Patient characteristics
A total of 115 patients with pneumonia were enrolled and divided into ICH group (n = 64) and ICO group (n = 51), respectively, according to the criteria we previously reported[25] (Table 1). Among ICH patients, hematological malignancy (34%) was the most common comorbidity, followed by solid organ malignancy (25%), immunosuppressive treatment (14%), and hematopoietic cell transplantation (14%). The median ages of ICH and ICO patients were 50 and 44 years old, respectively. A total of 339 specimens were collected from all patients. In detail, 32 BALF, 55 blood, 39 sputum, and 58 swab samples were collected from ICH patients, while 27 BALF, 43 blood, 41 sputum, and 44 swab specimens were collected from ICO patients.
Table 1
Clinical features | ICH | ICO |
Number of cases | 64 | 51 |
Sex, female | 24(38%) | 24(47%) |
Age (mean ± SD) | 50 ± 19 | 44 ± 20 |
ICH type | Solid organ transplantation | 3 (5%) | NA |
Hematopoietic cell transplantation | 9 (14%) | NA |
Solid-organ malignancy | 16 (25%) | NA |
Hematopoietic malignancy | 22 (34%) | NA |
Immunosuppressive treatment | 9 (14%) | NA |
Other types of immunodeficiency disease | 5 (8%) | NA |
PSI score | 1 | 2 (3%) | 27 (53%) |
2 | 16 (25%) | 17 (33%) |
3 | 26 (41%) | 4 (8%) |
4 | 14 (22%) | 2 (4%) |
5 | 6 (9%) | 1 (2%) |
Sample type | Balf | 32 | 27 |
Blood | 55 | 43 |
Sputum | 39 | 41 |
Swab | 58 | 44 |
PSI, Pneumonia Severity index; Balf, bronchoalveolar lavage fluid; EBV, Epstein-Barr virus. |
Microbial composition and diversity
Firstly, to characterize the microbial communities of ICH and ICO patients with pneumonia, the relative abundances of microbes in BALF and sputum samples were analyzed. For BALF samples, Mycoplasma pneumoniae, Pseudomonas aeruginosa, Prevotella melaninogenica, Human adenovirus 55, Rothia mucilaginosa, and Pneumocystis jirovecii were detected in both ICH and ICO patients (Figure S1A). Lower relative abundance of M. pneumoniae and higher relative abundance of P. jirovecii were found in ICH patients than in ICO patients (Figure S1B), which was consistent with our previous findings[25]. For sputum samples, the microbiota was mainly consisted of R. mucilaginosa, Lautropia mirabilis, P. melaninogenica, Veillonella parvula, Streptococcus I P16, and M. pneumoniae (Figure S1C). Compared with ICO patients, relative abundances of P. melaninogenica, M. pneumoniae, and V. dispar were lower in ICH patients (Figure S1D).
Both BALF metaDNA-seq data (Figs. 1A and 1B) and BALF metaRNA-seq data (Figs. 1C and 1D) revealed that ICH patients had higher richness and shannon indexes than ICO patients. Additionally, NMDS analysis using both BALF metaDNA-seq (Fig. 1E) and metaRNA-seq (Fig. 1F) data found clear variation in the microbiota profiles between ICH and ICO groups. However, both sputum metaDNA-seq data (Figs. 1G and 1H) and metaRNA-seq data (Figs. 1I and 1J) showed that ICH patients exhibited a significantly lower shannon index than ICO patients. Besides, both NMDS (Fig. 1K) and PCoA (Fig. 1L) showed that the total microbial compositions were much different between ICH and ICO groups.
Microbial diversity and Pneumonia Severity Index (PSI) score
PSI score is widely used for evaluating pneumonia severity, and the higher the value, the more serious the disease. The PSI score of the majority of ICH patients (n = 26, 41%) was 3, while patients with PSI score of 1 accounted for 53% (n = 27) of the ICO patients (Figure S2). Compared with patients with lower PSI scores, lower relative abundance of M. pneumoniae and higher relative abundance of P. jirovecii were found in BALF samples of patients with higher PSI scores, while lower relative abundance of P. melaninogenica and higher relative abundances of A. baumannii, Corynebacterium striatum, Streptococcus infantis, and C. glabrata were found in sputum samples (Figure S3).
The correlation between microbial diversity and PSI score was further assessed using metaDNA-seq data. For BALF samples, both richness and shannon indexes were positively correlated with PSI score (Figs. 2A and 2B). When PSI score was ≥ 2, the higher the PSI score was, the lower the richness (Fig. 2C) and shannon indexes were (Fig. 2D) in sputum samples. Furthermore, NMDS analysis using BALF metaDNA-seq data found a clear distance in the microbiota profiles between the patients with PSI score of 1 and the others (2, 3, or 4) (Fig. 2E), while microbial composition of patients with PSI score of 5 was different from that of patients with other PSI scores using sputum samples (Fig. 2F).
