c

TBB IKMONOBIOLOGY OF FETAL RBSORPTIOR IR MICB

BY

Robert L. Gendron

Department of Physiology McGill University Montreal, Quebec July, 1991

A thesis submitted to the Faculty of Graduate studies and Research in partial fulfillment of the requirements for the degree of Doctor of Philosophy c R.L. Gendron, 1991 c al c ABSTRACT

Ph. D. Robert L. Gendron Physiology

Human pregnancy failure is traumatic. Allogeneic concepti can be considered as grafts due to the presence of potentially foreign paternal antigens expressed in fetal tissue. Current thought on spontaneous pregnancy loss is centred around the role of the maternal immune response against the fetal tissues. However, the cellular immune events which mediate spontaneous fetal loss are at present only partially characterized. This thesis contributes to the knowledge of these events by addressing the role of non-specific immune reactions in fetal resorption in the mouse. Utilizing immunohistochemistry, we demonstrated that a large increase in the numbers of natural killer

(NK) cells infiltrating the decidual-ectoplacental cone junction during days 6-9 of gestation precedes fetal resorption (occurring at a frequency of 25-30%) in CBA/J X DBA/2 concepti. The NK-modulatory treatments polyinosynic cytidilic acid (POLY I:C) and anti-asialo GMl (aAGMl), which alter the spontaneous resorption frequency, also alter the frequency of NK-infiltrated c a2

concepti. From these studies we concluded that NK cells may be important cellular effectors of fetal resorption. Since eicosanoids are thought to be instrumental in NK regulation and feto-maternal immunobiology, we focused on arachidonic acid metabolites in characterizing the compounds regulating fetoplacenatal NK activity. Individual fetoplacental units from normal and CBA/J X DBA/2 pregnancies were analysed for NK regulatory activity at stages corresponding to the increase in NK cell number. Using silica column extraction and testing for NK activity modulation in vitro, we found that CBA/J X DBA/2 fetoplacental units were deficient in a lipophilic NK inhibitory factor that was expressed at peak levels on days 7-8 in normal fetoplacental units. The deficiency in NK suppression correlated with NK infiltration and with resorption frequency in CBA/J X DBA/2 concepti. These results demonstrated that CBA/J X DBA/2 fetoplacental units may be inherently predisposed to NK-mediated fetal resorption. In a subsequent paper, we addressed the phenotype and functional significance of the NK inhibitory factor(s) in normal versus resorption-prone pregnancies. By using silica column extraction and specific radioimmunoassay in individual fetoplacental units, we 0 a3 c found that CBA/J X DBA/2 concepti were deficient in the production of the lipoxygenase metabolite leukotriene B4 (LTB4). Furthermore, we demonstrated that LTB4 may be functionally important in averting NK mediated fetal resorption. Supplementation in vitro of fetoplacental extracts from resorption-prone fetoplacental units with exogenous LTB4 resulted in the restoration of the extract's NK suppressive capacity. These results demonstrated that LTB4 may be important in averting NK­ mediated fetal resorption. In order to more fully characterize the role of non-specific immune reactions in fetal resorption, we utilized lipopolysaccharide (LPS) in an attempt to trigger activated macrophages in the fetoplacental unit. We demonstrated an extreme sensitivity to very low concentrations of LPS in early pregnancy. LPS-induced fetal resorption was associated with the intra-uterine production of tumor necrosis factor alpha (TNF-a). In the course of our analysis of the sites of fetoplacental TNF-a production, we found that a novel form of TNF-a with a molecular weight of 50 kilodaltons is expressed in the developing vertebrate nervous system. These findings suggest that TNF-a may play a plieotrophic role in both fetal development and resorption. The results presented 0 a4

in this thesis provide an in-depth analysis of some of the functional deficiencies and immunological events predisposing allogeneic pregnancies to spontaneous fetal loss. The major focus is in providing examples of how non-major histocompatibility complex (MHC)-specific immune reactions can determine fetal survival.

c c RESUME

PhD Robert L. Gendron Physiologie

Chez l'homme, l'infecondite est une condition traumatique. Le foetus exprime des

antigenes paterneles, et peut done etre considere comme un greffon. Tout laisse penser, en

fait, que l'avortement spontane origine d'une reponse immunitaire de la mere encers son

foetus. Cette these analyse les consequences de l'immunite non-specifiques sur le taux

d'avortement spontane chez la souris, et les correle au evenements cellulaires et biochemiques

observe.

Utilisant les techniques d'immunolcalisation sur coups de tissue, nous avons demontre

un grand nombre de cellules de type natural killer (NK) infiltrant la jonction du cone

decidua-placentaire lors de croissements entre les souches de souris CBA/J X DBA/2. Les

cellules NK apparaissent entre la 6 ieme et la 9 ieme journee de grossesse et precede la

resorption des foetus (frequence de 20-30%). La modulation des cellules NK durant la

grossesse avec l'agent POLY I:C ou des anticorps anti-asoalo GMl altere la frequence

d'avortement spontane, et le nombre d'unite placentaire infiltre de cellules NK. Nous

concluons de cette etude que les cellules NK jouent un rOle important dans l'avortement

spontane.

Les eicosanoides sont associes a la modulation de l'activite NK et de fa~on generate a

l'immunobiologie de la grossesse. Nous avons done etudie le rOle des metabolites de l'acide

arachidonique sur l'activite NK a l'interface foetoplacentaire tors de grossesses normales ou resultant du croissement CBA/J X DBA/2. Nous avons prepare des extraits d'unite

foetoplacentaires et purifie les compose lipophiliques par chromatographie sur silice. Par la

suite, ces composes furent analyse in-vitro pour leurs effects sur l'activite NK. Nous avons

observe qu'un facteur lipophilique present auxjours 8-8 des unites foetoplacentaires normales 0 est absent des extraits CBA/J X DBA/2. Ces resultats demontrent que les unite foetoplacentaires CBA/J X DBA/2 possedent probablement un default primaire les c predisposant a l'activite NK et ultimement a la resorption des foetus. Une analyse subsequente a permit de definir le phenotype et !importance biologique du

compose lipophilique pour le succes de la grossesse. La substance fut identifie par

chromatografie sur silice et par dosage radio-immunologique comme un metabolite de la

lipoxygenase, le leukotriene B4 (LTB4). L'addition du L TB4 aux extraits foetoplacentaires

CBA/J X DBA/2 permet de contr6le ractivite NK in-vitro. Ces travaux suggerent que

I'absence du L TB4 represente une anomalie biochemique probablement responsable du taux

eleve d'avortement spontane chez le croissement CBA/J X DBA/2.

Pour mieux definir les mechanismes immunologiques non-specifiques en cause dans le

rejet du foetus, nous avons etudie la contibution des macrophages. Le lipopolysaccharide

(LPS) fut utilise pour activer les macrophages. Nous avons demontre que le LPS a faible dose

pertube fortement la viabilite de la grossesse. L'effect du LPS sur la resorption du foetus est

associe a la production intra-uterine de TNF-a. Aussi, nous avons observe qu'un forme

nouvelle de TNF-a ayant un poids moleculaire de 50 Kilodaltons est presente dans le systeme

nerveux foetus.

En conclusion, cette these apporte une meilleure connaissance des anomalies

biochemiques susceptible d'initier l'avortement spontane, et contribue a definir la

participation des processus immunitaires non-specifiques a la destruction du foetus. as c ACDOWLBDGBHBN'l'S

I extend my sincere gratitude to my research supervisors, Or. M.G. Baines and Dr. R. Farookhi for their support, guidance, encouragement and cooperation. I thank the faculty members with whom I have collaborated, in particular, Dr. o.w. Blaschuk, Dr. w.s. Lapp, and Dr. w.s. Powell. I thank the student and staff members of the Physiology Department for contributing to my scientific and instructional development. I thank the members of the McGill Centre for the Study of Reproduction for support and intellectual guidance. A very warm thanks to the students and staff of the Microbiology & Immunology Department for providing scientific assistance in addition to making me really feel like "one of the gang". All of the above has made my stay a truly memorable and enjoyable experience. Among the many people I wish to thank are Ginette Berube, Richard Boismenu, Josee Brisbois, Feng Cai, sung Cha, Paul Carrier, Cosimo Oemitri, Ralph Diorio, Alain Ouclos, Francine Gravelle, Sylvie Gravel, Alan Lazarus, Ben Lee, Ursula McGuinness and Fred Nestel. A very special thanks to Emilia Anteka, Shelly Feran, Lori Haughn, Marina Pascali, Stev Thomas 0 a6

and Nicolai Van Oers. Finally, I thank my parents, brothers and grandmother for their continued support and endless encouragement.

0 a7 c CLAIM OF CONTRIBUTIOR TO LITERATURE

1. Demonstration that spontaneous fetal resorption in mice is preceded by early NK infiltration at the decidual-ectoplacental cone area. 2. Characterization and time course analysis of the expression of a lipophilic NK regulatory activity in normal mouse fetoplacental units. Partial abrogation of the activity by ASA suggested the involvement of a cyclooxygenase metabolite. 3. Identification of a deficiency of fetoplacental lipophilic NK suppressive activity in the CBA/J X DBA/2 conceptus which correlates with NK infiltration and fetal resorption. Phenotype analysis of the lipophilic NK regulatory fetoplacental extract by radioimmunoassay showed that CBA/J X DBA/2 concepti were deficient in the production of the lipoxyqenase metabolites LTB4 but not the cyclooxygenase metabolite PGE2. 4. Increased fetal viability was observed in CBA/J X DBA/2 second matings. These second matings showed LTB4 and PGE2 concentrations that more closely resembled those observed in normal pregnacies, suggesting that the uterus may retain the memory of 0 as c a "priming" event even from a non-compatible male. 5. In vitro experiments suggested that the local concentration of LTB4 may be critical in averting NK mediated fetal resorption. 6. Early stages of normal pregnancy (day 7-8) are extremely sensitive to low doses of bacterial LPS, with 0.1 ug inducing lOO% fetal resorption within 24 hours. 7. NK depleting treatments can protect against LPS induced fetal resorption. s. LPS induced fetal resorption involves the release of TNF-a into the amnionic fluid. 9. Both LPS induced and spontaneous fetal resorption can b prevented by pre-treatment of the pregnant mice with the TNF-a suppressing drug pentoxifylline. 10. A novel form of TNF-a with a molecular weight of so,ooo daltons is transiently expressed in the developing nervous systems of vertebrates.

0 a9 c TABLE 01' COB'l'EN'l'S

Abstract al

Acknowledgments as

Claim of Contribution To Literature a7

List of Abbreviations a15

List of fiqures a17

List of Tables a19

CHAPTER I. LITERATURE REVIEW 1

A. J:KMUNOLOGY Ol!' PREGNANCY AND l!'ETAL LOSS IN .MAMMALS

1.. Historical Development and Introduction 1

B. THE ROLE 01' MBC IN FETAL SURVIVAL

1. MHC Antigen Sharing in Human Couples 4 0 alO c

2. MHC Antigen Expression in the Conceptus 6

3. MHC Antigen Expression in the Uterus 9

C. CELLULAR LINEAGES IN 'l'BE PREGNAN'l' UTERUS AND 'l'BEIR

RELEVANCE '1'0 FE'l'AL SURVIVAL

1. Cells of the Uterus and Conceptus a) Uterine Epithelium 11 b) Decidual Cells 12 c) Granulated Metrial Gland Cells 13 d) Trophoblast Cells 14

2. Cells of the Immune System

a) Myeloid Lineages 16

b) T and B Lymphocytes 18

c) Null Lymphocytes and Killer Cells 21

D. IMMUNOMODULA'l'ORY FAC'l'ORS AND THEIR RELEVABCE '1'0 FE'l'AL

SURVIVAL

1. Protein Factors

a) Alphafetoprotein 24

b) Transforming Growth Factor Beta 26 0 all c c) Partially Identified Immunosuppressive

Factors . . • . • • • . • . . • • • . • • • • • . • . • . • • • . • • • • 27

2. Sex Steroids

a) Sex Steroids and Fetal Loss 28

b) Progesterone 28

3. Arachidonic Acid Metabolites 30

E. EXOGENOUS PACTORS AND PETAL LOSS

1. Pathogens

a) Bacteria, Bacterial Products

and Parasites ...... 3 3

2. Water 37

P • SPONTANEOUS ABORTI:OH I:H HUMAN'S AHD ADBQUATB AN:IMAL MODELS

1. Introduction 38

2. Human Spontaneous Abortion 39

0 al2 c 3. Fetal Resorption in Animal Models a) Interspecific Matings 42 b) The CBA/J X DBA/2 Model of Spontaneous Fetal Resorption •••••••••••••••••••••••• 46

G. RATIONAL FOR THIS THESIS 50

RESULTS

CHAPTER II PREFACE. THE ROLE OF NK CELLS IN THE CBA/J X DBA/2 MURINE SPONTANEOUS RESORPTION MODEL 53

A. Infiltrating Decidual Natural Killer Cells Are Associated With Spontaneous Abortion in Mice ••.•. 55

B. Morphometric Analysis of the Histology of Spontaneous Fetal Resorption in a Murine Pregnancy •••••••••.. 62

c. Alteration in the Infiltration of Natural Killer Cells in CBA/J X DBA/2 Concepti by POLY I:C and Anti- Asialo GMl Treatment •••••••••.••.•••••••.•••••••• 72

CHAPTER III PREFACE. THE REGULATION OF BK ACTIVITY IB

THE FETOPLACENTAL UNIT AND DEFICIENT NK SUPPRESSION IB 0 all

SPONTANEOUSLY RESORPTION-PRONB CB.A/J X DBA/2 CONCEPTI: THE ROLB OF EICOSANOIDS. 78

A. Resorption of CBA/J X DBA/2 concepti in CBA/J Uteri Correlates With Failure of the Fetoplacental Unit to Suppress Natural Killer Cell Activity ••••••••••• 80

B. Murine Pregnancies Predisposed to Spontaneous Fetal Resorption Show Alterations in the Concentrations of Leukotriene B4 and Prostaglandin E2 •.•.•.••...••• 87

CHAPTER IV PREFACE. THE DUAL ROLE OF TOMOR NECROSIS FACTOR ALPHA IN EMBRYONIC DEVELOPMENT AND FETAL RBSORPTION 111

A. Lipopolysaccharide Induced Fetal Resorption in Mice is Associated With the Intra-Uterine Production of Tumor Necrosis Factor Alpha ....•....•.•.•••••.•• 113

B. Expression of Tumor Necrosis Factor Alpha in the Developing Nervous System •..•..•.•..•.•••••••••• 121

CHAPTER V. GENERAL CONCLUSIONS 148

0 a14 c CHAPTER VI. REJ'EREBCES 160

0 al5 c LIST 0~ ABBREVIATIONS

UP Alpha-fetoprotein

AGMl Asialo-GMl

APC Antigen presenting cell

ASA Acetyl salycilic acid

BCG Bacillus Calmette Guerin

Con A Concanavalin A CSF-1 Colony stimulating factor 1

CTL Cytotoxic T lymphocyte

E2 Estradiol

GM-CS~ Granulocyte macrophage colony stimulating factor GMGs Granulated metrial gland cells

GVBD Graft versus host disease hCG Human chorionic gonadotrophin

HETE Hydroxyeicosatetraenoic acid

HPF High power field

HPLC High pressure liquid chromatography LAX Lymphokine-activated killer cell LPS Lipopolysaccharide

LTs Leukotrienes MHC Major histocompatibility complex

MHV Murine hepatitis virus c a16 0 MLR Mixed lymphocyte reaction mol Mole

NC Natural cytotoxic cell

NK Natural killer cell

NS Natural suppressor cell

ODS Octadecylsilyl

OCT Optimal cryogenic temperature compound

PBS Phosphate buffered saline

PGs Prostaglandins

POLY I:C Polyinocynic cytidilic acid

PMNs Polymorphonuclear leukocytes

PXF Pentoxifylline

SCID Severe combined immune deficiency

SDS Sodium dodecyl sulphate

SDS-PAGE SDS-Polyacrylamide gel electrophoreses

TGFs Transforming growth factors

TNFs Tumor necrosis factors TXs Thromboxanes

0 a17 0 LIS'l' 01' J'IGORBS PAGB

1. Frequency of fetoplacental units with respect to number of asialo-GMl (GMl)-positive cells per high 'power field and gestational stage in days ••••••• 57 2. Mean number of lymphocytes as a function of gestational stage in normal and resorbing fetoplacental units .•.••••.••••••••••••••••••••• 58 3. Normal and preresorptive fetoplacental units stained for the localization of NK cells •.•.••.•••••.••• 60 4. Decidual-ectoplacental cone of normal and resorbing embryonic units at day 8 •••••••••••••••••••••••• 65 5. Placental structure of normal and resorbing embryonic

units at day 10 ...... 66 6. Nuclear density per high power field versus days of gestation in normal and resorbing embryonic

units ...... 67 7. Nuclear area per high power field versus days of gestation in normal and resorbing embryonic

units ...... 68 8. Differential expression of NK suppressive activity as a function of day of gestation in normal CFW/SW X DBA/2 pregnancies • • • • • • • . • • • • • • • • • • • • • • • • • • • • • • • 82 9. Representative data showing the NK inhibitory 0 a18

activity of individual fetoplacental units from single pregnancies •••••••••••••••••••••••••••••• 84 10. NK-modulating effects of LTB4 and fetoplacental unit

extracts ...... ·· ...... 110 11. Resorption frequency in normal pregnant female mice as a function of administration of LPS on different days of gestation ••••••••••••••••••••••.••.•.•• 116 12. Expression of TNF-a in the brain of normal mouse and chick embryos ...... 13 2 13. Expression of TNF-a in the spinal cord and peripheral mixed spinal nerves in normal mouse embryos ...... 133 14. Western immunoblot analysis of TNF-a in normal embryonic mouse and chick brain ••••••••.••••.•• 134 a19

L:IST OP TABLES PAGE

1. Nuclear density per hiqh power field in the decidual and ectoplacental cone reqions on days 8 and 9 in normal and resorbinq fetoplacental units •••••••• 68 2. NK inhibitory activity of CBA/J X DBA/2 and CBA/J X BALB/c preqnancies • • • • • • • • • • • • • • • • • • • • • • • • • • • • • . 83 3. Correlation between resorption, NK suppression deficiency and NK infiltration in mouse

pregnanci~s ...... 83 4. LTB4 and PGE2 concentrations in primiqravid CBA/J X DBA/2, second CBA/J X DBA/2 and CBA/J X BALB/c

matings ...... 108 5. TNF-a in amnionic fluid of normal preqnant mice treated with lipopolysaccharide and inhibition by pentoxifylline ...... •....• 115 6. Protection aqainst LPS-induced resorption by pentoxify11ine in normal pregnant mice •..•...••. 116 7. Prevention of LPS-induced resorption by treatment with anti-asialo GMl in normal preqnant mice •••• 117' c 1 c CBAP'l'BR I. LITURATURB RBVIBW

A. TBB IMMUNOLOGY OF PREGNANCY AND l!'BTAL LOSS Ill MAMMALS

1. Historical Development and Introduction

With the exception of syngeneic in-breeding in MHC-identical laboratory animals, the majority of mammalian pregnancies in the natural world are allogeneic. As a result of codominant parental gene expression, both maternal and paternal histocompatibility antigens are expressed on conceptus tissue during allogeneic gestation. Grafts of tissue from these offspring are promptly rejected by the maternal immune system. Hence the maternal tolerance of the potentially foreign "fetal allograft" remains one of the most fascinating and partially unsolved problems in reproductive biology. Peter Medawar was the first modern scientist to document in the literature a hypothesis to explain the non-rejection of the fetus in viviparous pregnancy (Medawar, 1937). Although parts of Medawar's initial hypothesis have since been disproved, his reasoning that the maternal anti-fetal immmune response may be locally suppressed within the uterus has been 0 2 c supported in recent years. In 1966, Kirby, Billington and James demonstrated that the uterus can prevent maternal immunological destruction of the embryo. Embryos were transplanted to the kidney and uterus of hyperimmunized mice. Development of allogeneic embryos failed in the kidney while control syngeneic embryos developed normally in this area. When placed in the uterus of hyperimmunized mice, allogeneic embryos survived (Kirby et al, 1966). Thus, it became clear that the uterus can provide potentially foreign allogeneic embryonic tissue local protection from maternal immunologically mediated destruction. During the time between blastocyst implantation into the steroid-primed uterine wall and mid-gestation (full placental development), a number of events take place in which immunological reactions could be potentially factorial in maternal-anti-fetal rejection. Among the factors determining the outcome of the apposition of maternal and fetal tissues are the expression of MHC antigens, the presence in the uterus of potentially reactive maternal leukocytes, and the invasive nature of the fetal tissue itself. In the course of implantation and during the early differentiation of the invading fetal tissue, maternal 0 3

blood vessels in the uterine wall are invaded by trophectodermally-derived cells. As discussed below, these trophoblast cells may express codominantly inherited paternal antigens that may stimulate maternal immune reactions. After implantation, a complex process of tissue organization in both the maternal and fetal compartments results in the formation of the placenta, which functions in a nutritive and maintenance role as well as an immunological barrier to the maternal immune system. Some of the most critical developmental events related to tolerance of the fetal allograft are likely to occur during the early phases of gestation during which time there is the most direct contact between fetal and maternal tissue. The most reasonable explanation for the survival of allogeneic concepti is the suppression of.maternal­ anti-fetal immune reactions. Theoretically, these immune reactions could either be specific MHC-reactive T and B cell responses or non-specific responses comprised of events mediated by NK, LAK cells or macrophages. Much of the liturature in human spontaneous fetal loss concentrates on specific cellular immune reactions, and therapies designed to modulate them. In contrast, the liturature on fetal resorption in mice points mainly to 0 4

non-specific immune reactions. The contents of the results section of this thesis will support the view that non-specific reactions are important in fetal resorption in mice. The immunoregulatory mechanisms present in the murine fetoplacental unit are particularly suited to provide the local suppression of killer cells. The deficiency of these regulatory factors in spontaneously resorption-prone pregnancies emphasizes their importance. The purpose of this liturature review is to familiarize the reader with both views (specific vs non­ specific immune reactions) on the role of the immune system in fetal survival. I will first describe the role of MHC-related events and the tissues and factors related to the fetoplacental unit. This will be followed by a discussion of human fetal loss and the development of animal models to address this problem.

B. THE ROLE OF MHC XH FETAL SURVIVAL

1. MHC antigen sharing in human couples As described in more detail below, spontaneous abortion in humans has been demonstrated to be associated with the sharing of MHC antigens between partners (Beer et al., 1981: Reznikoff-Etievent et al., 1984: Faulk and 0 5

Mcintyre, 1986) •.There is however a considerable degree of controversy over the importance of class I vs class II antigen sharing as well as whether such couples are ever normally fertile. A recent study by Ober et al. (1985) of Hutterite couples provides some rare insight into this problem. The Hutterite colonies were established by a very limited group of founding couples almost 200 years ago. Since colonized Hutterites do not practice contraception, marry monogamously from within their own colony and generally have large families (median family size·= 9), these people present a valuable example of multiparity and MHC antigen sharing in a fertile population. Hutterite couples sharing more than one class I or class II MHC antigen showed an increased median interval between marriage to first and fifth births. In addition, couples who shared HLA-DR class II antigens showed longer intervals that were associated with a 27% spontaneous abortion frequency as compared to couples sharing class I (12%) or none (9%). There were fewer offspring in couples sharing class II antigens as compared to those sharing class I antigens (5.0 vs 8.5). Interestingly, some couples sharing class II antigens did not suffer from spontaneous abortion although family sizes in this group were 10 or higher. This study 0 6

I"·· ·.~ clearly demonstrates that in some cases, sharing of class II MHC rather than class I antigens is indeed associated with increased spontaneous abortion. However, since other couples sharing class II antigens did not show increased abortion rates, there are probably undefined genes associated with the HLA locus that may be more important.

2. MHC Antigen Expression in the Conceptus Peter Medawar's initial analogy of the fetus to an allogeneic graft implied the presence of potentially foreign paternal MHC antigens on conceptus tissue. Although both the embryo and placenta codominantly inherit both sets of parental MHC-encoding genes, MHC antigen expression is functionally limited in the conceptus. The trophectodermally-derived tissues of the blastocyst and placenta are directly exposed to the maternal immune system during gestation, and thus constitute the interface between mother and fetus. In 1976, Page Faulk and Temple demonstrated using immunofluorescence cytochemistry that human villus trophoblast lacks the expression of B2-microglobulin and HLA antigens. In addition, amniotic epithelial cells express extremely low levels (detectable only by 0 7

sensitive radiobiological techniques) of B2- microglobulin, HLA class I and class II antigens (Adinolfi et al., 1982). In both the human and mouse, only class I MHC antigen expression can be detected on cytotrophoblast tissue (Lala et al, 1983). Since allograft survival is more successful! with class II MHC antigen matching, it is possible that the lack of paternal class II antigens on trophoblasts minimizes the potential for maternal anti-fetal reaction. Class I antigens expressed on pre-implantation embryos disappear immediately after implantation (Searle et al., 1986; Hakansson et al., 1975). The re-expression of class I remains undetectable until the mid-somite stage at day 10-12 of gestation (Jenkinson and owen, 1980; Ozato et al., 1985). Class I gene expression in embryonic tissue shows a gestational-stage specific inducibility by interferon (Ozato et al., 1985). Since placental tissues produce interferons (Fowler et al., 1980), it is conceivable that there exists an endogenously controlled regulation of fetal MHC antigen expression. A recent report demonstrates complex patterns of class I mRNA

expression in ecto-plac~ntal cone tissue as early as day 7.5 in the mouse (Head et al., 1987; Hedley et al., 1989). In addition, trophoblast cells are thought to 0 8 c express a truncated form of the class I heavy chain, which may contribute to its ability to evade the maternal immune response (Earl et al., 1985). Croy and Rossant (1987) have demonstrated that murine embryonic cells become susceptible to CTL lysis only after midgestation. Non-susceptibility to cellular attack at earlier stages may thus be related to the functional limits of MHC antigen expression during the first half of gestation. By inducing the expression of class II antigens on labyrinthine trophoblast with 5-azacytidine at times before ectoplacental cone formation in pregnant mice, Athanassakis et al. (1990) oberved a dramatic increase in fetal loss. In addition, the abortive effect of 5- azacytidine treatment can be blocked by pretreating allo­ pregnant females with anti-MHC class II antibody. It is thus possible that the potential maternal anti-fetal CTL response is averted in the initial stages of gestation by the absence or modification of specific MHC target structures on trophoblast cells. During the later stages of gestation when class I antigens begin to be re­ expressed on conceptus tissues, the maternal uterine resistance mechanisms such as blocking antibodies and local immunosuppressive factors may prevent responses directed against paternal MHC antigens. 0 9

With regards to whether MHC antigens are necessary for a viable pregnancy, the recent demonstration of healthy B2-microglobulin deficient transgenic progeny mice clearly shows that class I MHC antigen expression on fetal tissues is not needed for viable birth (Zijlstra et al., 1990). On the other hand, the absence of MHC target structures on trophoblast cells may explain why some allogeneic concepti are rejected via an NK-related mechanism. Bix et al. in a recent report (1991) utilizing B2-microglobulin gene disruption demonstrated that NK cells in normal mice reject bone marrow transplant cells deficient in cell-surface MHC expression. Thus, if trophoblasts are functionally comparable to these MHC-deficient haemopoietic cells in their lack of B2-microglobulin and MHC, they may present as "rejectable" targets under certain conditions.

3. MHC Antigen Expression in the Uterus An important parameter for the immune response against paternal or conceptus antigen is the processing of such antigens by class II positive APC in the fetoplacental unit and their presentation to the cells of the maternal immune system. Pregnancy uterine epithelium has been shown to lack the expression of MHC class I and 0 10

II antigens in immunohistological studies done with antibodies specific for monomorphic determinants and beta-2 microglobulin (Johnson and Bulmer, 1984). Studies carried out in the rodent and human indicate that both non-pregnant and pregnant non-epithelial uterine tissues contain appreciable numbers of class II positive cells. Class II positive cells appear in endometrial stroma and in trophoblast-infiltrated decidua (Head and Gaede, 1986; Bulmer and Sunderland, 1984; Goldsobel et al., 1986). In addition, macrophages have been identified in mouse decidua and placenta using specific markers such as F4/80 (Matthews and Searle, 1985). The potential thus exists for the processing and presentation of paternal or fetal antigen by APC in the pregnant uterus. Peripheral blood samples of women suffering from chronic habitual abortion do not show any significant change in the numbers of class II positive lymphocytes (Fiddes et al., 1986). However, there are no studies to date that present a clear analysis of the numbers of class II positive cells in human uterine tissue during spontaneous fetal loss.

C. MAJOR CELLULAR LINEAGES IN THE PREGNANT UTERUS AND THEIR RELEVANCE TO FETAL SURVIVAL

0 11

1) Cells of the Uterus and conceptus

a) Uterine Epithelium The first tissue that the implanting blastocyst encounters is the maternal uterine epithelium. Uterine epithelial cells produce large amounts of colony stimulating factor-1 in response to the steroid hormones secreted during pregnancy (Pollard et al., 1989). The CSF-1 cytokine has been hypothesized to be involved in the development of the placenta due to the simultaneous expression of CSF-1 receptors (c-fms) in trophoblast tissues (Regenstreif and Rossant, 1989). The exact function of CSF-1 is not presently known. However, the relevance of CSF-1 expression to fetal viability is exemplified by the osteopetrotic (opjop) mouse. This mouse completely lacks CSF-1 due to a nonsense mutation in the CSF-1 gene (Wiktor-Jedrzejczak et al., 1990). In addition to the systemic effects of CSF-1 and macrophage deficiency, the opjop mice exhibit very low fetal viability (Marks and Lane, 1976). Also proposed to be embryotrophic during gestation, GM-CSF is produced by the uterine epithelium in response to steroid hormones and a non-MHC, non-sperm associated component of seminal fluid. Treatment of spontaneous resorption-prone CBA/J X DBA/2 0 12

pregnancies with recombinant GM-CSF has been reported to significantly decrease the resorption frequency in these mice (Chaouat et al., 1990). The exact mechanisms involved in the protective effect of GM-CSF remains to be determined. It is conceivable that a deficiency in the embryo-trophic properties of the uterine epithelium may play a role in fetal loss, although early immunological fetal losses appear to occur sometime after implantation (Croy et al., 1982).

b) Decidual Cells The decidual tissue of the pregnant or pseudopregnant uterus consists of differentiated, hormonally regulated, bone marrow derived cells (Kearns and Lala, 1982). Decidual tissue arises after a deciduogenic stimulus (naturally by a blastocyst or artificially by oil) of hormone-primed endometrial stromal fibroblasts. The decidua surrounds the implanted embryo and trophoblast, protecting the uterine wall from trophoblast invasion (Lala et al., 1987; Weitlauf, 1988). Decidual tissue provides local uterine immunosuppression as allogeneic skin grafts survive longer when placed into a decidualized uterus compared to other locations in the body (Beer and Billingham, 1970). Recent evidence 0 13

demonstrates that decidual cell-derived immunoregu1atory molecules normally suppress maternal anti-fetal immune responses (Lala et al., 1990; Clark et al., 1990). Among the major lymphocyte subsets present in decidual tissue, cells of the null phenotype appear to be the largest in number. In addition, class II, MAC-1 positive phagocytic antigen presenting cells are present in significant numbers (Reviewed by Lala et al., 1987). Since it is invaded by trophoblast in hemochorial placental development, decidual tissue comes into direct contact with potentially foreign paternal and fetal. antigens. Thus, the determinant of the efficacy of defence against maternal anti-fetal reactions lies in the immunoregulatory properties in the decidua itself.

c) Granulated Metrial Gland Cells Granulated metrial gland (GMG) cells are large PAS (periodic acid schiff) positive bone marrow derived cells originating in the metria1 triangle (located in the mesometrial aspect of the implantation site) which are known to migrate throughout the fetoplacental unit during early stages of gestation (Stewart and Peel, 1980). GMG cells are thought to provide a nutritive role in the developing fetoplacental unit since they secrete a 0 14

variety of trophic cytokines including GM-CSF, TGF-b and Interleukins (Croy et al., 1989). However, GMG cells can also kill particular trophoblast cells in vitro, and thus have been hypothesized to regulate excessive or aberrant trophoblast growth in the fetoplacental unit (Stewart, 1987). Recent evidence demonstrates that GMG cells contain high levels of the cytolytic protein perforin (Parr et al., 1990: croy et al., 1991). Thus, GMGs possess the machinery of killer cells. Indeed, these cells can kill the NK tumor target cells YAC-1 if activated by interferon or POLY I:C (Croy et al., 1991). Interestingly, pregnancies obtained by matings of SCID/Beige mutant mice, which show a severe lack of decidual tissue, appear to contain significant numbers of normal GMG cells (Croy, 1990). At present, the possible role of GMG cells as mediators or effectors of fetal resorption remains obscure.

d) Trophoblast Cells The fetally-derived trophectoderm of the blastocyst is responsible for attaching to and invading the uterine epithelium during implantation. cytotrophoblast tissue located in the early ectoplacental cone develops into the placenta. The further 0 15

differentiated syncytialtrophoblast serves as the actual interface between the fetus and the maternal blood. The central importance of the trophoblast in the immunological status of the fetoplacental unit was elegantly demonstrated in the reports by Croy and

collegues utilizing the Mus musculus <-> Mus caroli xenogeneic pregnancy model. By constructing embryonic chimaeras they showed that maternal (Mus musculus) xeno- reactivity to Mus Caroli embryos could be averted by the presence of Mus musculus trophoblast in the implant sites containing the Mus caroli embryos. The presence of the correct genotype trophoblast also led to the presence of adequate suppressor cell activity in the fetoplacental unit (Croy and Rossant, 1986). These studies led to the thought that trophoblast tissue may direct (a) specific signal(s) to decidual cells in order to establish the necessary amount of local immunosuppression required for averting maternal anti-fetal attack. Indeed, the experiments by Kiger et al. (1985) involving vaccination of CBA/J female mice with BALB/c splenocytes to prevent spontaneous fetal resorption in matings with DBA/2 males have been hypothesized to be an example of an artificial replacement of the normal signalling between trophoblast and decidua (Clark, 1989). 0 16

Trophoblast cells produce cytokines and express cytokine receptors which may be involved in the development of the fetoplacental unit (Hunt, 1989). Trophoblastic giant cells of the placenta express message for the c-fms proto-oncogene, the product of which is the recaptor for Colony stimulating factor-1 (CSF-1) (Regenstreif and Rossant, 1989; Arceci et al., 1988). Since CSF-1 is expressed and produced by the uterine epithelium, it is likely that there is considerable cross-talk between the maternal and fetal compartments via CSF-1 and its recaptor. Other cytokines produced by fetoplacental tissue which can interact with receptors on

trophobl~st cells include IL-l, TGF-b, GM-CSF and TNF-a (Reviewed by Hunt, 1989). Murine trophoblasts from midgestation can be killed by both LAK cells and IL-2 activated NK cells (Drake and Head, 1989; Lala et al., 1990). Both LAK and NK cells are present in the fetoplacental unit (Lala et al., 1990: Croy et al., 1985). Thus, the potential exists in the fetoplacental unit for cellular anti-trophoblast immune reactions.

