Main

The chronic sequelae of premature birth place an increasing burden on health care services: BPD is now the major cause of infantile pulmonary fibrosis(1). Although immaturity of pulmonary architecture and deficient surfactant production are thought to be central in initiating neonatal RDS, there is much evidence to suggest that long-term outcome may depend on the extent of inflammatory responses. Infants who develop BPD have a more vigorous inflammatory reaction, with increased numbers of inflammatory cells and excess proteases in lung lavage fluid(2). High concentrations of the macrophage-derived cytokines TNF-α and IL-1 have been reported(3, 4), both of which may induce fibroblast collagen production(5, 6) and have been shown to cause pulmonary fibrosis in animal models(7–9). However, recent studies have suggested that the development of BPD in surfactant-treated infants may relate primarily to neutrophil involvement(10). Because of the importance of macrophage/fibroblast interactions in the pathogenesis of fibrosis(5, 6), and the difficulty of definitive macrophage identification using conventional staining techniques, we have used immunohistochemistry to delineate the inflammatory response in neonatal RDS.

One recently recognized initiating mechanism for inflammatory responses is by production of chemokines, related families of small molecular weight peptides that are specifically chemotactic for individual cell types. Theα (C-X-C) chemokines, notably IL-8, induce neutrophil chemotaxis while reducing fibroblast collagen production(11), whereas the β (C-C) chemokines are chemotactic for monocytes and T cell subsets without effect on collagen turnover. Among the β chemokines, MIP-1α and RANTES, which share a common receptor, attract and activate monocyte/macrophages and polymorphonuclear neutrophils(12). Increased local concentrations of IL-8 have been reported early in the course of neonatal RDS(13), whereas highly elevated levels of MIP-1α are found in BAL fluid in idiopathic pulmonary fibrosis(14). We have thus determined whether excess β chemokine production also occurs in RDS, whether such production may be associated with clinical outcome, and whether systemic corticosteroid therapy affects local production.

METHODS

Sequential BAL. Thirty consecutively born intubated preterm infants, with a clinical and radiologic diagnosis of respiratory distress syndrome, and seven infants born at the Homerton Hospital during the study period (February to September 1992) who required ventilatory support for other conditions (two for convulsions, one for congenital heart disease, and four with a clinical and radiologic diagnosis of transient tachypnea) were studied using serial (BAL, according to a protocol derived from routine endotracheal tube care procedure, which was approved by the Research Ethics Committee of St. Bartholomew's and Homerton Hospital. Lavages were performed within the first 48 h of birth and then at 1-2-d intervals while intubated, to a maximum of 21 d. Clinical details of the studied infants are given in Table 1. All infants were ventilated with the same model ventilator (Sechrist, UK), and pressures were monitored by digital analyzer(Digilog, EME, UK). Infants were classified, retrospectively, by outcome, as RDS (extubated and weaned into room air by 28 d), or BPD (remaining intubated or with requirements for supplemental oxygen beyond 28 d). All non-RDS infants were extubated into room air by 5 d, whereas 13/17 RDS and 13/13 BPD infants were ventilated beyond d 6, and 9/17 RDS and 13/13 BPD infants beyond d 12. Three infants in the RDS group and four in the BPD group received two doses of exogenous surfactant (Exosurf, Wellcome) on the first day of life via the endotracheal tube, according to the unit policy at that time (which required an arterial/alveolar ratio <0.22). These infants could not be differentiated in terms of ventilatory status, BAL results, or outcome. All BPD infants had radiologic evidence of chronic lung disease.

