Diagnosis

Abstract

The invention relates to a panel of biomarkers, comprising or consisting of: benzene, benzothiazole, decane, isothiocyanato-cyclohexane, methylstyrene, phenol, styrene, toluene, 1-hexanol, 2-butoxyethanol, and p-xylene. In particular to the use of this panel in i) diagnosing eosinophilia in a subject, ii) treating an asthmatic subject with eosinophilia and / or who is suffering severe exacerbations, or iii) monitoring treatment efficacy in a subject with eosinophilia and / or who is suffering severe exacerbations.

Description

[0001] DIAGNOSIS

[0002] FIELD OF THE INVENTION

[0003] The invention relates to a panel of biomarkers and their use in diagnosing eosinophilia in asthmatic subjects.

[0004] BACKGROUND

[0005] Asthma is a complex chronic disease characterised by type-2 inflammation in 40-60% of patients, associated with cognate elevation of relevant immune cells. According to GINA guidelines, approximately 3.7% of patients with asthma have severe disease, characterised by treatment failure to high dose inhaled steroids and current add on therapies, and require high cost targeted biologic therapies - for example, eosinophil depleting therapies targeting IL-5 and its receptor; eosinophils are believed to contribute to asthma exacerbations. Furthermore, several orally active biologic agents are now in development for type-2 asthma in patients that have a high unmet need but are pre-biologic care.

[0006] Cognate with the development of high-cost therapies in moderate-severe asthma, are the development of biomarkers that can be readily translated for stratification, and disease monitoring, as well as to allow a deeper understanding of disease mechanisms.

[0007] As an example, phenotyping of airway inflammation in asthma using induced sputum was originally described by Schleich et al. (BMC Pulm Med. 2013;13: l l). The identification of a sputum eosinophilia in severe asthma has now been shown to enable titration of corticosteroids with a view to preventing asthma exacerbations, and to identify patients with severe asthma that are more likely to respond to anti-IL-5 therapies.

[0008] However, whilst sputum phenotyping has proven to be invaluable in the characterisation of airway inflammation in severe asthma, sputum sampling is time consuming, biased towards patient that can generate a sample at any given time, and costly - these all limit its application across centres. The development of a non-invasive biomarker of sputum based inflammatory content would offer the potential for non-invasive and patient phenotyping of airway inflammation in severe asthma. The present invention seeks to address the problems mentioned above.

[0009] SUMMARY OF INVENTION

[0010] In a first aspect, the invention provides a panel of biomarkers, comprising or consisting of: benzene, benzothiazole, decane, isothiocyanato-cyclohexane, a-methylstyrene, phenol, styrene, toluene, 1-hexanol, 2-butoxyethanol, and p-xylene. Such biomarkers are volatile organic compounds (VOCs).

[0011] The panel may comprise a further one or more, two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, or more VOCs.

[0012] The panel may further comprise or consist of one or more or, or all of: acetone, benzaldehyde, decanal, nonanal, hexanal, 2-ethylhexanal, tridecane, and methylene chloride.

[0013] The panel may therefore comprise or consist of: acetone, benzene, toluene, phenol, p- xylene, styrene, a-methylstyrene, benzothiazole, benzaldehyde, decanal, nonanal, hexanal, 2-ethylhexanal, decane, isothiocyanato-cyclohexane, tridecane, 1-hexanol, 2- butoxy-ethanol, methylene chloride.

[0014] The panel may be for use in a method of diagnosing eosinophilia in a subject. Eosinophilia as described herein may be confirmed when the eosinophil count in a biological sample accounts for about 3% or more of the cells in the sample and / or when the eosinophil concentration is about 0.3xl09or more cells / L in a biological sample.

[0015] In a second aspect, the invention provides a method of diagnosing eosinophilia in a subject, comprising: i. providing a biological sample obtained from the subject; ii. determining the level of the panel of biomarkers defined in the first aspect of the invention; iii. comparing the level of the panel of biomarkers in the sample with a reference level of the same panel of biomarkers; iv. using the results from (iii) to provide a diagnosis to the subject. The eosinophilia may indicate that the subject is suffering from severe asthmatic exacerbations and / or that the subject has stable-severe or severe asthma.

[0016] The subject of any aspect may have been previously diagnosed with asthma. The asthma may have acute asthma, stable-severe asthma, or severe asthma.

[0017] The panel of biomarkers may be detected in a biological sample obtained from a subject. A sample obtained from the subject may be a breath sample. The sample may be sputum or blood sample. A blood sample may be a plasma or serum sample.

[0018] The panel of biomarkers used in any method of the invention may be detected in the headspace of the sample obtained from the subject.

[0019] Some or all of the steps of the method(s) of the invention may be carried out in vitro.

[0020] A method of the invention may be used to monitor the progression of asthma in a subject.

[0021] In another aspect, the invention, provides a method of treating an asthmatic subject with eosinophilia and / or who is suffering severe exacerbations comprising: i. providing a biological sample obtained from the subject; ii. determining the level of a panel of biomarkers of the first aspect of the invention; iii. comparing the level of the panel of biomarkers in the sample with a reference level of the same panel of biomarkers; iv. using the results from (iii) to determine if the subject has eosinophilia and / or is suffering severe exacerbations; and v. administering to the subject a therapeutic agent to treat or reduce the symptoms of the eosinophilia and / or severe exacerbations if the subject is determine in (iv) to have eosinophilia and / or is suffering severe exacerbations

[0022] In another aspect, the invention provides a method of monitoring treatment efficacy in a subject with eosinophilia and / or who is suffering severe exacerbations comprising: i. providing a first biological sample obtained from the subject prior to the administration of treatment; ii. determining the level of a panel of biomarkers of the first aspect of the invention in the first biological sample; iii. providing a second biological sample obtained from the subject after the administration of treatment; iv. determining the level of the same panel of biomarkers in the second biological sample as those in the first biological sample; v. comparing the level of the panel of biomarkers in the first sample with the level in the second sample to determine if the treatment is effective.

[0023] The level of the panel of biomarkers may give a VOC score. When the panel of biomarkers used in any method of the invention is detected in the headspace of the sample, a VOC score of about 2.6 or lower may indicate that the subject has eosinophilia, or a VOC score of about 3.0 or higher may indicate that the subject does not have eosinophilia. When the panel of biomarkers used in any method of the invention is detected in a breath sample, a VOC score of about 89 or higher may indicate that the subject has eosinophilia, or a VOC score of about 137 or lower may indicate that the subject does not have eosinophilia.

[0024] In the method of monitoring treatment efficacy according to the invention, if the VOC score is increased in the second sample compared to the first sample it may be concluded that the treatment was effective.

[0025] In another aspect of the invention, there is provided a kit for performing a method of the invention. The kit may comprise means for determining the level of the panel of biomarkers of the invention. The kit may contain sample collection apparatus. The kit may contain specialised storage container(s) for the sample. Preferably a container is gas impermeable. The kit may contain instructions to use the kit.

[0026] The inventors have surprisingly determined that VOCs in the headspace of biological samples taken from asthmatic subjects can be used to help stratify the subject accordingly, allowing those with eosinophilia and who are suffering from severe exacerbations to be identified. Thus, the invention provides a non-invasive method for stratifying a subject based on the type of asthma that they have, with a high degree of sensitivity and specificity. Importantly, the methods of the invention allow phenotyping of airway inflammation. This can then be used to determine the most appropriate types of treatment, such as eosinophil depleting therapies.

[0027] It will be appreciated that the level of any VOC in the panel of biomarkers may be determined using any suitable method / technique / technology known in the art. For example, determining the level may comprise using two-dimensional gas chromatography coupled with mass spectrometry (GCxGC-MS), Thermo-Desorption Gas Chromatograph-Mass Spectrometry (TD-GC-MS), gas chromatograph - ion mobility spectrometry (GC - IMS) technology, Gas Chromatograph (GC), Gas Chromatograph - Mass Spectrometry (GCMS), Mass Spectrometry (MS), Ion Mobility Spectrometry (IMS), Differential Mobility Spectrometry (DMS), light absorption Spectrometry, Field Asymmetric Ion Mobility Spectrometry (FAIMS), Electronic Nose, Selective-Ion Flow Tube Mass Spectrometry (SIFT-MS), Protein-transfer-reaction-MS, Optical absorbance / Non-dispersive Infra-red and gas sensors (individual or in an array). Preferably, TD-GC-MS is used.

[0028] The term "headspace ” as referred to herein can refer to the gaseous constituents of a sample, which may be in a closed space. Thus, the sample may be stored in a closed space.

[0029] Determining the level of the panel of biomarkers may comprise quantifying the presence and / or concentration of each biomarker, in the panel of biomarkers, in the sample. The level of the panel of biomarkers in the sample may be compared relative to that of a control / reference sample, or a predetermined standard level. The control / reference sample may be from a healthy non-asthmatic individual, or an asthmatic individual who is not suffering from severe exacerbations and / or who does not have eosinophilia, or from a subject prior to treatment. Multiple reference samples may be taken and averaged to obtain a reference level. Preferably the subject and the reference are species matched. Preferably the subject and the reference are age matched. Preferably the subject and the reference are gender matched.

[0030] By way of example, the level of one or more of, such as all of, the panel of biomarkers in the sample may be higher (or increased by at) least about 5%, 10%, 15%, 20%, 25%, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95% or 100% or more compared to the level in the reference sample. The level of one or more of, such as all of, the panel of biomarkers in the sample may be lower by (or reduced by at) least about 5%, 10%, 15%, 20%, 25%, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95% or 100% or more compared to the level in the reference sample. By way of example, the level of one or more of, such as all of, the panel of biomarkers may be at least 5, 10, 15 or 20 or more fold higher than the reference level. By way of example, the level of one or more of, such as all of, the panel of biomarkers may be at least 5, 10, 15 or 20 or more fold lower than the reference level.

[0031] In an embodiment, the level of a panel of biomarkers as referred to herein may refer to a VOC score. The VOC score may refer to a scoring system (as shown in figure 8C) which scores the biomarkers in a panel of biomarkers in a sample based on the peak area and the average difference between eosinophil enriched and non-eosinophil enriched sputa samples. In each case the normalised peak value for each VOC in a sample is multiplied by the average regression coefficient for that VOC. Values for all the VOCs are summed together and added to the regression model intercept (constant). The VOC score may be used to diagnose eosinophilia.

[0032] As described in the examples, the inventors determined the concentration of various VOC biomarkers in the headspace of biological samples taken from asthmatic patients with stable-severe or severe asthma. They demonstrated that there was a statistically significant change in the VOC score in the patients being monitored.

[0033] The term “severe exacerbations ” can refer to subjects with asthma where exacerbations are episodes characterised by a progressive increase in symptoms of shortness of breath, cough, wheezing or chest tightness and progressive decrease in lung function, i.e. they represent a change from the patient's usual status that is sufficient to require a change in treatment. Such a definition is based on symptoms (“talks in words, sits hunched forward, agitated”), clinical findings (respiratory rate >30 per min, heart rate >120 per min, oxygen saturation <90%, use of accessory muscles) and lung function (PEF <50% pred). Alternatively, the severe exacerbations may be defined as events that require urgent action on the part of the patient and physician to prevent a serious outcome, such as hospitalisation or death from asthma. Such action may include use of systemic corticosteroids or an increase from a stable maintenance dose, for at least 3 days. The action may include hospitalisation or emergency department visit because of the exacerbations, requiring systemic corticosteroids.

[0034] The term “stable severe asthma” or “stable severe asthmatic subjects” can refer to asthma or a subject with asthma that requires treatment with high dose inhaled corticosteroids plus a second controller and / or systemic corticosteroids to prevent it from becoming “uncontrolled” or that remains “uncontrolled” despite this therapy.

