Spatial transcriptomic profiling of Respiratory
Syncytial Virus (RSV) infection
Joseph Bergenstråhle
1, Alexandra Dondalska
2, Lovisa Franzén
1, Sandra
Axberg Pålsson
2, Alexandros Sountoulidis
2, Laura Sedano
3, Marie-Anne
Rameix-Welti
4, Jean-Francois Eleouet
3, Ronan Le Goffic
3, Marie Galloux
3,
Christos Samakovlis
2, Joakim Lundeberg
1, and Anna-Lena Spetz
21Department of Gene Technology, Royal Institute of Technology, Sweden
2Department of Molecular Biosciences, The Wenner-Gren Institute, Stockholm University, Sweden
3INRA UR892 Unité de Virologie et Immunologie Moléculaires, France
46INSERM U1173 Infectionet Inflammation Chronique (2IC), France
Abstract
Despite the fact that the human Respiratory Syncytial Virus (RSV) was first discov-ered back in 1956, it remains one of the leading causes of morbidity and mortality in young children. Transcriptome-wide spatially resolved transcriptomics is a technology under rapid development that introduces a new modality for exploratory examination of cellular behavior. With this modality, we examine how RSV infection changes the local cellular environment in the lung by infecting mice with RSV and comparing it to control samples four days after infection. We find viral presence in all compart-ments of the tissue, well-defined induced tertiary lymphoid tissue within some of the samples, compartmentalized infiltration of innate immune cells, as well as functional enrichment of airway epithelial repair pathways.
1
Introduction
The human Respiratory Syncytial Virus (RSV) is an enveloped, single-stranded, negative-sense RNA virus of the Pneumoviridae family. It is the most common viral pathogen found in children diagnosed with acute lower respiratory infection and is a leading cause of mortality within this age group [1]. It exerts a disease burden comparable to that of non-pandemic Influenza A in elderly and immunocompromised individuals [2], [3]. To date, there are still no RSV vaccines available, despite a long history of continued efforts since the virus was originally isolated and characterized in 1956 [4]. Furthermore, there is still a lack of knowledge about the exact routes of viral entry and reasons for the diverse outcomes of infected individuals.
Despite several reasons that could explain the slow progress of vaccine development, including suboptimal animal models, mice models are extensively used and led to the initial development of an approved humanised mAb. Here, we report a spatial transcriptomics (ST) profiling of RSV infected BALB/c mouse lungs, four days after intranasal inocula-tion, which has been suggested as the preferred way to induce robust pathology [6]. By polyadenylated- (polyA-) based in-situ capture, probing of the entire mRNA portion of the transcriptome is possible without choosing targets a priori. This type of approach for untargeted explorative analysis has been extensively used to characterize the tumor microen-vironment [7], neuropathologic molecular signatures [8], and developmental processes [9] among many other application areas.
By comparing to non-infected control mice, we characterize infection-activated gene expression profiles within clustered cell types. Furthermore, we use an external annotated data set, Tabula Muris [10], to map scRNAseq profiles onto our spatially resolved data grid. While the method for spatially resolved whole-transcriptome data we used lacks the resolution of single-cells, the integration between ST and scRNAseq allows us to perform an automatic mapping of where cells of certain transcriptomic profiles are most likely to be located. We find spatial compartmentalization of different immune cells, as well as localized behaviour of acute phase responses. Finally, we also profile the RSV genes themself, to get a better understanding of the spread of the viral infection across the tissue.
2
Results
2.1
Experimental setup and global gene expression differences
Mice were challenged with hRSV or PBS intranasally and euthanized four days later (figure
1A). For each animal, two sections were cross sectioned for ST analysis. After filtering
gene counts (methods), a total of 19126 genes across 43336 capture areas were evaluated. After batch correction (methods), clustering of the individual capture-areas revealed several clusters unique to the non-infected and infected conditions (figure 1B). Global differen-tial gene expression analysis showed several highly upregulated genes involved in immune responses and extracellular matrix production in response to infection (figure 1B, supple-mentary table 1). Since capture areas contain cell mixtures, we use marker genes to compute enrichment scores for cell types (figure 1C-D, supplementary figure S1). As expected, the RSV-infected samples demonstrated enrichment of several immune cell types not found in the PBS-treated samples. While the major clusters contain mixtures of cell types, we also identify clusters with a more homogenous composition. For example, the airway epithelium is associated with several clusters (clusters 5, 6, 7, and 9) but has an uneven distribution between conditions (section 2.3.2). Other clusters specific to RSV include alveolar tissue with markers for acute immune response and innate immune cell infiltration (clusters 1 and 2, section 2.3.1), and tertiary lymphoid organs exclusively found in the RSV samples (cluster 18, section 2.3.3).