2. Cells of the Immune System

a) Myeloid Lineages 0 17

Cells of the monocyte/macrophage lineage are present in high numbers in the decidua in both the human and murine pregnant uterus (Bulmer and Johnson, 1984; Matthews et al., 1985). Macrophages in the fetoplacental unit produce a number of cytokines such as CSF-1, GM-CSF, IL-l, TGF-b and TNF-a which may be involved in the development of the placenta or in the regulation of trophoblast cells. Uterine macrophages have recently been implicated in infection-associated pregnancy failure due to their ability to receive signals from bacterial products resulting in the production of cytolytic molecules (Reviewed by Hunt, 1989). LPS-induced early fetal resorption in mice involves the local production and release of TNF-a (Gendron et al., 1990). In addition, other groups have reported that infusion of recombinant TNF-a can cause placental necrosis and fetal abortion in the rodent, while infusion of GM-CSF can prevent spontaneous resorption in CBA/J X DBA/2 pregnancies (Chauoat et al., 1990). Another putative role for macrophages in the fetoplacental unit is the production of arachidonic acid metabolites such as prostaglandins. Prostaglandin E2 has been shown to be responsible for the suppression of maternal anti­ trophoblastic activity in the decidua by down-regulating 0 18

the expression of IL-2 receptors on decidual LAK cells (Lala et al, 1990) •. The myeloid cells in the fetoplacental unit may therefore be central to the regulation of events leading to either fetal viability or fetal resorption. The opjop mouse, discussed above, is also severely deficient in macrophages (Wiktor­ Jedrzejczak et al., 1990). Whether macrophage deficiency itself is related to the low fertility in the opjop mice remains to be determined.

b) T and B Lymphocytes The maternal immune system does not appear to mount any significant alloreactive cytotoxic T cell response during pregnancy although anti-paternal humoral immunity is detectable (Billington et al., 1983; Bell and Billington, 1983). Pregnancy is known to cause a significant hormone-mediated involution of the thymus

(Baines et al., 1973~ Chambers and Clark, 1979). Some groups have also reported a decrease in the numbers of T helper lymphocytes during early human pregnancy (Sridama et al., 1982; Barnett et al., 1983). A consequence of pregnancy-associated thymic involution is impairment of the IgG thymus-dependent humoral immune response (Baines et al., 1982). Not surprisingly, the alloantibodies 0 19

induced by placental tissue during pregnancy show low levels of complement mediated cytotoxic activity, being mainly of the IgG1 subclass (Baines et al., 1976; Bell and Billington, 1980). In addition, Wegmann has proposed that the placenta acts as an "immunabsorbant", taking up and degrading maternal anti-paternal-MHC antibody. In experiments using radiolabeled mouse anti-H2d antibody, the authors showed that placentae from female C3H/HEJ (H2k) mice mated with BALB/cCR (H2d) took up the antibody in a non-Fc manner while the same females mated with syngeneic males did not (Wegman et al., 1979). This finding prompted more specific studies utilizing abortion-prone matings. In subsequent experiments dealing with the BALB/c immunization protection of CBA/J X DBA/2 pregnancies the same group demonstrated that the protective IgG1 antibodies generated upon immunization of the CBA/J females were specifically absorbed by the placenta (Chaouat et al., 1986). With respect to the maternal humoral immune status, allogeneic pregnancy has been demonstrated to proceed uncomplicated in mice lacking a humoral immune response by complete B lymphocyte depletion (Rodger, 1985). In 1987, Wegman et al. proposed that allogeneically stimulated maternal T lymphocytes in the 0 20

pregnant uterus produced growth factors which were "immunotrophic" to placental cells. This hypothesis was based on the finding that the pre-immunization of pregnant mice with splenocytes to avert fetal loss is associated with increased fetal and placental weight (Wegmann et al., 1987). In addition, depletion during allogeneic pregnancy of maternal CD4 and cos positive T cells using monoclonal antibody treatments causes a decrease in placental cellular proliferation and phagocytosis. However, fetal viability is not affected in these T cell depleted pregnant mice (Athanassakis et al., 1987). This hypothesis was subsequently questioned when Croy and Chapeau demonstrated that SCID-SCID/Biege­ Biege mutant mice, completely lacking T lymphocytes, were able to gestate normal litters (Croy and Chapeau, 1990). Recent work from Clark's laboratory has demonstrated a role for a population of hormone-induced preimplantation Ly-2+ uterine cells in the establishment of decidual suppressive activity in pregnant mice. These Ly-2+ suppressor cells show a non-MHC specific suppressive activity against CTL generation and appear to be regulated by a separate Fe receptor-positive population of uterine cells. Furthermore, the administration of anti-Ly-2 antibody to resorption-prone CBA/J X DBA/2 0 21 c pregnant mice can increase the spontaneous fetal resorption frequency (Clark et al., 1990). Tezabwala et al., (1989) have presented data showing that administration of the T cell derived lymphokine IL-2 to pregnant mice induces fetal resorption. The requirement for full maternal T cell competance as a prerequisite for successful! pregnancy may thus be supportive rather than absolutely essential.

c) Null Lymphocytes and Killer Cells Lymphocytes of the null phenotype have been the subject of intensive research in reproductive immunology over the past 5-6 years. Croy's laboratory in Guelph first demonstrated that normal murine decidual cell suspensions contain high levels of NK activity (cell mediated lysis of YAC-1 targets) during the early stages of pregnancy between days 6-8 of gestation that reach negligible levels by days 9-12 (Gambel et al., 1985). High levels of NK activity are also found in non-decidual porcine pregnancy endometrium (Croy et al., 1988). These uterine NK cells do not appear to attack fetal or trophoblast tissues, and do not seem to be detrimental to normal pregnancy. However, NK activity is absent from the decidua of beigejbeige mutant mice, which can 22 c reproduce normally (Croy et al., 1985). Recent work presented in this thesis demonstrates that normal pregnancies contain a local NK suppressive activity peaking on the days corresponding to a decrease in decidual NK activity (Gendron et al., 1990). These studies suggest that NK cells present in the fetoplacental unit may either serve a dispensable role or be suppressed promptly enough to prevent detrimental cytotoxic activity in normal pregnancy. When assessed by quantitative immunohistochemistry, the numbers of NK cells in putatively resorbing fetoplacental units in CBA/J X DBA/2 concepti are significantly increased. The administration of NK-depleting antibodies such as aAGM1 is known to reduce fetal loss and decrease fetoplacental NK cell numbers in cases of spontaneous and induced resorption. Conversely, boosting systemic NK activity in pregnant mice with POLY I:C induces increased resorption and increased fetoplacental NK cell numbers (deFougerolles and Baines, 1987; Gendron and Baines, Unpublished observation: see results section). Other investigators have confirmed and extended these findings. Chaouat et al. (1989) reported that POLY I:C can induce NK-mediated fetal resorption in various strains of mice. CBA/J X 0 23 c DBA/2 concepti gestating in CBA/J uteri show NK cell infiltration of the fetoplacental unit. This deficiency correlates with a deficiency in a lipophilic NK suppressive activity. Our most recent results, reported in this thesis, indicate that the deficiency in NK suppressive activity in CBA/J X DBA/2 concepti involves a decreased local production of leukotriene B4 (Gendron et al., 1990; see results section). Recent work from Lala's laboratory has clearly demonstrated the importance of decidua-derived prostaglandin-E mediated NK and LAK cell regulation in the mouse and human fetoplacental unit. Killer cell activity can be rescued from decidual cell suspensions in the presence of indomethacin, anti-PGE2 antibody and IL-2. Such treatment leads to the expression of IL-2 receptors on LAK cells and the ability to kill trophoblast and NK targets. In addition, administration of indomethacin or IL-2 to pregnant mice leads to high rates of fetal resorption (Scodras et al., 1990; Lala et al., 1990). These studies taken together suggest that decidual NK or LAK cells may become detrimental to the fetoplacental unit in states of activation or in cases of compromised local regulation. Natural suppressor (NS) lymphocytes have been shown to represent another source of local nonspecific 0 24 suppressive activity in the fetoplacental unit. The spleens and decidua of both iso- and allo-pregnant mice contain suppressor cells of a null phenotype which can suppress mixed lymphocyte reactions (Hoskin et al., 1985). Administration of monoclonal antibodies directed against NS cell markers can induce fetal resorption in mice (Gronvik et al., 1987). Pope and collegues have demonstrated that pregnancy-associated NS cells produce suppressive factors in response to prostaglandins originating from a second population of NS-inducer cells (Pope et al., 1988). It remains unclear whether the non- T suppressor cells reported by Clark's group are the same cells as those reported as natural suppressors by other groups.

D. IMMUNOMODULATORY FACTORS AND THEIR RELEVANCE TO FETAL SURVIVAL

1. Protein Factors a) Alpha-Fetoprotein AFP is a 70,000 dalton immunoregulatory protein produced by the mammalian fetal liver and yolk sac (Reviewed by Murita and Wigzell, 1981). In addition to 25 c its immunoregulatory roles, AFP also acts as a carrier protein for various waste products and metals in the amniotic sac and embryo. The immunoregulatory properties of AFP include suppression of CTL generation, allogeneic MLR, NK activity and T cell dependent antibody responses. In addition, AFP is .known to induce natural and non-T suppressor cells (Reviewed by Van Oers, 1990). The suppression of MLR by AFP in the mouse shows a high degree of variability depending on the strain combinations comprising the reactions. Peck et al. reported in 1978 that the AFP regulation of MLR was

directed spe~ifically against MHC class II incompatibilities between the stimulator and responder cells. Interestingly, in some strain combinations AFP induced a stimulation of the MLR. Among the strain combinations showing AFP regulated MLR augmentation were CBA/J anti-DBA/2 responses. Whether this AFP-mediated regulation of MLR is associated with the decreased immunosuppression of T cell reactions and increased fetal resorption in CBA/J X DBA/2 matings is not known. AFP can potently suppress NK activity stimulated by interferon and interleukin, but has no appreciable effect on resting NK activity (Cohen et al., 1986). Thus, the modulation of NK activity by AFP in the fetoplacental 26 c unit may become important in a regulatory manner under conditions involving increased local lymphokine production. The administration of anti-AFP antibodies to pregnant mice has been shown to induce fetal resorption (Mizejewski and Vonnegut, 1983). However, it is not known whether anti-AFP induced fetal resorption involves cellular immune reactions in the fetoplacental unit or an interuption of the other physiological roles of AFP such as waste removal.

b) Transforming Growth Factor Beta Members of the transforming growth factor beta family of cytokines are developmentally regulated proteins showing close structural and functional similarity (Cheifetz et al., 1987: Whitman and Melton, 1989). In 1988, Clark et al. reported that the IL-2 dependent CTL suppressive activity associated with non-T cells in murine allopreqnancy decidua was neutralizable with an anti-TGF-b antibody. Using HPLC analysis of decidual supernatants, Clark et al. showed that the suppressive activity was concentrated in a 13,000 dalton fraction under acidic conditions, and in several different fractions under neutral conditions. These data suggested that the TGF-b-like factor normally associates . 0 27

with carrier proteins in vivo. A later report further characterized the molecular nature and immunosuppressive activity of this unique member of the TGF-b family (Clark et al., 1990). In 1990, Altman et al. described a 25,000 dalton TGF-b-like protein present in mouse amniotic fluid that complexed with AFP and exhibited both TGF-b-like activities as well as suppression of antigen induced T cell proliferation. In addition, peak levels of TGF-b mRNA were detected in pregnant uterine tissue between days 9 and 15 of gestation. These studies suggested that TGF-b may serve as a local immunosuppressive factor during early-mid gestation.

c) Partially Identified Immunosuppressive Factors Various groups have reported on immunosuppressive factors present in decidual cell supernatants with wide­ ranging molecular weights. Badet et al., (1983) describe immunosuppressive factors in mouse decidual supernatants with molecular weights of <1500, 60,000 and >300,000. Clark et al., (1985) described allopregnant mouse decidual supernatants as having a 100,000 dalton immunosuppressive factor. Finally, Matsui, et al., (1989) demonstrated that serum-free supernatants from purified human decidual cells contain a factor with 0 28 molecular weight between 43,000-60,000 daltons that suppresses IL-2 and interferon-gamma production as well as IL-2 receptor expression.

2. Sex steroids a) Sex Steroids and Fetal Loss Among the steroid hormones, estradiol and progesterone exhibit the most crucial roles in pregnancy­ associated uterine changes in addition to being immunosuppressive (Psychoyos, 1973; Pavia et al., 1979). Most of the studies on immunologically-related human spontaneous abortion have been reported for patients showing no underlying endocrinological abnormalities (Beer et al., 1981; Mcintyre et al., 1986; Reznikoff-· Etievant et al., 1984). In addition, maternal serum progesterone concentrations during resorptio-prone pregnancies are not significantly different from those observed in non-resorbing pregancies (Gendron and Farookhi, unpublished observation). It is thus unlikely that changes in systemic concentrations of steriods are involved in immune-mediated spontaneous fetal loss. b) Progesterone 29 c Among the sex steroids, the immunoregulatory role of progesterone has been studied in sufficient depth to warrent discussion. The major immunoregulatory effect of progesterone is the suppression of the proliferative and cognitive phases of the immmune response (Pavia et al., 1979). Szekeres-Bartho and Pacsa demonstrated in 1983 that progesterone binding capacity is significantly increased in lymphocytes from pregnant women. In a subsequent report, it was shown that the suppressive effect of progesterone on cellular cytotoxic activity is diminished in lymphocytes from women with threatened abortion as compared to normal pregnant women (Szekeres­ Bartho and Pacsa, 1985). However, the subjects in this report were studied at a very late stage of gestation. Since serum progesterone concentrations are considerably lower in the first trimester, it is impossible to relate these findings to early fetal loss. Nevertheless, a recent follow-up study utilizing a monoclonal anti­ progesterone receptor antibody demonstrated that COS+ lymphocytes from pregnant women express progesterone receptors, while lymphocytes from non-pregnant women.do not (Szekeres-Bartho et al., 1990). Van Vlasselaer et al., (1986), demonstrated that progesterone produced by murine trophoblast cells in culture can suppress 0 30

lymphocyte aggregation in vitro. However, the concentration of progesterone in the trophoblast culture supernatants was too low to show immunosupppresive effects by itself. Using an antiserum to progesterone or by adding IL-2 to the cultures, Van Vlasselaer et al. were able to abrogate the observed suppressive effects. Their findings suggest that progesterone may interact with other factors in the fetoplacental unit to effect suppression. Direct measurement by radioimmunoassay in fetoplacental units of the rat (Sio, 1984) demonstrates that the local progesterone concentration is indeed significantly lower than concentrations shown to be immunosuppressive in vitro. Kita et al., recently reported that progesterone treatment of con-A stimulated Ly2+ T cells results in the production of a 40,000-55,000 soluble factor that suppresses MLR and CTL generation by acting on T cells only (Kita et al., 1990).

3. Arachidonic Acid Metabolites

There is substantial evidence demonstrating that prostaglandins are synthesized by decidual cells in the pregnant uterus in both the rodent (Malathy et al., 1986; Parr et al, 1988) and the human (Smith and Kelly, 1988; 0 31 c Mitchell et al., 1990; 1991). A series of recent studies from Lala and collegues demonstrate that decidually­ derived prostaglandin E2 plays a major role in suppressing both MLR and non-specific maternal anti-fetal immmune reactions in the mouse and human. Parhar et al. (1988a) reported that human decidual cells purified by Ficoll-Paque and plastic adherence can suppress MLR and CTL responses in a non-MHC specific manner. This suppressive activity can be abrogated in the presence of indomethacin and can be restored by subsequent addition of pure PGE2. In a simultaneous paper, Lala et al., (1988b) reported that the mechanism behind PGE2-induced suppression of lymphocyte reactivity involves the suppression of both Il-2 receptor expression and IL-2 production. These findings were extended in a later study by the demonstration that decidual PGE2 also suppresses NK and LAK dependent maternal anti-trophoblast reactivity through modulation of IL-2 receptor expression and Il-2 production (Parhar et al., 1989). The importance of PGE2 was later confirmed by the demonstration that 89-100% of pregnant mice receiving chronic (post-implantation) indomethacin treatment showed fetal resorption. Administration of anti-AGM1 antibody to indomethacin-treated pregnant mice decreased 0 32

resorption to 7%, indicating that NK-like cells are involved in this process (Lala et al., 1990). Prompted largely by the early findings of Lala and colleques coupled with the known modulatory effects of eicosanoids on killer cells (Brunda et al., 1987; Rola­ Pleszczynski et al., 1986), we have hypothesized that concentrations of prostaglandins and leukotrienes may also be important in the regulation of NK activity in spontaneous fetal resorption. CBA/J X DBA/2 concepti show a deficiency in a lipophilic NK suppressive factor that correlates with resorption and fetoplacental NK infiltration. In addition, this suppressive activity can be partially abrogated by treatment of the pregnant mice with ASA (Gendron et al., 1990). Further phenotypic analysis of extracts containing this lipophilic NK suppressive activity revealed that resorption-prone pregnancies show significant changes in the fetoplacental concentrations of PGE2 and LTB4. In addition, the NK suppressive activity of extracts from resorption-prone CBA/J X DBA/2 pregnancies can be regained by adding sufficient amounts of pure LTB4 to approximate concentrations observed in normal fetoplacental units (Gendron et al., 1991; see results section). These results indicate that arachidonic acid metabolites play a 0 33

significant role in the regulation of NK activity in the normal fetoplacental unit. In addition, immune mechanisms mediating spontaneous fetal resorption may be closely associated with the capacity for the local production of arachidonic acid metabolites.

B. BXOGBNOUS FACTORS AND FBTAL LOSS

1. Pathogens

a) Bacteria, Bacterial Products and Parasites Early studies on infection-associated fetal loss concentrated mainly on the histo-pathological analysis of aborted conceptus tissue (Reviewed by Benirschke and Robb, 1987). Zahl and Bjerknes in 1943 first documented the abortifacient effect of endotoxins from gram-negative bacteria. These and other early studies were accomplished with relatively high doses (10-100 ug) of endotoxin at mid-late stages of gestation. Skarnes and Harper suggested in 1972 that PGF2a is involved in the fetal death and abortion in endotoxin-treated pregnant mice. Their conclusions were based on the hypothesis that abortifacient prostaglandins may be stimulated endogenously by bacterial products. However, they showed 0 34

in a subsequent report that indomethacin was unable to prevent fetal death induced by endotoxin, suggesting that another factor was involved (Harper and Skarnes, 1972). Bacterial endotoxin stimulates the release of TNF-a from primed macrophages (Beutler and Cerami, 1986). Furthermore, a recent study demonstrated that endotoxin and TNF-a can stimulate prostaglandin production by human decidual tissue (Romero et al., 1989). The hemhorragic, necrotic nature of the fetal death induced by endotoxin prompted us to measure intrauterine TNF-a levels during endotoxin induced fetal resorption in mice. Significant increases in TNF-a levels were detected in the amniotic fluids of mice administered 25 ug LPS at day 11 of gestation. In addition, the sensitivity of pregnant mice to low concentrations (0.1 ug) of LPS was found to be maximal at very early times in gestation (Gendron et al., 1990). Intrauterine infection has been reported to be associated with preterm labor in human pregnancy (Romero and Mazor, 1988). In a recent clinical study of pregnant women suffering from preterm labor, it was found that the amniotic fluid contained significantly elevated levels of LPS and TNF-a. The presence of these factors in amnionic fluid correlated with pre-term intrauterine infection (Casey et al., 1989). Pretreatment of pregnant mice with 0 35 c the TNF-a suppressing drug pentoxifylline can prevent both endotoxin induced and spontaneous fetal resorption in mice, suggesting that TNF-a may be a central effector molecule in fetal resorption (Gendron et al., 1990; See results section). However, the functions of TNF-a during gestation may be more complicated than the attributed role as a pathological mediator. Gendron et al., (1991), demonstrated that normal murine and avian embryonic nervous tissues express TNF-a at specific times during embryogenesis, suggesting a normal developmental role for TNF-a. This finding coupled with the reports discussed above raise the interesting possibility that detrimental exogenous stimuli coupled with placental insufficiency may also interfere with normal embryonic development. The interactions between the maternal and fetal compartment before and during spontaneous fetal loss must be carefully addressed in later studies. Parisitic infections such as malaria, toxoplasmosis and listeriosis during pregnancy are also known to be associated with increased fetal loss. However, the placenta appears to present a significant barrier as parisitemia often cannot be detected in the fetal compartment of infected aborters (Reviewed by Benirschke and Robb, 1987). In cases of abortion 0 36

associated with listeriosis, severe placental abscess formation has been hypothesized to be related to the breakdown of the placental barrier leading to fetal death. A recent report demonstrates the induction of fetal death by injection of small quantities of TNF-a in pregnant mice infected with malaria, whereas 10 times this dose of TNF-a has no effect in noninfected pregnant mice (Clark and Chaudhri, 1988). It is therefore possible that parisitemia may prime the maternal immune response, making abortifacient stimuli more effective. Placental villous tissue expresses TNF-a receptors with a dissociation constant almost an order of magnitude higher than that determined for various TNF-a sensitive cell lines (Eades et al., 1988). Since a wide range of pathogens appear to lead to the production of TNF-a during pregnancy, it is probable that one of the innate resistance mechanisms of the placental barrier is the binding and inactivation of released TNF-a. There is precedent for such an innate specific protective effect in dealing with released TNF-a in the body. Urine contains a specific soluble receptor that has been demonstrated to bind and effectively inactivate TNF-a (Englemann et al., 1989). TNF-a has been shown to be inhibitory to the proliferation of trophoblast cells 0 37

in vitro (Berkowitz et al., 1988). Whether the placental binding of TNF-a is protective or detrimental to the conceptus itself remains to be determined. The fetal expression of this protein may indicate a cross-talk between embryo and placenta. In a report utilizing the murine CBA/J X DBA/2 spontaneous resorption model, Hamilton and Hamilton (1987) presented data on the influence of the environment on fetal loss rates. By housing Jackson mice (sendai-negative, MHV-negative on arrival to facility) in either a specific pathogen-free facility or a conventional facility, the CBA/J X DBA/2 resorption rates were observed to be significantly lower in pathogen free surroundings. However, sentinal mice in both the pathogen-free and conventional rooms were reported to be MHV positive. Since MHV is known to be highly suppressive to cellular immune responses, it is conceivable that the MHV-free mice entering an MHV- positive facility may present a different immune status than the mice housed conventionally with exposure to other immunostimulatory pathogens.

2. Water Although there is a lack of extensive liturature on the subject, the presence of certain trace metals 0 38 c (arsenic, potassium and silica) in drinking water corellates with a small but significant increase in the spontaneous abortion rate in humans (Aschengrau et al.,

1989). An unpublished study from California demonstrates that women drinking tap water show an increased incidence of spontaneous abortion compared to women drinking bottled water (Water Exposure and Pregnancy Outcomes. Epidemiological Studies Section, California Department of Health Services, May, 1988: and personal communication). Bottled water is generally highly purified while most tap water contains significant levels of dissolved organic solids possibly including bacterial endotoxin. It is thus conceivable that the increased rate of spontaneous abortion in tap water drinkers may be related with exposure to small but possibly abortifacient concentrations of LPS.

P. SPONTANEOUS ABORTION IH JIUKl\HS AHD ADEQUATE ANIMAL

MODELS

1. Introduction

A surprisingly high number of allogeneic conceptions are not successful. The actual rate of 0 39

spontaneous fetal loss throughout the full duration of gestation in human pregnancies has been reported to be as high as 60%, based on clinical pregnancy diagnosis indicated by the increased concentration of the beta subunit of hCG in urine detected by radioimmunoassay (Edmonds et al., 1982; Wilcox et al., 1988). The rates of spontaneous fetal loss in mammals can vary greatly depending on the maternal environment, maternal age and the genetic combination of the mating (Bronson, 1985; Holinka et al., 1985; Kiger et al., 1982). Matt et al. reported in 1986 on the ratio of uterine implantation scars and litter size (dead and live) as an indirect measure of intra-uterine fetal resorption in ageing rats. It was found that the age-related failures in fecundity were due to events early in gestation, as the ratio of implantations to number of pups was significantly increased with increased maternal age.

2. Human Spontaneous Abortion

The most complete study of human spontaneous fetal loss was published by Wilcox et al., in 1988. This study reported a spontaneous fetal loss rate of up to 60% in pregnancies of normally cycling women. Earlier studies 0 40 had reported lower rates (15%) in humans (Roth, 1963). However, many fetal losses occur early during the first trimester of pregnancy arid may be undetected by the patient. The Wilcox study was based on measurement of the urinary concentration of the beta subunit of hCG, and thus represents the most accurate estimate of the number of detected and undetected combined losses. One of the most interesting findings in the Wilcox study is that women showing previous pregnancy failure are more likely to carry successful! pregnancies subsequently. Human couples suffering from chronic habitual spontaneous abortion of unknown etiology have been shown by a number of groups to exhibit an increased incidence of HLA class I and class II antigen sharing (Beer et al., 1981; Reznikoff-Etievant et al., 1984; Faulk and Mcintyre, 1986). These observations led to the hypothesis that sharing of MHC alleles contributed to an inability of the maternal immune system to recognize and produce blocking anibodies to paternal antigens expressed on conceptus tissue (Beer, 1986). Immunotherapies for human spontaneous abortion were subsequently developed consisting of paternal leukocyte transfusions prior to pregnancy (Mcintyre et al., 1986). The increased pregnancy success rate in treated aborters was correlated 41 with an increase in the maternal production of anti­ paternal blocking antibodies (Faulk and Mcintyre, 1986; Unander, et al., 1986; Takakuwa et al., 1986). Leukocyte transfusion therapy is relatively successful! in preventing habitual abortion in a highly select group of patients. such therapies have been reported to complicate cases of autoimmune disease but do not appear to endanger the immunological development of the fetus (Unander et al., 1986; Takakuwa et al., 1989). However, Beer et al. report an increased incidence (30%) of congenital anomalies in offspring of treated habitual aborters {Beer et al., 1987). The main obstacle to determining the etiology contributing to spontanequs fetal loss in humans is the realtive unavailability of adequate tissue specimens from times early enough in gestation. This combined with the ethical problems involved in the use of human conceptus tissues for research purposes has necessitated the development of suitable animal models of spontaneous abortion. It is noteworthy to define the terminology used in discussions of fetal loss. In singleton human pregnancies, fetal loss is usually referred to as "abortion", because of the sloughing of conceptus tissue 42

from the uterus. However, in polytocous mammals such as rodents, the gravid uterus contains more than one conceptus positioned linearly within each horn. The intrauterine death of part of a litter is thus dealt with

by the 11 resorption11 of conceptus tissue by cellular components of the maternal immune system. In the human, many early fetal losses may occur undetected by the patient due to the complete intrauterine resorption of small amounts of dead conceptus tissue (Wilcox et al, 1988). Therefore, the term "resorption" is synonymous with "abortion".

3. Fetal resorption in animal models

a) Interspecific Matings During the early 1980's, a series of papers were published describing the immunology of fetal loss in murine interspecific pregnancy. These studies provided crucial insight into the immunological mechanisms of xenogeneic fetal resorption which became valuable in the design of later experiments involving allogeneic spontaneous fetal loss. Transfer of Mus caroli embryos into the pseudopregnant uteri of xenogeneic Mus musculus mice results in implantation followed by complete fetal 0 c 43 resorption. All implanted embryos become infiltrated by pre-sensitized cytotoxic T lymphocytes on approximately day 10.5 of gestation and resorb rapidly thereafter (Croy et al., 1982). In addition, the Mus caroli embryos in Mus musculus uteri do not recruit the local uterine suppressor cells that are recruited in normal pregnancy (Clark et al., 1983). The reciprocal pregnancies, consisting of implanted Mus musculus embryos in Mus caroli uteri are successful (Croy et al., 1985). In a separate report, Rossant et al. (1982) elegantly demonstrated that the genotype of the trophoblast is critical for the gestational survival of interspecific embryonic tissue. By constructing aggregation chimaeras between Mus caroli and Mus musculus embryos, the authors showed that Mus caroli can survive successfully in Mus musculus uteri when Mus musculus trophoblast is present in the same fetoplacental unit. In order to further define the immune mechanisms associated with the death of Mus caroli in M. Musculus, Croy and colleagues performed the Mus caroli embryo transfers to immunodeficient Mus musculus recipients. Mus caroli embryos also failed when

transferred to chimaeric (Mus caroli <-> Mus musculus) recipients, NK cell deficient beige mutants, T cell deficient nude mutants and cyclosporin-A treated c c 44 recipients. In addition, the Mus caroli embryos are rejected more rapidly when the recipients are pre­ sensitized. The authors describe cells present in the failing implants with lytic activity against a broad range of targets including recipient cells, cells of 3rd party genotype and NK targets such as YAC-1 and P815 (Croy et al., 1985). The authors showed in a subsequent report that Mus caroli embryonic failure in immunodeficient recipients may be accounted for by the immunological mechanisms inherent in each case. For example, nude recipients possess activated NK cells which, interestigly, mediate an earlier anti-embryonic response. Finally, Mus caroli embryos recruit significantly less suppressive activity than Mus musculus embryos, even when gestating in a Mus caroli unterus (Clark et al., 1986). Crepeau and Croy (1988) confirmed that specific cellular immunity is not involved in the failure of Mus caroli embryos in Mus musculus uteri by observing Mus caroli embryo failure in scid recipient mice which do not respond specifically to Mus caroli xenoantigens. In 1989, Crepeau et al. published a report that more precisely defined the mechanism by which Mus caroli embryos die in Mus musculus uteri. The authors chose to 0 45 histologically examine the embryos at day 8.5 of gestation. Both Mus caroli and Mus musculus embryos implanted in the same Mus musculus uterus appear normal at day 7.5 and begin to show signs of embryonic death at day 9.5. Thus, the choice of day 8.5 observation offers a window at which changes may be clearly visible between the two cases. Mus caroli embryos in the Mus musculus uteri fail to show uniform stages of development and are deficient in trophoblast development. In addition, Mus caroli implants show structural abnormalities such as deficiency of trophoblast cell granularity, trophoblast giant cell degeneration and PMN infiltration of the giant cell area. The information presented in the studies of the Mus caroli - Mus musculus interspecific model of fetal loss were crucial in the development of the literature in two major areas. The first is the demonstration of the nature of the cellular immune responses as_sociated with fetal loss. The determination that non-specific cellular reactions largely mediate the rejection of xenogeneic embryos (Clark et al., 1986; Crepeau and croy, 1988) corroborates the later studies in which NK and LAK cells were shown to be important in allogeneic fetal loss (deFougerolles and Baines, 1987; Gendron and Baines, 46

1988: Lala et al., 1990). The second area of knowledge to which the interspecies model contributes key information is the importance of maternal-trophoblast relation and pregnancy success. The initial study by Rossant et al. (1982) raises interesting questions about whether embryo-trophoblast chimaerization may protect fetal loss in allogeneic models such as the CBA/J X DBA/2 system (see below).

b) The CBA/J X DBA/2 Model of Spontaneous Fetal Loss Murine models of spontaneous fetal loss offer the convenience of large numbers of small animals with a short gestation and a fully characterized congenic MHC. In 1982, Clark and collegues described the first murine model of spontaneous abortion (Clark et al., 1982). The model consists of pregnancies obtained by mating virgin CBA/J female mice with DBA/2 males. The resulting

pregnancies exhibit a 25-35% spontaneous fetal abortion frequency observable on day 10 of gestation as compared to rates of less than 10% in matings with either CBA/J or BALB/c males. Interestingly, the BALB/c possesses an MHC (H-2d) identical to the DBA/2 male. The reverse mating of CBA/J males with DBA/2 females is not characterized by increased fetal loss. The CBA/J X DBA/2 resorption 0 47

frequency decreases upon previous pregnancy with BALB/c, but increases in direct proportion to maternal age. A subsequent study advanced the understanding of the CBA/J resorption phenomenon in immunological terms by demonstrating that pre-immunization of the CBA/J female with BALB/c splenocytes could decrease the fetal loss in CBA/J X DBA/2 matings (Kiger et al., 1985). This decreased fetal loss is associated with the production in the CBA/J female of anti-H-2d specific IgGl antibody and increased splenic suppressor cell activity against CTL generation (Chaouat et al., 1984: 1985). However, pre­ immunization with DBA/2 or CBA/J splenocytes has no significant effect on the resorption frequency (Kiger et al., 1985). The pre-immunization protection against abortion is sex-specific as only male BALB/c cells are effective. Genetic studies utilizing H-2b or H-2k congenic BALB males to determine the effect of the H2 haplotype of the splenocytes used for immunizations of CBA/J females revealed that only BALB/c (H-2d) male splenocytes are effective. These findings demonstrated that the abortion protection effect is conferred by the presentation of an H-2d haplotype against a BALB/c male background. These studies emphasized two important points about the CBA/J X DBA/2 spontaneous resorption 0 48

model. The first demonstrates the importance of the maternal component in pregnancy success and emphasizes the interesting possibility that some females may be inherently compromised in the ability to successfully gestate pregnancies resulting from certain genetic combinations. Secondly, there exists a male-specific MHC-related antigenic stimulus that promotes fetal viability, the lack of which may be associated with increased fetal loss. In light of the studies by surani and colleagues demonstrating that development of extra­ embryonic tissue is dependent on the paternal genome, it is conceivable that such a male-specific resorption protection may be an example of MHC associated genomic imprinting (Surani et al., 1984). Recent studies in this laboratory (see results section) have partially addressed the former point, by demonstrating that the primigravid versus the multigravid CBA/J (X DBA/2) female is deficient in a leukotriene B4-related fetoplacental immunosuppressive factor. Following the initial findings by Clark, Kiger and collegues of the increased fetal resorption in CBA/J X DBA/2 matings, a series of reports were published which explored the reproductive physiology in both normal matings and the CBA/J X DBA/2 mating. These studies 0 49 c extended the concept of variable levels of local immunosuppression initially demonstrated in xenopregnancy to cases of allogeneic pregnancy. A population of small, non-T lymphocytes present in the uterine lymph nodes and decidua produces a soluble factor which suppresses maternal-anti-paternal CTL generation (Slapsys and Clark, 1982). Significantly lower numbers of suppressor lymphocytes are present in decidua of CBA/J X DBA/2 spontaneous resorption-prone pregnancies. In addition, the deficiency of suppressor cells was shown to be associated with the infiltration of maternal leukocytes into the fetoplacental units of resorbing CBA/J X DBA/2 pregnancies (Clark et al., 1986). Further studies contributed to identifying the nature of the maternal cellular immune response and immunosuppressive activity associated with failing CBA/J X DBA/2 concepti. In 1987, Gendron and Baines found that infiltration of asia1o-GM1 positive natural killer (NK)-like cells into the decidual-ectoplacental cone junction is associated with spontaneous resorption in the CBA/J X OBA/2 mating. This finding was supported by another report from the same laboratory showing that the modulation of NK activity by anti-asialo GMl antibody depletion and POLY I:C interferon-boosting treatments can significantly 0 50

alter the spontaneous resorption frequency (deFougerolles and Baines, 1987). In addition, NK-modulating treatments alter the infiltration of NK cells into fetoplacental units of CBA/J X DBA/2 pregnancies (Gendron and Baines, Manuscript in preparation; see results). The involvement of NK cells in fetal resorption became more generalised when Chaouat reported the resorption-inducing effects of ·various polyinosinic cytidilic acid analogues on a number of different murine matings (Chaouat et al., 1989). The argument for the involvement of NK cells and NK regulation in the CBA/J X DBA/2 resorption process was further strengthened by the demonstration that CBA/J X DBA/2 concepti were deficient in an NK inhibitory lipophilic factor (Gendron et al., 1990). our most recent studies suggest that this NK-suppression deficiency is related to a decreased production of LTB4 by the primigravid CBA/J (X DBA/2) uterus (Gendron et

al., 1991; see results).

