The clinical research project was undertaken with sponsorship from Cardiff University, support from the University of Wales teaching hospital haematology and Public Health Wales Microbiology departments, and ethical approval from Health and Care Research Wales (NISCHRC CRC 1351-14). REC reference 14/WA/1119 and IRAS project ID 151136. Written informed consent was obtained from all patients in the study. Whole blood samples were collected from 42 acute-myeloid leukaemia and stem cell transplant patients upon admission to hospital for their primary disease treatment. Fungal disease investigations were undertaken according to local health board guidelines. Of the 42 patients enrolled in the study, 9 developed probable IA according to the European Organization for Research and Treatment of Cancer/Invasive Fungal Infections Cooperative Group and the National Institute of Allergy and Infectious Diseases Mycoses Study Group (EORTC/MSG) consensus definitions (32). Patient’s primary disease, anti-fungal treatment and survival are included in Table 1 . This study tracked 42 patient’s fungal disease status and survival from 7^th September 2015 until 9^th April 2018. Each patient was anonymised and assigned a number which was consistent throughout the study. The results in this study show patients 1 to 43, as the samples from patient 18 were not processed. Not all assays were undertaken for each patient and statistical analysis was only completed on patient results that met all investigated parameters. Therefore, patient numbers throughout results and analysis vary.

17.5ml Whole blood was collected from each patient into EDTA tubes (BD) and 2ml whole blood from was collected from each patient into PAXgene blood RNA tubes (Preanalytix) following initial admission to hospital. The 17.5ml whole blood sample was immediately processed for use in the functional assays. The 2ml whole blood in PAXgene blood RNA tubes was stored according to manufacturer’s instructions for up to 6 months before being processed in batches.

Patient RNA samples were processed in batches of 8 and were not stored for longer than 6 months at -80°C. PAXgene blood RNA tubes were equilibrated at room temperature for 2 h following storage at -80°C before RNA was extracted from the whole blood within the PAXgene blood RNA tubes according to the manufacturer’s instructions. The optional DNA digestion step was completed for every patient sample. Typically, ≥3μg RNA was obtained from each patient sample and quantified by nanodrop (Thermo Scientific). The integrity of extracted RNA was confirmed by running patient RNA samples on a 1% agarose gel and visualising the 18S and 28S rRNA bands.

Patient RNA was used to generate cDNA using a Quantitect Reverse Transcription Kit (Qiagen). Typically, 2000ng of cDNA was generated for each patient at a final concentration of 100ng/µl for immediate use in Real-Time qPCR. Successful cDNA generation was confirmed by running patient RNA and cDNA samples on a 1% agarose gel. The absence of the 18S and 28S rRNA bands in the cDNA sample indicated successful cDNA generation.

Gene expression levels of the CLRs Dectin-1 (CLEC7A), Dectin-2 (CLEC6A), Mcl (CLEC4D) and Mincle (CLEC4E) was determined by Real-Time qPCR using the Taqman qPCR Mastermix (Thermo Scientific) and CLR gene-specific primer and probe sets (Thermo Scientific) detailed in Supplementary Table 1 . 100ng of patient cDNA was used in each CLR qPCR reaction. Gene expression normalization was performed against HPRT1. Patient CLR exon sequences were determined by PCR amplification of each CLR gene from patient cDNA using the primers detailed in Supplementary Table 1 . CLR DNA was purified from PCR reaction mixtures using a PCR purification kit (Qiagen) and sent for sequencing at GATC Biotech.

17.5ml blood collected into EDTA tubes was added to 50ml conical tubes on top of an equal volume of Ficoll Plus (Sigma) and centrifuged at 400 x g for 30 min with the centrifuge brake reduced to its lowest setting. After centrifugation the layer of PBMCs was removed, washed once with PBS (Life Technologies) and then three times with RPMI 1640 (Life Technologies) before being counted and resuspended in RPMI 1640 supplemented with 10% FBS (Life Technologies), 2% Human Serum (Sigma), 10mM L-glutamine (Life Technologies), 10mM Sodium Pyruvate (Life Technologies) and 100µg/ml Gentamycin (Life Technologies). 100µl of 5x10⁶/ml PBMCs were added to each required well on a 96-well plate (Thermofisher) and rested at 37°C for 4 h prior to the functional assay.