A B Time (d) 0 4 hRSV (n=2) PBS (n=2) ST (n=16) FISH (n=2) Intranasal challange Euthanization
C PBS RSV D 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
B cell−25% B cell−50% B cell−75%
ciliated columnar cell−25%
ciliated columnar −50%
ciliated columnar cell −75%
classical monocyte−25% classical monocyte−50% classical monocyte−75% epithelial cell of lung−25% epithelial cell of lung−50% epithelial cell of lung−75%
leu kocyte−25% leu kocyte−50% leu kocyte−75%
lung endothelial cell−25% lung endothelial cell−50% lung endothelial cell−75% monocyte−25% monocyte−50% monocyte−75%
m yeloid cell−25% m yeloid cell−50% m yeloid cell−75% natu
ral killer cell−25%
natu
ral killer cell−50%
natu
ral killer cell−75% stromal cell−25% stromal cell−50% stromal cell−75% T cell−25% T cell−50% T cell−75%
Cluster 0 4 17 1 8 12 5 10 11 3 14 19 6 13 7 18 15 16 2 9 0 4 17 1 8 12 5 10 11 3 14 19 6 13 7 18 15 16 2 9 Scaled pred. score E 0.2 0.4 0.6 0.8 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 0.1 0.2 0.3 0.4 0.5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Ciliated B cell Eef1a1Reg3g Ftl1 Lyz2 Nppa−1Chil1 H2−Ea−psSaa3 B2m Lcn2 Ifi27l2aRetnla Apoe Scgb3a2Clu C3 Lgals3bpLy6e MgpMt1 Irf7 Mfap4C1qa Alas1 Lgmn Erdr1−1Hp Ifitm3 H3f3b Col3a1C1ra Lyz1 Ccl6 Klf13Ctsb Cd74 Col1a2Wfdc2 Ctss PsapMt2 Col4a1Ly6a Vim Serping1Clec7a Cybb C1qc Oasl2 Col1a1Ctsh Aldh1a1Lgals3 Ly6i Ctsc Ccn2Calu Tubb4bH2afz Tmsb10Ces1d Rgcc Ltbp4 Hilpda AhnakCalcrl Hba−a2Inmt Malat1Scd1 Klf2 Hbb−b1 Scgb1a1 −2 −1 0 1 2 Expr. PBS RSV Cluster Cluster UMAP 1 UM AP 2 UMAP 1
Figure 1: A) Experimental setup. B) Global differential expression between non-infected (PBS) and infected (RSV). C) UMAP embedding of gene expression in capture areas split by condition. Labels and colors indicate cluster membership. D) Enrichment scores for ciliated cells and B-cells for each cluster. E) Annotated scRNAseq data mapped onto the spatially resolved data. Each capture area is given a prediction score for each cell type. For each cluster and cell type, quantiles across the capture areas within the cluster are calculated. Prediction scores are normalized by column-wise scaling.
2.2
Integration of single-cell RNA and spatial transcriptomics profiles
The resolution of the ST data implies that each individual capture area consists of a set of cells that are measured together. While enrichment scores of cell markers can give us a picture of cell type composition within the capture area, we are basing it on a manually chosen set of canonical marker genes. Another approach is to transfer labels from annotated scRNAseq to the spatially resolved data by automatically identifying matching cell pairs (“anchors”) across datasets [45]. To this end, we used the annotated Tabula Muris data set as an alternative method for single-cell mapping and identification of cell type localization. Lung endothelial/epithelial cells were broadly mapped across the entire tissue sections while ciliated cells were exclusively mapped to areas where cilia could be found by histology (figure 2A). For the interest of this study, we specifically investigated the presence of immune cells and their localization.
Manuscript, December 2020
A
B
Classical monocyte Natural Killer cell
Cilia ted c ell
b
500 µm 0.25 0.50 0.75 value leukocyte 500 µm 0.25 0.50 0.75 value leukocyte 500 µm 0.25 0.50 0.75 value classical monocyte 500 µm 0.25 0.50 0.75 value natural killer cell500 µm
0.25 0.50 0.75
value ciliated columnar cell of tracheobronchial tree
b
500 µ m 0.25 0.50 0.75 value T cell 500 µm 0.2 0.4 0.6 0.8 value T cell 500 µm 0.2 0.4 0.6 0.8 value T cell 500 µm 0.25 0.50 0.75 value B cell B cell 500 µm 0.25 0.50 0.75 value B cell Pred.score Leukoc yt e 500 µm 0.25 0.50 0.75 value leukocyte 500 µm 0.25 0.50 0.75 value natural killer cella
a
Low High A B CFigure 2: Label transfer of annotated single-cell data to spatially resolved data. The mapping score reflects the confidence in the mapping. Only capture areas with observed counts greater than zero are plotted A) Ciliated cells, where (a) and (b) show a close-up on the vascular endothelium and ciliated airway epithelial cells, respectively. B) Leukocytes, exhibiting a compartmentalized pattern across the RSV samples. C) Other immune cells and their mappings. Notably, B- and T-cells are mapped to different areas of the lymphoid aggregate. See figure 4 for a close-up of this area.