G. RATIONAL FOR THIS THESIS

At the time this work began in 1986, one largely unanswered question in reproductive immunology concerned the nature of the cellular immune response involved in c 51

allogeneic pregnancy failure. our initial work was thus aimed at identifying the phenotype of this cellular response in spontaneously resorbing mouse fetoplacental units (Gendron and Baines, 1988: 1989). Based on our findings that NK cells were associated with fetal resorption, we set out to determine the nature of the NK regulatory factors in pregnant mice. A lipophilic NK suppressing activity was partially purified from early fetoplacental units (Gendron et al., 1990). In addition, we determined how this activity changed in resorption- prone versus normal pregnancies. These studies led to the identification of an LTB4 deficiency in spontaneously resorbing fetoplacenatal units. In addition, LTB4 was found to be important in the NK suppressive activity of fetoplacental unit extracts (Gendron et al., 1991). In order to characterize the role of non-specific immune reactions in fetal resorption in more detail, we utilized

the LPS induced fetal resorption model described by Zahl and Bjerknes in 1943. From these studies, we found that TNF-a may be a major mediator of the effector stages of resorption (Gendron et al., 1990). During the course of our analysis of TNF-a production in the fetoplacental unit during resorption, we discovered that TNF-a is expressed in the normally developing embryonic nervous 0 52

..,-, ·~ system (Gendron et al., 1991).

0 53 c CHAPTER II PRBFACB. THE ROLE OF HK CELLS IN THE CBA/J X

DBA/2 MURINE SPOHTAHBOOS aBSORPTION MODEL

The following section consists of the work done in the initial phases of our studies on the CBA/J X DBA/2 spontaneous resorption model. The rational for these studies was to clarify the nature of the cellular immune response occurring in the early stages of spontaneous resorption. Our findings demonstrated that cells of the NK phenotype predominate in failing CBA/J X DBA/2 fetoplacental units as early as day 6 of gestation. NK­ modulatory treatments which affect the resorption frequency also show a correlative effect on the numbers of NK cells infiltrating the CBA/J X DBA/2 conepti. These studies provided a partial but essential identification of the cellular immune response in CBA/J X DBA/2 fetal resorption. The involvement of NK cells in the CBA/J X

DBA/2 model of fetal loss led us to the analysis of the factors regulating fetoplacental NK cells in normal versus resorption prone-pregnancies. This section includes three manuscripts. The first, Infiltrating decidual natural killer cells are associated with spontaneous abortion in mice, and the second, Morphometric analysis of the histology of 0 54

spontaneous fetal resorption in a murine pregnancy are published papers (Cellular Immunology, 113, 261-268, 1988 and Placenta, 10, 309-318, 1989, respectively). The third, Alteration of the infiltration of natural killer cells in CBA/J X DBA/2 concepti by POLY I:C and anti- asialo GM1 treatment consists of unpublished observations. All the research reported in these three manuscripts was conducted by R.L. Gendron under the supervision of Dr. M.G. Baines and Dr. R. Farookhi.

0 55

CELLULAR IMMUNOLOGY 113, 261-267 ( 1988)

Infiltrating Decidual Natural Killer Cells Are Associated with Spontaneous Abortion in Mice

R. L. GENDRON AND M. G. BAINES

Department ofMicrobiology and Immunology, McGill University, Lyman Duff Medical Sciences Building, 3775 University Street, Montreal, Quebec, CanadaHJA-2B4

Received June 25. 1987; accepted December 7, 1987

Immunohistochemistry was used to study a murine model which spontaneously aborts at a frequency of25 to 30%. Our results show that natural killer (NK) cells are not only the predomi­ nant infiltrating cells in aborting feto-placental units, but that they also appear in a similar pro­ portion of feto-placental units before abortion is detectable. The frequency of feto-placental units with significantly elevated NK infiltrates corresponds to the subsequent abortion fre­ quency, indicating a causal relationship. Immunization ofthe mother with BALB/C splenocytes prevents these NK infiltrates and decreases the abortion frequency to normal levels. These re­ sults suggest for the first time that maternal NK cells may have an instrumental role in the etiology of spontaneous abortion. e 1938 Academic Press. Inc.

INTRODUCTION The rejection of xenogeneic embryo implants in mouse uteri involves infiltration of the grafted tissue by maternal cytotoxic T lymphocytes ( l ). Little is known about the possible role of the immune system in the spontaneous rejection ofthe allogeneic fetus during pregnancy. Several strains of mice have been shown to spontaneously · abort or resorb a significant proportion of their litter during first pregnancy (2). That this high resorption rate decreases both with parity and with maternal immunization (3) may indicate a weakness of the immunoregulatory mechanisms which ensure fetal survival. Our approach was to study a murine model of spontaneous abortion chronologically at the.in situ level with the objective of revealing immunopathologi­ cal changes preceding the actual resorption event. We find a correlation between the frequency of feto-placental units containing significantly elevated levels of natural killer (NK) cells and the resorption frequency up to 4 days before resorption is de­ tectable.

MATERIALS AND METHODS Animals. Virgin CBA female mice (Jackson, Bar Harbor, ME) were mated with DBA/2J males (Charles River, St. Constant, Quebec, Canada) to provide first preg­ nancies. Females were placed with males overnight and examined the next morning for the presence of a copulatory plug. The day ofdetection of the plug was designated as Day 0 of gestation. Following sacrifice at various stages ofgestation, feto-placental units were removed, snap-frozen, and stored at -so·c until cryosectioning. Male

. i 261 ! I I 0008-8749/88 $3.00 ! Copyricht C 1988 by Academic Press, Inc. 0 All righ!S of reproducti011 in any form reserved 56

262 GENDRON AND BAINES mice of the BALB/C strain (Jackson) were used to immunize CBA virgins and as fathers in CBA X BALB/C matings. Immunizations. Virgin CBA females were injected intraperitoneally with 107 BALB/C spleen cells in 0.2 ml RPMI 1640 as previously described (3). After 8 days, immunized CBA females were mated with DBA/2J mates and monitored for pregnancy. Immunohistochemistry. Serial frozen sections (5 ~tm thick) of each feto-placental unit were air-dried, fixed in acetone for tO min, and stained for the localization of IgM (B lymphocytes), Thy-1 (T lymphocytes), asialo-GMl (NK cells), and Ly-2.1 (cytotoxic/suppressor lymphocytes). Rabbit polyclonal antisera to IgM (Bio-Can, Mississauga, Ontario, Canada), Thy-1 (Cedarlane, Homby, Ontario, Canada) and asialo-GM l (Wako, Dallas, TX) were used in the indirect immunoperoxidase proce­ dure as described previously (4), while mouse monoclonal anti-Ly-2.1 (Cedarlane) was used in the avidin-biotin immunoperoxidase procedure as described by the Vec­ tor methodology (V ecta-Stain Kit, Burlingame, CA). Blocking ofendogenous ')'-glob­ ulin in the localization of Ly-2.1 was accomplished by pretreatment of the sections with a 1:200 dilution of sheep anti-mouse IgG Fah fragments (Cedarlane). Positive controls consisted of the same procedures on mouse spleen sections while negative controls consisted of the procedure minus primary antibody both on mouse spleen and on sample sections of units which showed positive staining. Asialo-GM !-positive cells are provisionally referred to as NK cells on the basis of surface expression of asialo-GM 1 but not ofThy-1 as previously described (5). Lymphocyte counts. Positive cells were scored by counting 5 to 10 high power fields 2 (0.185 mm ) per feto-placental unit section at a magnification of 500. Serial sections from the same feto-placental unit were used in the four different lymphocyte staining procedures.

RESULTS Resorption Frequency Resorption (at a rate of 25-30% per litter) was detectable only after Day tO of gestation and was characterized grossly by smaller feto-placental units, fetal necrosis, and placental hemorrhage. Figure 1 shows the frequency distribution of feto-placen­ ta] units with respect to NK infiltration and gestational stage. As shown, the frequency of feto-placental units containing less than 2 NK cells per high power field remains at a relatively constant value of 70% as gestation proceeds. A second population, occurring with an average frequency of30% at each stage ofgestation, is characterized by NK infiltrates greater than 2 cells per high power field. Within this second popula­ tion, the most frequently occurring infiltrate amount at the 6- to 7-day stage consisted of 4-7.9 NK cells per high power field (16%). At the 8- to 9-day stage, the same percentage of feto-placental units shows NK infiltrates ofgreater than 2 cells per field, but the predominant infiltrate size has increased from 4-7.9 NK cells per field to 8- 15.9 NK cells per field. By Day 10, at which point resorptions are clearly detectable by microscopic criteria, a similar population of units shows the majority of infiltrates containing 8-15.9 NK cells per field and are all confirmed as resorptions. The higher modal frequencies (those units containing more than 2 NK cells per high power field) appearing at both the 6- to 7- and 8- to 9-day gestational stages were statistically indistinguishable from the higher modal frequencies appearing at the 10- ·57

SPONTANEOUS ABORTION AND DECIDUAL NK CELLS 263

100

90

80

70 ~z :::> 60

~ 50 a..~ I 0 40 1-.... ~ 30

20

10

0 6-7 e-9 1o-13 DAYS OF GESTATION

FIG. I. Frequency offeto-placental units with respect to number of asialo-GM 1 (GM I)-positive cells per 2 high power field and gestational stage in days. One high power field = 0.185 mm • 0 0-1.9 GM 1+ cells/ field; m2-3.9 GMI+ cells/field; 114-7.9 GM!+ cells/field; 11118-15.9 GM!+ cells/field; • 16-32 GM!+ cells/field. Solid circles above bars represent confirmed resorbing populations of feto-placental units. This analysis was based on the examination of 396 feto-placental units from 39 gravid female CBA mice. to 13-day stage as assessed by x2 analysis of population distribution (P > 0.05). Thus, on the basis of statistical comparison, the populations occurring at a frequency of 30% at both stages 6-7 and 8-9 are the same population that all resorbing feto-placen­ tal units fall into at the 10- to 13-d.ay stage. The shape of the bimodal distribution in Fig. I changes such that the distance between the modes increases aS gestation pro­ ceeds. This behavior is indicative of an expansion of the NK infiltration which ap­ pears to begin at Day 6 at an average 4-7.9 cells per field and which reaches a plateau of an average 8-15.9 cells per field between Day 8 and 13. No other lymphocytic subpopulation shows these characteristic kinetics during this gestational period (data not shown).

Infiltrate Size and Phenotype Figure 2 shows the comparative difference in amount of infiltration between NK cells, thymus--derived lympbocytes, and B lympbocytes in normal, preresorbing, and resorbing units. Natural killer cell infiltrates in preresorbing and resorbing units were as high as fivefold greater than infiltrates ofT lymphocytes in serial sections of the same units. Furthermore, the NK cell infiltrates appeared up to 4 days earlier than T -cell infiltrates (Figs. 2a and 2b). The T -cell infiltrates in preresorbing and resorbing units contained few cytotoxic/suppressor cells (Figs. 2b and 2d) and the cells in gen­ eral had large nuclear/cytoplasmic ratios, indicating a resting state. In contrast, NK cells infiltrating the same units appeared large and activated. There was no evidence of a significant B-lymphocyte infiltration in preresorbing and resorbing units (Fig. 2c), and B cells· present did not appear to be activated p1asma cells. Since average litter size remained constant at each gestational stage studied, resorptions did not occur at irregular intervals, but occurred at a specific stage and were preceded by stage specific NK infiltrates. 58 0 264 GENDRON AND BAINES

11.0 • 10 b 11.1 ...... Q. • Q. J: J: 12.0 in • -' in 10.6 7 -' w-' w-' /r---1--i u u 8.0 / 'i • 7.1 > I :::!: / J: Cl "( 1- ... 1.0 ... .. z 4.6 z a ....,t.... < < w w ,"" --2 :::!: 3.0 :::!: 2 1.1 ~ 0 0 &-7 IQ-11 12-13 &-7 .... 1G-II 12-1a DAYS·-· GESTATION DAYS GESTATION

10 10 c d • ... • Q. ... J: Q. • in • iE 7 7 If) -' -' -'w w-' u • u • I I :::!: .. g > .. -' ...... z a a < z w < 2 w 2 :::!: :::!: -~-- 0 0 1Q-II &-7 - to-ll 12-13 &-7 ...... DAYS GESTATION DAYS GESTATION FIG. 2. Mean number of lymphocytes as a function of gestational stage in normal feto-placental units {e), preresorptive feto-placental units <•), and resorbing feto-placental units {0) ofCBA X DBA/21 mat­ 2 ings. One high power field= 0.185 mm • Between 8 and 13 females were studied at each gestational stage shown.

Effect ofImmunization with BALB/C on NK Infiltrates Immunization of both CBA females with BALB/C spleen cells and mating the CBA females with BALB/C fathers prevented the NK infiltration associated with abortion between Day 7 and 9. Numbers ofGM1-positive cells in all sections of the 81 embryonic units from four BALB/C-immunized CBA X DBA/2J pregnancies and four CBA X BALB/C pregnancies never exceeded the 0 to 1.9 cells per high power field range shown in Fig. 1 for unimmunized CBA X DBA/2J matings (6). The pre­ sumptive abortion rate for these matings would therefore be less than I 0% as shown by other investigators (2, 3).

~ ' .'. . . . Spatial Characteristics ofNK Infiltrates The anatomical locations of the infiltrating NK. cells differed between the prere­ sorptive and resorbing groups. NK. infiltrates at both the 6- to 7- and 8- to 9-day stages were consistently localized in the area of the decidua which was directly underneath the embryo-ectoplacental cone region of the feto-placental unit (antimesometrial de­ cidua). Infiltrates were generally more concentrated and slightly greater in quantity Q in the 8- to 9-day group than the 6- to 7-day group (see Fig. I) and the shape of the

' . c 59 SPONTANEOUS ABORTION AND DECIDUAL NK CELLS 265

infiltrate by the latter stage had assumed a conical form which appeared to encapsu­ late the embryonic region of the unit. Early stage infiltrates (Day 6-7) were distributed throughout the subembryonic decidual area and did not seem to concentrate coni­ cally around the embryo. Infiltrates of NK cells found in the confirmed resorbing feto-placental units (Day 10-13) were evenly dispersed throughout the notably de­ ranged tissue that was once the embryo-ectoplacental cone area, although the infil­ trates never appeared in the endometrial or uterine muscular aspects. NK infiltrates were seldom found in the trophoblast, embryonic, or ectoplacental cone area of the preresorptive feto-placental units even in cases where infiltrate density was high at other sites. In cases where the general anatomy of rerorptive units was discernable, infiltrates did appear in the trophoblast/ectoplacental cone area. Figure 3 shows the tissue structure of normal and preresorptive feto-placental units.

DISCUSSION

This study shows evidence for the first time that natural killer cells may have an important role in the etiology of spontaneous pregnancy failure. Previous studies have shown high levels of NK activity in early murine decidua which were reported to peak at Day 6.5 and then decline to negligible levels by Day 12 due to regulation by other decidual cells (7, 8). Further experiments are needed to explore the effective activity of the cells described in this model. Although we did not detect high NK counts in 70% of the units studied at this gestational stage it is possible that the basal or normal counts we report correspond to the activities found in the above-mentioned studies. High NK counts found in the 30% population that we consider as resorptive or preresorptive may correspond to lytic activity levels far above what may have been considered as normal in previous studies (7, 8). Although studies have shown that NK cells are not detrimental to early embryos (8), the target of NK attack does not necessarily have to be embryonic or trophoblastic tissue to adversely affect the normal development of nutrient delivery to the embryonic unit. Recent work in our laboratory has shown that modulation of NK activity can significantly alter the resorption rate in mice (9). That the NK infil­ trates can be prevented by immunization with BALB/C strain cells strengthens our argument implicating these cells in this abortion process. IfNK cells are natural resi­ dents ofthe decidual tissue (7) mechanisms must be present which regulate these cells such that any possible lytic activity against local tissue is inhibited. Once this inhibi­ tion has been impaired these NK cells may be capable of attacking other decidual cell types which are thought to include bone marrow-derived (I 0) and unusual T-lineage cells (11 ). A population of mononuclear uterine cells has recently been partly charac­ terized which suppresses the proliferation of maternal lymphocytes in response to fetal antigens in a nonspecific manner (12-14). Considering the apparent "normal" presence of null Iymphocytes in decidual tissue (7) and the similarity of pregnancy to other special immunological cases such as tumor growth and organ grafting, one is led to speculate about the roles of nulllymphocytes in this process. When fully under­ stood, the immunoregulatory mechanisms of pregnancy may be simulated as part of a local treatment of grafted tissues with the hope of promoting grafts to "take" in cases where they would otherwise be rejected.

j. ;.. 60 •

FIG. 3. (a) Normal (Day 8) and (b) preresorptive (Day 8) feto-placental units stained for the localization of NK cells. Large arrows show the location of the embryo and smaller arrows the location of the NK infiltration at a magnification of 60. Lymphocyte counts were taken at a magnification of 500. 61

SPONTANEOUS ABORTION AND DECIDUAL NK CELLS 267

ACKNOWLEOOMENTS We thank Mr. Kevin GJassow for technical assistance and Dr. R. Farookhi and Mr. A. R. de Fougerolles for critical commentS. This work was supported by the Medical Resean:b Council of Canada and the Centre for the Study ofR.eproduction, McGill University.

REFERENCES 1. Croy, B. A., Rossant, J., and Oarlc, D., J. Reprod. Immunol. 4. 277, 1982. 2. Kiger, N., Cbaouat, G., Kolb, J. P., Wegmann, T. G., and Guenet, J. L., J. Immunol. 134, 2966, 1985. 3. Cbaouat, G., Kolb,J. P., K.iger, N., Stanislawski, M., and Wegmann, T. G., J. /mmunol. 134, 1594, 1985. 4. Sternberger, L. A., and Sternberger, N. H., J. Histochem. Cytochem. 34, 599, 1986. 5. Kasai, M., lwamori, M., Nagai, Y., Okamura, K., and Tada, T., Eur. J. /mmunol. 10, 175, 1980. 6. Baines, M. G., manuscript in preparation. 7. Gambel, P., Croy, B. A., Moore, W. D., Hunziker, R. D., Wegmann, T. G., and Rossant, J., Cell. /mmunol. 93,303, 1985. 8. Croy, B. A., Gambel, P., Rossant, J., and Wegman, T. G., Cell. lmmunol. 93,315, 1985. 9. deFougerolles,A. R.,and Baines, M. G.,J. Reprod. lmmuno/.11,141, 1987. 10. Keams,M.,andLala,P. K.,J.Exp.Med.l55, 1537,1982. 11. Bulmer,J. N., and Sunderland, C. A.,Jmmuno/ogy5:Z, 349, 1982. 12. Slapsys, R. M., and Oarlc, D. A., J. Reprod. lmmunol. 4, 355, 1982. 13. Nakayama, E., Asano, S., Kodo, H., and Miwa, S., J. Reprod.Jmmunol. 8, 25, 1985. 14. Tawfi.k, 0. W., Hunt, J. S., and Wood, G., J. Reprod. Immunol. 9, 213, 1986. 62 c

Morphometric Analysis of the Histology of Spontaneous Fetal Resorption in a Murine Pregnancy

R. L. GENDRON" & M. G. BAINESb

Departments of Physiology• and Microbiology 11.nd Immunology•, Centre for the Study of Reproduction, McGill University, 3775 University St., Montretl.l, Quebec H3A-:zB4, Canadtl. Paper accepted J.:z.I!j&j

INTRODUCTION

Throughout the course of normal allogeneic pregnancy, maternal tissue exists in close apposi­ tion to fetal tissues. Trophoblast cells and fetal tissue have been shown to express paternal class I major histocompatibility complex (MHC) antigens that paradoxically do not induce maternal anti-fetal rejection (Bulmer and Johnson, 1985; Lata et al, 1983; Beer and Billingham, 1976). Furthermore, the invasion of trophoblast into human decidual tissue has been shown to in­ crease the expression of class II major histocompatibility antigens on leukocytes in the invaded areas (Bulmer and Sunderland, 1984). Uterine stromal cells in the hormone primed rat uterus expressing high levels of la antigen (class II MHC) (Head and Gaede, 1986), may be capable of (paternal) antigen presentation. It is believed that active immunosuppression within the locale of the feto-placental unit prevents the activation of maternal mononuclear leukocytes and sub­ sequent fetal rejection (Oark, McDermott and Szewczuk, 198o). This immunosuppressive activity has been attributed to several morphologicaJJy distinct cell types that have been shown to suppress maternal anti-fetal immune responses including non-T lymphocyte suppressor cells in both murine and porcine pregnancies (Tawiik, Hunt and Wood, 1986; Oark, Chaput and Tutton, 1986; Croy, Wood and King, 1987). Decidual suppressive activity in the mouse has been shown to decrease cytolytic function by blocking the action of IL-z (Oark et al, 1988). It has been hypothesized that deficient immunosuppressive activity in the feto-placental unit leads to embryo death via leucocytic infiltration through the trophoblast (Oark, Chaput and Tutton, 1986). An abnormally high rate of spontaneous resorption (25-30 per cent per litter) can be observed in CBA/J female mice mated with DBA/J male mice (Qark, McDermott and Szewczuk, 1980). These pregnancies have been shown to be deficient in non-T lymphocyte suppressive activity. Such suppressive activity can be augmented by vaccination before preg­ nancy with spleen cells of the Balb/c allogeneic strain (Oark et al, 1987). Spontaneously resorbing CBA/J x DBA/J fetoplacental units become infiltrated as early as z days post im­ plantation by naturaJ killer (NK) cells but not by cytotoxic T lymphocytes (CI'L) (Gendron and Baines, r988). Pregnancy resorption in the CBA x DBA/J model therefore does not re­ semble the events mediating the rejection of xenogeneic fetal implants or allogeneic grafts in which specific CTL activity initiates the rejection of the implanted fetal or grafted tissue, re-

© 1989 Bailliere Tindall Ltd 63

310 Placmta (Hj89), Vol. ro spectively (Croy, Rossant and Oark, 1982, 1984; Soulillou, 1987). Although normal murine and porcine embryos have been shown to recruit NK cells to the uterus, decidual NK activity cannot be demonstrated in the resorption prone CBA/J x DBA/J pregnancy (Croy et al, 1988). However, systemic NK activity, either reduced by anti-asialo GMr antibody (a-AGMt) or augmented by poly inosinicfcytidylic acid (Poly I:C), correlates directly with the resorption rate in CBA/J x DBA/J pregnancies (de Fougerolles and Baines, 1987). These findings sug­ gest that CBA/J x DBA/J pregnancies are predisposed to a reversible immunological failure to regulate a seemingly non-specific maternal anti-fetal response. Recent studies of xenogeneic embryo resorption in recipient scid (severe combined immunodeficiency) C.B.-17 mice reveal that resorption proceeds despite the lack of specific cellular immunity (Crepeau and Croy, 1988). Since SCID C.B.-17 mice have functional NK cells (Dorshkind et al, 1985), resorption of xenoembryos in SCID recipients may be initiated by NK cells in this model (Crepeau and Croy, 1988). Although we have shown that NK infilt­ ration precedes resorption (Gendron and Baines, 1988), there are no studies which attempt to monitor decidual and ectoplacental cone structure during the course of the spontaneous resorption process. Such information would reveal whether resorption involves feto-placental damage only in areas ofNK infiltration or in areas remote to the initial NK infiltrate. We have undertaken the morphometric analysis of CBA/J x DBA/J pregnancies to answer the follow­ ing questions: does infiltration of the decidua of fetoplacental units by NK lymphocytes represent the only histological change which precedes resorption? Are there consequent or concurrent early changes in decidual and ectoplacental cone structure and cellularity? The answers we present here further our understanding of the mechanisms contributing to the initiation and mediation of resorption by demonstrating significant morphological disturbances in specifically defined areas of fetoplacental units at early and late stages of spontaneous resorp­ tion.

MATERIALS AND METHODS

Animals Primigravid pregnancies were obtained by caging virgin CBA female mice (Jackson, Bar Harbor, ME) with DBA/J male mice (Charles River, St. Constant, Qpebec). Day of vaginal plug detection was defined as day o of gestation. Pregnant females (n = 17} were sacrificed on each day for days 7 to 10 of gestation. Feto-placental units were removed, snap frozen by immersion in isopentane (-8o0 C) and stored. Feto-placental resorption, detectable after day 10 of gestation only, was characterized macroscopically by smalJer embryonic unit size, fetal necrosis and placental haemorrhage.

Immunohistochemistry NK lymphocytes were detected on frozen sections of the feto-placental units by immunohisto­ chemistry using a-AGM1 (Gendron and Baines, 1988). Briefly, sections were air dried for 1 h, fixed in acetone for 10 min and subjected to indirect immunoperoxidase staining of the asialo­ GMI antigen using rabbit polyclonal anti asialo-GM1 antibody (Wako, Dallas, TX). All sec­ tions were then counterstained with hematoxylin for exactly 20 sec to provide uniform nuclear staining intensity.

Morphometric analysis Feto-placental units at each gestational stage were separated into two groups according to a 64 c

Geruiron, Bt~ints: lmmur~opt~thology ofSpontaneous Abortior~ Jll bimodal distribution described previously (Gendron and Baines, 1988). The first group, desig­ nated as normal, had no evidence of a significant increase in NK lymphocyte density (between o and z NK cells per 0.185 mm 2). The second group, designated as resorbing, had clear evi­ dence in each section of significandy increased NK lymphocyte density {between 4 and 32 NK cells per o.r8s mm 2). Both groups of feto-placental units were then analyzed for the total number of nuclei (nuclear density) and total nuclear area per high power field ( x soo). The number of nuclei was assessed by manually counting nuclei in an area of 5000 p.m 2 ad de­ lineated by the boundaries of an eyepiece grid. Five to ten randomly chosen grid areas were counted per feto-placental unit section. Nuclear area was assessed by utilizing a microscope­ mounted color video camera interfaced with an IBM-AT mounted image analysis program (I MAN) (developed by Or H. Riml of the McGill Medical Physics Unit, Montreal, Q!lebec, Canada). Briefly, sections were mounted under the video camera and the image was projected on the monitor screen. The color of representative nuclei was sampled (oudined) with the aid of a 'mouse' and the image was then automatically scanned to yield a dichromatic pictorial rep­ resentation of the nuclear area on each section. This scanned image was then automatically analyzed to give percent of total image area occupied by nuclear color in an area of 30 ooo p.m 2 per high power field ( x 400). Measurements of both nuclear density and nuclear area were taken at days 7-10 from the decidual-ectoplacental cone junction where NK cells have been shown to appear in high numbers prior to resorption (Gendron and Baines, 1988). Measurements of nuclear density were also taken from the decidual area proper and the ectoplacental cone proper at days 8 and 9 (after the anatomical distinction between decidual and ectoplacental cone tissues becomes pos­ sible and before the appearance of definitive placental tissue) (Theiler, 1983). Measurements taken from the decidual or ectoplacental cone areas separately were used to establish any pos­ sible cellular changes in each tissue alone. Data were analyzed by student's t-test and by two way analysis of variance with a 95 per cent least significant interval test (Sokal and Rohlf, 1970).

RESULTS

Anatomical aspects of resorption The most visually apparent indication of impending resorption in the early stages (days 7-8) is the appearance of infiltration (increased overall cell density) and loss of cellular contact in the decidual-ectoplacental cone area in the resorbing units. Late stage resorption (days 10-u) is characterized by the loss of embryonic tissue and the pronounced presence of small mono­ nuclear and PMN leukocytes throughout the interior of the embryonic unit. Failure of normal placental development between days 8 and 9 in the resorbing population of units is evidenced by the lack of typical placental syncytia. Figure 1 shows the difference in cellular contact and tissue structure between normal, presumptively resorbing and resorbing fetoplacental units as resorption progresses from initial cellular infiltration [Figure 1(b)] to tissue damage and necro­ sis [Figure t{d)].

Nuclear density in normal versus resorbing units Figure z shows the mean nuclear density found in normal and resorbing feto-placental units at each day of gestation between days 7 and 10. The mean nuclear density per high power field in the normal feto-placental units declines as gestation proceeds with significant decreases between day 7 to days 8 and 9· In addition, the nuclear density at day to is significandy lower ,

65 •

Figure 1 . (a) Decidual-ectoplacental cone junction of a normal embryonic unit at day 8 of gestation. This unit contains low numbers ofNK cells and shows normal intact cell-cell contact (a-AGM1 and haematoxylin; x 250). (b) Decidual­ ectoplacental cone junction of a resorbing embryonic unit at day 8 of gestation. This unit shows increased numbers of K cells in addition to an AGM1 negative cellular infiltrate. Note abnormal tissue spaces (•) (a-AGM1 and haema­ toxylin; x 250) . • 66 ••

Figure 1. (c) Normal placental structure at day 10 of gestation showing normal blood lacunae (L) and syncytial tropho­ blast (S). This unit shows low numbers of NK cells (a-AGM1 and haematoxylin; x 100). (d) Interior of a resorbing embryonic unit at day 10 of gestation. This unit shows increased numbers of NK cells and large numbers of mono­ nuclear and polmorphonuclear leukocytes. Uterine lumen (UL); resorbing embryonic tissue (E) (a-AGM1 and haema­ toxylin; x 100).

/~ - 67

• Plawrta (u}89), Vol. ro 31<4

451

>...... ,.J Vl J d Cl) 0 s.. «S _..Cl) 0 15 ;j z

0 7 8 9 10 Gestational Age (days)

Figurt 2 . Nuclear density per high power field versus days of gestation in normal (solid) and resorbing (open) embry­ onic units. !'ormal units contained between o and 2 NK cells per 0.185 mm' while resorptive units showed signific­ antly increased numbers ofNK cells of between 4 and 32 NK cells per 0.185 mm'. This data is based on the analysis of 59 embryonic units. One high power field= 5000 Jlm' (•P

than at all other time points. Conversely, the mean nuclear density per high power field in the resorbing feto-placental units plateaus at levels which are significantly greater than normal at each stage of gestation. Between days 9 and 10, the mean nuclear density in the resorbing population is more than double that in the normal population. Significant differences are observed in the mean nuclear densities between days 7-10 within the resorbing population of units. The difference is greatest at day 10 at which point the nuclear density is lowest compared to all other days in the resorbing population.

Nuclear density in decidual tissue versus ectoplacental cone tissue in normal and resorbing units Significant increases in nuclear density are observed in resorbing embryonic units in both the decidual area and ectoplacental cone. Increases are observed at days 8 and 9 although at day 9 the changes are amplified. The percent increase in nuclear density (normal versus resorbing) is more substantial in the decidual tissue than in the ectoplacental cone at the earlier time (day 8); see Table 1. However, the percent increase in nuclear density (normal versus resorbing) in the ectoplacental cone increases at day 9 to levels which are more comparable to those observed for decidual tissue at both days 8 and 9· Table 1 shows the mean nuclear densities in decidual and ectoplacental cone areas of the embryonic units at days 8 and 9, per cent change and signific­ ance.

Nuclear area in normal versus resorbing units Figure 3 shows the mean nuclear area in normal and resorbing feto-placental units between • days 7 and 10 of gestation. The mean nuclear area decreases significantly between days 8 to 10 • 68 Gardron, Baines: Immunopathology ofSpontaneous Abortion JIS

Table 1. Nuclear density per high power field in the decidual (D) and ectoplacental cone (EC) regions assessed at days 8 and 9 in normal (N) and resorbing (R) embryonic units. Distinction between normal and resorbing embryonic units is the same as in Figure 2. Mean nuclear density (X), standard error (s.e.), sample size (n) and percent change(~ per cent) as calculated by the visual assessment of the same tissue sections used for data presented in Figures 1-4. One 2 high power field = sooo JJm •

Tissue N/R Day X SE n ~% Significance ·

D N 8 21.9 1.21 25 D R 8 39·6 0.98 20 81 p < 0.05 EC N 8 13.7 1.04 25 EC R 8 19.6 0.74 20 43 p < 0.05 D N 9 22.7 o.68 40 D R 9 49·0 2.34 5 116 p < 0.05 EC N 9 15.8 0.97 25 EC R 9 30.0 3·27 5 90 p < 0.05

in normal embryonic units. However, in the resorbing embryonic units, the mean nuclear area increases significantly between days 7 and 8. Following a plateau between days 8 and 9, the nuclear area then decreases significantly between days 9 and 10 in the resorbing group. There is no significant difference in mean nuclear area between normal and resorbing embryonic units at day 7· At days 8-10, mean nuclear area is significantly greater in the resorbing group (Figure 3). There was no evidence of nuclear hypertrophy in normal versus resorbing units. The average area per nucleus in the normal group at all stages of gestation was calculated to be 53.31 ( ± 6.22 s.e.) IJ.m 2 while the average area per nucleus in the resorbing group was 52.27 2 ( ± 4·47 s.e.) 1J.m , indicating no significant difference between the two groups (P > 0.05).

