The coating Aspergillus GM antibody was obtained from KeraFAST (Shanghai, China). The labeled Aspergillus GM antibody was acquired from ThermoFisher Scientific (Shanghai, China). GM antigen (purity, >95%) stock was provided by our laboratory. The magnetic beads (1 μm diameter) and the desalting column were from Thermo Fisher Scientific (Shanghai, China). N-(3-(dimethylamino)propyl)-N’-ethylcarbodiimide hydrochloride (EDC), N-hydroxysulfosuccinimide (NHS) and, 4-morpholineethanesulfonic acid (MES), bovine serum albumin (BSA) were purchased from Sigma-Aldrich (Shanghai, China). Alkaline phosphatase (AP) for labeling was purchased from Roche Diagnostics (Shanghai, China). APS-5 was bought by Lumigen (Michigan, USA). Other chemicals unless stated otherwise, were sourced from Solarbio (Beijing, China). All reagents were of analytical or higher grade. Ultrapure water (18.2 MΩ·cm) was made by a Millipore Milli-Q system. Ultraviolet absorption spectra were acquired using a NanoDrop 2000c system (ThermoFisher Scientific). The automated chemiluminescence analyzer (DF200) was provided by Danda Biotech ((Beijing, China).

Between February 2024 and June 2025, we enrolled a total of 241 participants at Tianjin Chest Hospital. The cohort comprised 86 patients with probable IPA classified according to the 2020 EORTC/MSG Criteria (mycologic evidence based on positive Platelia ELISA results, Aspergillus PCR detection or positive culture result), 90 patients without evidence of IPA, and 88 healthy volunteers. Participants had a median age of 45 years (range 21–86 years), with 108 (45%) being female. Medical records were available and reviewed for the 176 patients in the cohort (excluding healthy volunteers). Within this patient subgroup, clinical characteristics included underlying hematological malignancies in 58 patients (33%), solid organ transplantation in 35 patients (20%), ICU care requirement for various underlying diseases in 39 patients (22%), and a history of long-term corticosteroid or immunosuppressant use in most of the remaining patients. Patients meeting any of the following exclusion criteria were not included: 1) Aspergillus colonization or contamination; 2) Unavailability of required diagnostic markers. Based on recent CT scans, BALF was collected from targeted bronchial segments by instilling ~50 ml of room temperature saline, repeated 2–3 times. The fluid was collected sterilely and sent for prompt analysis (within 30 minutes). Both BALF and serum samples were collected within 24 hours of patient admission and prior to antifungal treatment. Of these samples, sera (n=131) and BALF (n=110) with sufficient volume remaining after routine clinical analyses were aliquoted and stored at -80 °C for future use. Repeated freeze-thaw cycles were strictly avoided. The study protocol and all study-related procedures was approved by the Ethics Committee of Tianjin Chest Hospital (Grant No: 2023YS-019-1) and a waiver of informed consent was granted.

GM antibody-functionalized MBs were prepared according to supplier provided protocols along with a laboratory-developed modification method. Briefly, 2 mg of Mbs were washed twice with deionized water using a magnetic separator. The carboxyl group on the surface of MBs were next activated in 1.0 mL of MES buffer solution (50 mM) containing 6 mg EDC and 4 mg NHS for 30min under a shaker. The activated MBs-COOH were purified via three MES buffer washes and resuspended in 0.5 mL of MES buffer. Following that, 100 μL of Ab (1 mg/mL) was added to the above suspension and allowed to react for 4h with gentle agitation, after which blocking was performed using 100 μL of BSA (10 mg/mL) for an additional 2h. Finally, the obtained MBs-Ab were washed, redispersed with 1 mL of PBS buffer containing 0.2 mg/mL BSA, and stored at 4°C for next use. Mbs-Ab formation was tracked using the Bicinchoninic acid (BCA) protein quantification method and the lack of residual antibody in the coupling buffer supernatant validated bioconjugation.