Aspergillus fumigatus (isolate 13073) was cultured on potato dextrose agar (Sigma) for 7 days before resting conidia (RC) were harvested by vigorous washing with PBS 0.05% Tween 20 (Sigma). Harvested RC were counted and stored at 4°C in PBS 0.05% Tween 20 for a maximum of 6 weeks before a new culture was started. For functional assays, RC were grown in RMPI 1640 supplemented with 0.2mg/ml Polymixin B (Sigma) at 37°C 5% CO₂ for 6 hours to generate swollen conidia (SC). Aspergillus fumigatus SC were counted and resuspended at 5x10⁶ SC/ml for use in functional assays.

100µl of 5x10⁶/ml PBMCs were stimulated with 100µl of 1µg/ml LPS (Sigma) or 100µl of 5x10⁶ Aspergillus fumigatus swollen conidia/ml for 24 h at 37°C 5% CO₂. An unstimulated media-only control was included. After 24 h, supernatant was removed from each well and TNF and IL-6 levels in the supernatants were quantified by ELISA (eBioscience).

Significance was determined using contingency multivariate statistical analysis with Fisher’s exact test. If two variables were analysed *p=0.05, **p=0.005. Where more than two variables were analysed Bonferroni’s correction was applied and adjusted p values are described in the Figure legend.

Previous studies have associated SNPs in Dectin-1, Dectin-2 and CARD9 with IA (33–36); therefore, we first aimed to determine the sequence and expression levels of four fungal binding CLRs in our cohort of SCT and AML patients. We screened 42 patients for mutations in the exon coding regions of four CLRs (Dectin-1, Dectin-2, Mincle and Mcl) and identified a Dectin-2 mutation (N170I) (rs1334241354) (34) present in 1 patient and an Mcl mutation (S32G) (rs4304840) present in 17 patients. The Dectin-2 mutation resulted in an early stop codon and loss of the carbohydrate binding region whereas the Mcl mutation only resulted in a single amino acid substitution. We next quantified gene expression of CLEC7A (Dectin-1), CLEC6A (Dectin-2), CLEC4D (Mcl) and CLEC4E (Mincle) by Real-Time qPCR for each patient ( Figure 1 ). Expression levels of the four CLRs varied considerably from patient to patient and no clear association between CLR expression level and the incidence of IA was observed.

In order to investigate whether CLR expression was associated with the incidence of IA, we grouped each patient’s CLR results into high (above the median) or low (below the median) and analysed whether high or low CLR expression was associated with IA incidence ( Figure 1 ). Here we found that AML patients who developed IA were 10.2 times more likely to have low Dectin-1 expression. In contrast, SCT patients who developed IA were 7.1 times less likely to have low Mcl expression, and 10 times less likely to have low Dectin-1 expression. However, the overall results associating CLR expression with IA incidence across total, AML and SCT patient groups did not provide significant associations.

We next determined whether CLR mutations were associated with IA incidence ( Table 2 ). We did not find any association between the incidence of the Mcl S32G mutation and the incidence of IA. Patients with the mutation were equally distributed between the IA group and the no evidence of fungal infection (NEF) group. In addition, as low Mcl expression in SCT patients reduced the likelihood of developing IA within the study, we also examined whether the Mcl S32G affected Mcl gene expression and found no association. The single patient with the Dectin-2 N170I mutation did develop IA (34); however, as this mutation was only present in one patient statistical correlation analysis was not appropriate. Overall, we did not observe any significant association between the CLR mutations and expression identified in this study and the incidence of IA.

