Histologic characterization of the lymphoid tissue of the human nasopharynx

First, a histologic result of NP biopsy from a 21-year-old male patient without history of SARS-CoV-2 infection was provided from the department of pathology in Seoul national university college of medicine (co-author YK Jeon) and investigated the histology of human NP lymphoid tissue. The histologic results revealed that a multilayered pseudostratified columnar epithelium and a squamous epithelium were observed at the surface of NP lymphoid tissue. This tissue showed distinctive histologic characteristics compared with nasal mucosa, and lymphoid follicles, germinal center, and intercellular T-cell zones were observed beneath NP epithelium (Fig. 1a). Histologic examination of nasal mucosa revealed a multilayered pseudostratified columnar epithelium, ciliated cells, and secretory cells. Both submucous glands and venous vessels were located beneath the nasal epithelium and were separated from the epithelium by a basement membrane (Fig. 1b). The germinal center, lymphoid follicles, and T-cell zone were not observed in the nasal mucosa. We estimated that the NP tissue had a unique structure similar to that of a lymph node and possessed a respiratory epithelium, suggesting that epithelial-derived innate immune responses could be activated when SARS-CoV-2 contacts the NP.

Fig. 1

Histologic findings of NP and nasal mucosa and differentially expressed genes and cell proportions in RNA-seq datasets of CoV2 + patients. (a) The patient’s nasopharyngeal lymphoid tissue was observed and sampled through an intranasal endoscope by an ENT specialist. The histologic structure of hematoxylin and eosin–stained NP lymphoid tissue. Pseudostratified columnar epithelium (black arrow), squamous epithelium (black triangle), enlarged lymphoid follicles (white arrow), germinal center (#), and intercellular T cell zones (*) were characteristic of NP lymphoid tissue. Scale bar; 100 µM. (b) Histologic structure of hematoxylin and eosin–stained nasal mucosa. Pseudostratified columnar epithelium (black arrow), submucous glands (black triangle), and venous sinusoid (white arrow) were characteristic of nasal mucosa. Scale bar; 200 µM. The histologic results are representative of NP tissues and nasal mucosa from three adults. (c) Volcano plot of regulated genes on nasopharyngeal swabs of 3 mild-CoV2 + patients and 3 healthy controls. Cut-off values of differentially expressed genes are log2-fold change > |2.0| and adjusted p < 0.05. (d) Regulated genes on nasopharyngeal swabs of 3 severe-CoV2 + patients and 3 healthy controls. Cut-off values of differentially expressed genes are log2-fold change > |2.0| and adjusted p < 0.05. (e) Heatmap of average differentially expressed genes of controls, mild, and severe CoV2 + patients. The 35 most common upregulated and downregulated genes of CoV2 + patients compared with controls (n = 3 for each group)

DEGs and cellular composition in the NP of mild or severe COVID-19 patients

We identified the DEGs in the NP of CoV2 + patients through bulk RNA sequencing (bulk-seq) as well as the differences between mild (n = 3) and severe (n = 3) cases and compared the transcription profiles with NP of healthy subjects (those who tested negative for SARS-CoV-2 [CoV2−], n = 3). Bulk-seq data identified 554 upregulated and 869 downregulated DEGs in the NP of CoV2 + patients, with a cutoff value of (Log2FC) ≥ 2 and an adjusted p < 0.05. Among them, 498 upregulated and 809 downregulated DEGs were found in CoV2 + patients with mild disease compared to CoV2 − patients (Fig. 1c), and 546 upregulated and 843 downregulated DEGs were found in CoV2 + patients with severe disease (Fig. 1d). We found striking differences between CoV2 + and CoV2 − patients regarding the transcriptions of IFN-related innate immune responses in the NP and the top 35 upregulated DEGs in CoV2 + patients compared with the comparable DEGs in CoV2 − patients, with a focus on ISGs (Fig. 1e). The heterogeneity of individual sample data was shown in Supplementary Fig. 1. These findings indicate potent induction of ISGs in the NP following SARS-CoV-2 infection the induction of ISGs was also identified in the NP of severe-CoV2 + patients.