[0035] Stable asthma can be defined as no asthmatic exacerbation in the preceding period. The preceding period may be one week or more, two weeks or more, three weeks or more, 4 weeks or more.

[0036] A single VOC in the panel of biomarkers may be used to perform the invention. However, it will be appreciated that the more VOCs that are used in the panel of the invention, the more reliable and statistically accurate the outcomes of the invention will be. The more VOCs that are used in a panel of invention, the higher the sensitivity and the higher the specificity of the invention.

[0037] A VOC (volatile organic compound) may be referred to as an organic compound that has a boiling point between about 50°C and about 250°C at a standard atmospheric pressure of 101.3 kPa.

[0038] The ‘subject ’ or ‘individual ’ may be a vertebrate, mammal or domestic mammal. Hence, the method according to the invention may be used to diagnose or treat any animal, for example, pigs, cats, dogs, horses, sheep or cows. Preferably, the subject is a human.

[0039] The term ‘treating ’ may refer to reducing the severity of asthma in a subject, such as reducing symptoms, exacerbations and their severity and / or reducing the eosinophil level in a subejct.

[0040] All of the embodiments and features described herein (including any accompanying claims, abstract and drawings), and / or all of the steps of any method or process so disclosed, may be combined with any of the above aspects or embodiments in any combination, unless stated otherwise or where at least some of such features and / or steps are mutually exclusive. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Reference will now be made, by way of example, to the accompanying drawings, in which:

[0042] Figure 1 (Upper Panel) is a summary of the in vitro discovery study used to develop a volatile biomarker signature of sputum eosinophilia in patients with severe asthma. The study started with the collection of the VOCs present in headspace of 22 eosinophil enriched (>3% ) and 14 non-eosinophil enriched sputum (<3%) samples from severe asthmatics. Headspace samples were analysed by gas chromatography coupled mass spectrometry (GC-MS). Multivariate statistical analysis using elastic net (eNET) regression (100 runs, 10-fold cross validation) identified a panel of VOCs which were able to discriminate between eosinophil enriched and non-eosinophil enriched sputum headspace samples. Equation used to calculate the sample scores: = intercept, = average regression coefficient for the ithVOC, xt= normalised peak value for the ithVOC, p = VOCs.

[0043] (Lower Panel) shows a summary of the clinical validation studies sought to fit the eNET model coefficients for VOC and the intercept from the in vitro study to different clinical cohorts. Cohort 1 : cohort of acutely exacerbating adult asthma patients admitted to hospital with severe exacerbations from the EMBER cohort (n=65) (13), characterised with blood eosinophil levels, and GCxGC-MS analysis of breath samples acquired on admission to hospital. Sub-Cohort 1 : fraction of adult acute exacerbating asthmatics from Cohort 1 (n=24), characterised with sputum eosinophil count and GCxGC-MS analysis of breath samples acquired on hospital admission. Cohort 3: stable severe asthma patients characterised with blood eosinophil levels (n=22). Created with BioRender.com

[0044] Figure 2 is (A) a histogram of the normalised peak area values of headspace VOC biomarkers identified as discriminatory for sputum eosinophilia > 3% using elastic net regression (summarised as median and IQR). Comparisons are between eosinophil- enriched (>3%) and non-eosinophil enriched (< 3%). Mann-Whitney test (significance at p< 0.05*, p<0.01**). 1-hexanol *p= 0.02, styrene *p= 0.017, **phenol p= 0.005, *decane p= 0.019, benzothiazole *p= 0.03; (B) a box plot of the headspace VOC biomarker score generated by eNET regression in eosinophil-enrichedand non- eosinophil-enriched (<3%) sputum samples (median, QI, Q3, min and max) (Mann- Whitney test: *p<0.0001). Headspace VOC biomarker score in the box plots are represented in three different colours according to asthma treatment intensity defined by the Global Inititaive for Asthma Guidelines(34). GINA 4, GINA 5 (maintenance oral steroids - mOCS) or GINA5 (mepolizumab±mOCS); (C) a receiver operating characteristic (ROC) curve to evaluate the diagnostic accuracy of the eNET derived headspace VOC biomarker score for detecting a sputum eosinophilia > / =3% in severe asthma; (D) a Spearman’s correlation between eNET derived headspace VOC biomarker score and the sputum eosinophil % (rs= -0.71; two-tailed t-test: p<0.0001); and (E) a Spearman’s correlation between eNET derived headspace VOC biomarker score and the total sputum eosinophil count (rs= - 0.54; two-tailed t-test: p=0.0006).

[0045] Figure 3 shows (A) Log transformed peak area values (median, IQR) of 16 / 19 selected VOCs which were detected by GCxGC-FID-MS in exhaled breath samples of acute asthmatics from cohort 1, which were grouped as patients with blood eosinophil levels > 0.3X109 / L (n=13) and patients with blood eosinophil levels < 0.3xl09 / L (n=52) . Mann Whitney test: decanal (p=0.04), hexanal (p=0.02); (B) Log transformed peak area values (median, IQR) of 16 / 19 selected VOCs which were detected by GCxGC-FID-MS in exhaled breath samples of a sub-cohort of acute asthmatics from cohort 1, which were grouped according to the sputum eosinophil count > 3% (n=9) and < 3% (n=15). Mann Whitney test: hexanal (p=0.02), styrene (p=0.04); (C) Breath VOC scores (median, IQR) derived using the same model fitted to the in-vitro headspace data (median, IQR) in acute asthmatics with blood eosinophils >0.3xl09 / L, and acute asthmatics with blood eosinophil count < 0.3xl09 / L; (D) Breath VOC scores (median, IQR) derived using the same model fitted to the in-vitro headspace data (median, IQR) in acute asthmatics with sputum eosinophils >3%, and acute asthmatics with sputum eosinophil count < 3%; (E) Receiver operating characteristic (ROC) curve to evaluate the ability of the breath VOC biomarker scores fitted from the in vitro model - to discriminate between acute asthmatics with blood eosinophils >0.3xl09 / L versus acute asthmatics with blood eosinophils <0.3xl09 / L; and (F) ROC curve to evaluate the ability of the breath VOC biomarker scores fitted from the in vitro model - to discriminate between acute asthmatics with sputum eosinophils >3% versus acute asthmatics with sputum eosinophils <3%. Figure 4 shows (A) Log transformed peak area values (median, IQR) of 17 / 19 selected VOCs, which were detected by GCxGC-FID-MS in exhaled breath samples of stable severe asthmatics from cohort 2. Patients were grouped as subjects with blood eosinophil levels > 0.3xl09 / L (n=12) and subjects with blood eosinophil levels < 0.3X109 / L (n=10); (B) Breath VOC scores (median, IQR) derived using the model fitted to the in-vitro headspace data in severe stable asthmatics classified as patients with and without blood eosinophilia (blood eosinophil threshold : 0.3X109 / L); and (C) Receiver operating characteristic curve to evaluate the ability of breath VOC biomarker scores fitted from the in vitro model to discriminate between severe stable asthmatic with and without blood eosinophilia.

[0046] Figure 5 is a schematic drawing the headspace sampling and analytical system used in the current study (Peltrini et al, 2020).

[0047] Figure 6 is a picture and schematic representation of a ReCIVA sampler (Owlstone Medical, Cambridge, UK).

[0048] Figure 7 is an example of a GCxGC chromatogram for an exhaled breath sample showing volatile organic compounds separation, the deuterated internal standards and the chromatographic region occupied by different chemical classes. Adapted from (Wilde et al, 2019).

[0049] Figure 8 is: (A) Histogram of the normalised peak area values summarised as median and IQR, and compared by Mann-Whitney test (significance at p< 0.05); (B) a heat map of the average regression coefficients for the 19 selected VOCs detected in sputum headspace of severe asthmatics as feature able to discriminate eosinophil-enriched from non-eosinophil-enriched sputa; and (C) an equation used to calculate the sample scores ( / ?0= intercept, pt= average regression coefficient for the ithVOC, xt= normalised peak value for the ithVOC, p = VOCs) and calculation of the highest (10.05) and lowest (- 1.6) sample scores.

[0050] Figure 9 shows a bar chart of the abundance values (median; IQR) of the 19 selected biomarkers in sputum and control headspace samples. Two-tailed non-parametric test compared the peak area values of each compound between sputum and control headspace samples. The comparison revealed significant differences in peak area values for acetone (p<0.0001), hexanal (p= 0.013), styrene (p<0.0001), phenol (p= 0.002), benzothiazole (p= 0.04) and 2-ethylhexanal (p= 0.006); benzothiazole was the only compound to show a significant increase in control headspace samples compared to sputum headspace samples.

[0051] EXAMPLES

[0052] Material and methods

[0053] Patient cohorts

[0054] Discovery cohort

[0055] Thirty-six patients were recruited from the adult severe asthma service at the Glenfield Hospital (UK) between 2018 and 2019. All patients had a diagnosis of severe asthma according to the ATS / ERS 2014 consensus criteria (11) and were recruited at least six weeks post exacerbation. Participants were asked to provide a spontaneous sputum sample, which was immediately transferred to the lab for the collection of IL of headspace onto sorbent tubes (Carbograph 1TD and Tenax TA 60 / 40 - Markes International, Llantrisant), using a previously validated and reported protocol (see “Sputum headspace collection”) (12).

[0056] Subsequently, within 1 hour of sputum collection, it was processed for the differential cell counting using the protocol described in the section “Total and differential cell count in sputum samples” in order to group sputum headspace samples as eosinophil enriched (n=22) and non-eosinophil enriched (n=14) according to the sputum eosinophil threshold of 3%, which represented the sputum eosinophil upper limit value for asthmatics and a marker of uncontrolled disease and of high risk of asthma exacerbations (Green et al, 2002).

[0057] Prior to each sputum headspace collection, a control headspace sample (in absence of sputum) was collected onto separate sorbent tube. Sputum samples whose headspace was collected the same day at the same moment shared the control headspace samples; therefore, for 36 sputum headspace samples there were 26 controls.

[0058] Both control and sputum headspace samples were then analysed by thermal desorption gas chromatography coupled to mass spectrometry (TD-GC-MS). An elastic net regression was fitted to the sputum headspace data matrix in order to identify stable VOC biomarkers of sputum eosinophilia >3%, and to develop a statistical model based on the selected VOCs, which was used to generate a VOC biomarker score for each participant. The developed model was then validated in different clinical cohorts.

[0059] Clinical validation cohorts

[0060] Separate clinical validation studies for the in vitro developed volatile biomarker-based model were performed in order to evaluate its accuracy to predict eosinophilic asthma exacerbations (cohort 1), identify sputum eosinophilia in a sub-cohort of exacerbating asthmatics (sub-cohort 1), and to detect blood eosinophilia in a stable independent severe asthma cohort (cohort 2). Breath samples were collected between 2018 and 2019 onto sorbent tubes using a ReCIVA sampler (Owlstone Medical, Cambridge, UK) as described in the section titled “Exhaled breath collection”, and analysed by two- dimensional gas chromatography with flame ionisation detector and mass spectrometry (GCxGC-FID-MS).

[0061] Cohort 1 (n=65): Participants from an acute breath phenotyping cohort (13) were admitted to hospital for moderate-severe asthma exacerbations (Glenfield Hospital, UK). Breath samples were acquired within 24 hours of hospital admission in parallel to blood samples, which were used to estimate the eosinophil levels (section “Peripheral blood eosinophils”) to classify breath samples as eosinophilic (blood eosinophils > 0,3X109 / E; n=13) and non-eosinophilic (blood eosinophils < 0,3xl09 / E; n=52) according to the blood eosinophil threshold of 0,3xl09 / E, reported by the Global Initiative for Asthma as the blood eosinophil threshold associated to type 2 inflammation and increased risk of asthma exacerbations (2).

