Seminal CD38 is a pivotal regulator forfetomaternal toleranceByung-Ju Kima,b, Yun-Min Choia,b, So-Young Raha,b, Dae-Ryoung Parka,b, Seon-Ah Parka, Yun-Jo Chunga,Seung-Moon Parkc, Jong Kwan Parkd, Kyu Yun Jange, and Uh-Hyun Kima,b,f,1

aNational Creative Research Laboratory for Ca2+ Signaling Network and Departments of bBiochemistry, dUrology, and ePathology, Chonbuk NationalUniversity Medical School, Jeonju, 561-180, Korea; cDivision of Biotechnology, College of Environmental and Bioresource Sciences, Chonbuk NationalUniversity, Iksan, 570-752, Korea; and fInstitute of Cardiovascular Research, Chonbuk National University, Jeonju, 561-180, Korea

Edited by John J. Eppig, The Jackson Laboratory, Bar Harbor, ME, and approved December 22, 2014 (received for review July 16, 2014)

A successful pregnancy depends on a complex process that estab-lishes fetomaternal tolerance. Seminal plasma is known to inducematernal immune tolerance to paternal alloantigens, but the seminalfactors that regulate maternal immunity have yet to be characterized.Here, we show that a soluble form of CD38 (sCD38) released fromseminal vesicles to the seminal plasma plays a crucial role in inducingtolerogenic dendritic cells and CD4+ forkhead box P3+ (Foxp3+) reg-ulatory T cells (Tregs), thereby enhancing maternal immune toleranceand protecting the semiallogeneic fetus from resorption. The abortionrate in BALB/c females mated with C57BL/6 Cd38−/− males was highcompared with that in females mated with Cd38+/+ males, and thiswas associated with a reduced proportion of Tregs within the CD4+

T-cell pool. Direct intravaginal injection of sCD38 to CBA/J pregnantmice at preimplantation increased Tregs and pregnancy rates in miceunder abortive sonic stress from 48 h after mating until euthanasia.Thus, sCD38 released from seminal vesicles to the seminal plasma actsas an immunoregulatory factor to protect semiallogeneic fetuses frommaternal immune responses.

seminal plasma | CD38 | regulatory T cells | dendritic cells | fetomaternaltolerance

Seventy-five percent of pregnancies that are lost representfailure of implantation and are therefore not clinically rec-

ognized as pregnancies (1). Recurrent miscarriage (the sponta-neous loss of three or more consecutive pregnancies) is asignificant health issue for 1–2% of women, with no identifiablebiological cause and no effective treatment. During early stagesof pregnancy, complex processes help to create a uterine envi-ronment that is conducive to a successful pregnancy. Theseinclude immunological adaptation to the semiallogeneic fetus.Tolerance to paternal alloantigens is critical for successful re-production in placental mammals (2, 3). Many studies haveproposed that regulatory T cells (Tregs) play an essential role inthe development of fetomaternal tolerance in mice and humans(4–7). Seminal plasma contains potent immunoregulatory mol-ecules that contribute to the induction of tolerogenic DCs(tDCs) and ultimately Treg expansion, which is necessary toestablish maternal tolerance against paternal antigens (8–10).However, the specific molecules in semen that are responsiblefor expansion of Tregs and establishment of maternal toleranceremain undefined.CD38, a mammalian prototype of ADP ribosyl cyclases

(ADPRCs), is a type II transmembrane (TM) glycoproteinexpressed in many cell types and seminal fluid (11–16). CD38produces calcium-mobilizing second messengers, cyclic ADP ri-bose, and nicotinic acid adenine dinucleotide phosphate (11, 12).We previously showed that intact CD38 in prostasomes assistsprogesterone-induced sperm Ca2+ signaling (13). In addition toits enzymatic role for Ca2+ signaling, CD38 may also have anonenzymatic role through its interaction with CD31 (17, 18).CD31, a type I TM homophilic or heterophilic receptor, isexpressed in endothelial cells and a variety of immune cells (19)and is involved in attenuating the inflammatory response in

a variety of inflammatory diseases (20–23). For example, recombi-nant CD38 inhibits LPS-induced inflammatory signals in mousemacrophages and human DCs through an interaction with CD31(24, 25).In this study, we found that CD38 is truncated and released

into the seminal plasma from seminal vesicles (SVs) as a solubleform (sCD38) in humans and mice. This finding prompted us toexamine whether CD38 plays a role in maternal immune toler-ance during pregnancy. We found that sCD38 present in seminalplasma was crucial for the induction of uterine tDC and Tregs,which are responsible for the development of the fetomaternaltolerance.

ResultsCD38 Is Truncated and Released from SVs to the Plasma as a SolubleForm. ADPRC activity was detected in human seminal fluid intwo proteins with molecular weights of 45 kDa and 37 kDa asidentified in in-gel assays (Fig. 1A). Notably, the 37 kDa proteinbound to a CD38 immunoaffinity column (Fig. 1B, lanes 1 and3), whereas the 45 kDa protein did not (Fig. 1B, lanes 2 and 4).This suggests that the 37 kDa protein was a truncated form ofCD38, whereas the 45 kDa protein was full-length CD38, whichis encapsulated within prostasomes (13), making the CD38 (45kDa) physically inaccessible for binding to the affinity column.Aminoterminal sequencing of the 37 kDa protein identified itscleavage site as Arg58 close to the TM domain of CD38 (Fig.1C). Western blot analysis with two different antibodies againstCD38 peptides (CD3845–57 and CD38171–292) confirmed that thiswas the case (Fig. 1D). The soluble form, 37 kDa CD38 (sCD38),was detected in all 19 seminal plasma samples tested (range,

Significance

In natural matings, semen delivers spermatozoa and immuno-regulatory fluids to the female reproductive tract. Here, a sol-uble form of CD38 (sCD38) is shown to play an important rolein facilitating maternal immune tolerance against the fetus byinducing the development of uterine tolerogenic DCs andforkhead box P3+ (Foxp3+) regulatory T cells. Deficiency ofsCD38 in seminal fluid increased the rates of loss of allogeneicfetuses, and this loss was rescued by a direct injection ofrecombinant sCD38 into the uterus. Thus, seminal sCD38 actsas a pivotal immune suppressor for establishing maternalimmune tolerance against the fetus. sCD38 could potentiallybe used to prevent failed pregnancies.

