Main functions of the lung as oxygen delivery and decarboxylation require unique structures within the lung tissue, providing physiological oxygen levels necessary for all human cell functions. Pulmonary functions are based on gas-exchanging units, i.e. alveoli mainly covered by epithelial cells. Together with the pulmonary capillary endothelium and the basal lamina inbetween, these alveolar epithelial cells (AECs, comprising two main cell types: pneumocytes type I, and pneumocytes type II) form a tight barrier. This so-called air-blood barrier enables an efficient gas exchange between air and blood, prevents the influx of protein-rich fluid into alveolar spaces, and protects the body from pathogenic microorganisms and environmental pollutants (1). Undisturbed gas exchange is crucial for maintaining oxygen supply and cell survival. Therefore air-blood barrier dysfunction is a critical pathological event leading not only to lung injury but potentially to patient morbidity and death (2, 3). Physiologically, maintenance of the barrier is guaranteed by anti-inflammatory reactions, involving complex interactions between and among immune cells and structural lung cells (4, 5). In particular, the interaction of AECs with alveolar macrophages (AMs) maintains a balance in the air-blood barrier, and enables self-repair processes at and of the barrier (6). However, aerogene stimuli like bacterial, viral, and fungal agents, as well as inhaled toxic particles and physical stress factors as pressure waves, trigger local and systemic pro-inflammatory reactions mediated by immune cells, such as dendritic cells (DCs) and AMs (7–9). This provoked release of pro-inflammatory mediators (e.g. tumor necrosis factor (TNF)-α, interleukin (IL)-1β, IL-8) as well as the production of reactive oxygen and nitrogen species (RONS) (10, 11) outweigh the anti-inflammatory processes at the air-blood barrier. Pulmonary inflammation plays an important role in the pathogenesis of acute pulmonary diseases, like pneumonia, as well as in the development and progression of chronic inflammatory disorders such as asthma and chronic obstructive pulmonary disease (COPD) (12, 13). Moreover, damage of AECs contributes to the development of acute respiratory distress syndrome (ARDS), one of the most destructive inflammatory processes occurring in the lung (14). Due to the vital role of the lungs in gas exchange, excessive inflammation in the lung tissue often leads to life-threatening conditions (15).

Lung injury is the primary cause of patient morbidity and mortality in many diseases, including coronavirus disease (SARS-CoV-2). The histopathological studies demonstrated a significant and profound alveolar damage and pneumonia, which progresses to ARDS and in the long-term to lung fibrosis in a multitude of patients (16). Meanwhile, the clinical manifestation of SARS-CoV-2 is paralleled by a sudden surge in pro-inflammatory cytokines known as “cytokine storm,” parenchymal loss, immune infiltration, and fluid-filled alveoli, altogether potentially resulting in an acute pulmonary failure and death (17, 18). Thus, to develop safe and effective therapies for infectious and inflammatory pulmonary diseases, it is crucial to understand cell-type-specific changes as well as the mechanisms and sequalae of the humoral immune responses in the lung. Additionally, reliable, well-defined experimental models with standardized conditions can provide a high-throughput therapeutical screening tools. Apart from in vivo models, several in vitro approaches have been developed to mimic the response to both pro-inflammatory stimulation and anti-inflammatory treatment, which range from simple cellular monolayers, multicellular models with primary or immortalized cell lines, primary tissue-derived organoids, alveolosphere cultures, to 3D multicellular systems of lung epithelial tissue. These approaches serve as a versatile and realistic toolkit for modeling and studying the pulmonary barrier function, and as screening tools for drug efficiency and safety. In the current review, we highlight different in vivo and in vitro models to examine pulmonary barrier, particularly during acute inflammatory conditions leading to lung injury.