[0062] Sub-cohort 1 (n=24): Subgroup of 24 participants from cohort 1 that were selected as able to provide spontaneous sputum samples at exacerbation for differential cell counts. Breath samples were acquired within 24 hours of hospital admission, and were grouped according to the sputum eosinophil count as acute asthmatics with sputum eosinophils > 3% (n=9) and acute asthmatics with sputum eosinophils < 3% (n=15).

[0063] Cohort 2 (n=26): Participants with stable severe asthma, classified as GINA 5 (14), which were recruited from a single centre (Glenfield Hospital, UK) at least six weeks post exacerbation. In parallel to exhaled breath collection, blood samples were collected in order to classify patients as eosinophilic (blood eosinophils > 0,3X109 / E; n=12) and non-eosinophilic (blood eosinophils < 0,3xl09 / E; n=10) stable severe asthmatics. Analytical methods and modelling

[0064] TD-GC-MS headspace analysis and statistical modelling

[0065] 36 sputum and 26 control headspace samples were separately analysed by TD-GC-MS using the analytical parameters summarised in Table 4 and in the section called “TD- GC-MC analysis”. Chromatographic peak deconvolution, integration and alignment across the headspace samples were performed by AnalyzerPro (SpectralWorks, Runcorn, UK - vertion5.7), as reported by Peltrini et al (12) and in the section called “Data analysis and chemometrics”, which generated a data matrix containing the detected chromatographic feature identities and their chromatographic peak area values normalised by the peak area value of deuterated toluene, an internal standard used for the TD-GC-MS analysis. Furthermore, the data matrix was amended removing the features deriving from the chromatographic column bleed, such as the siloxanes, and those present in only one sample.

[0066] Table 4 Analytical method parameters for TD-GC-MS headspace analysis

[0067] THERMAL DESORPTION

[0068] Tube desorption temperature 300 °C

[0069] Tube desorption time 5min

[0070] Tube desorption flow 45 mL min1

[0071] Cold trap temperature -10°C

[0072] Trap desorption temperature 300 °C

[0073] Trap desorption time 5 min

[0074] Trap desorption flow 2 mL min1

[0075] Flow path temperature 200 °C

[0076] Mode Splitless

[0077] GAS CHROMATOGRAPHY

[0078] Column Rxi-5ms 60mx0.25mm i.d.x0.25 pm

[0079] Initial oven temperature 40°C

[0080] Carrier gas Helium

[0081] Carrier gas flow 2 mL min1

[0082] Oven temperature ramp 5 °C min1

[0083] Final oven temperature 300 °C

[0084] Final temperature holding time 5 min

[0085] MASS SPECTROMETRY

[0086] Scan type Full scan (+ve) Mass range 40 to 550 m / z

[0087] Ionisation type El

[0088] Scan frequency 3Hz

[0089] Transfer line temperature 300°C

[0090] Quadrupole temperature 150°C

[0091] Manifold temperature 230°C

[0092] Solvent delay time 5 min

[0093] An elastic net regression (eNET) was fitted to the sputum headspace data matrix in which the samples were classified according to the sputum eosinophil threshold of 3%, in order to select those features with a non-zero regression coefficients in at least the 80 of 100 runs of a 10-fold cross validation. The estimated average value of the regression coefficient for each of the selected features (VOC biomarkers) and the intercept term were used in developing a score which was able to discriminate between eosinophil-enriched and non-eosinophil enriched sputum headspace samples. The biomarker score for each headspace sample was obtained by summing the regression intercept value and the sum of the products between the averaged regression coefficients of the selected features and their respective normalised peak area values (see Figure 8). The eNET was fitted using the cv.glmnet function from the glmnet package in R.3.6.1, (R Core Team, https: / / www.R-project.org). This research used the SPECTRE High Performance Computing Facility at the University of Leicester.

[0094] The identities of the selected chromatographic features were compared to data acquired from TD-GC-MS analysis of the corresponding chemical standards (Sigma-Aldrich), assigning so to the features a level 1 of identification according to the Metabolomics Standard Initiative (MSI) guidelines.

[0095] GC^GC-FID-MS breath analysis and model clinical validation

[0096] Exhaled breath samples from cohort 1, sub-cohortl, and cohort 2 were analysed by GCxGC-FID-MS as described by Wilde et al (15) and in the section called “TD-GC- cGC-FID-MS analysis”, using the analytical parameters reported in Table 6. Data analysis was performed by MassHunter GC-MS Acquisition B.07.04.2260 (Agilent Technologies Ltd, Stockport, UK) and by GC Image™ v2.6 along with GC Project and Image Investigator (JSB Ltd, Horsham, UK) as reported by Wilde et al (16). A targeted analysis was performed in order to detect in exhaled breath samples the previously in vitro selected VOCs (Data analysis and chemometrics). Peak area values of the selected VOCs were log transformed and batch adjusted dividing the product between VOC abundance and VOC abundance average by the VOC abundance standard deviation in all the samples. The in vitro developed model was used to generate a biomarker score for each exhaled breath sample of each clinical cohort by adding the previously estimated model intercept value to the sum of the products between the selected VOC peak area values in breath samples and the VOC averaged regression coefficients, which were estimated in vitro.

[0097] Table 6 TD-GCxGC-FID-qMS analytical method parameters.

[0098] DRY PURGE AND THERMAL DESORPTION

[0099] Tube dry purge helium flow 50mL min1

[0100] Tube dry purge time 1 min

[0101] Tube desorption temperature 300 °C

[0102] Tube desorption time 5min

[0103] Tube desorption flow 45 mL min1

[0104] Cold trap temperature -10°C

[0105] Trap desorption temperature 300 °C

[0106] Trap desorption time 2 min

[0107] Trap desorption flow 2 mL min1

[0108] Mode Splitless

[0109] TWO-DIMENSIONAL GAS CHROMATOGRAPHY

[0110] First chromatographic column 5% phenyl 95% dimethylpolysiloxane

[0111] (30 m x 0.25 mm x 0.25 pm Rxi-5Sil MS)

[0112] Second chromatographic column Polyethylene-glycol

[0113] (4 m x 0.25 mm x 0.25 pm DB-WAX)

[0114] Carrier gas Helium

[0115] First column-Carrier gas flow 0.6 mL min1

[0116] Modulation period 3 sec

[0117] Load time 2.799 sec

[0118] Injection time (to second column) 0.201 sec

[0119] Second column-Carrier gas flow 0.3 mL min1

[0120] Initial oven temperature 30°C - held for 5min

[0121] Oven temperature ramp 3 °C min1

[0122] Intermediate oven temperature 80 °C - held for 5min

[0123] Oven temperature ramp 5 °C min1

[0124] Final oven temperature 250 °C - held for lOmin

[0125] RESTRICTORS

[0126] Restrictor to FID 1.2 m x 0.25mm; fused silica Carrier gas flow rate 23 mLmin1

[0127] Restrictor to single quadrupole mass spectrometer 0.76 m x 0.10 mm; fused silica

[0128] FID

[0129] FID heater temperature 250°C

[0130] Make-up gas flow Nitrogen; 25mL min1

[0131] Pure air flow 400 ml min1

[0132] Hydrogen flow 35 mL min1

[0133] Acquisition rate 100 Hz

[0134] SINGLE QUADRUPOLE MASS SPECTOMETER

[0135] Mass scan range 40 to 300 m / z

[0136] Ionisation type El

[0137] Acquisition rate 21.5Hz

[0138] Transfer line temperature 300°C

[0139] Ion source temperature 250°C

[0140] Quadrupole temperature 150°C

[0141] Extended sampling methods

[0142] Sorbent tube preparation for VOC collection

[0143] VOCs from both headspace and exhaled breath samples were concentrated onto stainless steel sorbent tubes, which were packed with a hydrophobic dual-bed sorbent mixture of Carbograph 1TD and Tenax® TA 60 / 40 (Markes International, Llantrisant), respectively a medium and a weak strength adsorbent material.

[0144] Prior to sample collection, sorbent tubes were weighed and conditioned. Uncapped tubes were weighed before each conditioning procedure, to check for possible sorbent mass degradation, and then were conditioned at 330°C in a nitrogen flow at a pressure of 1.5 bar, for 150 minutes. The tube caps and their combined polytetrafluoroethylene (PTFE) ferrules were washed using an appropriate neutral detergent, rinsed in deionized water and dried in oven at 200°C. Capped tubes were stored at room temperature and used for headspace collection within 15 days from their conditioning.

[0145] Prior to the TD-GC-MS analysis for the in vitro discovery study, sorbent tubes containing the VOCs collected from headspace samples were removed from the storage at 4°C and placed at room temperature to warm up. In order to ensure acceptable levels of inter-samples variability, an internal standard solution was loaded into the sorbent tube. This was performed using a calibration solution-loading rig (CSLR, Markes International Ltd, Llantrisant, UK). The sorbent tube was attached to an injector and 0.6 j l of internal standard solution were introduced in a purified nitrogen flow (Zero grade, BOC) of 100 ml / min for 2 minutes to ensure an efficient transfer of the standard onto the adsorbent packing. The internal standard solution comprised a mixture of deuterated toluene (D8), octane (DI 8) and phenanthrene (DIO) at a concentration of 10 pg / ml.

[0146] Within 72 hours of breath collection, sorbent tubes were dry purged in a flow of nitrogen (CP grade with an in-line trap; BOC, Leicester, UK) at a rate of 50 ml / min for 2 minutes, in the sampling direction. The dry purge was performed to reduce the amount of condensed water contained in them, and therefore to increase sample stability. GC-MS technologies are normally very sensitive to water, which can affect the integrity, and therefore reduce the working life, of some components such as the chromatographic column and the mass spectometer detectors. Furthermore, water can affect the analysis outcome quenching the detector response for those compounds that co-elute with it (Woolfenden, 2010a). The sorbents tubes had higher affinity for organic molecules (including highly polar compounds like light alcohols) than water, therefore during the dry purge step, water was selectively removed to vent without loss of the most volatile / polar compounds of interest. Dry purged tubes were placed at 4°C to be then analysed by GCxGC-FID / MS within 4 weeks. Before GCxGC / MS analysis, the tubes were loaded with 0.6 pL of the same internal standards solution used for headspace sample.

[0147] Sputum collection

[0148] Spontaneously produced sputum samples were acquired at least six weeks post exacerbation from the severe asthmatic cohort employed in the in vitro study, and from cohorts 1 and 2, which were used for the biomarkers validation studies. A spontaneous sputum production was preferred to sputum induction by inhalation of hypertonic saline solution in order to reduce possible contaminating sources, to standardize the sputum collection method, and to avoid patients’ submission to this invasive procedure, encouraging therefore their participation.

[0149] Sputa were collected into small Petri dishes (Merck Chemicals Ltd., Nottingham, UK) in the clinical research facility testing room at the Glenfield hospital in Leicester.

[0150] Sputum headspace collection

[0151] For the in vitro biomarker discovery study, VOCs were collected from headspace of sputum samples, or just from control samples (blank). Collection was based on the generation of a dynamic headspace by flushing pure air (BTCA 178 air, BOC) above the uncovered Petri dish respectively in presence or absence of sputum. The air flow dragged the volatiles from the headspace onto a stainless steel sorbent tube, where VOCs were concentrated to be subsequently analysed by TD-GC-MS.

