Silent Infection of B and CD8⁺ T Lymphocytes by Influenza A Virus in Children with Tonsillar Hypertrophy

Influenza A virus (IAV) infections are important threats to human health worldwide. Although extensively studied, some aspects of virus pathogenesis and tissue tropism remain unclear. Here, by different strategies, we describe the asymptomatic infection of human lymphoid organs by IAV in children. Our results indicate that IAV was not only detected and isolated from human tonsils but displayed unique genetic features in comparison with those of contemporaneous IAVs circulating in Brazil and detected in swabs and nasal washes. Inside the tissue microenvironment, immune cells were shown to be carrying IAV antigens, especially B and T CD8⁺ lymphocytes. Taken together, these results suggest that human lymphoid tissues can be sites of silent IAV infections with possible impact on virus shedding to the population.

ABSTRACT

Influenza A viruses (IAVs) cause more than 2 million annual episodes of seasonal acute respiratory infections (ARI) and approximately 500,000 deaths worldwide. Depending on virus strain and host immune status, acute infections by IAV may reach sites other than the respiratory tract. In the present study, IAV RNA and antigens were searched for in tissues of palatine tonsils and adenoids removed from patients without ARI symptoms. A real-time reverse transcriptase PCR (RT-PCR) screening revealed that 8 tissue samples from 7 patients out of 103 were positive for IAV. Positive samples were subjected to next-generation sequencing (NGS) and 3 of 8 tissues yielded complete IAV pH1N1 genomes, whereas in 5 samples, the PB1 gene was not fully assembled. Phylogenetic analysis placed tonsil-derived IAV in clusters clearly segregated from contemporaneous Brazilian viruses. Flow cytometry of dispersed tissue fragments and serial immunohistochemistry of paraffin-embedded sections of naturally infected biopsies indicated that CD20⁺ B lymphocytes, CD8⁺ T lymphocytes, and CD11c⁺ cells are susceptible to IAV infection. We sought to investigate whether these lymphoid tissues could be sites of viral replication and sources of viable virus particles. MDCK cells were inoculated with tissue lysates, enabling recovery of one IAV isolate confirmed by immunofluorescence, reverse transcriptase quantitative PCR (RT-qPCR), and NGS. The data indicate that lymphoid tissues not only harbor expression of IAV proteins but also contain infectious virus. Asymptomatic long-term infection raises the possibility of IAV shedding from tonsils, which may have an impact on host-to-host transmission.

IMPORTANCE Influenza A virus (IAV) infections are important threats to human health worldwide. Although extensively studied, some aspects of virus pathogenesis and tissue tropism remain unclear. Here, by different strategies, we describe the asymptomatic infection of human lymphoid organs by IAV in children. Our results indicate that IAV was not only detected and isolated from human tonsils but displayed unique genetic features in comparison with those of contemporaneous IAVs circulating in Brazil and detected in swabs and nasal washes. Inside the tissue microenvironment, immune cells were shown to be carrying IAV antigens, especially B and T CD8⁺ lymphocytes. Taken together, these results suggest that human lymphoid tissues can be sites of silent IAV infections with possible impact on virus shedding to the population.

INTRODUCTION

According to the WHO, up to 5 million cases of acute respiratory infections (ARI) and approximately 500,000 human deaths worldwide are attributable to influenza virus annually (1). In addition to annual seasonal outbreaks, influenza can also cause sporadic pandemics when novel virus lineages are introduced in human populations. The latest pandemic influenza A virus (IAV) strain of the H1N1 subtype in 2009 resulted from complex reassortment events among avian-, swine-, and human-origin influenza virus strains (2).

In humans, influenza A infections occur preferentially in the respiratory tract, although several lines of evidence indicate that the virus may infect other tissues (3–6). Studies conducted in animal models have detected viral RNA in brain, heart, lung, thymus, spleen, liver, kidney, and pancreatic cells (4, 6). In hospitalized patients with avian H7N9, viral RNA was detected in urine, feces, and serum samples (7). Extrapulmonary detection of H5N1 viral RNA in liver, spleen, intestines, heart, and bone marrow of fatal human cases was also reported (8). Secondary lymphoid tissues, such as nasal-associated lymphoid tissue (NALT) and lymph nodes, are also possible sites of infection, as shown by detection of antigens by immunohistochemistry (IHC) and RNA replicative intermediates by in situ hybridization (ISH) (3, 5). The role of migrating dendritic cells (DCs) in the spread of IAV was indicated by the finding that infectious virions are delivered to draining lymph nodes in a mouse model (9, 10). A previous study by our group demonstrated that genomes of IAV and a variety of other respiratory viruses were detected by PCR in samples of human palatine tonsils and adenoids from asymptomatic children (11).

Secondary lymphoid organs are important sites where lymphocytes are continuously primed to multiple antigens and immune responses are mounted. In the human upper respiratory tract, the lymphoid structures around the pharynx are one pair of palatine tonsils, one pair of tubal tonsils, the adenoid, and the lingual tonsil. These are sites where the first responses against respiratory pathogens are organized (12). The cellular and humoral immune responses against IAV have been extensively studied in animal models, especially the roles of antibodies and CD8⁺ T lymphocytes (13–15). Studies carried out with human peripheral blood mononuclear cells (PBMCs) indicated that IAV can efficiently infect CD14⁺ monocytes and, to a lesser extent, CD4⁺ and CD8⁺ T and CD19⁺ B lymphocytes. In addition, avian influenza virus H7N9 induced higher levels of proinflammatory cytokines than those induced by common seasonal lineages (16, 17). Nonetheless, the different cells involved in the immune response against IAV and the roles played by infection of human lymphoid cells by IAV are still poorly understood.

The present study was conducted to assess the replicative state of IAV in adenoids and palatine tonsils from children without ARI symptoms. The results indicate that tonsils can harbor infectious IAV in susceptible immune cells and suggest that these secondary lymphoid organs may contribute to the maintenance of IAV in human populations, with possible impacts on human IAV evolution, initiation of transmission in the community, and mounting of cellular and humoral immune responses in tonsils.

RESULTS

Patients and samples.

