The Molecular Basis of Pulmonary Alveolar Proteinosis

GM-CSF Deficient Mice

The first major clue regarding the pathogenesis of PAP was the discovery mice deficient in GM-CSF develop a lung phenotype [5; 6] that is biochemically, pathologically, physiologically, histologically, and ultrastructurally similar to human PAP. It begins at birth with development of foamy alveolar macrophages and progresses slowly over time to include a patchy distribution of lesions comprised of well-preserved alveoli filled with a fine granular, eosinophilic, periodic acid-Schiff positive material [5; 6]. The phenotype also comprises an accumulation of lymphocytes in a patchy peri-broncho-vascular distribution. The accumulated material was structurally, biochemically and functionally similar to normal surfactant [13]. Importantly, accumulation was due to reduced clearance (not increased production [5]) caused by reduced catabolism in alveolar macrophages, not AEC-II [12]. Increased levels of several cytokines (monocyte chemotactic protein 1 and macrophage-colony stimulating factor) serve as useful biomarkers of PAP [14; 41; 42; 43]. GM-CSF deficient mice have increased mortality from uncontrolled pulmonary and systemic infections [44] and increased susceptibility to a wide variety of microbial infections including Gram negative and Gram positive bacteria, malarial parasites, virus, and M. tuberculosis [44; 45; 46; 47; 48; 49].

The disruption of surfactant homeostasis, increased microbial susceptibility, and abnormal alveolar macrophage morphology and function in GM-CSF-deficient mice were corrected by replacement [50; 51; 52]. The effects of GM-CSF on surfactant homeostasis and alveolar macrophage function were found to be mediated primarily by PU.1, a transcription factor required for terminal differentiation, surfactant catabolism and multiple immune functions of alveolar macrophages [14].

GM-CSF Receptor β Deficient Mice

Mice deficient in the GM-CSF receptor β chain develop a lung phenotype similar to that of GM-CSF deficient mice [53]. GM-CSF levels are increased in serum and lung in these mice due to reduced clearance [32], a molecular observation of translational significance (see below) [54]. The disruption of surfactant homeostasis were corrected by bone marrow transplantation, demonstrating that the site of the molecular pathogenesis was cells of hematologic and not pulmonary origin, i.e., alveolar macrophages not pulmonary epithelial cells [55]

Autoimmune PAP in Humans

A second major clue regarding the pathogenesis of PAP was the strong association of high levels of neutralizing GM-CSF autoantibodies with idiopathic PAP [31], the most common clinical form representing 90% of cases [56]. This rare disease has a prevalence of 6–7 individuals per million in the general population, occurs in all ethnic groups and is approximately twice as common in males, presumably due to the association with smoking. Individuals usually present in the third to fourth decades of life with progressive dyspnea of insidious onset, diffuse bilateral lung infiltrates on radiological evaluation, and pulmonary pathology comprised of well-preserved alveoli filled with a granular, eosinophilic, periodic acid-Schiff (PAS)-positive material. Secondary infections can occur and account for 18% of the reported mortality in PAP [1], presumably due to myeloid cell defects (see below).

The GM-CSF autoantibodies are polyclonal, comprised primarily of IgG1 and IgG2, with only small amounts of IgG3 and IgG4 [57]. They recognize human GM-CSF with high avidity (20±7.5 pM) and high specificity [58] and are capable of neutralizing far more GM-CSF than is physiologically present [58]. Interestingly, GM-CSF autoantibodies have been consistently detected and comprise the major anti-cytokine activity in pharmaceutical immunoglobulin prepared from pooled serum from healthy individuals [59]. Further, GM-CSF autoantibodies were also present in healthy donors, albeit at levels (median [interquartile range] = 1.04 [0.63–1.72] μg/ml) [57] far lower than those present in PAP patients (59.8 [27.4–116.5] μg/ml). These results suggested the hypothesis that the risk of PAP is increased when GM-CSF autoantibody levels are increased above a critical threshold, which has been reported to be ~5 μg/ml [60] (Figure 3). The function of GM-CSF autoantibodies in healthy individuals is uncertain. However, the finding that GM-CSF levels in serum are higher than earlier reports suggested and that more than 99% of serum GM-CSF in healthy individuals and PAP patients exists in the form of immune complexes suggesting that GM-CSF autoantibodies may have a scavenging role [57].

