Effects of mesenchymal stromal cell-conditioned media on measures of lung structure and function: a systematic review and meta-analysis of preclinical studies

Lung disease is a leading cause of morbidity and mortality. A breach in the lung alveolar-epithelial barrier and impairment in lung function are hallmarks of acute and chronic pulmonary illness. This review is part two of our previous work. In part 1, we demonstrated that CdM is as effective as MSCs in modulating inflammation. Herein, we investigated the effects of mesenchymal stromal cell (MSC)-conditioned media (CdM) on (i) lung architecture/function in animal models mimicking human lung disease, and (ii) performed a head-to-head comparison of CdM to MSCs. Adhering to the animal Systematic Review Centre for Laboratory animal Experimentation protocol, we conducted a search of English articles in five medical databases. Two independent investigators collected information regarding lung: alveolarization, vasculogenesis, permeability, histologic injury, compliance, and measures of right ventricular hypertrophy and right pulmonary pressure. Meta-analysis was performed to generate random effect size using standardized mean difference with 95% confidence interval. A total of 29 studies met inclusion. Lung diseases included bronchopulmonary dysplasia, asthma, pulmonary hypertension, acute respiratory distress syndrome, chronic obstructive pulmonary disease, and pulmonary fibrosis. CdM improved all measures of lung structure and function. Moreover, no statistical difference was observed in any of the lung measures between MSCs and CdM. In this meta-analysis of animal models recapitulating human lung disease, CdM improved lung structure and function and had an effect size comparable to MSCs.


Background
Pulmonary illness is a leading cause of morbidity and mortality [1]. In children, acute respiratory exacerbations are a common reason for primary care visits and are often implicated in hospitalizations [2,3]. Many of these pulmonary conditions result in impairments in lung function that may last into adulthood [4,5]. Consequently, identifying novel therapies for lung disease is highly warranted.
A unifying theme in many lung diseases includes inflammation [6][7][8]. While some inflammation is necessary to combat new disease and for proper wound healing, chronic inflammation may result in altered lung structure and function. During an acute illness, current therapies focus on restoring lung function by abating inflammation [9][10][11]. For instance, glucocorticoids are the mainstay therapy for reducing inflammation during acute exacerbations of asthma [12]. More recently, mesenchymal stromal/stem cells (MSCs) have shown encouraging outcomes in animal models of lung inflammation [13][14][15].
MSCs are promising agents as they are easily harvested, can be rapidly expanded, and can secrete factors (exosomes, microvesicles, microRNA) known to reduce inflammation [16][17][18]. The "secretome" or "conditioned media" of MSCs is considered biologically active and can be easily collected from the surrounding fluid of propagating cells [19][20][21]. Remarkably, preclinical studies suggest MSC conditioned media (CdM) may be as restorative as the MSCs themselves [22,23]. We supported this observation in a previous systematic review and meta-analysis demonstrating that CdM is as effective as MSCs in modulating inflammation [24].
This review is an extension of our previous work. In this review, we examined the effects of CdM on (i) lung architecture/function in animal models recapitulating lung disease and (ii) compare these findings to MSCs. Given that the therapeutic benefit of MSCs is attributed to a paracrine fashion, we believed CdM would have comparable effects to MSCs.

Overview and literature search
The methods in our review abide to those outlined by the Systematic Review Centre for Laboratory Animal Experimentation (SYRCLE) [25]. Our protocol was registered through the Collaborative Approach to Meta-Analysis and Review of Data from Experimental Studies (CAMARADES) [26]. Details are described in our previous publication.
We conducted a literature search in five databases using the following terms: mesenchymal stem cellconditioned media, lung disease, and animal. The last search was performed on March 17th, 2020. Three independent investigators evaluated titles and abstracts, followed by full-text review.

Inclusion criteria and outcomes of interest
We included studies administering MSC-CdM to animal models of acute lung injury or acute respiratory distress syndrome (ALI/ARDS), asthma, bronchopulmonary dysplasia (BPD), chronic obstructive pulmonary disease (COPD), cystic fibrosis (CF), pneumonia, pulmonary fibrosis (PF), and pulmonary hypertension (PH). Refer to Supplementary File 1 for the list of included studies.