Depending on the type of RONS that is generated, they can either cause oxidative damage to biomolecules and biological structures or contribute to intracellular signaling cascades (19). Early phase RONS, such as superoxide, hydrogen peroxide and nitric oxide (NO) are predominantly involved in signaling, while secondary RONS, such as peroxynitrite (ONOO) and hydroxyl radical are associated with oxidative damage to biological structures (19). Excessive production of RONS and oxidative damage may initiate a loop forward vicious cycle causing further damage to host cells (20, 21). This is due to the fact that induction of oxidative stress can further accelerate inflammation via NF-kB mediated pathway (22). Enhanced RONS generation and subsequent activation of oxidative stress-related pathways are the key processes causing damage to the pulmonary barrier via several specific pathological mechanisms (23). Thimmulappa et al. suggested eight oxidative stress-related mechanisms underlying pulmonary barrier damage: (1) lipid peroxidation, a process which disturbs the integrity of lipid bilayer of epithelial cells increasing their permeability, (2) RONS-induced activation of inflammatory response, (3) RONS-mediated activation of pathways involved in programmed cell death and more recently ferroptosis (24), (4) oxidative stress-mediated mitochondrial dysfunction, (5) endoplasmic reticulum stress-mediated disruption of protein synthesis (6) RONS-mediated epigenetic modifications via direct interaction with DNA/RNA, (7) activation of profibrotic mechanisms associated with endothelial cell dysfunction, and (8) airway mucus hypersecretion (23). AMs play a key role in the induction of these RONS-mediated pathways, since they are the first immune cells dealing with pathogens or foreign substances in the lung. They respond to bacterial/viral pathogens or their components such as pathogen-associated molecular pattern (PAMPs) or to substances released from damaged host cells (damage-associated molecular pattern, DAMPs) by releasing pro-inflammatory cytokines such as CXCL8, which upon release recruit neutrophils, the strong generator of ROS, into the lungs. AMs, AECs and neutrophils generate RONS predominantly via activation of NADPH oxidase (NOX)-2 located in the cellular membrane of these cells (25, 26). The activity of this enzyme is tightly linked to the intracellular generation of mitochondrial RONS. In addition to RONS derived from NOX and mitochondria, also xanthine oxidase and uncoupled NO-synthase (NOS) can contribute to excessive RONS generation occurring upon inflammatory conditions (27). Activation of both NOX-2 and NOS in turn stimulates the production of mitochondrial RONS and their release in cytoplasm via two discrete feed-forward loops (28). This interplay between intracellular and extracellular RONS generators can be beneficial in elimination of bacteria (28). However, the overproduction of intracellular RONS uncouples endothelial isoform of (e)NOS, which leads to formation of harmful peroxynitrite (ONOO^-) instead of NO. ONOO^- further aggravates oxidative stress and endothelial dysfunction (29). In addition to PAMPs and DAMPs, angiotensin (Ang)-II also activates NOX-2 via protein kinase (PK)C pathway (30), which leads to excessive RONS generation and pulmonary barrier damage particularly during SARS-CoV-2 infection. Therefore, oxidative stress and exaggerated inflammation are the major contributors to pulmonary barrier disruption which plays critical role in the pathogenesis of pulmonary diseases, including ARDS. Although pulmonary immune and oxidative homeostasis will not be further addressed in this review, it is of utmost importance to note that the pulmonary homeostasis is maintained by a complex network of tissue-resident cells as well as recruited immune cells. Incorporation of this complexity and orchestration into in vitro models is highly challenging but essential to determine und identify novel strategies for disease prevention and treatment. Therefore, it is essential to unfold this complexity by understanding different cellular and factoral entities of the pulmonary barrier and their incorporation into the experimental models.

Pulmonary microcirculation is an important component of the pulmonary barrier, and it is characterized by unique features differing from those of the systemic microvasculature. The pulmonary circulatory bed belongs to the low-resistance circulation area that must adopt to the cardiac output. Lung capillaries have extremely high density (comprising of ~600 billion capillaries) (31, 32) with considerably low flow and narrow diameter (33). Lung capillaries originate directly from relatively large arterioles and not only from arterioles with precapillary sphincters (34), therefore, they are less protected from increases in pulmonary arterial pressure and prone to edema formation (35). In contrast to the systemic circulation, pulmonary circulation responds to hypoxia by vasocontriction and not with vasodilation (36). Pulmonary microvessels temporarily entrap polymorphonuclear leukocytes (PMNs) during a physiological process called PMNs margination, whereby 2-3 times higher number of PMNs reside in lung microvessels than in the systemic circulation, reaching dynamic exchange and an equilibrium between the two compartments (37). Unlike systemic circulation, PMNs emigration into the alveolar space takes place not only in postcapillary venules, but also through the capillary network (38) and larger arteries (39). Another unique feature of the lung microvasculature is constitutive expression of adhesion molecules (P-selectin and ICAM-1) (40–42). Nonetheless, typical intravascular rolling of PMNs was not reported in lung capillaries (43), and the role of classical adhesion receptors involved in neutrophil recruitment (selectins and integrins) remains to be debate in the inflamed lungs. As opposed to the systemic circulation, PMNs adhesion occurs through both CD11/CD18-dependent and -independent ways, depending on the cause (e.g. the source of infection) (44, 45). In addition, the role of dipeptidase-1-dependent PMNs adhesion has been recently reported (46). Therefore, it is important to optimally model pulmonary vasculature in the experimental set ups and choose an appropriate method to examine changes induced upon acute injuries. In further chapters of this review, we provide an overview about possible in vitro models addressing this complex issue.