45 * *

* as Q) 30 <'"' as '"'Q) ' 0 -~ z 15

0 7 8 9 10 Gestational Age (days) Figure J . Nuclear area per high power field versus days of gestation in normal (solid) and resorbing (open) embryonic units. Distinction between normal and resorbing embryonic units is the same as in Figure 2. This data is based on the 2 • analysis of 51 embryonic units. One high powerfield = 30 ooo JJm (•P < o.os). 69

Pkumtil (u/J9), Vol. 10 DISCUSSION

Here we report that presumptively resorbing embryonic units which become infiltrated by NK lymphocytes also show significant increases in nuclear density and total nuclear area per tissue section. These changes represent an increase in the total numher of cells per volume of tissue in resorbing embryonic units. The increase in ceUularity correlates with the previously described decidual NK infiltration with respect to gestational stage and anatomical location (Gendron and Baines, 1988). Nuclear density and area peak at days 8-9 in the resorbing units, and remain significantly greater than normal until day to. However, increases in nuclear density in resorbing units are also observable in areas of the unit remote to initial NK infiltration {ecto­ placental cone). These results suggest that NK infiltration may only be one component of an uncharacterized cellular response (possibly including other mononuclear and polymorpho­ nuclear leucocytes) occurring throughout the embryonic unit during resorption. The increase in the nuclear density and area in NK infiltrated embryonic units suggests either infiltration of leucocytes from the maternal circulation or proliferation of local cells (causing the appearance of infiltration). Since cells observable within the decidua basalis appear very similar to mononuclear leucocytes in cellular morphology (Kearns, Parhar and Lala, 1985), it is possible that the increased nuclear density and area in resorbing fetoplacental units results from local proliferation of the cells in the decidua basalis of the units themselves. Such a possibility is not unreasonable considering that some cells of the decidua are of bone marrow origin (Kearns and Lala, 1982) and may therefore be capable of stimulus induced pro­ liferation in vivo. Because the area per nucleus does not change appreciably between the nor­ mal and resorbing cases, it does not seem likely that the increased nuclear area (as a percent of total field) refiects nuclear hypertrophy but rather an increase in the total number of cells per volume of tissue. Significant decreases in both the nuclear density and the nuclear area per volume of tissue in the normal embryonic units may reflect developmental changes such as blood space formation and cell membrane surface area expansion during growth of the syncytial trophoblast. In the resorbing embryonic units, the nuclear density and area both increase with gestation (through days 7--9) indicating that the normal development of the unit has been significantly altered. Significant within-group variation was also observed for nuclear density between day 7--9 in the resorbing population of embryonic units for which we have no explanation. However, sig­ nificant within-group decreases in nuclear density and nuclear area observed at day IO in the resorbirig population of embryonic units can be explained by the state of the tissue at this stage. At day 10, embryonic units undergoing resorption are already necrotic and contain only rem­ nants of the feto-placental unit (see results). Cells in the tissue at this state of degeneration are therefore smaller and less numerous (due to autoloysis) than the earlier points at which the inflammatory reaction is at a peak. Natural killer cells initially infiltrate only the decidual-ectoplacental junction of resorbing embryonic units (Gendron and Baines, r988). However, we report here that increased total cell density occurs also in the ectoplacental cone proper. Increased cellular density in the ecto­ placental cone of resorbing units indicates infiltration of cells into an area not associated with increased NK cell density. The percent increase in nuclear density is greater at day 9 versus day 8 in both tissues, suggesting that a progressive increase in cell density occurs in areas of the resorbing embryonic unit remote to the site of initial change. Natural killer cells of donor origin have been implicated in the induction of histopathologic lesions in graft versus host dis­ ease (GVHD), and the lesions have been reported to be most severe in cases where NK activity 0 peaked early in the course of the disease (Ghayur et al., 1987; Ghayur, Seernayer and Lapp, 70 c

1987). Our results suggest a similar pattern of NK activity (as assessed by increased NK cell density) followed by histopathologic lesions (increased total cellularity) in both the decidua and ectoplacental cone. The chronology of this process correlates with preliminary evidence we have, suggesting deficiency of immunosuppressive activity in some embryonic units leads to resorption (unpublished observations). Studies are now underway in which we are assessing individual embryonic units for NK cell density, cellularity changes and the absence of immunosuppressive activity. The findings reported here support the hypothesis that a naturally occurring inflammatory response may contribute to spontaneous pregnancy failure and resorption. We think that early resorption involves a non-specific inflammatory response induced initially by natural killer cells and accompanied by other non-specific leukocytes against antigens which do not normally induce such responses. If such is the case in human spontaneous abortion, similar cellular changes may also be detectable in feto-placental tissue that is aborted very early. The auto­ mated morphometric approach reported in this study, if applied to similar cases in humans could be useful in furthering our understanding of the mechanisms involved in early spon­ taneous abortion or other gynaecological problems associated with the immune system.

SUMMARY

Immunopathology of the spontaneous resorption phenomenon in the CBA x DBA/J murine model was explored using morphometric analysis. Accompanying the previously reported presence of natural killer (NK) cells in resorptive feto-placental units we find major changes in tissue morphology indicating that early infiltration of the feto-placental unit by maternalleuko­ cytes plays a direct role with NK cells in fetal demise. Total number of cell nuclei per field and total nuclear area per field were significantly elevated in feto-placental units containing abnor­ mally increased NK cell presence before detectable resorption as early as day 7 of gestation. This difference persisted throughout all stages of early gestation up to and including the final resorption event at day 10 to 12. Increases in cell density were also detected in areas of the embryonic unit not associated with NK infiltration. These results demonstrate that the spon­ taneous resorption phenomenon in this model involves: (i) Early (day 7-8) c~UW-infiltration of the decidual-ectoplacental cone junction associated with tlie pr*nce in this area of NK cells. (ii) Late (day 8-;j) cellular infiltration of the ectoplacental cone.

ACKNOWLEDGEMENTS

This work was supported by the Medical Researcll Council of Canada. We thank Dr R. Farookhi for critical com­ ments.

REFERENCES

Beer, A. E. & Billingham, R. E. (1976) The ImmuNJiJiology of Matllmlllilm ReproiiiKtion. Englewood Oiffs: Prentice­ Hall. Huhner, J. N. & Johnson, P. M. (tC}Ss) Antigen expression by ttophoblast populations in the human placenta and their poss~"ble immunological relevance. Plattlfla, 6, 1.z7-140. Bulmer, J. N. 11: Sunderland, C. A. (1CJ84) Immunohistological characterization of lymphoid ceiJ populations in the c early human placental bed. J_,ology, sa, .349-357· 71 c

Pkumt11 (11}89), Vol. 10

Oark, D. A., Chaouat, G., Guenet, J. & Kiger, N. (1987) Local active suppression and successful vaccination against spontaneous abortion in CBA/J mice. Jourul ofReprodru:tive lmmUMiogy, 10, 7cr-8S· Oark, D. A., Olaput, A. & Tutton, D. (1986) Active suppression of host vs graft reaction in pregnant mice. VII Spontaneous abortion of allogeneic CBA/J x DBA/a feruses in the uterus of CBA/J mice correlates with deficient non-T Suppressor cell activity.JollnJid oflmt~~Unology, 136, J668-t67S· Oark, D. A., Falbo, M., Rowley, R. B., Banwatt, D. & Stedronska-Oark, J. (11}88) Active suppression of host­ vs-graft reaction in pregnant mice IX. Soluble suppressor activity obtained from allopregnant mouse decidua that blocks the cytolytic effector response to I L-a is related to transforming growth factor-B. }olirulllflmmtmology, 141, J8JJ-J84Q. Oark, D. A., McDcrmott, M. R. & Szewczuk M. R. (1C)8o) Suppression of host versus graft reaction in pregnant mice. II Selective suppression of cytotoxic T cell generation correlates with a soluble suppressor cell activity and with successful pregnancy. Cellular lmm~~Miogy, 52, Jo6-tt8. Crepeau, M. A. & Croy, B. A. (tl}88) Evidence that specific cellular intmunity cannot account for death of Mus C11ro/i embryos transferred to M us M~~KUllis with severe combined intmune deficiency disease. Trllruplantlltiplacentation.JourM/ of1-/ogy, 139, Io8&-10CJS· deFougeroUes, A. R. & Baines, M. G. ( 1987) Modulation of the natural killer cell activity in pregnant mice alters the spontaneous abortion rate. JourMI ofRtprod~~~:rivt Immunology, 11; 147-1 53· Dorshkind, K., Pollack, S. B., Bosma, M. J. & Phillips, R. A. (1985) Natural Killer (NK) cells are present in mice with severe combined immunoditiency (scid). JourMIIIf Immunology, I ]

Alteration of the infiltration of natural killer cella in

CBA/J X DBA/2 concepti by POLY I:C and anti-aaialo GX1 treatment

A) INTRODUCTION Spontaneous fetal resorption in the CBA/J X DBA/2 murine model involves early infiltration of the decidual­ ectoplacental cone junction by natural killer cells (Gendron and Baines, 1988). Previous results from our laboratory have demonstrated that stimulation of NK activity by POLY I:C treatment or the depletion of NK activity by aAGMl treatment in pregnant mice on day 7 can increase or decrease, respectively, the subsequent spontaneous resorption frequency (deFougerolles and Baines, 1987). In order to determine whether NK modulatory treatments also alter the level of NK infiltration into the fetoplacental unit before resorption, we have used immunohistochemistry to measure fetoplacental NK cell numbers after POLY I:C and aAGMl treatments.

B) MATERIALS AND METHODS Female CBA/J mice were purchased from Jackson (Bar Harbor, ME) while male DBA/2 mice were purchased from 0 73 c Charles River (St. Constant, QUE). Mice were housed in a standard specific pathogen free facility, kept under a 12 hour light -12 hour dark cycle, and given free access to food and water. Females were paired overnight with males and checked in the morning for the presence of a vaginal plug. The day of plug detection was designated day zero of gestation. On day 7 of gestation, mice were given single intraperitoneal injections of either 20 ug POLY I:C, 10 ug aAGM1, or PBS vehicle. Each treatment group consisted of 4 or 5 pregnant CBA/J females. Treated mice were then killed on day 8 of gestation, the uteri were removed and frozen in OCT embedding compound. Frozen sections were cut at 5 um thickness and stored at -70°C until immunohistochemistry. Immunohistochemical staining for asialo GM1 and cell quantitation was carried out exactly as described in the previous manuscript (Gendron and Baines, 1988). Briefly, fetoplacental units containing greater than 2 aGM1 positive cells per high power field were designated infiltrated, while those containing less than 2 AGM1 positive cells per high power field were designated non­ infiltrated. Positive controls consisted of staining of the AGM1 epitope on sections of mouse spleen, while c 74 negative controls consisted of sections incubated with normal rabbit serum instead of primary anti-AGMl antibody.

C) RESULTS Table 1 shows the percentage of fetoplacental units containing elevated NK cell numbers and the mean number of NK cells per high power field for each treatment group. As expected, treatment with the NK activating polynucleotide POLY I:C increased the frequency (62%) vs 37% for PBS treated) of fetoplacental units with elevated NK counts. Conversely, treatment with aAGMl decreased the frequency of fetoplacental units with elevated NK counts (3.8%). Interestingly, the mean number of NK cells per high power field calculated for fetoplacental units with elevated NK cell numbers ( >2 NK cells/HPF) was lower for the POLY I:C treated group as compared to the PBS treated group.

0 75

Table 1 Quantitation of NK infiltrates in CBA/J X DBA/2 fetoplacental units from mice receiving NK modulatory treatments

Treatment Frequency (%) Mean NK cells/HPF ±SE n

PBS 37 9.6 ±1.5 11 29

POLY I:C 62 3.4 ±0.5 24 39

aAGM1 3.8 0.6 1 26

The frequency of fetoplacental units (n/nT) and the mean number of NK cells per high power field in fetoplacental units showing elevated NK infiltrates (defined as those implant sites containing greater than 2 AGMl positive cells per high power field as previously described; Gendron and Baines, 1988). All implantation sites in pregnant uteri from each of the 3 groups were examined for AGM1 positive infiltrates.

0 c 76 D) DISCUSSION

Treatment of pregnant CBA/J X DBA/2 pregnancies with POLY I:C and aAGM1 led to an alteration in the frequency of infiltrating NK cells that was consistent with the effect of such treatments on the resorption frequency. As previously reported, POLY I:C and aAGM1 treatments on day 7 of gestation result in 60% and 16% resorption frequencies, respectively, in the CBA/J X DBA/2 pregnancy (deFougerolles and Baines, 1987). our observation that 62% of the POLY I:C treated fetoplacental units contained elevated NK infiltrates corroborates the previous data on resorption. However, only 3.8% of the aAGM1 treated fetoplacental units contained elevated NK cells. In addition, a decreased mean number of NK cells per high power field was measured in these fetoplacental units. This suggests that 12% of the fetoplacental units in the aAGM1 treated group as well as untreated groups may resorb by mechanisms other than those involving increased NK infiltration. These results demonstrate that systemic modulation of NK activity by treatment of CBA/J X DBA/2 pregnacies with POLY I:C and aAGM1 alters the number and frequency of infiltrating NK cells in the fetoplacental unit. As POLY 77

I:C increases the resorption frequency, it also increases the frequency of NK infiltrated fetoplacental units. Conversely, as aAGMl decreases the resorption frequency, it also decreases the frequency of NK infiltrated fetoplacental units. These results strengthen the argument that NK cells are involved in spontaneous fetal resorption in the CBA/J X DBA/2 model.

0 c 78 CHAPTER Ill PREPACE. TBB REGULATION OP NK ACTIVITY IN TBB PBTOPLACBHTAL UNIT AND DEPBCIENT NX SUPPRESSION IN SPONTANEOUS RESORPTION-PRONE CBA/J X DBA/2 CONCEPTI: TBB ROLB OP EICOSANOIDS.

Our· identification of the presence of NK cells in spontaneous fetal resorption raised the question of how these cells are normally regulated in the fetoplacental unit. In addition, we sought out to find whether there was a disturbance of such NK regulatory activity in the fetoplacental units of CBA/J X DBA/2 pregnancies. The rational for choosing to analyse eicosanoids in fetoplacental tissue was based on two main points. First, the activity of NK cells is known to be potently modulated by eicosanoids (Brunda et al., 1980; Ramstedt et al., 1985). Second, prostaglandins have been shown to mediate the regulation of killer cells in mouse decidua (Lala et

al., 1986; 1990). We therefore analysed the NK-modulatory capacity of eicosanoid extracts from resorption-prone fetoplacental tissue. CBA/J X DBA/2 fetoplacental tissue was found to be deficient in NK-suppressive activity. Further analysis of the NK suppressive activity revealed, surprisingly, that LTB4 may play an important NK regulatory role in early gestation. 0 c 79 The following section consists of two manuscripts. The first, Resorption of CBA/J X OBA/2 mouse concepti in CBA/J uteri correlates with failure of the fetoplacental unit to suppress natural killer cell activity is a published paper (J. Reprod. Fertil. 89, 277-284, 1990). The second, Murine pregnancies predisposed to spontaneous fetal resorption show alterations in the concentrations of leukotriene B4 and prostaglandin E2, has been submitted for publication. All the research reported in these two manuscripts was conducted by R.L. Gendron under the supervision of or. M.G. Baines and or. R. Farookhi.

c 80 c Prillt«l ill Gr«JI BTittlilt J. Reprod. Fert. ( 1990) 89, 277-284 IC> 1990 Journals of Reproduction .t Fertility Ltd

Resorption of CBA/J x DBA/2 mouse conceptuses in CBA/ J uteri correlates with failure of the feto-placental unit to suppress natural killer cell activity*

R. L. Gendront, R. Farookhit and M. G. Baines Departments oftPhysiology. Microbiology & Immunology and the Centre for the Study of Reproduction, McGill University. Montreal, Quebec, Canada HJA 284

Summary. Lipid extraction was used to study the natural killer (NK) suppressive activity of individual feto-placental units. Normal pregnancies showed a lipophilic NK cell suppressive factor that was gestational day specific. Feto-placental units from CBA/J x DBA/2 pregnancies were deficient in the NK cell suppressive factor when compared to normal CBA/J x BALB/c pregnancies. The frequency of non-suppressive feto-placental units from CBA/J x DBAI2 pregnancies correlated with the frequency of feto-placental units infiltrated with NK cells and the frequency of spontaneous resorption. Our results implicate a deficiency of NK suppressive activity in the feto-placental unit as a contributing factor in spontaneous fetal resorption.

Keywords: pregnancy; resorption; natural killer cd.ls; immunosuppression; mouse

Introduction

In viviparous pregnancy. the apposition of allogeneic fetal tissue with the maternal immune system evinces the need for local uterine immunosuppression (Beer&. Billingham, 1974; Slapsys &. Oark, 1982; Hsi et al., 1984; Redman et al., 1984; Wells et al., 1984; Nakayama et al., 1985; Croy et al., 1987). Most allogeneic pregnancies progress to term with no evidence of rejection. In mice, spon· taneous fetal resorption frequencies are below lOo/o per pregnancy (Rugh &. Wohlfromm, 1967). A significantly increased murine fetal resorption frequency of approximately 30•;. is observed in CBA/J females mated with DBA/2 males (Clark et al., 1987). The reasons for the increased fetal resorption from this mating are unclear although a deficiency of non-T suppn:ssor a:Us has been implicated (Qark et al., 1986a). Spontaneous fetal resorption in CBA/J x DBA/2 pregnancies may involve natural killer (NK) cells (Gendron &. Baines, 1988). We have observed NK-Iike cells in the junction between the decidua and ectoplacental cone of resorbing feto-placental units as early as Day 6 of gestation. Cytotoxic T Jymphocytes are not involved in this process (Gendron & Baines, 1988). Augmentation of the NK activity by polyinosinic cytidilic acid (POLY I:C) significantly increased the resorption frequency whereas abrogation of the NK activity by anti-asialo-GMI antiserum (aAGMI) signifi­ cantly decreased the resorption frequency (deFougerolles & Baines, 1987). We have therefore argued that regulation of uterine NK activity may play an important role in preventing ~Uular immune reactions against the fetal and trophoblast tissue of the developing feto-placental unit. Natural killer cell activity can be inhibited by arachidonic acid metabolites (Brunda et al., 1980; Ramstedt et al., 198S). It has been suggested that NK cell activity is regulated in vivo through production of prostaglandins (PG) by monocytes and macrophages (Koren et al., 1981 ).

*Reprint requests to: R. L Gendron. Room 404, Lyman Duff Medical Sciences BuiJdina, 377S University SI. c Montnlal, P.Q., Canada H3A 284. 81 c 278 R. L. Gendron et al.

Prostaglandins are produced by the pregnant human, rat and rabbit uterus during the pen­ implantation period (Pakrasi et al., 1985; Malatby et al.• 1986; Parr et al., 1988; Smith & Kelly. 1988). Mouse decidual NK activity can be inhibited by decidual cell-derived PGE-2, and this suppression of decidual NK activity can be reversed by treatment of the mixed population of ceUs with indomethacin {lala et al., 1986). Inhibition by indomethacin of PG synthesis in pregnant mice induces partial or complete litter resorption (Lala er al., 1986). Furthermore, the suppressive mechanism in normal human decidual tissue involves PGE-mediated suppression of lymphokine­ activated killer (LAK) cell activity (Parhar et al., 1988a, b). These findings suggest that arachidonic acid metabolites may be intimately involved in the regulation of NK activity in the feto-placental unit. We have suggested that a failure in decidual PG production in the CBA/J x DBA/2 pregnancy leads to NK cell activation and resorption. Here we present evidence for the deficiency of a lipophilic NK suppressive factor in spontaneous fetal resorption.

Materials and Methods

Allim4ls a4 liuu proct!JM,. CBA/J female mice were purchased from Jackson Laboratories (Bar Harbor. ME. USA). CfW Swiss Webstcr female mice, BALB/c male mice and DBA/2 male mice were purchased from Charles River (St ConstaDt, Quebec, Canada). All animals were 6-8 weeks old. Animals were given free access to chow and water and housed in an animal care facility maintained at 2s•c with a 12-b licbt/dark cycle. Preanancies were obtained by caging S CBA/J or CFW females with either I DBA/2 or I BALB/c male overnight (the major MHC complex of BALB/c is identical to that of DBA/2; K.iger ~~ al., 1985). Tlie morning of detection of a vaginal plug was designated as Day I of gestation. Animals were killed by cervical dislocation on Day 9 of gestation. uteri were removed quickly and frozen immediately on solid C02 . Individual feto-placcntal units were separated by cutting through the areas between implantation sites with a razor blade while the tissue remained on the solid C02• Frozen feto-placcntal units were then weighed individually and stored at - so·c. . . Pregnant uteri from CFW x DBA!J pregnancies were proCessed its described above but feto-placcntal units from each pregnant uterus were combined. This was necessary to allow for the analysis or immunosuppressive activity as a function of gestational day in normal pregnancies. Non-pregnant uteri from 2 CBA/J females were removed. frozen and cut into fragments of various sizes (23-4-33· 7 mg) and processed as described below for individual feto-placcntal units. To examine the role of cyclo-oxygenase products on the putative NK suppressive activity. 4 CBA/J ( x DBA '21 pregnant females were injected (i.p.) with I 1'1 acetyl salicylic acid (Sigma. St Louis, MO, USA) dissolved in phosphate-buffered saline (PBS), 18 b before tissue collection; 4 control (CBA/J x DBA/2) pregnant females received PBS (i.p.) only. This dose of acetyl salicylic acid has been shown to restore NK activity in tumour-bearing mice which have increased PO production (Brunda er al., 1980). ldelltijk«iott a4 processiJt6 offni)-}JI«ea~IIIMits c0111.U.U., aUdD GMJ J(JiiliH «"""' ilt,/ilmllts. An indirect immunoperoxidase staining procedure was uillizcd on froZJ::D SCCiioDs ofCBA/J x DBA/J feto-placcntal units at Day 9 of gestation. Feto-placcntal units were then sorted into normal or prcsumptivdy resorbing groups accordins to the observed NK cell density (Gmdron .t. Baincs, 1988). The tissue blocks were placed on solid C02 and the infiltrated feto-placcntal units were separated from those with no infiltration. All infiltrated feto-placcntal units were pooled in one group and an equal number of normal feto-placcntal units was pooled in a separate group. Pooled tissue in each group was weighed. processed and extracted as described below. f&sru: extnctiM. The prostaglandin-thromboxane lipophilic fraction was extracted from the tissue by methods described previously (Powell, 1982). Weiglled, frozen tissue was homogenized at 4"C for 30 sec by hand in 0· S ml PBS. pH 7·4, containing ts•;. ethanol. Tbe bomogcnatcs were centrifuged at 4001 for 10 min at 4°C. Tbe supernataDts were collected and adjusted to a pH of H with IN-Ha. Tbe supemataDts were applied to SEP-PAK Cl8 columns (Millipore, Bedford, MA, USA) which bad been pre-wet with absolute ethanol and distiDed water. Following washes with 20 ml of IS% ethanol and 20 mJ absolute petroleum ether, the columns were eluted with 10 ml absolute methyl formate. The methyl formate eluants were dried under a stream of nitrogen and reconstituted in PBS, pH 7·4, COD· taining I% ethanol. Reconstitution volume for aU tissues was adjusted according to the initial frozen wcipt (0.01 mJ perms tissue). Extracts were apportioned to freezer tubes., purged with nitrogen and stored at -80"C untll assayed. Extracts were assayed within 3 days. Cytoto:ddq IISIIlF a41Mi11itt1f71111ih. TISSue extracts were tested for NK suppressive activity by serial dilutioa. Suppressive activity was assessed by the degree to which each extract inhibited lysis by virgin female CBA/1 splee:o ceJJs(effectorceJJs)ofYAC-1 taqet ceJJslabeDed withchromjwn SI (Pross & Baines, 1979). Assays were carried out in 96-weU V-bottom plates (Flow Laboratories, McLean, VA, USA) in RPMI 1640 supplemented with 10% beat· inactivated fetal bovine serum (Flow Laboratories. McLean., VA, USA). Effector ccDs (2 x 106 cells per well), 82 c Deficient NK suppression in spontaneous resorption in mice 279

prepared as whole spleen cell suspensions from 2-3 spleens of CBAJJ virpns, were added to wells in .which feto­ placental unit extracts had been diluted serially from 1!2 to 1{32. Chromium SI-labelled target cells (4 x let per well I were then added to all wells to yield a constant effector cell: target ceU ratio of SO: I and a final well volume of 200 ,.U. Control wells contained blank extracts (all extraction steps without tissue). Plates were centrifuged at 200 6 forS min.. incubated at 37cc for 4 h. recentrifuged and ISO pi supematant were harvested from each well. Supematants were counted in a Beckman Gamma counter and the percentage of cell mediated lysis Wo CML) was calculated according to the following formula: (c.p.m. test - c.p.m. spontaneous) 0 /t CML = X IOIWo (c.p.m. max. - c.p.m. spontaneous) Maximum release (c.p.m. max.) was determined by counting aliquants of resuspended target cells alone while spontaneous release (c.p.m. spontaneous) was determined by counting samples of supematant medium from wells containing target cells only. Samples of experimental wells were counted to measure c.p.m. test. Assays of each extract were carried out in triplicate. Control % CML at an effector: target ratio or SO: I was determined by averaging tbc values of% CML from wells without extract. Suppressive activity of tbc feto-placental unit extracts was determined by constructing linear dilution curves of% CML vs extract dilution. lnhlbitory units (IU) were defined as the titre of extract yielding SO% of control cytotoxicity, and were normalized to 100 mg tissue.

Dtlt11 IJIHIIysis. Analysis of variance (ANOVA) was used to distinguish signific8nt differences in the gestational stage specificity of NK suppressive activity in normal pregnancy. Data from experiments on NK suppressive activity in individual feto·placental units from CBA/J x DBA/2 and CBAiJ x BALB/c pregnancies were analysed by Mann-Whitney Wilcoxon non-parametric ranlt tests. Frequencies of non-suppressive feto-placental units. NK-infiltrated feto-placental units and resorption rate were analysed by x2 population analysis. Data from exper­ iments on the effect of acetyl salicylic acid on NK suppressive activity were analysed by Student's 1 test. Significancr was assessed at the P s; 0·05 leYel.

Results

Gestation day speciftdty of NK suppressive activity iD oormal pregnancy The formate-eluted fraction from pooled feto-placental units ofCFW x DBA/J pregnant uteri (normal pregnancy) showed NK inhibitory activity. The inhibitory activity showed a dependence on day of gestation (Fig. I). NK inhibitory activity increased sharply between Days 7 and 8 of gestation. Maximal NK inhibitory activity occurred at Day 8. The activity .then declined through Day 9, and reached negligible levels by Day 10 of gestation. The mean resorption frequency for 9 CFW x DBA/2 pregnancies was 2·2 ± 1·47% when assessed on Day 12 of gestation.

~100 8 ~50

OL-----~7----~~~--~9~--~~====~ Day of gestation fi&. 1. Differential expression ofNK suppressive activity (mean ± s.e.) as a function of day of gestation in normal CFW x DBA/2 pregnancies. Between 4 and 7 animals per day of gestation were analysed by extracting whole uteri. Data were analysed by ANOV A. Data points with different letter designations diJfer significantly. 83

280 R. L. Gendron et al.

Inhibition of NK activity by CBA/J x DBA/1 aod CBA/J x BALB/e feto-plattatal units The mean inhibitory activity of extracts from CBA/J x DBA/2 feto-placental units v;as significantly lower than the mean inhibitory activity for CBA/J x BALB/c feto-placental units (Table 1). When feto-placental units containing no inhibitory activity were excluded from tbe analysis, the mean inhibitory activity ofCBA/J x DBA/2 feto-placental units was still significantly less than that of tbe CBA/J x BALBic feto-placental units (Table I). The frequency of occurrence of feto-placental units with no inhibitory activity was significantly lower for CBAiJ x BALD c pregnancies compared to the CBA/J x DBA/J pregnancies (Table 2).

Table 1. NK inhibitory activity of CBA/J x DBA/2 and CBA/J x BALB/c pregnancies

Inhibitory units/lOO mg

All Suppressive No. of feto·placental feto-placental Mating implantations units units

CBA/J x BALB/c 5·8 ± 0.2 210·21 ± 21·4 217-71 ± 20·8 . CBA/J x DBA/2 8·6± 1-1 91·30 ± tJ-4• 130·86 ± 13-9*

Values are mean ± s.e.m. for 5 individuals at Day 9 of gestation. *Significantly different when compared to control (CBA/J x BALB/c).

Table 1. Correlation between resorption, NK suppression deficiency and NK infiltration in mouse pregnancies

o/o infiltrated % non-suppressive % resorption feto-placental units feto-placental units Mating at Day 12 at Day9 at Day9

CBA/J x BALB/c 4·50 ± 4·5 (4, 23) 5·00 ± 5·0 (5. 33) H ± 5·3 (5. 29) CBA/J x DBA/2 28·54 ± 31)• (8, 81) 30·16 ± 7·3· (6, 44) 29·0 ± 2-J• (5, 43)

Values are mean ± s.e.m. for the no. of animals and individual feto-placental units indicated in parentheses. 1 • P < o-os compared to CBA/J x BALB!c values (X ).

Figure 2 illustrates the differences in the NK inhibitory effect observed for extracts from individual feto-placental units between a CBA/J x DBA/2 and a CBA/J x BALB/c pregnancy. The fet6-placental units from the CBA/J x DBA/2 pregnancy were less inhibitory for NK activity than wtre units from the CBA/J x BALB/c pregnancy. Furthermore, 3 out oftbe 9 feto-placental units in tbe CBA/J x DBA/2 pregnancy contained no NK inhibitory activity. All feto-placental units from the CBA/J x BALB/c pregnancy showed inhibitory activity. When non·pregnant CBA/J uteri were analysed, only 1 of 11 extracted tissue fragments inhibited NK activity. and the level of inhibition (20 IU) was low. Analysis of extracts from 2 whole non-pregnant CBA/J uteri in separate experiments also showed no inhibition of NK activity. No inhibition ofNK activity was observed in any oftbe experiments for blank extracts.

Correlation between NK suppressive acthity, resorptioo rates and asialo GMI-positive ceUular hdiltrates The frequency of feto-ptacental units with oo inhibitory activity in the CBA/J x DBA/l pregnancies was similar to the resorption frequency observed at Day 11 of gestation (Table 2). 84

Deficient NK suppression in spontaneous rt!sorption in mice 281

400 r (a) j

200.

0 4

,llLI_LI_.___ , 2 3 4 5 6 7 8 9 Feto-placental unit Fie. 1. Representative data showing the NK inhibitory activity of individual feto-placentaJ units from single pregnancies. (a) NK inhibitory activity of all (6) individual units from a CBA/ J x BALB/c pregnancy at Day 9 of gestation. (b) NK inhibitory activity of aU (9) individual units from a CBA/J x DBA/2 pregnancy at Day 9 of gestation.

Furthermore, the frequency of feto-placental units with no inhibitory activity was similar to that of units infiltrated by asialo GM l positive NK cells at Day 9 of gestation (Table 2). In the CBA/J x BALB/c pregnancies, the frequency of non-inhibitory feto-placental units, the resorption frequency (at Day 11), and the frequency of feto-placental units with asialo GMI-positive cellular infiltrates were similar and

Etrect of a cyclo-oxy1eDase lobibitor oa fete-placental NI suppressive ldMty Treatment of CBA/J x DBA/2 pregnant females with acetyl salicylic acid resulted in a significant decrease of the NK suppressive activity (23·3 x 12·SIUfJOOmg) compared to CBA/J x DBA/2 pregnant females treated with PBS (88 ± 20-0 IU/100 mg).

Discussioa

These results demonstrate that the deficiency of a lipophilic factor in the feto-placental unit may predispose CBA/J x DBA/2 pregnancies to increased resorptioo. Our aaalysis of normal CFW x DBA/J pregnancies reveals that the NK suppressive activity is gestatioaal day specific. The peak in NK suppressive activity in nonnal pregnancies cx:curs at Day 8-9 ofgestation. The NK suppressive activity reaches neefiaible le¥cls by Day 10-11 of gestation. The demonstration of 0 differential expression ofclass I MHC genes during mid-gestation in the mouse (Hedley et al., 1989) 85 c 282 R. L. Gendron et al.

shows that antigenic presentation in the feto-placental unit may be tightly stage-specific. The early peak in NK suppressive activity during normal pregnancy may therefore correspond to a time of gestation during which fetally derived antigenic stimulation may be maximal. Indeed, the forma­ tion of the ectoplacental cone by invasion into the decidual tissue occurs between Days 6 and 8 (Tbeiler, 1983). Other groups have reported immunosuppressive uterine compounds of various molecular weights (Davies & Browne, 1985; Daya et al., 1987; Oark et al., 1988; Rolfe et al., 1988 ). We were able to abrogate a substantial portion of the NK suppressive activity ofCBA/J x DBA/2 extracts by pretreatment of the pregnant animals with acetyl salicylic acid. This suggests that the suppressive activity may be a product of the cyclo-oxygenase pathway. We have preliminary evidence that NK suppressive factors in normal pregnant uterine tissue can be separated into 2 distinct fractions of small and large molecular weight compounds. Natural killer cell activity present at high levels in normal early mouse decidua (Days 6-8) declines rapidly with gestation and is not associated with resorption (Gambel et al., 1985; Croy et al., 1985). The deficiency of NK suppression in the CBA/J x DBA/2 pregnancy which we observe at Day 9 may permit decidual NK activity to increase rather than decline. Our previous results implicate NK phenotype cells in the induction of resorption between Days 6 and 10 in the CBA/J x DBA/2 pregnancy (deFougerolles & Baines, 1987; Gendron & Baines, 1988). The present results suggest that deficient NK suppressive activity precedes or accompanies the infiltration of NK cells and subsequent spontaneous resorption of a predictable percentage of feto-placental units. Our data suggest that a predictable proportion of CBA/J x DBA/2 feto-placental units cannot provide normal1evels of NK suppressive activity. Previous studies on xenogeneic embryo transplants have revealed that suppressor cell recruitment depends upon the type of trophoblast­ decidual combination and that failure of the transplant occurs due to a lack of sufficient immunosuppressive activity with subsequent infiltration of maternal lymphocytes (Ciark er al., 1986b). There is good evidence for the role of arachidonic acid metabolites in the suppression of NK activity (Brunda et al., 1980; Koren et al., 1981; Rola-Pieszczynski et al., 1982, 1983, 1985; Ramstedt et al., 1985). Recent studies of the modulation of deciduallymphokine-activated killer (LAK) cell activity and IL-2 receptor expression by prostaglandins demonstrate the importance of these compounds in regulating the local immune response within the feto-placenta1 unit (Parhar et al., 1989). lt is therefore likely that the compounds within the feto-placental unit regulating' NK activity are also prostaglandins. Experiments using specific inhibitors of lipoxygenase, cyclo­ oxygenase and arachidonate metabolites during early gestation are underway and should clarify the mechanisms regulating NK activity in the spontaneous resorption model. Our results implicate a deficiency in feto·placental suppression ofNK activity as a predisposing factor for spontaneous resorption in the CBA/J x DBA/2 mouse pregnancy. We propose that the absence or deficiency ofNK suppressive factors, possibly theE series prostaglandins, is associated with regulation of NK activity and subsequent feto-placental resorption.

We thank Dr W. S. Powell for technical discussions. This work was supported by the Medical Research Council of Canada. R.L.G. is the recipient of a FRSQ Doctorat Bourse de Formation.