Tissue fragments of palatine tonsils and adenoids were obtained from 103 patients (57% males) aged 3 to 13 years (mean age, 6.33 ± 5.22 years) from 2014 to 2016. None of the patients had symptoms of ARI at the time of surgery nor in the 4 weeks prior to surgery, and none had received nasal live-attenuated influenza vaccine. Neither nasopharyngeal swabs nor serum samples were collected at the time of surgery.

Detection of influenza viruses in tonsils.

All tissue samples were screened by real-time reverse transcriptase PCR (RT-PCR) for the M gene of IAV and the hemagglutinin (HA) gene of Influenza B virus (IBV). Of 103 children, 7 (6.8%) had at least one tissue sample positive for the IAV M gene (patient number 432 was positive in both palatine tonsil and adenoid tissues), totalizing 5 palatine tonsils and 3 adenoids (Table 1). The IBV HA gene was not detected in any of the samples tested.

TABLE 1.

Influenza viruses detected in human hypertrophic tonsillar tissues

Sample	Collection date (yr-mo)	Lymphoid tissue	Strain	GISAID accession no. (length of genomic segment)
				PB2	PB1b	PA	HA	NP	NA	M	NS
339 ADN	14-5	Adenoid	A/Ribeirao Preto/01-399ADN/2014(H1N1)	EPI1319701 (2,341 nt)	i	EPI1319702 (2,233 nt)	EPI1319703 (1,779 nt)	EPI1319704 (1,565 nt)	EPI1319705 (1,458 nt)	EPI1319706 (1,027 nt)	EPI1319707 (890 nt)
416 AMG	14-6	Palatine tonsil	A/Ribeirao Preto/02-416AMG/2014(H1N1)	EPI1319708 (2,341 nt)	i	EPI1319709 (2,232 nt)	EPI1319710 (1,779 nt)	EPI1319711 (1,565 nt)	EPI1319712 (1,458 nt)	EPI1319713 (1,027 nt)	EPI1319714 (890 nt)
418 AMG	14-6	Palatine tonsil	A/Ribeirao Preto/03-418AMG/2014(H1N1)	EPI1322831 (2,341 nt)	i	EPI1322832 (2,233 nt)	EPI1322833 (1,779 nt)	EPI1322834 (1,565 nt)	EPI1322835 (1,458 nt)	EPI1322836 (1,027 nt)	EPI1322837 (890 nt)
432 AMGa	14-7	Palatine tonsil	A/Ribeirao Preto/05-432ADN/2014(H1N1)	EPI1322845 (2,341 nt)	i	EPI1322846 (2,233 nt)	EPI1322847 (1,779 nt)	EPI1322848 (1,565 nt)	EPI1322849 (1,458 nt)	EPI1322850 (1,027 nt)	EPI1322851 (890 nt)
432 ADNa	14-7	Adenoid	A/Ribeirao Preto/04-432AMG/2014(H1N1)	EPI1322838 (2,341 nt)	i	EPI1322839 (2,234 nt)	EPI1322840 (1,779 nt)	EPI1322841 (1,565 nt)	EPI1322842 (1,458 nt)	EPI1322843 (1,027 nt)	EPI1322844 (890 nt)
443 AMG	14-8	Palatine tonsil	A/Ribeirao Preto/06-443AMG/2014(H1N1)	EPI1322852 (2,341 nt)	EPI1322853 (2,341 nt)	EPI1322854 (2,233 nt)	EPI1322855 (1,779 nt)	EPI1322856 (1,565 nt)	EPI1322857 (1,458 nt)	EPI1322858 (1,027 nt)	EPI1322859 (890 nt)
445 AMG	14-9	Palatine tonsil	A/Ribeirao Preto/07-445AMG/2014(H1N1)	EPI1322860 (2,003 nt)	EPI1322861 (2,341 nt)	EPI1322862 (2,229 nt)	EPI1322863 (1,778 nt)	EPI1322864 (1,565 nt)	EPI1322865 (1,458 nt)	EPI1322866 (1,027 nt)	EPI1322867 (890 nt)
448 ADN	14-9	Adenoid	A/Ribeirao Preto/08-448ADN/2014(H1N1)	EPI1322868 (2,341 nt)	EPI1322869 (2,341 nt)	EPI1322870 (2,233 nt)	EPI1322871 (1,779 nt)	EPI1322872 (1,565 nt)	EPI1322873 (1,458 nt)	EPI1322874 (1,027 nt)	EPI1322875 (890 nt)

^aTissue samples from the same patient.

^bi, Incomplete segment.

Genome sequencing and phylogenetic analysis.

After the initial screening by real-time RT-PCR, the 8 positive tissue samples were amplified by multisegment RT-PCR (MS RT-PCR) and sequenced by next-generation sequencing (NGS) (2, 18). The quality of sequencing was high and coverage was good for most segments, leading to the assembly of 3 complete genomes, plus 5 partial genomes from which only the PB1 segment was missing (Table 1). The NGS results revealed that, in 6 of 8 IAV, there was comparatively lower coverage of sequences in the internal portions than in the extremities of the three polymerase segments (PB2, PB1, and PA), suggesting potential internal deletions (Fig. 1). This sequence profile in the polymerase gene segments has been previously linked to the presence of defective interfering RNAs (DI-RNAs), both in laboratory adapted IAV lineages (19, 20) and in clinical samples (21). The consensus sequences of all gene segments were compared to other sequences available in GenBank using BLAST (22), revealing that all 8 matched to strains of the 2009 pandemic H1N1 lineage (A/H1N1pdm09). All influenza-positive samples were collected between May and September 2014, although in Brazil, H1N1 circulation occurred throughout the year (23) (Table 1). Phylogenetic analyses of all segments were performed to find evolutionary relationships between Ribeirao Preto IAV sequences and other IAV virus sequences from GenBank and the Global Initiative on Sharing All Influenza Data (GISAID). Gene sequences obtained from the Ribeirao Preto viruses were aligned to sequences deposited between 2009 and 2017, with emphasis on Brazilian sequences and the period from 2010 to 2014. Additionally, representative sequences of each previously defined clade of pH 1 hemagglutinin (24–26) were included in the alignment (see Table S1 in the supplemental material).

FIG 1.

