The study included individuals with MPM, treated with nonradical surgical treatment, such as partial pleurectomy/decortication (P/D) or biopsy only, and platinum-based plus pemetrexed chemotherapy for MPM, followed by high-dose radiation therapy with curative intent, and followed up prospectively for development of (pulmonary) toxicity. Other inclusion criteria were age ≥18 years, proven gross residual disease after surgery, stages I–IVA (according to TNM stage 7th edition), Eastern Cooperative Oncology Group (ECOG) performance status scores 0–2, pulmonary function of at least 50% of predicted, technical feasibility for delivery of radical hemithoracic radiotherapy (RHR), satisfactory bone marrow function (white blood cells ≥2,000/μl, platelets ≥100,000/μl, haemoglobin >10 g/dl). Exclusion criteria were pathologic contralateral mediastinal nodes (N3), metastatic MPM (stage IVB), or intra-scissural disease. Tumour histology was classified as epithelioid and non-epithelioid (sarcomatoid and biphasic). All patients were staged by lung and abdomen contrast-enhanced computed tomography (CT) scans and a 18F-fluorodeoxyglucose (18FDG)-positron emission tomography (PET)/CT.

Between August 2014 and May 2018, 49 patients who received RHR were included in this study. Patients were treated with helical IMRT, delivering the dose with Accuray Tomotherapy System. The radiation therapy technique has previously been described in detail (7, 8, 11). Patients received 50 Gy in 25 fractions (except for one case, later described), plus an eventual boost to 60 Gy on PET-positive areas.

OS was defined as the time (years) intercourse since randomization to death from any causes, or the last follow-up (until December 2020, 1st), and estimated with the Kaplan-Meier method, p-value was calculated with Log-rank test.

This prospective study was conducted according to the ethical principles of the Declaration of Helsinki and approved by the local Ethical Committee (Comitato Etico Indipendente del CRO di Aviano, CRO-2013-38) and written informed consent was obtained from all the patients. The CRO-biobanking service managed and stored all biological samples before use for the present project (authorization for analyses obtained through protocol number 6825/D).

Toxicity was assessed using the Common Terminology Criteria for Adverse Events version 3.0, and divided into early toxicity (during treatment), acute toxicity (1–6 months from the end of RHR), and late toxicity (>6 months from the end of RHR).

Early toxicity was assessed weekly during radiation. Following completion of radiation treatment, patients were reviewed at 1, 3, 6 (acute toxicity), 8, and 12 months in the first year (late toxicity) and every 4 months from the start of second year or before for clinical need. All respiratory toxicities outcomes were analyzed: cough, dyspnea, fibrosis, pulmonitis, and pulmonary thromboembolic events.

Patients’ characteristics are summarized in Table 1 . A total of 48 patients completed the treatment as prescribed. One patient, requiring a treatment break due to anemia and trombocytopenia, globally received 32 Gy in 16 fractions. One patient developed infective pneumonia independent of RT, and discontinued the treatment, completing the prescribed course at a later moment. Forty-five patients received the simultaneous boost.

Reported adverse effects are summarized in Table 2 , divided in events observed during treatment (early toxicity), between 1 and 6 months after the end of RT (acute toxicity), and after 6 months of follow-up (late toxicity), and classified as nonpulmonary or pulmonary toxicities.

Globally, for acute and late toxicities, the number of events in the nonpulmonary toxicity group tended to diminish, except for the rise in liver gamma-glutamyltransferase (GGT), commonly observed in patients treated on the right hemithorax (not shown). Fractures and pericardial effusions are typically late events and were observed in a small number of patients, 2 and 3 cases, respectively. Conversely, pulmonary toxicity events tended to develop after treatment and to last in time. One-third of the patients developed low-grade cough by 6 months after RT, or at a later follow-up. Dyspnea was reported in 37% of patients as acute toxicity, mainly low grade, but became a severe toxicity in 8% of patients, as a late event. Fibrosis developed in 3 patients (2 severe), and pneumonitis in 7 cases, 4 of which were grade 3. Finally, 3 thromboembolic events were reported after treatment.