[0152] The system to collect the headspace from sputum and control (blank) samples consisted in a custom vessel (University of Leicester, UK) with a volume of 0.2 L, made of polyether ether ketone (PEEK), a semi-crystalline thermoplastic material with excellent mechanical and chemical resistance properties. The uncovered Petri dish that contained the sputum or a blank sample was placed into the vessel which was then sealed with a screw cap fitted with two ports; an inlet port, joined to a pure air cylinder (BTCA 178 grade, BOC, UK), and an outlet port which was connected to the sorbent tube during the headspace collection (Figure 5). The decision to flush the sample with pure air rather than nitrogen or other gasses originated from the intent to avoid possible modification in sputum microbiome, preventing so potential chemical alterations in the volatome composition.

[0153] In order to generate the dynamic headspace, the uncovered Petri dish in the vessel was flushed with pure filtered air at a flow rate of 200 ml / min for 5 minutes, sampling each time 1 L of air.

[0154] Pure air was chosen in other to reduce possible contaminations that could originate from interaction of different gasses with biological sample and / or by the interaction of the carrier gas with the sorbet material into the sampling tube (Woolfenden, 2010b).

[0155] The airflow displaced the headspace above the Petri dish and concentrated VOCs from the headspace onto the sorbent tube, which were then capped and stored at 4°C for no longer than 15 days.

[0156] The flow rate of 200 ml / min min1was selected based on a previous optimisation of the sampling parameters for standardised the sampling procedure of volatile compounds. A flow rate of 200 ml / min, was proven to enable optimum collection of VOCs whilst minimising sample collection time and the impact of environmental contamination. Furthermore, a speed of 200 ml / min was proved to reduce the risk of volatile back diffusion (when the speed is too low) or of volatiles lost (when the speed is too high and the compounds do not properly adsorb to the sorbent material). (Doran et al, 2017).

[0157] Headspace samples were collected also in absence of sputum samples with the aim to estimate the presence in the background signal of the selected volatile biomarkers.

[0158] Exhaled breath collection

[0159] Exhaled breath samples for the volatile biomarkers validation studies were collected using the ReCIVA sampler (Owlstone Medical, Cambridge, UK) shown in Figure 6, which collected exhaled breath samples from a silicon mask into thermal desorption stainless steel sorbent tubes for their subsequent analysis by GCxGC-FID / MS.

[0160] ReCIVA sampler allowed up to four sorbent tubes, which were held within four ports located above two pumps and pressure sensors. When the patient started breathing into the mask, the pressure sensors were able to monitor patient’s breath phases allowing the software to activate the pumps at the specific phase of breath cycle that was meant to collect. A method was set up to collect multiple alveolar breaths from severe asthmatics.

[0161] The samples collection and analysis took place in a dedicated room in the NIHR Respiratory Biomedical Research Unit at the Glenfield hospital (Eeicester). The patient was sat down, with the mask appropriately fitted to avoid leaks, and was asked to breath normally. During the sampling procedure, the patient breathed into the ReCIVA silicone mask. Sample collection lasted as long as it took to collect 1 E of exhaled breath at a pump flow rate of 200ml / min for a maximum of 15 minutes. Patients’ normal breathing, during the collection, was supported by a flow of 30 L / min of filtered purified air. The filters consisted in activated charcoal to adsorb environmental VOCs, and HEPA filters, which retained the particulate matter released from the charcoal. Only two sorbent tubes were used for each breath collection; they were placed in the two ReCIVA back ports, and two solid aluminium tubes were located in the two front ports. A sample of pure air was always collected before each patient’s breath collection.

[0162] Silicon facemask (Owlstone Medical Ltd) was conditioned at 180 °C overnight before being used in order to reduce the background level of siloxanes detected by TD-GCxGC- MS. The unused masks were reconditioned after two weeks. Extended analytical methods

[0163] TD-GC-MS analysis

[0164] The analytical method parameters which were used for headspace analysis, are summarised in Table 4.

[0165] After being spiked with the internal standard solution, the sorbent tube containing VOCs from sputum / blank headspace sample was placed into the thermal desorption Unity-2 (Markes International, Cardiff, UK) where the volatiles were desorbed at 300 °C for 5 minutes in a helium flow rate of 45 ml / min in splitless mode. The volatiles desorbed onto a ‘hydrophobic, general’ cold trap (matching the sorbent in the sample tube) which was held at -10 °C. The focussing trap was then heated at the maximum heating rate to 300 °C for 5 min with a desorption flow rate of 2 ml / min splitless and the compounds were introduced into the GC column.

[0166] The thermal desorption unit was linked to a gas chromatograph (Agilent, 7820 A). Volatile compounds were separated on the fused silica capillary chromatographic column (Rxi-5 m s 60 m x 0.25 mm ID x 0.25 pm film - (Restek, Bellefonte, PA, USA)) in a helium flow of 2 ml / min, with an initial oven temperature of 40 °C and a final oven temperature of 300 °C held for 5 minutes (oven temperature ramp of 5 °C / min). The intensity of the signal generate from a component passing through the detector, was reported as function of the compound RT generating a chromatogram. The gas chromatograph was coupled to a single quadrupole mass spectrometer (Agilent, 5977B). Mass spectrometer was fitted with an electron ionisation (El) ion source operated at 70 eV in which a beam of electrons ionized the sample molecules at vacuum condition. Once ionized, the ions were repelled out from the ionization chamber and moved to a single quadrupole mass analyser, in which were separated based on their mass / charge ratio (m / z). The mass scanning was performed for a range of 40 - 450 m / z at a frequency of 3 Hz, and the separated ions entered an output detector that recorded and converted the electrical impulses into visual displays.

[0167] Data analysis and chemometrics

[0168] Raw data generated from the TD-GC-MS analysis of headspace samples were analysed by AnalyzerPro (SpectralWorks, Runcorn, UK - version 5.7), a spectral pre-processing package for denoising, deconvolution, integration and alignment of the chromatographic peaks across the samples. AnalyzerPro is a vendor-independent software, which can perform the analysis of TD-GC-MS files using qualitative processing and proprietary algorithms to detect components. Before each headspace sample analysis a clean sorbent tube was spiked with 0.2 pl of a solution made of different chemical standards with well-known retention index (RI) and analysed by TD-GC-MS before the actual sample analysis. The RI standards solution (Table 5) comprised a range of n-alkanes (C8 to C20), alcohols and chlorinated compounds with known RIs, which were used to allocate the RI for each chromatographic features isolated from the sample and monitor instrument performance. The data analysis process always started with the analysis of all the TD-GC-MS raw data of the RI standards solutions. The software attributed a RT value to each compound with a known RI and spectra that was contained in the solution. This analysis allowed building a library of RIs, RTs and mass spectra (Table 5) that was subsequently used by the software to allocate RIs to the chromatographic features detected in the headspace samples.

[0169] Table 5 Library of RTs, RIs, molecular weights, formula, CAS numbers and base peaks for the components of the RI standards solution.

[0170] Name RT RI Formula Molecular CAS Base weight number peak

[0171] Toluene D8 5-50 777-9 C7D8 100 2037-26-5 98 n-Octane 6-10 800-0 C8H18 114 111-65-9 43

[0172] 1-Chlorohexene 7-33 856-3 C6H13CI 120 544-10 91

[0173] 1-Hexanol 7-62 867-4 C6H14O 102 111-27-3 56

[0174] 1-Chloroheptane 10-03 957-0 C7H15CI 134 619-06-1 91 n-Decane 11-18 1000-0 C10H22 142 124-18-5 57

[0175] 1-Chloroctane 12-93 1056-4 C8H17CI 148 111-85-3 91

[0176] 5-Nonanol 13-74 1087-2 C9H200 144 623-93-8 69 n-Undecane 14-04 1100-0 C11H24 156 1120-21-4 57

[0177] 1-Chlorononane 15-91 1163-6 C9H19CI 162 2473-01-0 91 n-Dodecane 16-95 1200-0 C12H26 170 112-40-3 57

[0178] 1-Chloroundecane 18-79 1267-7 C10H21CI 190 1002-69-3 91 n-Tetradecane 22-35 1400-0 C14H30 198 629-59-4 57 n-Heptadecane 29-.44 1700-.0 C17H36 240 629-50-5 57 n-Eicosane 35-52 2000-0 C20H42 282 17312-73-1 57

[0179] Dioctyl-phthalate 44-75 2435-1 C16H22O4 390 4376-20-9 149 The data analysis process continued with the simultaneous analysis of all the raw data generated from the TD-GC-MS analysis of the headspace samples. The software allowed to use the RI library previously generated to attribute a RI to each detected feature. AnalyzerPro detected all the different chromatographic features and aligned them across different samples; each feature was annotated with its RI, the first and second mass spectrum quantifier ion (m / z) and its retention time (RT), which together constituted a unique compound label. AnalyzerPro headspace data analysis generated a data matrix containing the chromatographic features (rows) identified across all the headspace samples (columns) and their chromatographic peak area values.

[0180] The data matrix was optimised through a refining process for the subsequent multivariate statistical analysis. The peak area value of each chromatographic feature was normalised by the internal standard toluene D8 - the most representative internal standard across all the samples, compared to the others spiked in sorbent tubes before thermal desorption.

[0181] Additionally, the data matrix was edited by removing the deuterated features deriving from the internal standard solution, the siloxanes, which are mostly volatiles that derived from the chromatographic column bleed, and the features present only in one headspace sample, in order to work on more stable and consistent compounds.

[0182] TD-GCxGC-FID-MS analysis

[0183] Breath samples analysis was performed by a two-dimensional gas chromatograph (Agilent 7890 A) which was linked to a flow modulator (G3486 A CFT) and a three- way splitter plate coupled to a flame ionisation detector and to a single quadrupole mass spectrometer (HES 5977B) with electron ionisation ion source (Agilent Technologies Ltd, Stockport, UK). The GCxGC-FID / MS analytical method and the used parameters are summarised in Table 6.

[0184] For all the volatile biomarkers validation studies, chemical standards for the in vitro selected VOCs - which discriminated non-eosinophil-enriched sputa from eosinophil- enriched sputa -were analysed by TD-GCxGC / MS and the chromatographic data files were used for VOCs targeted analysis in breath samples. Sampled tubes containing exhaled breath samples were capped with diffusion locking (DiffLok) caps (Markes International Ltd) after being spiked with the internal standard solution, and then placed on trays into a thermal desorption auto-sampler (Markes TD- lOOxr thermal desorption auto-sampler - Markes International Ltd, Llantrisant, UK). Diffusion locking caps prevented sample loss and contaminants entrance onto the tube. A tray comprised six samples, a reference mixture of n-alkanes and aromatics, and an analytical blank sample.

[0185] Reference solution containing saturated alkanes at a concentration of 10 pg / ml and aromatics calibration standards at a concentration of 20 pg / ml diluted in methanol was analysed by TD-GCxGC-FID / qMS in order to monitor the retention behaviour and to facilitate separation of different VOCs and their chemical identity assignment.

[0186] The auto-sampler loaded a tube which was purged with helium at a flow rate of 50 ml / min for 1 minute, and then desorbed at 300 °C for 5 minutes into a general purpose hydrophobic trap (Markes International Ltd, Llantrisant, UK) which was held at -10 °C and that matched the tube sorbent material. The trap was purged for 2 minutes at 2 ml / min and then was heated for 5 minutes to reach 300 °C with a split flow rate of 2 ml / min. An empty tube with no sorbent material was loaded by the auto-sampler between each breath sample and a bake-out method was performed, with the primary column flow increased to 1.5 ml / min and the oven was held at 250 °C for 30 min. After the analysis of each tray batch, a trap blank was always analysed by purging the trap for 2 min at 20 ml / min and then desorbing it at 300 °C for 5 min with a split flow rate of 2 ml / min (Wilde et al, 2019).

