Tumor necrosis factor-a in midtrimester amniotic fluid is associated with impaired intrauterine fetal growth

Kent D. Heyborne, MD: Steven S. Witkin, PhD: and James A. McGregor, MD·

Denver, Colorado, and New York, New York

OBJECTIVE: To investigate whether abnormal immune system activation is involved in the pathogenesis of some instances of impaired fetal growth, we compared tumor necrosis factor-a levels in midtrimester amniotic fluid samples obtained from appropriate-for-gestational-age and small-for-gestational-age pregnancies. STUDY DESIGN: In a case-control study with the sensitive and specific WEHI cell assay, bioactive tumor necrosis factor-a levels in amniotic fluid samples from 24 gestations resulting in small-for-gestational-age infants were compared with levels in 35 samples obtained from gestations resulting in the birth of a term, appropriate-for-gestational-age infant. The two groups were not significantly different with regard to indication for amniocentesis, maternal age, race, smoking history, parity, or other factors. RESULTS: Elevated amniotic fluid tumor necrosis factor-a activity was associated with small-for-gestational-age birth, p = 0.028. With a threshold of 10 pg/ml, elevated amniotic fluid tumor necrosis factor-a had a sensitivity of 48% for the detection of small-for-gestational-age birth, with a specificity of 83%. CONCLUSION: Elevated tumor necrosis factor-a in midtrimester amniotic fluid is associated with impaired intrauterine fetal growth. Abnormal immune system activation, as manifested by increased amniotic fluid tumor necrosis factor-a activity, may mediate impaired fetal growth in some cases. (AM J OBSTET GVNECOL

1992;167:920-5.)

Key words: Tumor necrosis factor-a, intrauterine fetal growth, cytokines

Low birth weight (birth weight <2500 gm) remains the leading cause of perinatal morbidity and mortality worldwide. 1 Although most low-birth weight infants re­sult from preterm delivery, fully one third are term and small for gestational age (SGA, birth weight < 10th percentile for gestational age)'; other infants may be born preterm with impaired growth.

Impaired fetal growth is etiologically diverse. Drug exposure, congenital infection, genetic disorders, and chronic utero placental insufficiency are commonly as­sociated factors. 2 Demographic risk factors include black race, low socioeconomic status, and smoking.' Pla­cental abnormalities, including reduced placental size4

and inadequate invasion of the maternal spiral arteries by trophoblast,' characterize SGA gestations in many cases. Abnormal maternal-fetal immune interaction, triggered by maternal recognition of fetal alloantigens, has been suggested as a factor in such deficient pla­centation. An inflammatory lesion of the placenta,

From the Department of Obstetrics and Gynecology, University of Colorado Health Sciences Center: and the Department of Obstetrics and Gynecology, Cornell University Medical College.' Presented at the Twelfth Annual Meeting of the Society of Perinatat Obstetricians, Orlando, Florida, February 3-8, 1992. Reprint requests: Kent D. Heyborne, MD, University of Colorado Health Sciences Center, 4200 E. 9th Ave., Box B-198, Denver, CO 80262. 616139880

920

termed chronic placental villitis, has been reported in association with SGA gestations, supporting this hy­pothesis. 6 Some SGA infants are considered merely "constitutionally small," with no recognized underlying pathologic condition.

To investigate whether abnormal immune system ac­tivation may be involved in the pathogenesis of some instances of SGA, we conducted a case-control study comparing tumor necrosis factor-a (TNF-a) levels in mid trimester amniotic fluid specimens from pregnan­cies resulting in appropriate-for-gestational-age (AGA) and SGA births. TNF-a was selected for investigation because of its association with impaired cell growth.

Material and methods

Amniotic fluid specimens were collected from Jan­uary 1989 through September 1990 from mid trimester amniocenteses performed in a large, private practice genetics clinic. All patients had private health insur­ance. At the time of amniocentesis demographic and medical information was obtained by experienced ge­netic counselors. Amniotic fluid specimens were cen­trifuged (500 revolutions/min for 10 minutes) to re­move cellular components for karyotype determina­tion, and supernatants were frozen at - 20° C for later TNF-a assay. At the expected date of delivery follow­up questionnaires were sent to all patients. Responses were obtained from 2580 patients, representing an 80%

Volume 167 Number 4, Part I

Amniotic fluid tumor necrosis factor-a and impaired fetal growth 921

Table I. Maternal characteristics and indications for amniocentesis.