NGS coverage read map through each segment of one example of influenza virus from Ribeirao Preto [A/Ribeirao Preto/08-448ADN/2014(H1N1)]. The nucleotide position along the segment is shown on the x axis, and the coverage is in on the y axis, which indicates the number of reads per position. The valley-like shape in the graphs of the polymerase genes (PB2, PB1, and PA) suggests the presence of DI-RNAs.

Sequences of IAV from tonsils detected in Ribeirao Preto in 2014 were more closely related to HA clade 3 viruses (Fig. 2), including Brazilian IAV sequences from Santo Angelo and Sao Paulo. The influenza virus sequences from tonsils clustered together and clearly separated from other sequences of A/H1N1pdm09 viruses circulating in Brazil within the same time frame. Interestingly, the cluster formed by Ribeirao Preto sequences was relatively distant from the Brazilian viruses that circulated in the same year and the previous 4 years, all of them clustered in the clades 6B, C, and 7. Similar clustering was also noted in the phylogenetic trees of the neuraminidase (NA) gene, as well as the internal genes PB2, PB1, PA, NP, M, and NS (Fig. 2).

FIG 2.

Phylogenetic analysis of whole genomes of influenza A (H1N1) viruses detected in human tonsils. The trees were constructed using the Bayesian Markov chain Monte Carlo (MCMC) temporal approach, and analysis was run for 50 million states with sampling every 1,000 states. A total of 91 (HA) and 70 (NA) sequences from 2009 to 2017 available in GenBank and GISAID were used to estimate the phylogenetic relationships with the Ribeirao Preto viruses (green branches, square-labeled), including representative sequences of each clade of HA. A/California/04/2009, A/California/07/2009, and A/California/04/164/MA/2009 (red branches) were used as roots for the trees. The BLAST closest hit sequences used in protein alignments were also included in the trees under the black dots. Brazilian sequences from 2009 to 2015 are displayed in blue branches. Subtrees for HA and NA showed in detail the branch times and the phylogenetically closest sequences of tonsil-derived viruses.

Genetic analysis and amino acid substitutions.

Despite the clear segregation indicated by the phylogenetic analysis, the viruses detected in adenoids and palatine tonsils showed slight internal genetic differences among themselves. Whole-genome analysis showed genetic distances among Ribeirao Preto tonsil viruses varying between 0 and 0.00438 for all 8 segments. The segments with the biggest distance variation were segments 4 (HA) (0 to 0.004183) and 6 (NA) (0 to 0.00438). When comparing with whole genomes from 39 influenza reference strains (2009 to 2015), the distance values were much more prominent (Fig. 3), corroborating the phylogenetic results. To probe further into specific features possibly related to this genetic segregation, all 8 viruses detected in tonsils had their predicted protein sequences aligned with the respective BLAST closest hits and with available Brazilian virus sequences from 2009, 2011, and 2015. The predicted protein sequences showed mutations common to Brazilian viruses that circulated in 2009, 2011, and 2015, indicating some degree of similarity with local contemporary viruses. A considerable number of amino acid mutations in all 11 predicted proteins were exclusive of Ribeirao Preto tonsil viruses (see Table S2 in the supplemental material). The protein sequences of tonsillar IAV with higher diversity in comparison with the reference BLAST hits were HA, PA, and PB1 with a total of 19, 14, and 9 nonsynonymous mutations, respectively (see Table S2). The protein sequences with the lowest diversity were M1 and NEP with 3 and 2 nonsynonymous mutations, respectively (see Table S2). Importantly, the HA segment had 5 out of 19 mutations mapping to the HA1 epitopes relevant for vaccine efficacy, namely, A (N133T), D (V202A, R208K), and E (I66V, S78P) (27, 28).

FIG 3.

Genetic distances pairwise between all 8 genomic segments of Ribeirao Preto influenza A (H1N1) viruses. The heatmaps were created using R program and Ape package. The Tamura Nei evolution model was used to calculate the genetic distances between 39 whole genomes and the genomic sequences of the viruses detected in this study (in bold). Each segment was compared individually. The color scheme indicates distance and ranges from white to red, with white indicating zero and red 0.06. For genomic segment 2 (PB1), only fully assembled sequences (3 in total) were included in the alignments (upper heatmap).

Immunohistochemistry for influenza virus nucleoprotein.

All 8 tonsillar tissues positive for IAV by RT-PCR were also positive for IAV nucleoprotein (NP) by IHC. All 14 control tissues that tested negative for IAV by RT-PCR tested negative for NP by IHC. In IAV-positive tissues, NP staining was restricted to epithelial cells in two distinct palatine tonsils (Fig. 4). Sections from the remaining 6 tissue samples showed more disperse NP-positive (NP⁺) cells, both in the epithelium of crypts and throughout the lymphoid tissue, both in follicles and inter-follicular areas (Fig. 4). This indicates that, in addition to IAV RNA, a virus structural protein is being produced in tonsillar tissues from children without ARI symptoms. Furthermore, viral protein production was not only located on the epithelium but also in lymphoid cells, suggesting that IAV infection may be productive in tonsil tissues from children without clinical manifestation of influenza.

FIG 4.

Immunohistochemistry for IAV nucleoprotein (NP) in representative tissue sections from human tonsils. (A) IAV-negative tissue as negative control. (B and C) Epithelial staining for IAV NP. (D) NP staining in lymphoid tissue. Dashed line indicates boundary between lymphoid follicle and interfollicular area. (E and F) Insets from panel D with arrowheads indicating positively stained cells in both compartments of tonsils. Scale bars, 25 μm.

Viable IAV can be rescued from human tonsil tissues.

To confirm whether the presence of IAV RNA and NP viral antigen in tonsillar tissues correlated with the presence of infectious virus particles, lysates of 8 tonsil tissues from 7 patients who were positive for IAV by real-time RT-PCR were inoculated onto MDCK cell monolayers. After 2 blind passages, intense cytopathic effect (CPE) was noted on the 4th day postinoculation in one tissue sample. Immunofluorescence (IF) of MDCK cells with antibodies to IAV NP confirmed isolation of infectious IAV (Fig. 5). Sequencing of segments 6 and 7 confirmed that the isolated virus was A/H1N1pdm09, with 95.7 and 98% nucleotide identity, respectively, to the sequences obtained directly from the tonsil tissue (Fig. 6). Moreover, this IAV isolate was obtained from an adenoid tissue sample (number 448ADN) that yielded a complete genome sequence by NGS.