Figure 1 provides an overview on general mechanisms of pulmonary barrier damage.

Direct assessment of pulmonary microcirculation in septic patients is difficult to implement, thus suitable in vivo models are used to analyze several aspects of sepsis-related pulmonary dysfunction. In vivo examination of changes in microcirculation of the whole lung can be conducted by several methods including micro-computer tomography (CT) (47), and dynamic approaches such as single photon emission computed tomography (SPECT) (48) or microsphere techniques (49). If acquisition of spatiotemporal changes in circulatory and cellular inflammatory features of the microcirculatory compartment are of interest, different forms of intravital microscopy (IVM) could probably be a tool of choice (50, 51). These methods are based on conventional, one/two-photon or confocal fluorescence microscopic imaging methods as well as polarized light- and sidestream dark-field-based approaches in nearly all experimental animal species. During examination of the lung with IVM, observation of superficial microcirculation network of subpleural alveoli can be made possible via a surgically implanted thoracic window approach (52) supplemented with suctioning devices for stabilization purposes (53, 54). Yet, pulmonary motions and the beating heart cause fluctuations in vascular pressure and motion artifacts. Pulmonary movements could be overcome by taking recordings during end-expiration or after cessation of mechanical ventilation (55). Although the penetration depth of IVM methods is limited (<100 μm) (54), time-wise changes in microvascular diameters and capillary recruitment (55, 56), in cell-to-cell interactions (57–59), as well as alterations in endothelial glycocalyx (GX) integrity (60, 61) can be detected. Once IVM is combined with other methods, e.g. optical coherence tomography or multispectral oximetry, alveolar/airway dynamics (62, 63) and alveolar dynamics in parallel with capillary perfusion (55) can also be examined. Furthermore, subcellular process can be observed with two-photon microscopy at high resolution (54).

Leukocytes represent a particularly important first line of defense against airborne pathogens in the lungs. To date, IVM provides one of the best real-time methods to assess PMNs-endothelial and PMNs-platelet-endothelial interactions with regards to physiological PMNs margination (57, 64, 65). IVM-based studies have revealed the impact of leukocyte-platelet interactions (66) and their role in endothelium activation in the lungs (67, 68).

Although current experimental models are incomplete to recapitulate the major features of an acute lung injury and its clinical manifestation, the ARDS (69), the microcirculatory manifestations, ventilation- and sepsis-induced acute lung injury have been thoroughly examined in several studies. Numerous reports highlighted the impact of ventilation on microcirculatory processes in the lungs under experimental circumstances. Accordingly, respiration with high positive end-expiratory pressure (PEEP) values cause deterioration in capillary perfusion (56), and high respiration volumes evoke an exacerbation of endothelial-inflammatory cell interactions (70, 71). Ventilation-induced lung injury is also associated with the formation of neutrophil extracellular trap (NET) or NET-osis (72). Furthermore, increases in the lung microvascular pressure have been shown to increase cytosolic and intramitochondrial Ca²⁺ levels of endothelial cells (64, 73) with subsequent increase in pro-inflammatory cytokines and endothelium-derived adhesion molecule expression (64, 74). Looney et al. recently suggested that alveolar epithelium and vascular endothelium form a cellular syncytia favoring the spread of Ca²⁺ signal in a vectorial manner so as to achieve inflammatory communication in the multicellular environment within the lung, in response to mechanical forces (50).