For each category of immune cells, we find differences in spatial localization (figure 2
B-C). Furthermore, localization patterns of each class were largely consistent across samples
(figure 2B). Adaptive immune cells were highly enriched in two lymphoid aggregates (section 2.3.3), while natural killer cells did not show any clear aggregation except in a small region in one of the samples (figure 2C). Classical/Circulating monocytes were mapped broadly but with some spatial restriction, while the leukocyte mapping demonstrates clear compartmentalization (figure 2C). The signal for these immune cells in the control samples were overall low and without spatial patterns (data not shown). To explore the transcriptional context of these spatial patterns, we correlated the prediction score from the scRNAseq label transfer to the gene expression within each capture area for the sections that demonstrated presence of the various immune cell types. The top-ranking genes (figure 3, supplementary table S1-S5) based on Pearson correlation were for Leukocytes Cxcl2, Ccl4, S100a9, Il1rn, Il1b; for classical monocytes Fth1, Ctss, Ccl6, Tyrobp, Ccl9; for natural killer cells Gzmb, Gzma, Ccl5, Klrk1, Nkg7; for B cells Cd22, Cd79a, Cd19, Fcrl1, Ighm-1; and for T cells Lat, Cd3d, Ms4a4b, Dnase1l3, Coro1a. Functional enrichment analysis using the identified top-ranking genes suggests enrichment of neutrophil activity within both the classical monocytes and leukocyte captures areas but with high cytokine activity (most pronounced IL-10) only within the leukocyte areas. Other areas were enriched for pathways associated with the dominating cell type. These pathways include, for example, Natural killer cell mediated cytotoxicity, B cell receptor signaling, and T cell activation activity.
2.3
Spatial autocorrelation reveals genes with restricted expression
localization
The entire data set was probed for genes that demonstrate spatial patterns of expression using an autocorrelation-based approach [11]. Apart from marker genes for cell types with clear spatial restriction (e.g. airway epithelial cells and genes such as Scgb1a1, Scgb3a2, Scgb3a1, or endothelial and smooth muscle cells and genes such as Myl4, Myl7, Tnni3), identified genes show extensive overlap with genes that were upregulated in response to infection. In other words, infection-altered expression is mainly found within distinct clusters of the tissue and not induced evenly across the entire section. Naturally, these spatially restricted genes are also the primary drivers for the data-driven clusters and also correlate with the localization of the different immune cell types mapped with the scRNAseq data. Among the differentially expressed genes with the highest autocorrelation scores, we find, for example, Saa3, Lcn2, and Retnla among others (supplementary table S6). In this section, we highlight the gene profiles driving the most notable clusters and describe how they are associated with different cell types.
2.4
Areas of acute phase inflammation
Cluster 2 is strongly expressed after RSV challenge and displays distinct localization in parts of the alveolar epithelium in 5 out of 8 RSV samples. It is driven by genes such as Saa3, Lcn2, C1qa and Chil1 (supplementary figure S2). Of these, Saa3 is the single most discriminative gene to the overall cluster profile. It is an isoform of the acute-phase serum amyloid A protein, an inflammatory protein that correlates with CRP and has been suggested as a clinical marker marker for infectious disease monitoring [12]. Furthermore, it has been shown to be critical for lung homeostasis and survival after influenza infection [13], although the exact mechanism as a key immune response mediator is unclear. Lcn2 (Ngal) has historically been viewed as a neutrophil marker but has later also been shown to be expressed by epithelial cells in the lung [14]. The complement system plays an important role in the innate immune response, and its function in viral infections is multifaceted. While the complement system can provide protective effects, by neutralization of the virus both extra- and intracellularly, or via promoting other immune responses, it can also exacerbate disease [15]. For example, it has been shown to contribute to the risk of developing acute respiratory distress syndrome in SARS-Cov-2 infections [16]. The classical pathway of the complement system is initiated by the action of C1q, and the genes for all subcomponents are elevated in response to infection in our data. The importance of complement activation in RSV-infected mice have been studied by disabling the system prior to infection in either RSV-naive mice or mice given RSV-anti serum [17]. These data suggest that complement activation is not critical for restricting viral replication in naive mice but enhances the antibody-mediated response to infection. Furthermore, as an enveloped virus, there seems to be no activation of complement intracellular signaling pathways during RSV infection [18]. In light of the complement activity noted from the expression of both the C1q subcomponents and C3, we probed the genes for the other pathways, but found only low levels and no differential expression between conditions with regard to Masp, Mbp, Factor B and Factor D (supplementary figure S3). Thus, the expression data suggests a predominant activation via the classical pathway. Finally, Chil1 has been correlated with the severity of asthma, chronic lung inflammation, and RSV infection. Specifically, when challenging RSV to WT and Chil1-KO mice, it was seen that Cili1-KO mice did not experience any decreased viral load compared to WT but did show an attenuated airway inflammation [19]. It was further suggested that Chil1 promotes Th2-type response and IL-13-dominant inflammation, which could be an underlying cause for severe disease progression following infection. However, we were unable to detect any evidence of a stronger Th2-type than Th1-type response.
Leukocyte Classical monocytes Natural killer cells A B 500 µm 0.0 0.5 1.0 1.5 2.0 value 500 µm 0.0 0.5 1.0 1.5 2.0 value 500 µm 2.0 2.5 3.0 3.5 4.0 4.5 value 500 µm 0 1 2 value 500 µm 0 1 2 3 value 500 µm 0.0 0.5 1.0 1.5 value Cxcl2 Fth1 Ctss Ccl5 Gzmb Ccl4
Acute phase inflammation
0 1 2 3 4 Saa3 0 1 2 3 Lcn2
Airway epithelial remodeling
C PBS RSV 0 2 4 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Cluster Expression L ev el Retnla 0 1 2 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Cluster Foxj1
Figure 3: Spatial distribution of genes. A) Highest-ranking genes by correlation between gene expression and prediction score of the scRNAseq mapping for Leukocytes, Classical monocytes and Natural killer cells, respectively. B) Top driving genes for cluster 2, involved in acute-phase inflammation processes. C) Cluster 9, which predominantly consists of measurements from RSV-infected mice (figure 1C) and ciliated cells (figure 1D), shows enrichment of genes involved in tissue repair and remodeling.