Innate immune responses in the NP following SARS-CoV-2 infection

As a next step, we performed active-subnetwork-oriented enrichment analysis using on GO-BP terms. Enrichment p values were adjusted with Bonferroni method and filtered the enriched terms by adjusted p value under 0.05 [26]. The GO categories of “response to virus” and “defense response to virus” were most significant in the NP of both mild- (Fig. 2a) and severe-CoV2 + patients (Fig. 2b) compared to controls. An analysis of the genes involved in the enriched terms of mild- (Fig. 2a) and severe-CoV2 + patients (Fig. 2b) showed common or distinct genes between terms. Transcription of diverse ISGs and IFN-related genes was elevated in the NP of mild and severe-CoV2 + patients, but no induction of IFNs was detected in bulk-seq data. In mild-CoV2 + patients, 23 DEGs were included in the “defense response to virus” GO category and 19 in the “response to virus” category compared with CoV2 − patients (Fig. 2b). In addition, 25 DEGs included in the “defense response to virus” category and 19 DEGs in the “response to virus” category was found in the NP of severe-CoV2 + patients (Fig. 2b). These included IFI44L, MX1, MX2, RSAD2, IRF7, BST2, and APOBEC3A, while ISG15 showed the largest increase in the NP of both mild- and severe-CoV2 + patients.

Fig. 2

Enrichment analysis of CoV2 + patients compared to healthy controls using gene ontology–biological process gene sets (n = 3 for each group). (a) An enrichment chart and a term gene heatmap of mild-CoV2 + patients (n = 3). (b) An enrichment chart and a term gene heatmap of severe-CoV2 + patients (n = 3). Enriched terms were filtered by an adjusted p value < 0.05 (Bonferroni method)

We then measured 18 cytokines, including IFNs and ISGs, in the NP of mild- (n = 30) or severe-CoV2 + patients (n = 30) at hospitalization (acute = AC) and discharge (convalescent = CV). PCR results showed that transcription of both TNFA and IL1B was elevated in the NP of mild-CoV2 + patients (p = 0.026 and 0.050) and severe-CoV2 + patients (p = 0.031 and 0.046) at AC, while mRNA levels were downregulated at CV (Fig. 3a, b). Expression levels of IFNA, IFNL1, and IFNL2/3 were increased in the NP of mild-CoV2 + patients at AC and were lower at CV. In contrast, IFNB mRNA level was not induced in the NP of CoV2 + patients at AC and CV. Although the expression level was lower than in mild-CoV2 + patients, the mean mRNA levels of IFNA, IFNL1, and IFNL2/3 were significantly higher in the NP of severe-CoV2 + patients at AC compared with that of the CoV2 − group, but IFNL4 and IFNG expression levels were significantly decreased in severe patients compared with mild patients (p = 0.005, and 0.013, respectively) (Fig. 3c-h).

Fig. 3

mRNA transcription levels of IFNs and ISGs, including (a) TNFA, (b) IL1B, (d) IFNA, (d) IFNB, (e) IFNG, (f) IFNL1, (g) IFNL2/3, (h) IFNL4, (i) CXCL10, (j) MX1, (k) IFIT1, (l) IFIT2, (m) IFIT3, (n) RSAD2, (o) USP18, (p) ISG15, (q) IFI27, and (r) IFI44L, were assessed in the control group, as well as in mild and severe patients during both the acute and convalescent phases (mean ± SEM; Mann-Whitney test; n = 10 for control, n = 30 for each CoV + group). All mRNA transcription levels were normalized to that of GAPDH. The significance levels are represented as follows: ns indicates p value > 0.05, * indicates p value < 0.05, ** indicates p value < 0.01, *** indicates p value < 0.001, and **** indicates p value < 0.0001