FIG 5.

Isolation of infectious IAV from human tonsils. (A) Mock-inoculated cells. (B) Cytopathic effect induced in MDCK cells 4 days postinoculation with tonsillar tissue lysate. (C and D) Evans blue counterstaining and NP IF staining on mock-infected cells. (E and F) Evans blue counterstaining and IF staining for NP on MDCK cells 4 days postinoculation with tissue lysate. Scale bars, 25 μm.

FIG 6.

Nucleotide alignments of the NA and M genome segments sequenced by NGS from the IAV isolate obtained from adenoid tissue sample number 448ADN. Sequences of the NA (A) and M (B) RNA segments were aligned to the homologous sequences obtained directly from the tissue sample.

Immunophenotyping of tonsillar cells susceptible to IAV.

Lymphoid follicles are composed mainly of subsets of B cells, T cells, macrophages, and dendritic cells (29, 30). To determine the immunophenotypes and the frequencies of cells expressing IAV NP in naturally infected tonsils, cell suspensions prepared from dispersed tissues were analyzed by flow cytometry. Individual experiments were performed in three different tissue samples positive for IAV by real-time RT-PCR. Cell suspensions of tonsils from three children negative for IAV by real-time RT-PCR were used as negative controls (3 palatine tonsils and 3 adenoids) (Fig. 7). The cell immunophenotypes most frequently associated with detection of IAV NP in dispersed tonsillar tissues were CD20⁺ B lymphocytes (mean, 9.33% NP⁺ from all CD20⁺ cells in palatine tonsils and 8.63% from all CD20⁺ cells in adenoids), CD3⁺ CD8⁺ T lymphocytes (mean, 5.16% in palatine tonsils and 3.89% in adenoids), and CD11c⁺ cells (mean, 0.56% in palatine tonsils and 1.44% in adenoids) (Fig. 7). NP staining in T CD4⁺ lymphocytes was not significantly different from background (Fig. 7). These findings indicate that the immune cell subpopulations in human tonsillar tissues containing IAV antigen were mainly B and CD8⁺ T lymphocytes, with minimal signal in CD11c⁺ cells.

FIG 7.

Immunophenotyping of IAV NP-positive cells. (A) Flow cytometry of naturally infected dispersed human tonsils indicating the gating strategies used to assess the frequencies of IAV NP detection in CD20⁺ B, CD3⁺ CD8⁺ T, and CD3⁺ CD4⁺ T lymphocytes and CD11c⁺ cells. Cell suspensions prepared from IAV-negative tonsils were used as negative controls. (B) Frequencies in percentages (left) and cell count (right) of cells positive for IAV NP in palatine tonsils and adenoids. The proportion of IAV NP-positive CD20⁺ cells was significantly higher than that of CD8⁺ T cells and CD11c⁺ cells. Experiments were performed in triplicate using tonsillar tissues from three different patients, with acquisition of 10⁵ events. Multiple comparisons were made and statistical significance was determined using the Holm-Sidak method. *, P ≤ 0.05 was considered significant. (C, F, I, L, O) Immunohistochemistry for IAV NP with positive cells pseudocolored in green and hematoxylin counterstaining. (D, G, J, M, P) Same tissue sections as panels C, F, I, L, O, stained for CD8a, CD20, CD4, CD11c, and cytokeratin after erasing the previous staining, pseudocolored in red. (E, H, K, N, Q) Merged layers showing overlay of previous antigens in yellow. Arrowheads indicate cells expressing IAV NP. Scale bars in panels C, D, and E are 100 μm. Scale bars in the remaining panels are 50 μm.

To assess the spatial distribution of NP-expressing cells in the lymphoid tissue, sequential immunoperoxidase labeling and erasing (SIMPLE) (31) was used to stain for IAV NP and subsequently for the cell surface markers CD3, CD4, CD8, CD20, CD11c, or cytokeratin on the very same tissue sections. The overlaid composite images revealed individual cells simultaneously expressing IAV NP/CD8 (T cells), and IAV NP/CD20 (B cells) (Fig. 7), reinforcing the flow cytometry results. No substantial IAV NP signal was detected on cells expressing CD4 or CD11c surface markers (Fig. 7). Apart from lymphoid cells, epithelial cells expressing cytokeratin were also positive for IAV NP (Fig. 7). NP-expressing B lymphocytes were found both in follicular and interfollicular areas (Fig. 8), while NP-expressing T CD8⁺ lymphocytes were found most frequently in interfollicular areas (Fig. 8).

FIG 8.

Detailed localization of B and CD8⁺ T cells stained for IAV NP in palatine tonsil. (A) Human palatine tonsil tissue section stained for IAV NP with positive cells pseudocolored in green. (B and E) Staining for CD20 (B) and CD8 (E), with positive cells pseudocolored in red. (C and F) Overlay of the two previously stained sections, showing that B lymphocytes expressing IAV NP were present both in interfollicular areas (C′) and lymphoid follicles (C′′), while NP-positive CD8⁺ T cells were found mainly in interfollicular regions (F′) and not in lymphoid follicles (F′′). Arrowheads indicate IAV NP-positive cells. Scale bars, 100 μm.

Nuclear localization of IAV NP in B and T lymphocytes.

The presence of nuclear localization sequences in IAV NP is a key factor for its nuclear import (32) and the presence of NP in the nuclei of infected cells is indicative of active infection. To achieve this level of magnitude, IF was performed in Carnoy’s-fixed, paraffin-embedded (CFPE) tissue sections from palatine tonsils and adenoids and analyzed by confocal microscopy to assess the intracellular localization of NP in B and T CD8 lymphocytes. IAV NP was detected in both CD20⁺ and CD8⁺ cells (Fig. 9). Although present in the cytoplasm, NP (red channel) was also detected in the nuclei (blue channel) of both cell subsets (Fig. 9G and H). When analyzed individually, nuclei of infected cells showed considerable percentages of colocalization between red and blue channels (Fig. 9I), indicating that NP protein is present in the nuclei of infected B and T CD8⁺ lymphocytes. Our data set suggests that beyond detection of viral RNA and NP, infection of lymphoid cells by IAV is active in human tonsils.

