight mice (12–14 mo old) pergroup. The area of immune infiltrates of H&E-stained liver tissues wasquantified using ImageJ software. Five images per mouse were examined.The y axis indicates the ratio of the area of immune infiltrates over totalarea of the tissue in each image. The results are presented as means ± SE ofthe means. *P < 0.05.

Shim et al. PNAS | July 5, 2016 | vol. 113 | no. 27 | 7599

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

0

significant effect on plasma IL-12(p40) in wild-type mice. IFNγlevels decreased somewhat in both wild-type and Pip4k2c−/−

mice at 24 h after the first treatment with rapamycin. However,after 2 wk of treatment with rapamycin, IFNγ levels in both wild-type and Pip4k2c−/− mice returned to the basal levels (the levelbefore any treatment). On the other hand, IL-2 and IL-12(p70)levels did not significantly change in the Pip4k2c−/− or wild-typemice in response to rapamycin treatment (Fig. S1H). Morestrikingly, we found that the area of immune infiltrates in thelivers of rapamycin-treated Pip4k2c−/− mice was dramaticallydecreased after 2 wk of rapamycin treatment (Fig. 4D). Thesedata collectively indicate that inhibition of mTORC1 byrapamycin partially reduced the inflammatory phenotypes inPip4k2c−/− mice.

DiscussionHere we report the generation of Pip4k2c−/− mice and show thatPip4k2c−/− mice have hyperactivated immune systems. Pip4k2c−/−

mice were viable with a normal lifespan and did not show anyspecific abnormality until they were older than 8 mo. But amongthe older mice at ages between 8 mo and 14 mo, Pip4k2c−/− micedisplayed increased immune infiltrates in various tissues, includingliver, intestine, kidney, and lungs. These infiltrating immune cellswere mostly T cells and B cells. Moreover, we found that plasmaof Pip4k2c−/− mice contained high levels of proinflammatoryTh1-type cytokines such as IFNγ, IL-12, and IL-2. Importantly,the increase in Th cells and the decrease in Treg cells in Pip4k2c−/−

mice reflect the inflammatory phenotype of these mice. Further-more, the increase in CD44+ T-cell population (central memoryT cells) in these mice also supports the hyperactivation of theirimmune system. Interestingly, mTORC1 downstream compo-nents p70-S6K and SREBP1 were activated in Pip4k2c−/− mousetissues. Because mTORC1 signaling directs the immune systemby regulating diverse immune cell types, a possibility that otherimmune cells in addition to T cells contribute to the inflammatoryphenotype of these mice would be interesting to investigate further.Moreover, after 2 wk of rapamycin treatment, the inflamma-

tory phenotypes as well as the mTORC1 signaling that wereenhanced in the Pip4k2c−/− mice decreased. These results sug-gested that increased activation of mTORC1 signaling in Pip4k2c−/−

mice could be responsible for the chronic inflammation inPip4k2c−/− mice. These results are in agreement with the cor-relation between the SNP (rs1678542) in the PIP4K2C locusand familial autoimmunity in humans and suggest that the SNPmay suppress expression of PI5P4Kγ protein. Our currentmodel is that loss of PI5P4Kγ results in activation of mTORC1signaling and that this results in increased inflammation, par-tially through up-regulation of Th cells.Whereas high rheumatoid factor levels are often associated

with autoimmune diseases, they do not always correlate. Wemeasured the levels of rheumatoid factor from Pip4k2c−/− miceand found that the levels of rheumatoid factor were not elevatedin the case of Pip4k2c−/− mice (Fig. S3B). In addition, the sizes oforgans of Pip4k2c−/− mice were not different from wild-typemice. However, we found that Pip4k2c−/− mice older than 12 mooccasionally exhibit pale livers (Fig. S3C). So we analyzed liverfunction parameter proteins and metabolites using plasma fromPip4k2c−/− mice (Fig. S3A). There was an increase in aspartatetransaminase (AST) and a decrease in blood urea nitrogen(BUN), without a significant change in alanine transaminase(ALT) and other metabolites. These high AST and low BUNresults suggest liver damage in Pip4k2c−/− mice, the mechanismof which would be intriguing to investigate further.The mechanism by which the immune system is hyperactivated

in Pip4k2c−/− mice has not yet been fully elucidated. Our studiessuggest that mTORC1, which regulates the immune system in di-verse aspects, including modulating T-cell differentiation and ac-tivation (12), is required to maintain the inflammatory phenotype

in Pip4k2c−/− mice. The simplest explanation for the observedautoimmunity is that the protein encoded by Pip4k2c (PI5P4Kγ)plays a role in suppressing the function of the related and muchmore highly active enzymes PI5P4Kα and PI5P4Kβ when mTORC1signaling is too high and that this provides feedback suppression ofmTORC1 signaling to maintain homeostasis in the T-cell pop-ulation. However, because these studies are based on germlinedeletion of Pip4k2c in all tissues, it is possible that non–T-cellautonomous events in other tissues contribute to the autoim-munity observed. This question is currently being addressed bygeneration of mice with T-cell–specific deletion of Pip4k2c.The biochemical mechanisms by which PI5P4Kγ might pro-