EBV and disease severity
Microbes with average relative abundance of > 1% detected by metaDNA-seq were selected to evaluate the association between microbial abundance and PSI score. It was observed that the relative abundances of EBV and M. pneumoniae in BALF samples, Human betaherpesvirus 5 (CMV) in blood samples, C. striatum, EBV, M. pneumoniae, and P. jirovecii in sputum samples, and A. graevenitzii, EBV, P. melaninogenica, P. nanceiensis, and S. infantis in swab samples were significantly correlated with PSI score, respectively (Fig. 3A). Positive correlations between EBV abundance and PSI score in sputum and swab samples were also found using metaRNA-seq data (Fig. 3B). Moreover, EBV abundances in ICH patients were notably higher than those in ICO patients (Figure S4).
Given the strong correlation between EBV abundance and PSI score, we further divided the enrolled patients into EBV-detected (samples, n = 112) and EBV not-detected (samples, n = 227) groups (Figure S5). ICO patients with EBV detection by BALF metaDNA-seq had high PSI scores and disease severity (Fig. 4A). Additionally, EBV detected by sputum and swab metaDNA-seq was significantly associated with high PSI score and increased disease severity (Figs. 4C and 4D). However, using BALF metaDNA-seq data, there was no significant correlation between EBV and disease severity in ICH patients (Fig. 4A), and a same trend was found in both ICO and ICH patients using blood metaDNA-seq data (Fig. 4B).
A slight difference in microbial communities was observed between EBV-detected and not-detected groups, manifested by higher relative abundance of L. mirabilis in BALF samples and lower relative abundances of M. pneumoniae and S. infantis in sputum samples, compared to EBV not-detected samples (Figure S6). Additionally, reactivated EBV has been extensively reported to regulate the host immune response and microbial composition to promote viral persistence[39–41]. The aforementioned results raised a question that whether EBV was reactivated in ICH and ICO patients. We hypothesized that combined with other microbes, EBV might play an important role in exacerbating pneumonia.
Host response
MetaRNA-seq data were further analyzed (Figs. 5 and S7). The expression of 801 genes were significantly changed in EBV-detected group compared to EBV not-detected group (Figure S7 and Table S1). For sputum samples, there were 670 up-regulated genes and 94 down-regulated genes in EBV-detected ICH patients compared with EBV not-detected group, while 157 genes and 245 genes were respectively up-regulated and down-regulated in ICO patients (Fig. 5A). For swab samples, few genes were significantly up-regulated or down-regulated in both ICH and ICO patients (Fig. 5B). For BALF samples, there were 183 up-regulated genes and 85 down-regulated genes in EBV-detected ICH patients compared with EBV not-detected group, while no significant differences were found between EBV-detected and EBV not-detected ICO patients (Fig. 5C). The above findings indicate that EBV might be reactivated in ICH patients, while EBV might be latent in ICO patients.
KEGG enrichment analysis of immune related genes (IRGs) using BALF samples (Fig. 5D) identified 18 differential pathways in ICH patients. Cytokine-cytokine receptor interaction had 30 DEGs, followed by Chemokine signaling pathway (8 DEGs) and JAK-STAT signaling pathway (7 DEGs). Furthermore, 33 IRGs (Table S2) and 40 interferon stimulated genes (ISGs) were differentially expressed (Table S3). A total of 33 IRGs were differentially down-regulated in ICH patients. CSF1R, CXCR6, IL10, IL16, and TNFRSF25 were significantly down-regulated in ICH patients. Additionally, the expression of 37 ISGs were significantly down-regulated in ICH patients, while CNP and GEM were significantly down-regulated in ICH patients.
For innate immune response, the macrophage M0 and neutrophils accounted for the majority of the cells in both ICH and ICO patients (Fig. 5E). The ratios of neutrophils, monocytes, and dendritic cells activated in EBV-detected ICH patients were lower than those in EBV not-detected ICH patients. For adaptive immune response, high ratio of T cells CD4 memory resting was detected in both ICH and ICO patients, followed by NK cells resting (Fig. 5E). Additionally, the ratios of T cells CD4 memory resting and T cells CD4 memory activated were lower in EBV-detected groups than in not-detected groups in both ICH and ICO patients. Hence, we hypothesized that presence of reactivated EBV might inhibit the immune response of hosts to increase the possibilities of infections caused by other pathogens, exacerbating the pneumonia severity in EBV-detected group.
To verify our hypothesis, co-occurrence network was further constructed to evaluate the relationship between EBV and other microorganisms. We found that EBV had strong correlations with other 48 microbial species, most of which were negative (blue line) (Fig. 6 and Table S4). Positive correlations between EBV and Citrobacter freundii or Campylobacter concisus were found, indicating that synergistic effects on exacerbating the severity of pneumonia might exist between EBV and these two microbes.