As the cytokines IL-6 and TNF are vital for a protective anti-Aspergillus immune response (37, 38), we next decided to investigate each patient’s functional anti-Aspergillus immune response. To this end, we isolated patient PBMCs and challenged them with LPS or Aspergillus fumigatus SC for 24 h before quantifying the IL-6 and TNF cytokine response ( Figures 2A–C ). Here, our results suggest the majority of patients were able to generate LPS-induced TNF and IL-6; however, fewer patients were able to generate Aspergillus-induced TNF and IL-6. Interestingly, whilst all 9 of the IA positive patients produced LPS-induced cytokines, only three IA positive patients (all in the AML cohort) generated Aspergillus-induced TNF or IL-6, and only one of these patients produced both cytokines ( Figures 2A, C ). Additionally, only 11 out of 17 AML patients were able to produce any TNF or IL-6 response ( Figure 2C ). This was likely due to AML patient’s highly disrupted haematopoietic compartment and lack of mature myeloid cells. However, for SCT patients our data suggests that IA patients may lack a specific anti-Aspergillus response. All 5 IA positive SCT patients produced LPS-induced TNF and 4 of the 5 IA positive SCT patients produced LPS-induced IL-6, whilst none of these IA positive patients produced either cytokine following Aspergillus challenge ( Figure 2B ). Similarly, the majority (all but one) of the NEF patients produced LPS-induced TNF and/or IL-6, but in contrast to the IA patients, 13 out of 18 NEF patients generated Aspergillus-induced TNF and/or IL-6 ( Figure 2B ). The samples used to generate these results were isolated from patients upon admission to hospital for their primary disease treatment. At the time the assays were performed all IA patients in the study had monocyte counts within the normal range. Therefore, this assay may stratify SCT patients according to their risk of developing IA prior to the patients becoming highly susceptible to invasive fungal disease.

In order to determine whether a patient’s functional immune response could be used as a risk factor for IA, we associated each patient’s functional response results with the incidence of IA ( Table 3 ). Each patient’s functional response results were included as the single parameter no Aspergillus-induced TNF or IL-6 response. Here, we identified a significant association between patient’s functional immune response and the incidence of IA. The strongest association was identified in SCT patients who lacked an Aspergillus-induced IL-6 response. This factor defined 5 out of 5 SCT patients who developed IA and only 5 out of 18 SCT NEF patients. Patients who fulfilled these criteria were 27 times more likely to develop IA when compared to SCT patients who generated Aspergillus-induced IL-6. SCT patients who did not produce an Aspergillus-induced TNF response were also significantly more likely to develop IA, with over 21 times higher IA risk within our study. Here, we identify novel IA functional risk factors that could be used to stratify patients according to their IA susceptibility.

Whilst we have described the significant association between Aspergillus-induced TNF and IL-6 and the incidence of IA, it was important to determine whether these results were specific for predicting IA or a more general indicator of poor patient prognosis. Therefore, we associated the incidence of IA and patient’s functional immune response results with mortality ( Supplementary Table 2 ). Here, we found no association between the incidence of IA or patient’s functional immune response results with the incidence of mortality. These results suggest the TNF and IL-6 functional immune response results are specific risk factors for the incidence of IA within our cohort and not simply identifying patients with a high incidence of mortality.