2.5
Changes in airway epithelium
We find changes in the transcriptomic profile of airway epithelial cells in response to RSV infection 4 days post challenge. The most pronounced upregulation is seen for genes such as Retnla, Foxj1, and Tubb4b. Retnla has been shown to be expressed by “Alternatively Activated” macrophages (AA-M) along with genes such as Arg1, Nr3c2, and Chil3. AA-M is thought to be an early response during RSV infection to mitigate lung damage, and its induction has been suggested to be promoted by IL-4, IL-13, and IL-4Ra [20]. However, of these, our data only shows Retnla as highly expressed while the rest is found at levels barely above detection, contradicting any activity and presence of such macrophages. Furthermore, the function of Retnla appears to be highly pleiotropic in nature [21]. Forkhead transcription factor, Foxj1, is a maintenance protein of airway epithelial cells that regulates ciliogenesis. Interestingly, when susceptible mice were challenged to Sendai Virus, ciliated cells were infected and decreased Foxj1 expression, which was also shown for the response to RSV in humans, resulting in loss of cilia. B-tubulin-IV (Tubb4b), another cilia-specific protein, showed the same behavior [22]. The expression of these proteins then increased after the infection, promoting regeneration of the cilia. Our data demonstrate localized expression of Foxj1 and Tubb4b in the epithelial cells and an overall higher expression of these genes in RSV-infected mice compared to controls, thus pointing to an ongoing repair process within ciliated cells at the time of sampling.
2.6
Induced Bronchus-associated lymphoid tissue
At the time point of observation, we find two well-defined Bronchus-associated lymphoid
tissue (BALT) structures within the lungs of two out of four mice (figure 4A, supplementary
figure S4). Within this area, we find a high abundance of known makers for B-cells (Ms4a1/Cd20), T-cells (Ccr7, Cd8), and FDCs (Cr2/Cd21). These markers, together with the mapping of external scRNAseq profiles to our spatially resolved transcriptomics data, suggests a division between B/T-cells in different compartments within the overall BALT structure (figure 2C). Follicular dendritic cells (FDCs) typically reside within the B-cell follicle to aid in presentation of antigens. We find high expression of complement receptor expression within the same area where the B-cell markers are most pronounced (figure 4). Furthermore, highly elevated expression levels within the BALT area are seen for Mki67 and Pcna, suggesting the formation of germinal centers.
A B Cr2 Mik67 Ccr7 Sell 500 µm 0.0 0.4 0.8 1.2 1.6 value Cr2 500 µm 0.0 0.4 0.8 1.2 1.6 value Mki67 500 µm 0.0 0.5 1.0 1.5 value Ccr7 500 µm 0.0 0.5 1.0 value Sell Cxcl13 Itgal 500 µm 0 1 2 value Cxcl13 500 µm 0.0 0.5 1.0 value Itgal 500 µm 0 1 2 3 value Ccl5 500 µm 0 1 2 value Ccl8 Ccl5 Ccl8 500 µm 0 1 2 value Cxcl13
Figure 4: Induced BALT. A) Hematoxylin and eosin stain of one of the two RSV samples containing BALT. B) Genes with upregulated expression within or adjacent to the BALT area.