Author contributions: B.-J.K. and U.-H.K. designed research; B.-J.K., Y.-M.C., S.-Y.R.,D.-R.P., S.-A.P., Y.-J.C., S.-M.P., J.K.P., and K.Y.J. performed research; B.-J.K., K.Y.J.,and U.-H.K. analyzed data; and B.-J.K. and U.-H.K. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. Email: uhkim@chonbuk.ac.kr.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1413493112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1413493112 PNAS | February 3, 2015 | vol. 112 | no. 5 | 1559–1564

MED

ICALSC

IENCE

S

0.5–10.6 μg/mL), including one from a vasectomised individual(7.8 μg/mL).We next used a mouse model to examine the role of seminal

CD38 in more detail. We began by asking whether sCD38 waspresent in mouse seminal fluid. Due to technical difficulties inobtaining mouse semen, we collected uterine lavage fluid fromunmated estrous females and mouse SV fluid. sCD38 was notdetected in uterine lavage fluid from unmated estrous females,whereas sCD38 was present in mouse SV fluid (Fig. S1). Toidentify the organ that secreted the sCD38, we collected uterinelavage fluid from the female reproductive tract 1 h after mating.Western analysis demonstrated that the uterine lavage fluidcollected from female mice mated with normal or vasectomisedmice contained sCD38 (34 kDa). Mouse splenocytes expressedonly the full-length CD38 (42 kDa) (Fig. 1E) (26). sCD38comigrated with recombinant sCD38 lacking the cytoplasmic andTM domains on SDS/PAGE gels (Fig. 1E). By contrast, uterinelavage fluid collected from females mated with SV-deficient(SVX) males did not contain sCD38. These findings indicate thatsCD38 in seminal fluid originates from SV and that the trunca-tion of CD38 in SV is common to both humans and mice.

Seminal sCD38 Is Crucial for Fetomaternal Tolerance During Pregnancy.As sCD38 was present in mouse seminal fluid, we next investigatedwhether seminal sCD38 affected pregnancy, particularly with re-spect to fetomaternal immune tolerance. We generated an alloge-neic pregnancy model based on BALB/c (H-2d) female mice mated

with Cd38−/− or wild-type C57BL/6 (H-2b) male mice. Mice wereexamined at 12.5 d postcoitum (dpc), at which time both implan-tation and resorption would be visible. We found that female micemated with Cd38−/− males (Cd38−/− matings) showed higher ratesof fetal resorption at 12.5 dpc than females mated with Cd38+/+

A B

C

1 2

– 45 kDa – 37 kDa

HC45 kDa CD38

37 kDa sCD38

171-292 45-57 Epitope :

D E

1 2 3 4

– 45 kDa – 37 kDa

42 kDa

CD38 ID

sCD38 RFPETV

45-57 171-292

TM ED58

34 kDa

HC

Fig. 1. Identification of sCD38 in seminal fluid by in-gel ADPRC assay. (A)Visualization of human seminal fluid proteins with ADPRC activity (lanes 1and 2 for two healthy volunteers). (B) Coomassie blue staining of proteinseluted from a CD38 immunoaffinity column (lane 1) and proteins that passedthrough the column (lane 2). In-gel ADPRC activity assay for the proteinseluted from the column (lane 3) and passed through the column (lane 4). (C)Upper diagram shows the intracellular domain (ID), TM domain, and extra-cellular domain (ED) of CD38, and the lower diagram shows the N-terminalamino acid (aa) sequence (highlighted in red) of sCD38. The segments (aa45–57 and aa 171–292) denoted by green lines represent epitopes of theantibodies used for immunoblotting. The dotted line indicates the cleavagesite within CD38. (D) Proteins immunoprecipitated from seminal fluid andprostasome lysates using anti-CD38 antibodies were blotted with antibodiesspecific for two different epitopes: amino acids 45–57 and 171–292. HC,heavy chain of Ig. (E) Mouse seminal plasma was collected from B6 femalesby uterine lavage within 1 h of mating with B6 intact (CON), vasectomised(VAS), and SVX males immunoprecipitated with anti-CD38, and the immu-noprecipitates were analyzed for CD38 by Western blotting. Mouse sple-nocyte lysate (SP) and recombinant mouse sCD38 (sCD38) were included inthe experiment as size references of full length and truncated sCD38.

A C

0

10

20

30

40P = 0.015

% R

esor

ptio

n

n=

13

n=

9

0

2

4

6

8

10

12

Tota

l im

plan

tatio

n si

te

n=

9

n=

13

n.s.

MHCII D

100 101 102 103 104

CD80

100 101 102 103 104

Even

ts

P = 0.04P = 0.015

0

10

20

30

40

50

60

0

50

100

150

200

MH

CII+

/CD

11c+

(MFI

)

CD

80+ /C

D11

c+(M

FI)