Sepsis models are frequently used as acute lung injury models, and their relevance is marked by the fact that ARDS is a critical prognostic factor for mortality of clinical sepsis (75). In sepsis, lower pulmonary infections represent the leading cause of death (76). Further, diminished functionality of the microcirculatory perfusion is associated with increased mortality in septic patients. Notably, the ProCESS trial demonstrated an association between vascular density and De Backer score of sublingual microcirculation at 72 hours with 60-day mortality in septic patients (77). In experimental sepsis, CD11b/CD18-dependent PMNs accumulation in lung microvessels with a resultant mismatch in pulmonary ventilation-perfusion ratio could be visualized using IVM (59). Moreover, a link between heparanase-dependent process and the increase in PMNs adhesion has been demonstrated with IVM in the lungs of septic patients.

Moreover, IVM method allows the examination of endothelial surface layer GX which is in direct contact with the blood (78). Schmidt et al. showed that GX in the pulmonary endothelial cells appears to be thicker than in the systemic circulation, and is more vulnerable to sepsis than in other organs (60). Indeed, experimental endotoxemia/polymicrobial sepsis was associated with the degradation of GX (61). The relevance of these findings is underlined by observations whereby shedding of GX constituents from the injured endothelia into the blood stream was found to be important biomarker of pulmonary failure in patients undergoing septic shock (79). Elevated plasma levels of GX degradation products were also demonstrated in SARS/CoV/2 patients at a severe stage of the disease, and these changes were accompanied by GX damage in sublingual microvessels (80).

In summary, the functionality of the pulmonary microcirculation can be examined by several sophisticated methods of which IVM seems to be the method of choice for examination of pulmonary microcirculatory barrier upon lung injury. Despite an increasing understanding on the importance of microcirculatory changes in the progression of pulmonary diseases, so far, no specific microcirculation-targeted therapy reached a clinical significance. Nonetheless, real-time monitoring of disease progression and pulmonary microcirculatory manifestations upon therapeutic interventions may represent an important tool towards successful treatment of acute lung injury.

The structural extensity of airway epithelial cells and their crucial functions in the pulmonary barrier highlight the need to perform a comprehensive assessment of pulmonary mechanics in the in vivo studies. Animal models cover a wide range of pathologies including pulmonary fibrosis (81), smoke-induced chronic obstructive lung disease (82), neonatal chronic lung disease (83, 84), ventilator-induced lung injury (85, 86), viral and bacterial infections (86, 87). These models involve both airway and parenchymal alterations forming a complex scenario. Although the airway and lung parenchymal compartments are structurally connected, and operate together to maintain the pulmonary mechanics (88), characterizations of airway/parenchymal mechanics in terms of resistance (R) and elastance (E, the reciprocal of compliance) remained underrated.

Methodological precision and non-invasiveness tend to oppose each other, and this is referred to as the “phenotyping uncertainty principle” (89). At one extreme, the popular whole-body plethysmography in unrestrained animals was suggested to measure airway reactivity (90); however, this method has been shown to reflect more the changes in breathing pattern than the mechanical properties of the lungs (91). In contrast, the low-frequency oscillometry technique (LFOT) (92), which is considered the most sophisticated and selective assessment of the airway and tissue mechanics to date (93, 94) requires anesthesia and tracheostomy or intubation/cannulation of the trachea.

There are several other methods overviewed (89, 94). A robust traditional approach in mice (95) is the fitting of the transpulmonary pressure to a multiple-linear model of flow- and volume-dependence to obtain values of R and E. The main limitation of this technique is that the airway and tissue resistances are combined in the value of R. Attempts have been made to implement the widespread human pulmonary function tests such as the measurement of forced vital capacity (FVC) and forced expiratory volume in 1 s (FEV1) in small animals (96). However, the value of this implementation is dubious as increasing evidence challenges the sensitivity and structural specificity of the FEV1/FVC test (97).