Outcomes of interest
Measures of lung structure and/or function were our primary endpoint. Lung architecture and function were assessed under the following categories: alveolarization, vasculogenesis, right ventricular hypertrophy, fibrosis, permeability, pulmonary pressures, compliance, and lung injury. Although the pathogenesis of the included lung diseases are heterogeneous, we combined all processes irrespective of disease. This was conducted to obtain a scoping overview of the impact of CdM on biologic processes implicated in lung disease. Subsequently, we assessed lung structure/function by disease in our subgroup analysis. Excluded studies were those which did not provide data concerning our primary outcome of inflammation.

Data extraction
Three groups of investigators were used (ED and CE; RN and JM; ME, DM, and SM) to collect data. Uniformity of data was assessed by the primary author. This data included general study design, animal model characteristics, conditioned media characteristics, and outcomes of interest.

Data analysis
A random effects model was used to generate forest plots. A minimum of three studies were required for each outcome to proceed with a meta-analysis. The estimated effect size of CdM or MSC on lung architecture/ function was determined using standardized mean difference (SMD) with a 95% confidence interval (CI). Statistical heterogeneity between studies was calculated using the I 2 metric, and funnel plots were used to examine publication bias. If more than six articles were included per outcome, we conducted a subgroup analysis for disease, animal species, and route and dose of CdM administration. All statistical analyses were performed in R version 3.6.2; packages used included dmetar, metafor, and meta.

Study selection
Our literature search resulted in 245 articles. After removing duplicates and viewing the titles and abstracts, 55 articles underwent full-text review. Twenty-nine articles met inclusion (refer to Supplementary Figure 1). Table 1 summarizes the relevant study characteristics. Articles included in the review were published between the years 2009 to 2020. BPD was the most common animal model (n = 8), followed by ALI/ARDS (n = 5) and asthma (n = 5). All of the studies used rodents to induce their lung model. CdM vs. MSC: not applicable as less than three studies performed a head-to-head comparison.

All outcomes for lung structure and function combined
CdM: Supplementary Figure 8A shows the SMD of − 1.38 (with 95% CI of − 1.57, − 1.19) favoring CdM over control.

Subgroup analysis
Stratification of data was performed by lung disease, tissue source, dose, and route of delivery of CdM. Evaluation was performed if more than 6 studies had data.

Alveolarization
Supplementary Figure  animal models (SMD 1.67) and when the media was derived from cord blood (SMD 2.89), given at a dose of 7 μl/g (SMD 2.89), and delivered via the intraperitoneal route (SMD 1.56).

RVH
Supplementary Figure 10A-D depicts that CdM significantly improved RVH in BPD animal models (SMD − 0.93) and only when the media was derived from adipose tissue (SMD − 1.05), given at a dose of 100 μl (SMD − 1.14) and delivered intravenously (SMD − 0.86).

Fibrosis
Supplementary Figure

Vascularization
Supplementary Figure 12A-D shows that CdM had the greatest impact in animal models of COPD (SMD − 8.09), when the media was derived from adipose tissue (SMD − 2.61), given at a dose of 300 μl (SMD − 8.09) and delivered intravenously (SMD − 3.65).

Risk of bias
No study was judged as low risk across all ten domains. Eight studies stated that the allocation selection was random. Most studies (n = 25) had similar groups at baseline. Risk of bias was large regarding allocation concealment, whether authors mention random housing of animals, and blinding of caregivers or random selection of outcome. All studies were found to sufficiently report complete data and being free from other bias. Refer to Supplementary File 2 [27].

Publication bias
Supplementary Figures 13, 14, 15, 16, 17, 18, 19, and 20 illustrate publication bias through funnel plots. Overall, a b Fig. 7 Effect size of CdM (a) and MSC (b) on histologic lung injury. Forest plots demonstrate SMD with 95% confidence interval publication bias was low in all the outcomes except for lung permeability.