B

CD

4

Foxp3

E

100 101 102 103104100101102103104

11.7

14.2

Wild

x 400

Wild

-type

mat

ings

mat

ings

x 40 x 200

G P = 0.039

% F

oxp3

+of

CD

4+

FP < 0.05

% F

oxp3

+of

CD

4+

% F

oxp3

+of

CD

4+

P = 0.008

Isotype Wild-type matingsmatings

0

2

4

6

8

10

12

Viab

le im

plan

tatio

n si

te P = 0.003

n=

9

n=

13

De

jz

Ib3 mm

De

Ib

jz

jzIb Ib

jzIb / jz Ib3 mm

500 m

500 m

100 m

100 m

50 m

50 m 50 m

50 m

Cd38-/-

Cd38-/-

Cd38-/-

Cd38-/-

Cd38-/-

Cd38-/-

Cd38-/-

Cd38-/-

Cd3

8-/-

Cd38-/-

Cd38-/-

Fig. 2. Pregnancy impairment in female mice mated with Cd38−/− malemice. Wild-type or Cd38−/− C57BL/6 males were mated with BALB/c females,and plugged mice were killed at 12 dpc. (A) Resorption rates. (B) Histologicalevaluation of placental tissue. In wild-type mating, the placenta and thejunctional zone (jz) area of the placenta located between the labyrinth (lb)and decidual tissues (De) are intact, and all of the cellular components, in-cluding trophoblasts (empty arrows), are viable. In Cd38−/− mating, theplacenta is necrotic (stars), and the junctional zone area shows vascularcongestion (empty arrow heads), hemorrhage, and inflammatory infiltration(arrows). Insets show low-magnification images of the embryos. (C) Thenumber of total and viable implantation sites in wild-type and Cd38−/−

matings. Bars, mean ± SEM. (D) Phenotypic analysis of CD11c+ uterine DCsfrom wild-type or Cd38−/− mating mice at 3.5 dpc. MHC-II and CD80 ex-pression was examined in gated CD11c+ uterus cells. Mean fluorescent in-tensities (MFIs) of MHC-II and CD80 were plotted as bar graphs (Rightpanels). Bars, mean ± SEM; number of mice tested for each experiment, 8–9.(E, Left panels) Representative flow cytometric analysis of CD4+Foxp3+ T cellsin the PALN in wild-type and Cd38−/− matings at 3.5 dpc. Percentage of CD4+

Foxp3+ T cells in the PALN in wild-type and Cd38−/− matings was plotted forstatistical evaluation at 3.5 (E, Right panel) and 6.5 dpc (F). (G) Percentage ofCD4+Foxp3+ T cells in the uterus in wild-type and Cd38−/− matings at 3.5 dpc.

1560 | www.pnas.org/cgi/doi/10.1073/pnas.1413493112 Kim et al.

males (wild-type matings) (Fig. 2A). In agreement with this finding,histochemical analysis revealed visible signs of inflammation, in-cluding thrombosis and massive immune cell infiltration into theplacentas of Cd38−/− matings, none of which were present in wild-type matings (Fig. 2B). These findings indicate that a lack of CD38in seminal fluid induces embryo resorption, presumably due to theimpairment of fetomaternal tolerance (3–6). The number ofimplantations in Cd38−/− matings was similar to that in wild-typematings, whereas the number of surviving fetuses in Cd38−/− mat-ings was reduced compared with that in wild-type matings (Fig. 2C).To exclude the possibility that mouse pregnancy failures can beovarian-based due to progesterone insufficiency, we com-pared maternal plasma progesterone levels at 12.5 dpc, butfound no significant difference between the two groups (Fig.S2). This suggests that sCD38 reduces the maternal immuneresponse toward the fetus after implantation, but does notaffect the implantation itself.Previous studies suggest that seminal factors promote the gen-

eration of tDCs (9, 10). Therefore, we next examined whetherseminal sCD38 inhibited the insemination-induced maturationof uterine DCs by comparing phenotypic maturation markers onuterine DCs in wild-type and Cd38−/− matings. In agreementwith the high resorption rates in Cd38−/− matings, the uterineCD11c+ cells from Cd38−/− mating showed higher levels ofMHC-II and CD80 expression at 3.5 dpc than those from wild-type matings (Fig. 2D). Because tDCs promote the generation ofTregs (27), we next asked whether sCD38 was necessary forTregs expansion. Expression of the transcription factor, Foxp3, isa hallmark of mature Tregs expansion (28). We found that para-aortic lymph nodes (PALNs) in Cd38−/− matings were signifi-cantly larger (198 ± 60.4%) and contained more CD4+ cells thanthose in wild-type matings at 3.5 dpc. The numbers of Foxp3+

T cells in PALNs in Cd38−/− and wild-type matings were notdifferent at 3.5 dpc and 6.5 dpc (Fig. S3). However, the per-centage of CD4+Foxp3+ T cells in the PALNs from Cd38−/−

matings was significantly smaller than that from wild-type mat-ings at both 3.5 (Fig. 2E) and 6.5 dpc (Fig. 2F), indicating thatsCD38 induced Foxp3+ Treg-mediated T-cell suppression. Thisis consistent with previous reports (29, 30) that Tregs are able tosuppress effector T cells and thereby cause an expansion ofCD4+ cells in the presence of a decreased proportion of Tregs.Likewise, the percentage of CD4+Foxp3+ T cells in the uterus inCd38−/− matings was also significantly less than that from wild-type matings at 3.5 dpc (Fig. 2G). By comparison, syngeneicmatings of C57BL/6 females with wild-type and Cd38−/− malesshowed no differences in resorption rates, total number of im-plantation sites, viable implantation sites, and CD4+Foxp3+

T-cell population (Fig. S4). In addition, Cd38−/− females matedwith C57BL/6 wild-type and Cd38−/− males showed no differencein fetal loss, indicating that a CD38 deficiency in maternal and/orplacental tissue does not affect fetal development, at least insyngeneic pregnancy (Fig. S4). Thus, by inducing Foxp3+ Tregsin females, seminal sCD38 contributes to a successful pregnancyby supporting fetomaternal tolerance.

sCD38 Induces Fetomaternal Tolerance Through Expression of tDCsand Foxp3+ Tregs. Because the properties of uterine DC inpregnant female mice are markedly influenced by seminalsCD38, we examined whether sCD38 induced tDC, which ischaracterized by expression of immature DC surface phenotype.To this end, we generated DCs by culturing bone marrow (BM)cells in vitro with GM-CSF in the absence (BM-DCcontrol) orpresence (BM-DCsCD38) of recombinant sCD38 for 6 d. UponLPS stimulation, BM-DCcontrol displayed a typical mature phe-notype, including increased expression of MHC-II, CD80, andCD40; however, this was not the case for BM-DCsCD38 (Fig. 3A).Moreover, proteinase K-treated recombinant sCD38 was in-capable of suppressing DC maturation (Fig. S5). These data

suggest that sCD38 directly suppresses the maturation, causingDCs to maintain an immature phenotype. We also found thatupon exposure to LPS, BM-DCsCD38 secreted a higher level ofIL-10 (TH2 cytokine) and expressed higher levels of TGF-βmRNA and protein than did BM-DCcontrol (Fig. 3B). Conversely,BM-DCsCD38 inhibited the production of both IL-12p70 andTNF-α (TH1 cytokines) (Fig. 3C), indicating that sCD38skewed DCs toward an anti-inflammatory phenotype. However,

0

5

10

15

20

25

30

35

Prol

ifera

tion

(103

c.p.

m.)

