Introduction

Pneumocystis jirovecii is a fungus that causes pneumonia in immunocompromised patients. Clinically P. jirovecii pneumonia is characteristically associated with fever, shortness of breath and hypoxia and radiologically ground glass opacities are often observed. In general, it causes a relatively milder disease with lower (10–12%) mortality in human immunodeficiency virus (HIV)-positive patients but more severe disease with higher (30–50%) mortality in other immunocompromised patients who are HIV-negative [1]. Traditionally, laboratory diagnosis of P. jirovecii pneumonia was achieved by direct detection of P. jirovecii asci in respiratory tract specimens by microscopic examination after Grocott-Gomori methenamine silver (GMS) or immunofluorescence staining [2]. In recent years, polymerase chain reaction (PCR) has also been used for the detection of P. jirovecii [3,4,5,6]. Although it has improved the sensitivity of detection, it is not able to distinguish between genuine P. jirovecii pneumonia and P. jirovecii colonization of the respiratory tract [7, 8].

In the last few years, next-generation sequencing (NGS) has emerged as a technology for laboratory diagnosis of many culture-negative infections [9, 10]. We have recently reported its application in confirming the first case of listeria meningitis in a patient with autoantibody against interferon gamma as well as understanding the spectrum of Q fever, fungal infections and culture-negative meningitis and encephalitis [10,11,12,13]. It is notable that we have shown, in our recent review, that in fact P. jirovecii is the commonest fungal organism detected by NGS in clinical specimens [10]. Furthermore, others have also shown that NGS is much more sensitive than direct GMS staining and microscopy for the detection of P. jirovecii in respiratory samples [14,15,16]. In this study, we describe the epidemiology of P. jirovecii infection and colonization diagnosed by NGS in our hospital and discuss the reasons that may account for such phenomena. In addition, we also explore the usefulness of the number of P. jirovecii sequence reads for the diagnosis of P. jirovecii pneumonia and discuss how to interpret NGS results.

Materials and Methods

Ethical Statement

This study was approved by the Institutional Review Board of The University of Hong Kong—Shenzhen Hospital ([2022]120), and the requirement of obtaining informed consent was exempted.

Patients

This study was conducted over a 12-month period (January to December 2022) in The University of Hong Kong—Shenzhen Hospital, Shenzhen, China. This 1,400-bed multi-specialty hospital was established in 2012 and provides primary to tertiary medical services to the residents of Shenzhen city in both inpatient and outpatient settings. Supported through the policy from the government of Shenzhen, the hospital is established as a reform model medical institution in China, and many new medical technologies can be introduced to the hospital first. The laboratory reports of all respiratory samples submitted for NGS were examined. The clinical details, laboratory data and radiological findings of all patients with P. jirovecii sequence reads detected in their respiratory samples were retrieved from the hospital electronic record system and analysed.

Case Definitions

According to the Consensus Definitions of Invasive Fungal Disease of the European Organization for Research and Treatment of Cancer and the Mycoses Study Group [17], a case of P. jirovecii infection is defined as a P. jirovecii NGS-positive patient, whom after careful consideration of his/her clinical, radiological and laboratory findings, the clinician-in-charge has decided to prescribe specific anti-P. jirovecii treatment. A case of P. jirovecii colonization is defined as a P. jirovecii NGS-positive patient, whom after careful consideration of his/her clinical, radiological and laboratory findings, the clinician-in-charge has decided not to prescribe specific anti-P. jirovecii treatment.