FIG 9.

IAV NP in the nuclei of infected cells in human tonsils. Representative confocal microscopy of tissue sections stained with DAPI (A and D) and antibodies to CD20 (green) (B), CD8 (green) (E), and IAV NP (red) (C and F). (G and H) Merged channels indicating CD20⁺ and CD8⁺ cells expressing IAV NP. Insets are shown in the upper right corner of each image (g′ and h′). (gx′, gy′, hx′, hy′) Orthogonal views of z-stack images (total of 30 layers) show NP in nuclei of CD20⁺ and CD8⁺ cells. (I) Manders coefficients. Colocalization thresholds were calculated from analysis of NP/nuclei (red/blue channels) in z-stack images of 5 individual nuclei for each condition. Scale bars, 10 μm. Z stacks width, 0.49 μm.

DISCUSSION

Influenza virus, a most relevant pathogen worldwide, frequently causes severe human disease but may cause asymptomatic infections, depending on the interactions of multiple virus and host factors (33). Most studies that have estimated rates of asymptomatic influenza infection have been based on RNA detection by RT-PCR in respiratory secretions (34). Attempts to determine the tissue sources of viral RNA have not been reported. The present study addressed that issue, showing that a proportion of hypertrophic tonsils are sites of IAV infection in asymptomatic children, proposing that not only epithelial cells but also tonsillar lymphoid cells, mainly B and CD8⁺ T lymphocytes, are probable sources of infectious IAV in secretions. Considering the high prevalence of tonsillar hypertrophy, reflected in the high frequency of tonsillectomy worldwide (35–38), the present study is relevant for influenza epidemiology and pathogenesis.

The present finding of IAV RNA detected by real-time RT-PCR in tonsil tissues from children with tonsillar hypertrophy in the absence of symptoms of ARI reiterates previous reports by our group (11, 39). This real-time RT-PCR assay is sensitive enough to detect 10 copies of influenza RNA (11), and although rates of asymptomatic IAV infection vary considerably depending on study design and detection methods (40), the 6.8% IAV detection rate is similar to what has been reported by others in asymptomatic individuals tested by real-time RT-PCR (40, 41). Conversely, the in situ detection of IAV nucleoprotein in all tissues positive by real-time RT-PCR is an advance. When analyzed by confocal microscopy, NP subcellular localization indicated multiple cells in different stages of infection. In addition, NP colocalized with the nuclei of infected lymphocytes, which strongly suggests active infection in those cells as opposed to the mere presence of remnant viral RNA and NP.

It has been known that IAV can be detected in lymphoid tissues. For example, mouse pulmonary dendritic cells (DCs) are infected by IAV and deliver viable virus to draining lymph nodes (9, 10). Also, A/H1N1pdm09 virus has been shown in diverse tissues by IHC and in situ hybridization (ISH) in human fatal cases (5), where antigenome was detected in association with undetermined small mononuclear lymphoid cells by ISH in mediastinal lymph nodes. However, the present study is the first to show the presence of IAV structural protein in naturally infected human tonsils, both in epithelial and lymphomononuclear cells, in lymphoid follicles, and interfollicular areas. Thus, comparably to mouse lung-draining lymph nodes, human tonsils may be secondary lymphoid sites of infection after IAV replication in upper airway epithelia.

Apart from epithelial, endothelial cells, and fibroblasts, palatine tonsils and adenoids are composed mainly by resident memory T and B cells, naive lymphocytes, follicular DCs, macrophages, and migratory antigen-presenting cells (APCs) originated in peripheral sites of antigen recognition (29). This study showed that IAV can infect numerous cells of multiple immune phenotypes in tonsils, especially in lymphoid follicles, which are important sites of lymphocyte maturation where follicular and migratory dendritic cells, and other APCs, bring antigens to naive lymphocytes, resulting in the mounting of an adaptive immune response (12).

The tonsillar lymphomononuclear cells that were more frequently found infected by IAV were CD20⁺ B lymphocytes and CD8⁺ T lymphocytes and, in smaller proportion, CD11c⁺ cells. Also, CD20⁺ B lymphocytes were by far the most abundant infected cells. This is remarkable since B cells are key effectors during IAV infections, producing neutralizing antibodies to IAV. It is noteworthy that IAV NP was detected also in cytokeratin-expressing epithelial cells that may contribute for the production of infectious viral progeny in tonsils, a finding that is in keeping with the previously reported detection of IAV in nasopharyngeal secretions from children with tonsillar hypertrophy (11).

While the detection of IAV antigen in CD11c⁺ APCs is not surprising considering their role in antigen transportation and processing, the presence of IAV in CD8⁺ T cells is rather surprising since these cells, once activated by APCs, undergo expansion, activation, and differentiation in cytotoxic T lymphocytes (CTLs), crucial effectors of the adaptive immune response against many viruses (42). Considering the essential antiviral roles of B and CD8⁺ T lymphocytes, it is reasonable to speculate that infection of these cells by IAV must have impact on the mounting of immune responses that depend on the recruitment of IAV-infected cells in secondary lymphoid organs. Upon prolonged antigen exposure and upregulation of inhibitory receptors, T-cell exhaustion may occur (43). While mechanisms of CD8⁺ T lymphocyte infection by IAV are unknown, some level of CD8⁺ T cell impairment may be in place in IAV-infected secondary lymphoid tissues, allowing for a more prolonged virus stay.

It is noteworthy that only A/H1N1pdm09 was found in tonsils, and neither A/H3N2 nor IBV were detected. This could be due to sampling site selection bias, since the distribution of infection in the tonsil may not be homogeneous and, therefore, may be affected by serendipity. Therefore, it may also indicate that the tropism of influenza viruses for tonsil lymphoid cells is lineage specific, which is further supported by the fact that the most frequent influenza viruses detected in Brazil in 2014 were by far H3N2 and B, whereas pH1N1 occurred only modestly (23).