Reproduction of Autoimmune PAP in Healthy, Non-Human Primates

In the context of data from GM-CSF deficient mice, the strong association of neutralizing GM-CSF autoantibodies in patients with PAP is compelling evidence suggesting a potential role in pathogenesis. However, their presence in healthy individuals and pharmaceutical immunoglobulin prepared from healthy individuals was troubling. This problem was addressed in translational studies in which GM-CSF autoantibodies were isolated from a patient with PAP and injected into healthy, non-human primates [61]. Serum GM-CSF autoantibody levels were maintained above 40 μg/ml for up to ten months resulted in the reproduction of the biochemical, cellular, and histopathologic features of PAP. The first abnormality observed was the accumulation of surfactant within alveolar macrophages and was followed by progressive accumulation of intraalveolar surfactant. This provided a ‘first look’ at the initiation of PAP, since clinically, patients only present after surfactant has accumulated to sufficient levels to impair gas exchange (a late complication). These data confirm GM-CSF autoantibodies as the pathogenic driver of the common clinical form of PAP, suggesting it can be appropriately called autoimmune PAP.

Hereditary PAP

PAP associated with non-detection of GM-CSF receptor β chains on blood leukocytes was reported in three infants presenting with respiratory failure [62]. All had the typical radiological manifestations and two had histopathologic findings similar to GM-CSF receptor β chain deficient mice and autoimmune PAP. Altered GM-CSF signaling was demonstrated by ligand binding studies and progenitor clonogenic assays. CSF2RA mRNA and protein was detectable in all three patients. A point mutation within CSF2RB encoding the GM-CSF receptor β chain was detected in one patient, but was subsequently shown to be present in 6% of 184 multinational healthy individuals (NCPI single nucleotide polymorphism (SNP) Database ID rs1801122 (http://www.genecards.org)) and thus, was not pathogenic. Notwithstanding, this report suggested the GM-CSF receptor β chain was critical in surfactant homeostasis in humans and in the pathogenesis of hereditary PAP.

Hereditary PAP caused by abnormalities or absence of GM-CSF receptor α was reported in three children [32; 33]. One child [32] developed progressive dyspnea of insidious onset by 3–4 years and diffuse lung disease and respiratory insufficiency by 6 years. Lung histopathology was similar to autoimmune PAP and GM-CSF deficient mice. However, GM-CSF autoantibodies were absent and GM-CSF levels were elevated in the lungs and serum. The GM-CSF receptor β chain was present and had a normal nucleotide sequence. However, the α chain had a reduced molecular weight, abnormal glycosylation pattern, reduced GM-CSF binding, and impaired activation of STAT5 [32]. Karyotyping, genomic chromosomal hybridization, and high resolution SNP analysis revealed a 1.6 Mb deletion in the pseudo-autosomal region at Xp22.33 encompassing the maternal CSF2RA allele, and a point mutation (G196R) in the paternal CSF2RA allele that altered glycosylation of the GM-CSF receptor α peptide and markedly reduced GM-CSF binding. This child’s asymptomatic 8-year-old sister had increased serum GM-CSF and SP-D (both markers of PAP lung disease severity) and subsequent radiological and pulmonary function abnormalities demonstrated she had PAP. A 4 year old child developed PAP in association with Turner’s syndrome and loss of both CSF2RA alleles [33]. Subsequent screening of pediatric patients with unexplained PAP (i.e., who were negative for GM-CSF autoantibody) for an elevated serum GM-CSF level resulted in identification of 5 additional patients aged 3–11 years of age with PAP caused by various CSF2RA mutations (B.C.T., unpublished results). Together, these cases indicate that loss of GM-CSF receptor function causes hereditary PAP, but that additional factors determine the severity of lung disease in PAP.