The GM antibodies and AP were coupled using glutaraldehyde. Firstly, 2 mg of AP was dissolved in 0.5 mL of antibody solution (2 mg/mL), followed by dilution to a final volume of 2 mL with ultrapure water and thorough mixing. Then, 100 μL of 5% w/v glutaraldehyde solution was slowly added to the mixture under continuous stirring for 2h in dark. Lastly, the labeled antibody mixture was dialyzed in phosphate-buffered saline (PBS, pH 7.2) (0.2 M Na₂HPO₄ and 0.2 M NaH₂PO₄) overnight and purified using desalting column. The resulting AP-conjugated GM antibody (AP-Ab) was mixed with an equivalent volume of glycerol, and stored at 4°C. UV spectral analysis of AP-Ab showed a shift toward shorter wavelengths compared to AP and anti-GM, verifying AP-antibody conjugation and its potential for diagnostic use.

As shown in Figure 1 , GM antigen was detected by sandwich immunoassay as follows. Prior to testing, all samples, including the calibrator control (2.5 ng/mL Aspergillus GM in BSA solution), serum, and BALF specimens underwent a 3:1 (v/v) dilution in 4% EDTA-PBS, followed by thermal denaturation (120°C, 5min) to eliminate protein interference. The treated samples were then centrifuged (10,000 × g, 5min) to remove precipitates. Equal volumes of anti-GM-AP (50 μL) and anti-GM-Mbs (50 μL) were added to 100 μL the above supernatant in reaction tubs. The obtained mixture was vortexed, incubated at 37°C for 20 mins, then subjected to three wash cycles via magnetic separation. Finally, the luminescent substrate solution (APS) was added to the sandwich Mbs-GM-AP immunocomplex, and the resulting chemiluminescent reaction was measured by the commercial chemiluminescence analyzer. The calculation of CLEIA Units was performed by dividing the chemiluminescent intensity of each sample by that of the calibrator control.

Duplicate measurements were performed for all test samples. The experimental results were statistically analyzed with the GraphPad Prism 9 software. Receiver operating characteristic (ROC) analysis was performed to establish an optimal positivity threshold and assess the overall diagnostic performance of CLEIA. Qualitative agreement between CLEIA and ELISA was assessed by calculating observed agreement and Cohen’s kappa statistic. Quantitative agreement was evaluated using Spearman’s rank correlation between the GM indices derived from both assays. Median values were compared using the Mann-Whitney U test. Repeatability was evaluated by the coefficient of variation (CV). A two-sided p value of less than 0.05 was considered to be statistically significant.

The detection performance of the assay was found to depend critically on four parameters: the dilution ratio of Mbs-Ab, the concentration of AP-Ab, the incubation time, and the volume of APS solution. To optimize the testing conditions, the signal-to-noise (S/N) ratios derived from the two calibrators with the GM concentration of 1.0 ng/mL and 0 ng/mL were analyzed through the developed assay.

Initially, a checkerboard titration experiment was performed involving Mbs-Ab and AP-Ab. If the quantity of antibody conjugates is too low, it will be insufficient to capture all the target analytes present in the sample. Conversely, an excessive amount of these conjugates can interfere with luminescence or cause high background noise due to nonspecific interactions between the signal probe and the capture probe. Both scenarios would negatively impact the signal-to-noise ratio. As indicated by Figure 2A , the working parameters established were a 1:200 dilution for the Mbs-Ab and a concentration of 0.15 μg/mL for the AP-Ab.

Second, the incubation time for the immunoreaction between the antigen and antibody was also investigated. Analysis of incubation time ( Figure 2B ) revealed an initial S/N ratio increase over the first 20min, stabilizing (or marginally declining) between 20 and 30min. The maximum S/N was attained at 20min, confirming immunoreaction equilibrium. Considering both the absence of further S/N gains and the potential for increased nonspecific adsorption during prolonged incubation, 20min was chosen as the optimal reaction time.