A

B P < 0.001

P < 0.005P < 0.001

P < 0.05E

p-STAT3

STAT3

sCD38Control

p-ST

AT3

(RE)

D

F

Rel

ativ

e m

RN

A(T

GF-

1)

-actinTGF-

TGF-

1 (R

E)

C

P < 0.001

P<0.05

P < 0.001

84.9

CD11cMHCII CD80 CD40

34.6

79.1 62.0

39.4

73.0

42.7

87.1

TNF–

(pg/

ml)

IL-1

2p70

(pg/

ml)

IL-1

0 (p

g/m

l)

Even

ts

Foxp3

CD

4An

ti-TG

F-C

ontr

ol

P < 0.01

n.s.

0

2

4

6

8

10

12

% F

oxp3

+of

CD

4+

P < 0.01

4.9 10.8

4.6

100 101 102 103 10410

010

110

210

310

4

5.6

100 101 102 103 104100 101 102 103 104100 101 102 103 104100 101 102 103 104

100 101 102 103 104

BM-DCcontrol

BM-DCsCD38

BM-DCcontrol BM-DCsCD38

Fig. 3. sCD38-mediated induction of tDC. (A) DCs were generated frommouse BM cells by culture with GM-CSF for 6 d in the absence (BM-DCcontrol,line histogram) or presence (BM-DCsCD38, filled histogram) of 200 ng/mLsCD38 and then activated with 1 μg/mL LPS for 24 h to induce DC matura-tion; gray histogram, isotype-matched control. Data are representative ofnine experiments. (B and C) The levels of cytokines of LPS-stimulated BM-DCswere measured in IL-10, IL12p70, and TNF-α ELISAs. TGF-β1 mRNA and pro-tein were measured by real-time PCR and immunoblotting, respectively.(Right Bottom) Band intensity of TGF-β1 relative to that of actin. RE, relativeexpression. (D) Western blot analysis of phosphorylated and total STAT3 inBM-DCcontrol and BM-DCsCD38. (Bottom Panel) Relative band intensity (RE).Data of TGF-β and STAT3 represent the mean ± SD of three experiments. (E)LPS-stimulated BM-DCcontrol and BM-DCsCD38 (2 × 104) were cocultured withallogeneic CD4+ T cells (2 × 105), and proliferation was determined by [3H]thymidine incorporation. Data represent the mean ± SD of three in-dependent experiments. (F) Representative flow cytometric analysis showingthe CD4+Foxp3+ T cells in the mixed leukocyte reaction. OVA 323–339 pep-tide-stimulated BM-DCcontrol or BM-DCsCD38 were cocultured with OT-II CD4+

CD25− T cells in the presence or absence of anti–IL-10 or anti–TGF-β mAbs for5 d. Bar graph represents mean percentage of Foxp3+ among gated CD4+

cells. Bars, mean ± SD of three experiments.

Kim et al. PNAS | February 3, 2015 | vol. 112 | no. 5 | 1561

MED

ICALSC

IENCE

S

no distinct expression levels of indoleamine 2,3-dioxygenase (IDO)and BM-DCsCD38 and BM-DCcontrol were observed (Fig. S6).Previously, CD31 ligation has been reported to inhibit NF-κBsignaling (24, 31). However, sCD38 did not affect NF-κB, Erk, andp38 signaling pathways (Fig. S7). Instead, sCD38 triggered thephosphorylation of a signal transducer and an activator of tran-scription 3 (STAT3) (Fig. 3D), a transcription factor involved inregulating tDCs (32, 33), indicating that sCD38 induced the dif-ferentiation of tDCs via STAT3-dependent pathways. Consistentwith this mechanism, BM-DCsCD38 showed impaired activity ina mixed leukocyte reaction containing allogeneic CD4+ T cellscompared with BM-DCcontrol (Fig. 3E).Next, we investigated whether tDCs induced by sCD38 have

the potential to induce Foxp3+ Tregs. We observed a substantialreduction in the percentage of Foxp3+ in CD4+ cells when sortedOT-II naïve T cells were cocultured with BM-DCcontrol (4.7 ±0.8%) versus with BM-DCsCD38 (9.9 ± 1.3%) (Fig. 3F). Fur-thermore, the degree of induction of Foxp3+ Tregs by BM-DCsCD38 was lower in the presence of anti–TGF-β than that inthe absence of anti–TGF-β or anti–IL-10 mAbs. When anti–IL-10 was present instead of TGF-β, the induction of Foxp3+ Tregsby BM-DCsCD38 was not altered. These results indicate thatTGF-β released by BM-DCsCD38 was primarily responsible forthe induction of Foxp3+ Tregs. Thus, BM-DCsCD38 show featuresof tDCs, as BM-DCsCD38 are able to induce the differentiation ofTregs, which are crucial for fetomaternal tolerance.