vide feedback inhibition of mTORC1 signaling is not clear. Thefact that PI5P4Kγ has very low activity compared with PI5P4Kαand PI5P4Kβ but forms heterodimeric complexes with theseactive enzymes suggests that its major role is to regulate thelocalization and/or activity of PI5P4Kα and PI5P4Kβ. As dis-cussed in the introduction, it was reported by Mackey et al. (6)that PI5P4Kγ is phosphorylated by mTORC1 on S324 and S328.Interestingly, Mackey et al. (6) found that expression of a S324A/S328A double-mutant form of PI5P4Kγ in HeLa cells enhancedmTORC1 signaling, whereas expression of a phosphomimeticmutant (S324D/S328D) suppressed mTORC1 signaling. Theseresults are consistent with a model in which PI5P4Kγ can facil-itate mTORC1 signaling (perhaps by recruiting the more activeenzymes, PI5P4Kα and/or PI5P4Kβ to lysosomes where mTORC1is activated) and that phosphorylation of PI5P4Kγ at S324 andS328 by mTORC1 provides a negative feedback loop to shut offmTORC1 signaling (perhaps by preventing recruitment of PI5P4Kαand PI5P4Kβ to lysosomes). According to this model, deletion ofPI5P4Kγmight impair basal mTORC1 signaling, although PI5P4Kαand/or PI5P4Kβ homodimers or heterodimers could be capable oflocalizing to lysosomes independent of PI5P4Kγ to maintainmTORC1 signaling in tissues from Pip4k2c−/− mice. In any event,in the absence of PI5P4Kγ the negative feedback control wouldbe eliminated, thereby explaining increased mTORC1 activity inmultiple tissues of the Pip4k2c−/− mice.With respect to the abilities of PI5P4Kγ, PI5P4Kα, and

PI5P4Kβ to heterodimerize with each other (17–20), the geneticinteraction among PI5P4Kγ, PI5P4Kα, and PI5P4Kβ is particu-larly thought provoking. Interestingly, Pip4k2a−/− Pip4k2b−/−

double KO mice (5) and Pip4k2b−/− Pip4k2c−/− double KOmice (Fig. S3D) are not viable, whereas Pip4k2a−/− Pip4k2c−/−

mice are viable. Thus, mice that only have Pip4k2b are viable,but if this gene is deleted both Pip4k2a and Pip4k2c are criticalfor viability, indicating that these genes do not have redundantfunctions and must both be expressed to replicate the functionof Pip4k2b. The respective roles for PI5P4Kγ, PI5P4Kα, andPI5P4Kβ are complex. Pip4k2b−/− mice exhibited enhancedinsulin sensitivity, smaller body size, and decreased adiposity ona high-fat diet. In contrast, Pip4k2c−/− mice were not differentfrom wild-type mice in these features (Fig. S1 A–E). In additionto the synthetic lethality for loss of Pip4k2a and Pip4k2b, andfor loss of Pip4k2b and Pip4k2c, distinct phenotypes of each ofthe Pip4k2a−/−, Pip4k2b−/−, and Pip4k2c−/− mice indicate thateach isoform has a unique role in vivo.Of particular interest is the effect of the various knockouts on

signaling through the PI3K–Akt–mTORC1 pathway. Deletion ofPip4k2b enhances Akt activation but, surprisingly, does not resultin enhanced mTORC1 signaling (4, 5). Previous studies frommany laboratories have shown that impaired mTORC1 activa-tion results in smaller cells and smaller mice (10, 21). Consistentwith the failure of Pip4k2b deletion to link Akt activation tomTORC1 activation, the Pip4k2b−/− mice are smaller than wild-type littermates. Although deletion of Pip4k2a results in no ob-servable phenotypes, deletion of a single allele of Pip4k2a in thecontext of deletion of both alleles of Pip4k2b results in evensmaller mice. These data indicate that Pip4k2a and PIP4k2b

7600 | www.pnas.org/cgi/doi/10.1073/pnas.1600934113 Shim et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

0

suppress PI3K-Akt activation but facilitate mTORC1 activation.This model is consistent with the observation that deletion of thesingle form of PIP4K2 in flies causes suppression of TORC1activation and suppression of growth (7). To determine theepistasis among the multiple mammalian enzymes, phenotypesof the viable double knockout mice will need to be betterexamined.Finally, the results that we present here support the associa-

tion of a SNP in the PIP4K2C locus with autoimmunity, sug-gesting that PI5P4Kγ expression is probably low in autoimmunepatients with the SNP near the PIP4K2C locus. Our observationthat treating Pip4k2c−/− mice with rapamycin reduced the in-flammatory phenotype by decreasing the activation of mTORC1indicates that drugs that target mTORC1 signaling are likely tobe effective for patients with familial autoimmunity that corre-lates with the SNP (rs1678542) near the PIP4K2C locus.

