In recent years, the prevalence of pulmonary fungal disease has exhibited a significant upward trend due to an increase in high-risk groups requiring immunosuppressive therapy. Invasive pulmonary aspergillosis (IPA) is still an important cause of morbidity and mortality (Herbrecht et al., 2002; Gangneux et al., 2010; Schauwvlieghe et al., 2018). However, the non-specific clinical features of IPA, especially in non-neutropaenic patients, often lead to delayed initiation of treatment and higher mortality (Cornillet et al., 2006; Gangneux et al., 2010). Studies have shown that mortality rates can reach up to 90% if the treatment is not administered until 10 days after the first clinical or radiological sign of the disease, but fall to 40% when treatment is administered early (von Eiff et al., 1995). Therefore, timely and accurate diagnosis is of great significance.

Unfortunately, the diagnosis of Aspergillus infections poses a significant challenge. While histopathology is often recommended as a ‘gold standard’, obtaining a sufficient quantity of biopsy samples from the infected site can be difficult, resulting in a high false-negative rate (Bialek et al., 2002). Fungal culture is currently the most commonly used diagnostic tool, but is time-consuming and offer a low yield (Perfect et al., 2001; Arvanitis and Mylonakis, 2015). Although smear by microscopy can identify Aspergillus hyphae more quickly than culture, it requires a heavy pathogen burden in the lung and experienced microbiologists to ensure the detection, making it insensitive and biased. In the past decade, the galactomannan (GM) test has been widely used in the diagnosis of IPA, but its sensitivity in serum samples from non-neutropenic hosts is limited due to low fungal load in the lesions and a low chance of antigen appearing in the bloodstream (Stergiopoulou et al., 2007; Cordonnier et al., 2009). BALF GM test is superior to serum GM in the diagnosis of IPA, but its results are affected by standardization of the sampling, diverse mycological procedures, antimicrobial treatment before sampling and various other factors, making its role in non-neutropenic patients uncertain (Maertens et al., 2009; Racil et al., 2011; Brownback et al., 2013; Hage et al., 2019). While polymerase chain reaction (PCR) based methods are more rapid, sensitive, and specific, they require specific primers and are not routinely performed in many hospitals. Therefore, there is an urgent need for new technology with a higher sensitivity to improve Aspergillus infection diagnosis.

Metagenomic nextgeneration sequencing (mNGS) is considered a promising microbial identification technology, enabling direct pathogen detection without primers or probes. Although mNGS has been widely used in infectious diseases, there is still a scarcity of clinical experience in pulmonary fungal infections. Previous studies have reported mixed results, with some showing that mNGS outperformed conventional microbial tests (CMT) for fungal detection (Li et al., 2018; Miao et al., 2018; Yang et al., 2021), while others found the opposite. (Xie et al., 2019; Fang et al., 2020). It was worth noting that these studies without exception encompassed various fungi, including Aspergillus, Cryptococcus, P. jirovecii and Candida, but different techniques may have their advantages in identifying specific fungal species (Miao et al., 2018; Xie et al., 2019; Fang et al., 2020). Identifying filamentous molds such as Aspergillus spp. through mNGS remains challenging due to the difficulty of DNA extraction from the thick polysaccharide cell wall, as well as the relatively low fungal load in BALF (Bittinger et al., 2014; Clarke et al., 2018; Han et al., 2019). To date, limited evidence is available, mainly from case reports and small case series (He et al., 2019; Bao et al., 2022; Yang et al., 2022). Our previous study demonstrated the excellent performance of mNGS in identifying pneumonia pathogens in immunocompromised patients (ICP), but subgroup analysis did not reveal its advantage in Aspergillus detection due to the small sample size and immature technology (Shi et al., 2022). As an extension of our earlier study, we conducted research in patients with IPA to contribute to this growing field of research by comparing the diagnostic performance between mNGS and CMT, including smear, cultures, serum and BALF GM test.