sCD38 Induces the Differentiation of tDCs Through a CD31 IndependentPathway. CD31 has been reported to be a counterreceptor forCD38, and CD38-mediated negative regulation of TLR4 sig-naling was undertaken by CD31 (24, 25). Thus, we evaluated thepossibility of CD31 involvement in the sCD38 effects on sus-taining immature phenotypes of BM-DC as seen in Fig. 3A.Although a recent study suggested that DCs lacking CD31 favorimmunogenic potential (31), our data showed that LPS-stimulatedwild-type BM-DCsCD38 and Cd31−/− BM-DCsCD38 expressed similarlevels of MHC-II and CD80 (Fig. 4A). Similarly, upon exposure toLPS even without CD31 expression, mRNA levels of IL-10 andTGF-β were higher, whereas the mRNA level of IL-12p40 waslower in sCD38-treated BM-DCsCD38 than those in BM-DCcontrol(Fig. 4B). Regardless of CD31 expression, BM-DCsCD38 maintainedweaker allogeneic T-cell proliferative responses than BM-DCcontrol(Fig. 4B). Collectively, these data show that CD31 is not involved insCD38-driven differentiation of tDCs.

sCD38 Reduces Stress-Challenged Fetal Resorption by Expansion ofFoxp3+ Tregs. Because sCD38 induces tDCs and then Foxp3+

Tregs, we examined whether intravaginal injection of sCD38maintained pregnancy in abortion-prone mice under sonic stress(34). Direct injection of sCD38 to the vagina of stress-challengedCBA/J female mice mated with DBA/2J male mice greatly re-duced the incidence of fetal resorption (Fig. 5A), although thenumber of fetal implantations was unchanged (Fig. 5B). To ex-amine the inhibitory effects of sCD38 on the aberrant maturationof uterine CD11c+ cells induced by sonic stress, we characterizedthe phenotype of CD11c+ cells. Sonic stress induced the maturationof uterine CD11c+ cells with increased levels of MHC-II and CD80in pregnant CBA/J females mated with DBA/2J males, comparedwith those in control mice. Administration of sCD38 decreased thenumber of phenotypically matured uterine CD11c+ cells in stress-challenged mice (Fig. 5C). In addition, the PALNs of female miceinjected with sCD38 contained a higher proportion of Foxp3+ cellsat 6.5. dpc than Foxp3+ cells of control females (Fig. 5D). Takentogether, the results of the present study show that sCD38 causesinduction of tDCs and Foxp3+ cells and thereby protects embryosfrom maternal immune rejection.

DiscussionThe role of Tregs in fetomaternal tolerance is essential fornormal pregnancy. Studies have shown that the adoptive transferof Tregs prevented fetal loss in the embryo resorption model(DBA/2J-mated CBA/J female), and the depletion of Tregs ledto high levels of embryo resorption in the allogeneic pregnancymodel (35, 36). Foxp3+ Tregs are generated in the thymus andperiphery from naïve CD4+ T cells (5). Interestingly, recentstudies reported that extrathymic Foxp3+ Tregs play a pivotalrole in fetomaternal tolerance and that those memorized fetal-specific Tregs rapidly expanded to induce tolerance during thesubsequent pregnancy (5, 6). Even in the absence of fertilization,exposure of the female genital tract to seminal plasma inducedtolerance to the paternal alloantigen via expansion of Foxp3+

Tregs (8, 37). Seminal plasma confers DCs with immunoregu-latory potential, and these DCs are crucial for embryo implan-tation and the generation of Tregs (9, 10, 34, 38). Seminal fluidcontains high levels of TGF-β (39, 40) and prostaglandins (41),both of which have been proposed to contribute to the inductionof Foxp3+ Tregs and/or the differentiation of tDCs (9, 42).However, it has been suggested that other additional moleculesare involved in the induction of tDCs (9). In the present study,we demonstrated that sCD38-treated BM-DCs transform into

A

B

Cd31-/- BM

Cd31-/- BM

Cd31-/- B

M-D

C

Cd31-/- BM

Cd31-/- BM

Fig. 4. CD31 is not involved in sCD38-mediated induction of tDC. (A) BM-DCs were produced from BM cells of wild-type and Cd31−/− mice by culturein the medium containing GM-CSF for 6 d in the presence (BM-DCsCD38) orabsence of sCD38 (BM-DCcontrol) and stimulated with 1 μg/mL LPS for 24 h.Expression of MHC-II and CD80 on BM-DCcontrol and BM-DCsCD38 derived fromwild-type and Cd31−/− mice was plotted as line histogram (Left panels). MFIsof MHC-II and CD80 were plotted as bar graphs (Right panels). Data repre-sent the mean ± SD of 15 experiments. (B) The mRNA levels of IL-10, TGF-β1,and IL-12p40 in LPS-stimulated BM-DCcontrol and BM-DCsCD38 derived fromCd31−/− mice. Allogeneic T-cell proliferative responses were measured by[3H]thymidine incorporation after coculture with LPS-stimulated BM-DCcontrol and BM-DCsCD38 derived from Cd31−/− mice. Data are expressedas the mean ± SD of four experiments.

1562 | www.pnas.org/cgi/doi/10.1073/pnas.1413493112 Kim et al.

tDCs, which in turn promoted the generation of Foxp3+ Tregs. Adeficiency of sCD38 in semen led to the immunogenic matura-tion of uterine DCs and the diminished proportion of Foxp3+

Tregs, resulting in fetal loss (Fig. 2). In addition, the adminis-tration of sCD38 restored tolerance through the expansion ofFoxp3+ Tregs and prevented fetal loss.Regarding fetomaternal tolerance induced by sCD38-medi-

ated Foxp3+ Treg expansion, it is likely to be developed vianumerous immunomodulatory mechanisms. The expression ofMHC-II and costimulatory molecules, which are essential fortolerogenic ability of DC, was low in DCsCD38. In addition,DCsCD38 secrete IL-10, which inhibits DC maturation andreduces their capacity to stimulate CD4+ T-cell–mediated im-munity. DCsCD38 produce TGF-β, which induces the differentiationof naïve T cells to Foxp3+ Tregs; DCsCD38 inhibit the production ofTH1 cytokines, such as IL-12p70 and TNF-α (Fig. 4).Although the proportion of Foxp3+ Tregs in the CD4+ T cell

