Evaluation of the use of anti-TNF-α in an LPS-induced murine model

1. Introduction 

Preterm birth (<37 completed weeks gestation) is a major unresolved problem in modern obstetrics, affecting 12.5% of live births in the United States. In addition, morbidities associated with prematurity cause more than 70% of fetal and neonatal deaths in the U.S. annually (Arias et al., 2003). Despite considerable research and clinical efforts, the incidence of prematurity has increased by 20% over the past 20 years (Hamilton et al., 2003). Although some of this increase is related to the increased use of assisted reproductive technologies and advances in the care of preterm neonates (Ananth et al., 2005), the condition persists in large part because of an incomplete understanding of the multifactorial pathophysiology of preterm labor.

Many cases of preterm labor are associated with infection and/or inflammation. Infections, both involving the genital tract and other areas of the body, like appendicitis and pyelonephritis, have been associated with preterm birth (Romero et al., 1989). Intra-amniotic markers of infection have been found in up to 50% of cases of preterm birth (Gravett et al., 2004, Fortunato et al., 1996). Activation of inflammatory mediators is the alleged mechanism by which infection causes preterm birth. Indeed, placental inflammation is present in up to 60% of placentas of patients experiencing preterm birth (Hillier et al., 1988). An important inflammatory mediator is tumor necrosis factor α (TNF-α). TNF-α is elevated in the amniotic fluid of women with term and preterm labor in both animal models and human pregnancy (Romero et al., 1992, Romero and Mazor, 1988). In addition, TNF-α increases the production of other inflammatory mediators, matrix metalloproteinases that lead to amnion degradation, and uterotonins such as prostaglandins and endothelin (Crider et al., 2005, Tanaka et al., 1998). Finally, the direct administration of TNF-α causes preterm labor and fetal death in a murine model (Silver et al., 1994).

We demonstrated previously that antibodies against TNF-α can reduce preterm birth in a murine model of infection (Silver et al., 1994). Over the past decade, a commercially available antibody against TNF-α has been used successfully to treat inflammatory conditions associated with TNF-α activity such as Crohn's disease and rheumatoid arthritis (Arnott et al., 2001, Blumenauer et al., 2003). Thus, we sought to determine if this antibody could abrogate the effects of lipopolysaccharide (LPS) in a murine model. Our second objective was to determine whether anti-TNF-α alters the uterine inflammatory response to LPS administration.

2. Methods 

3. Results 

3.1. Obstetric outcomes 

There were 72 resultant pups in the LPS+IgG group, and 91 in the group receiving LPS+anti-TNF-α. Obstetric outcomes are depicted in Table 1. Systemic pretreatment with anti-TNF-α significantly reduced the rate of fetal death, with 25/72 (34.7%) deaths in the LPS+IgG group and only 7/91 (7.6%) in the LPS+anti-TNF-α mice (p<0.01). In the group receiving the LPS+IgG, 6/9 (66.7%) ‘pregnancies’ experienced preterm delivery, with only 2/12 (1.7%) in the group treated with anti-TNF-α (p<0.01).

Table 1. Clinical outcomes

		Control IgG+LPS (n=72)	Anti-TNF-α+LPS (n=91)	p-Value
	Dead fetuses in utero	25 (34.7%)	7 (7.8%)	<0.001
	Alive fetuses in utero	24 (33.3%)	73 (80.2%)	<0.001
	Preterm fetuses delivered	23 (31.9%)	11 (12.1%)	<0.001
	Dams experiencing preterm delivery	6/9 (66.7%)	2/12 (16.7%)	<0.05
	Dams experiencing intrauterine fetal demise	6/9 (66.7%)	2/12 (16.7%)	<0.05

3.2. Inflammatory mediator expression profile 

Gene expression levels for IL-6, TNF-α, Cox-1, Cox-2, IL-10, TLR-2, TLR-4 and IL-1β were analyzed for all groups (Group 1: LPS+IgG; Group 2: LPS+anti-TNF-α; Group 3: saline+anti-TNF-α; and Group 4: saline+IgG). In total, there were 12 animals analyzed per treatment group. A Ct value was computed for each gene of interest and, by utilizing a comparison to GADPH ‘housekeeping gene’ expression, a ΔCt was calculated as was a ΔΔCt. From these, relative copy numbers of each gene analyzed were computed and are displayed in Fig. 1.