Refereoces

Beer, A.L lt B!Diop1m, R.L (1974) Host responses to mice. VD. Spontaneous abortion of alloaencic CBA/ intra-uterine tissue, c:dlular and fetal allografts. J. J X DBA/2 fetuses in the uterus or CBAJJ mice kprod. Fert., Suppl. 21, 59-88. con:elatcs with deficient non-T suppressor c:eiJ ...... M.J., Hflllenua, R.B. 4 Hokles, H.T. (1980) activity. J.lmmunol. 136, 1668-1675. Inhibition of murine natural tiller c:dl activity by Clut.D.A..,Crey,B.A.,Rossut,J.II:CIIIoaat.G.(l986b) prostaglandins. J. Immunol. 124, 2682-2687. Immune presensitization and Jocai intrauteriue Clut. D.A.., Clalpat, A. .t T~~ttoa, D. (1986a) Acave defeocesudcterminantsofsuccessorfai!Ureofmurine suppression of host-vs-craft reaction in preanant interspecies pregnancies. J. Reprod. Fert. 71, 633-643. 86 c Deficient N K suppression in spontfJlleous resorption in mice 283

Out, D.A., Claaoat, G., Gaeaet, J.·L 11 Kipr, N. uterus during periimplantation period. Prostagllln­ (1987) Local active suppression and suocessful vacci­ dins 31. 60~14. nation against spontaneous abortion in CBA/J mice. Nakayama, E., Slaiptaka, A., Kodo. H. & Mlwa, S. J. Reprod.lmmunol. 10,79-85. (1985) Suppression of mixed lymphocyte rr.action C1ariL, D.A., Falbo, M., Rowley, R.B., Baawatt, D. 11 by cells or human first trimester pregnancy Stedroaska..Qark., J. (1988) Active suppression of endometrium. J. Reprod.lmmuno/.1, 25-31. host-vs-graft reaction in pregnant mice. IX. Soluble Pakrui, P.L., Bedl.a, R. & Dey, S.L (1985) Cyclo­ suppressor activity obtained from allopregnant oxygenase and lypoxygenase pathways in the mouse decidua that blocks the cytolytic effector periimplantation rabbit uterus and blastocyst. response to IL-2 is related to transforming growth Prostag/alulins 19, 481-495. factor-beta. J. lmmunol. 141, 3833-3840. Partaar, R.S., Keaaedy, T.G. & Lala, P.L (1988a) Croy, B.A., Gamllel, P .. ROSSIUil, J. & Weamaa, T.G. Suppression of lymphocyte alloreactivity by early ( 1985) Characterization of murine decidual natural gestational human decidua. I. Characterization of killer (NK) cells and their relevance to the suocess of suppressor ceUs and suppressor molecules. Cell. pregnancy. Cell. Jmmunol. 93,315-326. lmmunoi. 116, 392-410. Croy, B.A., Wood. W. & Kill&. G.J. (1987) Evaluation of Padw, R.S., Keaaedy, T.G. & Lala, P.L (1988b) intrauterine immune suppression during pregnancy Suppression of lymphocyte alloreactivity by early in a species with epitheliochorial placentation. J. gestational human decidua. 11. Characterization of Immunol. 139, 1088-1095. the suppressor mechanisms. Cell. Jmmunol. 116. Dal·ies, M. & BroWDC, C.M. (1985) Pregnancy-associated 411-422. nonspecific immunosuppression: kinetics of the Parllar, R.S., Yapl, S. & I.ala, P.L (1989) POEl generation and identification of the active facton. mediated immunosuppression by· first trimester Am. J. Reprod. Jmmunol. Microbiol. 9, 77-83. human decidual cells blocks activation of maternal Daya, S., Roseudaal. LL & Clark, D.A. ( 1987) lmmuno­ leukocytes in the decidua with potential anti­ suppressor factor(s) produced by decidua-associated trophoblast activity. Cel/.lmmunol. 118., 61-74. suppressor cells: a proposed mechanism for fetal Pan, M.B., Pll't, E.L, Mwaarette. K., C1ariL. M.R. & J. 1!6, allograft survival. Am. Obstet. G;vrwcol. Dey, S.L (1988) lmmunohistochemicallocalization 344-350. of prostaglandin synthase in the rat uterus and & defougerolles, A.R. Baiaes, M.G. (1987) Modulation embryo during the periimplantation period. Bioi. of natural killer cell activity alters the spontaneous Reprod. 31, 333-343. abortion rate in mice. J. Reprod. lmmunol. 11, Powtll. W.S. ( 1982) Rapid extraction or arachidonic acid 147-153. metabolites from biological samples using octadecyl­ Gamllel, P., Croy, B.A., Moon, W.D., HIIIIZiker, R.D., silyl silica. In Methods in &zymology, Vol. 86, T.G. & Roasaat, J. (1985) Characterization w....., pp. 467-477. Eds W. E. Lands .t W. L. Smith. of immune effector cells present in early murine Academic Press, New York. decidua. Cell. Jmmunol. 93,303-314. GeDdroa, R.L & Baiaes, M.G. (1988) Infiltrating Pr-. H.F. & Babies, M.G. (1979) Spontaneous human decidual natural killer cells are associated with lymphocyte mediated cytotoxicity against tal'p:t spontaneous abortion in mice. Cell. lmmunol. 113, cells. I. The effects of malignant disease. /ru. J. 261-267. Cancer.te, 113-119. Hetlley, M.L, Drake. B.L. Hea4, J.R., Tlder, P.W. & RUIISte4t. lJ., Na. J., WipeD, H., Serhaa, C.N. & Fonua, J. (1989) Differential expression of the class Samuelaoa, B. ( 1985) Action of novel eicosaDOids I MHC genes in the embryo and placenta durin& mid­ lipoxin A and B on human natural killer cell cytotoxi­ gestational development in the mouse. J. lmmunol. city: effects on intracellular c:AMP and taraet cell 141,4046-4053. binding. J. /mmtiiiOI. 1.35, 3434-3438. lW, B.·L, Yek, C...J'.G. & Faullr., W.P. (1984) Class l Redmu, C. W.G., McMlcllael, A.J., Sdmat., G.M., antigens of the major histocompatibility complex SuDdedallll. C.A. A 1111&. A. (1984) Class I major on cytotrophoblast of human chorion laeve. histocompatibility complex antigens on human atra­ Immunology 51, 621-629. villous trophoblast. Immunology 52, 457-468. ltices'. N., Claaoat, G., Kolh, J.P., Wecma-, T.G. & Rola-~ M., Gapoa, L & Slrois, P. (1982) G..-. J.L. (1985) Immunogenetic studies of Leukotriene B. induces human suppressor lym­ spontaneous abortion in mice. Preimmunization of phocytes. Biocllem. Biophys. Res. Commun. lttl, females with aDogeneic males. J. lmmunol. 134, 1531-IS37. 2966-2970. Rola-Pieszczyasld. M., Gapoa, L & Sirois, P. (1983} Korea, H.S., Aadersoa. s.J., Fiscber, D.G., Copelull, Leukotriene B" augments human natural cytotoxic C.S. & Jaaa, P.J. (1981) Regulation of human cell activity. Biodll!m. Biophys. Res. Commun. 113, natural tilling. I. Tbc role of monocytes, interferon 531-537. and prostaglandins. J. Immunol. 117, 2007-2013. Rola·l'leszczyasld,M., Gapoa, L.Boduc.D. & LeB.retoll, Lala., P.K., Padw, R.S..., Kearus, M., J..._, S. A G. (1985) Evidence for the involvement of the throm­ Sc:oclru, J.M. (1986) Immunologic aspects of the boxane synthase pathway in human natural cytotoxic decidual respoDSC. In Reprodllt:ti.e Immunology, cell activity. J.Imrmmol. 135,4114-4119. 1986, pp. 190-198. Eels D. A. Clarll: .t B. A. ClOy. Rol'e, B.E., Ca....P, A.C., Qui-. K.A. A Mono.,IL Elsevier, New y ork. (1988) Identification or two suppressor factors 0 Mala6J, P.V., C11eec. ILC. & Dey, S.L (1986) Produc­ induced by early pregnancy factor. Clilt. up. tion of leukotrialel and prostasfandins in the rat lnrtrwnDI. 73.219-225. 87

284 R. L. Gendron et al.

Rugb, R. & Woblf.._, M. (1967) The reproductive 'l'lleiler. K. (1983) The MOI.IM ilf Biomedical Research. 111. perfonnana: of the laboratory mouse: maternal age, lVOI'mati~ Biology. Immunology and Husbandry. litter size and sex ratios. Proc. Soc. exp. 8io/. Med. pp. 121-136. Academic Press. Toronto. 116.685-687. w• M •• Hsi. 8.-L A Falllk. W.P. (19841 The Slapsys., R. & Oarl., D.A. (1982) Active suppression of expression of class I antigens of the major histocom· host-vs-graft reaction in pregnant mice. IV. Local patibility complex on cytotrophoblastic shell cells of suppressor cells in decidua and uterine blood. J. the placental basal plate. Am. J. Reprod. lmmunol. 6. Reprod. ln'llf'IWtol. 4, 355-364. 175-181. Smith, S.L A l:eUy. R.W. (1988) The release of PGy2 alpha and PG£ 2 from separated cells of human endo­ metrium and decidua. Prosiag/andins Uukotrienes and Essential Folly Acids 33.91-96. Received IS September 1989

• 88 c Murine preqnancies predisposed to spontaneous resorption show alterations in the concentrations of leukotriene B4 and prostaqlandin E21

R.L. Gendron2, 4 , R. Farookhi2 and M.G. Baines3

Departments of Physioloqy2 ' Microbiology ' Immunoloqy3 and the McGill Centre for the study of reproduction. Lyman Duff Medical Sciences Building, 3775 University St. Montreal, Quebec, H3A 2B4, Canada.

1 This work was supported by a grant from the Medical Research Council of Canada. R.L.G. is supported by a Doctoral Bourse d'Etude from the Fonds de la Recherche en Sante du Quebec. 4 Reprint Requests 89

ABSTRACT

Female CBA/J mice mated with DBA/2 males exhibit an increased spontaneous resorption rate (30-35%) in their first pregnancy. Second pregnancies show a decreased resorption rate (15-20%). In contrast, resorption in CBA/J females mated with BALB/c males (identical to DBA/2 at the H-2 major histocompatibility locus) occurs with a frequency of 5-10%. Resorption is preceded by fetoplacental infiltration of natural killer (NK)-like cells and a deficiency in a lipophilic NK-suppressive activity. The eicosanoids leukotriene 84 (LTB4) and prostaglandin E2 (PGE2) are known to modulate NK activity in vitro. We have measured the concentrations of LTB4 and PGE2 in extracts of individual fetoplacental units at day 8 of gestation from 1) Primigravid CBA/J X DBA/2 resorption-prone matings (RES); 2) Second CBA/J X DBA/2

matings (SEC); 3) Primigravid CBA/J X BALB/c control matings (CON). We detected a significant decrease in the mean concentration of LTB4 in RES fetoplacental units (176.4 ±11.8 pg/ml; n=42) compared with CON and SEC fetoplacental units ( 570.2 ±45. 5 pgjml; n=21 and 420.2 ±59·. 5 pgjml; n=39, respectively). We also detected a small but 0 90

significant increase in the mean concentration of PGE2 in RES fetoplacental units (18.0 ±0.3 ngjml) compared with CON and SEC fetoplacental units (15.0 ±1.3 and 12.4 ±0.9 ngjml, respectively). In order to confirm that the LTB4 deficiency is associated with decreased NK suppression in RES matings, we supplemented RES extracts, in vitro, with exogenous LTB4 (O- 500 pgjml). Addition of LTB4 to RES extracts increased the extract's NK suppressive capacity, whereas LTB4 alone either stimulated NK activity or was without effect. These results suggest a critical role for LTB4 in averting NK-mediated early spontaneous fetal resorption.

0 91

INTRODUCTION

Spontaneous and induced fetal resorption in mice is associated with the infiltration and activation of a population of GM1-positive natural killer (NK)-i'ike cells (de Fougerolles and Baines, 1987: Gendron and Baines, 1988, 1990: Chaouat et al., 1988: Lala et al., 1990). There is ample evidence for the involvement of arachidonic acid metabolites in the local regulation of non-specific maternal anti-fetal immune reactions. Decidual natural killer and lymphokine-activated killer cell activity in pregnant mice is diminished by decidual-derived PGE2. Activity can be restored by treating the cells with the prostaglandin synthesis inhibitor indomethacin (Lala et al, 1990). Furthermore, spontaneously resorbing CBA/J X DBA/2 fetoplacental units are deficient in an NK- inhibitory activity recoverable in eicosanoid fraction extracts during early stages of normal pregnancies (Gendron et al., 1990). These observations suggest that eicosanoids may be important in regulating the activity of a population of nonspecific killer cells in the fetoplacental unit during early stages of normal pregnancy. In addition to the inhibitory effects of the E-series of 0 c 92 prostaglandins, other eicosanoids are known to modulate NK activity in vitro. Leukotriene B4 (LTB4), a product of the 5-lipoxygenase pathway, has been reported to augment NK and natural cytotoxic activity at concentrations of 10- lOM (Bray and Brahmi, 1986; Gagnon et al, 1986; Rola­ Pleszczynski et al., 1983). However, the role of LTB4 in NK activation is not entirely understood; some studies have shown no augmenting effects (Samuelsson et al.,· 1985). Both LTB4 and PGE2 have been detected in uterine tissue during early pari-implantation stages in rat pregnancy (Malathy et al., 1986). These studies, however, were not extended further than day 6 of gestation. We demonstrated previously that lipid extracts from day 8 fetoplacental units of CBA/J X DBA/2 primigravid matings were deficient in an NK-inhibitory activity (Gendron et al., 1990). The absence of this activity occurred with a frequency similar to the resorption frequency (Gendron et

al., 1990). We undertook the present investigation to determine if the local fetoplacental concentrations of LTB4 and PGE2 differed in normal versus resorption-prone matings. The present results demonstrate decreased concentrations of LTB4 in resorption-prone primigravid CBA/J X DBA/2 fetoplacental units. Since LTB4 concentrations were 93 decreased in RES matings, we examined the effects of adding exogenous LTB4 to RES extracts. LTB4 supplementation increased the NK suppressive activity of RES extracts. These results suggest that LTB4 may be necessary for the normal regulation of NK activity during early gestation.

MATERIALS AND METHODS

Animals Virgin female mice of the CBA/J strain were purchased from Jackson Laboratories (Bar Harbor, ME). Male mice of the DBA/2 and BALB/c strains were purchased from Charles River (St. Constant, Quebec). Mice were used at 8 to 9 weeks of age. Two different groups of matings were conducted. In the first, virgin CBA/J females were mated with either DBA/J or BALB/c males. In the second, CBA/J females which had delivered CBA/J X DBA/2 litters one-two weeks previously were mated with DBA/2 males to yield second CBA/J X DBA/2 pregnancies. Females paired overnight with males were checked the following morning for the presence of a vaginal plug. The day of plug detection was designated day 0 of gestation. Pregnant mice were killed by cervical dislocation on day 8 of c 94 gestation. Uteri were removed immediately, snap frozen on

solid co2 and fetoplacental units were separated from the frozen uteri. Individual fetoplacental units were placed in individual freezer vials, weighed and stored at -ao0 c until extraction (not more than 1 day later). Resorption frequency was assessed at day 11-12 of gestation in seperate groups of CBA/J X DBA/2 first and second matings by gross inspection of the uterine contents.

Chemicals, materials and reagents Eicosanoids (LTB4 and PGE2) were purchased from Sigma Chemical (St. Louis, MO), and diluted into culture medium from stock solutions (1mgjml absolute ethanol) immediately before use. Tissue culture medium (RPMI 1640) and fetal bovine serum were purchased from Flow Laboratories (McLean, VA). Petroleum ether and methyl formate used in the extraction procedure were purchased from Sigma Chemical (St. Louis, MO). The SEP-PAK C18 octadecylsylil silica (ODS) cartridges were purchased from Waters, Bedford, MA).

Eicosanoid extraction and measurement Eicosanoids were extracted from individual fetoplacental 0 95 c units with SEP-PAK C18 ODS cartridges as described previously (Gendron et al., 1990). Briefly, weighed frozen tissue was homogenized at 4°c for 30 sec by hand in PBS pH 7.4 containing 15% ethanol. The homogenates were centrifuged at 400g for 10 min at 4°C. The supernatants were collected and adjusted to a pH of 3.5-4 with 1 N-HCl. The supernatants were applied to SEP-PAK C18 columns which had been pre-wet with absolute ethanol and distilled water. Following washes with 15% ethanol and absolute petroleum ether, the columns were eluted with absolute methyl formate. The methyl formate eluants were dried under a stream of nitrogen. The extracts were reconstituted in PBS pH 7.4 containing 1% ethanol at volumes adjusted according to the initial weight of the fetoplacental unit (0.01 mljmg tissue) and stored at -ao 0 c until assay (not more than 1-2 days later). The concentrations of LTB4 and PGE2 in the extracts were measured using radioimmunoassay kits purchased from Advanced Magnetics (Cambridge, MA). Specific antisera supplied with the kits were reported by the supplier to show less than 1% cross-reactivity with other related and non-related arachidonate metabolites. Assay sensitivities were 4 and 8 pgjml for LTB4 and PGE2, respectively. The inter- and intra-asssay coefficients of variation were c 96

less than 10% for either assay.

HK activity measurements NK activity was measured by a 4 hour micro-assay in which mouse spleen cells are mixed with radiolabeled target cells at fixed effector-target cell ratios. (Pross & Baines, 1981). Assays were conducted in 96 well v­ bottom plates (Flow Laboratories, McLean, VA, U.S.A.) in RPMI 1640 medium supplemented with 5% heat inactivated fetal bovine serum. Effector cells (5 x 105 cells per well), prepared as whole spleen cell suspensions from spleens of CBA/J virgins were added to wells containing various concentrations of extract andjor LTB4. Chromium 51 labelled YAC-1 target cells (1 x 104 per well) were then added to all wells to yield a constant effector cell:target cell ratio of 50:1 in a final well volume of 200 ul. Plates were centrifuged at 200 g for 5 min,

incu~ated at 370c for 4 h, recentrifuged and 150 ul

superna~ant were harvested from each well. Supernatants were counted in a Beckman Gamma counter and the percentage of cell mediated lysis (% CML) was calculated according to the following formula:

0 97 c (c.p.m. test - c.p.m. spontaneous) % CML = ------X 100% (c.p.m. max - c.p.m. spontaneous)

Maximum release (c.p.m. max) was determined by counting aliquots of resuspended target cells alone while spontaneous release (c.p.m. spontaneous) was determined by counting samples of supernatant medium from wells containing target cells only. Samples of experimental wells containing lymphocytes and targets with extract and/or LTB4 were counted to measure c.p.m. test. Control % CML at an effector:target ratio of 50:1 was determined by averaging the values of %CML from wells containing either medium alone or extract buffer (PBS containing 0.5% ethanol). All assays were carried out in triplicate.

Statistics Differences in fetal resorption frequency were analysed by Student's t Test. LTB4 and PGE2 concentrations in individual fetoplacental units were analysed by the Mann­ Whitney-Wilcoxon rank test, A minimal significance was determined at p<0.05.

RESULTS 98 c

The mean resorption frequency in 5 primigravid DBA/2- mated CBA/J females was 35.1 ± 6.9 %, while the mean resorption frequency in 6 CBA/J X DBA/2 second pregnancies was significantly lower (16.7 ± 3.9 %). The mean resorption frequency in CBA/J X BALB/c control matings previously reported in this and other laboratories is below 5% (Gendron et al., 1990; Clark et al., 1987). Primigravid CBA/J X DBA/2 (RES) fetoplacental units contained significantly lower mean concentrations of LTB4 than second mating CBA/J X DBA/2 (SEC) or control CBA/J X BALB/c (CON) fetoplacental units (Table 1). RES fetoplacental units showed a small but significant increase in the mean PGE2 concentration (Table 1). There was no significant difference in LTB4 or PGE2 concentrations of fetoplacental units from the second CBA/J X DBA/2 compared with those of the CBA/J X BALB/c matings. Extracts from RES matings contained significantly less NK suppressive activity than in CON matings, as we have described previously (Figure 1; Gendron et al., 1990). In order to confirm that decreased LTB4 concentration is associated with the deficient NK suppressive activity of RES matings, we added exogenous LTB4 to RES extracts c 99 c before testing their NK modulatory capacity. Supplementation in vitro with pure LTB4 of the extracts from primigravid CBA/J X DBA/2 fetoplacental units resulted in significantly increased NK suppressive activity as compared to extract alone (Figure 1). However, the increased extract suppression with LTB4 was dose-dependent. LTB4 was effective in increasing the suppressive activity of the extract in the concentration range of 30 to 125 pgjml. At higher concentrations of LTB4 (250 to 500 pg/ml), in the presence of extract, augmentation of NK activity was observed. Addition of LTB4 (50 pgjml) to CON extracts resulted in a further increase of CON NK suppressive activity (data not shown). Addition of LTB4 alone to the cytotoxicity assay at concentrations similar to those measured in the extracts of the normal matings resulted in significant augmentation of NK activity while concentrations effective in increasing the suppressive activity of the extract had no significant effect (Figure 1).

DISCUSSION

· These results suggest that LTB4 may be an important regulator of NK cell activity in the fetoplacental unit. 0 100

At an early stage prior to visible fetal death, fetoplacental units from primigravid CBA/J X DBA/2 matings predisposed to spontaneous fetal resorption contained decreased concentrations of LTB4 compared with fetoplacental units from second CBA/J X DBA/2 or control

CBA/J X BALB/c matings~ Fetoplacental unit extracts from resorption-prone primigravid CBA/J X DBA/2 matings supplemented in vitro with exogenous LTB4 suppressed NK activity significantly more than the same extracts alone. These results also suggest that the increased viablity of CBA/J X DBA/2 fetoplacental units in second matings may be related to the increased capacity of the CBA/J multigravid pregnant uterus to produce LTB4. The fetoplacental LTB4 concentration in early allogeneic pregnancy may thus be critical in the regulation of NK mediated maternal anti- fetal cellular immune responses. The early transient peak of decidual NK activity previously reported in normal murine gestation has been hypothesized to be promptly down-regulated to negligible levels by day 10-12 (Croy et al., 1985; Gambel et al., 1986; Croy et al., 1988). Our previous results demonstrated that the NK suppressive capacity of lipophilic extracts from normal fetoplacental units peaked at day 8 of gestation while CBA/J X DBA/2 RES matings were 101

deficient in NK suppression on day 8 (Gendron et al.,

1990). In addition, the NK suppressive activity was partially inhibited by acetylsalicilate, suggesting that (one of) the factor(s) responsible for the effect may be a cyclooxygenase-derived eicosanoid (Gendron et al., 1990). The present data however suggest that the deficiency in NK suppression capacity is not due to a deficiency in the ability of the RES fetoplacental unit to produce PGE2, since the concentrations of PGE2 were either close to normal or increased in RES matings. Rather, these results suggest an inability of RES fetoplacental units to produce sufficient concentrations of LTB4 and thus favoring the augmentation of NK activity. The ratio of LTB4 to PGE2 may be important in RES fetoplacental units, as the more successful SEC and CON matings show LTB4/PGE2 ratios over

three fold higher (350 and 380, respectively) than in RES

matings (103). The concentration of LTB4 in the SEC and CON fetoplacental units is approximately 1o-10M, and hence within the putative stimulatory range reported by others (Rola-Pleszczynski et al., 1983; Bray & Brahmi, 1986;

Gagnon et al., 1986). CON extracts, however, containing these concentrations of LTB4 suppress NK activity. Extracts from LTB4-deficient RES matings show decreased 0 102 c

NK-suppression. These findings suggest that the amount of LTB4 in the extract may be critical for expression of suppressive capacity. Extracts from RES matings supplemented with LTB4 at concentrations from 30 to 125 pgjml (lo-10M) suppress NK activity significantly more than the same extracts without exogenous LTB4. LTB4 either alone or in the presence of extract significantly augmented mouse NK activity at concentrations of 250 pgjml or greater (lo-1° to 10-9M). Thus, concentrations of LTB4 ranging within an order of magnitude of one another appear to have a unique modulatory effect on NK activity. The NK modulatory activity of LTB4 may be dependent on the presence of additional factors, possibly other eicosanoids, in the extracts. Two products arising from the interaction between the 5 and 15-lipoxygenase pathways, Lipoxins A and B (LXA and LXB) are potent suppressors of NK activity (Ramstedt et al., 1985). In contrast, the hydro-peroxy precursor to LTB4, 5-HPETE (5, hydroperoxy-eicosatetraenoic acid) is known to potently stimulate NK activity at sub-micromolar concentrations in vitro (Bray and Brahmi, 1986). Recent results from our laboratory show that the lipoxygenase metabolite 12-HETE is also present in the extracts. Preliminary experiments suggest that 12-HETE may interact with LTB4 to modulate NK 103

activity in the fetoplacental unit (manuscript in preparation) • The results presented here implicate fetoplacental LTB4 deficiency as a predisposing factor in CBA/J X DBA/2 murine pregnancies showing spontaneous fetal resorption. Furthermore, fetal resorption in CBA/J X DBA/2 first matings does not appear to correlate with a deficiency in the fetoplacental production of PGE2. We propose that the presence of LTB4 during early gestation may be important in the regulation of decidual NK cells in the developing fetoplacental unit.

ACKNOWLEDGMENTS

We thank Emilia Antecka and Alain Duclos for their deft technical assistance and Dr. w.s. Powell of the Department of Medicine, McGill University for technical discussions and critical comments.

c 104

R.BPERENCES

Bray, R.A., Brahmi, z., 1986. Role of lipoxygenation in human natural killer cell activation. J. Immunol. 136:5: 1783-1790. Clark, D., Slapsys, R.M., Croy, A., Rossant, J., 1983. Suppressor cell activity in uterine decidua correlates with success or failure of murine pregnancies. J. Immunol. 131: 540-546. Clark, D.A., Chaouat, G., Guenet, J.-L., Kiger, N., 1987. Local active suppression and successful vaccination against spontaneous abortion in CBA/J mice. J. Reprod. Immunol. 10: 79-85. Clark, D.A., Chaput, A., Tutton, D., 1986. Active suppression of host-vs-graft reaction in pregnant mice VII. Spontaneous abortion of allogeneic CBA/J X DBA/2 fetuses in the uterus of CBA/J mice correlates with deficient non-T suppressor cell activity. J. Immunol. 136:5: 1668-1675. Croy, B.A., Gambel, P., Rossant, J., Wegman, T., 1985. Characterization of murine decidual natural killer (NK) cells and their relevance to the success of pregnancy. Cell. Immunol. 93:315-326. Croy, B.A., Waterfield, A., Wood, W., King, G.J., 1988. 105

Normal murine and porcine embryos recruit NK cells to the uterus. Cell. Immunol. 115:471-480. de Fougerolles, A.R., Baines, M.G., 1987. Modulation of natural killer cell activity alters the spontaneous abortion rate in mice. J. Reprod. Immunol. 11:147-153. Gambel, P., Croy, B.A., Moore, W.D., Hunziker, R.D., Wegman, T., Rossant, J., 1985. Characterization of immune effector cells present in early murine decidua. Cell. Immunol. 93:303-314. Gagnon, L., Girard, M., Sullivan, A.K., Rola-Pleszczinski, M., 1986. Augmentation of human natural cytotoxic cell activity by leukotriene B4 mediated by enhanced effector-target cell binding and increased lytic efficiency. Cell. Immunol. 110:243-252. Gendron, R.L., Baines, M.G., 1988. Infiltrating decidual natural killer cells are associated with spontaneous abortion in mice. Cell. Immunol. 113:261-267. Gendron, R.L., Farookhi, R., Baines, M.G., 1990. Resorption of CBA/J X DBA/2 mouse conceptuses in CBA/J uteri correlates with failure of the feto-placental unit to suppress natural killer cell activity. J. Reprod. Fertil. 89:277-284. Gendron, R.L., Nestel, F.P., Lapp, w.s., Baines, M.G., 1990. Lipopolysaccharide induced fetal resorption in 0 106

mice is associated with the intra-uterine production of tumor necrosis factor alpha. J. Reprod. Fertil. 90:395-402. Lala, P.K., Scodras, J.M., Graham, c., Lysiak, J.J., Parhar, R.S., 1990. Activation of maternal killer cells in the pregnant uterus with chronic indomethacin therapy, IL-2 therapy or a combination therapy is associated with fetal demise. Cell. Immunol. 127:344- 351. Malathy, P.V., Cheng, H.C., Dey, S.K., 1986. Production of leukotrienes and prostaglandins in the rat uterus during periimplantation period. Prostaglandins 32:605- 614. Pross, H.F., Baines, M.G., 1979. Spontaneous human lymphocyte mediated cytotoxicity against target cells. 1. The effects of malignant disease. Int. J. Cancer 40:113-119. Rola-Pleszczynski, M., Borgeat, P., Sirois, P., 1982. Leukotriene B4 induces human suppressor lymphocytes. Biochem. Biophys. Res. Comm. 108:4:1531-1537. Rola- Pleszczynski, M., Gagnon, L., Sirois, P., 1983. Leukotriene B4 augments human natural cytotoxic cell activity. Biochem. Biophys. Res. Comm. 113:2:531-537. Ramstedt, u., Ng, J., Wigzell, H., Serhan, c., Samuelsson, c 107

B., 1985. Action of novel eicosanoids lipoxin A and B on human natural killer cell cytotoxicity: effects on intracellular cAMP and target cell binding. J. Immunol. 135:5:3434-3438. Samuelsson, B., Dahlen, S.-E., Lindgren, J.A., Rouzer, C.A., Serhan, C.N., 1987. Leukotrienes and lipoxins: Structures, biosynthesis, and biological effects. Science 237:1171-1176.

c 108

Table 1. LTB4 and PGE2 Concentrations in Primigravid CBA/J X DBA/2, Second CBA/J X DBA/2 and CBA/J X BALB/C Matings.

Mating LTB4 pgjml PGE2 ngjml n

RES 176.4 ±11.8* 17.9 ±0.32* 42 SEC 420.2 ±59.5 12.4 ±0.94 39 CON 570.2 ±45.5 15.0 ±0.13 21

The mean (±SE) conce~trations of LTB4 and PGE2 in fetoplacental unit extracts from (n) individual implantation sites at day 8 of gestation in 6 primigravid CBA/J X DBA/2 (RES) matings, 4 second CBA/J X DBA/2 (SEC) matings and 4 control CBA/J X BALB/c (CON) matings (* p<0.05).

0 109

Piqure 1. Mean (+/-SE) percent of control cell-mediated lysis (CML) of YAC-1 mouse tUmor target cells by CBA/J spleen cells (effector:target ratio of 50:1) as modulated by: fetoplacental extract from resorption prone primigravid CBA/J X DBA/2 matings (RES) at a 1/8 dilution; fetoplacental extract from CBA/J X BALB/c control matings (CON) at a 1/8 dilution. The lines represent % of control CML in the presence of either: medium alone (horizontal line), a 1/8 dilution of RES extract supplemented in vitro with a range from 30 pgjml to 500 pgjml of LTB4 (*), or the same range of LTB4 alone (+). Each point represents the mean of triplicate determinations from three separate experiments.

0 110

~­ ~'""""""''

0 0 .. 10 e-...

-

0 (I) ....1 :E 0 ....1 0 a: 1- z z 0 0 (,) 0 LL 0w 0 a: fli.l I 0 0 0 0 0 (\I ... ~-- 111

CHAPTER IV PREFACE. THE DUAL ROLE OF TUMOR NECROSIS

FACTOR ALPHA IH EMBRYONIC DEVELOPMENT AND FETAL RESORPTIOH

The presence of NK cells in the fetoplacental units of spontaneously resorbing pregnancies suggested that the maternal reaction during resorption was a non-MHC restricted, non-specific immune response. The other major non-specific cellular immune component in the fetoplacental unit is the macrophage. The tempo of fetal resorption is rapid and much like the TNF-a mediated necrosis of tumors observed in endotoxin-treated animals (Carswell et al., 1975). We therefore investigated the role of TNF-a by expanding the LPS-induced fetal resorption model first described by Zahl and Bjerknes in 1943. Our results revealed a period of extreme sensitivity to LPS during early gestation. We also were able to detect TNF-a in the amnionic fluids of LPS treated mice. During the course of our investigation, I attempted and was unable to definitively localize TNF-a in resorbing placentae by immunohistochemistry (only slightly above background-level TNF-a immunoreactivity was detected). Instead, I noticed intense TNF-a immunoreactivity in the embryonic brain. Our further analysis of the TNF-a expression in the embryo suggests that this protein may 112

have a role in the development of the normal embryonic nervous system in vertebrates. These results demonstrate that TNF-a may be involved in non-specific effector roles in fetal resorption and also exemplify the plieotrophic nature of this protein through developmental regulation. The following section consists of two manuscripts. The first, Lipopolysaccharide induced fetal resorption in mice is associated with the intra-uterine production of tumor necrosis factor alpha, and the second, Expression of tumor necrosis factor alpha in the developing nervous system are published papers (J. Reprod. Fertil. 309, 1990; Int. J. Neurosci. In Press, respectively). The research reported in the first manuscript was conducted by R.L. Gendron, with the exception of the TNF-a assay (conducted by F.P. Nestel), in collaboration with Dr. w.s. Lapp of the Physiology Department at McGill. The research reported in the second manuscript was conducted by R.L. Gendron, with the exception of the activated spleen tissue preparation (conducted by F.P. Nestel), in collaboration with Dr. w.s. Lapp. All research was conducted under the supervision of Dr. M.G. Baines and Dr. R. Farookhi.

0 113 c Prillt~d ill Grtat lkitaill J. Reprod. Fert. (1990) 90, 395-402 Cl 1990 Journals of Reproduction&: Fertility Ltd

Lipopolysaccharide-induced fetal resorption in mice is associated with the intrauterine production of tumour necrosis factor-alpha

R. L. Gendron*, F. P. Nestel*, W. S. Lapp* and M. G. Bainest Departments of• Physiology and t Microbiology & Immunology. McGi/1 University, Montreal. Lyman Duff Medical Sciences Building. Room 404,3775 University St, Montreal. Quebec, Canada H3A-2B4

Summary. Certain strains of mice display an increased frequency of fetal resorption, but little is known about the effector mechanisms involved. We have examined the events associated with lipopolysaccharide (LPS}-induced fetal resorption in mice. Administration of 25 J.18 LPS on Day 12 of gestation resulted in the appearance of tumour necrosis factor-alpha (TNF-a) in the amniotic fluid and fetal resorption. Levels ofLPS-induced TNF-a were reduced by 90% after pretreatment with the TNF­ a-suppressing drug pentoxifylline (PXF). Treatment of p~gnant mice during early gestation with 0·1 JJg LPS resulted in fetoplacental resorption which was maximal when the LPS was given on Day 8. Resorption induced by 0·1 J.t8 LPS on Day 8 of gestation was significantly reduced by pretreatment with PXF. Infiltration of asialo-GM 1- positive cells was observed in the decidual-ectoplacental cone area of embryonic units from LPS-treated mice. In addition. treatment with anti-AGM I antiserum prevented the LPS-induced resorption. Our results suggest that TNF-a and asialo-GMJ-positive cells are involved in LPS-induc:ed fetal resorption.

Keywords: abortion; lipopolysaccharide; TNF-a; natural killer cells; pentoxifylline; mouse

lntroductioo

Spontaneous fetal resorption in CBA/J x DBA/2 mice has been shown to be associated with early decidual infiltration of natural killer (NK)-Hke cells (Gendron & Baines, 1988). Previous results have shown that treatment of pregnant CBA/J x DBA/2 mice with polyinosinic cytidylic acid (poly I:C). which is known to activate NK cells through macrophage interferon production, increases the resorption frequency (deFougerolles & Baines, 1987). These findings suggest that activation of non-specific immune mechanisms in the fetoplacental unit can ioftuence fetal survival. It has been previously shown tbat bacterial endotoxins induce fetal resorption in mice and rats (Zahl & Bjerknes, 1943; Rieder & Thomas, 1960; McKay & Wong, 1963). However, endotoxins do not appear to interupt the normal endocrine functions of pregnancy and do not cross the placental barrier (Chedid et al., 1962; Parant & Chedid, 1964; Gasic et al., 1975). One of the characteristic effects of bacterial endotoxins such as LPS is the triggering of tumor .. necrosis factor-alpha (TNF-a) release from primed macrophages (Beutler et al., 198.Sa). If macro­ phages are primed first with gamma interferon they can be stimulated to produce TNF-a after exposure to very small amounts of LPS (Gifford & Lohmann-Matthes, 1987). Furthermore, LPS has been shown to stimulate interferon-gamma production by human peripheral blood mono­ nuclear cells and mouse splenocytes (Le et al., 1986; Blanchard et al., 1986). Thus LPS can botb trigger TNF-a release by interferon-primed macropbages and induce interferon production, result­ ing in an additive effect on the production of TNF-a. Although macrophages an: considered a 114

396 R. L. Gendron et al.

major source ofTNF-a, production ofTNF-a by large granular lymphocytes (NK phenotype cells) has also been demonstrated (Peters et al., 1986). Exposure to excessive quantities of TNF-o can elicit numerous systemic toxic and lethal effects (Tracey et al .. 1986: Kiener et al., 1988). In this study. we have examined LPS.induced resorption of the mouse conceptus.