Another significant finding of the present study was the striking phylogenetic segregation of IAV genome sequences obtained from tonsils into clusters well separated from the rest of the sequences in phylogenetic trees prepared for all of the eight IAV RNA segments. The phylogenetic distances indicate that this is a very unique group of sequences, suggesting independent evolution from the other A/H1N1pdm09 viruses circulating in Brazil in the same time period, and that the proneness of these viruses to infect tonsillar lymphoid cells is favored by some genomic sequence features. Despite these unique genomic features in comparison with other circulating strains, we cannot exclude the possibility that the patients were infected with IAV variants already carrying the genetic mutations observed in this study.

Moreover, IAV virus sequences from palatine tonsils and adenoids had a considerable number of exclusive nonsynonymous mutations that were not found in viruses circulating before and after 2014. Keeping in mind that differences in structural barriers between tissues may positively select for viruses that can spread more easily, and thus generate tissue-specific subpopulations (44, 45), it is reasonable to hypothesize that the mutations seen in tonsil virus sequences, common to seven different patients, may have been selected for their effects on the virus fitness in lymphoid cells. Alternatively, this selection of mutants could be due simply to parallel evolution, lacking actual tissue specificity. IAVs with virtually identical RNA sequences were detected simultaneously in palatine tonsil and adenoid from one child, raising the possibility that IAV variants in anatomically different lymphoid tissues might be lymphoid tissue specific. However, that finding was present in only one case, and therefore, it should be interpreted with caution.

The lower coverages of internal regions in the polymerase genes are indicative of the presence of DI-RNAs as reported by previous studies (21, 46). This was observed in viruses sequenced from tonsils from different individuals, suggesting a common trait for influenza A virus infecting this type of tissue. Virus DI-RNAs have been reported in different viral infections, including IAV (19, 21). IAV DI-RNAs arise preferentially for segments 1, 2, and 3, although all 8 segments can generate DIs (19, 47). The main features of IAV DI-RNAs are internal regions significantly shortened by large deletions, with maintenance of the 3′ and 5′ extremities (48). Under certain conditions, DI-RNAs rather than full-length genome segments are preferentially incorporated into new viral particles, which consequently cannot produce viable progeny, except with the help of a virus with fully functional polymerase (49). Interestingly, it has been shown that cloned DI-RNA of A/PR/8/34 (H1N1) segment 1 was replicated and packaged with the help of the cognate virus whose replication is inhibited by the DI-RNA in respiratory cells in vitro as well as in animal models (50).

The full implications of DI-RNAs in natural IAV infections are not known, but it is reasonable to hypothesize that they could play a role in maintaining low levels of replication of infectious IAV in nonepithelial sites of infection, such as secondary lymphoid organs. Accordingly, in such noncanonical IAV host cells as B and CD8⁺ T lymphocytes, the low abundance of replicating viruses with complete genome may be the reason why infectious IAV was recovered from only a single IAV-positive tissue. Unfortunately, nasopharyngeal swabs were not collected at the time of surgery in parallel with the tissue sampling, and thus, it was not possible to verify shedding of infectious IAV from hypertrophic tonsils into respiratory secretions.

In summary, children without ARI symptoms can harbor IAV in hypertrophic palatine tonsils and adenoids, with production of infectious virus, and hence they can be silent sources of IAV for human populations. Further studies will be needed to assess the impact of IAV infection on the functions of naturally infected human lymphoid cells.

MATERIALS AND METHODS

Ethics statement.

This study was conducted in compliance with the Ethics Review Committee of the University of Sao Paulo Clinical Hospital (protocol 10466/2008). All parents or guardians signed an informed consent form before enrollment.

Patients and sample processing.

This cross-sectional prospective study included 3- to 13-year-old children who underwent adenotonsillectomy for the treatment of tonsillar hypertrophy or recurrent tonsillitis at the State Hospital of Ribeirao Preto, Brazil. All patients had enlarged adenoids occupying 70% of the oropharyngeal airways and tonsils of grades III and IV according to the Paradise criteria (51). Exclusion criteria were the presence of ARI symptoms and antibiotic treatment in the 30 days before surgery. Tissue fragments obtained from adenoids and palatine tonsils removed during surgery were placed in RPMI 1640 supplemented with 10% fetal bovine serum (FBS) and 15% antibiotic-antimycotic solution (penicillin-streptomycin, 10,000 U/ml, and amphotericin B, 200 μg/ml) (Gibco, Gran Island, NY, USA) and kept on ice until arrival in the laboratory. Tissue specimens were washed in RPMI to remove debris and blood and then split in four aliquots. Approximately 120 mg of tissue was homogenized in TRIzol (Invitrogen, Carlsbad, CA, USA) using a TissueLyser LT (Qiagen). Another portion was cut in pieces of approximately 0.25 cm³, treated with collagenase (100 U/ml) and dispase (0.6 U/ml), and passed through cell strainer nylon mesh (Falcon) to obtain a cell suspension. Another piece of tissue was fixed in Carnoy’s fixative and embedded in paraffin for histology, and the last piece of tissue was kept in viral transport medium consisting of minimum essential medium (MEM) with 20% FBS and 15% glycerol. Except for paraffin-embedded tissues, all samples were stored at −70°C. Unfortunately, neither nasopharyngeal swabs nor serum samples were collected from the patients at the time of surgery.

Detection of influenza viruses.

Total RNA was extracted from homogenized tissue fragments with TRIzol reagent according to the manufacturer’s instructions. Reverse transcription was done with a high-capacity reverse cDNA transcription kit (Life Technologies, Carlsbad, CA, USA) using 1 to 2 ng of extracted RNA and random hexamers following the manufacturer’s instructions. Newly synthesized cDNA was subjected to TaqMan real-time PCR (quantitative PCR [qPCR]) for IAV, IBV, and the housekeeping gene β-actin in separate tubes using specific primers and probes (Table 2). The qPCR assays were done in a final volume of 10 μl with 3 μl of cDNA, 10 μM forward and reverse primers, 5 μM specific probe and TaqMan universal PCR master mix (Applied Biosystems, Foster City, CA, USA) in a StepOnePlus real-time PCR system (Applied Biosystems) with the following parameters: 50°C for 2 min, 95°C for 10 min, and 45 cycles of 95°C for 15 s, 55°C for 30 s, and 60°C for 1 min. All experimental steps described were done in separate rooms to avoid contamination, and each plate contained positive and negative controls. Plasmids containing PCR products defined by each primer set, as well as virus stocks, were used as positive controls, and ultrapure distilled water and uninfected MDCK cell supernatant were used as negative controls.