Age (yr) Gravidity Parity Race

White Black Hispanic Other

Smoker (> V2 pack/ day) Gestational age at amniocentesis (wk) Amniotic fluid a-fetoprotein (multiples of median) Indication for amniocentesis

Maternal age High maternal serum a-fetoprotein Low maternal serum a-fetoprotein Other

Group 1: SGA (rl = 24)

35.1 2.4 0.5

20 1 I 2 3

16.3 1.12

18 4 I I

Group 2: Control (n = 35)

35.9 2.7 0.9

34 o I o o

16.6 0.94

30 2 I 2

Difference

NS* NSt NSt NS*

NSt NSt NS* NSt

Age, gravidity, parity, gestational age at amniocentesis, and amniotic fluid a-fetoprotein are expressed as means. Characteristics are expressed as numbers of patients. No statistically significant differences exist between the two groups with regards to these characteristics. NS, Not significant.

*By Student t test.

tBy Wilcoxon rank-sum test. tBy Fisher's exact test.

response rate, and study patients were selected from this cohort.

Patient information was stored on a computer data base (Revelation Technologies, New York City) for later retrieval. From this data base a total of 24 specimens was identified as coming from gestations that subse­quently resulted in SGA births (birth weight < 10th per­centile for gestational age as determined by the Colo­rado Intrauterine Growth Curve).7 Thirty-five samples (five samples from each week of gestation at time of amniocentesis, from 14 to 20 weeks) obtained from gestations that resulted in AGA births at term were randomly selected as controls. All specimens used in the study had normal karyotypic analysis, and gesta­tional age as determined by ultrasonography agreed with gestational age as determined by last menstrual period (within 1 week) in all cases. No patient in the study had a history of maternal immune disease; one patient in the SGA group had chronic hypertension.

Bioactive TNF-IX was measured with a sensitive and specific WEHI cell assay, as described. B

.g To determine

if measured TNF -IX bioactivity was attributable to TNF­IX, murine monoclonal antibody to human TNF-IX (ob­tained from Dr. L. Moldawer, Cornell University Med­ical College) was added to undiluted amniotic fluid sam­ples at a final concentration of 20 I-Lgi ml and incubated for 30 minutes at room temperature. All TNF-IX bioac­tivity was neutralized by the anti-TNF-IX antibody. In­traassay and interassay coefficients of variation were 8.5% and 10.1 %, respectively.

The Student t test, Wilcoxon rank-sum analysis, Fish­er's exact test, and logarithmic and logistic regressioll

analysis (SAS, Cary, N.C.) were used, as appropriate, for statistical comparisons.

Results

Tables I and II compare the two groups with regard to maternal characteristics, indications for amniocen­tesis, and pregnancy outcome. Birth weight percentile was calculated for each individual patient with the ges­tation-specific mean and 10th percentile weights from the Colorado Intrauterine Growth Curve,7 assuming a normal distribution. The groups were similar regard­ing maternal factors and amniocentesis indications. Birth weight, birth weight percentile, gestational age at delivery, length, and ponderal index (birth weight x 100 -;- length') were all significantly lower in the SGA group, and pre term delivery and cesarean section were more common in the SGA group.

The distribution of amniotic fluid TNF-IX levels in SGA and AGA pregnancies are presented in Fig. 1. Levels of amniotic fluid TNF-IX in SGA pregnancies (median 10.5 pg/ml) were elevated over those from AGA pregnancies (median 4.0 pg/ml). Comparison by Wilcoxon rank-sum analysis demonstrated that ele­vated mid trimester amniotic fluid TNF -IX is significantly associated with subsequent SGA birth, with p = 0.028. With a threshold of > 10 pg/ml to define an abnormal test, elevated TNF-IX in midtrimester amniotic fluid identifies SGA births with a sensitivity of 48% and a specificity of 83%.