Materials and Methods

AlfinNis. Mice were maintained in an animal care facility with 12 h light/dark cycles (lights on 07:00-19:00 h l and were given free access to food and water. Female CBA/J mice were purchased from Jackson Laboratories (Bar Harbor. ME, USA). Female CFW!SW and male DBA/2 mice were purdwed from Charles River (St. Constant. Quebec. Canada). Pregnancies were obtained by housing 4 female mice with I male mouse per cage:. Day of vaginal plus detection was designated as Day I of gestation. Animals were killed by cervical dislocation. · Treatme•ts. For the induction of resorption, pregnant mice received tail vein injections (i.v.) ofO· I pg £. coli(ctSS: 85) lipopolysaccharide (LPS) dissolved in PBS (Sigma No. L-2880. St Louis, MO. USA) on various days of gestatJOn. Treatment with LPS was used alone or in combination with a prior single (i.v.) 0·2 ml injection of sterile anti-asialo­ GM I antibody (aAGM I) (Wako, TX, USA). at a dilution of 1120 {previously shown to decrease the spontaneous fetal resorption frequency in mice; deFougerolles & Baines, 1987). Control groups received (i.v.) o-2 ml injections ofPBS. To suppress the production ofTNF-a (Strieter et al.. 1988). daily injections of pentoxifylline (PXF) were administered to pregnant mice intra peritoneally (i.p.) at a dose of 1·15 mg per mouse per day. PXF was dissolved in PBS vehicle and filter sterilized. PXF was kindly provided by Hoechst Inc. (Sommerville. NJ, USA). Control groups received equal volumes ofPBS (i.p.).

ReSOf'ptioll fiNI tiutH procenm,. Fetoplacental resorption was assessed by viewing the contents of the pregnant uterus. Resorbing uterine contents were characterized by haemorrhage and tissue maceration. CBA/J x DBA '2 rree· nant uteri were removed on Day 10 after LPS treatment on Day 9, embedded and frozen in OCT compound (Miles Scientific, Elkhan. IN. USA) for immunohistochemical analysis (see below). Resorption was assessed 1-2 days after treatment with LPS or PBS. CFW /SW x DBA/2 pregnant mice were used on Day 12 of gestation in the analysis of amniotic fluid for TNF-a. In these pregnancies. amniotic fluid was harvested 2 h after treatment with 25pg LPS or PBS by aspirating the amniotic fluid from the amniotic sac using a 25-gause needle and syringe:. Amniotic fluid was then centrifuged to remove debris and red blood cells and stored at - SO"C until assayed. The 2·h time point was chosen to attain peak TNF-a levels (Beutler et al., J985b). · lmtr1111101ti:stocllmtistry. Cryostat sections of uteri from LPS-treated CBA/J x DBA/2 pregnancies were analysed for the infiltration of NK·lili:e cells utilizing an indirect immunoperoxidase technique with anti-asialo-GM I as pri· mary antiserum as previously described (Gendron &. Baines. 1988). Presnant mice were intravenously treated with 0·1 pg LPS or PBS vehicle on Day 9 of gestation and killed 24 h later. Stained sections were scored for the density of 2 asialo-GMI (AGMI)-positive cells by counting 5-10 high-power fields (0·185mm ) pooled from the mesometrial decidual and ectoplacental cone areas of each fetoplacental unit. All fetoplacental units were counted and used to calculate the mean number of AGM !-positive cells per high-power field. Metria! gland areas were not included in the cell counts. TNF..,_ aNJ'. For the analysis ofTNF-a in amniotic fluid. samples were titrated on actinomycin D-treated L929 cells as previously described (Fisch & Gitrord. 1983) TNF-a was determined by the MTI [3-4,S.ddmetbylthiazo1..2-yl)- 2.5-4iphenyltetrazolium bromide] dye reduction assay using recombinant murine TNF..,_ (4 x )()" uniUipg) (Genzyme, Boston, MA, USA) for creatin& the standard curves (Mosman, 1983). Dye reduction was quantified 011 an EAR 400 AT plate rcaderSLT-Labinstruments(Austria). To confirm that the cytotoxin was iDdccd TNF-a, amniotic ftuid samples were also tested at 1:2 dilutions in the presence of a 1:20 dilution of rabbit-anti-TNF-a antiserum (Genzyme).

Dt#t~IIIUI/ysb. Analysis ofvariance was used to assess frequency data from LPS-induced resorption as a functioa of gestational day. The modulation of LPS-induced resorption frequency by aAGMl treatment. amniotic ftuid TNF-a levels and PXF modulation of resorption frequency were analysed by Student's 1 test. Data were considered significant at a probability level of P < O·OS.

Results

Production ofTNF-a during LPS-induced fetal resorptioa and inhibition by PXF LPS has been shown to trigger tbe release ofTNF-a from primed macrophages (Beutler et al., Q 1985a). Since we could not detect TNF-a in the serum or in supernatants of decidual cell suspen­ sions from LPS-treated pregnant mice, we analysed amniotic ftuid for TNF-a at the earliest time ;,j 115

Lipopolysaccharide induced abortion 397 point (Day 12) at which a sufficient volume of this fluid could be collected. Animals received 25 ~~ LPS to induce maximal release ofTNF-a into the amniotic fluid. Treatment of pregnant mice with 25 ~g LPS on Day 12 of gestation resulted in the appearance of significant amounts ofTNF-a in the amniotic fluid 2 h after treatment (Table 1). The cytotoxic activity in the amniotic fluids from LPS­ treated pregnant mice was shown to be due to TNF-a as neutralization with anti· TNF-a antiserum completely abrogated the bioactivity. TNF-a concentrations in control PBS-treated pregnant mice were below the detection limit of the assay. The dose of LPS used in the analysis of amniotic fluid TNF-a levels (25 ~g on Day 12) resulted in 100% resorption in all animals by 24 b after treatment in separate experiments. There was no detectable increase (above background levels) in TNF-a production in the serum of pregnant mice 2 h after treatment with 25 1'8 LPS on Day 8 or Day 11 (data not shown). Pentoxifylline is a methylxanthine which has been shown to suppress TNF-a m RNA expression and TNF-a production (Strieter er al., 1988). Pretreatment of pregnant mice with PXF on Days 9- 12 followed by treatment with 25 IJ8 LPS on Day 12 resulted in a significant decrease in amniotic fluid TNF-a levels compared with LPS treatment alone (Table 1).

Table I. TNF-a in amniotic fluid of CFW/ SW x DBA{J pregnant mice treated with lipopoly­ saccharide (25 Jl8) and inhibition by pentoxifylliDe

Amniotic TNF-a No. of Treatment/days (U/ml) mic::e

LPS Day 12 37-4 ± 6·7 6 PBS Day 12 <2·0 s LPS Day 12 30·6 ±5-9 LPSDay 12 3 ( +a-TNF-a Ab) <2.0 LPS Day 12 34·4 ±s-o s PXF Days 9, 10. 11, 12 +LPSDay 12 4·4 ± 1.()0' 6 Values are mean ± s.e. •p < 0·05.

LPS-ioduced fetoplaceotal resorptioo varies witb day or gestatioa

Since little is known about the effects of LPS on early pregnancy, we studied the response to LPS as a function of gestational day. A differential effect of LPS on fetoplacental resorption in normal pregnancy depended on the day of gestation on which the LPS was administered. As shown in Fig. I, sensitivity to O·l11g LPS was maximal on Day 8 of gestation (100% resorption) and then declined between Days 9 and 11. The frequencies ofLPS-indued resorption on Days 9 and 10 were not significantly different. Resorption frequencies after control injections of PBS on Days 7 through 11 were 2-Jo/o. Resorptions occurring in response to treatments with O·l.~g LPS on Day 8 were characterized by extensive haemorrhage and complete maceration of the uterine contents within 24 h after LPS administration, but were not accompanied by any other noticeable effects in the treated gravid female mice. Resorptions occurring in response to treatments with 0·1 ~g LPS on Days 7 and 9-11 were not grossly detectable until48 h after LPS administration and appeared to be characterized by a slower time course. Abortion could be induced in later pregnancy, but required higher doses (lo-25 Jig) of LPS. Higher doses in late gestation (Days 15-20) were associated with systemic effects such as severe diarrhoea, pilo-erection and hunched posture. 116

398 R. L. Gendron et al. ] +

-'i~ + oL-----~7~--~~~e======s~==----1~0------,~, Day of gestation

Fag. 1. Mean resorption frequency {±standard error) in CFW/SW ( x DBA 2) pregnant females as a function of administration of 0·1 J.l! LPS (+)or PBS (0) on different days of gestation: 4-9 pregnant animals were studied for each treatment group per day. • P < 0·05 rs PBS control.

LPS-induced fetoplacental resorptioa can be prevented by pretreatment with pentoxifyiUoe Since PXF decreased concentrations ofTNF-a induced by LPS in amniotic fluid. we studied the effect of PXF on LPS-induced early resorption. As shown in Table 2. PXF treatment on Days 5 through 8 of gestation completely prevented the 100% resorption frequency observed for Day 8 LPS treatment alone. Pregnant animals in this group (killed on Day 18 of gestation) appeared to contain viable, healthy fetuses. There was no significant difference in the mean number of implantation sites in PXF-protected pregnancies compared with normal. untreated pregnancies. Pretreatment of pregnant mice with PXF on Days 7 and 8 of gestation decreased the LPS-induced resorption frequency considerably but was less effective than daily PXF pretreatment on Days 5-8 (Table 2).

Table 1. Protection against LPS-induced resorption by pentoxifylline (PXF) in CFW/SW x DBA:2 pregnant mice

Gestational day of treatment

PBS PXF LPS lmplalltation No. of (1-tSm&fday) (0·1 pg) sites % Resorptionst mice

S, 6. 7, 8 8 100" 4 5,6, 7,8 8 JO -·± 0.4 0" 4 7,8 8 H±D-95 18 ± IJ• 4 s. 6, 7, 8 8-6 ± 0.94 6·S ± J.Sb 7 7-7 ± HS 2·2 ± 1·4,. 9 Values are mean ± s.e.m. *Not assessed because or extensive maceration or the uterine contents. tAssessment on Day IIJC"l or Day lP.

The reladou lletweea LPS seosidvity and aslalo GMI'"fOSitive cells 0 Our previous studies bad revealed that spontaneousresorptions in CBA/J x DBA/2 pregnancies were associated with the infiltration or AGMI-positive cells into the decidual--ectoplacental cone 117

Lipopolysaccharide induced abortion 399 junction area of approximately 30°/o of the fetoplacental units from each pregnancy. We suggested that this 30% frequency distribution would change after LPS administration if the effects of LPS were related to AGM I ~positive cells. We therefore measured the frequency of AGM 1-positive cells in the decidual-ectoplacental cone region of CBA/J x DBA/2 pregnant mice treated with LPS. LPS was administered on Day 9 to avoid the complete uterine maceration observed for Day 8 treatment. Treatment with Q-1 J.18 LPS led to a significant increase in the density of AGM !-positive cells in the decidual-ectoplacental cone region of 92% of the fetoplacental units 24 b after treat­ ment (13·0 ± 0·8 cells per high-power field (mean± s.e.) for 39 fetoplacental units from 4 LPS­ treated pregnant mice and 1·9 ± 0· 3 cells per high-power field for 23 fetoplacental units from 3 PBS-treated pregnant mia:). In the PBS-treated pregnant mice, 31% of the fetoplacental units contained AGMI-positive infiltrates, as predicted by our previous studies of spontaneous resorp­ tions (Gendron & Baines. 1988). The infiltrating AGM !-positive cells displayed the morphology of large granular lymphocytes. similar to the resorption-associated NK-Iike cells described previously. Pretreatment of normal pregnant mice with anti· AGM I was effective when given 24 h before LPS administration on Day 9. resulting in significant protection from LPS-induced resorption (Table 3). Pretreatment ·with PBS had no effect on the LPS-induced resorption frequency. Pretreat­ ment with anti-A GM I on Days 7 or 9 followed by LPS on Days 9 or 10, respectively, had little or no effect on the LPS-ind~ resorption frequency.

Table 3. Prevention of LPS-induced resorption by treatment with anti-AGMI in CFW/SW x DBA/2 pregnant mice

Gestational day of treatment

PBS Anti-AGM I LPS •;. Resorptiont No. of (Ool pg) (I!IICall ± s..e.) mice

7 9 37·5 ± 11.0 13 7 9 28·2 ± 18·3 5 8 9 2·5 ± 2-3* 6 9 10 32·8 ± 13-2 9 9 10 28·3 ± 12·6 9

• P < Q-05 compared with PBS alone. t Assessed 2 days after LPS treatment.

Evidence presented in this report indicates that the normal mouse conceptus is exquisitely sensitive to LPS during early gestation. The sensitivity to LPS-induced resorption was maximal on Day 8 of gestation and declined thereafter. Resorption induced by LPS was associated with the appearance ofTNF-u in the amniotic ftuid. The effects of LPS in early pregnancy appeared to be site-specific as no other tissues showed any gross signs of pathology. Treatment of pregnant animals with the TNF-a-suppressing drug PXF d~sed LPS-induced amniotic fluid TNF-u levels and decreased the LPS-induced resorption frequency. Histopathologjcal examination of fetoplacental tissues demonstrated high densities of AGM 1-positive ceUs infiltratin& the decidual-ectoplacental cone area of fetoplacental units of LPS-treated pregnancies. Furthermore, the LPS-induced resorption was prevented by treat!J:lalt with anti-AGMl antiserum. Our results suggest that LPS-induccd resorption may involve fetoplacental necrosis in response to Joc::aJ TNF-u production by primed macropbages or NK. ceUs in the decidual-ectoplacental cone tissue. 118

400 R. L. Gendron et al.

Previous studies on the abortifacient effect of endotoxins on pregnancy in mice have concen· trated on late gestation (Days 10-20) (Zahl & Bjerknes. 1943; Rieder & Thomas. 1960; Chedid et al., 1962; Parant & Chedid, 1964; Gasic et al., 1975). Our results indicate that awindow during early gestation is the most sensitive period for LPS-induced resorption (Days 8-10). The dose of LPS used for the induction of resorption in early gestation was I 0-90% less than doses used in previous studies (Rieder & Thomas, 1960; Parant & Chedid, 1964: Rioux-Darrieulat et al.• 1978). The presence oflow concentrations of LPS associated with bacterial infection therefore may result in the induction of resorption at a critical time in gestation. The mechanism of LPS-induced fetoplacental resorption may be due to the direct effect of TNF-a on the placental vasculature resulting in haemorrhage and necrosis. A similar mechanism for the action of TNF-a has been described in the haemorrhagic necrosis and regression of an established sarcoma (Carswell et al., 1975: Havell et al., 1988; North & Havell, 1988). Both LPS and TNF-a have been detected in the amniotic fluids of pregnant women with preterm labour (Casey et al., 1989). Furthermore, placental necrosis has been demonstrated in rats treated with recombinant human TNF-a (Silen et al., 1989). Minor infections of the genito-urinary tract during a critical period in early gestation may be sufficient to trigger the production of TNF-a and may therefore lead to resorption. TNF-a has been demonstrated to inhibit DNA synthesis and prolifer· ation in trophoblast cells (Hunt. 1989). lt is therefore also possible that TNF-n may act through the inhibition of trophoblast proliferation after LPS treatment. causing disruption of placental development. Pretreatment of pregnant mice with PXF prevented resorption induced by LPS. PXF has been shown to inhibit TNF-a mRNA expression and TNF-a production by primed macrophages (Strieter et al., 1988). and also decreases the activation of neutrophils by LPS. TNF-a and other cytokines (SuUivan et al., 1988; Hammerschmidt et al., 1988). Our previous results demonstrated that the decidual-ectoplacental cone areas of early spontaneously resorbing fetoplacental units show increased total cellular densities indicative of leukocytic infiltration (Gendron & Baines. 1989). The mechanism of PXF protection against LPS-induced resorption may involve inhibition of TNF-n production and suppression of neutrophil function in the fetoplacental unit. The infiltration of NK-like lymphocytes appears to precede spontaneous resorption in CBA J x DBA/2 pregnant mice (Gendron & Baines, 1988). As shown here, LPS administration increases AGM 1-positive cellular infiltrates in the same location of the fetoplacental unit as the AGMI-positive cellular infiltrates found in spontaneous resorption. We have previously demon­ strated that decidual AGM !-positive cellular infiltrates increase sharply between Days 7 and 8 during spontaneous resorption (Gendron & Baines, 1988). Furthermore, spontaneous fetal resorp­ tions are significantly reduced by anti-AGMl treatment on Day 8 (deFougerolles & Baines. 19871. This suggests the existence of a short period during early gestation in which the presence of a population of AGM !-positive cells in the fetoplacental unit could affect fetal survival. Our data showing the maximal effectiveness of anti-AGM 1 treatment on Day 8 (Fig. 1) correlate with this finding. The exact phenotype and role of the resorption-associated AGM 1-positive cells remains to be determined. AGMI-positive cells may participate in direct cell killing through lymphokine acti­ vation. Recent evidence demonstrates that the toxic effects of recombinant interleukin-2 in mice are mediated by AGM I -positive cells (Gately et al., 1988). Fetal resorption can be induced by adminis­ tration of interleukin-2 to pregnant mice (Tezabwala et al., 1989) and this may be related to the activation of decidual AGMI-positive cells. Alternatively, AGMI-positive NK cells may act indirectly through the release of gamma interferon, which primes macropbages for TNF-u produc­ tion (Gifford & Lohmann-Matthes, 1987). Since large granular lymphocyteS (NK cells) are capaf:tle of TNF-a production (Peters et al., 1986), it is possible that infiltrating AGM 1-positive cells may themselves produce TNF-a during the early effector stages of resorption. Since activated macro­ phages are known to express surface AGM1 and are present in decidua (Mercurio et al.• 1984; Matthews et al., 1985), it is possible that some of the cells being depleted by anti-AGM I treatment 119

Lipopolysaccharide induced abortion 401 are of the monocyteimacrophage lineage. Macrophages that have been primed by interferon­ gamma produce and release TNF in response to small doses of LPS that are insufficient to induce TNF production by unprimed macrophages (Gifford & Lohmann-Matthes. 1987). Decidual macrophages may be naturally primed during early pregnancy. This priming could perhaps occur through the production of interferon by early placental tissue (Fowler et al., 1980). Priming of decidual macrophages may explain the exquisite sensitivity to LPS observed on Day 8 of gestation. Treatment with anti-AGM I may therefore partly deplete a population of AGM I -positive activated macrophages that could be involved in the effector stages of fetal resorption. Since granulated metria) gland (GMG) cells have also been shown to be AGMJ-positive (Redline & Lu, 1989), it is possible that this cell population may be involved in LPS-induced resorption. However, GMG cells are not able to kill either classical NK targets or targets sensitive to a TNF-a-related cytotoxin (Parr et al .• 1990; Liu et al .• 1987), suggesting that, if GMG cells are involved in the effector stages of LPS-induced resorption, the mechanism may be a novel one. Our present findings demonstrate that LPS-induced resorption involves an AGMI-positive cellular effector component in addition to the production and release ofTNF-a within the pregnant uterus. These two components may act synergistically in the effector stages of fetal resorption in response to bacterial endotoxin.

We thank R. Farookhi and N. S. C. Van Oers for critical comments and Z. Ali-Khan for financial support. R. G. was supported by the Fonds de Recherche en Saute du Quebec.

Refereoces Beutler, 8., Greenwald. D., Hulmes, J.D., Qua, M., Gask.,G.J.,Gask, T.B . .tStrauss,J.f.,lll(l97S)Aborti· Pao. Y-C., Mathisoa, J .. L'lnitda, R. .t Cerami, A. facienl effecl5 of Vibrio cholercw exo-enlerotoxin and (1985a) Identity of tumor necrosis factor and the endotoxin in mice. J. Reprod. Fen 45.315-322 . . macrophage-secreted factor cachectin. Nature, Lond. Gately, M.. AlldersoD. T.D . .t Hayes. T.J. (1988) Role of 316,552-554. asialo-GMI positive lymphoid cells in mediating the Beutler, B.A., Milsark. I.W. & Cerami, A. (198Sb) toxK: effecl5 ofrecombinantll-2 in mice. J.lmm1111ol. Cach~tin/lumor necrosis factor: production. distri- 141, 189-200. bution. and metabolic fate in vivo. J. lmm1111ol. 135, Gendroo. R.L .t Baines, M.G. (1988) Infiltrating 3972-3977. decidual natural killer cells are associated with Blandlarcl, D.K., Djeu, J.K., KltiD. T. \\'.., Friedmao. H• .t spontaneous abortion in mice. Cell. lmm1111ol. 113, Stewan. W.£.,11 (1986) Interferon-gamma induction 261-267. by lipopolysaccharide: dependence on interleuldn 2 Gendroo. R.L .t Baioes. M.G. (1989) Morphometric and macrophagcs. J. /mmunol. 136, 963-970. analysis of the histology of spontaneous fetal CIIJ'SWell. E.A., 0W. W.., Kusel., R.L., C..-. S.., Fiore, rcsorption iD a murine pregnancy. Placenta 10, N. .t WilliUIISOII, B. (1975) An endotoxin-induced 309-318. serum factor that causes necrosis of tumors. Proc. Gifl'oN.. G.E. & l..ollmua-Matthes, M.-L (1987) Gamma natn. Acad. Sci. USA 72. 3666-3670. interferon priming of mouse and human macro- Casey, M.L, Cox. S.M,. Beuder, B.. Milewida, L .t phages for production oftumor necrosis factor pro- MacDooald, P.C. (1989) CachectiDttumor necrosis duction by bacterial lipopolysacdwide. J. Mtn. factor alpha formation in human decidua. Potential Ct111cer hut. '71. 121-124. role of cytokines in infection-induced pretenn labor. Hammersduuidt, D.E., Kotasek, D., McCartlly, T., Hub, J. clin. Invest. 83, 430-436. P.-W., FreyiJar&er, G. .t Ven:eltotd, G.M. (1988) Checlld, L, Bayer, F. & Paraat, M. (1962) Etude de Pentoxifylline inhibil5 granulocyte and platelet fuac- l'action abortive des endotoxines injectees a la souris tion, including granulocyte primina by platelet gravide normale. castree ou bypophsectomisee. Annls ~vatin& factor. J. Lllb. Clin. Med. 112, 254-263. Inst. Pasteur 102,77-84. Havel, E.A., FJers, W• .t Nordt. R.J. (1988) The anti- deFouaerolles, A.R. & Baioes, M.G. (1987) Modulation tumor function of tumor necrosis factor. 1. Thera- of the natural killer ceD activity in p~ant mice peutic action of TNF against an established murine alters the spontaneous abortion rate. J. Reprod. sarcoma is indirect, immunologically dependent, and /mmllll()(. 11, 147-153. limited by extreme toxicity. J. ~xp. Mt!tl. 167, Fisda, H • .t Gifl'ord, G.E. (1983) In vitro production of 1067-IOBS. rabbit ..macrophage tumour ceJI cytotoxin. lnt. J. Huat, J.S. (1989) Cytokine networks in the uteroplacen- Ct~~~cu 32, 105-112. tal unit: macrophages as pivotal regulatol}' cells. J. Fowler, A.K., Reell, c.D. 4 Gfrea, D.J. (1980) ldeatifi- Reprod.IIMfllltlil. 16. 1-17. cation of an interferon in murine placentas. Natur~. ICieaer, P.A., Muel, F., Rodpn. G.. U., P., Warr, G. Lond. 286, 266-267. A Desiderfo, J. (1988) Induction o( tumor necrosis 120

402 R. L. Gendron et al.

factor. IFN gamma. and acute lethality in mice by stimulate production of TNF-alpha but not TNF· toxic and non-toxic fonns ortipid A. J./mmunol. 141, beta (lymphotoxin) by highly purified human 87j}-874 peripheral blood large granular lymphocytes. J. Lt. J.., LiD. J.-X., Hemikseo-DeStefuo, D. & Vikek. J. /mmunol. 13'7, 2592-2598. (1986) Bacteriallipoloysaccharide-induced interferon­ Redlille. R.W. & La, C.Y. (1989) Localization of fetal pmma production: roles of interleukin I and major histocompatibility complex antigens and interleukin 2. J. lmm1111ol. 136. 4525-4530. matemalleukoeytes in murine placenta. Lab. lnres1. U.. C. .C ... Sfdea, M., Kiac, F. & Dia&·E Youaa, J. 61,27-36. (1987) Identification. isolation. and characterization ltie4er,R.F.ano.as.L.(I960)Studiesofthemecbamsm of a novel cytotoxin in murine cytolytic lymphocytes. involved in the production of abortion by endotoxm. Cell SI, 393-403. J. Immunoi.IA. 189-193. Matdlews, c.J ... Adams, A.M. a Searle, R.F. (1985) Riou-Darrieulat.. F., Paraat, M. & Cbeclid. L (19781 Detection of macrophages and the characterization Prevention of endotoxin-induced abortion by treat· of Fe receptor-bearing cells in the mouse decidua. ment of mice with antisera. J. Infect. Dis. 137, 7-13. placenta and yolk sac using the macrophage-specific SOea. M.L., F'll'pO. A., MoraeUo. s .. Lowry, S.F. & monoclonal antibody F4/80. J. Reprod. lmmunol. 7, Ff'IIIICaS, T. (1989) lnterleukin-1 alpha and tumor 315-323. necrosis factor alpha cause placental injury in the rat. ~.A..\t..,SdnriJ'tiD&, G.A. a Robbias,P.W.(I984) Am. J. Pathol. 135, 239-244. Glycolipids of the mouse peritoneal macrophage. J. Strid«, R.M.., Remid, D.G., Ward. P.A., Speacler, exp. Med. 160, 1114-1125. L'li., Lyudt m, J.P., l.anick, J. & Kunkel, S.l... McKay, D.G. & WOD&t T..C. (1963) Effect of bacterial (1988) Cellular and molecular regulation of tumor endoto:ltins on the placenta of the rat. Am. J. Parh. necrosis factor-alpha production by pentoxifylline. 42, 357-372. Bi«hem. Bioph_vs. Res. Comm1111. 155, 12Jj}-J236. M--, T. (1983) Rapid oolorimetric assay for cellular SulliYM, G.W., Carper, H.T., Nowiek Jr., W.J. & growth and survival: application to proliferation and Maaddl, G.l... ( 1988) Inhibition of the inflammatory cytotoxicity assays. J. Imm~~~~ol. Me rh. 65, 55-63. action of interleukin-1 and tumor necrosis· factor Nora. R.J. & Havell., E.A. (1988) The antitumor action (alpha) on neutrophil function by pentoxifylline. of tumor necrosis factor (TNF). 11. Analysis of the 111/ect./mmun. 56, 1722-1729. role of endogenous TNF in endotoxin-induced Tez:Uwala, 8.1: ... Jolmsoa. P.M. & Rees. R.C. (1989) hemonhagic necrosis and regression of an established Inhibition of pregnancy viability in mice following sarcoma. J. exp. Med. 167, 1086-1092. 1L·2 administration. lmm1111ology61, 115-119. Parut. M. A Cbeclid.l... (1964) Protective effect ofchlor· Traa!J, K.J., Beuder, B., Lowry, S.F.., Men')'Weadter, J ... ·promazine against endotoxin-induced abortion. Wolpe, S., Milsark. I.W., Hariri, R.J., Faltey, T .J., Proc. Soc. exp. Bioi. Med. 116, 906-909. 01, Zetatella., A.., Alhert, J.D., Slain!s, G. T. & Cmlllli. Pur, E.L.; Szary, A. & Parr, M.B. (1990) Measurement A. (1986) Shock and tissue injury induced by human of natural killer activity and target cell binding by cacbcctin. Scinlce. NY 134, 47~74. mouse metria! gland cells isolated by enzymic or ZlliL P.A. & Bjerbes, C. (1943) Induction of ~ecidua­ mechanical methods. J. Reprod. Fert. 81, 283-294. placental hemorrhage in mice by the endotoxins of Petm., P.M.. Ortalclo. J.R., Shalaby, M.R., Sveclersky, certain gram-negative bacteria. Proc. Soc. exp. Bioi. LP... Ntftin, G.E., IJriDpuua. T .S., Halk, P.G., A/«1. 54, 329-332. Agll"ltal. 8.8., Hertlermaa. R.B., Goe*l, D.V. 11 Pllll.ldiaD. M.A. (1986) Natural killer sensitive tarsets Rec:eil'ed 3 Jaaaary 1990 121 c Expression of tumor necrosis factor-alpha in the developinq nervous system

R.L. Gendron1,3, P.P. Neste11 , w.s. Lapp1 and K.G. Baines2 •

Department of Physiologyl and Department of Microbiology & Immunology2 , McGill University. Lyman Duff Medical Sciences Building, 3775 University St. Montreal, Quebec, Canada,

H3A 2B4.

3 To whom correspondence should be addressed

Key Words: TNF-a, Development, Nervous system

Acknowledqments We thank R. Farookhi, A. Peterson, S.T. Thomas and R. Head for critical comments and M. Ratcliffe, & A. Shrier for providing avian tissue. we also thank o.w. Blaschuk for use of his photographic equipment. R.L. Gendron is the recipient of a F.R.S.Q Bourse de Formation doctorat. This work was supported by a grant from the Medical Research Council of Canada. 0 122 c ABSTRACT

we present evidence that tumor necrosis factor alpha (TKF­

a) is transientl~ expressed at specific times during embryogenesis in precisely defined areas of the nervous system in two different classes of vertebrates. In murine embryos, TKF-a was detected in the brain, neural tube and peripheral mixed spinal nerves. In the chick embryo, TKF­ a was observed in the brain neuroepithelium and in the developinq Purkinje neurons of the cerebellum. Western immunoblot analysis revealed that brain tissue from both mouse and chick embryos contained a so kDa protein showinq immunoreactivity with ant-TNF-a antibody. These results sugqest that TNP-a participates in the normal development of the vertebrate brain and spinal cord.

0 - 123 INTRODUCTION

Tumor necrosis factor alpha (TNF-a) is a protein produced by activated macrophages, T cells and astrocytes that mediates tumor cytotoxicity, septic shock and cachexia (Carswell et al., 1975; Kinkhabwala et al., 1990; Lieberman et al., 1989; Beutler and Cerami, 1986). In addition to its role in pathophysiology, TNF-a has also been implicated in mediating cell growth, differentiation and tissue remodeling in the adult (Vilcek et al., 1986; Takeda et al., 1986; Vlassara et al., 1988). A recent study has shown that TNF mRNA and a 26 kDa TNF- a immunoreactive protein is present in normally developing mouse conceptus tissue (Ohsawa and Natori, 1989). In light of its various effects as a cytokine, it is conceivable that TNF-a may serve as a physiological mediator in normal embryonic development. In this study, we describe the spatial and temporal distribution of TNF-a during development of the normal murine and avian embryo by immunohistochemistry utilizing a rabbit anti-mouse TNF- a antibody.

MATERIALS AND METHODS

0 124

Immunohistochemistry Air dried, acetone fixed cryostat sections of snap­ frozen embryos from CFW/SW X DBA/2 outbred murine pregnancies or incubated fertilized chicken eggs were stained for the localization of TNF-a utilizing a rabbit anti-murine-TNF-a primary antibody generated against purified recombinant mouse tumor necrosis factor-alpha (Genzyme) in an indirect immunoperoxidase procedure as previously described (Sternberger, 1979). In murine pregnancies, the day of vaginal plug detection was designated day 0 of gestation, while day 0 of incubation was the day on which· fertilized chicken eggs were placed in a 38°C rocking incubator. Negative specificity controls were incubated with either primary anti-murine­ TNF-a antiserum that was preabsorbed with recombinant, purified mouse TNF-a (Genzyme) prior to immunohistochemistry or with normal rabbit serum.

Western Immunoblot Analysis Equivalent weights of brain samples from mouse or chick embryos and mouse spleen were homogenized in distilled water at 4°c (weight to volume ratio of 1:5). After centrifugation at 20,000g, an equal volume of 2X sodium dodecyl sulphate (SDS) sample buffer was added to aliquots 125 c of the supernatants. Samples were then boiled for 3 minutes. Approximately 10 ug of protein per sample was loaded onto a 50S-polyacrylamide gel (15%), and electrophoresed. Proteins were transferred to PVC­ Imobilon filter paper (Millipore), blocked with gelatin and probed for 1 hour with rabbit anti-murine-TNF-a (Genzyme) or normal rabbit serum as negative control. Following washes, the paper was incubated with alkaline phosphatase conjugated goat-anti-rabbit IgG (Bio-rad) for 1 hour. The reaction was then developed using nitro blue tetrazolium as chromogen.

RESULTS

Expression of TNF-a in The Developing Brain Tumor necrosis factor-alpha was present in the developing brain tissue of mouse embryos beginning on day

10 of gestation, and was not observed on days 8 and 9. TNF-a appeared in the forebrain, midbrain and hindbrain regions. The distribution of TNF-a in transverse sections of day 10 mouse embryonic brain formed a radial pattern with the heaviest expression occurring in the ependymal

and outer aspects of the neuroepithelium (figure la). An identical distribution of TNF-a was also observed in the 0 126

brain neuroepithelium of chick embryos on day 8 of incubation (data not shown). By day 15 in the mouse embryo, TNF-a was observed only in hindbrain regions, and consisted of a columnar pattern of TNF-a positive cells distributed tangentially to the hindbrain vesicle (figure lb). In chick embryonic brain on day 16 of incubation, TNF-a was present in what clearly appeared to be Purkinje neurons within the cerebellum. TNF-a expression in putative Purkinje cells appeared to be cytoplasmic and appeared to extend along the dendrites radiating into the molecular layer of the cerebellar lobes (figure le). Preabsorption of the primary anti-murine-TNF-a antiserum with recombinant TNF-a completely inhibited staining of the brain tissue in the mouse and chick embryos, confirming the specificity for TNF-a (figure ld,e,f). No staining of embryonic brain tissue was observed when sections were incubated with normal rabbit serum as primary antibody.

Expression of TNF-a in the Developing Spinal Cord TNF-a could not be detected in early neural tube tissue in mouse embryos on days 7-9 of gestation. However, beginning on day 10, TNF-a was present in the outer and ependymal aspects of the entire length of the neural tube 0 127

epithelium (not shown). The area of TNF-a expression in the spinal cord diminished after day 11, becoming progressively localized to the ventral aspect of the neuroepithilium by day 12 (as shown in figure 2). TNF-a was not present in the spinal cord on day 15 nor at stages between day 15 to term.

Expression of TNF-a in the Developing Mixed Peripheral Spinal Nerves. Transverse sections of the spinal column of mouse embryos revealed that TNF-a was selectively expressed in peripheral mixed nerve bundles beginning on day 11 (figure 2). TNF-a was not observed in the dorsal or ventral spinal nerve ganglia. At higher magnification, TNF-a was observed to be associated with individual nerve processes within the peripheral mixed nerve bundles. Oblique sections through the spinal column confirmed that TNF-a was associated with all the paired peripheral spinal nerves extending ventrally away from the spinal cord (figure 2c). The expression of TNF-a in peripheral spinal nerves was diminished on day 13, and was absent by day 15. Preabsorption of the primary anti-murine-TNF-a antiserum with recombinant TNF-a inhibited staining of the spinal cord tissue, confirming that the reaction was specific for 0 128

TNF-a (figure 2b). Furthermore, no staining of embryonic spinal cord tissue was observed when sections were incubated with normal rabbit serum as primary antibody.

