Proposal of selective wedge instillation of pulmonary surfactant for COVID-19 pneumonia based on computational fluid dynamics simulation

It is known that SARS-CoV-2 binds angiotensin converting enzyme 2 (ACE2) on the pulmonary epithelial cells and causes acute interstitial pneumonia. ACE2 and TMPRSS2 (Type II transmembrane serine protease) are necessary for invasion into a cell; the two enzymes coexist in Type II pneumocytes, ileal absorptive enterocytes, and nasal goblet secretory cells [1]. Type II pneumocytes are the only cells that produce and secrete pulmonary surfactant (PS), which prevents alveoli from collapsing by reducing the surface tension of the alveolar liquid layer [2]. Iwasawa et al. reported that ultra-high-resolution X-ray CT images in coronavirus disease of 2019 (COVID-19) pneumonia showed volume loss of the lung parenchyma because of alveolar collapses [3]. The alveolar collapse causes mechanical injury of capillaries located in the alveolar wall and allows the virus to bind ACE2 on the surface of endothelial cells in the alveolar capillary.

Acute respiratory distress syndrome (ARDS) was first reported in 1967 by Ashbaugh et al., who inferred that its essential pathology was alveolar collapse caused by PS dysfunction and likely infant respiratory distress syndrome [4]. Currently, ARDS is defined as non-cardiogenic pulmonary edema with vascular hyper-permeability; the role of PS dysfunction is regarded as subtle [5]. This conceptual change may be partly because of the clinical ineffectiveness of PS instillation therapy for adults. Several clinical trials indicated that PS instillation therapy failed in reducing mortality in spite of improvements in blood oxygenation [6, 7]. However, there is a possibility that this ineffectiveness is caused by insufficient delivery owing to the significant difference in size and structure between newborn and adult lungs [7, 8]. Additionally, several researchers have reconsidered classical concepts of interstitial pneumonia and stated that alveolar collapse is the predominant mechanism of diffuse alveolar damage, which is a typical pathologic change in ARDS, acute interstitial pneumonia, and acute exacerbation of idiopathic pulmonary fibrosis [9, 10]. Steffen et al. recently demonstrated the experimental therapeutic effects of surfactant replacement for acute lung injury in a rat bleomycin model [11].

Therefore, it is apparent that the loss of PS because of the infection of Type II pneumocytes causes diffuse alveolar damage in COVID-19. Gattinoni et al. reported that the respiratory system compliance is almost normal in the initial stages of COVID-19 pneumonia [12]. It may be because regional low distensibility due to alveolar collapses is compensated by over-inflation of the surrounding intact lung parenchyma. Hence, supplying PS to the diseased alveolar regions to recover alveolar collapses is one of the most important strategies for ARDS by COVID-19. Filoshe et al. performed simulations of PS instillation and concluded that current PS delivery methods should be improved [13]. The airway models in their study were a combination of cylinders of varying calibers representative of the cascading lung bifurcations without a continuous surface. In this study, we simulated PS instillation in a realistic three-dimensional (3D) human airway model [14] using computational fluid dynamics (CFD).

Fluid flow in the liquid layer covering the airway wall driven by PS instillation was computed using a CFD solver (acuSolve, Altair Engineering, USA). Because applying CFD to the entire airway tree from the trachea to respiratory bronchioles requires substantial computer resources, two types of 3D human airway models were prepared. One was from the trachea to the lobular bronchi (Fig. 1, light blue part), and the other was from a subsegmental bronchus to respiratory bronchioles corresponding to the posterior subsegment of the basal medial segment in the right lung (rtS7a, Fig. 1, dark blue part), which is the smallest subsegment in the human lung. The number of end segments in the entire model was 531, and their mean diameter was 1.92 mm (total lung capacity). The subsegmental model had 511 end segments corresponding to respiratory bronchioles with a mean diameter of 0.35 mm. Because the algorithm for the 3D airway tree model was devised to assign the air-supplying region of lung parenchyma for each segment [15], the parenchymal volume in the subsegmental model was automatically recognized as 1.5 % of the entire lung volume. The entire model consisted of 1,929,371 nodes and 8,531,760 finite elements (tetras), whereas the subsegmental model consisted of 1,802,620 nodes and 7,960,320 tetras.

According to the simulated results by Filoshe et al. [13], the thickness of the liquid layer was 14 % of the inner radius of the airway segment. The liquid layer was assumed to be replaced by an instilled PS with the same thickness. The fluid surface facing the airway wall (green or blue parts in Fig. 1) were set as non-slip condition (velocity is zero for all directions), and the fluid surface facing air (white parts in Fig. 1) were set as slip (velocity is zero for only the normal direction) condition. The flow rate of the instilled PS was assigned a constant value, which was determined by the total amount and instillation time in clinical use according to Filoshe et al. [13]. For simplicity, the PS instillation was assumed to be evenly distributed at the proximal end surface of the liquid layer (pink part in Fig. 1). The assigned flow rate of the subsegmental model was proportional to the volume of the subsegment. The PS concentration of the liquid layer during instillation was computed by solving the advective–diffusion equation at each time step (0.01 s). The viscosity of the instilled PS was set as 30 cP [13]. The diffusivity of PS to existing liquid layer was assigned at 10^− 10 m²/s. The densities of the liquid layer and PS were assigned 1.0 kg/L.