The diagnosis of IPA is challenging due to the lack of characteristic clinical manifestations, making mycological evidence crucial for diagnosis. The slow growth of Aspergillus or its isolation under the lens or capsular bag can complicate pathogen detection (Bittinger et al., 2014; Clarke et al., 2018). Additionally, empirical antifungal therapy is increasingly common in high-risk groups of ICP, resulting in a low pathogen detection rate of CMTs (Perfect et al., 2001; Racil et al., 2011; Li et al., 2021). Therefore, there is a pressing need to explore new sensitive technology to improve Aspergillus identification.

mNGS has emerged as a promising tool for the diagnosis of infectious diseases, however, its clinical experience in diagnosing IPA is largely based on case reports and small case series. For instance, Yang et al. (Yang et al., 2022) and He et al. (He et al., 2019) reported cases of IPA in which mNGS quickly detected the pathogen, while other tests failed to identify it. Similarly, the largest study to date by Miao et al. (Miao et al., 2018) suggested that mNGS exhibited better performance than culture with regard to fungal detection, but the difference was not significant in Aspergillus detection (OR 3.7, 95% CI 0.97–20.5; P = 0.057) due to the small sample size of IPA (26 cases). Therefore, further validation by larger studies is necessary to confirm these preliminary findings.

In this study of 133 ICP, we foud that IPA was associated with severe clinical presentations and poor outcomes in critically ill patients. mNGS exhibited excellent performance in detecting Aspergillus. Its sensitivity outperformed any single CMT method, and can be comparable to the combination of multiple CMTs. Furthermore, mNGS performed well in identifying fungal species and co-pathogens, indicating that mNGS may have a valuable role in guiding antimicrobial therapy.

Upon reviewing relevant literature, it is noted that the vast majority of published studies only compared the performance of mNGS with a single CMT method for aspergillosis identification. For example, a series of studies found that the sensitivity of mNGS was better than that of culture (Chen et al., 2020; Bao et al., 2022); Additionally, the BALF GM test (range from 54.5% to 80%) outperformed serum GM (range from 24.3% to 46.9%) in the diagnosis sensitivity, which was widely reported and supported our results (Zhou et al., 2017; Dai et al., 2021; Ghazanfari et al., 2022). However, there have been few studies comparing the differences between BALF GM and mNGS. Ao et al. reported that BALF GM (57.7%) had the highest sensitivity in detecting Aspergillus, followed by mNGS (42.3%), culture (30.8%) and smear (7.7%) in 26 IPA patients (Ao et al., 2023). In contrast, our study observed improved sensitivity of mNGS compared to BALF GM. This divergence may be attributed to differences in the study population, sampling methods, empirical antifungal therapy, and even cutoff value (Stergiopoulou et al., 2007; Racil et al., 2011; Dai et al., 2021; Li et al., 2021). The aforementioned factors may reduce the sensitivity or increase false-positive of BALF GM (Cornillet et al., 2006; Pfeiffer et al., 2006; Stergiopoulou et al., 2007). Furthermore, the optimization of nucleic extraction methods from respiratory specimens may improve the ability of mNGS for Aspergillus identification (Zinter et al., 2019; Chen et al., 2020), which may explain the higher sensitivity of mNGS in this study. Notably, our study found a positive correlation between Aspergillus unique reads and fungal loads, and false-negative results of BALF GM or culture typically corresponded to smaller reads by mNGS, indicating that despite the difficulty of nucleic acid extraction, mNGS can still be a useful detection tool for low fungal load through technological improvements (Tsang et al., 2021; Jiang et al., 2022).

As is well known, accurate identification of strains is of great significance in guiding antifungal treatment. With the genomes of more than 1000 fungal species publicly available, mNGS offers a more accurate diagnostic analysis of pathogenic strains, which is even more specific than other methods (Chen et al., 2020; Etxebeste and Espeso, 2020; Yang et al., 2021). Our study found that only 39% of patients exhibited positive Aspergillus cultures, whereas mNGS not only identified strains that were highly consistent with those detected by culture, but also detected Aspergillus and even Mucor genera that were not detected by culture. Case reports highlighted the favorable prognosis resulting from antifungal drug changes based on mNGS results (Zhang et al., 2020; Yang et al., 2022). With an increase in high-risk populations receiving prophylactic antifungal therapy, there may be an increase in azole-resistant strains. However, mNGS has the potential to revolutionize the management of fungal infections by providing rapid and accurate identification of pathogenic strains, and even information on virulence and resistance to guide appropriate antifungal therapy (Diao et al., 2022; Jiang et al., 2022).