Table 1 Clinical and outcome details of infant groups in bronchoalveolar lavage study
Full size table

BAL protocol. After disconnection of the ventilator, 1.0 mL/kg sterile 0.9% saline was instilled into a distal airway via a wedged fine-bore(6FG) tube. The lavage fluid was aspirated under suction (20-40 cm H2O) into a bronchoscopy suction trap, and the suction tube was removed. After 2-3 min, providing the infant had fully stabilized, the tube was reinserted and the procedure was repeated, with the aspirates pooled. Full monitoring, including continuous measurement of oxygen saturation, was maintained throughout the lavages, which were all performed by the same investigator(S.H.M.). No adverse events were detected during or after the procedures. The average recovered represented 40-55% of the instilled volume. The lavaged fluid was maintained on ice until separated by centrifugation within 45 min(300 × g for 10 min at 4 °C). The supernatant was frozen at -70 °C, and the cell pellet was resuspended in 2.0 mL of RPMI and filtered through a 70-μm filter before cell counting was performed using a hemocytometer. Viability, assessed by trypan blue exclusion, was around 90-95%. Cells were transferred to slides by cytospin and then processed by differential counting (Diffquick, Dade, Germany) or immunohistochemistry. Differential counting was performed over at least five high power fields.

Bacteriologic exclusions. We excluded lavages from infants within the study who developed clinical, bacteriologic, or radiologic evidence of systemic or focal pulmonary infection. To try to detect covert infections, all infants underwent initial screening for congenital infection, including culture of amniotic fluid and oropharyngeal secretions, and urine was sent for cytomegalovirus screening. All lavages were examined microscopically for bacterial contamination, and additional lavages were twice weekly cultured for pathogens. However, specific cultures for Ureaplasma urealyticum and Mycoplasma hominis were not available.

Corticosteroid administration. Ten of the infants were commenced on dexamethasone therapy (initial dose, 0.5 mg/kg/day i.v., reducing sequentially after 4 d), at ages ranging from 12 to 21 d, because of poor clinical progress. The decision to treat with dexamethasone was made on entirely clinical grounds, without reference to the study and without knowledge of the results of cytologic examination. These infants were sampled on a daily basis for 48 h before and after therapy was commenced, and were omitted from the main study after dexamethasone administration.

Immunohistochemical analysis. Serial slides from each lavage were stained with MAb recognizing epithelial cells (cytokeratin, Dako, High Wycombe, UK), T lymphocytes (UCHT-1, a gift from Peter Beverley, UCH), macrophages (CD68, Dako), the intermediate-stage macrophage activation marker RM/3-1 (BMA Biomedicals AG, Switzerland), TNF-α (CB0006, a gift from Celltech Ltd., Slough, UK)(15), HLA-DR (Dako), proliferating cells (Ki67, Dako) and CD25 (Dako). Staining was developed using the alkaline phosphatase anti-alkaline phosphatase technique with Fast Red substrate, as previously described(15). The percentage of immunoreactive cells was obtained by counting over at least five high power fields, and overall numbers of cells were calculated using the initial volume and cell density. A good correlation of macrophage numbers was obtained between the different antibodies used.

Supernatant analysis. Supernatant fluid was diluted as appropriate in Tris-buffered saline, and assayed by commercial ELISA for the chemokines MIP-1α and RANTES (R&D Systems, Abingdon, Oxford), the cytokines TNF-α and IL-1β (Lab Impex, Teddington, Middlesex), and by in-house ELISA for elastase/α-1 antitrypsin, as previously described(16), using sheep anti-human elastase (Binding Site, Birmingham, UK) as capture antibody and peroxidase-conjugated sheep anti-humanα-1 antitrypsin as secondary. The sensitivities of the assays were 2 pg/mL (RANTES), 3 pg/mL (MIP-1α), 10 pg/mL (IL-1β and TNF-α), and 2 ng/mL (elastase/α-1 antitrypsin). Values were expressed per mL of initial supernatant.

Statistical analysis. Analysis between groups of the BAL data, which were not normally distributed, was performed using the two-tailed Mann-Whitney U test. Medians and 95% confidence intervals were calculated by the Wilcoxon method. The effects of dexamethasone upon inflammatory parameters were studied using the Wilcoxon signed rank test.