Microbiological and Other Laboratory Tests

Clinical specimens were collected and handled according to standard protocols [18]. Direct detection of P. jirovecii and acid-fast bacilli were performed by GMS stain and Ziehl–Neelsen stain, respectively. The identities of bacterial and fungal isolates were confirmed by matrix-assisted laser desorption ionization–time-of-flight mass spectrometry [19]. Cryptococcal antigen detection was performed using lateral flow assay (Norman, USA). 1,3-β-D-glucan detection was performed using Test Kit for the Detection of Fungus 1,3-β-D-Glucan (Photometric Assay) (A & C Biological Ltd, Zhanjiang, China). Real-time PCR for Mycobacterium tuberculosis was performed using M. tuberculosis DNA Fluorescence Diagnostic Kit (PCR-Fluorescence Probing) (Sansure Biotech, Hunan, China); and real-time RT-PCR for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) was performed using 2019-nCoV Nucleic Acid Test Kit (Biogerm, Shanghai, China). Real-time RT-PCR for influenza virus A and real-time PCR for herpes simplex virus (HSV) were performed by KingMed Diagnostics company.

NGS

The sputum and bronchoalveolar lavage (BAL) samples were sent to KingMed Diagnostics company, Sagene company, Vision Medicals company, Dinfectome company or GensKey company for targeted NGS (tNGS) or metagenomics NGS (mNGS) analysis.

Statistical Analysis

A comparison of characteristics between the P. jirovecii infection and colonization groups was performed. Chi-square test was used for categorical variables and unpaired Student’s t-test or Mann–Whitney U test was used for continuous variables. P < 0.05 was considered as statistically significant.

Results

Patients

During the 12-month study period, a total of 285 respiratory samples from 241 patients were submitted for tNGS or mNGS analyses. Among these 285 samples, P. jirovecii sequence reads were detected in 56 samples from 53 patients. For these 53 patients, the male to female ratio was 35:18. The median age was 61 (range 30 to 85) years. Thirty-eight (71.7%) of the 53 patients had underlying immunocompromised conditions, the commonest being malignancies (n = 19), followed by connective tissue and autoimmune diseases on corticosteroid and/or other immunosuppressive treatment (n = 15), solid organ transplant recipients on corticosteroid and/or other immunosuppressive treatment (n = 4) and HIV infection (n = 3) (Table 1).

Table 1 Demographic and clinical characteristics of patients in whom Pneumocystis jirovecii was detected by NGS
Full size table

NGS Analysis

In 44 of the 53 patients, P. jirovecii was detected by tNGS; whereas in nine patients, it was detected by mNGS. In samples collected from 31 (58.5%) of the 53 patients, sequence reads of other potential respiratory pathogens were detected (Table 2). These included bacteria (Acinetobacter baumannii, Bordetella pertussis, Chlamydia psittaci, Haemophilus influenzae, Klebsiella pneumoniae, Legionella pneumophilia, Pseudomonas aeruginosa, Staphylococcus aureus, Stenotrophomonas maltophilia and Streptococcus pneumoniae), mycobacteria (Mycobacterium abscessus, Mycobacterium chelonae, Mycobacterium intracellulare, Mycobacterium kansasii and M. tuberculosis), viruses (adenovirus, influenza virus and rhinovirus) and fungi (Aspergillus fumigatus, Cryptococcus neoformans and Trichosporon asahii). In samples collected from 26 (49.1%) of the 53 patients, sequences that were considered as contaminants or colonizers were present. Most of them were bacteria and yeasts present in the oral cavity of immunocompetent or immunocompromised hosts (Supplementary Table 1). In samples collected from eight (cases 5, 17, 29, 30, 31, 32, 37 and 51) of the 53 patients, Tropheryma whipplei, a bacterium of doubtful clinical significance in the respiratory tract [21], was detected (Table 2).

Table 2 NGS analysis and other key laboratory results of patients in the present cohort
Full size table

Other Microbiology Tests

Serum cryptococcal antigen was positive in one patient (case 46), and acid-fast bacilli were detected in two patients (cases 20 and 39) and M. tuberculosis DNA in one patient (case 39) (Table 2). HSV DNA was positive in the BAL of one patient (case 43). Influenza A virus and SARS-CoV-2 RNA were each positive in the nasopharyngeal swab of two patients respectively (cases 23 and 51) (Table 2).