Due to the impact of grade 3 pulmonary toxicities on the quality of life, for further genetic analyses patients were divided into two groups: no severeTox (NoSTox), none or tolerable toxicity (grade 0 to 2) (grade ≤2) (N = 40), and severeTox (STox) (grade = 3) (N = 9) based on the development of at least one among the following: cough, dyspnea, fibrosis, pneumonitis, and/or thromboembolic event. STox patients’ characteristics and reported adverse effects are summarized in Table 3 . No grade 4 or 5 toxicities were recorded during treatment.

Among the 49 patients included, median OS was 25 months, and 2-year OS rate was 51.4%. No significative survival differences were observed between the NoSTox group and the STox group (data not shown).

The global genetic analysis results and all exome detailed reports are shown in Supplementary Datas 1, 2 . Supplementary Data 1 contain the descriptive information about exomic analysis of the 49 patients with MPM; the distribution of variants ( Supplementary Table S1A ), distribution (%) of variant location ( Supplementary Figure S1A ), function ( Supplementary Figure S1B ), and distribution of genes carrying variants ( Supplementary Table S1B ) are reported in the exome report of both groups (STox and NoSTox). Supplementary Data 2 contain a comparative description about genes and common variants of 49 patients with MPM; NoSTox and STox common variants were filtered to retrieve only the common FunctVar ( Supplementary Table S2A ) and ImpactVar ( Supplementary Table S2B ) variants exclusive of the STox group; similarly, NoSTox and STox common variants were filtered to retrieve the common FunctVar ( Supplementary Table S2C ) and ImpactVar ( Supplementary Table S2D ) variants present just in the NoSTox group. Noteworthy, none of the genes reported by Hylebos and colleagues (28), to exhibit molecular alterations in MPM, were related with the development of pulmonary toxicity in the subjects involved in this study (data not shown).

Based on the results obtained considering variants present just in the STox patients, we retrieved the pathways possibly disrupted by the identified genetic variants, without, however, detecting a specific pathway signature associated with the clinical phenotypes observed (data not shown).

Furthermore, with the aim of applying a wider approach to enrich the pathway analysis, we performed a variant enrichment analysis based on functional and impact variants; 47 pathways showed a significantly increased number of functional variants exclusively in the NoSTox group ( Table 4 ). For the STox group, we observed four exclusive pathways ( Table 5 ). Regarding the impact variants, we observed 19 and 3 exclusive pathways for the NoSTox and STox groups, respectively ( Tables 6 , 7 ).

Tracing back the variants possibly involved in those pathways, we noticed 243 functional variants (along 64 genes), for the NoSTox group, and 9 variants for the STox (in 5 genes) ( Supplementary Data 3 : Table S3A, B ). When considering the pathways with impact variants, we identified 2,908 variants (in 645 genes) present in the pathways exclusive of the NoSTox group, while only one variant (one gene) was observed in the pathways of the STox group ( Supplementary Data 3 : Tables S3C, D ).

Despite the genetic differences and the number of diverse variants found between NoSTox and STox, principal component analysis (PCA) showed a homogeneous distribution of variants in both groups. Thus, no specific genetic signature was observed in NoSTox and STox ( Supplementary Figure S3 ). However, in all analyses, patient 53 showed relevant significant genetic distance compared with all other patients regardless of whether they were NoSTox or STox, suggesting the presence of a specific genetic signature, but not influencing the response to radiotherapy.

Four pathways, related to inflammatory and fibrotic processes, possibly involved in the development of lung toxicity, were in silico disrupted in the noSTox patients. Other disrupted pathways were related to more general biological processes not specifically related to the patients’ phenotypes. The four pathways are as follows:

1. TNFR1-induced NFkappaB signaling pathway (R-HSA-5357956)

Tumor necrosis factor alpha (TNF-α) is rapidly and persistently expressed in irradiated and adjacent tissue (34). It can trigger multiple signal pathways (35), including the signaling activity of NFkB, which has a strong proinflammatory action/function, and thus needs to be strictly controlled to prevent persistent inflammation. One mechanism that contributes to ensuring proper control of NFkB includes ubiquitination (36, 37). TNF-α has been implicated in radiation mucositis, enteritis, and dermatitis (30, 38–40), and its deficiency in a lung injury model has been shown to prevent symptoms of radiation pneumonitis (41).