The mean mRNA levels of ISGs, including MX1, IFIT1, RSAD2, and USP18, were increased in the NP of mild-CoV2 + patients at AC (p < 0.001, 0.012, 0.017, and < 0.001, respectively), and no induction of ISGs was detected in mild-CoV2 + patients at CV (Fig. 3i-r). MX1, RSAD2, and USP18 mRNA levels were relatively high in severe-CoV2 + patients at AC (p = 0.046, 0.400, and 0.012, respectively), but the transcription levels of MX1 and RSAD2 in severe patients were lower than those in mild-CoV2 + patients (p = 0.018 and 0.043). Transcription of IFIT1 increased only in mild acute-phase CoV2 + compared with CoV2 − patients (p = 0.012). Although the mean mRNA level of IFI27 was different from bulk-seq data, a decrease in transcription level was seen in the NP of severe-CoV2 + patients compared with CoV2 − patients, and no significant induction of ISG15 and IFI44L was observed in the NP of mild- or severe-CoV2 + patients. Based on these findings, we estimated that significant induction of IFNs, except IFNB and ISGs including MX1, IFIT1, RSAD2, and USP18, occurred in the NP of mild-CoV2 + patients, and that IFN-related innate immune responses were activated in the NP of severe-CoV2 + patients at the onset of infection. However, it was difficult to find a correlation between the genes included in the downregulated GO categories and the immune response (Supplementary Fig. 2).

IFN-related immune cells in the NP of CoV2 + patients

Last, we used the CIBERSORT algorithm to calculate immune cell proportions and compared gene expression associated with 22 types of immune cells in the NP of CoV2 + patients. We used t-test to compare fraction of each immune cell type between control and CoV + patients. A significant increase was evident in the transcriptional proportion of M1 macrophages (MФs) in the NP of mild and severe-CoV2 + patients compared with CoV2 − subjects (p = 0.026), and a significant decrease was evident in the transcription of M2 MФ (p = 0.005). In addition, expression of genes associated with CD4 + memory T cells, activated dendritic cells (DCs) was greater (p = 0.079 and 0.017) in the NP of mild and severe-CoV2 + patients, while that of neutrophils was higher among mild-CoV2 + patients (Fig. 4a). In contrast to these immune cells, the transcription of CD8 + T cells was significantly attenuated in the NP of mild- and severe-CoV2 + patients (p = 0.009). Heatmaps show the relative expression of genes related to M1 MФ, activated DCs, neutrophils, CD4 + memory T cells, and CD8 + T cells from the LM22 [27] in the NP of CoV2 + patients (Fig. 4b-f). When investigating the COVID-19 IFN response, we found that transcriptions of ISGs which were correlated with the prognosis of CoV2 + patients and were significantly associated with higher transcriptions of M1 MФ and monocytes -related genes in the NP of CoV2+ (Fig. 4g). These results suggested that the transcriptions of immune cells were altered in the NP of CoV2 + patients and IFN-related innate immune response might be characteristic in M1 MФ of the NP, when the is trying to control the viral infection in upper airway.

Fig. 4

Average immune cell proportions and related DEGs among controls and CoV2 + patients (n = 3 for each group). (a) Average immune cell proportions for controls, mild-, and severe-CoV2 + patients (t-test). M1 and DCs are significantly elevated in CoV2 + patients. DEGs for (b) M1 macrophage, (c) Dendritic cells activated, (d) Neutrophils, (e) CD4 memory T cells (activated), (f) CD8 T cells according to LM22 were shown as heatmap. (g) Pearson correlation test between immune cell proportions of each sample and ISGs demonstrated the relevance of M1 macrophages and monocytes with ISGs. MX1, IFIT1, IFIT2, IFIT3, RSAD2, USP18, ISG15, IFI27, and IFI44L were log-transformed due to skewed distribution. The significance levels are represented as follows: * indicates p value < 0.05, ** indicates p value < 0.01, and *** indicates p value < 0.001

Induction of IFNs and ISGs in the NP is correlated with the clinical outcome of COVID-19 treatment