To investigate the relationship of TNF-IX activity in amniotic fluid and preterm birth, the entire study pop­ulation was restratified by preterm birth (gestational

922 Heyborne, Witkin, and McGregor

Table II. Pregnancy outcomes by group

Gestational age at delivery (wk) Infant sex (male/female ratio) Birth weight (gm) Birth weight percentile Birth length (cm) Ponderal index [(Birth weight x 100) I Length3]

Cesarean section Stillbirth Neonatal death Preterm delivery «37 wk)

Group 1: SGA (n = 24)

35.9 10: 14

1812 4.3

45.2 2.16

14 1 2

12

Group 2: Control (n = 35)

39.4 18: 17

3357 61.3 51.6

2.46 10 o o o

October 1992 Am J Obstet Gynecol

Difference

p < 0.01* NSt

p<O.OI* p < 0.01* P < 0.01* P < 0.01* P = 0.030t

NSt NSt

p < O.Olt

The groups were statistically different as indicated. NS, Not significant. *By Wilcoxon rank sum test. tBy Fisher's exact test.

age at delivery <37 weeks, n = 12) or term birth (n = 47). Analysis by Wilcoxon rank-sum analysis failed to reveal a significant difference in TNF-o: levels be­tween these groups, p = 0.3. Similarly, the SGA group was stratified by term versus preterm birth (n = 12 for each group); there was a trend toward elevated am­niotic fluid TNF-o: levels in the term SGA group, p = 0.19.

To evaluate further the relationship between TNF-0: levels and infant birth weight, logarithmic regression analysis was performed. The hypothesis under inves­tigation in this phase of the analysis was whether there is a dose-response relationship between increased am­niotic fluid TNF-o: activity and decreased birth weight. Three outlying values from the control group (all greater than the median + 3.5 SD) were eliminated from the data set for the regression analysis. Possible explanations for these discrepant values are discussed below. Fig. 2 demonstrates the logarithmic regression analysis; to control for the effect of preterm birth, ges­tational age-specific birth weight percentile was used as the dependent variable. This model is significant, p = 0.013, r = -0.38.

Comment This study demonstrates that elevated TNF-o: in mid­

trimester amniotic fluid is significantly associated with impaired intrauterine fetal growth. There is significant overlap in TNF-o: levels between the AGA and SGA groups, and this is to be expected, inasmuch as SGA birth is multifactorial. Some infants may have been small for reasons unrelated to mechanisms investigated in this study, and some SGA infants may be constitu­tionally small without underlying pathologic condi­tions. Elevated TNF-o: levels in some control patients may represent intercurrent maternal illness, technical difficulties in specimen collection or TNF-o: assay (such as endotoxin contamination), or other factors. TNF-o: levels are markedly less than those previously reported

in conjunction with intraamniotic infection 10 but are similar to previously reported levels of bioactive TNF-0: in mid trimester amniotic fluid specimens."

The groups studied were drawn from a cohort of patients attending a private practice genetics clinic with well dated gestations, minimal maternal medical illness (one patient with chronic hypertension), and a low prevalence of demographic risk factors commonly as­sociated with SGA infants. Reliable information re­garding pregnancy complications associated with SGA, such as preeclampsia, is lacking because of the patient­provided follow-up. The only patient-provided data used in this study were date of birth, birth weight and length, infant sex, route of delivery, and occurrence of stillbirth or neonatal death. We considered that this information could reliably be obtained by patient ques­tionnaire.

Measurements of TNF -0: are known to vary mark­edly, depending on the assay technique used,! 1 and most TNF-o: in amniotic fluid is biologically inactive. Because of this, we chose the sensitive and specific WEHI bioassay for our analysis.9 The cause of elevated TNF-o: levels in SGA samples is unknown but may be a marker of abnormal immune system activation. Sub­clinical infection, maternal response to the invading fetal allograft, or disordered cytokine regulation are possible explanations. There are multiple potential sources for TNF -0: found in amniotic fluid. Tropho­blast and decidua have both been shown to produce TNF-0:.10·12 Macrophages present at the maternal-fetal interface are known producers ofTNF-0:. '3 Finally, the fetus itself may contribute to TNF -0: production. I.