was reduced in Cd38−/− matings (Fig. 2 E and F), the number ofFoxp3+ Tregs was not reduced (Fig. S2). The explanation for thisfinding would be that increased mature DCs in Cd38−/− matings(Fig. 2D) induce naïve T cells to produce IL-2, a paracrine Tregsgrowth factor. Thus, tDCsCD38 may positively regulate the Treg-effector T-cell balance in favor of Tregs through inhibitionof T-cell–mediated immunity and induction of Foxp3+ Treg

differentiation, suggesting that the percentage of Foxp3+

Tregs may be more important than the sheer number ofFoxp3+ Tregs in fetomaternal tolerance. Consistent withreports (36) that the contribution of Tregs at the time of earlypregnancy impacts fetal survival later on in gestation, ourresults showed that the Foxp3+ Tregs proportion was higher inwild-type matings than in Cd38−/− matings at 3.5 dpc, and thisdifference persisted through 6.5 dpc. These findings suggestthat a consistently high proportion of Foxp3+ Tregs, inducedby sCD38 during the early phase of pregnancy, may play animportant role in fetomaternal tolerance.Previous studies have reported that maternal inflammation

decreased Foxp3+ Treg accumulation, which caused fetal loss,and that the decidua of the surviving fetus exhibited abnormalspiral artery modification (5). Abnormal maternal inflammationrestricted the growth of surviving fetuses. Abnormal spiral arterymodification in the decidua was associated with complications ofpregnancy, including preeclampsia, placental abruption, pre-term, and fetal loss (43, 44). In addition, immunosuppressivefunction of maternal Tregs in pregnant mice was fetal antigen-specific, and depletion of Tregs led to a reduction in body weightof the surviving embryos (6, 45). In our histological analysis ofsurviving fetuses from Cd38−/− matings, although there wassome variability of individual fetuses, hemorrhage and immunecell infiltration could be observed within the decidua. In somecases, spiral arteries clustering in decidua were also observed(Fig. S8A). Moreover, the number of Foxp3+ cells within alllayers of decidua from Cd38−/− matings was reduced comparedwith that from wild-type matings (1.57 ± 0.64 vs. 3.36 ± 1.2) (Fig.S8B). Consistent with histological evaluations, the mean weightof fetuses was reduced more in Cd38−/− matings than in wild-type matings (Fig. S8C). These observations indicate that Cd38−/−

matings may show other complications of pregnancy, including low-birth-weight babies, besides fetal loss.Seminal plasma interacts not only with epithelial cells of the

vagina and cervix but also with immune cells such as DCs andcauses the recruitment of DCs and macrophages into the en-dometrium (46, 47). The seminal antigen is processed and dis-played on class II MHC cells and transported to draining lymphnodes, activating CD4+ T cells and promoting the acquisition ofa regulatory phenotype. Although the present study shows thatallogeneic mating with sCD38-deficient mice causes maturationof uterine DCs and results in fetal resorption, this may be due tomultiple mechanisms exerted by sCD38, for instance, on otherimmune cells such as macrophages. Therefore, understandingthe mechanisms by which sCD38 affects immune responses is ofutmost importance.Clement et al. (31) recently showed that CD31 is a coinhibi-

tory receptor in the development of tDCs, thus supporting theirprevious finding that the lack of T-cell CD31 signaling increasesautoimmune responses (21–23). The disruption of CD31 sig-naling favored immunogenic maturation. Also, ligation of CD31using a homotypic CD31 peptide reduced the expression ofcostimulatory molecules and proinflammatory cytokines and in-creased the expression of anti-inflammatory cytokines (31). Therole that CD31 plays in inhibitory signaling in effector-adaptiveimmune cells was explained through the action of Src homology2 domain tyrosine phosphatases SHP-1/SHP-2, which are re-cruited by its cytoplasmic immunoreceptor tyrosine-based in-hibitory motif (19). The CD31 peptide can also elicit the sig-naling pathway (48). Contrary to this report, our data showedthat the extent of maturation of Cd31−/− and wild-type BM-DCsupon stimulation with LPS was similar (Fig. 4A). Moreover, aninhibitory signaling on DC maturation by a homotypic CD31 pep-tide was observed in Cd31−/− BM-DCs (Fig. S9). These resultssuggest that the action of sCD38 in driving the differentiation oftDCs is through an as-yet-unidentified target molecule(s) otherthan CD31.

B

0

1

2

3

4

5

6

7

8

Impl

anta

tion

site

n=

8

n=

10

n.s n.sP < 0.05

0

10

20

30

40

no itproseR

%pe

r mot

her

P < 0.05An

= 10

n=

8

n=

10

DP = 0.013

%Fo

xp3+

of C

D4+

P = 0.011

ControlStressStress + sCD38

MHCII

CD80

Even

ts 0

10

20

30

40

0

10

20

30

40

MH

CII+ /C

D11

c+(M

FI)

CD

80+ /C

D11

c+(M

FI)

C

n=

10n

= 7

n=

7

n=

6

n=

7

n=

7

n=

6

P < 0.05P < 0.05

P < 0.05P < 0.05

100 101 102 103 104

Isotype ControlStress Stress + sCD38

ControlStressStress + sCD38

100 101 102 103 104

Fig. 5. Direct injection of sCD38 into the vagina protects fetuses fromabortion. sCD38 or PBS was injected to the pregnant CBA/J females matedwith DBA/2J males intravaginally at 1.5 and 3.5 dpc. Sonic stress was appliedfrom 2.5 dpc until sacrifice. (A) Resorption rates and (B) total number ofimplantation sites were measured at 12.5 dpc in the mated female miceexposed to sound stress with or without sCD38. Bars, mean ± SEM. (C)Phenotypic analysis of CD11c+ uterine DCs from nonstressed mating miceinjected with PBS (Control), stressed mating mice injected with PBS (Stress),and stressed mating mice injected with sCD38 (Stress + sCD38) at 6.5 dpc.Mean fluorescent intensities (MFIs) of MHC-II and CD80 were plotted as bargraphs (Right panels). Bars, mean ± SD. (D) Percentages of CD4+Foxp3+ Tcells in the PALNs of Control, Stress, and Stress + sCD38 were calculated byflow cytometry analysis at 6.5 dpc.