Homeostatic chemokines, such as the CXCL13, CCL19, and CCL21, have been shown to be required for optimal immune responses. Following Mycobacterium tuberculosis infection, CCL19 and CCL21 were required for effective T-cell priming, and, of these, CCL19 were locally induced in the lung tissue [23].Other studies have shown CCL21 to promote migration of T-cells to the lung [24] and that the loss of CCL19 and CCL21 impaired BALT formation following influenza infection and resulted in delayed viral clearance when examining mice with removed lymphoid structures (spleen, LNs, and Peyer’s patches) [25]. Chemokine/receptor pairs of Cxcl13/Cxcr5 and Cxcl12/Cxcr4 have also been shown to be critical for the induction and maturation of BALT [26].Here, we observe the FDC-excreted
chemokine Cxcl13 to demonstrate clear upregulated expression in RSV-infected samples. This is in line with earlier reports of induced expression after influenza infection, where antibody staining detected elevated Cxcl13 expression within the follicular areas containing B cells and FDCs [27]. We find no difference between conditions for the expression of Cxcr5, Cxcl12, or Cxcr4. Furthermore, neither Ccl19 nor Ccl21 are detected in our data. Instead, we find other chemokines, such as Ccl5 and CCl8, induced within the BALT area. The precise location within and around the lymphoid organ varies (figure 4B); while Ccl5 expression is confined to the center of the BALT structure, Cxcl13 and Ccl8 is expressed at the outer edges of the aggregate. While the section-wide chemokine profile is similar in the other BALT-containing sample, spatial patterns are less evident. The location of the BALT structure in this sample is located at the edge of the array and close to a larger airway, possibly causing mRNA diffusion, which makes it challenging to pinpoint the exact location of the transcripts (supplementary figure S4). A previous study identified two alternative pathways for homing of blood-derived lymphocytes to the BALT, whereas the route via high endothelial venules (HEV) ochastred by integrins and chemokines was viewed as the primary route of entry [28] .To this end, entry via HEVs was shown to be dependent on L-selectin (Sell) and CCR7. Both of these genes are upregulated within the BALT area in our data (figure 4B). Other genes that have been reported to be involved in lymphocyte homing during spontaneous BALT formation are Vcam1 and Itgal [29]. Itgal displays a slight increase in our BALT area, while Vcam1 does not show any particular upregulation. The levels of discriminative markers for T-cell subsets were overall low and insufficient to profile the composition of the lymphoid aggregates. However, we note a slightly higher localized expression of Cd8 compared to Cd4 (data not shown). When CD8+ T-cells are ac-tivated following respiratory viral infection, activation-associated markers (e.g., Itgal, Il2ra, Klrc1, and Cd44) are upregulated while homeking receptors (e.g., Sell) are downregulated, and the T-cell subsets express different levels of characteristic activation markers (e.g., Ptprc and Ccr7) as well as effector and inflammatory cytokines like Ifn-y, Tnf, Gzmb, and others [30]. Most of these genes are found to be upregulated within the BALT area in our data, but their overall expression levels and ratios are not associated with any particular T-cell subpopulation. Furthermore, polarizing cytokines and effector cytokines usually used to differentiate between the various T-cell subclasses (e.g. Il4, Il5, Il6, Il12, Il13, Il17, Il21, and Il22, among others) are only observed at very low levels. The same observation is made between the master transcriptional regulators (Foxp3, Roryt, GATA3, T-bet, etc). Thus, no robust characterization could be made of the lymphocytic composition.
2.7
Localization of viral genes
The ST data was used to map the expression of RSV genes onto the spatial grid. The viral genes were found in all infected samples and were undetectable in the control samples, as expected. The genes were dispersed across the entire tissue sections, without any clear spatial localization. The ST method quantifies expression from a mixture of cells within each
capture area. With the dispersed expression of viral genes across the entire tissue sections, we could not determine if the viral genes were preferably located within specific cell types. Accordingly, we used a padlock probe-based in-situ hybridization imaging method [31] to obtain higher-resolution maps of RSV expression (figure 5). We used probes for three viral genes: the nucleocapsid, the nonstructural protein 1, and the non-structural protein 2. For characterization of cell types, we used probes targeting Scbg1a1 (airway club cells), Scgb3a1 (proximal-large club cells), Sftpc, Cd74 (alveolar type 2 cells), Ager (alveolar type 1 cells), Ptprc (inflammatory cells), and Pecam1 (endothelial cells). The RSV genes were exclusively localized in or adjacent to the pulmonary alveolar epithelium. No RSV genes were detected in airway epithelium. In the PBS-treated lung, no RSV genes were detected above noise level, as expected.
Figure 5: In situ hybridization four days post RSV infection. Top images show an overview of RSV- and PBS-treated lungs. Close-up images, RSV-treated (A) and PBS-treated (B) lungs, showing markers used to identify cell types and viral genes: Scgbg1a1 (airway club cells), Scgb3a1 (proximal-large club cells), Sftpc, Cd74 (alveolar type 2 cells), Ptprc (inflammatory cells), Pecam1 (endothelial cells), Ager (alveolar type 1 cells), and the viral genes belonging to the nucleocapsid, the nonstructural protein 1, and the non-structural protein 2.
3
Conclusions
In this study, we used inbred BALB/c mice, the most common experimental non-primate model when studying RSV infection. Following RSV challenge, this model has been shown
to display a symmetric bell shape of viral load in the lungs with a peak around day 4 at similar levels of p.f.u [32]. It usually displays mild to moderate bronchiolitis pathology [33], even though there are variability between viral strains and animals [34]. The primary location for viral replication in humans is within ciliated airway epithelial cells, AT1 and AT2 cells, while laboratory hRSV-challenged mice seem to demonstrate replication primarily in AT1 cells [33]. Our in-situ hybridization data suggest no viral replication within ciliated airway cells, but within, or in proximity to, pulmonary alveolar epithelium. Even if viral antigen previously has been shown to be located exclusively in the alveolar cells and not in the bronchial epithelium, instances of lung lesions from histology suggest that bronchial epithelial cells might have been infected [35]. In-situ hybridization has been used to contrast hRSV to pneumonia virus of mice (PVM) infection, for which the mice is a natural host and more permissive. Here, the data suggests the alveoli cells as the first to be infected, which subsequently spread to columnar epithelia of the terminal bronchioles [36]. While the course of PVM infection cannot be extrapolated to RSV, it might suggest that the reason we do not find any viral genes in the airway epithelial cells could be that the infection is still spreading. However, it could also be a result of a more effective innate immune response towards hRSV, which prevents the virus from entering these cells. Nevertheless, the fact that we find an active metagene signature of airway epithelial remodeling and repair suggests that these cells either were infected at an earlier time point or that an earlier epithelial cell differentiation with loss of cilia might have taken place regardless of viral infection. It has previously been shown that Foxj1 null mice, which lack cilia, have impared clearance of pathogens [22], which suggests an inherent defense mechanism that, in theory, could be activated even if the cells themselves are not infected. Further studies with samples from several time points during the infection are warranted to conclude where the viral infection begins and subsequently spreads.
