New insights on congenital pulmonary airways malformations revealed by proteomic analyses

Lung development is a complex process allowing parenchymal architecture to evolve along the bronchial organization. To establish correct bud elongation and airway branching, cellular interactions between epithelial, endothelial and mesenchymal cells are required. These interactions are dependent on the paracrine secretion of different growth factors or transcription factors. Growth factors are classified into different groups based on their cell of origin, such as fibroblast growth factors (FGF), vascular growth factors (VEGF), and epithelial growth factors (EGF). Transcription factors, such as SOX2 and SOX9, are recognized to play a role in lung development and in particular during branching morphogenesis [1‐5]. During the canalicular stage, expression of SOX2 and SOX9 differ in their localization. Indeed, SOX 2 is expressed in the proximal airways surrounded with smooth muscle cells (SMCs) and SOX9 is restricted to the distal epithelial buds [1]. SMCs surrounding epithelial cells are crucial in this process due to their ability to contract and to allow SMCs later to extrude into branches [6, 7].

Congenital lung anomalies (CLA) are a group of developmental lung alterations thought to result from different external factors occurring during pregnancy, such as toxic exposure, or are associated to preterm birth. In these cases, cellular crosstalk can be altered or interrupted leading to the impairment of lung branching and alveolar formation [8‐12]. Congenital pulmonary airway malformations (CPAM) belong to a group of rare CLA whose pathological origin is still poorly understood [13]. In Western Europe, CPAM have an estimated prevalence between 0.87 and 1.02/10,000 live births [14]. Depending on timing of routine ultrasound, CPAM are often detected around 16 to 20 gestational weeks (GW). CPAM were initially classified by Stocker et al. in 3 different subtypes of cystic lung lesions (1 to 3), differing both macroscopically (cyst size) and on histology [13]. Despite further attempts at refining the categories, a type 0, or congenital acinar dysplasia and a type 4 category, representing pneumopulmonary blastoma instead of CPAM were added [15]. Langston preferred the denomination “large cyst and small cyst-types”, i.e. type 1 and 2, the definition used in this paper [16]. It remains as yet unclear whether or not CPAM 1 and 2 share the same origin.

Based on these considerations, the current research project aims at studying by several approaches the cellular origins of the two most frequent CPAM, CPAM types 1 and 2 (0.85/10,000 and 0.2/10,000 live births, respectively). We stained surgically removed CPAM specimens and analyzed markers of alveolar, muscular and bronchial cell differentiation on these samples. Adjacent healthy lung parenchyma served as control. We found that cystic epithelium from both CPAM subtypes expresses several bronchial markers. On the other hand, SPC, a marker of alveolar epithelial type 2 cells (AECII), was expressed in CPAM 1, but barely seen in CPAM 2. We then assessed ACTA2 expression and its distribution in CPAM. Here again, we observed similarities in terms of ACTA2 expression in SMCs of both CPAM 1 and bronchi, whereas ACTA2 positive SMCs were less prevalent in CPAM 2. These results were further reinforced by proteomic analysis performed on CPAM cysts, and healthy adjacent normal-appearing lung, as well as on fetal airspace and bronchial epithelium, after microlaser dissection. CPAM 2 protein profile was clearly distinct from all the other samples. Furthermore, CPAM protein profiles overlapped partially with those of fetal samples. Our data provide important insights into CPAM origin and demonstrate some differences between CPAM types 1 and 2, suggesting that these malformations might occur at different stages of embryogenesis.

In this study, we used two different approaches, IHC and proteomics, to characterize CPAM lesions. Several studies have been dedicated to the analysis of the growth and transcription factors implicated in the cellular proliferation of altered lungs [8]. The results have highlighted the role of different molecules, such as FGF-7, FGF-10, PDGF BB and HOXB5, in abnormal lung development [22, 23]. All these studies were performed in animal models mimicking CPAM, but only few experiments have been realized to date on human samples [24‐29]. In humans, the characterization and classification of the different types of CPAM is based on histopathological evaluation, as a first guide for assessing phenotypic variation and subtyping. CPAM types 1 and 2 differ in both cyst size and histology (cell types lining the cysts, muscle wall, presence or absence of cartilage among others) [13, 15, 16]. CPAM are therefore mainly classified according to gross findings, histological features and the structures along the respiratory tract they most resemble. However these descriptions do not integrate embryological pathogenesis.

To provide better insight in possible links with the embryological development of the lung, we analyzed the expression of SOX2 and SOX9 in growing buds during lung canalicular developmental stage and also observed an antero-posterior gradient in SOX expression, as previously reported [1]. This gradient was not present in control postnatal bronchi, neither in CPAM cysts. Indeed, SOX2 and SOX9 staining was similar along CPAM epithelium and adjacent bronchi. Notwithstanding, CPAM 2 lesions expressed significantly less SOX2 than CPAM 1.

SOX2 positive cells are tightly regulated by the presence of ACTA2 positive cells that allow for branching in parallel to a decrease in SOX2/ SOX9 positive cells during the canalicular stage [1]. Danopoulos and colleagues suggested an interaction between SOX2/ SOX9 cells and SMCs that could possibly influence cell proliferation in the growing airways of the human fetal lung [1]. Indeed, SMCs are essential to regulate epithelial branching through peristalsis, paracrine-signalling pathways and secondary lung septation [7]. In CPAM, although we observed significant differences in SMC distribution, with CPAM 1 being closer to bronchi, we could not find a correlation between the SMC thickness and epithelial cell proliferation.