Murine Models of Secondary PAP

Murine models of secondary PAP have been reported. Rats exposed to inhaled silica develop PAP [90] although the mechanism was not determined. PAP also develops in mice with severe combined immunodeficiency, again without determination of the mechanism involved [91]. A recent study reported that use of dichloromethylene diphosphonic acid-encapsulated liposomes to reduce alveolar macrophage numbers increased pulmonary surfactant levels [92]. Although limited, these data support the hypothesis that secondary PAP occurs as a consequence of disorders that reduce either the numbers or functions of alveolar macrophages.

Secondary Infections In PAP

Microbial infections account for 18% of reported deaths attributable to PAP [1]. These infections occur at pulmonary and extrapulmonary sites, indicating that the predisposition to infection is systemic, rather than confined to the lungs [73; 107; 108; 109]. Although organisms common in community- and hospital-acquired infections have been reported in PAP (e.g., Streptococcus, Klebsiella, Hemophilus, Staphylococcus, Pseudomonas, Serratia, Proteus, Escherichia), more often reports have identified ‘unusual’ or opportunistic microbial pathogens (Mycobacteria, Aspergillus, Nocardia and others) [1]. GM-CSF deficient mice also have increased mortality from uncontrolled pulmonary and systemic infections and increased susceptibility to a wide variety of microbial pathogens including Gram positive and Gram negative bacteria (Streptococcus, Pseudomonas. aeruginosa, Listeria monocytogenes), fungi (Pneumocystis jiroveci), mycobacteria (M. Tuberculosis), malarial parasites (Plasmodium chabaudi), and virus (adenovirus) [44; 45; 46; 47; 48; 49]. The increased pulmonary infection risk in GM-CSF deficient mice can be rescued by expression of GM-CSF specifically in the lungs [47]. Interestingly, overexpression of GM-CSF in the lungs of mice reduced the susceptibility to bacterial lung infection beyond that of wild type mice resulting in mice that were ‘super-resistant’ to infection [47]. Together, studies in man and mice with disrupted GM-CSF signaling demonstrate that GM-CSF plays an important role in immunity through regulation of myeloid cell maturation and activation and the recognition, molecular signaling responses, phagocytosis, and clearance of microbial pathogens.

GM-CSF Regulation of Alveolar Macrophage Immune Functions

Alveolar macrophages from GM-CSF deficient mice have a number of functional and mechanistic defects, including impaired adhesion, expression of pathogen recognition- and other receptors, phagocytosis, reactive oxygen species generation, microbial killing, and pathogen-stimulated proinflammatory and chemokine secretion. These defects are rescued by expression of GM-CSF in the lungs of transgenic mice, demonstrating that the presence of GM-CSF in the lungs is required for normal alveolar macrophage immune functions [14]. Alveolar macrophages from GM-CSF deficient mice have reduced expression of PU.1, a macrophage ‘master’ transcription factor involved in macrophage differentiation [14; 17]. The level of GM-CSF in the lungs determines the level of PU.1 expression in alveolar macrophages [14]. Expression of PU.1 in alveolar macrophages from GM-CSF deficient mice rescues the alveolar macrophage innate immune functions [14]. The presence of elevated levels of GM-CSF autoantibodies in patients with autoimmune PAP or healthy non-human primates reduces PU.1 mRNA levels and innate immune functions of alveolar macrophages (97 and 98). GM-CSF augmentation therapy rescues PU.1 expression and innate immune functions in humans [110]. Together, these and other results strongly support the conclusion that GM-CSF, via PU.1, stimulates the terminal differentiation of alveolar macrophages in the lungs and for the acquisition of normal immune functions [14; 57; 61; 111].