TNF-o: may mediate impaired fetal growth in some cases; the logarithmic regression analysis demonstrates a dose-response relationship between TNF-o: activity in amniotic fluid and birth weight percentile, supporting biologic links between TNF-o: and fetal growth beyond the simple association shown by the two-sample infer­ence. Several mechanisms linking TNF-o: and growth

Volume 167 Number 4, Part I

Amniotic fluid tumor necrosis factor-a and impaired fetal growth 923

70~---------------------------------------'

.. .! CL

60

50

E 40 I c ~ 30 .,.

20

10

o 0-5 5·10

35~----------------------------------------~

30

25 .. .. C. E 20 .. .. ~ 15 U) .,.

10

5

o 0-5 5·10 10-15 15·20 20-25 25-30 30-35 35-40 > 40

Amniotic fluid TNF-alpha (pg/ml) Fig. 1. Frequency distribution of amniotic fluid TNF-a levels in AGA and SGA gestations.

924 Heyborne, Witkin, and McGregor October 1992 Am J Obstet Gynecol

100.-------------------------------------,

CD 80 -;: C CD () .. CD a. -.c C)

60

'j 40

== .c 1:: ,-m 20

. ..

.. O~--~~~~--_r----~--_r----~--~--~

o 5 10 15 20 25 30 35 40

Amniotic fluid TNF-alpha (pg/ml)

Fig. 2. Logarithmic regression analysis demonstrating a dose-response relationship between elevated TNF-a levels and decreased birth weight percentile (birth weight percentile = 58.1 - 11.81n [TNF­al; significance level for the modelp = 0.013, r = -0.38).

impairment are possible. TNF-a, also known as cach­ectin, causes wasting in association with sepsis, chronic infection, and cancerl5

; this phenomenon is mediated in part by interference with lipid metabolism. The an­thropomorphic characteristics of infants with intra­uterine growth retardation may reflect TNF-a me­diated cachexia. 16 The SGA group had a significantly lower ponderal index than the control group, sup­porting a component of intrauterine growth retarda­tion in this group. Alternatively, impaired growth may be mediated by abnormal placental function. 17 TNF-a either administered directly or induced by lipopolysac­charide injection has been demonstrated to cause pla­cental hemorrhage and abortion in rabbits, mice, and rats. 18-20 Specific inhibition of TNF-a with pentoxy­phylline inhibits this process, suggesting a specific effect ofTNF-a.20 TNF-a receptors have been demonstrated on trophoblast,21 and TNF-a inhibits growth of tro­phoblast as measured by deoxyribonucleic acid metab­olism and cellular proliferation.22.23 It is interesting that these samples are obtained at a time in gestation when placental development and spiral artery invasion is cul­minating; increased TNF-a activity may be related to disorders of this process.5 TNF-a is directly toxic to endothelium and may damage the decidual vascula­ture.24 Finally, TNF-a interferes with the anticoagulant system and may induce placental thrombosis. 25

In summary, elevated TNF-a activity in mid trimester

amniotic fluid is significantly associated with impaired fetal growth. Multiple mechanisms have been suggested by which TNF-a may mediate some cases of SGA. Fu­ture studies repeating this work, measuring TNF -a and other cytokines within the uteroplacental unit and am­niotic fluid, and correlating these abnormalities with placental pathologic conditions are under way. In­creased understanding of mechanisms involved in im­paired fetal growth may allow for the development of etiology-based prevention and treatment strategies.

We thank George Henry, MD, Reproductive Genet­ics, P.C., Denver, Colorado, for providing amniotic fluid samples and patient information for this study.

REFERENCES I. McCormick MC. The contribution of low birth weight to

infant mortality and childhood morbidity. N Engl J Med 1985;312:82.

2. Creasy RK, Resnik R. Intrauterine growth retardation. In: Creasy RK, Resnick R, eds. Maternal-fetal medicine: principles and practice. Philadelphia: WB Saunders, 1989:547.

3. Shiono PH, Klebanoff MA, Graubard BI, Berendes HW, Rhoads GG. Birth weight among women of different eth­nic groups. JAMA 1986;255:48.

4. Molteni RA, Stys SJ, Battaglia FC. Relationship of fetal and placental weight in human beings: fetal/placental weight ratios at various gestational ages and birthweight distributions. J Reprod Med 1978;21 :327.