A functional variation is found in TNFR1-induced NfkappaB signaling pathway: a frameshift insertion in the UBC gene. The protein encoded by this gene is a ubiquitinase (ubiquitin C). Although overall changes in ubiquitin cellular levels in response to ionizing radiation are still little known, Tang and colleagues showed that downregulation of ubiquitin C reduces radiation-induced expression of NF-kB and inhibits its translocation into the nucleus (42), suggesting that upon irradiation, de-ubiquitination silencing may suppress NF-kB-induced cellular response to radiation.

2. The AIM2 inflammasome pathway (R-HSA-844615)

A frameshift deletion in the AIM2 gene was detected in patient #16; also, a variant on PYCARD was observed in patient #84. AIM2 is an innate immune sensor that mediates assembly and activation of the inflammasome in response to double-stranded DNA breaks (43, 44). The DNA-sensing AIM2 inflammasome controls radiation-induced cell death and tissue injury (45). In a rat model of radiation pneumonitis (RP), radiotherapy increased the mRNA expression level of AIM2, which further triggered the release of IL-1β and induced RP (40), suggesting that the activation of the AIM2 inflammasome by radiotherapy may contribute to the pathogenesis of RP. Interestingly, AIM2 inflammasome deficiency has been shown to protect mice from radiation-induced small intestine syndrome as well as hematopoietic failure (45). Furthermore, the PYCARD (also known as ASC) gene encodes an adaptor protein that mediates assembly of large signaling complexes representing a critical step in initiating inflammasome responses (46).

3. The insulin receptor signaling cascade (R-HSA-74751) pathway was disrupted in 4 NoSTox patients

There is a high degree of structural homology between the insulin receptor (IR) and the insulin-like growth factor 1 (IGF-1) receptor (IGF-1R), which signals via many common mediators (47). Ionizing radiation activates several tyrosine kinase receptors involved in DNA damage response, including IGF-1R (48), possibly promoting a profibrotic state. Furthermore, dysregulation of the insulin-like growth factor (IGF) axis has been implicated in the pathogenesis of fibrosis in lung diseases (49–52). Insuline-like growth factor 2 (IGF-2) promotes fibrosis through IGF1R, IR, and IGF1R/IR, differentiates fibroblasts into myofibroblasts, decreases protease production and extracellular matrix degradation, and stimulates expression of two transforming growth factor β (TGF-β) isoforms, suggesting that IGF-2 may exert profibrotic effects via multiple mechanisms (53).

4. Plasmalogen biosynthesis (R-HSA-75896)

Recent evidence suggests that cholesterol-sphingolipid rafts might play a role in insulin signaling via IR (54). The potential contribution of lipids to insulin signaling is intriguing since the plasmalogen biosynthesis pathway has also been shown to have a potential impact on NoSTox patients’ phenotypes, carried by patients #50, #66, and #88 who did not develop severe pulmonary toxicities. Plasmalogens represent a class of phospholipids ubiquitously found in considerable amounts as constituents of mammalian cell membranes and significantly enriched in lipid rafts (55, 56), functional microdomains in cell membranes, which may affect signal transduction. Therefore, disrupting plasmalogen biosynthesis could affect trans-membrane signaling of insulin, thus making a functional bridge between this pathway and the pathogenetic mechanisms of radiation toxicity. Furthermore, Laiakis and colleagues (57) investigated the effects of ionizing radiation on the blood level of metabolites and lipids, showing that acute exposure to γ radiation in a mouse model induced specific mobilization of plasmalogens. Though the pathological role of plasmalogens and the biochemical pathways underlying their upregulation after radiation exposure remain to be elucidated, Braverman and Moser suggested a specific serum lipidomic biosignature as an indicator of radiation exposure (58).