Western Blot Analysis of Embryonic TNF-a Western blot analysis of mouse and chick embryonic brain tissues revealed the presence of anti-TNF-a immmunoreactive protein bands showing a molecular mass of approximately 50 KDa (figure 3}. Two distinct anti-TNF-a immunoreactive bands appeared in mouse embryonic brain tissue. Control lanes loaded with either recombinant TNF- a or with homogenates of splenic tissue from mice treated with gamma-interferon and lipopolysaccharide (LPS) showed TNF-a immunoreactive bands with molecular masses of 17 kDa and approximately 50 kDa, respectively. Identical blots incubated with normal rabbit serum showed no stained protein bands.

orscussroN

These results demonstrate that TNF-a is spatially and

temporally distributed in a specif~c manner in the developing nervous systems of normal mouse and chick embryos. The presence of TNF-a was observed in early 0 129 murine and avian neuroepithelium, developing murine spinal nerves and cerebellar PUrkinje-like neurons in the chick embryo. We propose that the transient expression of TNF-a during embryogenesis may participate in events mediating the development of specific components of the central and peripheral nervous systems. In the present investigation we describe that the molecular mass of TNF-a found in both mouse and chick embryonic brain tissue is approximately 50 kDa. Whereas a single anti-TNF-a immunoreactive band was seen in chick brain, two similar yet clearly distinct bands were observed in the mouse. The cleaved form of TNF-a has a molecular mass of 17 kDa and also associates into multimers. Membrane associated TNF-a in activated macrophages has a molecular mass of 26 kDa while that observed in murine CTLL cells and activated T lymphocytes is 50-60 kDa (Carswell et al., 1975; Luettig et al., 1989;

Liu et al., 1989; Kinkhabwala et al., 1990). We were not able to clarify whether the TNF-a found in the developing nervous system is an integral membrane protein or is receptor bound. The presence of TNF-a in both murine and avian embryos suggests an evolutionarily conserved requirement for this protein in fetal nervous system development. 130

The production and release of TNF-a in macrophages and astrocytes is induced by bacterial cell wall components (Carswell et al., 1975; Lieberman et al., 1989). Since the developing embryo is not exposed to enteric bacteria, it seems unlikely that bacterial products are involved in triggering the expression of TNF-a in the embryo. It would therefore appear that the cells involved and the signals required for fetal TNF-a production are different from those in the adult. An identical pattern of TNF-a expression was detectable in early brain tissue of both mouse and chick embryos, although putative Purkinje cells were found to be TNF-a immunoreactive only in the chick. It is possible that the appearance of TNF-a in Purkinje cells of the chick and not the mouse embryo is related to the precocious nature of the former. This would suggest that the appearance of TNF-a in Purkinje cells may be related to their earlier functional maturation in the chick. The regulation of embryonic TNF-a may be similar to the developmental regulation of proto-oncogenes in the nervous system (Wilkinson, 1987; Wiestler and Walter, 1988). Control of gene expression of many proto-oncogenes and cytokines including TNF-a is mediated by the derepression of a J' untranslated TTATTTAT element allowing for c ·- 131 transient expression following an appropriate stimulus (Han et al., 1990). Elucidation of the regulation of TNF­ a in the developing nervous system may help to define other mechanisms by which the expression of this plieotrophic protein is regulated. 132

~iqure 1 TNF-a expression in frozen cryostat sections of the brain of normal mouse and chick embryos. a, Expression of TNF-a in a sagittal section through the forebrain (Fb) of a mouse embryo on day 10, XlOO, phase contrast. b, Expression of TNF-a in a hindbrain region of a mouse embryo on day 15; XlOO. Anterior (A) and posterior (P) directions show that columns of TNF-a positive cells occur tangentially to the brain vesicle. c, Expression of TNF-a in putative Purkinje neurons of the cerebellum of a chick embryo on day 16 of incubation, X400, phase-contrast. Note TNF-a positive dendrites (d) radiating into the molecular layer (M) of the cerebellar lobe. d,e,f, Serial section negative controls from tissue shown in a, b, c, respectively, stained with rabbit anti­ murine-TNF-a preabsorbed with recombinant murine TNF-a as described in methods. (n= 4 pregnant mice or 4 chick embryos per gestational or incubation time point,

respectively).

0 133

Figure 2 TNF-a expression frozen cryostat sections in the spinal cord and peripheral spinal nerves of normal mouse embryos on day 12 of gestation. a, TNF-a expression in mixed peripheral spinal nerve bundles on day 12 in a transverse section through the spinal column, lightly counterstained with hematoxylin to illustrate the location of the spinal cord (SC) & notocord (NC), XlOO. c, TNF-a expression in peripheral mixed spinal nerves in an oblique section through the spinal cord region of a day 12 embryo, illustrating segmented peripheral nerve bundles which were always laterally paired (pairing not shown in photo), X100. b, Serial section negative control from tissue shown in a, stained with rabbit anti-murine-TNF-a preabsorbed with recombinant murine TNF-a as described in methods.

0 134

Figure 3 Western immunoblot analysis of TNF-a in normal embryonic mouse and chick brain. Lane 1, 5000 units of recombinant murine TNF-a. Lane 2, Mouse embryonic brain on day 15 of gestation. Lane 3, Chick embryonic brain on day 8 of gestation. Lane 4, Normal mouse splenic tissue. Lane 5, Splenic tissue from a mouse intravenously injected 4 hours previously with 10ug lipopolysaccharide (E Coli; 0111:B4, Calbiochem) and 24 hours previously with 60,000 units of recombinant murine gamma interferon (Holland Biotechnology). Identical blots incubated with normal rabbit pre-immune serum showed no staining.

0 135

Beutler, B., Cerami, A. (1986). cachectin and tumor necrosis factor as two sides of the same biological coin. Nature 320, 584-588. Carswell, E.A., Old, L.J., Kassel, R.L., Green, s., Fiore, N. and Williamson, B. (1975). An endotoxin-induced serum factor that causes necrosis of tumors. Proc. Natn. Acad. Sci. U.S.A. 72, 3666-3670. Han, J., Brown, T., Beutler, B. (1990). Endotoxin­ responsive sequences control cachectinjtumor necrosis factor biosynthesis at the translational level. J. Exp. Med., 171, 465-475. Kinkhabwala, M., Sehajpal, P., Skolnik, E., Smith, D., Sharma, V.K., Vlassara, H., Cerami, A., Suthanthiran, M. (1990). A novel addition to the T cell reperatory. Cell surface expression of tumor necrosis factorjcachectin by activated normal human T cells. J. Exp. Med. 171, 941-946. Lieberman, A.P., Pitha, P.M., Shin, H.S., Shin, M.L. (1989). Production of tumor necrosis factor and other cytokines by astrocytes stimulated with lipopolysaccharide or a neurotropic virus. Proc. Natn. Acad. Sci. U.S.A. 86, 6348-6352. Liu, C-C., Detmers, P.A., Jiang, S., Ding-E Young, J. (1989). Identification and characterization of a 0 136 c membrane-bound cytotoxin of murine cytolytic lymphocytes that is related to tumor necrosis factorjcachectin. Proc. Natl. Acad. Sci. U.S.A. 86, 3286-3290. Luettig, B., Decker, T., Lohmann-Matthes, M.-L. (1989). Evidence for the existence of two forms of tumor

necrosis factor: An integral protien and a molecule attached to its receptor. J. Immunol. 143, 4034- 4038. Ohsawa, T., Natori, s. (1989). Expression of tumor necrosis factor at a specific developmental stage of mouse embryos. Dev. Biol. 135, 459-461. Sternberger L.A. (1979). Immunocytochemistry. 2nd ed. N.Y. John Wiley & Sons. Takeda, K., Iwamoto, s., Sugimoto, H., Takuma, T., Kawatani, N., Hoda, M., Masaki, A., Morise, H., Arimura, H., Kono, K. (1986). Identity of differentiation inducing factor and tumor necrosis factor. Natura 323, 338-340. Vilcek, J., Palombella, V.J., Henriksen-DeStefano, D., swenson, c., Feinman, R., Hirai, M., Tsujimoto, M. (1986). Fibroblast growth enhancing activity of tumor necrosis factor and its relationship to other polypeptide growth factors. J. Exp. Med. 163, 632- 0 137 c 643. Vlassara, H., Brownlee, M., Manoque, K.R., Dinarello, D.A., Pasagian, A. (1988). Cachectin/TNF and IL-l induced by gluc·ose-modified proteins: Role in normal tissue remodeling. Science 240, 1546-1548 ( 1988) •. Wiestler, O.D., Walter, G. (1988) Developmental expression of two forms of pp6oc-src in mouse brain. Mol. Cell. Biol. 8, 502-504. Wilkinson, D.G., Bailes, J.A., McMahon, A.P. (1987) Expression of the proto-oncogene int-1 is restricted to specific neural cells in the developing mouse embryo. Cell 50, 79-88. 138 •

.. .' . Fb I

Te · ·.

fig la

• 142

' -'. ·...... ,.

•

•

le

• • 143

• lf • 144

fig 2a

• • 145

• 2b • 146

2c

• • 147

1 2 3 4 5

fig 3

• 148

CBAPrBR V. GENERAL CONCLUSIONS

The contributions made by this thesis centre on three general findings which support the view that non- specific uterine immune reactions are determinants of fetal survival in mice. The first is our demonstration that the predominant mediators of spontaneous resorption are cells of the NK phenotype rather than classical T or B lymphocytes (Gendron and Baines, 1988; 1989; Gendron et al., 1990). The second is our identification of an eicosanoid-mediated NK regulatory activity in the pregnant uterus of normal mice and the deficiency of this activity in spontaneous resorption (Gendron et al., 1990; 1991). The third consists of our contrasting findings on the role of TNF-a in embryonic development and fetal resorption (Gendron et al., 1990; Gendron et al., 1991). Immune reactions determining fetal survival in mice appear to consist largely of innate responses acting in a non-MHC- specific manner (Croy et al., 1985; Bobe et al., 1986; deFougerolles and Baines, 1987; Gendron and Baines, 1988; 1989; Gendron et al., 1990; Chaouat et al., 1989; Lala et al., 1990; Parhar et al., 1990). The recent demonstration that scid-scid/biege-biege mutant mice lacking T, B and NK cells can successfully produce normal-sized litters 149

clearly demonstrates that mechanisms more fundamental than specific immunity are requisite to viviparity (Crepeau and croy, 1990). Our views are in sharp contrast with the those found throughout much of the human clinical liturature that spontaneous fetal loss is related primarily to specific MHC-reactive immunity (Beer et al., 1981: Faulk and Mcintyre, 1986; Mcintyre et al., 1986; Unander et al., 1986; Takakuwa et al., 1986). In order to more clearly define the events mediating fetal survival and resorption in mammals, we utilized the CBA/J X DBA/2 model of spontaneous resorption first documented by Clark and collegues in 1980. Our initial findings from the immunohistochemical analysis of . individual implantation sites revealed that the phenotype of the cellular infiltrate in resorption-prone fetoplacental units of CBA/J X DBA/2 concepti gestating in CBA/J uteri was of the null lineage. The majority of infiltrating NK-like cells are negative for B and T cell markers (Gendron and Baines, 1988; 1989). In addition, NK modulatory treatments such as POLY I:C and aAGM1 which alter the spontaneous resorption frequency (deFougerolles and Baines, 1987) also alter the numbers of NK cells infiltrating CBA/J X DBA/2 concepti (Gendron and Baines, see section II-C). Since monoclonal antibodies to CBA/J 150

NK antigens were unavailable, our initial studies of spontaneously resorbing fetoplacental units relied largely on results obtained with polyclonal antiserum to the AGM1 ganglioside present on mouse NK cells (Kasai, 1980). Other groups have recently characterized uterine GMG cells as being AGM1 positive (Redline and Lu, 1989). GMG cells are migratory and have been hypothesized to regulate excessive trophoblast growth in the early fetoplacental unit (Stewart, 1987). It remains possible that the cells we observed are GMG cells. Thus, further phenotypic analysis with monoclonal antibodies is needed to confirm the actual identity of the NK-like cells associated with spontaneous resorption. The involvement of NK cells in fetal resorption has been confirmed and generalized by Chaouat and collegues who showed that POLY I:C can be used to induce NK-mediated fetal abortion in any number of murine pregnacies (Chaouat et al., 1990). An analysis of the potential effects of NK-modulatory treatments in human spontaneous abortion may be helpful in developing newer and safer methods of preventing this problem. The early transient peak of decidual NK activity at day 6 (Croy et al., 1985; Gembel et al., 1986) present in normal mouse pregnancies suggests that such activity is regulated during early gestation. We hypothesized that 151

the NK regulatory process may be disturbed or attenuated in the resorption-prone CBA/J X DBA/2 pregnancy. our strategy for testing this hypthesis was thus to identify the class of compounds mediating fetoplacental NK suppression and to measure their concentrations in fetoplacental units during which time NK infiltrates are measurable but before the resorption has occurred. studies by Lala et al., (1986; 1990) in which fetal resorption was induced by indomethacin treatment suggested that arachidonic acid metabolites, particularly the E­ series prostaglandins of the cyclooxygenase pathway, may be important in promoting fetal viability. We therefore utilized a rapid small volume procedure developed by Powell in 1984, for solvent extraction of eicosanoids from tissue homogenates with octadecylsilyl silica (ODS) columns. Using this technique, we showed that eicosanoid fraction extracts from normal fetoplacental units contain considerable NK suppressive activity which peaks at day 7-8 of gestation. The identity of the compound(s) responsible for NK suppressive activity was unclear since the activity was present in a partially-purified lipohilic extract of the fetoplacental unit. The NK suppressive activity was, however, partially inhibited by treatment with ASA, suggesting that eicosanoids of the 0 152 cyclooxygenase pathway may be involved. We found that the deficiency of this activity in fetoplacental units from the resorption-prone CBA/J X DBA/2 mating correlates with resorption and NK infiltration (Gendron et al., 1990). Our results support the conclusions made by Croy and collegues (Croy et al., 1985; Gambel et al., 1986) that decidual NK activity is promptly down regulated during early stages of gestation in normal pregnancy. These results also suggested that the deficiency of NK suppression may be an inherent predisposing factor in the CBA/J X DBA/2 fetoplacental unit. Such deficiency in eicosanoid-mediated suppression may be factorial in addition to the other reported anomalies such as MHC antigen sharing in human habitual aborters. Based on our initial findings that suggested spontaneous fetal resorption may be related to a local deficiency of eicosanoids, we investigated the role of two compounds which were likely NK-regulators in the fetoplacental unit. Prostaglandin E2 and leukotriene B4 have been shown to inhibit and augment, respectively, NK activity in vitro, although these findings for LTB4 are controversial (Brunda et al., 1987; Rola-Pleszczynski et al., 1986; Ramstedt et al., 1985). our finding that resorption-prone CBA/J X DBA/2 fetoplacental units 153 contained decreased concentrations of LTB4 and increased concentrations of PGE2 as compared to CBA/J X BALB/c fetoplacental units was surprising. However, these results correlated with increased fetal viability, since CBA/J X DBA/2 second matings showed decreased resorption frequencies and a pattern of LTB4/PGE2 concentration more closely resembling that observed in control CBA/J X BALB/c matings (Gendron et al., 1991: see section III-B). The finding that LTB4 concentrations were significantly decreased in CBA/J X DBA/2 first matings suggested that these matings may be predisposed to NK mediated resorption due to a disturbance in NK-regulatory function. The increased viability and shift toward a normal pattern of eicosanoid concentrations in CBA/J X DBA/2 second matings raises some interesting questions about the counterbalancing of uterine function with paternal incompatibility. our findings suggest that a previous pregnancy, even with a non-compatible male, may result in the memory of a uterine priming event. Wilcox and collegues (1988) present evidence that women with previous spontaneous abortions are more likely to gestate subsequent successful pregnancies. A prior uterine stimulus may therefore elicit local non-specific reactions that could conceivably consist of immunosuppressive 154

eicosanoids. Subsequent in vitro experiments demonstrated that the local concentration of LTB4 may indeed be important in the regulation of NK activity in the pregnant uterus. LTB4 supplementation of extracts from CBA/J X DBA/2 fetoplacental units at concentrations comparable to those measured in control fetoplacental units increased the NK suppressive activity of the extract alone (Gendron et al., 1991: see section III-B). Since the immunomodulatory effects of lipoxygenase metabolites are known to be interactive (Bray and Brahmi, 1986), our finding suggested that LTB4 may act synergistically with other lipoxygenase metabolites in the fetoplacental unit extract. Alternatively, it is conceivable that the CBA/J X DBA/2 fetoplacental unit is generally deficient in lipoxygenase metabolism. Hence, we were prompted to analyze the capacity of fetoplacental tissue from normal and resorption-prone matings to produce such metabolites as the HETE derivatives of the lipoxygenase pathway by HPLC. These further experiments demonstrated that early fetoplacental tissue (day 8) predominantly produces the 12-HETE lipoxygenase metabolite. However, CBA/J X DBA/2 fetoplacental units show no deficiency in the ability to produce 12-HETE as compared to the CBA/J X BALB/c mating. 0 - 155 Our most recent results confirm the interactive role of LTB4 and 12-HETE by demonstrating that the presence of both compounds at critical concentrations in an NK assay in vitro can abrogate the stimulatory activity of either compound alone (Gendron et al., manuscript in preparation) • Fetoplacental resorption in mice occurs rapidly over a course of 24-48 hours (Croy et al., 1982; Clark et al., 1985; Gendron and Baines, 1989). The reaction itself is reminiscent of the TNF-a mediated hemmorhagic necrosis of tumors in mice treated with serum from endotoxin-primed animals initially reported by Carswell and collegues in 1975. We therefore undertook to determine whether TNF-a was involved in the events mediating resorption of the fetoplacental unit. Earlier studies had demonstrated that large doses of bacterial cell wall components such as LPS could induce fetal resorption during late gestation (Zahl

and Bjerknes, 1943; Chedid et al., 1964). Our work extended these findings by demonstrating that the pregnant uterus is extremely sensitive to low doses of LPS early in gestation (Gendron et al., 1990). LPS induced fetal resorption was found to be associated with the release of TNF-a into the amnionic fluid. Furthermore, LPS induced resorption could be prevented by the TNF-a inhibiting drug 0 156 c pentoxifylline. We also found that pentoxifylline was effective in decreasing spontaneous fetal resorption. In CBA/J X DBA/2 pregnant females receiving 1.15 mg pentoxifylline on days 8,9,10 and 11 of gestation, the spontaneous resorption frequency was significantly decreased to 11% versus 37.8% resorption in vehicle treated females (Gendron and Baines, unpublished observation). Casey et al. (1989) have reported the presence of TNF-a in the amnionic fluid of women suffering from infection-associated preterm labor. In addition, Chaouat and collegues recently reported that administration of TNF-a to pregnant mice can induce fetal resorption (Chaouat et al., 1990). Placental injury has also been obtained by administration of TNF-a in pregnant rats (Silen et al., 1989). These results taken together suggest that TNF-a may be an important mediator in the effector stages of fetal resorption. It remains to be determined whether TNF-a exerts its influence directly on trophoblast tissue via cytostasis (Hunt, 1986) or on the fetoplacental vasculature, as in cases of TNF-a mediated tumor regression (North et al., 1989). The exact source of uterine TNF-a produced during resorption also remains unclear. Since the fetoplacental unit contains monocytes 157 and macrophages, the most likely initial source is the maternal immune system. Other sources such as the embryo, as discussed below, may be important at later stages of fetal death. Our finding that TNF-a is associated with the developing nervous system in normal embryos (Gendron et al., 1991) raises the intriguing possibility that TNF-a has dual roles in defence and development in the pregnant uterus. The expression of TNF-a in the developing nervous system commences on day 10 of gestation, at which time spontaneous fetal resorption is first grossly visible. Since we detected released TNF-a in the amnionic fluids of LPS-treated pregnant mice (Gendron et al., 1990), it is likely that the source of this TNF-a is the embryo. Indeed, immunohistochemical staining of the embryos from LPS treated mice reveals that the normal nervous system expression of TNF-a is almost completely absent (Gendron, unpublished observation). This observation may have intriguing implications in light of the fact that LPS has been reported to be teratogenic, with major effects on the developing nervous system. In a 1983 report, Lanning et al. described the effects of low dose LPS administration to pregnant hamsters. In fetuses surviving the endotoxin treatment, the major malformations included spina bifida 158 and exencephaly. It is thus possible that the early damage putatively mediated by TNF-a leads to further lethal disruption of a developmental function of TNF-a in the embryo itself. Further studies are necessary to determine the role TNF-a plays in the early stages of fetal resorption as well as its precise function in the developing nervous system. Sarcophoga lectin has recently been shown to serve both defence and developmental roles at specific stages of sarcophaga peregrina ontogeny. Thus, it may be that other defence-associated cytokines are also expressed during embryonic development in vertebrates. There are some intriguing parallels in gestational chronology throughout the work presented in our studies which may shed light on the timing of early events related to fetal survival. There is a striking correlation in the time of NK infiltration in spontaneous resorption (day 6- 7; section II) and the time during gestation at which NK- depleting treatments are most effective in preventing LPS- induced resorption (day 7; section IV). Furthermore, the maximal expression of NK suppressive activity occurs on day 7 in normal pregnancy (section III). The coincidence of these events suggests that day 7 may be a critical time period for establishing the local response needed to 159

regulate potentially reactive cells in the fetoplacental unit. In conclusion, we have presented evidence that the predisposition of the CBA/J X DBA/2 murine pregnancy to spontaneous fetal resorption can be partially explained by changes in non-specific uterine immune reactions. The CBA/J uterus primigravid with CBA/J X DBA/2 concepti is deficient in the capacity to produce NK suppressive compounds observed in normal pregnant uteri. The deficiency of the primigravid CBA/J X DBA/2 fetoplacental unit to produce lipoxygenase metabolites, particularly LTB4, may lead to a local uterine micro-environment which favors the activation of local NK cells. The cytokine TNF-a appears to be involved in both the effector stages of fetal resorption as well as the normal development of the embryonic nervous system. These findings further our understanding of the innate mechanisms contributing to the

complex immunological relationships determining fetal survival.

-· - 160 CHAPTER VI. REFERENCES

Altaan, D.J., Schneider, S.L., Thompson, D.A., Cheng, H.- L. and Thomasi, T.B. (1990) A transforming growth factor-beta2 (TGF-B2)-like immunosuppressive factor in amniotic fluid and localization of TGF-B2 mRNA in the pregnant uterus. J. Exp. Med. 172, 1391-1401. Aschenqrau, A., Zierler, s. and Cohen, A. (1989) Quality of community drinking water and the occurrence of spontaneous abortion. Arch. Envir. Health 44, 5, 283- 290. Athanassakis, I., Chaouat, G. and Weqmann, T.G. (1990) The effects of anti-CD4 and anti-COS antibody treatment on placental growth and function in allogeneic and syngeneic murine pregnancies. Cell. Immunol. 129, 1, 13-21. Athanassakis, I., Galanopoulos, V.K., Grigoriou, M. and Papamatheakis, J. (1990) Induction of class II MHC antigen expression on the murine placenta by 5-Azacytidine correlates with fetal abortion. Cell. Immunol. 128, 2, 438-449. Badet, M.T., Bell, s.c. and Billington, W.D. (1983) Partial characterization of immunosuppressive factors from short-term cultures of murine decidual tissue. Ann. Immunol. 134, 3231-3237. 0 161

Baines, M.G. and Pross, B.F. (1982) Impairment of the thymus-dependent humoral immmune response by syngeneic or allogeneic pregnancy. J. Reprod. Immunol. 4, 337-348. Baines, M.G., Spears, E.A., Pross, B.F. and Millar, K.G. (1976) Characteristics of the maternal lymphoid response of mice to paternal strain antigens induced by homologous pregnancy. Immunol. 31, 363-369. Barnett, M.A., Learmonth, R.P., Pih1, E. and Wood, E.C. (1983) T helper lymphocyte depression in early human

pregnancy. J. Reprod. Immunol. 5, 55-57. Beer, A.E. and Bi1lingham, R.E. (1974) Host responses to intra-uterine tissue, cellular and fetal allografts. J. Reprod. Fertil. Suppl. 21, 59-88. Beer, A.E., and Billingham, R.E. (1976) The Immunobiology of Mammalian Reproduction. Englewood Cliffs: Prentice- Hall. Beer, A.E., Quebbeman, J.F., Ayers, J.W.T and Haines, R.P.

(1981) Major histocompatibility complex antigens, maternal and paternal immune responses, and chronic habitual abortion in humans. Am. J. Obset. Gynecol. 141, 987-999. Beer, A.E., Quebbeman, J.F., Hamazaki, Y. and Semprini, A.E. (1987) Pregnancy outcome in human couples with recurrent spontaneous abortions: The role(s) of HLA 0 162

antigen sharing, ABO blood group antigen profiles, female serum MLR blocking factors, anti-sperm antibody and immunotherapy. In: lmmunoregulation and Fetal survival, Ed. T.J. Gill, Ill and T.G Wegmann, 286-289. Oxford University Press, N.Y. Beer, A.B., Semprini, A.B., Xiaoyu, z. and Quebbeman, J.P. (1985) Pregnancy outcomes in human couples with recurrent

spontaneous abortions: 1) HLA antigen profiles; 2) HLA antigen sharing; 3) Female serum MLR blocking factors;

and, 4) Paternal leukocyte immunization. Exp. Clin. Immunogenet. 2(3), 1-12. Bell, s.c. and Bi1lington, W.D (1980) Major anti- paternal antibody induced by murine pregnancy is non- complement fixing IgG1. Nature 288, 387-388. Bell, s.c. and Billinqton, w.D. (1983) Anti-fetal alloantibody in the pregnant female. Immunol. Rev. 75, 5- 30.

Benirschke, K. and Rohb, J.A. (1987) Infectious causes of

fetal death. Clin. Obstet. Gynec. 30 1 2, 284-295. Berkowitz, R.s., Hill, J.A., Kurtz, e.G. and Anderson, D.J. (1988) Effects of products of activated leukocytes (lymphokines and monokines) on the growth of malignant trophoblast cells in vitro. Am. J. Obstet. Gynec. 158, 199-203. 0 - 163 Beutler, B., Greenvald, D., Bulmes, J.D., Chanq, M., Pan, Y- c., Mathison, J., Ulevitch, R. and cerami, A. (1985a) Identity of tumor necrosis factor and the macrophage­ secreted factor cachectin. Nature 316, 552-554. Beutler, B.A., Milsark, I.W., and cerami, A. (1985b) cachectinjtumor necrosis factor: Production, distribution, and metabolic fate in vivo. J. Immunol. 135, 3972-3977.

Beutler, B. & Cerami,· A. (1986). Cachectin and tumor necrosis factor as two sides of the same biological coin (1986) Nature 320, 584-588. Billinqton, w.D., Bell, s.c. and smith, G. (1983) Histocompatibility antigens of mouse trophoblast of significance in maternal-fetal immunological interactions. In: Immunology of Reproduction. Ed. Wegmann, T.G. and Gill, III, T.J. Oxford University Press, N.Y. 207-227. Bix, M., Liao, N.-s., Zijlstra, M., Loring, J., Jaenisch, R. and Raulet, D. (1991) Rejection of class I MHC

deficient haemopoeitic cells by irradiated MHC matched mice. Nature 349, 329-331. Blanchard, D.K., Djeu, J.Y., Klein, T.w., Friedman, B. and Stevart II, W.E. {1986) Interferon-gamma induction by lipopolysaccharide: Dependence on interleukin 2 and macrophages. J. Immunol. 136(3), 963-970. Bobe, P., Chaouat, G., stanislavski, M. and Kiger, N. 0 164

(1986) Immunogeneitic studies of spontaeous abortion in mice II. Anitabortive effects are independent of systemic regulatory mechanisms. Cell. Immunol. 98, 477-485. Bray, R.A., Brahmi, z., 1986. Role of lipoxygenation in human natural killer cell activation. J. Immunol. 136:5: 1783-1790.

Bronson, ~.B. (1985) Mammalian Reproduction: An ecological Perspective. Biol. Reprod. 32, 1-26. Bulmer, J.N. and Johnson, P.M. (1984) Macrophage populations in the human placenta and amniochorion. Clin. Exp. Immunol. 57, 393-403. Brunda, M.J., Herberman, R.B. and Holden, B.T. (1980). Inhibition of murine natural killer cell activity by prostaglandins. J. Immunol. 124,(6), 2682-2687. Bulmer, J.N. and Johnson, P.M. (1985) Antigen expression by trophoblast populations in the human placenta and their possible immunological relevance. Placenta, 6, 127-140.

Bulmer, J.N., and Sunderland, C.A. (1984) Immunohistological characterization of lymphoid cell populations in the early human placental bed. Immunology, 52, 349-357.

carswell, E.A., Old, L.J., Kassel, R.L., Green, s., ~iore,

N. and Williamson, B. (1975) An endotoxin-induced serum factor that causes necrosis of tumors. Proc. Nat. Acad. 0 165

Sci. 72(9), 3666-3670. Casey, K.L., cox, s.x., Beutler, B., Kilewich, L., and

KacDonald, P.C. (19~9) Cachectin;tumor necrosis factor alpha formation in human decidua. Potential role of cytokines in infection-induced preterm labor. J. Clin. Invest. 83, 430-436. Chambers, S.P. and Clarke, A.G. (1979) Measurement of thymus weight, lumbar node weight and progesterone levels in syngeneically pregnant, allogeneically pregnant, and pseudo-pregnant mice. J. Reprod. Fert. 55, 309-315. Chaouat, G., Kiger, N. and Wegmann, T.G. (1983) Vaccination against spontaneous abortion in mice. J. Reprod. Immunol. 5, 389-392. Chaouat, G., Kolb, J.-P., Chaffoux, s., Riviera, M., Lankar, o., Athanassakis, I., Green, D. and Wegmann, T.G. (1987) In: Immunoregulation and fetal survival. Ed. Gill III, T.J. and Wegmann, T.G. Oxford University Press,

N.Y., 246. Chaouat, G., Kolb, J.-P., Kiger, N., stanislawski, M. and Wegmann, T.G. (1985) Immunologic consequences of vaccination against abortion in mice. J. Immunol. 134, 1594-1598. Chaouat, G., Menu, B., Clark, D.A., oy, M., Mynkowski, M. and Wegmann, T.G. (1990) Control of fetal survival in 0 166

CBA/J X DBA/2 mice by lymphokine therapy. J. Reprod. Fert. 89, 447-458.

Chedid, L., Boyer, ~., and Parant, M. (1962) Etude de !'action abortive des endotoxines injectees a la souris gravide normale, castree ou hypophsectomisee. Annls. Inst. Pasteur. 102, 77-84. Cheifetz, s., weatherbee, J.A., Tsanq, M.L.-s., Anderson, J.K., Mole, J.E., Luoas, R. and Massaque, J. (1987) The tranforming growth factor-beta system, a complex pattern of ligands and receptors. Cell 48, 409-414. C1ark, D.A., Brier1ey, J., Banwatt, D. and Chaouat, G. (1990) Hormone-induced pre-implantation Lyt 2+ murine uterine suppressor cells persist after implantation and may reduce the spontaneous abortion rate in CBA/J mice. Cell. Immunol. 123, 334-343. Clark, D.A., Chaouat, G, Guenet, J., Kiger, N. (1987) Local active suppression and successful vaccination

against spontaneous abortion in CBA/J mice. J. Reprod. Immunol. 10, 79-85. Clark, D.A., Chaput, A., Tutton, D. (1986) Active suppression of host vs graft reaction in pregnant mice. VII Spontaneous abortion of allogeneic CBA/J X DBA/2 fetuses in the uterus of CBA/J mice correlates with deficient non-T Suppressor cell activity. J. Immunol. 0 167

136, 1668-1675. Clark, D.A., Chaput, A., Walker, c. and Rosenthal, K.L. (1985) Active suppression of host-versus-graft reaction in pregnant mice: VI. Soluble suppressor activity obtained from decidua of allopregnant mice blocks the response to IL-2. J. Immunol. 134, 1659-1664. Clark, I.A. and Chaudhri, G. (1988) Tumor necrosis factor in malaria induced abortion. Am. J. Trop. Med. Hyg. 39, 3, 246-249. Clark, D.A., croy, B.A., Rossant, J. and Chaouat, G. (1986b). Immune presensitization and local intrauterine defences as determinants of success or failure of murine interspecies pregnancies. J. Reprod. Fert. 77, 633-643. Clark, D.A, Palbo, H., Rowley, R.B., Banwatt, D., and stedronska-Clark, J. (1988) Active suppression of host­ vs-graft reaction in pregnant mice IX. Soluble suppressor activity obtained from allopregnant mouse decidua that

blocks the cytolytic effector response to IL-2 is related to transforming growth factor-B. J. Immunol. 141, 3833- 3840. Clark, D.A., Flanders, K.C., Banwatt, D., Miller-Book, w., Hanuel, J., Stedronska-Clark, J. and Rowley, G. (1990) Murine pregnancy decidua produces a unique immunosuppressive molecule related to transforming growth c 168 factor beta-2. J. Immunol. 144, 3008-3014. Clark, D.A., Head, J.R., Drake, B., Palbo, M., Brierley, J., Manual, J., Banwatt, D. and Chaouat, G. (1990) Role of a factor related to transforming growth factor-beta in successful pregnancy. In: Molecular and Cellular Immunology of The Fete-maternal Interface. Ed. Wegmann, T.G. and Gill III, T. Oxford University Press, N.Y. Clark, D.A., McDermott, M.R. and szewczuk. (1980) Suppression of host versus graft reaction in pregnant mice. II Selective suppression of of cytotoxic T cell generation correlates witha soluble suppressor cell activity and with successful pregnancy. Cell. Immunol. 52, 106-118. Clark, D., Slapsys, R.M., croy, A., Rossant, J., 1983. Suppressor cell activity in uterine decidua correlates with success or failure of murine pregnancies. J. Immunol. 131: 540-546.