The major strength of an unbiased whole-transcriptome platform is to allow for ex-ploratory analysis without prior choice of targets, allowing us to look broadly across the whole transcriptome in a manner unfeasible with current targeted approaches. However, cur-rent high-throughput methods for whole-transcriptome spatial transcriptomics lack single-cell resolution, which poses a challenge for classifying single-cell types or states within areas that contain multiple such types and detecting rare cell types or states that are by nature not a dominant constituent of a single capture area. Combining the spatially resolved data with data of higher resolution can in some instances be used to circumvent these limitations. For example, as in this study, the spatial grid can be leveraged to get positional information of non-spatially resolved single-cell data. To this end, public repositories can be leveraged to increase the power of the analysis. However, a drawback of using external single-cell data for integration is the inherent mismatch between the data sets. Only cell profiles that exist within the single-cell data can be used as reference for the spatially resolved query. Thus, any cell type missing from the single-cell data will not be found. Furthermore, if there exist differences in environmental conditions, the cell types will most likely exhibit
differences in transcriptomic state. Here, we used an annotated lung data set from Tabula Muris, which contains several immune cell profiles but none in the presence of viral infec-tion. Despite differences in environmental conditions, this analysis works well for giving a broad overview of spatial localization of major cell type classes, as shown in this study by the identified spatially restricted expression patterns along with co-localization of known marker genes. Nevertheless, detailed subclassification of cell types would require either (i) a complete scRNAseq data set containing all different cell types and states, preferably from a sample from the same tissue and condition as the spatial query data set to ensure efficient integration or label transfer, or (ii) a high-resolution spatial method without the need for integration.
ALT is a secondary lymphoid tissue located in the large airways. While these structures could be present under normal conditions, they can also be formed directly in response to pulmonary inflammation, where they convey and prime a local respiratory immune re-sponse. These Inducible BALT (iBALT) structures could preferably be viewed as a type of tertiary lymphoid tissue [37]. In fact, even after removal of the spleen, lymph nodes, and peyer’s patches, mice are able to generate an adaptive immune response against respi-ratory infection, probably initiated within BALT structures [27]. While BALT structures typically have a clearly defined morphology, composed of densely packed lymphocytes in a follicular structure within a stromal meshwork, the induced form could display less defined organization. Furthermore, these could form in different areas within the tissue and are not solely found in the larger bronchial airways. In particular, they can be formed around pulmonary arteries and the perivascular space. As such, both their structure and location are subject to high variability, and it is unclear to which extent the degree of structural formation confers functional alterations, but a suggested minimal criterion for the term BALT is that they contain a B-cell follicular structure with follicular dendritic cells (FDCs) present [37]. Infection-induced BALT often displays germinal center formation, indicating that B-cell responses can be sustained at the local level of the lung tissue. These types of tertiary lymphoid organs in atypical regions have been found in various disease states in humans [38], including several pathogenic conditions in lungs. Its role could, however, be a double-edged sword, as lymphoid aggregates can thicken the walls of airways and contribute to pathogenesis [39]. Extensive iBALT formation could thus be suspected of exacerbating pulmonary disease. Compression by lymphoid aggregates stemming from the BALT has also been suggested to drive progression of RSV in children [40]. As of today, the underlying mechanisms behind the creation, development, and maintenance of tertiary lymphoid organs are not fully understood. Our data clearly suggests the presence of perivascular iBALT structures, which contain FDCs and active germinal centers, as well as excretion of homeostatic chemokines. Several chemokines, their receptors, and other integrins have been shown to convey important functions in iBALT initiation, organization, and maintenance. While we find support for the existence of these molecules in our data, es-pecially for the receptors, there are notable absences of certain key cytokines, most notable
Ccl19/Ccl21. It is currently unclear how long it takes for iBALT structures to form and when they begin to generate a robust adaptive immune response due to infection. When mice were challenged with Influenza A virus, the iBALT volume peaked between 1 and 2 weeks dpi [27]. In another study, BALT formation was noted at the earliest point of sampling (5 days) post challenge with RSV [41]. The time points should be compared across studies with caution, given the differences in pathogenesis known to exist between animals, viruses, and doses. Furthermore, across our 8 sections, we detect two BALT located within the lung. These two sections are both from one of the two studied animals. The other animal has a lymphoid aggregate present in one section, but it seems to be part of non-lung tissue. Thus, the presence of iBALT can only be confirmed in one of the animals. However, it should be noted that we profiled four tissue sections per animal, and did not explore the entire lung. Differences in the role of iBALT formation in animal models and humans warrant further study, as these differences could explain alterations in immune responses with important implications for the usability of the models for translational research.