Kim et al. PNAS | February 3, 2015 | vol. 112 | no. 5 | 1563

MED

ICALSC

IENCE

S

sCD38 was first found in the amniotic fluid, suggesting thatsCD38 levels likely increase under certain conditions (49).sCD38 is present at extraordinarily high concentrations in sem-inal plasma, but its levels were below the detection limit of ourELISA (250 pg/mL) in normal serum. The immunoregulatoryrole of sCD38 may provide crucial insights into how inseminationregulates maternal immunity to allow successful pregnancies.Furthermore, this knowledge of the immunoregulatory role ofsCD38 is important for understanding the function of sCD38under pathogenic conditions. Our finding that sCD38 is an im-portant immune regulator for stimulating Treg cells may informstudies to develop novel treatments for recurrent miscarriagewith an immune etiology.

Materials and MethodsThe Institutional Review Board approved the collection of semen from normaland vasectomised volunteers. sCD38 was detected by Western blotting, an in-gel

activity assay, CD38 affinity chromatography, and ELISA. Pregnancy in femalesmated with wild-type or Cd38−/− male mice was determined by checking thecopulation plug (at 0.5 dpc). Embryo tissue sections from pregnant mice werestained with hematoxylin and eosin. Uterine DCs and Foxp3+ Tregs from preg-nant mice were analyzed at 3.5 or 6.5 dpc. To assess whether sCD38 induced tDC,BM cells were cultured in differentiation medium in the presence or absence ofsCD38. A mouse model of stress-induced fetal abortion was used to evaluate theprotective effects of sCD38. Details on materials, assays, and experimental pro-tocol are available in SI Materials and Methods.

All data were analyzed using the Student’s t test or ANOVA as appro-priate. A P value < 0.05 was considered statistically significant.

ACKNOWLEDGMENTS. We thank Dr. Young June Kim, Dr. Donghee Kim,and Mr. Chansu Park for critical reading of the manuscript. This study wassupported by National Research Foundation Grant 2012R1A3A2026453,funded by the Korean government (Ministry of Science, ICT & Future Plan-ning) (to U.-H.K.), and a BK21 grant recipient at Chonbuk National University(to B.-J.K.)

1. Norwitz ER, Schust DJ, Fisher SJ (2001) Implantation and the survival of early preg-nancy. N Engl J Med 345(19):1400–1408.

2. Arck PC, Hecher K (2013) Fetomaternal immune cross-talk and its consequences formaternal and offspring’s health. Nat Med 19(5):548–556.

3. Erlebacher A (2013) Mechanisms of T cell tolerance towards the allogeneic fetus. NatRev Immunol 13(1):23–33.

4. Aluvihare VR, Kallikourdis M, Betz AG (2004) Regulatory T cells mediate maternaltolerance to the fetus. Nat Immunol 5(3):266–271.

5. Samstein RM, Josefowicz SZ, Arvey A, Treuting PM, Rudensky AY (2012) Extrathymicgeneration of regulatory T cells in placental mammals mitigates maternal-fetal con-flict. Cell 150(1):29–38.

6. Rowe JH, Ertelt JM, Xin L, Way SS (2012) Pregnancy imprints regulatory memory thatsustains anergy to fetal antigen. Nature 490(7418):102–106.

7. Saito S, Sasaki Y, Sakai M (2005) CD4(+)CD25 high regulatory T cells in humanpregnancy. J Reprod Immunol 65(2):111–120.

8. Robertson SA, et al. (2009) Seminal fluid drives expansion of the CD4+CD25+ T reg-ulatory cell pool and induces tolerance to paternal alloantigens in mice. Biol Reprod80(5):1036–1045.

9. Remes Lenicov F, et al. (2012) Semen promotes the differentiation of tolerogenicdendritic cells. J Immunol 189(10):4777–4786.

10. Robertson SA, Prins JR, Sharkey DJ, Moldenhauer LM (2013) Seminal fluid and thegeneration of regulatory T cells for embryo implantation. Am J Reprod Immunol69(4):315–330.

11. Lee HC (2012) Cyclic ADP-ribose and nicotinic acid adenine dinucleotide phosphate(NAADP) as messengers for calcium mobilization. J Biol Chem 287(38):31633–31640.

12. Song EK, et al. (2012) NAADP mediates insulin-stimulated glucose uptake and insulinsensitization by PPARγ in adipocytes. Cell Reports 2(6):1607–1619.

13. Park KH, et al. (2011) Ca2+ signaling tools acquired from prostasomes are required forprogesterone-induced sperm motility. Sci Signal 4(173):ra31.

14. Mallone R, et al. (1998) Characterization of a CD38-like 78-kilodalton soluble proteinreleased from B cell lines derived from patients with X-linked agammaglobulinemia.J Clin Invest 101(12):2821–2830.

15. Zielinska W, Barata H, Chini EN (2004) Metabolism of cyclic ADP-ribose: Zinc is anendogenous modulator of the cyclase/NAD glycohydrolase ratio of a CD38-like en-zyme from human seminal fluid. Life Sci 74(14):1781–1790.

16. Kramer G, et al. (1995) High expression of a CD38-like molecule in normal prostaticepithelium and its differential loss in benign and malignant disease. J Urol 154(5):1636–1641.

17. Horenstein AL, Stockinger H, Imhof BA, Malavasi F (1998) CD38 binding to humanmyeloid cells is mediated by mouse and human CD31. Biochem J 330(Pt 3):1129–1135.

18. Deaglio S, et al. (1998) Human CD38 (ADP-ribosyl cyclase) is a counter-receptor ofCD31, an Ig superfamily member. J Immunol 160(1):395–402.