To better understand whether the induction of IFNs and ISGs in response to SARS-CoV-2 infection of the NP is correlated with the prognosis of COVID-19, we explored the correlations between IFN and ISG transcription and clinical factors in CoV2 + patients. First, we identified dynamic parameters over the prognosis of CoV2+, including age, sex, number of vaccine doses, previous history of COVID-19, hospitalization, length of hospital stays, presence of initial chest X ray abnormalities, SOFA score, and the incidence of comorbid diseases (Table 2). Next, we analyzed the correlations between the clinical parameters of CoV2 + patients and IFNs or ISGs (Fig. 5, Supplementary Tables 2, 3). The results revealed that age in CoV2 + patients was significantly negatively correlated with mean mRNA levels of IFNL1, IFNL2/3, IFNL4, IFNG, CXCL10, IL1B, MX1, IFIT2, RSAD2, ISG15, IFI27, and IFI44L. Hospitalization had a significant negative correlation with IFNL1, IFNL2/3, IFNL4, IFNG, MX1, IFIT2, RSAD2, ISG15, CXCL10, IL1B, IFI27, and IFI44L. The initial chest X-ray abnormality in CoV2 + patients was negatively correlated with the expression of IFNL1, IFNL2/3, IFNG, MX1, IFIT1, RSAD2, ISG15, and IFI27 (Fig. 5, Supplementary Tables 2, 3). The median duration of supplemental oxygen therapy of severe-CoV2 + patients was 8.5 days (Q1, 4.0 days; Q3, 15.0 days; mean ± standard deviation, 14.2 ± 23.2) (Fig. 5, Supplementary Tables 2, 3). SOFA score was negatively correlated with the expression of IFNL1, IFNG, MX1, RSAD2, IFI27, IFI44L. After adjusting for age (multiple linear regression), demographic value that showed a significant difference between mild and severe group, MX1 (coefficient: -0.455, p = 0.033) and RSAD2 (coefficient: -0.370, p = 0.024) expression levels significantly predicted the SOFA score (Table 3). We found that COVID-19 patients with significant induction of IFNs and ISGs in their NP showed relatively good clinical factors after treatment in the early stages of infection.

Table 2 The demographics of mild and severe COVID-19 patientsFig. 5

A heatmap of the correlations between clinical factors and markers of innate immune responses (n = 60). IFNs and ISGs were log-transformed for normalization. Pearson correlation test (point-biserial correlation test for binary variables) was performed. The significance levels are represented as follows: * indicates p value < 0.05, ** indicates p value < 0.01, and *** indicates p value < 0.001

Table 3 IFNs and ISGs related with SOFA score. Multiple linear regression adjusted for age and sex was performed. Log transformation was performed on IFNs or ISGs. IFNs or ISGs resulting in significant models are listedInnate immune signatures in the NP are associated with viral RNA level and the prognosis of COVID-19 treatment

We investigated the correlation between Ct value at admission or mortality predictions using SOFA scores, which is one of the criteria for determining the severity of disease more definitely. The results of the SARS-CoV2 PCR performed on NP samples when COVID-19 was confirmed showed that the viral RNA level did not exhibit a direct association with the severity of COVID-19 (t-test, p = 0.707) (Fig. 6a) or with the SOFA score of COVID-19 patients (Pearson correlation, p = 0.194) (Fig. 6b). After adjusting for age (multiple linear regression), demographic value that showed a significant difference between mild and severe group, the Ct value at admission influences approximately 1.8% of the SOFA score (multiple linear regression, coefficient: -0.054, p = 0.263, adjusted R-squared: 0.018). However, Innate immune signatures such as MX1, USP18, and ISG15 in the NP were associated with Ct value at admission of COVID-19 patients (Fig. 5, Supplementary Tables 2,3). The additional results revealed that the gene expression of IFNL1, IFNG, MX1, IFIT2, IFIT3, ISG15, RSAD2, IFI27, IFI44L, and USP18 in the NP was negatively correlated with Ct value at admission (Table 4), and CoV2 + patients with relatively low expression of innate immune factors in the NP had more severe disease. Based on these findings, we determined that MX1 and RSAD2 levels were significantly higher in the NP of CoV2 + patients with higher viral RNA level at initial stage of infection and the induction of both IFNs and ISGs in the NP of CoV2 + patients may be closely correlated to the good prognosis of COVID-19 treatment.

Fig. 6

Relationships between the Ct value at admission and the severity and innate immune response markers (mild n = 23, severe n = 28). (a) The Ct value at admission showed no difference between mild and severe-CoV2 + patients. The 2-tailed t-test was used to compare each group. ns indicates p > 0.05. Error bars indicate the SEM. (b) There was no significant correlation between the Ct value at admission and SOFA score (Pearson test)

Table 4 IFNs and ISGs related with ct value at admission. Multiple linear regression adjusted for age was performed. Log transformation was performed on IFNs or ISGs due to skewed distribution. IFNs or ISGs resulting in significant models are listed