GM-CSF has pleiotropic effects on myeloid cells, regulating survival, proliferation, differentiation, priming, and activation. Recent advances in our understanding of GM-CSF receptor structure and function now provide an explanation for this pleiotropy. The dose-dependent effects of GM-CSF are mediated in a mutually exclusive, reciprocal fashion through two β-chain residues: Ser585 at low GM-CSF levels and Tyr577 at high levels [112]. Thus, the GM-CSF receptor acts as a binary switch promoting cell survival and differentiation at low doses and proliferation and antimicrobial functions at high levels. GM-CSF bioactivity is at the center of the mechanism that regulates myeloid functions. Low levels of GM-CSF present in normal healthy tissues promote the terminal differentiation of tissue macrophages while high levels typically present at sites of inflammation promote macrophage priming and enhance the macrophage activation response.

GM-CSF is important in the regulation of molecular signaling pathways involved in the innate immune recognition of microbial pathogens. GM-CSF deficient mice have reduced local and systemic responses to lipopolysaccharide (LPS) exposure in vivo and in vitro [15]. Specifically, the LD50 of intravenous LPS administration is increased in GM-CSF deficient mice compared to strain-matched wild type mice [113]. Pulmonary administration also results in lower levels of LPS-stimulated proinflammatory cytokines (e.g., TNFα), chemokines (e.g., MIP-1), and pulmonary neutrophil recruitment in GM-CSF deficient mice compared to control mice [15]. Administration of neutralizing anti-GM-CSF antibodies to wild type mice also results in lower levels of LPS-stimulated molecular and cellular inflammation compared to controls [114]. These findings can be reproduced in vitro by exposure of isolated alveolar macrophages to LPS, which results in reduced expression of inflammatory cytokines and chemokines [15]. Further, the proinflammatory responses and mortality from experimental peritonitis or intraperitoneal LPS exposure were blunted in GM-CSF deficient mice compared to controls [115]. Normal responses were rescued by intraperitoneal instillation of wild type macrophages or macrophages from GM-CSF deficient mice retrovirally transduced to express PU.1. Since the pathways involved in the molecular recognition of LPS have been defined [116], the effects of GM-CSF on this innate response were studied in detail. The low levels of GM-CSF present physiologically were found to enable TLR-4 signaling in macrophages by stimulating expression of several components of this pathway (Figure 4). GM-CSF also regulates the expression of other pathogen associated molecular pattern (PAMP) receptors on macrophages including TLR2, mannose receptors, and scavenger receptors as well as receptors recognizing the Fc motif of ligand-bound immunoglobulin (FcγI, FcγII, and FcγIII) [14; 15; 16; 17; 18]. These results indicate that GM-CSF plays an important role in the response to infection by regulating the innate immune recognition and initial inflammatory signaling responses of macrophages to microbial infection.

GM-CSF also regulates molecular communications between innate and adaptive immune responses, e.g. interleukin-18 (IL-18)/interferon-γ (IFN-γ) axis. Normally, macrophages exposed to microbial pathogens (e.g. adenovirus) respond by secretion of IL-18, which in turn stimulates Th1 and NK cells to secrete IFNγ [16; 117]. IFNγ has pleiotropic effects on innate and adaptive immunity, for example, enhancement of Fcγ receptor expression, phagocytosis and other host defense functions of macrophages and T and B lymphocytes. In GM-CSF deficient mice, stimulation of IFNγ production by adenovirus-exposed macrophages was markedly reduced due to their reduced capacity for secretion of IL-18 and IL-12. GM-CSF deficient mice also have blunted responses in collagen-induced arthritis [118] and experimental allergic encephalitis [119], which has stimulated clinical interest in the potential use of GM-CSF autoantibodies to treat inflammatory and autoimmune diseases [120]. The molecular basis of such complex inflammatory responses involving macrophages is currently the subject of active investigation.