5. Sheppard BL, Bonnar J. An ultrastructural study of utero-placental spiral arteries in hypertensive and nor-

Volume 167 Number 4, Part I

Amniotic fluid tumor necrosis factor-a and impaired fetal growth 925

motensive pregnancy and fetal growth retardation. Br J Obstet Gynaecol 1981 ;88:695.

6. Labarrere C, Althabe 0, Telenta M. Chronic villitis of unknown aetiology in placentae of idiopathic small for gestational age infants. Placenta 1982;3:309.

7. Battaglia FC, Lubchenco LO. A practical classification of newborn infants by weight and gestational age. J Pediatr 1967;71:159.

8. Rollinghoff M, Warner NL. Specificity of in vivo tumor rejection assessed by mixing immune spleen cells with target and unrelated tumor cells. Proc Soc Exp Bioi Med 1973;144:813.

9. Espevik T, Meyer N. A highly sensitive cell line, WEHl 164 clone 13, for measuring cytotoxic factor/tumor ne­crosis factor from human monocytes. J Immunol Methods 1986;95:99.

10. Casey ML, Cox SM, Beutler B, Milewich L, MacDonald PC. Cachectin/tumor necrosis factor-a formation in hu­man decidua. J Clin Invest 1989;83:430.

II. Jaattela M, Kuusela P, Saksela E. Demonstration of tumor necrosis factor in human amniotic fluids and supernatants of placental and decidual tissues. Lab Invest 1988;58:48.

12. Yelavarthi KK, Chen HL, Yang Y, Cowley BD Jr, Fishback JL, HuntJS. Tumor necrosis factor-a mRNA and protein in rat uterine and placental cells. .I Immunol 1991; 146:3840.

13. Hunt JS. Cytokine networks in the uteroplacental unit: macrophages as pivotal regulatory cells. J Reprod Im­munoI1989;16:1.

14. Kutteh WH, Rainey WE, Beutler B, Carr BR. Tumor necrosis factor-a and interleukin-l[3 production by hu­man fetal Kupffer cells. AM .I OBSTET GYNECOL 1991; 165: 112.

15. Epstein FH. Mechanisms of disease. N Engl J Med 1987;316:379.

16. Miller JC, Hassanein K. Diagnosis of impaired fetal growth in newborn infants. Pediatrics 1971 ;48:511.

17. Dawes GS. The placenta and foetal growth. In: Dawes GS, ed. Foetal and neonatal physiology. Chicago: Year Book, 1968.

18. Silen ML, Firpo A, Morgello S, Lowry SF, Francus T. Interleukin-Ia and tumor necrosis factor a cause placen­tal injury in the rat. Am J Pathol 1989; 135:239.

19. Zahl PA, Bjerknes C. Effect of the endotoxin of Shigella paradysenteric on pregnancy in rabbits. Proc Soc Exp Bioi Med 1944;56: 153.

20. Gendron RL, Nestel FP, Lapp WS, Baines MG. Lipopoly­saccharide-induced fetal resorption in mice is associated with the intrauterine production of tumour necrosis fac­tor-alpha. J Reprod Fertil 1990;90:395.

21. Aiyer RA, Aggarwal BB. Characterization of receptors for recombinant human tumor necrosis factor-a from human placental membranes. Lymphokine Res 1990;9:333.

22. HuntJS, Atherton RA, PaceJL. Differential responses of rat trophoblast cells and embryonic fibroblasts to cytokines that regulate proliferation and class I MHC antigen expression. J Immunol 1990;145:184.

23. HuntJS, Soares MJ, Lei MG, et al. Products of Ii po poly­saccharide-activated macrophages (tumor necrosis factor­a, transforming growth factor-(3) but not lipopolysaccha­ride modify DNA synthesis by rat trophoblast cells exhib­iting the 80-kDa lipopolysaccharide-binding protein. J Immunol 1989;143:1606.

24. Nawroth PP, Stern DM. Modulation of endothelial cell hemostatic properties by tumor necrosis factor. J Exp Med 1986; 163:740.

25. Bevilacqua MP, Pober JS, M~eau GR, Fiers W, Cotran RS, Gimbrone MAJr. Recombinant tumor necrosis factor induces procoagulant activity in cultured human vascular endothelium: characterization and comparison with the actions of interleukin I. Proc Natl Acad Sci USA 1986;83:4533.