LFOT employs small external pressure or volume oscillations during apnoeic intervals to estimate respiratory impedance (Z) at multiple frequencies covering the respiratory rate of the species (92, 98). The “constant-phase” model (CPM) of lung mechanics (92) is fitted to the measured Z values to obtain the parameters of airway resistance (Raw) and inertance, tissue damping (G) and elastance (H). The unique feature of this model is that G and H characterize the viscous losses and energy storing, respectively, of the respiratory tissues over a wide range of breathing and oscillation frequencies. The CPM has been validated in multiple studies using different resident gases, demonstrating the clear separation of the airway and tissue compartments. It has also been shown that non-uniform behavior of the lungs lends a virtual component to G (99). Extension of the CPM to express inhomogeneity via distributions of peripheral airway resistances and/or parenchymal units (100–103) has added further flexibility to cover diverse pathological alterations in lung mechanics. A large number of studies have contributed to establishing normative values for the CPM parameters in different species (93). Over 4 decades of body weight from 2-wk-old mice to adult humans, Raw, G and H arrange on the same trajectories and the log-log plots display high correlation coefficients (r² = 0.91-0.98), while the G/H ratio expressing the mechanical efficiency or hysteresivity (103) of the lung tissue remains at the level of ~0.2, largely independently of the lung size (93). Thus, the LFOT-CPM method can be suggested as a universally accepted approach, yet with a parsimonious number of parameters, to evaluate lung mechanics in animal models of pulmonary diseases.

Majority of the methods on lung mechanics are used on the whole respiratory system. Therefore, the results include the mechanical impedance of the chest wall, unless thoracotomy or oesophageal pressure measurement is performed. In larger mammals, chest wall properties may mask the changes occurring in the lungs, whereas in small rodents with compliant chest wall structures, especially in mice, this error may be negligible.

Finally, in many models of pulmonary disease the temporal dynamics of the development of pathological changes may become a primary aspect. Repeated studies are technically feasible in mice (104) and rats (105) and they can be combined with high-sensitivity lung function techniques to properly address the questions raised in the animal models.

In the past years, several in vitro models of the air/blood barrier were developed. The principal aim was to enable the use of a solid method for investigations of the air/blood barrier and determine the potential impacts of harmful exposures. In contrast to submerged 2D culture models, the air/blood model mimics the air surface side of the alveolar epithelium. As cells of the pulmonary system show an environment-dependent differentiation behavior, it is crucial to mimic this air/interstitial environment. Airway cells cultured in the air-liquid interface (ALI) show significant morphology and barrier characteristics compared to the submerged culture condition (106). Therefore, ALI models are particularly exploited to investigate the impact of various exposures on the barrier integrity by using several immortalized cell lineages (107). In the ALI model, cells are grown on transwells or biological scaffolds (108), where the basolateral side is in contact with cell culture media, whereas the apical side of the cells is exposed to the air (109). The currently used cell lines are either of alveolar origin (A549 (110), hAELVi (111), hAEpC (112), and NCI-H441 (113), H358 (108), HCC827 (114) or of bronchial origin [Calu-3 (115), BEAS 2B and 16HBE 14o^- (116), primary human airway epithelial cells (pHAECs) (117)]. Interestingly, pHAECs were successfully grown in miniaturized transwell plates (118). These primary human cells are superior to immortalized cells in terms of cell behaviour. However, they lack comparability due to donor-dependent attributes. This limitation was overcome by ALI differentiation of human-induced pluripotent stem cells into AEC II (119). Given the lack of blood supply in ALI model, modification made by co-culturing capillary endothelial cells, including HPMEC (114, 120) and EA.hy 926 (121) with AECs to optimally recapitulate the air-blood barrier (122). Combination of both endothelial and alveolar components allows optimal assessment of the cell-cell interaction during pulmonary inflammation (114, 123). For instance, the impact of bacterial component lipopolysaccharide (LPS) (124, 125) or human inflammatory cytokines like initerferon (IFN) or TNF (118) on the barrier is evaluated by ALI model. In such models, it is possible to include leukocytes in addition to the alveolar cells (126, 127) and endothelial cells in a triple culture to optimally mimic disruptive barrier conditions during inflammation (120, 128).