View Large Image Download to PowerPoint

Fig. 1. Relative copy number of cytokine genes with administration of IgG/saline, anti-TNF-α/saline, IgG/LPS, and anti-TNF-α/LPS.

When compared to control injections of saline+IgG, there was increased expression of IL-6, IL-1β, TLR-2, CD14, TNF-α and IL-10 in mice treated with LPS+IgG (Fig. 1). Of the genes assayed, expression of IL-6, IL-1β, TLR-2, CD14 and COX-1 were found to be significantly reduced in mice treated with anti-TNF-α and LPS compared to LPS alone (Table 2). The effect was most profound for IL-6 and IL-1β. The expression of TNF-α, COX-2, IL-10 and TLR-4 was not significantly different between the two groups.

Table 2. Inflammatory gene response

	Gene	p
	CD14	0.028*
	COX-1	0.043*
	COX-2	0.106
	IL-10	0.225
	IL-1β	0.001*
	IL-6	0.001*
	TLR-2	0.003*
	TLR-4	0.119
	TNF-α	0.862

Wilcoxon Rank Sum test (a non-parametric version of the 2-sample t-test) comparing ΔCt between IgG+LPS and anti-TNF-α+LPS.

4. Discussion 

There is little doubt that proinflammatory cytokines play an important role in systemic and localized inflammatory-mediated preterm birth. However, the extent to which the expression and clinical effect of these cytokines can be modified is an important question when considering either treatment or prevention of preterm birth. In this study, we have demonstrated that administration of a TNF-α antagonist prior to administration of an inflammatory mediator in mice can significantly decrease both the proportion of fetal deaths and the proportion of pups delivered preterm. However, the antibody did not entirely abrogate the effects of LPS, suggesting that TNF-α plays an important but not obligatory role in the mediation of LPS-induced preterm birth and fetal death.

Our clinical results are similar to other studies of TNF-α blockade in murine models of infection-mediated preterm birth (Silver et al., 1994, Gendron et al., 1990, Xu et al., 2006). However, this is the first study to use an antibody analogous to one that has clinical use in humans. It is noteworthy that not all experiments using TNF-α antagonists ameliorated the effects of LPS (Fidel et al., 1997, Hirsch et al., 2002). Differences in results may be due to different methods of TNF-α blockade, different models and varied antibody specificities. Another factor is the tremendous complexity and redundancy in the cytokine network. Hirsch and colleagues demonstrated that mice lacking IL-1 and TNF-α receptors have reduced preterm birth in response to infection, whereas those lacking IL-1 alone do not (Hirsch et al., 2006). Thus, blockade of multiple cytokines may prove superior to a single cytokine antagonist.

The mechanism of LPS-induced fetal death and preterm birth appears to involve the intrauterine production of inflammatory mediators. LPS up-regulates several inflammatory cytokines in the uterus including TNF-α, IL-1α, IL-1β and IL-6 (Silver et al., 1997, Fidel et al., 1994, Lin et al., 2006). LPS also increases intrauterine prostaglandin production in a murine model, primarily by increasing COX-2 expression (Hirsch and Wang, 2005,). Other inflammatory mediators induced by LPS include Toll-like receptors and CD14 (Elovitz and Mrinalini, 2005, Rath et al., 1998).

TNF-α blockade appears to be protective against LPS-induced fetal loss and preterm birth by decreasing intrauterine production of inflammatory mediators in response to LPS. These effects were most profound for IL-1β and IL-6, two powerful cytokines known to be up-regulated by TNF-α (Sadowsky et al., 2006, Moore et al., 2001) In fact, administration of IL-1α causes murine death in a dose-dependent fashion and anti-IL-1 can also limit the effects of LPS in this murine model. IL-6 is strongly associated with preterm labor with and without intra-amniotic infection, and is intimately tied to both TNF-α and IL-1 (Robertson et al., 2006).