GM-CSF is also important in the capacity of macrophages to internalize a wide range of microbial pathogens and particulates. For example, macrophages from GM-CSF deficient mice have reduced phagocytic internalization of microbial pathogens [14; 16; 17]. Consistent with the reduction in Fc receptor levels in these mice, Fc receptor-mediated phagocytosis of latex microspheres by macrophages is reduced. However, this effect can not be completely explained by changes in Fc receptor expression since phagocytosis of unopsonized latex beads and zymosan is also impaired. Further, impaired uptake is more pronounced for larger (i.e. 2–10 μm) beads than for smaller (< 100 nm) beads and uptake of bacteria (~ 0.5 μm) is more severely affected than that of adenoviral virions (~100 Å diameter). Since macrophages internalize adenovirus by endocytosis, these results suggest GM-CSF may have a differential effect on phagocytosis and endocytosis. This hypothesis is supported by the observation that endocytosis of surfactant by alveolar macrophages is not impaired in GM-CSF deficient mice or patients with either autoimmune PAP or hereditary PAP. The phagocytic defect of alveolar macrophages in GM-CSF deficient mice was corrected in vivo by restoration of GM-CSF in the lungs [16] and in vitro by retroviral expression of PU.1 [14], demonstrating that GM-CSF, via PU.1, critically regulates phagocytosis of microbial pathogens by macrophages in mice. Phagocytosis of bacteria, zymosan and latex particles by alveolar macrophages from human PAP patients is similarly impaired [57]. While GM-CSF has marked effects on macrophage cytoskeletal organization and cell morphology, the precise mechanism(s) by which phagocytosis is regulated by GM-CSF is not known.

GM-CSF plays a critical role in the killing and pulmonary clearance of various extracellular and obligate intracellular microbial pathogens, including Streptococcus, Pseudomonas, Pneumocystis, Mycobacterium, and Adenovirus [14; 18; 45; 46; 47]. For each of these organisms, GM-CSF expression in the lungs is required to control the lung infection and to enable clearance. Further, increased levels of pulmonary GM-CSF in transgenic mice confer an increased capacity for pulmonary clearance by increasing both the numbers and intrinsic antimicrobial activity of alveolar macrophages. In isolated macrophages from GM-CSF deficient mice, defective killing of Gram negative and Gram positive bacteria are regulated by PU.1, independent of effects on phagocytosis [14]. Although the precise mechanism responsible is not known, GM-CSF regulates several important antimicrobial activities in macrophages including the production of reactive oxygen species (e.g., superoxide and hydrogen peroxide radicals), expression of antimicrobial enzymes (e.g., lysozyme, other proteases) and intracellular translocation to sites of destruction ([47], and unpublished observation). Adenoviral infection provides a useful example of the latter [18] (Figure 5).

GM-CSF Regulation of Neutrophil Immune Functions

Prompted by the implication of a possible defect in systemic immunity in patients with autoimmune PAP and in GM-CSF deficient mice, functional defects were identified in circulating neutrophils from both [111]. The defects comprise impaired cellular adhesion, reactive oxygen species production, phagocytosis, and bacterial killing [111]. Notably, the differential pattern of impairment in these various functions varied widely among the different functions but was strikingly similar between the two species, suggesting that GM-CSF regulates neutrophil functions similarly in man and mice (see figure 1d of reference [111]). Importantly, the phagocytic defect was reproduced in healthy mice by injection of anti-mouse GM-CSF antibodies and in healthy non-human primates by injection of PAP patient-derived GM-CSF autoantibodies. The defect in cellular adhesion was partially explained by complete inhibition of the GM-CSF-stimulated increase in cell surface levels of the adhesion molecule CD11b in PAP patients and in GM-CSF autoantibody-injected non-human primates [61]. The defects in autoimmune PAP patients were associated with loss of GM-CSF receptor signaling as judged by the inability of GM-CSF to stimulate phosphorylation of the immediately downstream signaling molecule STAT5. However, the precise mechanisms underlying these abnormalities are currently unknown. Notwithstanding, disruption of GM-CSF signaling does not appear to affect neutrophil differentiation because patients with autoimmune PAP had normal PU.1 levels in their neutrophils, cell surface differentiation markers, morphology, and numbers of circulating neutrophils. Thus, GM-CSF plays a critical role in regulating the basal functional capacity of mature circulating neutrophils, providing a potential explanation for the increased systemic infection risk in PAP patients and GM-CSF deficient mice.