In this secondary analysis of a double-blind placebo-controlled multicenter clinical trial, we investigated longitudinal proteomic perturbations associated with critical illness and assessed the effects of imatinib treatment in hospitalised patients with COVID-19. Proteomic analyses were performed using a wide-angle approach that measured 6385 unique proteins at hospital admission and three days thereafter. We found that critical illness was associated with a dysregulation of both canonical acute phase response pathways, as well as previously unreported pathways. These newly identified pathways included, but were not limited to, cell adhesion, extracellular matrix turnover, TGF-β and FGF signalling. Moreover, we found that the plasma proteome at hospital admission provides reliable and reproducible information on critical illness development and prognosis in hospitalised COVID-19 patients. Dysregulations in extracellular matrix turnover and focal adhesion were attenuated by imatinib treatment in the human plasma proteome. Analysis of external RNA-seq data on extracted lungs of SARS-CoV-2 infected hamsters validated that imatinib attenuates these dysregulations in the lungs. Taken together, this study reveals that the alveolar-capillary barrier is disrupted in critical COVID-19 illness, which is attenuated by imatinib treatment.
Numerous studies have investigated the association between disease severity and the blood proteome in COVID-1912,13,26–42. According to these studies, the most frequently observed and relevant disruptions observed in the proteome of critical illness include: neutrophil degranulation, cell cycle and transcriptional regulation and complement, coagulation, interferon, chemokine and interleukin signalling pathways (i.e., predominantly pathways involved in the acute phase response). However, few proteomic studies have identified proteomic profiles associated with tissue injury specifically related to ARDS34,36. The failure to detect these may result from low patient number and power to detect subtle changes or from the use of smaller protein panels that specifically detect certain subsets of proteins (e.g., targeted panels of inflammatory or cardiovascular-related proteins). Most studies do not collect serial samples, which is crucial since pathobiological processes in COVID-19 are time dependent1. In our study, patients from 13 hospitals throughout the Netherlands were systematically followed according to a strict prospective sampling and clinical data registration scheme. The patients in our study comprised a relatively homogenous population, with 92% of patients receiving oxygen with a nasal cannula or mask at hospital admission. Patients were systematically followed for 90 days after randomisation. This is the first study to test the effect of a pharmacological intervention in a randomised setting on pathways associated with critical outcomes. This permitted an evaluation of whether identified pathways are causally involved in imatinib treatment and critical disease development. Therefore, this study may serve as an example for future studies evaluating the change in biology underlying novel interventions.
The most significant and relevant novel pathways associated with critical illness in our study can be divided into three main groups according to their functions 1. cell-cell junctions and cell-matrix interaction (as evidenced by proteins regulating focal adhesion and adherens junctions such as integrins, ICAMs, IBSP, RGMA); 2. ECM turnover (e.g., collagens, biglycan, thrombospondin-2, nidogen-1, LAMA4), which are crucial for maintaining structure of alveoli and integrity of the alveolar-capillary barrier function; and 3. FGF (e.g., FGF4 and FGF8) and TGF-β (e.g., TGFB1, TGF-β1, TGF-β3, BMP10) signalling pathways that regulate tissue remodelling and repair. The alveolar-capillary barrier consists of epithelial cells, a thin collagen-rich basement membrane and a monolayer of endothelial cells43. Permeability across epithelial cells is tightly regulated by intercellular tight junctions and adherens junctions that anchor the cytoskeletons of neighbouring cells44. In severe COVID-19, the lung’s epithelial and endothelial barrier function is compromised resulting in accumulation of protein rich fluid into the alveoli, a phenomenon that similarly occurs in non-COVID ARDS24,45,46. While intercellular gap formation and hyperpermeability were deemed central in the development of inflammatory pulmonary oedema47, ultrastructural analyses in COVID-ARDS have shown additional and more severe forms of injury to the alveolo-capillary barrier including detachment and loss of alveolar endo- and epithelial cells as well as turnover of the extracellular matrix16,48,49. This results in the release of collagen, integrins, nidogens, proteoglycans and laminin from the alveoli into the bloodstream24,34,50. In our study, we found that these were measurable in plasma and were associated with the development of critical illness. Following alveolar-capillary destruction, repair and regeneration processes such as TGF-β and FGF signalling are required51–54. However, these repair processes are often impaired or overactivated in severe viral infections, contributing to the activation of coagulation and fibrosis51,55. Histopathological analyses confirm that the alveolar-capillary barrier is disrupted and that the FGF and TGF-β signalling repair mechanisms are impaired in severe COVID-1916. In line with previous studies, we observed that in critical illness the plasma proteome reveals enhanced extracellular matrix turnover, impaired intercellular adhesion and inappropriate impair processes16,34, 48–50.
We identified 538 proteins that had prognostic value for critical illness progression and could thus be used for risk stratification. We refined the prognostic signature to nine proteins with relatively good performance using LASSO regression, which imposes model sparsity and avoids collinear predictors that convey redundant information. Most importantly, this small protein panel was superior to conventional blood and clinical parameters for risk stratification in the validation cohort. Previous studies observed that using prediction models, the proteome can accurately classify patients with critical illness29,31,35,56. However, no studies have tested its performance for risk stratification at hospital admission or did not include a homogeneous population at baseline.