As we found a significant association for patient’s functional immune response and IA, we next analysed multiple risk factors in combination. Patient’s CLR status and functional immune responses were combined and associated with the incidence of IA ( Table 4 ). Here, our data shows SCT patients that produced LPS-induced TNF and/or IL-6, lacked Aspergillus-induced TNF and/or IL-6 and had high Mcl expression possessed a high risk of IA. Of the 6 patients that fulfilled these parameters, 4 (66%) went on to develop IA. Our data suggests the highest IA risk is present in patients that produced LPS-induced TNF and/or IL-6, lacked Aspergillus-induced TNF and/or IL-6 and had high Dectin-1 expression or both high Dectin-1 and high Mcl expression. These parameters stratified 5 patients from the total SCT cohort and 4 developed IA. Interestingly, patient 39 was positive for Aspergillus mycology but their radiological investigation did not show evidence of IA; therefore, the patient was not classified as probable IA in accordance with EORTC/MSG guidelines (32). In our study patient 39 had high Mcl and Dectin-1 expression, produced LPS-induced IL-6 and/or TNF and lacked Aspergillus-induced IL-6 and/or TNF, suggesting this patient had a high risk of IA. Patient 39 was the only patient stratified into the highest risk cohort that did not have proven IA.

In Table 4 we also display the incidence of IA within the SCT cohort according to the risk factor parameters used. Here, 5 of 11 (45%) SCT patients with no Aspergillus-induced TNF response developed IA, 5 of 10 (50%) SCT patients with no Aspergillus-induced IL-6 response developed IA, and 5 of 8 (62.5%) SCT patients with an LPS-induced TNF and/or IL-6 response but lacking an Aspergillus-induced TNF and/or IL-6 response developed IA. We next investigated whether the CLR status and functional immune response risk factors could be used to predict the incidence of IA in AML patients within the study. Here, we found no association between AML patients’ functional responses and the incidence of IA. Using low Dectin-1 expression as a risk factor stratified 9 patients from the total AML cohort of which 3 (33%) developed IA but this result was not significant.

Finally, we used the risk factors and associated IA incidence described for SCT patients to propose a strategy that would enable the pre-emptive stratification of SCT patients according to their IA susceptibility ( Figure 3 ). We show in the intermediate- and high-risk groups how the risk factors identified in Table 4 could be used to predict the incidence of IA within the study. We also demonstrate how the risk factors described in this study could be used as exclusion criteria to stratify patients at low risk of IA. Importantly, none of the SCT patients that produced Aspergillus-induced TNF and/or IL-6 developed IA.

In this study our investigations into patient’s CLR status and functional immune response have led to the identification of novel IA risk factors. We have demonstrated how these risk factors may be applied to stratify patients into low, intermediate, and high-risk cohorts. Crucially, the risk factors identified here would enable the pre-emptive stratification of patients and permit a personalised medicine approach to patient’s anti-fungal investigations and treatment.

In this study we investigated patient’s CLR status as deficiencies in anti-fungal immune components have previously been associated with IA. The first inherited or acquired genetic condition associated with IA susceptibility was identified in CARD9, a downstream adaptor molecule that transduces CLR signalling. CARD9-deficiency resulted in fungal infections without any immune suppression (35, 36, 39). The Dectin-1 Y238X mutation increases IA susceptibility in haematology patients through reduced fungal recognition and immune responses (33). Similarly, the newly identified Dectin-2 N170I mutant was shown to reduce fungal recognition and immune responses (34). Crucially, genetic deficiencies are present prior to the initiation of patient’s primary treatment and immune suppression, thereby offering an early indication of a patient’s susceptibility to IA. Whilst these risk factors have been associated with IA, they are not yet widely applied in clinical practice (22, 33, 35, 40). Using pre-emptive risk factors to stratify patients has significant promise and has recently been tested. Mutations in Dectin-1 and DC-SIGN, respiratory viral infection, allogeneic stem cell transplant and Aspergillus PCR positivity were used to stratify patients in a predictive disease model. Patients with no risk factors had a 2.4% probability of developing IA whilst patients with four or more risk factors had a 79% probability of developing IA (22). A future study combining the risk factors described by White P. L. et al. and those described in this study may further enhance patient stratification.