[0187] Two-dimensional gas chromatography allowed the separation of all the compounds according to their volatility and functional group along two GC columns containing different stationary phases. The primary column contained a non-polar stationary phase made of 5% phenyl 95% dimethylpolysiloxane (30 m x 0.25 mm x 0.25 pm Rxi-5Sil MS - (Restek Thames Ltd, Saunderton, UK)), and the second was a polar column whose configuration was a polyethylene glycol 4 m x 0.25 mm x 0.25 pm DB-WAX (Agilent Technologies Ltd, Stockport, UK). Data analysis and chemometrics

[0188] TD-GCxGC-MS data were acquired in MassHunter GC-MS Acquisition B.07.04.2260 (Agilent Technologies Ltd, Stockport, UK) and processed using GC Image™ v2.6 along with GC Project and Image Investigator (JSB Ltd, Horsham, UK). A reference mixture of M-alkane and aromatics was analysed at the beginning of each tray generating a chromatographic file, which was used to correct variation in the retention time position captured by the reference chromatograms in each sample batch (Wilde et al, 2020). Reference mixtures analyses helped to separate different chemical classes and to perform VOCs class assignment, based on the analysis of reference compounds, their elution order and mass spectral library matching and interpretation. An example of chromatogram (represented as colour plot) generated from a breath sample analysis in shown in Figure 7. VOCs chemical classes detected in breath samples by TD- GCxGC / MS had a typical distribution on the chromatograms. Branched and unsaturated alkenes and cycloalkanes are the most abundant non-polar compounds present in breath samples that were mostly separated on the first dimension. Polar compounds such as aromatic carbonyls, amines, alcohols and phthalates, were located at the top section of the chromatogram, which had a quite low peak intensity, as they eluted late in the second dimension. Saturated, unsaturated and cyclic carbonyls such as aldehydes ketones, furans, esters and sulphide occupied the middle region of the chromatogram, while siloxanes - which mostly originate from environmental air, facemask and chromatographic column bleeding - were located at the lower chromatographic region.

[0189] Chemical standards for the selected VOCs, which were identified in sputum headspace, were run on TD-GCxGC / MS and the chromatographic data files were used for VOCs targeted analysis in breath samples (in both the studies that involved exhaled breath analysis). Data matrix comprised the peak area values of the selected VOCs, which were log transformed and adjusted for batch effect dividing the product between VOC abundance and VOC abundance average by the VOC abundance standard deviation.

[0190] Extended statistical methods

[0191] Penalised regression methods: elastic net and biomarker sample score calculation An unsupervised statistical approach did not perform well in dealing with headspace and breathomic data. Indeed, sputum headspace and breath samples analysed by TD- GC / MS and TD-GCxGC / MS contained hundreds of chromatographic features, of which uninformative variables dominated a vast proportion. Too many redundant variables negatively affected the selection of projection directions, leading to overfitting and low prediction accuracy. Therefore, a supervised statistical approach was used to select those VOCs which were able to discriminate between eosinophil enriched and noneosinophil enriched sputa and so were able to predict airway eosinophilia.

[0192] For the in vitro volatile biomarker discovery study a supervised multivariate analysis was performed using penalized regression techniques and providing class information (eosinophil-enriched and non-eosinophil-enriched sputa) for each observation.

[0193] More specifically, an elastic net regression was fitted to a data set of 393 chromatographic features, which were detected in 36 sputum headspace samples, 22 classified as eosinophil-enriched and 14 as non-eosinophil-enriched sputa.

[0194] Elastic net has got the advantage of performing with variable multi-collinearity as Ridge regression, and the ability of variable selection performed by Lasso, by combining LI and L2 regularization penalties (Zou & Hastie, 2005) and therefore it was the chosen method for this study.

[0195] A 10-fold cross validation (CV) was used to estimate how the model was expected to perform when it was used to make predictions on data that were not used during the model training. Ten represents the number of groups in which the data were split; therefore the set of observations was divided into 10 groups. Each group was taken as test data set once while the others were treated as training data set, then the model was fitted on the training set and assessed on the test data set.

[0196] The elastic net model allowed estimating the regression coefficient of all the detected volatile features, in order to identify those with a non-zero regression coefficient in at least 80 out of 100 runs of a 10-fold cross validation. The selected chromatographic features represented the most highly discriminatory volatiles, able to predict the class to which a sample belonged. The model estimated the average coefficient of each selected volatile feature, which together with the respective peak area value was used to calculate a sample score. The score attributed to each sample derived from the equation shown in Figure 8. Where p0is the regression model intercept (constant), Pi is the average regression coefficient for the ithfeature obtained after running the regression model over 100 cross-validations, x, is the peak area value for the ithfeature, and p is the number of features (VOCs).

[0197] Comparison of the selected volatile biomarker peak area values between sputum headspace and control headspace samples

[0198] Figure 9 reports the comparative analysis of the abundance values of the selected volatile biomarkers between sputum headspace and control headspace samples.

[0199] Evaluation of the model classification performance: ROC curve analysis

[0200] ROC curve analysis was used for assessing the ability of the statistical model to discriminate samples between two classes (such as eosinophil-enriched and non- eosinophil-enriched sputa; exhaled breaths samples with blood eosinophils > or < 0.3xl09cell / L, and exhaled breaths samples with sputum eosinophils > 3%or < 3%).

[0201] A ROC curve was obtained by plotting on the y-axis the model sensitivity, and on the x-axis the model false positive rate (1 -specificity), which were both measured across the range of all the possible score values.

[0202] Area under the ROC curve (AUC) represented the predictive accuracy of the model to distinguish between classes, and it was used to summarise the performance of each score (classifier) into a single measure.

[0203] According to Swets’ classification to interpret the AUC values (Swets, (1988) ): a model with an AUC that ranges between 0.5 and 0.7 has a low discriminating accuracy, while an AUC oscillating between 0.7 and 0.9 indicates a model with a moderate discriminating accuracy, and an AUC comprised between 0.9 and 1.0 designates a model with a high discriminating accuracy.

[0204] Extended clinical methods

[0205] Clinical assessment, demographics and asthma diagnosis

[0206] A clinical assessment was performed to estimate the presence and recurrence of clinical symptoms and signs related to airway hyper-responsiveness and airway obstruction. The predictive value of a combination of different symptoms and signs was considered clinically relevant and more helpful in asthma diagnosis compared to the presence of single symptoms. Typical clinical symptoms and signs considered for asthma diagnosis were recurrent wheezes, cough and / or breathlessness. Furthermore, the presence of triggers that made symptoms worse was estimated by skin prick test to aeroallergens, by increased serum IgE (total and specific) and of peripheral blood eosinophils. Skin prick test required the patients to stop any antihistaminic assumption 72 hours before the test. The test was performed using a panel of common allergens (such as grass pollen, tree, dog and cat fur, Aspergillus fumigatus and Penicillum notatum etc.) and saline solution and histamine as controls. After 15 minutes the formation of a wheal larger than 2 mm compared to controls was considered a positive response. Additionally, several other pieces of information were recorded to characterise the asthmatic cohorts, such as baseline demographic information (age, gender, body mass index, smoking habits and exposure), disease history (age at onset, duration of asthma, exacerbation frequency and ITU admission), personal or family history of atopic disorders such as eczema and rhinitis that increase the probability of asthma, and current medications.

[0207] Spirometry and bronchodilator reversibility test

[0208] Obstructive spirometry was used to estimate the ratio between FEV1and FVC, where FEVi represented the exhaled volume at the end of the first second of a forced expiration, and FVC the vital capacity from a maximally forced expiratory effort. Patients were required to stop their ICS 6 / 8 hours prior the test and not to smoke. Their nose was closed so that they were forced to breath into a mouthpiece linked to the spirometer. They were asked to perform a full inhalation followed by a forced exhalation. The procedure was repeated eight times with one minute pause between each repetition. The best values for FEVi and FVC were recorded and used to calculate the reference value and the lower limit of normality for each patient according to the global lung function equation (Quanjer et al, 2012). According to GINA guidelines, the FEVi / FVC ratio cut-off of normality ranges between 0.75-0.80 in adults, and it usually decreases in asthmatics due to the airway obstruction.

[0209] The bronchodilator reversibility test was performed to measure, through spirometry, the improvement of the expiratory airflow 15 minutes after patient treatment with a bronchodilator (salbutamol). It was performed after the normal spirometry to assess patient expiratory airflow limitation and to perform a differential diagnosis of asthma and COPD, since flow-related bronchodilator reversibility is more strongly associated with asthma than COPD. A positive response to salbutamol was usually given by an increase of FEVi of 200 ml.

[0210] Asthma inflammatory biomarkers

[0211] FeNO

[0212] Fractional exhaled nitric oxide (FeNO) was measured in exhaled breath samples from all the four independent cohorts of asthmatics in order to assess the airway eosinophilic inflammation, although the amount of FeNO is not always strictly associated with airway eosinophilic inflammation. FeNO levels indeed can increase in cases of allergic rhinitis, rhinovirus infection in healthy individuals, in men, tall people and in case of consumption of dietary nitrates, and can be lower in children, in cigarette smokers or in cases of assumption of inhaled or oral corticosteroids.

[0213] FeNO levels were measured using an electrochemical analyser (Niox Vero, Aerocrine, Sweden), at a flow rate of 50 ml / sec before taking any inhaled drug.

[0214] Levels of FeNO higher than 40 parts per billion by volume (ppb) were considered a supportive but non-conclusive sign of asthma in steroid naive adults. While 27 ppb was the threshold to identify a sputum eosinophil count > 3% in patients that were treated with high dose of ICS (Schleich et al, 2016a).

[0215] Total and differential cell count in sputum samples

[0216] Sputum samples were used not only to collect the headspace in the in vitro study but also to estimate the sputum cellularity and to assess the granulocytic airway inflammatory phenotype.

[0217] After the immediate headspace collection for the in vitro study, and directly after sputum production for the validation studies, the sputa were processed to estimate the total and differential count of inflammatory cells.

[0218] Sputum plugs were selected, weighed and diluted in a homogenization solution made of 0.1% DTT (dithiothreitol), and after 15 minutes were diluted in cold PBS with Ca2+and Mg2+and centrifuged. The cells were stained with Trypan Blue and counted using a haemocytometer to record the cell viability (equation 1 : L= live cells, D= dead cells), the squamous cells count (equation 2: Sq. = squamous cells), and to calculate the number of total non-squamous cells, which was expressed as millions per gram of selected sputu

[0219] Total cells x 106Total cells x 106xl000

[0220] Eq. 3: — - = - - - g (selected sputum) sputumplugs weight (mg)

[0221] The sample was centrifuged and the pellet was re-suspended in PBS and then placed into a cytospin, where it was centrifuged again at 459xg for six minutes in order to have the cells laid on a slide. The slide underwent to Rapidoff-2 Romanovsky staining to perform the differential cell count for eosinophils, neutrophils, macrophages, lymphocytes and bronchial epithelial cells. Differential cell counts were reported as a percentage of the total non-squamous cell count. Sputa with a percentage of non- squamous cells lower than 20 and a viability higher than 50% were classified as good samples.

[0222] In the in vitro biomarker discovery study, sputum differential cell count was used to classify patients’ sputa in eosinophil-enriched and non-eosinophil-enriched sputa. The eosinophil and neutrophil upper limit values used for asthmatics were 3% and 65% respectively (Green et al, 2002). Therefore, sputum samples with an eosinophil differential count > 3% and a neutrophil count> 65% or < 65% were considered as eosinophil-enriched, while all the other samples that did not meet these criteria were classified as non-eosinophilic sputa.