Discussion
Preclinical studies reiterate the ability MSCs have on dampening lung inflammation. This capacity is largely due to the paracrine secretion of MSC factors (microvesicles, exosomes) that provide a basis for future cell-free therapies for human disease [28][29][30][31]. This is the first review to directly compare the effects of CdM vs MSCs on lung structure and function in animal models of diverse lung disease. Overall, we found that CdM improved measures of alveolarization, right ventricular hypertrophy, lung fibrosis, vasculogenesis and permeability. Furthermore, CdM reduced pulmonary pressures, ameliorated histologic lung injury, and increased lung compliance. We found that CdM was comparable to MSCs in all lung measures evaluated individually and when combined.
The bioactive factors contained in the CdM of MSCs have been the focus of multiple studies and review articles [32][33][34]. Congruent with the findings found in this review, Hansmann et al. show that MSC-CdM, compared to CdM from lung fibroblasts, reversed alveolar injury, normalized lung function (airway resistance), and reversed RVH [35]. Additionally, the same group recently demonstrated that MSC exosomes (molecular cargo found within CdM) restored lung architecture, stimulated pulmonary blood vessel formation, and modulated lung inflammation [22]. In an E. coli pneumoniainduced ALI mouse model, MSC microvesicles (also found in MSC-CdM) reduced lung permeability and histologic injury score and were equivalent to MSCs [36]. Together, these findings, and those in recent a b Fig. 8 Effect size of CdM (a) and MSC (b) on pulmonary compliance. Forest plots demonstrate SMD with 95% confidence interval reviews, substantiate the results found in this review [37,38].
This  [39]. Similar to their results, this review showed that CdM had among the largest impact on measures of alveolarization and vasculogenesis, processes critical for appropriate lung healing, development, and function [40]. Although vasculogenesis/angiogenesis is an important process to restore lung function/structure, it can also enhance remodeling and thus worsen outcomes in other lung diseases such as asthma or pulmonary fibrosis [41]. In Supplementary Figure 12A, we demonstrate that this process improved in BPD, pulmonary hypertension, and COPD but was not assessed in asthma/pulmonary fibrosis.
In the study by Hayes et al., they found that MSCs were superior to CdM in a rodent model of ventilatorinduced lung injury. However, our review suggests that when you compile the literature, there were no significant benefits of using cells over CdM. We cannot explain why CdM was not comparable in this study; however, an important challenge that remains in the field includes the rigorous testing of key variables (tissue source, dose, route, disease, etc.) that may impact the quality of CdM [42][43][44]. For instance, we found that the intravenous route provided optimal results. Moreover, multiple administrations of CdM may augment vascular development, as seen in the study by Huh et al (n = 10 intravenous injections). Conversely, the optimal source and dose of CdM is dependent on the variable or the lung disease. This brings to light that it will be incredibly challenging to find a single CdM product that is ideal for all lung diseases. Thus, the idea of "one-size-fits-all" does not hold true for regenerative cells or products. Illustrating this concept, Rathinasabapathy et al. showed greater improvement in measures of RVH compared to other studies measuring right ventricular size. Important differences seen in the study by Rathinasabapathy and colleagues was that they used a different animal model (PH vs. BPD) and age of rodents (adult vs. neonatal) [45].
As investigators, we should attempt to tease out these characteristics in order to have the ideal product(s) for our lung disease of interest. In this way, we may have translational success in future clinical studies. Refining these features will take time but will play a vital role in efficacy. Moreover, pinpointing small and large animal models of lung disease that will recapitulate what occurs at the patient bedside is essential if we want to move the needle in the field [46].
The plausibility of using a cell-free product as a therapeutic agent for lung disease is substantiated by newly registered human clinical trials. For instance, NCT04235296 and NCT04234750 are evaluating safety of MSC-CdM in regulating wound inflammation and promoting wound healing in burn injury. Another Phase I trial (NCT04134676) plans to study the therapeutic potential of umbilical cord tissuederived stem cell CdM on chronic skin ulcers. Trials valuing the safety of stem cell CdM constituents (exosomes) are also underway for ischemic stroke (NCT3384433) and ocular conditions (NCT04213248, NCT03437759).
There are several limitations to our systematic review and meta-analysis, many of which mirror those published in our previous report. We incorporated multiple animal models of lung disease that have diverse pathologic processes resulting in their etiology. Also, most of the studies lacked methodologic details rendering them with an unclear risk of bias. Moreover, although preclinical models of lung disease have been helpful in identifying targetable mechanisms/processes, they oftentimes lack the intricacies of human disease. Thus, meticulous efficacy studies in large animals may be one approach to mitigate translational failure in human trials.

Conclusion
This review demonstrates that the administration of CdM in animal models of lung disease improves lung architecture and function. When compared to MSCs, CdM is as efficacious and provides a basis that cell-free products are a viable option for future studies. However, mores studies are needed to identify how specific variables (tissue source, route of delivery, concentration, etc.) may impact/strengthen their therapeutic potential.