RESULTS

Cellular response. The infants with RDS and BPD had higher overall cell counts at all ages than control infants. Total cell counts ranged from median <10 × 104/mL supernatant in control infants to>100 × 104/mL in BPD infants. The use of immunohistochemistry allowed formal identification of cell types, and in particular enabled the pleomorphic macrophage population to be distinguished from pneumocytes, often of similar appearance (Fig. 1). Infants who developed BPD had higher overall counts at all ages than those with uncomplicated RDS (Fig. 2). Although overall median numbers of macrophages and neutrophils were largely similar (Fig. 2), it was noticeable that some infants had a predominant neutrophil response, whereas the cell pellet in others consisted almost entirely of macrophages.

Figure 1
figure 1

Immunohistochemical characterization of BAL-derived cells. (A) Cytokeratin-positive epithelial cell (red) with cytokeratin-negative cells of similar appearance on conventional staining(magnification × 500). (B) Cytokeratin-positive epithelial cells in a typical d 3 lavage from a BPD infant (× 200). (C) RM/3-1+ macrophages in matched slide from the same d 3 lavage (× 200). (D) TNF-α immunoreactive cells in d 5 lavage from a BPD infant (× 200).

Full size image
Figure 2
figure 2

Median cell numbers (× 104/mL), with upper quartile, in BALs obtained from seven ventilated infants without RDS(controls), 17 with uncomplicated RDS, and 13 who developed BPD. Samples were obtained on d 0-2, 3-5, 6-11, and 12-21 of life. All non-RDS infants were extubated by d 5. Statistical comparisons were made by the two-tailed Mann-Whitney U-test. Key: *vs non-RDS group,p < 0.05 (**p < 0.01, ***p< 0.001); ≈vs RDS group, p < 0.05(≈≈p < 0.01).

Full size image

Epithelial cells. In general, the numbers of cytokeratin-immunoreactive epithelial cells obtained were similar between groups at all ages (Fig. 2). The exception occurred in samples taken from the BPD infants during the first 48 h of life (mean 38.8× 104/mL supernatant), significantly higher than the RDS infants(11.0) and the controls (10.3).

Polymorphonuclear neutrophils. Significantly increased numbers of neutrophils were seen in the RDS and BPD infants at all ages, compared with controls, in whom medians <3 × 104/mL were obtained (Fig. 2). No significant difference could be detected between the RDS and BPD groups in the first 48 h of life, but thereafter neutrophil numbers were significantly higher in the BPD infants, with median counts >40 × 104/mL.

Macrophages. Numbers of immunohistochemically identified macrophages were significantly higher in RDS and BPD infants than controls at all ages (Fig. 2). In addition, numbers in BPD infants were significantly higher than RDS infants from birth. Overall numbers of macrophages were similar to neutrophil counts. There was great phenotypic variation among immunoreactive cells of the macrophage/monocyte lineage, making identification by routine staining difficult (Fig. 1). Almost all identifiable macrophages were immunoreactive for RM/3-1, an intermediate stage activation marker, and many showed clear evidence of activation on routine staining. Proliferating cells (Ki67+) comprised 2-7% of the recovered population.

T lymphocytes. There was no evidence of a T cell response to ventilation in RDS. Numbers of UCHT-1+ cells were uniformly below 0.5× 104/mL, with the majority of samples showing no identifiable UCHT-1+ cells. No increase was found in disease groups compared with non-RDS infants, or between groups.

CD25 expression. Despite the morphologic evidence of activation in many macrophages, and the near ubiquity of expression of the RM/3-1 activation marker, CD25+ cells generally comprised less than 5% of total macrophage numbers, and many specimens had minimal or no CD25 expression. Elevation of CD25+ numbers was, however, seen in several specimens obtained at times of bacterial infection (omitted from the main study), when up to 35% of macrophages expressed CD25.

TNF-α immunoreactivity. Numbers of TNF-α-immunoreactive (TNF-α+) cells were greater at all ages in both RDS and BPD infants than in controls. A significant increase in TNF-α+ cell numbers was found in BPD infants compared with the RDS group throughout, with recovery of median >30 × 104 TNF-α+ cells/mL at all times after 48 h in the former group.