IL-10 production in response to LPS was not significantly affected by antibodies against TNF-α. IL-10 is generally considered to be an anti-inflammatory cytokine, and serves to limit the effects of pro-inflammatory cytokines (Robertson et al., 2006, Pestka et al., 2004). Indeed, IL-10 serves to limit the effect of LPS on murine pregnancy, as demonstrated by an increase in preterm fetal loss and growth restriction as well as an increase in TNF-α and IL-6 in response to low doses of LPS in IL-10-deficient mice (Pestka et al., 2004). This cytokine is influenced by several pathways that do not involve TNF-α (Fujihara et al., 2003).

As expected, we did not find a significant difference in gene expression of TNF-α after LPS administration in the anti-TNF-α treated group. The TNF-α antagonist we used is a chimeric monoclonal antibody that binds TNF-α. Thus, it would be expected to directly inhibit TNF-α. However, the anti-TNF antibody likely does not function as a direct inhibitor of the TNF-α gene.

It was of interest that TNF-α blockade did not significantly decrease COX-2 expression in response to LPS. COX-2 is up-regulated in this murine model in response to LPS and is thought to play a major role in LPS-induced preterm birth/fetal loss. It is likely that COX is regulated by factors other than TNF-α in response to LPS. COX-2 expression, in response to LPS, is reduced in mice lacking receptors for both TNF-α and IL-1. These observations illustrate the tremendous redundancy of the cytokine network. COX-1 expression after LPS administration was not altered by LPS alone, anti-TNF-α or in the TNF-α/IL-1 knockout mice.

Both CD14 and Toll-like receptors are increased in the chorioamniotic membranes in spontaneous labor associated with chorioamnionitis (Muzio et al., 2000, Kim et al., 2004) and both can be up-regulated by inflammatory cytokines. LPS-induced expression of both CD14 and TLR2 were significantly reduced by anti-TNF-α, indicating that TNF-α likely plays a role in this pathway. It is of interest that Elovitz and Mrinalini (2005) demonstrated that medroxyprogesterone acetate can decrease the LPS-induced up-regulation of TLR-2 and TLR-4 mRNA in murine cervix and placenta. In part, this may be due to the ant-inflammatory effects of progesterone acetate. As expected, anti-TNF-α had little effect on TLR-4 expression, since this gene was not up-regulated by LPS in the tissues studied.

Our study has several limitations. First, a murine model was used which may limit the applicability to human pregnancy. Second, we assessed mRNA but not protein levels of inflammatory mediators. Thus, we cannot be certain that protein levels were actually altered by anti-TNF-α. Nonetheless, changes in mRNA correlated well with clinical effects and, thus, likely with protein levels. Third, since whole uterus was used, we cannot determine whether changes in gene expression occurred in myometrial cells, inflammatory cells or both.

These data support further evaluation of TNF-α blockade as a potential treatment, or (more likely) prophylactic agent for preterm labor in humans. Given the abundant data regarding the role of TNF-α in preterm labor, even in the absence of infection, antibodies against this cytokine make an attractive potential therapy. Moreover, Infliximab (a chimeric IgG 1 monoclonal antibody comprised of 75% human and 25% murine sequences, which has a high specificity and affinity for TNF-α) has been used during human pregnancy to treat Crohn's disease and rheumatoid arthritis without evidence of adverse fetal effects (Katz et al., 2004, Chakravarty et al., 2003). In fact, Inflixamab is a pregnancy category B drug. There is too little experience to be confident regarding the safety of Infliximab or other anti-TNF-α agents during pregnancy, and studies proving efficacy are required. Nonetheless, our data provide rationale and support for further research regarding the use of anti-TNF-α as a prophylactic or therapeutic tool against preterm birth.