In this study, we identified two CLR mutations in our patient cohort and investigated each patient’s CLR expression before associating these factors with the incidence of IA. The S32G mutation identified in Mcl has been previously described (rs4304840). This missense mutation results in the substitution of serine to glycine at position 32; this is not thought to have functional consequence. In agreement with this, we identified no association between the incidence of Mcl S32G and IA. Our study also identified a novel N170I mutation in Dectin-2. This mutation resulted in an early stop codon located in the carbohydrate binding domain of the CLR. We recently characterized this mutation and showed that it results in reduced receptor expression and deficient anti-fungal immune responses (34). This mutation was identified in a patient who developed IA and died; however, as this is only one patient the statistical association of this mutation and IA cannot be undertaken.

We determined that high (above the median in this study) Mcl and Dectin-1 expression in SCT patients and low (below the median in this study) Dectin-1 expression in AML patients may be associated with an increased risk of IA. Dectin-1 has been extensively shown to drive protective immune responses against fungal pathogens including Aspergillus (28, 41, 42). Therefore, as was found with the Dectin-1 Y238X mutant, it is unsurprising the low/deficient Dectin-1 expression in AML patients may increase IA susceptibility. However, it is not clear why high Dectin-1 expression in SCT patients may be associated with an increased risk of IA. There is limited research describing the functional role of Mcl. Mcl-deficient mice produced defective immune responses against TDM (mycobacterial trehalose dimycolate) (43). A collaborative role for Mcl and the Mincle heterodimerising and enhancing the recognition of carbohydrate and lipids molecules has been identified; however, the functional consequences of this heterodimer CLR complex remain largely unknown (44–46). An explanation for the high Mcl expression identified in SCT patients who developed IA is not clear and requires further investigation.

Previous studies aiming to describe risk factors associated with IA have not investigated patient’s functional response against inflammatory stimuli or Aspergillus. Patient PBMCs have been used to identify those most receptive to immunotherapy with positive results correlating assay outcome and patient outcome. However, these investigations were completed retrospectively after all patients had received treatment (47, 48). Here, we investigated whether patient’s functional immune response could be used to predict IA incidence. Our research describes how the results from a simple assay can be used as a novel IA risk factor that could drive a personalised medicine approach.

In this study we describe how LPS- and Aspergillus-induced TNF and IL-6 response can be used to identify the SCT patients most susceptible to IA. PBMCs produce pro-inflammatory cytokines and drive protective immune responses when challenged with Aspergillus-extracted chitin or live Aspergillus (49, 50). The cytokines IL-6 and TNF play a key role in anti-Aspergillus immunity with mice deficient in either cytokine possessing an increased susceptibility to Aspergillosis (37, 38). In agreement, the patients in this study who were unable to produce TNF and/or IL-6 against Aspergillus possessed enhanced susceptibility to IA. Similar results were attained with Candida albicans where fungal disease was associated with a delayed secretion of cytokines from myeloid cells and T cells (51).

Our functional assay results determined that patients able to produce an LPS-induced response but lacking an Aspergillus-induced response possessed the highest IA susceptibility. Anti-Aspergillus immune responses are complicated and require collaboration between numerous receptors, signalling molecules and cell types to produce a protective immune response. Patients with haematological malignancies often possess immune defects and have highly variable immune cell counts (52). Here, our functional assay may be able to discriminate between patients with high disruption to their anti-fungal immune response and those with minimal disruption. It is likely those with high disruption possessed deficiencies in anti-fungal immune components that we did not screen for. In contrast to the Aspergillus-induced response, the majority of patients were able to produce LPS-induced TNF and/or IL-6. The LPS-induced response requires only TLR4 and CD14 signalling to produce a robust pro-inflammatory response (53) and PBMCs produce large quantities of TNF and IL-6 within 4 h of LPS stimulation (54). Crucially, TLR signalling is retained in patients with a haematological malignancy even after SCT, radiotherapy or chemotherapy and is often responsible for graft versus host disease, gut toxicity and chronic pain (55–57). This maintenance of TLR signalling likely explains why most patients in this study produced LPS-induced responses.