Of the granulocytes, the gene profiles point towards heightened neutrophil presence in the infiltrated lung tissue. This finding is consistent with an early RSV infection [42]. Once an RSV infection is established, cellular T-cell mediated response is critical for viral clearance, and the balance between different T-cell populations have been suggested as a contributing factor to disease severity. A skew towards T helper 1 (Th1), in contrast to Th2, has been shown to be positively associated with viral clearance [43]. Furthermore, the opposing roles of Th17 and Treg also influence the overall Th1/Th2 activity [44]. Due to the low to undetectable levels of subclass markers and canonical polarizing/effector molecules, we were unable to robustly subclassify the T-cell populations in our data. The T-cell subclasses were originally defined by ex-vivo stimulation with antigen or certain cytokines, and the ability to resolve this heterogeneity within spatial in situ capture data is currently challenging due to the combination of overlapping cytokine profiles, observations of mixed cell types, and low expression of key marker genes. These limitations could also be the reason we were unable to determine if AA-M are induced and if they contribute to the Retnla signal we note within, or in proximity to, the epithelial airways.
In summary, we present an approach for unbiased exploratory transcriptomic profiling during the course of a viral infection. While the resolution of high-throughput untargeted spatial methods prohibits high-resolution subclassification of immune cell types and exact location of viral transcripts, the exploratory analysis can be followed up with targeted high-resolution approaches to answer newly generated research questions. We believe such multimodal approaches will be important for understanding the course of diseases within specific models and contribute to new therapeutic innovations.
4
Methods
4.1
Animal experiment
Four adult female BALB/c mice (> 8 weeks) were intranasal challenged by RSV as described before [5]
4.2
Spatial transcriptomics
4.2.1 Tissue optimization (TO)
To optimize the permeabilization time for the mouse lung tissue, we utilized the Visium Spatial Tissue Optimization Slide & Reagent kit (10x Genomics, Pleasanton, CA, USA). The manufacturer’s instructions were followed with exception of the hematoxylin and eosin (H&E) staining, where the staining was carried out with 1 min isopropanol, 4 min hema-toxylin, and 1 min eosin (1:20 dilution). Optimal values for tissue section thickness and permeabilization time were determined to be 12 µm and 30 minutes, respectively.
4.2.2 Visium Spatial Gene Expression Sequencing Libraries
The Visium Spatial Gene Expression protocol was performed largely according to the manufacturer’s instructions (10x Genomics, Pleasanton, CA, USA) for generating sequenc-ing libraries. Modifications to the protocol were based on the parameters optimized in the TO experiment regarding section thickness, permeabilization time, and H&E staining procedure, as described above.
4.2.3 Sequencing of Visium Libraries
The finished Visium libraries were sequenced in multiple rounds using the Illumina plat-forms NextSeq500, NovaSeq6000, and NextSeq2000. The number of bases sequenced were 28 bases for read 1, 120 bases for read 2, and 10 bases for each of the indexes. Sequencing on the NovaSeq platform was carried out with support from the National Genomics Infras-tructure in Genomics Production Stockholm. Computational infrasInfras-tructure was provided by SNIC/UPPMAX.
4.3
In-situ fluorescence
To characterize the cell types of RSV-infected cells, we applied SCRINSHOT, a multiplexed in situ hybridization method, as described in [31]. Briefly, we transferred the fresh-frozen lung sections from −80 °C to45 °C for 15 min to avoid moisture accumulation. We post-fixed them with 4% paraformaldehyde (PFA) in PBS 1X for 5 min at room temperature (RT) and permeabilized them with HCl 0.1M for 3 min at RT. We blocked unspecific DNA binding for 30 min at RT and then incubated the slides with the padlock probes (Table 1)
for the targeted genes for 2 hours at 45 °C. After thorough washes, we used SplintR ligase (NEB, M0375) to circularize the hybridized probes, O/N at 25 °C. To linearly amplify the padlock probe sequences, we used Phi29-polymerase (Lucigen, 30221-2) mediated rolling cycle amplification (RCA), O/N at 30 °C. Sections were washed and post-fixed with PFA 4% in PBS 1X for 15 min at RT. For the detection of the RCA products, we used fluorophore-labelled oligos (Supplementary Table S6) that recognize the gene-specific sequence of each padlock probe and incubated them for 1 hour at 25 °C. After washing the slides, we applied the SlowFade™ Gold Antifade mounting medium (Thermo, S36936) and put the cover-slip on them. Nuclei were counterstained with DAPI (Biolegend, 422801). To detect the 10 targeted genes, we used sequential hybridization cycles of fluorophore-labelled oligos and signal acquisition. After each detection cycle, the uracil-containing oligos were enzymatically fragmented by uracil-N-glycosylase (UNG) and removed by stringent washes. We repeated the procedure until all genes were detected. Image acquisition was done with a Zeiss AxioImager Z2 microscope, equipped with a Zeiss AxioCam 506 Mono digital camera, a 20x/0.75 Plan-Apochromat lens, and an automated stage. The microscope was equipped with a Zeiss LED Colibri7 light source (Zeiss, 423052-9770-000) with the following Chroma filters: DAPI (49000), FITC (49308), Cy3 (49304), Cy5 (49307), Texas Red (49310), and AlexaFluor 750 (49007). After image acquisition of whole lung sections, we used orthogonal projection and stitching in Zen Blue 2.5 free version. The alignment of the same-tissue images from all detection cycles was done using the DAPI channel (acquired in each cycle) using Zen Blue 2.5.