Specific epithelial cell marker analysis showed some similarities between bronchi and CPAM epithelium, suggesting abnormal epithelial proximo-distal differentiation. Interestingly, only the CPAM 1 epithelium contained SPC positive cells, a specific staining for late progenitors and differentiated AECII. The increased number of SPC positive cells can be due to the differentiation process. Indeed, bronchioalveolar stem cells express SPC and give later rise to AECII. Alternatively, the presence of pro-SPC in these cysts might suggest that the initial event happens later during embryogenesis stage when alveolar cells are already differentiated.

In contrast to the study by Swarr et al., no mucinous cell clusters were seen in the analyzed CPAM samples [30]. Furthermore, MUC5AC expression in CPAM 1 and CPAM 2 was significantly lower than in bronchi (Fig. 3b). This result suggests that although CPAM epithelium might derive from bronchial tree, significant differences are observed between CPAM and bronchial epithelium.

After microlaser dissection of the different epithelial linings, we detected distinct protein clusters in CPAM 2 and CPAM 1 corresponding to the preliminary pathologist classification. Indeed, CPAM 2 presented a distinctive proteomic profile compared to CPAM 1, bronchi and alveoli. CPAM 1 clustered with alveoli and bronchi, thereby corroborating our immunohistochemistry results, where CPAM 1 showed some similarities with bronchi concerning ACTA2, Krt17 and SOX2/9 staining, but also with alveoli.

These results call for two different hypotheses: either CPAM originates from the developing bronchi at different stages of development, or these lesions represent truly distinct entities resulting from a different etiology. The similarities between CPAM 1 and both bronchi and alveoli, and of CPAM 2 only with bronchi reinforce the hypothesis that CPAM 1 and CPAM 2 grow at different lung branching timelines. A subset of proteins is upregulated in CPAM 1 and fetal bronchi, as compared to CPAM 2 and to fetal canaliculi. This suggests a similarity between CPAM 1 and fetal bronchi. Contrariwise, no clear link was observed between CPAM 2 and fetal proteins. The comparison between fetal canaliculi, bronchi and CPAM identified clearly distinct protein profiles between fetal tissue and CPAM 2, whereas in CPAM 1 some fetal bronchial proteins remained detectable.

A modest reduction of PI3K-AKT-mTOR signalling pathway was suggested to influence CPAM 1 and CPAM 2 formation in transcriptomic data [30]. We also found in our proteomic analyses a down regulation of phosphatidylserine binding protein, an AKT activation modulator, supporting this finding (Fig. 4e) [31]. The same result was observed in most of fetal canaliculi (Fig. 4g). Similarly to Swarr et al., we report that CPAM 2 upregulated proteins were involved in cellular proliferation and differentiation (cytoskeleton organization, spermatogenesis and keratinocyte development). In addition, filament and microtubule organization are important to allow correct cellular arrangement. Upregulation of these biological process confirmed previous published transcriptomic data [30].

Finally, the heterogeneity in the CPAM 1 cluster compared to bronchi and alveoli could suggest the existence of intermediate phenotypes reinforcing the overlapping features often seen histologically. The upregulated proteins seen in both CPAM 1 and fetal bronchi, but not in CPAM 2 and fetal canaliculi have a role in mesenchyme-epithelial differentiation or cytoskeletal formation. These proteins have been involved in tumorigenesis due to their role in proliferation and differentiation pathways, as well as in cellular crosstalk during lung embryogenesis [32‐34]. This last result links CPAM and alterations in cellular crosstalks with abnormal desmosome communications that could suggest a different physiopathological etiology in CPAM 2. The link with tumorigenesis is unclear, since malignancy in CPAM is rather related to the presence of clusters of mucinous cells, seen in CPAM 1 but not in CPAM 2 Higher numbers of CPAM 2 samples should be analyzed to confirm these results.

Our results are thus in agreement with the transcriptomic analyses already published showing a distinction between cyst and control lung [30]. Nevertheless our experiments add a more precise distinction between the epithelium present in cysts and control tissue samples due to the microlaser dissected epithelium analysis.

Our study has some limitations. First, although our findings were confirmed by proteomic analysis, sample size is small and need to be enlarged. Second, the prevalence of CPAM 1 and CPAM 2 variants can differ depending of the center of recruitment with more cases of mucinous cell clusters leading to potential tumoral transformation [35, 36]. We observed the presence of different keratins in our proteomic results. Finally, according to different animal studies, SMCs may influence CPAM formation due to the secretion of growth factors [27, 28]. Unfortunately, even if we suspect a role of the mesenchyme in CPAM formation, proteomic analysis was not able to identify in this study the previously involved growth factors, and only a minority of the transcriptional factors described in lung malformations in animal models. Protein crosslink due to FFPE conservation could have influenced our results, by allowing for only partial protein detection, the most resistant being mainly structural proteins. However, our results are in accordance with the previous transcriptomic study by Swarr et al., who find differences between CPAM malformation types [30].

The results reported in our study provide a new step in the understanding of CPAM etiology. This study is the first on CPAM to our knowledge, to use a proteomic approach with lung samples obtained after microlaser dissection. This exciting method allows for the analysis of different compartments within the CPAM lesions. This methodology applied to CPAM lesion is innovative and the possible use of FFPE material will allow for the analysis of tissue samples from different biobanks, avoiding the shortage of frozen material. Interestingly, proteomic differences observed between CPAM 1 and 2 support to the initial pathological classification proposed by Stocker et al., and by the revision provided by Langston more than the recent classification using micro and macrocysts classification [30, 37]. This technique could also help in the diagnosis of CPAM subtypes in unclear clinical cases.

Future work including more patients and quantitative proteomic analyses could pave the way to a more in-depth delineation between CPAM types 1 and 2. In conclusion, the description and classification of CPAM lesions remains a real challenge, the main issues being adequate management decisions for these patients.