[0223] Peripheral blood eosinophils

[0224] Blood eosinophil count was performed for all the cohorts employed in the study. Patients with asthma and blood eosinophil count greater than 0.4xl09cells / L are usually subject to frequent exacerbations and poorer asthma control (Price et al, 2016). However, a slightly higher blood eosinophil threshold (0.5xl09cells / L) was used for the cohort of acutely exacerbating adult asthma patients (cohort 2 in validation studies) to classify them in acute eosinophilic and non-eosinophilic patients, in order to include in the first group patients with airway eosinophilia. Indeed, a higher blood eosinophil threshold was proved to have a better sensitivity and specificity in predicting sputum eosinophilia (sputum eosinophils > 3%) (Negewo et al, 2016). Asthma severity-GINA guidelines

[0225] All the patients were classified as severe or difficult to treat asthmatics, corresponding to GINA steps 5 and 4. Moderate asthmatics (steps 2 and 3) were included in the cohort 1 used for a validation study.

[0226] Asthma control and quality of life questionnaires

[0227] Patient answered the Juniper asthma questionnaire (ACQ) and the Juniper asthma quality of life questionnaire (AQLQ) in order to assess asthma controls. ACQ included seven questions, five related to symptoms, one on the rescue treatment use and one on pre-bronchodilator % predicted FEVi finding. In this study a modified and validated version of ACQ, which was characterised by the absence of point seven, was used as the pre-bronchodilator test was assessed only during the first visit. Both the control of symptoms and the % predicted FEVi were described on a severity scale of 0-6. The overall score was given dividing by six the total score of symptoms and rescue treatment usage questions. The minimal clinical significant difference in the score value between visits was 0.5.

[0228] AQLQ assessed the asthma-specific health-related quality of life measures. It comprised of 32 items that covered four domains: symptoms (11 items), activity limitation (12 items), emotional function (5 items), and environmental exposure (4 items). The overall score was given by the arithmetic average of the scores of each domain. 0.5 was the minimally relevant difference in score for overall quality of life and for each of the individual domains.

[0229] Statistical methods for discovery and validation studies

[0230] Receiver operating characteristic (ROC) curve analysis of the generated scores for the discovery and the clinical validation cohorts was performed in order to assess the developed statistical model sensitivity, specificity and accuracy to discriminate between

[0231] (I) headspace samples from eosinophil-enriched and the non-eosinophil enriched sputa,

[0232] (II) breath samples from acute asthmatics with and without blood eosinophilia, (III) breath samples from acute asthmatics with and without sputum eosinophilia, and (IV) breath samples from stable severe asthmatics with and without blood eosinophilia. Patient clinical characteristics, normalised peak area values of the selected VOCs, and VOC sample scores were summarised as median and interquartile range (IQR), and were compared within the respective subgroups of each cohort performing a two-tailed nonparametric t-test (Mann-Whitney test). For the discovery cohort, a Spearman’s correlation between the generated biomarker scores for eosinophil-enriched and noneosinophil enriched sputa, and both the percentage and the absolute value of sputum eosinophils was estimated. TD-GC-MS analysis of 26 control headspace samples was performed in order to estimate the presence of sputum headspace selected VOCs in the background volatile signal. Statistical comparative and descriptive analyses were performed using GraphPad Prism 8 (GraphPad Prism 8 Software Inc., La Jolla, CA, USA).

[0233] Results

[0234] Sputum headspace discovery study

[0235] Figure 1 summarises both the design of the sputum headspace volatile biomarker-based model discovery study to predict airway eosinophilia, and the model clinical validation studies in exhaled breath samples. Table 1 reports the clinical characteristics of the discovery cohort, while Table 2 summarises the sputum differential cell count results. Sputum sample classification according to sputum eosinophil count threshold of 3% revealed 22 eosinophil-enriched sputa (> 3%) and 14 non-eosinophil-enriched sputa (< 3%). The comparison of the differential and total sputum eosinophil count (median, IQR) between the two groups showed, as expected, a significant higher number of eosinophils in eosinophil-enriched sputa compared to the other group (p<0,0001; Table 2).

[0236] Table 1 - Clinical characteristics of severe asthmatic donors of sputum samples. Data are summarised as median (Q1-Q3) and a non-parametric test (Mann-Whitney) was performed for continuous variables.

[0237] ’Patients with+Patients with p-value eosinophil-enriched non-eosinophil sputa (n=22) enriched sputa (n=14)

[0238] Age (years) 67 (58-71) 58 (56-66) 0,23

[0239] BMI (Kg / m2) 29,6 (26,4-37,6) 31,1 (29,1-35,2) 0,49

[0240] Sex (% females) 45 28 0,31

[0241] Age of asthma onset 39 (7-53) 35 (29-49) 0,72 Asthma duration (years) 24 (8-33) 23 (12-38) 0,95

[0242] Smoking status 1 / 20 - 13 / 20 -8 / 20 0 / 12 - 7 / 12 - 7 / 12 0,56 (current-never-ex)

[0243] Post bronchodilator FEVi 75 (53,8-92,4) 53 (49,5-81) 0,17

[0244] (%predicted)

[0245] Post bronchodilator 89,1 (76,3-99,3) 73,9 (66,7-99,1) 0,47

[0246] FEVi / FVC (%predicted)

[0247] FeNOso (ppb) 37 (23-52) 26 (14-68.5) 0,40

[0248] Blood eosinophils (xlO9 / L) 0,21 (0,08-0,46) 0,10 (0,05-0,3) 0,30

[0249] ACQ 6-score 1,67 (1 - 3,17) 2,67 (0,67-4,08) 0,85

[0250] *Atopy (%Yes) 53 60 0,70

[0251] Number of patients on 13 / 22 7 / 14 0,31 daily dose of OCS Number of patients on 20 / 22 11 / 14 > 0,99 high dose of ICS Number of patients on 21 / 22 12 / 14 > 0,99 long-acting beta antagonist (LABA) Number of patients on 11 / 22 5 / 14 0,72 long-acting muscarinic antagonist (LAMA ) Anti-leukotrienes 5 / 22 5 / 14 0,24 (Montelukast) Number of patients on 8 / 22 5 / 14 0,71 anti-IL5 Anti-IL5 treatment 8 (6-12) 12 (4-16) 0,98 duration at the time of sputum headspace VOC sampling (week)

[0252] * Sputum eosinophils > 3%

[0253] I Sputum eosinophils < 3% f Positive skin prick test

[0254] Table 2 - Total and differential cell count of severe asthmatics’ sputum samples. Data are summarised as median (Q1-Q3) and a non-parametric test (Mann-Whitney) was performed for continuous variables. ’Patients with+Patients with non- p-value eosinophil- eosinophil enriched sputa enriched sputa (n=22) (n=14)

[0255] Total cells (xl06 / g) 3,02 ( 1,03-5,51) 2,86 ( 1,46-3,84) 0,88 Eosinophils (%) 5,37 (3,00-8,75) 0,37 (0,25-0,93) <0,0001 Eosinophils (xlOe / g) 0,09 (0,08-0,20) 0,012 (0,007-0,03) <0,0001 Neutrophils (%) 74,62 (65,01- 77,52 (64,43-91,95) 0,39

[0256] 83,37)

[0257] Neutrophils (xlOe / g) 1,82 (0,63-3,72) I,84 (0,97-2,85) 0,88 Macrophages (%) 14,33 (6,87-23, 12) II,37 (5,5-30,25) 0,73 Macrophages (xlOe / g) 0,40 (0, 13-1, 10) 0,27 (0, 14-0,82) 0,66 Lymphocytes (%) 0,39 (0,00-0,75) 0,87 (0,31-1,75) 0, 14 Columnar epithelial cells 3,75 (0,76-5,50) 3,25 ( 1,31-6,68) 0,57

[0258] (%)

[0259] Squamous cells (%) 1,91 (0,03-5,35) 5,35 (2,43-8,51) 0,08 Viability (%) 69,86 (61,45- 68,55 (55,28-74,85) 0,76

[0260] 83,50)

[0261] Plugs weight (mg) 189,5 ( 106,5- 141,5 (57,75- 0,51

[0262] 264,25) 213,75)

[0263] * Sputum eosinophils > 3%

[0264] I Sputum eosinophils < 3%

[0265] TD-GC-MS analysis of 36 sputum headspace samples detected 916 chromatographic features of which 513 were removed because they were present in just one sample or represented siloxanes, therefore, the eNET regression was fitted to a data matrix of 393 features. The eNET regression selected 19 features (VOCs).

[0266] The normalised peak area value of each VOCs was compared between eosinophil- enriched and non-eosinophil-enriched sputa, and significant differences between the groups were observed for 1-hexanol (p= 0.02), styrene (p= 0,017,), phenol (p= 0,005), decane (p= 0,019), benzothiazole (p= 0,03) (Figure 2A). The majority of the other VOCs showed higher normalised peak area values in headspace of non-eosinophil-enriched sputa compared to headspace of eosinophil-enriched sputa. In Figure 2B the biomarker scores (median, (IQR)) generated for each sputum headspace samples were compared between eosinophil-enriched and non-eosinophil-enriched sputa, showing a statistical significant difference (p= 0,0001) with score values higher for non-eosinophil-enriched sputum headspace samples compared to eosinophil-enriched ones.

[0267] Receiver operating characteristic (ROC) curve to estimate the model ability to discriminate between eosinophil-enriched and non-eosinophil-enriched sputa, showed an AUC of 0,88 (p<0,0001) (Figure 2C).

[0268] The assessment of the correlations between the biomarker sample scores and the percentage and the absolute sputum eosinophil count for each sample are reported in Figure 2D and E. The Spearman’s correlation coefficients and p values proved a decreasing monotonic trend between the percentage of sputum eosinophils and sample score values (rs= -0,71; p<0,0001) and between the absolute number of sputum eosinophils and the score values (rs= -0,54; p= 0,0006).

[0269] The selected VOCs are listed and grouped according to their chemical class in Table 3, which also reports the results of a literature search to check whether the selected VOCs had previously been identified as biomarkers of airway eosinophilia or neutrophilia, and the detected VOCs putative origin / metabolic pathway in human. Two of the 19 selected VOCs in our study, the nonanal and the 2-butoxy-ethanol, had already been reported in literature as potential biomarkers of airway neutrophilia. Nonanal was detected in breath samples of asthmatics as biomarker able to discriminate neutrophilic from eosinophilic asthma (6). 2-Butoxyethanol, instead, was reported as biomarker able to discriminate headspace of activated blood neutrophil cultures from headspace of activated blood eosinophil cultures (8).

[0270] Table 3 - Chemical classification, association to eosinophilic inflammation and asthma, and putative origin of the 19 sputum headspace volatile biomarkers on which is based the developed model predictive of sputum eosinophilia.

[0271] TD-GC-MS analysis of control headspace samples generated a data matrix of 969 chromatographic features, which were identified across all the 26 control headspace samples. After the editing process, the number of the features in the data matrix was lowered to 341. All the 19 volatile biomarkers, which were selected in sputum headspace samples, were identified also in control headspace samples, excluding nonanal and tridecane. Comparative analysis of the abundance values of the selected volatile biomarkers between sputum headspace and control headspace samples (Figure 9) revealed a significant difference in peak area values for acetone (p<0,0001), hexanal (p= 0,013), styrene (p<0,0001), phenol (p=0,002), benzothiazole (p=0,04) and 2- ethylhexanal (p= 0,006). Normalised peak area values of these components were all enriched in sputum headspace compared to control headspace samples, with the exception of benzothiazole, whose peak area values were higher in control headspace samples.

[0272] Clinical validation study for the developed volatile biomarker-based model

[0273] Demographics and clinical characteristics for the acute asthmatic cohort 1 are summarised in Table 7.