19. Privratsky JR, Newman DK, Newman PJ (2010) PECAM-1: Conflicts of interest in in-flammation. Life Sci 87(3-4):69–82.

20. Ma L, et al. (2010) Ig gene-like molecule CD31 plays a nonredundant role in theregulation of T-cell immunity and tolerance. Proc Natl Acad Sci USA 107(45):19461–19466.

21. Tada Y, et al. (2003) Acceleration of the onset of collagen-induced arthritis by a de-ficiency of platelet endothelial cell adhesion molecule 1. Arthritis Rheum 48(11):3280–3290.

22. Graesser D, et al. (2002) Altered vascular permeability and early onset of experi-mental autoimmune encephalomyelitis in PECAM-1-deficient mice. J Clin Invest109(3):383–392.

23. Maas M, et al. (2005) Endothelial cell PECAM-1 confers protection against endotoxicshock. Am J Physiol Heart Circ Physiol 288(1):H159–H164.

24. Rui Y, et al. (2007) PECAM-1 ligation negatively regulates TLR4 signaling in macro-phages. J Immunol 179(11):7344–7351.

25. Fedele G, et al. (2004) CD38 is expressed on human mature monocyte-derived den-dritic cells and is functionally involved in CD83 expression and IL-12 induction. Eur JImmunol 34(5):1342–1350.

26. Lund FE, et al. (1995) Signaling through murine CD38 is impaired in antigen receptor-unresponsive B cells. Eur J Immunol 25(5):1338–1345.

27. Morelli AE, Thomson AW (2007) Tolerogenic dendritic cells and the quest for trans-plant tolerance. Nat Rev Immunol 7(8):610–621.

28. Fontenot JD, Gavin MA, Rudensky AY (2003) Foxp3 programs the development andfunction of CD4+CD25+ regulatory T cells. Nat Immunol 4(4):330–336.

29. Schmidt A, Oberle N, Krammer PH (2012) Molecular mechanisms of Treg-mediatedTcell suppression. Front Immunol 3(51):1–20.

30. Shevach EM (2009) Mechanisms of foxp3+ T regulatory cell-mediated suppression.Immunity 30(5):636–645.

31. Clement M, et al. (2014) CD31 is a key coinhibitory receptor in the development ofimmunogenic dendritic cells. Proc Natl Acad Sci USA 111(12):E1101–E1110.

32. Rabinovich GA, Gabrilovich D, Sotomayor EM (2007) Immunosuppressive strategiesthat are mediated by tumor cells. Annu Rev Immunol 25:267–296.

33. Ilarregui JM, et al. (2009) Tolerogenic signals delivered by dendritic cells to T cellsthrough a galectin-1-driven immunoregulatory circuit involving interleukin 27 andinterleukin 10. Nat Immunol 10(9):981–991.

34. Blois SM, et al. (2007) A pivotal role for galectin-1 in fetomaternal tolerance. Nat Med13(12):1450–1457.

35. Zenclussen AC, et al. (2006) Regulatory T cells induce a privileged tolerant microen-vironment at the fetal-maternal interface. Eur J Immunol 36(1):82–94.

36. Shima T, et al. (2010) Regulatory T cells are necessary for implantation and mainte-nance of early pregnancy but not late pregnancy in allogeneic mice. J Reprod Im-munol 85(2):121–129.

37. Robertson SA, Guerin LR, Moldenhauer LM, Hayball JD (2009) Activating T regulatorycells for tolerance in early pregnancy—The contribution of seminal fluid. J ReprodImmunol 83(1-2):109–116.

38. Plaks V, et al. (2008) Uterine DCs are crucial for decidua formation during embryoimplantation in mice. J Clin Invest 118(12):3954–3965.

39. Loras B, et al. (1999) Seminal transforming growth factor-beta in normal and infertilemen. Hum Reprod 14(6):1534–1539.

40. Politch JA, Tucker L, Bowman FP, Anderson DJ (2007) Concentrations and significanceof cytokines and other immunologic factors in semen of healthy fertile men. HumReprod 22(11):2928–2935.

41. Templeton AA, Cooper I, Kelly RW (1978) Prostaglandin concentrations in the semenof fertile men. J Reprod Fertil 52(1):147–150.

42. Kali�nski P, Hilkens CM, Snijders A, Snijdewint FG, Kapsenberg ML (1997) IL-12-deficient dendritic cells, generated in the presence of prostaglandin E2, promote type2 cytokine production in maturing human naive T helper cells. J Immunol 159(1):28–35.

43. Renaud SJ, et al. (2011) Spontaneous pregnancy loss mediated by abnormal maternalinflammation in rats is linked to deficient uteroplacental perfusion. J Immunol 186(3):1799–1808.

44. Avagliano L, Bulfamante GP, Morabito A, Marconi AM (2011) Abnormal spiral arteryremodelling in the decidual segment during pregnancy: From histology to clinicalcorrelation. J Clin Pathol 64(12):1064–1068.

45. Kahn DA, Baltimore D (2010) Pregnancy induces a fetal antigen-specific maternal Tregulatory cell response that contributes to tolerance. Proc Natl Acad Sci USA 107(20):9299–9304.

46. Sharkey DJ, Tremellen KP, Jasper MJ, Gemzell-Danielsson K, Robertson SA (2012)Seminal fluid induces leukocyte recruitment and cytokine and chemokine mRNAexpression in the human cervix after coitus. J Immunol 188(5):2445–2454.

47. Robertson SA (2005) Seminal plasma and male factor signalling in the female re-productive tract. Cell Tissue Res 322(1):43–52.

48. Fornasa G, et al. (2010) TCR stimulation drives cleavage and shedding of the ITIMreceptor CD31. J Immunol 184(10):5485–5492.

49. Funaro A, et al. (1996) Identification and characterization of an active soluble form ofhuman CD38 in normal and pathological fluids. Int Immunol 8(11):1643–1650.

1564 | www.pnas.org/cgi/doi/10.1073/pnas.1413493112 Kim et al.