Beyond the aforementioned reaction conditions, substrate volume critically influences luminescent signal amplitude and analytical sensitivity. Based on the data in Figure 2C showing a peak S/N ratio at 150 µL, this volume of APS was standardized for all subsequent experimental procedures.

To determine cut-off values, 241 retrospective specimens (131 serum, 110 BALF) from 241individuals were evaluated under the above optimal conditions. Analysis of GM antigen levels (CLEIA Units) revealed a range of 0.01-26.74 ( Figure 3A ), with statistically distinct clustering between positive and negative groups (p<0.0001). ROC curve evaluation ( Figure 3B ) determined an optimal cutoff value of 0.20 CLEIA Units, yielding diagnostic sensitivities of 90.70% (95% CI, 82.70 to 95.21) and specificities of 85.16% (95% CI, 78.72 to 89.90).

To assess CLEIA analytical sensitivity, we prepared serum and BALF samples spiked with Aspergillus GM antigen (4.00 ng/mL) followed by eight serial 1:2 dilutions. Testing included 20 replicates per concentration level. The analytical sensitivity—defined as the concentration achieving ≥95% positive results—was calculated to be 0.50 ng/mL for both matrices.

In sandwich immunoassays, the hook effect represents a critical analytical artifact wherein excessive analyte concentrations saturate both solid-phase and labeled antibodies. This antibody saturation prevents complete formation of ternary immune complexes, generating false-negative results.

As presented in Table 1 , validation testing for hook effect revealed no false-negative CLEIA results in Aspergillus GM antigen-spiked human serum and BALF samples at concentrations up to 200 µg/mL.

Precision on the CLEIA was evaluated using a specimen panel comprising 4 pooled patient serum samples (negative, low positive, positive, high positive) and 4 pooled patient BALF samples spiked with purified GM (negative, high negative, low positive, positive). Each sample underwent 30 replicate measurements: two operators performed three independent runs per day over five consecutive days. The following metrics were calculated from CLEIA Units results to evaluate precision: mean, standard deviation (SD), percent coefficient of variation (%CV), percentage of positive identifications (% Positive), and percentage of negative identifications (% Negative). Comprehensive results are documented in Table 2 , which indicated that the reproducibility and precision of this method are acceptable.

The influence of endogenous serum components known to cause matrix interference (hemoglobin, bilirubin, triglyceride) was investigated. Spiking experiments showed no measurable interference effect in the assay system. Concurrently, the potential for cross-reactivity of the CLEIA was also examined using a diverse set of patient serum samples collected from individuals with clinically relevant conditions. Results of cross-reactivity testing are compiled in Table 3 . A Histoplasma serum specimen was found positive by the present chemiluminescent immunoassay. Notably, the single Histoplasma cross-reactive serum specimen was also positive on another commercially available Platelia GM-ELISA, with a GM Index of 4.20, compared to 0.36 CLEIA Units on the CLEIA.

A cohort of 241 specimens (86 cases, 155 controls) underwent parallel testing with both the FDA-approved Platelia GM ELISA test and the developed chemiluminescent immunoassay. Qualitative assessment between platforms showed high diagnostic concordance across sample cohorts: percent positive agreement = 71.68% (95% CI, 62.30–79.56%), percent negative agreement = 80.00% (95% CI, 72.79–85.73%), and overall agreement = 86.72% (95% CI, 81.63–90.61%). This corresponded to a Cohen’s kappa statistic of 0.72 (95% CI, 0.64–0.81), indicative of substantial agreement according to the Landis and Koch scale ( Table 4 ) (Landis and Koch, 1977). The overall quantitative correlation between the GM index values calculated by the candidate CLEIA and the Platelia ELISA was moderate (Spearman’s ρ = 0.48, P <0.0001; Figure 4A ). Significantly stronger agreement was observed in case-derived specimens (ρ = 0.75, P < 0.0001; Figure 4B ), while control samples showed no statistically significant correlation (ρ = -0.05, P=0.50; Figure 4C ).