[0274] Table 7 Clinical characteristics and demographics of acute asthmatics in cohort 1, grouped according to the blood eosinophil threshold of 0-3xl09 / L. Unpaired two-tailed t test was performed for continuous variables. Chi square test was performed for categorical variables. Statistically significant differences were reported for: baseline blood eosinophil count (p<0.0001)*, and for sputum eosinophil count (p=0.0002) **

[0275] N Acute asthmatics N Acute asthmatics p-value with blood with blood eosinophils eosinophils < 0-3xl09 / L (n = 52) > 0-3xl09 / L (n = 13) Age 13 36 (23-5-54) 52 43-5 (27-25-59-75) 0-45

[0276] Sex (%females) >0-99

[0277] 13 30 52 71-1

[0278] BMI (Kg / m2) 0-17

[0279] 12 27-2 (23-8-30-8) 51 30-8 (25-2-36-01)

[0280] Smokers-Non-Ex >0-99

[0281] 13 15-3- 46-1-38-4 52 23 - 50- 27

[0282] (%)

[0283] Blood Eosinophils <0-0001*

[0284] 13 0-77 (0-61-1-42) 52 0.13 (0-05-0-25)

[0285] (xlO9 / L)

[0286] Sputum

[0287] 6 30-2 (7-3-36) 18 0-25 (0-25-2-31) 0-0002** eosinophils (%)

[0288] Forced oscillation 0.76

[0289] 11 1-4 (0-3-1-9) 44 1-22 (0-5-2-08) technique R5-R19

[0290] AQLQ score 0-61

[0291] 8 21-5 (18-6-23-5) 24 18-7 (14-1-25-04)

[0292] Exacerbations in 0-06

[0293] 11 0 (0-3) 48 2 (1-4) past 12 months

[0294] OCS daily dose 0-86

[0295] 12 12-5 (5-20) 52 10 (5-20)

[0296] (mg)

[0297] ICS (%Yes) 12 83-3 52 69 0-48

[0298] According to the blood eosinophil count 13 patients were classified as acute eosinophilic (blood eosinophils > 0,3xl09 / L), and 52 as acute non eosinophilic asthmatics (blood eosinophils < 0,3xl09 / L).

[0299] 16 out of the 19 sputum headspace selected VOCs were detected in exhaled breath samples; the missing VOCs were phenol, 2-butoxyethanol and benzothiazole. VOCs peak area values are reported in Figure 3A, and compared between acute asthmatics with blood eosinophils > 0,3xl09 / L, and acute asthmatics with blood eosinophils < 0,3 X109 / L. Statistically significant differences for decanal (p= 0,04) and hexanal (p=

[0300] 0,02) levels between the acute groups.

[0301] Biomarker scores (median (IQR)) generated using the in vitro developed model were compared between the acute asthmatics groups in Figure 3C. No statistically significant differences were reported in exhaled sample scores between groups.

[0302] ROC curve analyses in Figure 3E showed that the selected volatile biomarker-based model had a high accuracy (AUC: 0,89, p<0,0001) to discriminate acute asthmatics with blood eosinophils > 0,3xl09 / L from acute asthmatics with blood eosinophils < 0.3xl09 / L.

[0303] The volatile biomarker-based model was then validated in the adult acute asthma sub- cohort of cohort 1 (Table 8), which comprised of 24 acute asthmatics grouped according to the sputum eosinophil threshold of 3%. The log transformed peak area values of the 16 selected VOCs are shown in Figure 3B. Hexanal and decanal levels were significantly increased in acute asthmatics with sputum eosinophils < 3%. The volatile biomarker scores (median (IQR)) were compared between the two acute asthmatic groups in Figure 3D, showing no statistically significant differences, furthermore they had a good discrimination accuracy (AUC: 0,79, p<0,0001) between the two acute groups (Figure 3F).

[0304] Table 8 Clinical characteristics and demographics of acute asthmatics in sub-cohort 1, grouped according to the sputum eosinophil threshold of 3%. Unpaired two-tailed t test was performed for continuous variables. Chi square test was performed for categorical variables. Statistically significant differences were reported for: baseline blood eosinophil count (p<0.0000)*, and for sputum eosinophil count (p=0.0001) ** .

[0305] N Acute asthmatics N Acute asthmatics p-value with sputum with sputum eosinophils eosinophils

[0306] > 3% (n=9) < 3% (n = 15)

[0307] Age 9 50 (33-54) 15 43 (29-5-54-25) 0-87

[0308] Sex (%females) 9 33-3 15 73-3 0-09

[0309] BMI (Kg / m2) 9 30-11 (25-2-38-8) 14 34-3 (26-6-43-59) 0-43

[0310] Smokers-Non-Ex 9 1 / 9-3 / 9-5 / 9 15 4 / 15 - 6 / 15 - 5 / 15 0-49

[0311] (%)

[0312] Blood Eosinophils 9 0-54(0-27-0-76) 15 0-07 (0-035-0-16) <0-0009*

[0313] (xlO9 / L)

[0314] Sputum 9 9-5 (6-35) 15 0-25 (0-25-0-46)

[0315] 0-0001** eosinophils (%) Forced oscillation 8 1-94 (1-19-2-35) 14 1-20 (0-67-2-31) 0.52 technique R5-R19

[0316] AQLQ score 2 24 (22-5-25-5) 8 15-33 (12-8-21-29) 0-26 Exacerbations in 8 0 (0-0) 12 0-5 (0-1) 0-11 past 12 months

[0317] OCS daily dose 9 5 (0-20) 15 43 (29-5-54-25) 0-90

[0318] (mg)

[0319] ICS (%Yes) 9 100 15 73-3 0-37

[0320] Demographics and clinical characteristics of the 22 non-exacerbating severe asthmatics from cohort 2 were summarised in Table 9, and were compared between stable asthmatics with blood eosinophil levels > 0,3xl09 / L (n=12) and with blood eosinophil levels < 0,3xl09 / L (n=10). 17 out of 19 VOCs were detected in exhaled breath samples of this severe asthmatic cohort 2; phenol and 2-butoxyethanol were missing. VOCs peak area values were reported in Figure 4A, and were compared between stable severe asthmatics in presence and absence of blood eosinophilia, reporting no significant differences in their levels. The box plot in Figure 4B showed that no statistically significant differences were reported for the volatile biomarker score between the two stable severe asthmatic groups.

[0321] Table 9 Clinical characteristics and demographics of stable severe asthmatics in cohort 2, grouped according to the blood eosinophil threshold of 0-3xl09 / L. Unpaired two- tailed t test was performed for continuous variables. Chi square test was performed for categorical variables. Statistically significant difference was reported for baseline blood eosinophil count (p<0.0001)* .

[0322] Severe Severe asthmatics asthmatics with with blood blood

[0323] N N p-value eosinophils > eosinophils >

[0324] 0-3X109 / L 0-3X109 / L

[0325] (n=12) (n=10)

[0326] Age (Years) 12 53 5 (45-57-5) 9 56 (52-63.5) 0 66

[0327] BMI (Kg / m2) 10 30-2 (27-6-36-3) 9 28-6 (25-5-29-9) 0 43

[0328] Sex (% Females) 12 57 10 80 0 64

[0329] Duration of Asthma(Years) 12 18-5 (14-25) 9 31-5 (23-33) 0 32

[0330] 12 0 / 12-10 / 12 - 10 0 / 10- 8 / 10-2 / 10 >0 99

[0331] Smokers-Non-Ex (%)

[0332] 3 / 12

[0333] Post bronchodilator FEVi 11 94 (65 9-104) 10 76 (67-85-2) 0 24

[0334] (Predicted %) Post bronchodilator 10 94 (88-5-97-5) 10 99 (86-100) 0 38 FEVi / FVC (Predicted %) FeNOso 11 38 (26-68-5) 5 27-5 (16-7-43) 1-15

[0335] ACQ 6 score 11 2-33 (1-25-3-25) 7 2 33 (1 08-4 17) 0 98

[0336] AQLQ score 11 22-8 (19-9-25-8) 10 23 9 (15-6-30) 0 77

[0337] Atopy (%Yes) 12 75 10 50 0 37

[0338] High dose ICS (%Yes) 12 100 10 100 >0 99

[0339] OCS daily dose (mg) 11 12 5 (7-5-15) 10 10 (10-13.7) 0-51 Exacerbations in past 12 12 3-5 (1-6-5) 10 3 (0-4) 0 78 months

[0340] Baseline blood eosinophils 12 0-35 (0-3 -0-63) 10 0 1 (0-1-0-20) 0.0001* (xlO9 / L)

[0341] Additionally, the biomarker scores showed a low accurac y to predict blood eosinophilia in stable severe asthmatics ( AUC: 0,70, p=0,01) (Figure 4C).

[0342] Refined VOC Model

[0343] The inventors then fit a penalized regression model (elastic net, binary dependent variable) to the discovery cohort. The discovery cohort (sputum headspace) subjects were grouped according to the sputum eosinophil count. The original 19 VOCs identified as above were included as potential predictors of eosinophil count (EA or NEA). 10-fold cross-validation alpha (elastic net penalty) =0.22, lambda. min=0.0072, was obtained from a fixed grid search. This identified a refined group of VOCs which are able to be used to carry out the invention as defined above: Benzene, Benzothiazole, Decane, Isothiocyanato-cyclohexane, Methylstyrene, Phenol, Styrene, Toluene, 1 -Hexanol, 2-Butoxyethanol, p-Xylene

[0344] Table 10 - refined VOC model

[0345] Discussion

[0346] 40-60% of severe asthmatics show type 2 inflammation, characterised by the presence of airway eosinophilia (17). Assessing the presence of eosinophilic airway inflammation allows a better stratification of severe asthma patients for eosinophil targeted therapies, or allow corticosteroid titration, and helps preventing asthma exacerbations. Although the sputum eosinophil count is a good predictor of airway eosinophilia in severe asthma, it requires a time consuming and costly procedure. There is therefore the need to develop novel non-invasive biomarkers of eosinophilic airway inflammation in severe asthma.

[0347] This study developed a predictive model of sputum and therefore airway eosinophilia > 3%, which was based on selected VOCs detected in sputum headspace samples from severe asthmatics.

[0348] The inventor have developed a model which showed a high accuracy (AUC: 0-88; p<0-0001) in discriminating eosinophil-enriched and non-eosinophil-enriched sputum headspace samples (Figure 2C).

[0349] The 19 initially selected VOCs in sputum headspace samples were mostly represented by aromatics hydrocarbons, aldehydes and alkanes, and in a small amount by alcohols and ketones (Table 3).

[0350] After the model development and based on the evidence that increased levels of blood eosinophils are usually associated to the presence of sputum (31) eosinophilia in asthma, this study validated the model predictive of sputum eosinophilia in exhaled breath samples of different clinical cohorts, and tested its accuracy to predict (i) blood (eosinophils > 0-3xl09 / L) and sputum eosinophilia (eosinophils > 3%) in acute severe asthmatics, and (ii) blood eosinophilia in a stable severe asthma cohort.

[0351] The volatile biomarker-based model of sputum eosinophilia showed a good predictive accuracy of both blood (AUC: 0-89, p<0-0001) and sputum (AUC: 0-79; p<0-0005) eosinophilia in acute asthmatics (Figure 3 E, F). A key finding for the sputum headspace volatile biomarker-based model validation against blood eosinophilia in a stable severe asthma cohort 2 were that 17 / 19 VOCs were detected in exhaled breath samples (missing compounds: phenol, and 2- butoxyethanol), and that the model was proven to have a good accuracy in predicting blood eosinophilia in stable severe asthmatics with an AUC of 0-70, and a p-value of 0-01 (Figure 4C).

[0352] References

[0353] 1. Fahy JV. Type 2 inflammation in asthma— present in most, absent in many. Nat Rev Immunol. 2015 ; 15 ( 1 ) : 57-65 .

[0354] 2. GINA: Global Initiative for Asthma. Global Strategy for Asthma Management and Prevention, www.ginasthma.org. 2022.