P. jirovecii Infection and Colonization

According to our case definitions, 37 (69.8%) of the 53 patients with P. jirovecii sequence reads detected in their respiratory samples had P. jirovecii infection, whereas the other 16 (30.2%) of the 53 were considered as colonization. P. jirovecii infection was associated with the presence of underlying disease with immunosuppression (35/37, 94.6%) compared to P. jirovecii colonization (3/16, 18.8%) (P < 0.05) (Table 3). Only two patients with P. jirovecii pneumonia did not have major immunosuppression. The first one (case 21, Table 1) was a 71-year-old man with membranous nephropathy, hypertension, diabetes mellitus and coronary heart disease. He refused to receive corticosteroid and other immunosuppressive treatment for his membranous glomerulonephritis. The second one (case 44, Table 1) was a 73-year-old man with chronic obstructive pulmonary disease, coronary heart disease, gout, hypertension, renal calculi and benign prostatic hyperplasia. In addition to underlying diseases, patients with P. jirovecii infection were associated with higher number of P. jirovecii sequence reads in their respiratory samples than those with P. jirovecii colonization (P < 0.005) (Table 3). Furthermore, there were significantly more patients with positive serum 1,3-β-D-glucan results in the infection (14/34, 41.2%) than in the colonization group (0/11) (P < 0.01) (Table 3). On the other hand, P. jirovecii colonization was associated with a predominance of the male sex (15/16, 93.8%) compared to P. jirovecii infection (20/37, 54.1%) (P < 0.01) (Table 3). Patients with P. jirovecii colonization was also associated with another definitive infectious disease diagnosis of the respiratory tract (7/16, 43.8%) compared to P. jirovecii infection (1/37) (P < 0.001) (Table 3). The definitive diagnoses of these seven patients were psittacosis (case 8), M. kansasii pulmonary infection (case 20), cryptococcosis (case 46), tuberculosis (case 39 and 47), influenza (case 23) and COVID-19 (case 51) (Table 2). Moreover, a significantly higher proportion of patients with P. jirovecii colonization (by definition did not receive specific anti-P. jirovecii treatment, 16/16, 100%) had improved compared to those with P. jirovecii infection (by definition received specific anti-P. jirovecii treatment, 25/37, 67.6%) (P < 0.01) (Table 3).

Table 3 Comparison of characteristics in patients with Pneumocystis jirovecii infection and colonization
Full size table

Discussion

In this study, 53 patients in our hospital with P. jirovecii sequences in their respiratory samples were detected by tNGS or mNGS analysis. Among these 53 patients, only three were HIV-positive (cases 12, 41 and 45, Table 1), whereas the other 50 were HIV-negative. This is very different from the general epidemiology of P. jirovecii infections, of which HIV infection is the single most important risk factor. The common reasons for immunosuppression in the HIV-negative patients in this cohort were solid tumour or haematological malignancies on chemotherapy and autoimmune diseases or solid organ transplant recipients on corticosteroid and/or other immunosuppressive treatment, which is consistent with the changing epidemiological profile of P. jirovecii infection in the past decades [22]. For HIV-positive patients with P. jirovecii infections, the fungal loads in their respiratory tracts are usually high and direct microscopic examination after GMS staining, sometimes even using induced sputum samples, is often sufficient for making a diagnosis. In contrast, for the other immunocompromised patients, the fungal load is usually low and bronchoscopic examination has to be performed to collect BAL samples so as to improve the yield. In fact, for all the three HIV-positive patients in the present cohort, their BAL samples were also positive for P. jirovecii by direct microscopic examination after GMS staining, whereas for the 29 BAL samples obtained from the HIV-negative patients that were submitted for microscopic examination after GMS staining, only three were positive for P. jirovecii (P < 0.001 by Fisher’s Exact test). This is in line with the high number of P. jirovecii sequence reads (84,000 for cases 12, 58,385 for case 41 and 64,629 for case 45) (Table 2) observed in the three BAL samples collected from the three HIV-positive patients examined by tNGS, which is significantly higher than the number of P. jirovecii sequence reads (median 655, range 27 to 48,609) in the BAL samples collected from the 21 HIV-negative patients examined by tNGS (P < 0.001 by Mann–Whitney U test).