TABLE 2.

Primers and probes used for RT-qPCR screening

Virus or housekeeping gene	Primer	Sequence	Target	Reference
Influenza	InfA for	GACCRATCCTGTCACCTCTGAC	M	61
	InfArev	AGGGCATTYTGGACAAAKCGTCTA	M
	InfAprobe	Fam-TGCAGTCCTCGCTCACTGGGCACG-BHQ1	M
	INFB-1	AAATACGGTGGATTAAATAAAAGCAA	HA	62
	INFB-2	CCAGCAATAGCTCCGAAGAAA	HA
	INFB-probe	Vic-CACCCATATTGGGCAATTTCCTATGGC-Tamra	HA
β-actin	β-actin F	CCCAGCCATGTACGTTGCTA	β-actin	63
	β-actin R	TCACCGGAGTCCATCACGAT	β-actin
	β-actin-probe	Fam-ACGCCTCTGGCCGTACCACTGG-Tamra	β-actin

Multisegment RT-PCR and flu genome sequencing.

Total RNA from samples that tested positive for IAV by RT-PCR were used to synthesize cDNA and amplify the IAV genome using SuperScript one-step III high-fidelity RT-PCR kit (Life Technologies, Carlsbad, CA, USA). Reactions were done according to the manufacturer’s guidelines using primers Opti1-F1 (5′-GTTACGCGCCAGCAAAAGCAGG-3′), Opti1-F2 (5′-GTTACGCGCCAGCGAAAGCAGG-3′), and Opti1-R1 (5′-GTTACGCGCCAGTAGAAACAAGG-3′). The next-generation sequencing (NGS) was performed using the platform MiSeq (Illumina) with library preparation with Nextera XT and the sequencing cartridge MiSeq v2 for 300 cycles. Genome assembly was performed using a pipeline previously described (2). Briefly, low-quality sequences and adapters were removed by Cutadapt (52) from FASTQ files storing paired-end reads 150 nucleotides (nt) long. An initial assembly was done using the Inchworm component of Trinity (53), and viral contigs bearing internal deletions were identified by BLAT (54) mapping against nonredundant Influenza Research Database (IRD) reference sequences. Afterward, breakpoint-spanning kmers from the assembly graph were removed by repeating the Inchworm assembly, and the resulting contigs were then oriented and trimmed to remove low-coverage ends and any extraneous sequences beyond the conserved IAV termini. To improve contiguity, the CAP3 (55) assembler was used, and contigs from the same segment were merged if their ends overlapped by at least 25 nt. Finally, assembly contigs and contiguity were assessed for all segments by mapping sequence reads back to the final assembly using Burrows-Wheeler alignment (56). Consensus sequences for each segment in every sample were generated, and then the tool BLAST (22) was used to confirm the similarity with other influenza genomes.

Phylogenetic analysis.

Sequences of whole genomes of influenza A/H1N1pdm09 deposited between 2009 and 2017 were obtained from GenBank and GISAID as follows: 94 of the HA segment; 73 of NA; 39 of NP, NS, and PA; 40 of PB1 and PB2; and 43 of M (see Table S1 in the supplemental material). The sequences were aligned with the software Seaview 4.6.2 (57), and temporal phylogeny was carried out using the Bayesian Markov chain Monte Carlo (MCMC) approach as implemented in BEAST 2.4.8 (58). The data were analyzed under a coalescent exponential-growth model as a prior on the tree, the GTR + Γ model of nucleotide substitution, and an uncorrelated exponential strict clock model. The analysis was run for 50 million states with sampling every 1,000 states to ensure an adequate sample size of all analysis parameters. The data obtained from BEAST were assessed using the effective sampling size (ESS) after a 10% burn-in using Tracer software v1.6. The consensus tree with the maximum product of posterior probabilities (maximum clade credibility tree) for analyzing the MCMC data set was annotated by TreeAnnotator v2.4.8 with a burning of 10% and visualized using MEGA X.

Genetic analysis.

All RNA sequences obtained by NGS from the Ribeirao Preto viruses were used to create an alignment with 39 whole genomes available in GenBank and GISAID. The analysis was made using R program, the Ape package with the Tamura and Nei evolution model (59), to calculate the genetic distance between Ribeirao Preto viruses and reference strains for each segment individually. Finally, the genetic distance values were plotted into a heatmap. Another alignment was made using the software Seaview 4.6.2 (57) to evaluate the predicted amino acid sequences from the Ribeirao Preto viruses. The variant position in the amino acid sequences was obtained using MEGA 7 (60) in the alignment with the sequences from Ribeirao Preto and the BLAST closest hits for each specific protein.

Immunohistochemistry and sequential immunoperoxidase labeling and erasing.

Carnoy’s-fixed, paraffin-embedded (CFPE) tissue sections of 3 μm were deparaffinized, rehydrated, and subjected to antigen retrieval by treatment with citrate buffer at 95°C for 15 min. After endogenous peroxidase blocking with 10% H₂O₂ and incubation with 5% horse serum in phosphate-buffered saline (PBS)-bovine serum albumin (BSA) 0.1%, tissue sections were incubated with primary mouse anti-influenza A NP monoclonal antibody (Chemicon, Merck/Millipore) for 1 h. Primary antibodies were detected with secondary biotinylated horse anti-mouse IgG antibody (Vector laboratories, Burlingame, CA, USA), and color development was done with the AEC kit (Vector laboratories, Burlingame, CA, USA), followed by counterstaining with Harris’s hematoxylin. Tissue sections were scanned with ScanScope VS120 (Olympus Life Sciences, Tokyo, Japan) in bright field using ×400 magnification. After that, slides were dehydrated, rehydrated, and the staining was erased as described previously (31). Sequential staining and erasing were then performed on the same tissue sections slices for CD3, CD4, CD8, CD11c, CD20, and cytokeratin (Abcam, Cambridge, UK) cell surface markers in order to determine the phenotypes of IAV-infected cells. Slides were completely scanned after each round of staining, and the sequential stainings were pseudocolored and overlaid on the first hematoxylin counterstain image using ImageJ v1.50b (NIH, USA) and Adobe Photoshop CS5 software (Adobe Systems, San Jose, CA, USA). Pseudocolored layers were overlaid using “normal” blending mode, and yellow color represented superposition of red and green colors. MDCK cells infected with A/Texas/36/91 (H1N1) were fixed in Carnoy’s paraffin-embedded and used as positive controls. Tonsil tissue sections negative for IAV by RT-PCR were used as negative controls.