Cohen, B.L., Orn, A., Gronvik, K.-0., Gid1und, M., Wiqze11, B. and Kurqita, R.A. (1986) Suppression by alpha-fetoprotein of murine natural killer cell activity stimulated in vitro and in vivo by interferon and interleukin 2. scand. J. Immunol. 23, 211-223. crepeau, M.A. and croy, B.A. (1988) Evidence that specific cellular immunity cannot account fro death of Mus 169

Caroli embryos transferred to Mus Musculus with severe combined immune deficiency disease. Transplantation 45, 1104-1110. crepeau, H.A., Yamashiro, s. and croy, B.A. (1989) Morphological demonstration of the failure of Mus Caroli trophoblast in the Mus Musculus uterus. J. Reprod. Fert. 86, 277-288. croy, B.A. and Chapeau, c. (1990) Evaluation of the pregnancy immunotrophism hypothesis by assessment of the reproductive performance of young adult mice of the genotype scidjscis.bgjbg. J. Reprod. Fertil. 88, 231-239. croy, B.A., Gambel, P., Rossant, J., Wegman, T., 1985. Characterization of murine decidual natural killer (NK) cells and their relevance to the success of pregnancy. Cell. Immunol. 93:315-326. Croy, B.A., Reed, N., Malashenko, B.-A., Kim, K. and Kwon, B.s. (1991) Demonstration of YAC target cell lysis by murine granulated metrial gland cells. Cell. Immunol. 133, 116-126. croy, B.A. and Rossant, J. (1987) Mouse embryonic cells become susceptible to CTL lysis after midgestation. Cell. Immunol. 104, 355-361. croy, B.A., Stinchcolb, D., Gough, N.M. and Guilbert, L.J. (1990) Cytokine production is a function of murine 0 170

granulated metrial gland (GMG) cells. Cytokine 1 croy, B.A., Rossant, J. and Clark, D.A. (1982) Histological and immunological studies of postimplantation death of Mus caroli embryos in the Mus Musculus uterus. J. Reprod. Immunol. 4, 277-293. Croy, B.A., Rossant, J., and Clark, D.A. (1985) Effects of alterations in the immunocompetent status of Mus Musculus females on the survival of transferred Mus Caroli embryos. J. Reprod. Fertil. 74, 479-489. croy, B.A., waterfield, A., and wood, w. (1988) Normal murine and porcine embryos recruit NK cells to the uterus. Cell. Immunol. 115, 471-480. croy, B.A., Wood, w., and Rinq, G.J. (1987) Evaluation of intrauterine immune suppression during pregnancy in a species with epitheliochorial placentation. J. Immunol. 139, 1088-1095. Davies M., and Browne, c.x. (1985). Pregnancy-associated nonspecific immunosuppression: kinetics of the generation and identification of the active factors. Amer. J. Reprod. Immunol. Microbiol. 9, 77-83. Daya s., Rosenthal R.L., C1ark, D.A. (1987). Immunosuppressor factor(s) produced by decidua-associated suppressor cells: A proposed mechanism for fetal allograft survival. Amer. J. Obstet. Gynecol. 156, 344-350. 0 171

deFouqerolles, A.R. and Baines, M.G. (1987) Modulation of the natural killer cell activity in preqnant mice alters the spontaneous abortion rate. J. Reprod. Immunol. 11, 147-153. Dorshkind, K., Pollack, s.a.,- Bosma, H.J., and Phillips, R.A (1985) Natural Killer (NK) cells are present in mice with severe combined immunodifiency (scid). J. Immunol. 134, 3798-3801. Drake, B.L. and Head, J.R. (1989) Murine Trophoblast can be killed by lymphokine-activated killer cells. J. Immunol. 143, 9-14. Eades, D.K., Cornelius, P. and Pekal, P.H. (1988) Characterization of the tumor necrosis factor recaptor in human placenta. Placenta 9, 247-251. Earle, u., Wells, M. and Bulmer, J.N. (1985) The expression of major histocompatibility complex antiqens by trophoblast in ectopic tubal pregnancy. J. Reprod.

Immunol. S, 13-24. Edmonds, D.K., Lindsay, K.s., Miller, J.P., Williamson, E., Wood, P.J. (1982) Early embryonic mortality in women. Fertil. Steril. 38, 447-453. Enqelmann, H., Aderka, D., Rubinstein, M., Rotman, D. and Wallach, D. (1989) A tumor necrosis factor bindinq protein purified to homogeneity from human urine protects c 172 cells from tumor necrosis factor toxicity. J. Biol. Chem. 264, 20, 11974-11980. Piddes, T.K., o•Reilly, D.B., cetrulo, c.L., Killer, w., Rudders, R., Osband, M. and Rocklin, R.B. (1986) Phenotypic and fuctional evaluation of suppressor cells in normal pregnancy and in chronic aborters. Cell. Immunol. 97, 407-418. Pisch, H. & Gifford, G.B. (1983) In vitro production of rabbit macrophage tumour cell cytotoxin. Int. J. Cancer 32, 105-112. Theiler, K. (1983) The mouse in Biomedical Research III, Normative Biology, Immunology and Husbandry. (Ed.) Poster, T.K., Small, J.D., and Fox, J.G. pp. 121-136. Toronto, Academic Press. Fowler, A.K., Reed, C.D. & Giron, D.J. (1980) Identification of an interferon in murine placentas. Nature 286, 266-267. Gagnon, L., Girard, M., Su11ivan, A.K., Rola-P1eszczinski, M., 1986. Augmentation of human natural cytotoxic cell activity by leukotriene B4 mediated by enhanced effector­ target cell binding and increased lytic efficiency. Cell. Immunol. 110:243-252. Gambel, P., Croy, B.A., Moore, W.D., Hunziker, R.D., Wegman, T., Rossant, J., 1985. Characterization of immune c 173 effector cells present in early murine decidua. Cell. Immunol. 93:303-314.

Gasio, G.J., Gasio, T.B., and strauss, III, J.~. (1975) Abortifacient effects of vibrio cholerae exo-enterotoxin and endotoxin in mice. J. Reprod. Fert. 45·, 315-322. Gately, M., Anderson, T.D. 'Hayes, T.J. (1988) Role of asialo-GM1 positive lymphoid cells in mediating the toxic effects of recombinant IL-2 in mice. J. Immunol. 141, 189-200.

Gendron, R.L. and ~aines, M.G. (1988) Infiltrating decidual natural killer cells are associated with spontaneous abortion in mice. Cell. Immunol. 113(2), 261- 267.

Gendron, R.L. and Baines, M.G. (1989) Morphometric analysis of the histology of spontaneous fetal resorption in a murine pregnancy. Placenta, 10, 309-318.

Gendron, R.L., ~arookhi, R., Baines, M.G., 1990. Resorption of CBA/J X DBA/2 mouse conceptuses in CBA/J uteri correlates with failure of the fete-placental unit to suppress natural killer cell activity. J. Reprod. Fertil. 89:277-284.

Gendron, R.L., Nestel, ~.P., Lapp, w.s., Baines, M.G., 1990. Lipopolysaccharide induced fetal resorption in mice is associated with the intra-uterine production of tumor 0 174

necrosis factor alpha. J. Reprod. Fertil. Gifford, G.E. and Lohmann-Matthes, M.-L. (1987) Gamma interferon priming of mouse and human macrophages for peoduction of tumor necrosis factor production by bacterial lipopolysaccharide. JNCI, 78, 121-124. Ghayur, T., Seemayer, T.A., Kongshavn, P.A.L., Gartner, J.G., and Lapp, w.s. (1987) Graft-versus-host reactions in the beige mouse. An investigation of the role of host and donor natural killer cells in the pathogenesis of graft-versus-host disease. Transplantation, 44:2, 261- 266. Ghayur, T., Seemayer, T.A., and Lapp, w.s. (1987) Kinetics of natural killer cell cytotoxicity during the graft- versus-host reaction. Relationship between natural killer cell activity, T and B cell activity, and development of histopathological alterations. Transplantation, 44:2, 254-260. Goldsobel, A., Ank, B., Spina, c., Giorqi, J. and Stiehm, E.R. (1986) Phenotypic and cytotoxic characteristics of the immune cells of the human placenta. Cell. Immunol. 97, 335-343. Gronvik, K.-o., Hoskin, D.W. and Murgita, R.A. (1987) Monoclonal antibodies against murine pregnancy associated natural null suppressor cells (NNS) induce fetal 0 175

resorption. Scand. J. Immunol. 25, 533-540. Hakansson, s. and sundquist, K.G. (1975) Decreased antigenicity of mouse blastocysts after activation for implantation from experimental delay. Transplantation 19, 479-485. Hamilton, M.s. and Hamilton, B.L. (1987) Environmental influences on immunologically associated spontaneous abortion in CBA/J mice. J. Reprod. Immunol. 11, 237-241. Hammerscbmidt, D.E., Kotasek, D., KcCarthy, T., Buh, P-w., Freyburger, G., Vercellotti, G.K. (1988) Pentoxifylline inhibits granulocyte and platelet function, including granulocyte priming by platelet activating factor. J. Lab. Clin. Med. 112(2), 254-263. Ban, J., Brown, T. & Beutler, B. (1990). Endotoxin- responsive sequences control cachectinjtumor necrosis factor biosynthesis at the translational level. J. Exp. Med., 171, 465-475. Harper, H.J.K. and Skarnes, R.C. (1972) Inhibition of abortion and fetal death produced by endotoxin or prostaglandin F2-alpha. Prostaglandins 2, 295-309. Havell, E.A., Piers, w.' and North, R.J. (1988) The antitumor function of tumor necrosis factor. I. Therapeutic action of TNF against an established murine sarcoma is indirect,immunologically dependent, and limited c c 176 by extreme toxicity. J. Exp. Med. 167, 1067-1085. Bead, J.R., Drake, B.L. and zuckermann, F.A. (1987) Major histocompatibility antigens on trophoblast and their regulation: Implications in the maternal-fetal relation. Am. J. Reprod. Immunol. Microbial. 15, 12-18. Bead, J., and Gaede, S.D. (1986) Ia antigen expression in the rat uterus. J. Reprod. Immunol. 9, 137-153. Bedley, M.L., Drake, B.L., Head, J.R., Tucker, P.W. and Forman, J. (1989). Differential expression of the class I MHC genes in the embryo and placenta during midgestational development in the mouse. J. Immonol. 142, 4046-4053. Holinka, c.P., Tseng, Y.-c. and Pinch, c.E. (1979) Reproductive ageing in C57BL/6 mice: Plasma progesterone, viable embryos and resorption frequency throughout pregnancy. Biol. Reprod. 20, 1201-1211. Hoskin, D.W. and Murgita, R.A. (1985) Increased maternal T cell autoreactivity associated with syngeneic murine

pregnancy. 3. Reprod. Immunol. 8, 187-196. Hsi, B.-L., Yeh, c.-J. G. and Paulk, w.P. (1984). Class I antigens of the major histocompatibility complex on cytotrophoblast of human chorion laeve. Immunol. 52, 621- 629. Hunt, J.s. (1989) cytokine networks in the uteroplacental unit: Macrophages as pivotal regulatory cells. J. Reprod. 0 c 177 Immunol. 16, 1-17. Jenkinson, E.J. and OWen, V. (1980) Ontogeny and distr­ ibution of major histocompatibility antigens on mouse placental trophoblast. J. Reprod. Immunol. 2, 173-179. Johnson, P.M. and Bu1mer, J.N. (1984) Uterine gland epithelium in human pregnancy often lacks detectable maternal MHC antigens but does express fetal trophoblast antigens. J. Immunol. 132, 1608-1614. Kawaquchi, N., Komano, H. and Natori, s. (1991) Involvement of sarcophaga lectin in the development of imaginal discs of sarcophaga peregrina in an autocrine manner. Dev. Biol. 144, 86-93. Kearns, M., and Lala, P.K. (1982) Bone marrow origin of decidual cell precursors in the pseudopregnant mouse uterus. J. Exp. Med. 155, 1537-1554. Kearns, M., Parhar, R.s., and Lala, P.K. (1985) Decidual cell-specific surface antigens recognized by monoclonal antibodies: Tissue and specific distribution. J. Immunol. 135, 1047-1052. Kiener, P.A., Marek, F., Rodgers, G., Lin, P., warr, G.,

and Desiderio, J. (1988) Induction of tumor necrosis factor, IFN gamma, and acute lethality in mice by toxic and non-toxic forms of lipid A. J. Immunol. 141(3), 870- 874. 0 178 c Kiger, B., Chaouat, G., Kolb, J.P., Weqmann, T.G. and Guenet, J.L. (1985) Immunogenetic studies of spontaneous abortion in mice. Preimmunization of females with allogeneic males. J. Immunol. 134, 2966-2970. Kinkhabwala, M., Sehajpal, P., Skolnik, B., smith, D., Sharma, v.K., Vlassara, B., cerami, A. and suthanthiran, M. (1990). A novel addition to the T cell reparatory. Cell surface expression of tumor necrosis factorjcachectin by activated normal human T cells. J. Exp. Med. 171, 941-946. Kinski, R., Delaqe, G., Rosi, H., Thang, M.-B., Boffmann, M. and Chaouat, G. (1990) A murine model of NK cell mediated resorption. Am. J. Reprod. Immunol. 23, 73-77. Kirby, D.R.s., Billington, W.D. and James, D.A. (1966) Transplantation 4, 713-718. Kita, B., Hamuro, A., Oku, D., Nishikawa, F., Yasui, K., Emoto, M., Kasui, N. and Kashiba, s. (1990) Hormonal regulation of suluble immune response suppressor (SIRS): A possible role of SIRS in the maintanence of pregnancy. Cell. Immunol. 130, 1, 92-105. Koran, H.s., Anderson, s.J., Fischer, D.G., Copeland, c.s. and Jensen, P.J. (1981). Regulation of human natural killing. I. The role of monocytes, interferon and

prostaglandins. J. Immunol. 127(5), 2007-2013. 0 179 c Lala, P.K., Chatterjee-Hasrouni, s., Kearns, M., Montgomery, B., and Colavincenzo, v. (1983) Immunobioloqy of the feto-maternal interface. Immunological Reviews, 75, 87-116. Lala, P.K., Kearns, M. and Parhar, R.s. (1987) Immunology of the decidual tissue. In: Immunoregulation and fetal survival. Ed. Wegmann, T.G. and Gill, T. Oxford University Press, N.Y. 78-95. Lala, P.K., Parhar, R.s., Kearns, M., Johnson, s. and Scodras, J.M. (1986). Immunologic aspects of the decidual response. In: Reproductive Immunology, 1986. Ed. Clark, D.A. and B.A. Croy. Elsevier. N.Y., 190-198. Lala, P.K., Scodras, J.M., Graham, c., Lysiak, J.J., Parhar, R.s., 1990. Activation of maternal killer cells in the pregnant uterus with chronic indomethacin therapy, IL-2 therapy or a combination therapy is associated with embryonic demise. Cell. Immunol. 127, 2, 368-381. Lanninq, J.c., Ri1be1ink, D.R. and Chen, L.T. (1983) Teratogenic effects of endotoxin on the golden hamster. Teratogen. Carcinogen. Mutagen. 3, 145-149. Le, J., Lin, J.-x., Henriksen-Destefano, D. and Vilcek, J. (1986) Bacterial lipoloysaccharide-induced interferon­ gamma production: Roles of interleukin 1 and interleukin 2. J. Immunol. 136(12), 4525-4530. c 180

Lieberman, A.P., Pitha, P.M., Shin, H.S. and Shin, M.L. (1989). Production of tumor necrosis factor and other cytokines by astrocytes stimulated with lipopolysaccharide or a neurotropic virus. Proc. Natn. Acad. sci. u.s.A. 86, 6348-6352. Liu, c-c., Detmers, P.A., Jiang, 8. & Ding-E Young, J. (1989). Identification and characterization of a membrane-bound cytotoxin.of murine cytolytic lymphocytes that is related to tumor necrosis factorjcachectin. Proc. Natl. Acad. Sci. U.S.A. 86, 3286-3290. Luettig, B., Decker, T. & Lohmann-Matthes, M.-L. (1989). Evidence for the existence of two forms of tumor necrosis factor: An integral protien and a molecule attached to its receptor. J. Immunol. 143, 4034-4038. Malathy, P.V., Cheng, H.C., Dey, S.K., 1986. Production of leukotrienes and prostaglandins in the rat uterus during periimplantation period. Prostaglandins 32:605- 614. Marks, s.c. and Lane, P.W. (1976) J. Hered. 67, 11-18. Matsui, s., Yoshimura, N. and Oka, T. (1989) Characterization and analysis of solube suppressor factor from early human decidual cells. Transplantation 47, 4, 678-683. Matt, D.w., Lee, J., sarver, P.L., Judd, H.L. and Lu, 0 181

J.K.B. (1986) Chronological changes in fertility, fecundity and steroid hormone secretion during consecutive pregnancies in aging rats. Biol Reprod. 34, 478-487.

Matthews, c.J., Adams,.A.M. and Searle, R.J'. (1985) Detection of macrophages and the characterization of Fe receptor-bearing cells in the mouse decidua, placenta and yolk sac using the macrophage-specific monoclonal antibody

F4/80. J. Reprod. Immunol. 7, 315-323. Mozntyre, J.A., Page Faulk, w., Niohols-Jobnson, v.R. and Taylor, e.G. (1986) Immunological testing and immunotherapy in recurrent spontaneous abortion. Obstet. Gynecol. 67, 169-175. MoKay, D.G. and Wong, T-c. (1963) Effect of bacterial

endotoxins on the placenta of the rat. Am. J. Path. 42,

357- 372.

Medawar, P.B. (1953) Some immunological and endocrinological problems raised by viviparity in

vertebrates. Symp. Soc. Exp. Biol. 7, 320.

Merourio, A.M., Sohwarting, G.A. and Robbins, P.W. (1984)

Glycolipids of the mouse peritoneal macrophage. J. Exp. Med. 160, 1114-1125.

Millar, K.G., Mills, P. and Baines, M.G. (1973) A study of the infl_uence of pregnancy on the thymus gland of the

mouse. Am. J. Obstet. Gynec. 117, 7, 913-918. c 182

Kitchell, K.D. (1989) Bacterial endotoxin and tumor necrosis factor stimulate prostaglandin production by human decidua. Prostagl. Leukotr. Essent. Fatt. Acids 37, 183-186. Kitchell, K.D., Bdwin, s. and Romero, R.J. (1990) Prostaglandin biosynthesis by human decidual cells: Effects of inflammatory mediators. Prostag. Leukotr. Essent. Fatty Acids. 41, 35-38. Mitcell, K.D., Romero, R.J., Avila, c., Poster, J.T. and Bdwin, s.s. (1991) Prostaglandin production by amnion and decidual cells in response to bacterial products. Prostaglandins, Leukot and Essent Fatty Acids. 42, 167- 169. Mizejewski, G.J. and Vonnegut, M. (1984) Mechanisms of fetal demise in pregnant mice immunized to murine alpha- fetoprotein. Amer. J. Reprod. Immunol. 5, 32-38. Mosman, T. (1983) Rapid colorimetric assay for cellular growth and survival: Application to proliferation and cytotoxicity assays. J. Immunol. Meth. 65, 55-63. Mukhtar, D.D.Y. and stewart, I. (1988) Migration of granulated metrial gland cells from cultured explants of mouse metrial gland tissue. Cell Tiss. Res. 253, 413-417. Murqita, R.A. and Wiqzell, H. (1981} Regulation of immune functions in the fetus and newborn. Prog. Allergy 29, c 183

54-133. Nakayama, B., Shiqetaka, A., Kodo, B. and Miwa, s. (1985). suppression of mixed lymphocyte reaction by cells of human first trimester pregnancy endometrium. J. Reprod. Immunol. 8, 25-31. North, R.J. and Bavel1, B.A. (1988) The antitumor action of tumor necrosis factor (TNF). II. Analysis of the role of endogenous TNF in endotoxin-induced hemorrhagic necrosis and regression of an established sarcoma. J. Exp. Med. 167, 1086-1092. Ober, c., Hauok, w.w., Kostyu, D.D., o•Brian, E.s., Simpson, J.L. and Martin A.o. (1985) Adverse effects of HLA antigen sharing in fertility: A cohort study. Fertil. Steril. 44, 227-232. Ohsawa, T. and Natori, s. (1989). Expression of tumor necrosis factor at a specific developmental stage of mouse embryos. Dev. Biol. 135, 459-461.

Ong Sio, J. (1984) Ph.D. Thesis, University of Texas Health Sciences Center, Dallas. Pakrasi, P.L., Beoka, R. and Dey, s.K. (1985). Cyclooxygenase and lypoxygenase pathways in the periimplantation rabbit uterus and blastocyst. Prostaglandins. 29(3), 481-495. Parhar a.s., Kennedy, T.G. and Lala, P.K. (1988a). c 184

Suppression of lymphocyte alloreactivity by early gestational human decidua. · I. Characterization of suppressor cells and suppressor molecules. Cell. Immunol. 116, 392-410. Parhar R.s., Kennedy, T.G. and Lala, P.K. (1988b). Suppression of lymphocyte alloreactivity by early gestational human decidua. II. Characterization of the suppressor mechanisms. Cell. Immunol. 116, 411-422.

Parhar, R.s., Yaqel, s. and Lala, P.K. (1989). PGE2 mediated immunosuppression by first trimester human decidual cells blocks activation of maternal leukocytes in the decidua with potential anti-trophoblast activity. Cell. Immunol. 120, 61-74. Parant, M. & Chedid, L. (1964) Protective effect of chlorpromazine against endotoxin-induced abortion. Proc. Soc. Exp. Biol. Med. 116, 906-909. Parr, M.B., Parr, E.L., Munaretto, K., Clark M.R. and Dey, s.x. (1988). Immunohistochemical localization of prostaglandin synthase in the rat uterus and embryo during the periimplantation period. Biol. Reprod. 38, 333-343. Parr, E.L., Younq, L.H.Y., Parr, M.B. and Dinq-E Younq, J. (1990) Granulated metrial gland cells of pregnant mouse uterus are natural killer-like cells that contain perforin and serine esterases. J. Immunol. 145, 2365-2372. 0 185 c Pavia, c., Siiteri, P.K., Perlman, J.D. and stites, D.P. (1979) Suppression of murine allogeneic cell interactions by sex hormones. J. Reprod. Immunol. 1, 33-38. Peck, A.B., Murqita, R.A and Wiqzell, H. (1978) Cellular and genetic restrictions in the immunoregulatory activity of alpha-fetoprotein. I. Selective inhibition of anti-Ia associated proliferative reactions. J. Exp. Med. 147, 667-683. Peel, S., Stewart, I. ana Bulmer, D. (1983) Experimental evidence for the bone marrow origin of gra.nulated metrial gland cells of the mouse uterus. Cell Tiss. Res. 233, 647-656. Peters, P.M., orta1ao, J.R., Sha1aby, M.R., sveaersky, L.P., Nedwin, G.E., Brinqman, T.S., Halk, P.G., Aqqarwa1, B.B., Herberman, R.B., Goede11, D.V., ana Palladino, M.A. (1986) Natural killer sensitive targets stimulate production of TNF- alpha but not TNF-beta (lymphotoxin) by highly purified human peripheral blood large granular 1ymphocytes. J. Immunol. 137, 2592-2598. Pollard, J.W., Bartocci, A., Arceci, R., Orlofski, A., Ladner, M.B. and Stanley, E.R. (1987) Apparent role of the macrophage growth factor, CSF-1, in placental development. Nature 330, 484-486. Powell, w.s. (1982). Rapid extraction of arachidonic acid c 186

metabolites from biological samples using octadecy1si1y1 silica. in Methods in Enzymo1ogy. Ed. Lands, w.B. and Smith, W.L. Academic Press, N.Y. Vol. 86, 467-477. Pross, H.P. and Baines, M.G. (1979). Spontaneous human lymphocyte mediated cytotoxicity against target cells. I. The effects of malignant disease. Int. J. Cancer. 40, 113- 119. Psychoyos, A. (1973) Endocrine control of egg implantation. In: Handbook of Physiology. Ed. Greep, R.O., Astwood, E.B., Geiger, S.R. Sect. 7, Vol2, part 2. Bethesda, MD: American Physiological Society, 187-215. Ramstedt, u., Nq, J., Wigzell, B., serhan, c.N. and Samuelsson, B. (1985). Action of novel eicosanoids lipoxin A and B on human natural killer cell cytotoxicity: Effects on intracellular cAMP and target cell binding. J. Immunol. 135(5), 3434-3438. Redman, C.W.G., McMichael, A.J., Stirrat, G.M., Sunderland, C.A. and Tinq, A. (1984). Class I major histocompatibility complex antigens on human extravillous. trophoblast. Immunol. 52, 457-468. Regenstreif, L.J. and Rossant, J. (1989) Expression of the c-fms proto-oncogene and of the cytokine, CSF-1 during mouse embryogenesis. Dev. Biol. 133, 284-294. Redline & Lu, (1989). Localization of fetal major c 187 histocompatibility complex antigens and maternal leukocytes in murin placenta. Lab. Invest. 61, 27-36. Reznokoff-Btievant, H.P., Bdelman, P., Huller, J.Y., Pinon, P. and sureau, c. (1984) HLA-DR locus and maternal-foetal relation. Tissue Antigens 24, 30-34. Rieder, R.P. and Thomas, L. (1960) Studies of the mechanism involved in the production of abortion by endotoxin. J. Immunol., 84, 189-193. Rioux-Darrieulat, P., Parant, M. and Chedid, L. (1978) Prevention of endotoxin-induced abortion by treatment of mice with antisera. J. Infect. Dis. 137(1), 7-13. Robertson, S.A. and seamark, R.P. (1990) Granulocyte macrophage colony stimulating factor (GN-CSF) in the murine reproductive tract: Stimulation by seminal factors. Reprod. Fertil. Dev. 2, 359-368. Rodger, J.c. (1985) Lack of a requirement for a maternal humoral immune response to establish or maintain successful allogeneic pregnancy. Transplantation 40, 4, 372-375. Rola-Pleszczynski, M., Gagnon, L., Boduc, D. and LeBreton, G. (1985). Evidence for the involvement of the thromboxane synthase pathway in human natural cytotoxic cell activity. J. Immunol. 135(6), 4114-4119. Rola-Pleszczynski, M., Gagnon, L. and Sirois, P. (1982). 188 c Leukotriene B4 induces human suppressor lymphocytes. Biochem. Biophys. Res. Commun. 108(4), 1531-1537. Rola-P1eszczynski, M., Gaqnon, L. and sirois, P. (1983).

Leukotriene B4 augments human natural cytotoxic cell activity. 1983. Biochem. Biophys. Res. Commun. 113, 531- 537. Rolfe B.B., cavanagb A.c., guinn K.A., Morton H. (1988). Identification of two suppressor factors induced by early pregnancy factor. Clin. Exp. Immunol. 73, 219-225. Romero, R., Mazor, M., wu, Y.K., Sirtori, M., oyarzun, B., Mitchell, M.D. and Bobbins, J.c. (1988) Infection in the pathogenesis of preterm labor. Seminars in Perinatology 12, 262-279. Rossant, J., Mauro, V.M. and croy, B.A. (1982) Importance of trophoblast genotype for survival ·of interspecific murine chimaeras. J. Embryol. Exp. Morph. 69, 141-149. Rotb, D.B. (1963) The frequency of spontaneous abortions. Int. J. Fertil. 8, 431-434. Rugb, R. and Woh1fromm, M. (1967) The reproductive performance of the laboratory mouse: Maternal age, litter size and sex ratios. Proc. Soc. Exp. Biol. Med. 126, 685- 687. Samuelsson, B., Dablen, S.-B., Lindgren, J.A., Rouzer, C.A., Serhan, C.N., 1987. Leukotrienes and lipoxins: 0 189 c Structures, biosynthesis, and biological effects. Science 237:1171-1176. saarla, R.F., Sallans, K.H., Elson, J., Jankinson, E.J. · a.nd Billinqton, E.J. ( 1976) Detection of alloantigens during pre-iplantation development and early trophoblast differentiation in the mouse by immunoperoxidase labeling. J. Exp. Med. 143, 348-354. scodras, J.K., Parhar, R.s., Jtannedy, T.G •. and Lala, P.K. (1990) Prostaglandin mediated inactivation of natural killer cells in murine decidua. Cell. Immunol. 127, 2, 352-367. Silen, K.L., Firpo, A., Morgello, s., Lowry, s.F. and Francus, T. (1989) Interleukin-1 alpha and tumor necrosis factor alpha cause placental injury in the rat. Am. J. Pathol. 135(2), 239-244. Skarnes, R.c. and Harper, J.Jt. (1972) Relationship between endotoxin induced abortion and the synthesis of prostaglandin F. Prostaglandins 1, 3, 191-203. S1apsys, R., and C1ark, D.A. (1982). Active suppression of host-vs-graft reaction in pregnant mice. IV. Local suppressor cells in decidua and uterine blood. J. Reprod. Immunol. 4, 355-364.

smith, S.K. and Kelly, R.W. (1988). The release of PGp2 alpha and PGE2 from separated cells of human endometrium 0 190

and decidua. Prostaglandins Leukotrienes and Essential Fatty Acids. 33, 91-96. Sokal, R.R., and Rohlf, F.J. (1969) Biometry. San Francisco: W.H. Freeman & Co., pp. 204-246. soulillou, J.P. (1987) Functional characteristics of cells infiltrating rejected allografts. Immunology Today, 8:10, 285-287. Sridama, V., Pacini, F., Yanq, S.-L., Moawad, A., Reilly, M. and DeGroot, L.J. (1982) Decreased levels of T helper cells. A possible cause of immunodeficiency in pregnancy. N. Eng. J. Med. 307, 6, 352-356. stanley-samuelson, D.W., Jensen, B., Nickerson, K.w., Tiebel, K., ogq, C.L. and Howard, R.W. (1991) Insect immune response to bacterial infection is mediated by eicosanoids. Proc. Natl. Acad. Sci. USA 88, 1064-1068. Sternberqer L.A. Immunocytochemistry. 2nd ed. N.Y. John Wiley & Sons (1979). stewart, z. (1988) Granulated metrial gland cells in the non-traumetized regions of the uterus of ovariectomized mice with deciduomata maintained on progesterone. J. Endocr. 116, 11-15. Stewart, I. and Peel, s. (1980) Granulated metrial gland cells at implantation sites of the pregnant mouse uterus. Anat. Embryol. 160, 227-238. 0 191

stewart, I., and Kukhtar, D.D.Y. (1988) The killing of mouse trophoblast cells by granulated metrial gland cells in vitro. Placenta, 9, 417-425. Strieter, R.K., Remiok, D.G., Ward, P.A., Spengler, R.R., Lynch III, J.P., Larriok, J. and Kunkel, s.L. (1988) Cellular and molecular regulation of tumor necrosis factor- alpha production by pentoxifylline. Biochem. Biophys. Res. Commun. 155(3), 1230-1236. sul1ivan, G.w., carper, H.T., Noviok Jr., w.J. and Kande11, G.L. (1988) Inhibition of the inflammatory action of interleukin-1 and tumor necrosis factor (alpha) on neutrophil function by pentoxifylline. Infect. Immun. 56(7), 1722-1729. Surani, K.A.H., Barton, s.c. and Norris, K.L. (1984) Development of reconstituted mouse eggs suggests imprinting of the genome during gametogenesis. Nature 308, 548-550. szekeres-Bartho, J., Hadnagy, J. and Paasa, A.S. (1985) The suppressive effect of progesterone on lymphocyte cytotoxicity: Unique progesterone sensitivity of pregnancy lymphocytes. J. Reprod. Immunol. 7, 121-128. Szekeres-Bartho, J., Szekeres, G., Debre, P., Autran, B. and Chaouat, G. (1990) Reactivity of lymphocytes to a progesterone receptor specific antibody. Cell. Immunol. 0 192

125, 2, 273-283.

~akakawa, x., Goto, s., Haseqawa, I., Ueda, H., Maruhashi,

~., Xajino, ~., Yoshizawa, H., Xanazawa, x. and ~akeuchi, s. (1989) Is immunotherapy for habitual aborters an immunologically hazardous procedure for infants? Am. J. Reprod. Immunol. 19, 53-56.

~akakuwa, x., xanazawa, x. and ~askeuchi, s. (1986) Production of blocking antibodies by vacqination with husband's lymphocytes in unexplained recurrent aborters: The role in successful pregnancy. Am. J. Reprod. Immunol. Microbial. 10, 1-9.

~akeda, x., Iwamoto, s., suqimoto, H., Takuma, T., xawatani, B., Hoda, K., Kasaki, A•, Korise, H., Arimura, H. and Xono, K. (1986). Identity of differentiation inducing factor and tumor necrosis factor. Nature 323, 338-340. Tawfik, o.w., Hunt, J.s. and wood, G.W. (1986) Partial charaterization of uterine cells responsible for suppression of murine maternal anti-fetal immune responses. J. Reprod. Immunol. 9, 213-224. Tezabwala, B.U., Johnson, P.K. & Rees, R.C. (1989) Inhibition of pregnancy viability in mice following IL-2 administration. Immunol. 67, 115-119.

~racy, K.J., Beutler, B., Lowry, S.F., Kerryweather, J., 0 193 c Wolpe, s., Kilsark, I.w., Hariri, R.J., Fahey ~II, T.J., zentella, A., Albert, J.D., Shires, G.T. and Cerami, A. (1986) Shock and tissue injury induced by human cachectin. Science 234, 470-474. Unander, A.K. and Lindholm, A. (1986) Transfusions of leukocyte-rich erythrocyte concentrates: A successful treatment in selected cases of habitual abortion. Am. J. Obstet. Gynecol. 154, 516-520. Van Oers, N.s.c. (1990) The biochemical and immunoregulatory properties of a distinct murine alpha­ fetoprotein isoform. Ph.D. Thesis, Dept. Microbial. and Immunol., McGill University, Montreal, Quebec. Van Vlasselaer, P., Goebels, J. and vandeputte, M. (1986) Inhibition of lymphocyte aggregation by progesterone. J. Reprod. Immunol. 9, lll-121. Vilcek, J., Palombella, V.J., Henriksen-DeStefano, D., swenson, c., Feinman, R., Hirai, K. and Tsujimoto, K. (1986). Fibroblast growth enhancing activity of tumor necrosis factor and its relationship to other polypeptide growth factors. J. Exp. Med. 163, 632-643. Vlassara, H., Brownlee, K., Kanogue, K.R., Dinarello, D.A. and Pasagian, A. (1988). cachectin/TNF and IL-l induced by glucose-modified proteins: Role in normal tissue remodeling. Science 240, 1546-1548 (1988). 0 194 c Weqmann, T.G., Singb, B. and carlson, G.A. (1979) Allogeneic placenta is a paternal strain antigen immunoabsorbant. J. Immunol. 122, 270-273. Weitlauf, B.H. (1988) Biology of implantation. In: The Physiology of Reproduction. Ed. Knobil, E. and Neill, J. et al., Raven Press, N.Y. 231-262. Wbitman, H. and Helton, D.A. (1989) Growth factors in early embryogenesis. Annu. Rev. Cell Biol. 5, 93-117. Wiestler, O.D. and Walter, G. (1988) Developmental expression of two forms of pp6oc-src in mouse brain. Mol. Cell. Biol. 8, 502-504. Wells, H., Bsi, B.-L.and Paulk, W.P. (1984). The expression of class I antigens of the major histocompatibility complex on cytotrophoblastic shell cells of the placental basal plate. Amer. J. Reprod. Immunol. 6:175-181. Wiktor-Jedrzejczak, w., Bartocci, A., Ferrante, A.w., Ahmed-Ansari, A., Sell, K.W., Pollard, J.W. and Stanley, E.R. (1990) Total absence of colony-stimulating factor 1 in the macrophage-deficient osteopetrotic (opjop) mouse. Proc. Natl. Acad. Sci. 87, 4828-4832. Wilcox, A.J., Weinberg, C.R., o•conner, J.F., Baird, D.D., Scblatterer, J.P., canfield, R.B., Armstrong, E.G. and Nisula, B.c. (1988} Incidence of early loss of pregnancy. 195 c N Engl J Med 319, 189-194. Wilkinson, D.G., Bailes, J.A. & McMahon, A.P. (1987) Expression of the proto-oncogene int-1 is restricted to specific neural cells in the developing mouse embryo. Cell 50, 79-88. Zahl, P.A. and Bjerknes, c. (1943) Induction of decidua- placental hemorrhage in mice by the endotoxins of certain gram-negative bacteria. Proc. Soc. Exp. Biol. Med. 54, 329- 332. Zijlstra, M., Bix, M., Simister, M.E., Loring, J.M., Raulet, D.H. and Jaenisch, R. (1990) B2-Microglobulin deficient mice lack CD4-8+ cytolytic T cells. Nature 344, 742-746.