MethodsGeneration of Pip4k2c−/− Mice. Protocols approved by Beth Israel DeaconessMedical Center’s Institutional Animal Care and Use Committee and WeillCornell’s Institutional Animal Care and Use Committee were used for thecare and use of the mice in this research. Pip4k2c-targeted embryonic stemcell clones, BO1, BO2, and DO1, were obtained from the KOMP repository.These cells were grown in our laboratory and the conditional knockout al-lele of each clone was confirmed by genomic DNA PCR. The clones were thenkaryotyped and the normal clones BO1 and BO2 were selected and injectedinto blastocysts at the Beth Israel Deaconess Transgenic Facility. Three chi-meric male mice were obtained and each was backcrossed with C57BL/6Jmice. Cassette-bearing mice were mated to Rosa-eFLP1 mice to remove thelacZ reporter. The critical exons 3 and 4 were removed by mating the lacZreporter-deleted mice to germline cytomegalovirus (CMV)-Cre deleter mice.Knockout mice were verified by PCR and Western blotting.

PCR Genotyping. For genotyping, the four primers belowwere used to amplifyregions of genomic DNA present in either wild-type samples or knockoutsamples:

pwtF: TGTCCCCAGGTCTTCAGGAACCT

pwtR: TGCCTTCAGTTTCGCTTGGGGG

pkoF: CACACCTCCCCCTGAACCTGAAAC

pkoR: AGCCGCTGGGGCCAGATGAT.

The primer pair pwtF/pwtR amplifies a fragment (∼0.5 kb) in wild type andthe primer pair pkoF/pkoR amplifies a fragment (∼0.5 kb) in knockout.

Preparation of Mouse Tissues for Immunohistochemistry. Tissues were re-moved from the killed mice and washed with PBS. The samples were fixed in10% (vol/vol) buffered formalin for 24 h and paraffin embedded. H&Estaining was performed at the Rodent Histopathology Core at the Dana-Farber/Harvard Cancer Center. Staining the immune infiltrates in liver tissuewas performed at the Laboratory of Comparative Pathology at MemorialSloan Kettering Cancer Center. The samples were microsectioned, depar-affinized, rehydrated, and heated with a pressure cooker to 125 °C for 30 s incitrate buffer for antigen retrieval and then incubated with peroxidase andprotein blocking reagents, respectively, for 5 min. Sections were then in-cubated with anti-CD3, anti-B220, and anti-Mac2 antibodies, respectively.

Blood Collection from Mouse and Plasma Preparation. Mouse tails were cut1mm from the tip with scissors and blood was collected into prechilled 1.5 mLEDTA-coated Eppendorf tubes (Microvette CB300, Sarstedt). The sampleswere centrifuged for 15 min at 825 × g at 4 °C. The supernatants weretransferred to new Eppendorf tubes and passed through 0.22-μm filters(centrifuging for 1 min at 5,000 rpm at 4 °C).

Measurement of Plasma Cytokines. Mouse plasma was prepared as describedabove. Plasma cytokine levels were measured using multiplex cytokine assayat Eve Technologies. To measure cytokines in our laboratory, the BD Bio-science ELISA kit for IL-12, IFNγ, and IL-2 (M1270, MIF00, and M2000, re-spectively) was used according to the manufacturer’s protocol.

T-Cell Proliferation. Cells were grown in DMEM supplemented with 10% FCS,β-mercaptoethanol, l-glutamine, gentamicin sulfate, and penicillin/strepto-mycin. For the thymidine proliferation assay, 5 × 105 cells per milliliter purifiednaïve CD4+ T cells were cultured for 48 h in flat-bottom 96-well plates in thepresence of various concentrations of anti-CD3 antibody (range, 0.1–10 μg/mL).Cells were pulsed with 1 μCi 3H thymidine for another 16 h of incubation.Mean thymidine incorporation in triplicate wells was measured using aβ-counter (LS 5000; Beckman Coulter).

Rapamycin Treatment of the Mice. For rapamycin treatment, stock solutions(50 mg/mL) were diluted into vehicle [5% (vol/vol) Tween-80, 5% (vol/vol)PEG 400 (polyethylene glycol, molecular weight 400)] in 1× PBS for 2 wk(3 mg·kg−1·d−1) treatments through i.p. injections. Mice were killed after2 wk of treatment.