Flow cytometry of dissociated lymphomononuclear tonsil cells.

Tonsil lymphomononuclear cells (TLMC) were separated from frozen cell suspensions by Ficoll-Paque Plus (GE Healthcare, Little Chalfont, UK). After isolation of the buffy coat layer, cells were washed twice and kept in RPMI medium supplemented with 10% FBS at 37°C with 5% CO₂. TLMC purified from IAV-positive tissues were initially stained for 40 min at 4°C using CD3-CF594, CD4-Cy5.5, CD8a-FITC, and CD20-FITC antibodies (BD Pharmingen) and then washed, permeabilized with 0.1% saponin, and fixed with 4% buffered paraformaldehyde. Intracellular staining was done with mouse anti-influenza A NP primary antibody for 1 h, followed by anti-mouse IgG-PE secondary antibody (BD Pharmingen) for 30 min. Cells were washed and run on a Guava easyCyte flow cytometer (Merck/Millipore, Darmstadt, Germany), with data analysis performed with Guava Easy Cytosoft software (Merck/Millipore, Darmstadt, Germany). Cell suspensions obtained from IAV RT-PCR-negative tonsils (3 palatine tonsils and 3 adenoids) were used to confirm the specificity of the IAV NP antibody used. Matching isotype controls for the surface antibodies were used as negative controls. A third negative control with only the IgG-PE secondary antibody was used for calibration of PE acquisition.

Cells and virus isolation.

Madin-Darby canine kidney (MDCK) cells were grown in MEM with 10% fetal bovine serum and 1% antibiotic-antimycotic solution and kept at 37°C with 5% CO₂. Lysates of IAV-positive tissues were fast frozen and thawed, clarified by centrifugation at 12,000 × g, and supernatants were diluted 1:1 in serum-free MEM containing 2 μg/ml trypsin and then inoculated onto 80% confluent MDCK cells. Infected cell cultures were kept at 37°C with 5% CO₂ and monitored daily for cytopathic effect (CPE). Three additional blind passages were done if no CPE was detected in 7 days, and regardless of the presence of CPE, cells were tested after each round of passage by immunofluorescence. All experiments were carried out in biosafety level 2 (BSL-2).

Immunofluorescence staining for influenza A nucleoprotein.

After fixation with acetone and permeabilization with PBS-BSA-saponin solution, infected MDCK cells were incubated with primary mouse anti-influenza A NP monoclonal antibody (Chemicon, Merck/Millipore) for 1 h. Primary antibody was detected using secondary anti-mouse-Alexa 488 antibody (Abcam).

For analysis of NP nuclear localization, CFPE tissue sections were deparaffinized, rehydrated, and subjected to antigen retrieval by treatment with citrate buffer at 95°C for 15 min. After blocking/permeabilization with 5% horse serum in PBS-BSA-saponin solution, tissue sections were incubated overnight with one of the following surface primary antibodies, rabbit anti-CD8a or rabbit anti-CD20 (Abcam), and pooled with mouse anti-influenza A NP monoclonal antibody (Abcam) and 4′,6-diamidino-2-phenylindole (DAPI) (Abcam). Primary antibodies were then detected using secondary horse anti-rabbit Alexa 488 and horse anti-mouse Alexa 647 antibodies (Abcam).

Confocal image and processing.

Tissue sections were stained as mentioned above, and images were obtained with confocal fluorescence microscopy on a Leica TCS SP8 confocal laser scanning microscope (Leica Microsystems, Wetzlar, Germany). For colocalization studies of IAV nucleoprotein with nuclei, analysis was performed on Z series images by quantification of Manders colocalization coefficients M1/M2 using Fiji software. The nuclei of CD20⁺ NP⁺ or CD8⁺ NP⁺ cells were selected using the free-hand selection tool. The colocalization threshold and the orthogonal view was carried out in the 30 z-stacks (0.49 μm each) of five cells in each condition (adenoid CD20 and CD8, palatine tonsil CD20 and CD8). M1 is the percentage of above-background pixels in the red channel that overlap above-background pixels in the blue channel and the inverse for M2. Staining of the nuclei was considered the blue channel and the IAV NP the red channel.

Statistical analysis.

Patient stratification, frequencies, and descriptive statistics were done using IBM SPSS Statistics, version 20. One-way analysis of variance (ANOVA), two-way ANOVA, and Holm-Sidak post-tests were performed using GraphPad Prism Software, version 6.0. P values of <0.05 were considered significant.

Data availability.

Nucleotide genomic sequences from all tonsil-derived viruses have been deposited at the GISAID EpiFlu Database under the following accession numbers: EPI1319701, EPI1319702, EPI1319703, EPI1319704, EPI1319705, EPI1319706, EPI1319707, EPI1319708, EPI1319709, EPI1319710, EPI1319711, EPI1319712, EPI1319713, EPI1319714, EPI1322831, EPI1322832, EPI1322833, EPI1322834, EPI1322835, EPI1322836, EPI1322837, EPI1322845, EPI1322846, EPI1322847, EPI1322848, EPI1322849, EPI1322850, EPI1322851, EPI1322838, EPI1322839, EPI1322840, EPI1322841, EPI1322842, EPI1322843, EPI1322844, EPI1322852, EPI1322853, EPI1322854, EPI1322855, EPI1322856, EPI1322857, EPI1322858, EPI1322859, EPI1322860, EPI1322861, EPI1322862, EPI1322863, EPI1322864, EPI1322865, EPI1322866, EPI1322867, EPI1322868, EPI1322869, EPI1322870, EPI1322871, EPI1322872, EPI1322873, EPI1322874, EPI1322875.