Detection of P. jirovecii sequence reads in respiratory samples has to be interpreted discreetly. Traditionally, P. jirovecii infection was diagnosed in the laboratory by direct detection of P. jirovecii asci in respiratory samples after GMS staining in immunocompromised patients with suspected clinico-radiological features, such as shortness of breath, hypoxia and ground glass infiltrates on chest radiographs. In the past decades, a number of PCR assays have been developed for the detection of P. jirovecii in respiratory tract specimens [3,4,5,6]. In some of these studies, colonization of P. jirovecii in the respiratory tract has been suggested [5,6,7,8]. In the present cohort, 16 (29.6%) of the 53 patients with P. jirovecii sequences in their respiratory samples detected by NGS analysis recovered without receiving specific anti-P. jirovecii therapy. In some of these 16 patients, other respiratory pathogens were present. For example, C. psittaci was detected from the sputum of Case 8 and he responded to doxycycline well; and in Cases 23 and 51, influenza A virus and SARS-CoV-2 RNA were detected in their nasopharyngeal and throat swabs respectively (Table 2). All of the 16 patients improved after receiving specific antimicrobial therapy to the other pathogens identified or just symptomatic treatment. In these 16 patients, P. jirovecii was considered as colonizers of the respiratory tract, rather than pathogens; and they were associated with the male gender, absence of underlying disease, negative serum 1,3-β-D-glucan, and a lower number of P. jirovecii sequence reads (Table 3). It is interesting to note that in our recent study on the detection of T. whipplei in respiratory samples by NGS, Whipple disease was never suspected to be a diagnosis in any of the patients before detection of the bacterium; and the presence of T. whipplei in the respiratory specimens of these patients was still elusive [21].

Although infection is associated with a significantly higher number of P. jirovecii sequence reads as compared to colonization, clinical judgement is still the most crucial in determining whether a particular case is genuine P. jirovecii pneumonia. When the number of P. jirovecii sequence reads in all respiratory (sputum and BAL) samples between the infection and colonization groups were compared, it was observed that the number of sequence reads was significantly higher in the infection than the colonization group (P < 0.005) (Fig. 1A). However, for example, if 79.5 reads were used as the cutoff for distinguishing between P. jirovecii pneumonia and colonization, only a sensitivity of 82.14% and a specificity of 68.75% could be achieved. Furthermore, when the analysis was performed for the sputum group, the number of sequence reads was still significantly higher in the infection than the colonization group (P < 0.05); but when the analysis was performed for the BAL group, there was no difference between the number of reads in the two groups, although there was still a trend towards a higher number of reads in the infection than the colonization group (Fig. 1B). All these showed that the number of sequence reads is not a reliable parameter to indicate whether a particular patient has P. jirovecii pneumonia or just P. jirovecii colonization. In contrast to the number of P. jirovecii sequence reads, it was shown in the present cohort that positive serum 1,3-β-D-glucan and direct GMS staining of respiratory samples were highly specific, although not sensitive, for P. jirovecii infection (Table 3); and hence would be useful for the prediction of P. jirovecii pneumonia if these results were positive. The final diagnosis of P. jirovecii pneumonia should be made using a combination of clinical, radiological and laboratory findings.

Fig. 1
figure 1

Distribution in number of P. jirovecii sequence reads in respiratory samples from patients in the present cohort detected by targeted next-generation sequencing (tNGS). Panel A: boxplot showing number of P. jirovecii sequence reads distribution in all respiratory samples from patients with P. jirovecii infection and colonization. Panel B: boxplot showing number of P. jirovecii sequence reads distribution in bronchoalveolar lavage (BAL) and sputum samples from patients with P. jirovecii infection and colonization

Full size image