ACKNOWLEDGMENTS. We thank Gina DeNicola, Florian Kerrath, and othermembers of the L.C.C. laboratory for helpful discussions. L.C.C. is supportedby NIH Grants R01 GM041890 and P01 CA120964. C.W. is supported byNational Multiple Sclerosis Society Career Transition Award TA 3059-A-2 andR00 NIH Pathway to Independence Award 4R00AL110649-02.

1. Clarke JH, Emson PC, Irvine RF (2008) Localization of phosphatidylinositol phosphatekinase IIgamma in kidney to a membrane trafficking compartment within specializedcells of the nephron. Am J Physiol Renal Physiol 295(5):F1422–F1430.

2. Ciruela A, Hinchliffe KA, Divecha N, Irvine RF (2000) Nuclear targeting of the betaisoform of type II phosphatidylinositol phosphate kinase (phosphatidylinositol 5-phosphate 4-kinase) by its alpha-helix 7. Biochem J 346(Pt 3):587–591.

3. Clarke JH, Emson PC, Irvine RF (2009) Distribution and neuronal expression of phos-phatidylinositol phosphate kinase IIgamma in the mouse brain. J Comp Neurol 517(3):296–312.

4. Lamia KA, et al. (2004) Increased insulin sensitivity and reduced adiposity in phos-phatidylinositol 5-phosphate 4-kinase beta-/- mice. Mol Cell Biol 24(11):5080–5087.

5. Emerling BM, et al. (2013) Depletion of a putatively druggable class of phosphati-dylinositol kinases inhibits growth of p53-null tumors. Cell 155(4):844–857.

6. Mackey AM, Sarkes DA, Bettencourt I, Asara JM, Rameh LE (2014) PIP4kγ is a substratefor mTORC1 that maintains basal mTORC1 signaling during starvation. Sci Signal7(350):ra104.

7. Gupta A, et al. (2013) Phosphatidylinositol 5-phosphate 4-kinase (PIP4K) regulatesTOR signaling and cell growth during Drosophila development. Proc Natl Acad SciUSA 110(15):5963–5968.

8. Raychaudhuri S, et al. (2008) Common variants at CD40 and other loci confer risk ofrheumatoid arthritis. Nat Genet 40(10):1216–1223.

9. Fung EY, et al. (2009) Analysis of 17 autoimmune disease-associated variants in type 1diabetes identifies 6q23/TNFAIP3 as a susceptibility locus. Genes Immun 10(2):188–191.

10. Laplante M, Sabatini DM (2013) Regulation of mTORC1 and its impact on gene ex-pression at a glance. J Cell Sci 126(Pt 8):1713–1719.

11. Saunders RN, Metcalfe MS, Nicholson ML (2001) Rapamycin in transplantation: Areview of the evidence. Kidney Int 59(1):3–16.

12. Delgoffe GM, et al. (2011) The kinase mTOR regulates the differentiation of helper Tcells through the selective activation of signaling by mTORC1 and mTORC2. NatImmunol 12(4):295–303.

13. Delgoffe GM, et al. (2009) The mTOR kinase differentially regulates effector andregulatory T cell lineage commitment. Immunity 30(6):832–844.

14. Zheng Y, et al. (2009) Regulatory T-cell suppressor program co-opts transcriptionfactor IRF4 to control T(H)2 responses. Nature 458(7236):351–356.

15. Mitsdoerffer M, et al. (2010) Proinflammatory T helper type 17 cells are effective B-cell helpers. Proc Natl Acad Sci USA 107(32):14292–14297.

16. Düvel K, et al. (2010) Activation of a metabolic gene regulatory network downstreamof mTOR complex 1. Mol Cell 39(2):171–183.

17. Bultsma Y, Keune WJ, Divecha N (2010) PIP4Kbeta interacts with and modulatesnuclear localization of the high-activity PtdIns5P-4-kinase isoform PIP4Kalpha.Biochem J 430(2):223–235.

18. Rao VD, Misra S, Boronenkov IV, Anderson RA, Hurley JH (1998) Structure of typeIIbeta phosphatidylinositol phosphate kinase: A protein kinase fold flattened for in-terfacial phosphorylation. Cell 94(6):829–839.

19. Burden LM, et al. (1999) The flattened face of type II beta phosphatidylinositolphosphate kinase binds acidic phospholipid membranes. Biochemistry 38(46):15141–15149.

20. Wang M, et al. (2010) Genomic tagging reveals a random association of endogenousPtdIns5P 4-kinases IIalpha and IIbeta and a partial nuclear localization of the IIalphaisoform. Biochem J 430(2):215–221.

21. Shima H, et al. (1998) Disruption of the p70(s6k)/p85(s6k) gene reveals a small mousephenotype and a new functional S6 kinase. EMBO J 17(22):6649–6659.

Shim et al. PNAS | July 5, 2016 | vol. 113 | no. 27 | 7601

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

0