Supernatant concentrations. β-Chemokines. Low concentrations of RANTES (<10 pg/mL) were found in all groups at all ages, and no increase could be detected in disease groups above controls. In contrast, concentrations of MIP-1α were high, exceeding 1-2000 pg/mL in many affected infants. A significant increase was found from birth in affected infants compared with controls (median concentrations <100 pg/mL), and concentrations were increased on d 0-2 and after d 6 in infants who developed BPD (Fig. 3).

Figure 3
figure 3

Median concentrations of chemokines, cytokines, and elastase, with upper quartile, in BAL supernatant fluid obtained from study groups as in Figure 2. All values are in pg/mL apart from elastase in ng/mL. Key: *vs non-RDS group,p < 0.05 (**p < 0.01, ***p< 0.001); ≈vs RDS group, p < 0.05(≈≈p < 0.01).

Full size image

Macrophage cytokines. Levels of TNF-α were increased compared with controls at all ages, with median concentrations >100 pg/mL found at all ages beyond 48 h in the BPD group. However, TNF-α concentrations were significantly increased only in the BPD infants compared with the RDS group after 12 d, although tending to be higher throughout (Fig. 3). Levels of IL-1β were substantially higher than TNF-α in both disease groups and controls (median concentrations>1000 pg/mL beyond 48 h in the BPD group). However, a similar distribution was found to TNF-α, with higher concentrations throughout in BPD infants, reaching significance only after 12 d.

Elastase/α-1 antitrypsin. A different pattern was found in elastase/α-1 antitrypsin concentrations, with high elevation above controls in the first 5 d, and then decreasing concentrations in later samples (Fig. 3). Although lower concentrations were found beyond 12 d, these were significantly increased in BPD infants compared with RDS.

The effects of dexamethasone administration. The administration of dexamethasone to 10 infants induced significant decrease within 24 h in mean BAL cellularity (macrophages, neutrophils, and TNF-α+ cells) and in the mean concentration of MIP-1α, TNF-α, IL-1β, and elastase/α-1 antitrypsin, but not RANTES, in supernatants (Fig. 4). These marked changes occurred at least 24 h before there was clinical improvement and were sustained for up to 5 d after administration began. Two infants developed frank bacterial colonization of secretions after 3-5 d of treatment and demonstrated an increase in cellularity and inflammatory mediators toward pretreatment values.

Figure 4
figure 4

The effects of dexamethasone administration in 10 infants. The hatched bars represent mean values (with SE) in the 48 h before treatment began, and the filled bars mean values during the first 48 h of treatment. Statistical significance is denoted by asterisks(*p < 0.05, **p < 0.01,***p < 0.001). (A) Concentrations of macrophage-associated chemokines and cytokines (pg/mL) and of elastase(ng/mL). (B) Numbers (× 104/mL) of macrophages, neutrophils, and TNF-α-immunoreactive cells.

Full size image

DISCUSSION

Using immunohistochemistry we have found evidence for significant inflammation very early in the course of neonatal RDS. This accords with reports of such an early inflammatory response by several groups(2, 3, 10, 17, 18), but the use of specific immunohistochemistry suggests that this response is more intense and probably of earlier onset than has been recognized. The detection of polymorphonuclear neutrophils by conventional staining techniques is not normally difficult, and their increase has been established(2, 4, 10). The pleomorphic nature of cells of the macrophage/monocyte lineage makes certain recognition more difficult, and thus the extent of macrophage involvement in RDS has been controversial(10). However, the centrality of macrophages and their products in inducing pulmonary fibrosis by alteration of extracellular matrix structures(7) and inducing fibroblast collagen production(5, 6, 8, 9) is well recognized, and we would contend that our findings of early macrophage infiltration and cytokine production in RDS are significant in both acute organ dysfunction and in the progression to fibrosis. Elastase/α-1 antitrypsin complex concentrations were high, although it has been shown that free elastase is not frequently found in RDS(18). We found no evidence of a cell-mediated response in RDS, and the lack of CD25 expression on activated monocytes appears to be a feature of prematurity(19).