Our study also revealed that the vast majority of patients (80%) had mixed infections of Aspergillus with other pathogens, particularly “uncultivable” microorganisms such as CMV and P. jirovecii. Establishing an early diagnosis of co-infections remains a great challenge. CMT methods rely on pre-evaluating possible pathogens to carry out targeted detection. However, this approach is highly susceptible to the subjective experience of clinical personnel, and has a narrow pathogen spectrum, leading to a high rate of missed diagnosis. In contrast, mNGS offers an unbiased sampling approach, allowing for the simultaneous identification of all potentially infectious agents in a single run, without defining targets for diagnosis beforehand (Chen et al., 2022; Diao et al., 2022; Jiang et al., 2022). This may explain the satisfactory performance of mNGS in identifying co-pathogens in this study, which has also been confirmed by numerous other studies (Pan et al., 2019; Wang et al., 2019; Jiang et al., 2021). We believed that the mNGS strategy can not only simplify clinical testing for diagnosing mixed infections, but also bring the greatest benefits to ICP who often suffer from various complicated pathogen infection. Furthermore, due to the common co-pathogens being P.jirovecii and CMV, mNGS is more likely to reveal the etiology of infections when multiple PCR tests are rarely performed on respiratory samples.

Although our study has preliminarily demonstrated the feasibility of diagnosing IPA using mNGS, we must acknowledge the limitations of this technology. Firstly, the difficulty of nucleic acid extraction varied by species, and the number of unique reads was substantially reduced when contrasting fungi with bacteria pathogens (Fang et al., 2020; Jiang et al., 2022). Additionally, samples with high levels of human nucleic acid further reduced the identification of scarce microorganisms (Jacob, 2013; Grumaz et al., 2016). Furthermore, filamentous fungi mainly spread on the surface of lung tissue, making it difficult to wash pathogens off using lavage, resulting in small reads of fungi detected by BALF-mNGS (Yang et al., 2021). Several case reports have suggested that the reads numbers of aspergillus detected by mNGS as clinically significant pathogens range from 11 to 57 (He et al., 2019; Yang et al., 2022). Due to the inability to exclude low abundance organisms as potential causative agents (Li et al., 2018; He et al., 2019; Yang et al., 2022), the interpretation of the complex and diverse data generated by mNGS is more challenging. Even to this day, there is still a lack of widely accepted standards and quantitative thresholds (Miao et al., 2018; Fang et al., 2020; Qin et al., 2021; Yang et al., 2021; Ao et al., 2023). The evaluation for mNGS using absolute values as criteria like unique reads number remained controversial since the difficulty of extracting nuclei acids varies from species to species and the total sequencing yield varies from sample to sample (Li et al., 2018). Thus, mNGS cannot completely replace CMTs, but rather serves as a valuable adjunctive tool, particularly in scenarios where CMTs prove inadequate or yield inconclusive results. In addition, the interpretation of mNGS results requires advanced bioinformatics tools and expertise, as well as careful consideration of patient-specific factors, such as immune status, underlying diseases, antifungal pretreatment and clinical manifestations (Langelier et al., 2018; Diao et al., 2022; Jiang et al., 2022).

This study had several limitations that should be acknowledged. First, it was a single-center retrospective study; thus, intrinsic bias was unavoidable. Second, the diagnostic performances of mNGS and PCR method were not compared as PCR was not routinely performed in our hospital. This line of query should be explored in future work. Third, we used the clinical composite diagnostic criteria as the gold standard, but it must be acknowledged that even for experienced physicians, clinical misjudgment cannot be completely avoided, leading to deviations in diagnostic performance between different methods. Finally, the generalizability of mNGS is subject to certain limitations, and the next big challenges will be the optimization and standardization of sampling and mNGS analysis, which promise to enhance the clinical practicality of genomic diagnosis. Also, the potential benefits of mNGS in management of IPA warrant further investigation.

In summary, mNGS is feasible and highly sensitive method for detecting Aspergillus in ICP. It also performed well in identifying species of fungi and co-pathogens, providing significant benefits to ICP. However, due to the lack of widely accepted standards and inherent technical limitations, mNGS cann’t yet replace CMT, and the interpretation of its results should be combined with patient-specific factors.

The datasets presented in this article are not readily available because of license/restriction from the affiliated institution. Requests to access the datasets should be directed to YS, pumchshi@sina.com.

The studies involving humans were approved by The ethics committee of Peking Union Medical College Hospital. The studies were conducted in accordance with the local legislation and institutional requirements. The ethics committee/institutional review board waived the requirement of written informed consent for participation from the participants or the participants’ legal guardians/next of kin because given the retrospective design.

YS conceived and designed this study. Material preparation and data collection by YS, J-MP and X-YH. Q-WY and YW was responsible for coordinating microbiological testing and interpretation. All authors were involved in drafting the article or revising it critically for important intellectual content. All authors contributed to the article and approved the submitted version.