Comparing pre-treatment (admission) versus post-treatment samples (day 3) in a randomised setting gave us the opportunity to reliably assess the effects of imatinib. Compared to profiles associated with critical illness, the most significant processes attenuated by imatinib were 1. TNF signalling and acute phase response (e.g., IL-6, PLAU and CXCL12); and 2. focal adhesion and ECM turnover (e.g., collagens, PLAU, thrombospondin-2 and biglycan). In a secondary analysis of our study in hospitalised patients treated with imatinib using multiplex assays, we observed that imatinib decreased plasma levels of IL-6, E-selectin and SP-D, indicating attenuation of inflammation, endothelial cell adhesion and epithelial injury.6 Dampening of inflammation in this study was evident from inhibition of MAPK8, MAPK9, IL-6, CXCL12 and the serine protease urokinase-type plasminogen activator (PLAU or its alias uPA). After binding of uPA to its receptor, uPAR, it converts plasminogen to plasmin which induces a proteolytic cascade resulting in extracellular matrix turnover57. Its action is modulated by other serine protease, of which serine protease 27 and serpin B13 levels were suppressed by imatinib treatment in our study57. Moreover, IL-6 and uPA are both modulators of vascular permeability and could therefore aggravate vascular leak58,59. Elevated activation of the uPA/uPAR system is associated with poor prognosis in COVID-19, but it could also indicate a good response to the recombinant IL-1 receptor antagonist anakinra60,61. Proteins with important involvement in extracellular matrix organisation and adhesion include thrombospondin-2 and collagens type I and III, which are most abundant in the alveolar wall and alveolar septa62. The effect of imatinib on type 4 collagen, the most abundant collagen in the basement membrane, was not measured because no aptamer was available62. Thrombospondin-1 and thrombospondin-2 are matricellular proteins involved in multiple processes including cell-cell and cell-matrix interactions, extracellular matrix turnover, wound repair, tissue remodelling and regulation of angiogenesis23,25. Thrombospondins interact with integrins, ECM components (e.g., fibronectin, laminin), matrix metalloproteins and FGF and TGF-β signalling23,25. In light of our findings of dysregulations preceding critical illness in COVID-19, thrombospondins could play a central role in the alveolar capillary barrier and repair mechanisms. The FGF and TGF-β pathways were not significantly affected by imatinib treatment, although some proteins affected by imatinib have the potential to regulate FGF and TGF-β signalling (i.e., biglycan, thrombospondin-2, RGMA, TGFBR3)23,25,63,64. The above findings indicate that imatinib attenuates SARS-CoV-2 induced inflammation and disruption of the alveolar-capillary barrier in severe COVID-19 patients.
In the study performed by Xia et al. imatinib attenuated SARS-CoV-2 induced inflammation and precluded death in Syrian hamsters14. In contrast, histopathologic evaluation mock-treated hamsters showed widespread destruction of the lung parenchyma, marked by accumulation of pulmonary edema, inflammation and increased transcription of ECM components. These processes reflect the acute inflammatory responses of ARDS including disruption of the alveolar–capillary barrier. During acute inflammation, macrophages, fibrocytes, fibroblasts and myofibroblasts accumulate in the alveolar compartment, leading to excessive deposition of fibronectin, collagens I and III and other components of the extracellular matrix65. This is referred to as the fibroproliferative response, which results in the formation of a provisional ECM65. Pathway analysis on the RNA-seq data indicated that imatinib attenuated enhanced extracellular matrix turnover, impaired cell adhesion and TGF-β and FGF signalling in the pulmonary compartment. In line with the findings from the human plasma proteomics analysis, imatinib treatment reduced uPA and IL-6 levels, indicating that imatinib attenuates both pulmonary and systemic inflammation. Taken together, the RNA-seq analysis validate that imatinib treatment attenuated alveolar capillary disruption and dampened the fibroproliferative response in the lungs.
Our study has two important limitations. Despite the large number of proteins measured, only those for which an aptamer was available were included. Considering that the human proteome consists of approximately 20.000 different proteins, we can potentially overlook important proteins66. Another limitation is that we only collected plasma samples, as the collection of bronchoalveolar lavage fluid (BALF) or tissue from human patients was either not appropriate or not feasible. The proteome in tissue and BALF is distinct from that in blood and provides more information on biological processes in the bronchoalveolar compartment67,68. To gain more insight into the effects of imatinib on the lungs, we analysed the RNA sequencing data from extracted lungs of Syrian hamsters who were treated with imatinib after SARS-CoV-2 infection. Pathological manifestations of COVID-19 show many similarities with severe COVID-19 in humans69, however hamster models can still only be used as a surrogate for humans. Moreover, the comparison is somewhat limited because transcriptomics rather than proteomics was applied.
Our findings could also have implications in other fields. Patients with critical COVID-19 share many similarities with non-COVID-19 ARDS patients, although COVID-19 ARDS is associated with more alveolar edema45. Impaired TGF-β signalling, ECM turnover and cell adhesion play pivotal roles in alveolar capillary barrier disruption and thus alveolar oedema formation70,71. Stratification between non-critical and critical COVID-19 closely approximates stratification between patients who do or do not develop a (COVID-19) ARDS phenotype. The effects of imatinib could also implicate a beneficial effect for the treatment of diseases with increased vascular remodelling such as pulmonary arterial hypertension (PAH).
Imatinib modulated a substantial number of proteins that are taking part in the pathobiology of PAH. These include thrombospondin-2, collagens, biglycan and members of the TGF-β superfamily, and are involved enhanced tissue remodelling and extracellular matrix turnover of pulmonary arteries in PAH72,73.