[0355] 3. Schleich FN, Manise M, Sele J, Henket M, Seidel L, Louis R. Distribution of sputum cellular phenotype in a large asthma cohort: predicting factors for eosinophilic vs neutrophilic inflammation. BMC Pulm Med. 2013; 13: l l .

[0356] 4. Wan XC, Woodruff PG. Biomarkers in Severe Asthma. Immunol Allergy Clin North Am. 2016;36(3):547-57.

[0357] 5. Ibrahim B, Basanta M, Cadden P, Singh D, Douce D, Woodcock A, et al. Non- invasive phenotyping using exhaled volatile organic compounds in asthma. Thorax. 2011;66(9):804-9.

[0358] 6. Schleich FN, Zanella D, Stefanuto PH, Bessonov K, Smolinska A, Dallinga JW, et al. Exhaled Volatile Organic Compounds Are Able to Discriminate between Neutrophilic and Eosinophilic Asthma. Am J Respir Crit Care Med. 2019;200(4):444- 53.

[0359] 7. Ibrahim W, Natarajan S, Wilde M, Cordell R, Monks PS, Greening N, et al. A systematic review of the diagnostic accuracy of volatile organic compounds in airway diseases and their relation to markers of type-2 inflammation. ERJ Open Res. 2021;7(3).

[0360] 8. Schleich FN, Dallinga JW, Henket M, Wouters EF, Louis R, Van Schooten FJ. Volatile organic compounds discriminate between eosinophilic and neutrophilic inflammation in vitro. J Breath Res. 2016; 10( l):016006.

[0361] 9. Yamaguchi MS, McCartney MM, Falcon AK, Linderholm AL, Ebeler SE, Kenyon NJ, et al. Modeling cellular metabolomic effects of oxidative stress impacts from hydrogen peroxide and cigarette smoke on human lung epithelial cells. Journal of Breath Research. 2019; 13(3):036014. 10. Yamaguchi MS, McCartney MM, Linderholm AL, Ebeler SE, Schivo M, Davis CE. Headspace sorptive extraction-gas chromatography-mass spectrometry method to measure volatile emissions from human airway cell cultures. J Chromatogr B Analyt Technol Biomed Life Sci. 2018; 1090:36-42.

[0362] 11. Chung KF, Wenzel SE, Brozek JL, Bush A, Castro M, Sterk PJ, et al. International ERS / ATS guidelines on definition, evaluation and treatment of severe asthma. European Respiratory Journal. 2014;43(2):343-73.

[0363] 12. Peltrini R, Cordell RL, Ibrahim W, Wilde MJ, Salman D, Singapuri A, et al. Volatile organic compounds in a headspace sampling system and asthmatics sputum samples. Journal of Breath Research. 2021; 15(2):027102.

[0364] 13. Ibrahim W, Wilde M, Cordell R, Salman D, Ruszkiewicz D, Bryant L, et al. Assessment of breath volatile organic compounds in acute cardiorespiratory breathlessness: a protocol describing a prospective real -world observational study. BMJ Open. 2019;9(3):e025486.

[0365] 14. GINA: Global Initiative for Asthma. Global Strategy for Asthma Management and Prevention. 2019. www.ginasthma.org. Global strategy for Asthma Management and Prevention. .

[0366] 15. Peltrini R, Cordell R, Ibrahim W, Wilde M, Salman D, Singapuri A, et al. Volatile organic compounds in a headspace sampling system and asthmatics sputum samples. J Breath Res. 2020.

[0367] 16. Wilde MJ, Cordell RL, Salman D, Zhao B, Ibrahim W, Bryant L, et al. Breath analysis by two-dimensional gas chromatography with dual flame ionisation and mass spectrometric detection - Method optimisation and integration within a large-scale clinical study. Journal of chromatography A. 2019; 1594: 160-72.

[0368] 17. Wilde MJ, Zhao B, Cordell RL, Ibrahim W, Singapuri A, Greening NJ, et al. Automating and Extending Comprehensive Two-Dimensional Gas Chromatography Data Processing by Interfacing Open-Source and Commercial Software. Analytical Chemistry. 2020;92(20): 13953-60.

[0369] 18. Bel EH, Ten Brinke A. New Anti-Eosinophil Drugs for Asthma and COPD: Targeting the Trait! Chest. 2017; 152(6): 1276-82.

[0370] 19. Belizario JE, Faintuch J, Malpartida MG. Breath Biopsy and Discovery of Exclusive Volatile Organic Compounds for Diagnosis of Infectious Diseases. Frontiers in Cellular and Infection Microbiology. 2021; 10(783). 20. Pereira J, Porto-Figueira P, Cavaco C, Taunk K, Rapole S, Dhakne R, et al. Breath analysis as a potential and non-invasive frontier in disease diagnosis: an overview. Metabolites. 2015 ;5 ( 1 ) : 3 -55 .

[0371] 21. Smolinska A, Klaassen EM, Dallinga JW, van de Kant KD, Jobsis Q, Moonen EJ, et al. Profiling of volatile organic compounds in exhaled breath as a strategy to find early predictive signatures of asthma in children. PLoS One. 2014;9(4):e95668.

[0372] 22. Samadi MT, Shakerkhatibi M, Poorolajal J, Rahmani A, Rafieemehr H, Hesam M. Association of long term exposure to outdoor volatile organic compounds (BTXS) with pro-inflammatory biomarkers and hematologic parameters in urban adults: A cross- sectional study in Tabriz, Iran. Ecotoxicol Environ Saf. 2019; 180: 152-9.

[0373] 23. Dallinga JW, Robroeks CM, van Berkel JJ, Moonen EJ, Godschalk RW, Jobsis Q, et al. Volatile organic compounds in exhaled breath as a diagnostic tool for asthma in children. Clin Exp Allergy. 2010;40( l):68-76.

[0374] 24. Buckley JP, Quiros-Alcala L, Teitelbaum SL, Calafat AM, Wolff MS, Engel SM. Associations of prenatal environmental phenol and phthalate biomarkers with respiratory and allergic diseases among children aged 6 and 7 years. Environ Int. 2018; 1 15 :79-88.

[0375] 25. Li J, Peng Y, Liu Y, Li W, Jin Y, Tang Z, et al. Investigation of potential breath biomarkers for the early diagnosis of breast cancer using gas chromatography-mass spectrometry. Clin Chim Acta. 2014;436:59-67.

[0376] 26. Neerincx AH, Vijverberg SJH, Bos LDJ, Brinkman P, van der Schee MP, de Vries R, et al. Breathomics from exhaled volatile organic compounds in pediatric asthma. Pediatr Pulmonol. 2017;52( 12): 1616-27.

[0377] 27. Porter L, Toepfner N, Bashant KR, Guck J, Ashcroft M, Farahi N, et al. Metabolic Profiling of Human Eosinophils. Frontiers in Immunology. 2018;9(1404).

[0378] 28. Filipiak W, Sponring A, Mikoviny T, Ager C, Schubert J, Miekisch W, et al. Release of volatile organic compounds (VOCs) from the lung cancer cell line CALU-1 in vitro. Cancer Cell International. 2008;8( l): 17.

[0379] 29. Contoli M, Pauletti A, Rossi MR, Spanevello A, Casolari P, Marcellini A, et al. Long-term effects of inhaled corticosteroids on sputum bacterial and viral loads in COPD. European Respiratory Journal. 2017;50(4): 1700451 .

[0380] 30. Gan WQ, Man SF, Sin DD. Effects of inhaled corticosteroids on sputum cell counts in stable chronic obstructive pulmonary disease: a systematic review and a metaanalysis. BMC Pulm Med. 2005;5 :3. 31. Lima-Matos A, Ponte EV, de Jesus JPV, Almeida PCA, Lima VB, Kwon N, et al. Eosinophilic asthma, according to a blood eosinophil criterion, is associated with disease severity and lack of control among underprivileged urban Brazilians. Respiratory Medicine. 2018; 145:95-100. 32. Caldeira M, Perestrelo R, Barros AS, Bilelo MJ, Morete A, Camara JS, et al.

[0381] Allergic asthma exhaled breath metabolome: a challenge for comprehensive two- dimensional gas chromatography. J Chromatogr A. 2012; 1254:87-97.

[0382] 33. Pizzichini MM, Popov TA, Efthimiadis A, Hussack P, Evans S, Pizzichini E, et al. Spontaneous and induced sputum to measure indices of airway inflammation in asthma. Am J Respir Crit Care Med. 1996; 154(4 Pt 1): 866-9.

[0383] 34. Reddel HK, FitzGerald JM, Bateman ED, Bacharier LB, Becker A, Brusselle G, et al. GINA 2019: a fundamental change in asthma management: Treatment of asthma with short-acting bronchodilators alone is no longer recommended for adults and adolescents. Eur Respir J. 2019;53(6).

Claims

CLAIMS1. A panel of biomarkers, comprising or consisting of: benzene, benzothiazole, decane, isothiocyanato-cyclohexane, methylstyrene, phenol, styrene, toluene, 1 -hexanol, 2- butoxyethanol, and p-xylene.

2. The panel of claim 1, further comprising or consisting of one or more or, or all of: acetone, benzaldehyde, decanal, nonanal, hexanal, 2-ethylhexanal, tridecane, and methylene chloride.

3. The panel of claim 1 or claim 2, for use in diagnosing eosinophilia in a subject.

4. The panel for use according to claim 3, wherein the panel is detected in a biological sample obtained from the subject.

5. A method of diagnosing eosinophilia in a subject, comprising: i. providing a biological sample obtained from the subject; ii. determining the level of the panel of biomarkers of claim 1 or claim 2; iii. comparing the level of the panel of biomarkers in the sample with a reference level of the same panel of biomarkers; iv. using the results from (iii) to provide a diagnosis to the subject.

6. The panel for use according to claim 3 or claims 4, or the method of claim 5, wherein the eosinophilia is indicative that the subject is suffering from severe asthmatic exacerbations and / or that the subject has a stable-severe or severe asthma.

7. The panel for use according to claim 3 or claim 5, or the method of claim 4 or claim 5, wherein the subject has previously been diagnosed with asthma.

8. The panel for use according to any of claims 4 or 6-7, or the method of any of claims 5-7, wherein the sample is a breath sample, sputum sample or blood sample.

9. The panel for use of claim 8, or the method of claim 8, wherein the panel is detected in the headspace of the sample obtained from the subject.

10. A method of treating an asthmatic subject with eosinophilia and / or who is suffering severe exacerbations comprising: i. providing a biological sample obtained from the subject; ii. determining the level of the panel of biomarkers of claim 1 or claim 2; iii. comparing the level of the panel of biomarkers in the sample with a reference level of the same panel of biomarkers; iv. using the results from (iii) to determine if the subject has eosinophilia and / or is suffering severe exacerbations; and v. administering to the subject a therapeutic agent to treat or reduce the symptoms of the eosinophilia and / or severe exacerbations if the subject is determined in (iv) to have eosinophilia and / or is suffering severe exacerbations.

11. A method of monitoring treatment efficacy in a subject with eosinophilia and / or who is suffering severe exacerbations comprising: i. providing a first biological sample obtained from the subject prior to the administration of treatment; ii. determining the level of the panel of biomarkers of claim 1 or claim 2 in the first biological sample; iii. providing a second biological sample obtained from the subject after the administration of treatment; iv. determining the level of the same panel of biomarkers in the second biological sample as those in the first biological sample; v. comparing the level of the panel of biomarkers in the first sample with the level in the second sample to determine if the treatment is effective.

12. A kit comprising means for determining the level of the panel of biomarkers of claim 1 or claim 2.

13. The kit of claim 12, wherein the kit comprises biological sample collection means.

Images