Based on the pathways findings in NoSTox subjects, we hypothesized that functional impairment in the control of the radiation-induced expression of NF-kB, mainly in/of the ubiquitination/deubiquitination mechanisms, together with an impairment of the AIM2 inflammasome pathway and a dysregulation of the insulin-like growth factor (IGF) axis (possibly correlated to a impaired plasmalogen biosynthesis), could modulate radiation-induced lung inflammation preventing development of acute radiation pneumonitis and chronic radiation-induced pulmonary fibrosis ( Figure 1 ).

The same approach based on the detection-disrupted pathways was used for the Stox group.

Two disrupted pathways, with potential impact on Stox patients’ phenotypes were found:

1. The dissolution of the fibrin clot pathway (R-HSA-75205)

One variant related to this this pathway was identified: a stopgain variant in the SERPINF2 gene in patient #2, who developed a thromboembolic event.

Irradiation activates procoagulant activity and reduces fibrinolytic activity (59) in animal models of thoracic irradiation, suggesting the involvement of coagulation and fibrinolysis in radiation-induced pulmonary injury (60). The SERPINF2 gene encodes a member of the serpin family of serine protease inhibitors, the alpha 2 antiplasmin (α2-AP), which, together with the plasminogen activator inhibitor-1 (PAI-1), is the principal direct inhibitor of fibrinolytic activity. Alpha2-antiplasmin plays a significant role in acute pulmonary embolism (61) and patients with pulmonary embolism have been shown to display higher rate of fibrin clot degradation (62). Impairment of Alpha2-antiplasmin activity by the stopgain variant in the SERPINF2 gene, by possibly enhancing fibrinolytic activity, could then be involved in the development of the pulmonary thromboembolic event observed in patient 2#.

2. Formation of editosomes by ADAR proteins (R-HSA-77042)

A high impact nonsynonymous variant in the ADAR gene, involved in the formation of editosomes by ADAR proteins, was carried by patient #94, who progressed to high-grade fibrosis.

The posttranscriptional modification of RNA is a key process controling the output of the genome, and the deamination of adenosines (A) to inosines (I) is the prominent RNA editing event in humans, catalyzed by the adenosine deaminase acting on RNA (ADAR) family of proteins (63). The effects of radiation on RNA editing are poorly understood; Liu and colleagues reported that upon/after α-particle radiation, RNA editing sites change greatly and their total amount decreases after radiation (64). It has been determined that the expression of ADAR is tissue specific, the lung being the second most highly expressing site, in terms of tissue expression (65, 66). The RNA editome (R-HSA-77042) analysis of idiopathic pulmonary fibrosis (IPF) and normal lungs revealed increased editing frequency in IPF compared with normal lungs, suggesting a role for dysregulated editing in IPF pathogenesis (67). It has been found that the expression of ADAR1 and ADAR2 is downregulated in fibroblasts from patients with IPF (68). The changes in expression levels of ADAR1 and ADAR2 may represent an important controling mechanism in IPF, regulating the processing of key miRNAs such as miRNA-21. Overexpression of miRNA-21 in lung tissue and pulmonary fibroblasts from patients with IPF may be due to defective editing by ADAR (66). This miRNA targets antifibrosing protein synthesis such as SMAD family member 7, TGF beta receptor 2 (TGF-βR2), TIMP metallopeptidase inhibitor 3 (TIMP3), and vascular endothelial growth factor A (VEGF-A) (69, 70). Crosstalk among impact pathways which control RNA editing (R-HSA-75064 and R-HSA-77042) could contribute to the induction of the severe lung fibrosis observed in patient #94.

Taken together the pathway findings of Stox patients, lead us to hypothesize that specific variants in genes affecting both the fibrinolytic activity and RNA editing pathways, could be responsible of the severe toxicity events reported, after irradiation, by some patients of the Stox group, who present these specific variants ( Figure 1 ).