Although 2D monolayers or well-differentiated airway epithelial cells in ALI are commonly used to investigate pulmonary inflammation (129, 130), these models undermine the impacts of inflammatory cells recruiting to the site of inflammation. To close this gap, primary human vascular endothelial cells (HMVEC-LB1), human white blood cells (WBC), and their co-cultures can be used to evaluate their inflammatory responses to various inflammatory stimuli (131). HMVEC-LB1 cells can be applied as an in vitro model to examine pro-inflammatory responses to inflammatory trigger by measuring the production and/or release of IL-6, soluble intercellular adhesion marker (sICAM)-1, or soluble vascular cell adhesion marker (sVCAM)-1 (131). Furthermore, an alveolo-capillary barrier in vitro model consisting of a co-culture system of human distal lung epithelial cells and HMVECs was established (113, 132) to assess epithelial-endothelial interactions in context of acute lung injury (113). Functional cellular junctions are formed in these models by a tight epithelial barrier similar to in vivo conditions. More specifically, the in vitro co-culture system consist of monolayers of human lung epithelial cell lines (A549 or NCI H441) and HMVECs on opposite sides of a permeable membrane. Although A549 failed to show sufficient differentiation with respect to formation of a tight epithelial barrier with intact cell-cell junctions (113), the co-cultures of NCI H441 and HPMEC established differentiated monolayers after stimulation with dexamethasone by forming tight junctional protein, ZO-1 and the adherens junction protein, E-cadherin (113). Furthermore, the model induced a polarized epithelial cell monolayer with typical junctional structures as confirmed by transmission electron microscopy. Taken together, such a co-culture system constitutes a suitable in vitro model to examine e.g. functional barrier epithelial, and epithelial/endothelial interactions in the pathogenesis of acute or toxic lung injury and inflammatory lung diseases.

Simple monoculture studies have enabled researchers to understand the abnormal phenotypes of lung cells in diverse pathologies. For example, 2D models are well established for the research of lung function, as they can fully differentiate to airway cells as well as AEC I-like and AEC II-like cells (133). Although 2D cultures are well established in pulmonary epithelial barrier studies, in vitro co-culture models are advantageous in recapitulating the complex structure and function of native lung by mimicking the multicellular interactions. Such co-culture models aiming to assess cellular communication in the lung include conditioned medium-exposure experiments, epithelial-mesenchymal co-cultures, and several 3D co-culture models. 3D cultures can be further differentiated into so-called “human lung organoids” (HLOs). These lung progenitors can give rise to proximal airway-like structures, as well as to alveolar cell types (134, 135). Furthermore, biological and synthetic scaffolds like decellularized rat and human lung slices or whole organs can be used to develop a more functional lung cell response to inflammation (136–138). These models provide insights into cellular communication either directly via assessment of cell-to-cell interactions or indirectly via assessment of the release of soluble mediators which mimic in vivo conditions in the pulmonary system. Nevertheless, various models require careful assessment of cell entities as well as used culture media to ascertain the viability and functionality of co-cultured cells.

Extensive understanding of mechanisms underlying pulmonary barrier dysfunction is pivotal for the development of therapeutical strategies in acute lung injuries. To better understand molecular mechanisms leading to both deterioration and consecutive dysfunction of pulmonary barrier several in vitro and in vivo models have been developed. In this review, we highlighted the development and applications of these models, including 2D mono- and multiple cell-types, co-culture models, 3D air-liquid interface models, 3D bioprinting models, lung-on-a-chip, precision-cut lung slices, and lung organoids. These models may improve our understanding of the complex pulmonary homeostasis under regular and pathophysiological conditions by mimicking the in vivo environment. Given the increasing number of studies applying these models, they represent valuable platforms to assess cell-to-matrix, cell-to-cell interactions as well as the role of exogenous and endogenous inflammatory or danger associated molecules in lung damage and repair mechanisms. The use of such complex 3D models will further improve the incorporation and assessment of specific cellular phenotypes and humoral factors by sustaining cell functionalities in a manner that resembles the in vivo conditions to elaborate their impact on morphology, toxicity, viability, barrier integrity, ciliary beating frequency, mucociliary clearance, as well as cytokine release. The understanding of complex interactions between lung structural cells, (immune) environment, and cellular matrix in the lung micromilieu will provide further options to investigate novel therapeutics for targeting different pathogenic mechanisms in acute lung injury.

The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author.

All authors listed have made a substantial, direct, and intellectual contribution to the work, and approved it for publication.

AVK was supported by the Ludwig Boltzmann Institute Traumatology. The Research Center in Cooperation with AUVA Austrian Cluster for Tissue Regeneration. ZH was supported by Hungarian NKFI-OTKA grant K128701. SG was supported by the Federal Ministry of Education and Research (BMBF), The STIMULATE Research Campus, grant number 13GW0473A. SMC was supported by the Federal Ministry of Education and Research (BMBF), grant number 03Z22JN12 to SC. BR was supported by the Federal Ministry of Education and Research (BMBF), The STIMULATE Research Campus, grant number 13GW0473A to BR.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.