We consider that there is currently no satisfactory endogenous marker of dilution for neonatal BAL studies, although we accept that other workers may choose to use such markers. The controversy over this decision has been well reviewed(20, 21). We chose to exclude albumin and urea on the grounds that transepithelial albumin leakage and tight junction integrity are highly dependent on epithelial sulfated glycosaminoglycans(22, 23), which we had shown to be disrupted in macrophage-mediated inflammation(24). We excluded secretory component on the basis that its nonsecretion is a recognised cause of immunodeficiency in the United Kingdom, and that its production has been shown to be up-regulated by inflammatory cytokines such as TNF-α(25) or decreased by protein deficiency(26). However, recent work suggests that this marker may provide a more constant denominator than albumin, and none of the reported infants was congenitally deficient in its production(27). Clearly the use of an exogenous marker would be ideal for precise quantification of cytokine production, but its detection in such high local concentration is likely to be significant in the pathogenesis of RDS.

The early elevation of MIP-1α concentrations, higher in those who developed BPD, may be a primary initiating mechanism for the macrophage recruitment and activation that we have demonstrated. MIP-1α production has been demonstrated in rat models of acute lung injury, in which anti-MIP-1α therapy markedly reduced both tissue damage and local TNF-α production(28). The cellular source of the MIP-1α in RDS is not yet clear, although its production has been documented in a pulmonary epithelial cell line, pulmonary fibroblasts, and alveolar macrophages(29). It is notable in this context that our BPD group was smaller and less mature at birth, and also showed higher numbers of epithelial cells on initial lavage, suggesting that the initial barotrauma may have been more damaging in this group. It is not yet known whether oxygen exposure or barotrauma can induce MIP-1α production by alveolar epithelium, although nanomolar endotoxin contamination of the humidified ventilatory gases may potentially do so(30). Other potential causes of macrophage recruitment include pulmonary production of other β-chemokines, including macrophage chemotactic protein-1(31), or granulocyte-macrophage colony-stimulating factor, which induces multifocal macrophage-mediated inflammation in granulocyte-macrophage colony-stimulating factor transgenic mice(32). The production of TNF-α and IL-1β at high concentrations will contribute independently to pathophysiologic disturbance in RDS. Systemic TNF-α administration in animals induced severe lung injury to an extent greater than did IL-1(33), whereas pneumocyte-specific expression of a TNF-α transgene caused either early death from severe respiratory distress or chronic progressive pulmonary fibrosis with leukocyte infiltration(34). Neutrophil infiltration and free radical damage is also characteristic of the pulmonary response to IL-1 administration(35). In addition to acute tissue dysfunction, progression to fibrosis in animal models can be prevented by blockade of either TNF or IL-1(8, 9).

This rapid temporal increase of BAL inflammatory cell numbers and activation status concords with our findings in postmortem specimens of dense interstitial TNF-α+ macrophage and neutrophil infiltration, which is associated with severe disruption of sulfated glycosaminoglycans(36). Despite the limitations inherent in neonatal BAL studies, they may thus be of direct relevance in the investigation of mucosal immunopathology in RDS.

Inflammatory cell numbers, concentrations of MIP-1α, macrophage cytokines, and elastase were reduced within 12-24 h of dexamethasone administration, whereas clinical improvement was not apparent for 48 h. This time course accords with reported studies of pulmonary cytokine production(3, 4), albumin leak(17), and lung mechanics(37), to suggest that reduction of local inflammation may be the primary effect of dexamethasone underlying its clinical benefit. These data suggest that very early introduction, possibly at birth, of agents that specifically attenuate the inflammatory response in RDS would be of clinical benefit. The metabolic consequences of dexamethasone administration counteract its beneficial effects, and more specific therapy, blocking chemokines, cytokines, or adhesion molecules is likely to have much more impact. The outcome in animal models is dramatically improved if cytokine antagonists are used before the damage is inflicted(8, 38). Neonatal RDS is perhaps the only human condition in which patients present without inflammation and will invariably and predictably develop it, to their serious disadvantage, within hours: it may be uniquely amenable to cytokine antagonism.