4.4
Filtering, normalization, batch correction and visualization of ST
data
All data analysis was conducted in R (4.0.3). First, the count data was filtered by removing mitochondrial, ribosomal, and predicted genes from the data set. The SCTransform function from Seurat (3.2.2.9011) [45] with default parameters was used to normalize the data. Prior to clustering, Harmony 0.1 [46] was used for batch correction between ST arrays using the normalized data as input. STUtility (0.1.0) [11] was used for visualizing the data on top of tissue images.
4.5
Cell marker scores
Cell marker scores were calculated by the AUCell approach [47]provided in the AUCell R package (1.12.0). Gene markers were obtained from PanglaoDB [48]. AUCell computes enrichment scores as the area under the curve of the overlap between marker genes and the k highest-expressed genes in each observation unit for increasing 𝑘 ≤ 0.05 × #𝑔𝑒𝑛𝑒𝑠 (in our case, observation units are capture areas). Instead of using the scores for label assignment, we collect all the scores across the tissue sections to visualize spatial patterns.
The enrichment strength was measured by computing the median score across all samples and dividing by the standard deviation.
4.6
Differential gene expression, cluster markers and correlations
The differential gene expression between conditions was estimated using the FindMark-ers() function provided by the Seurat package. Cluster markers were identified using the FindAllMarkers() function in the same package. Both functions were applied with default parameters to the normalized data. Correlations between scRNAseq prediction scores and gene expression were calculated by applying the cor() function in base R with default parameters. The correlation was calculated across capture areas for each gene. Spatial au-tocorrelations were calculated using the STUtility function CorSpatialGenes() with default parameters.
4.7
Integration between scRNAseq and ST data
Seurat V3 was used for integration following the provided vignette. Changes to the default settings were as follows: Anchors between datasets were found using the FindTransfer-Anchors function with canonical correlation analysis (CCA) by setting the “reduction” parameter to “cca”. The Tabula Muris scRNAseq data was used as reference and the ST data as query. Label predictions were computed using the TransferData function withCCA as the weight reduction method by setting the “weight.reduction” parameter to “cca”. The prediction scores for each capture area were used for plotting.
5
Additional information
5.1
Ethics statement
The in vivo studies were carried out in accordance with INRAE guidelines in compliance with European animal welfare regulation. The protocols were approved by the Animal Care and Use Committee at “Centre de Recherche de Jouy-en-Josas” (COMETHEA) under relevant institutional authorization (“Ministère de l’éducation nationale, de l’enseignement supérieur et de la recherche”),authorization number 201803211701483v2 (APAFIS14660). All experimental procedures were performed in a biosafety level 2 facility.
5.2
Author contribution
JB, AD, SAP, and ALS designed the experiments. AD, LS, MARW, JFE, RG and MH per-formed the animal handling and RSV challange. LF perper-formed the spatial transcriptomics (Visium) protocol, AS performed the in situ fluorescence protocol. JB analysed the data and wrote the manuscript.
5.3
Funding
This work was supported by the Swedish Research Council https://www.vr.se/ (521-2014-6718, K2015-99X-22880-01-6) awarded to ALS.
References
[1] T. Shi, D. A. McAllister, K. L. O’Brien, et al., “Global, regional, and national disease burden estimates of acute lower respiratory infections due to respiratory syncytial virus in young children in 2015: A systematic review and modelling study,”
Lancet (London, England), vol. 390, no. 10098, pp. 946–958, 2017, issn:
1474-547X 0140-6736. doi: 10 . 1016 / S0140 - 6736(17 ) 30938 - 8. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/28689664%20https://www.ncbi.nlm.nih.gov/pmc/ articles/PMC5592248/.
[2] P. L. Collins, R. Fearns, and B. S. Graham, “Respiratory syncytial virus: Virology, reverse genetics, and pathogenesis of disease,” Current topics in microbiology and
immunology, vol. 372, pp. 3–38, 2013, issn: 0070-217X. doi:
10.1007/978-3-642-38919 - 1 _ 1. [Online]. Available: https : / / pubmed . ncbi . nlm . nih . gov / 24362682 % 20https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4794264/.
[3] A. R. Falsey, P. A. Hennessey, M. A. Formica, C. Cox, and E. E. Walsh, “Respiratory syncytial virus infection in elderly and high-risk adults,” N Engl J Med, vol. 352, no. 17, pp. 1749–59, 2005, issn: 0028-4793. doi: 10.1056/NEJMoa043951.
[4] J. Blount R. E., J. A. Morris, and R. E. Savage, “Recovery of cytopathogenic agent from chimpanzees with coryza,” Proc Soc Exp Biol Med, vol. 92, no. 3, pp. 544–9, 1956, issn: 0037-9727 (Print) 0037-9727. doi: 10.3181/00379727-92-22538. [5] S. A. Pålsson, A. Dondalska, J. Bergenstråhle, C. Rolfes, A. Björk, L. Sedano,
U. F. Power, M.-A. Rameix-Welti, J. Lundeberg, M. Wahren-Herlenius, P. Mas-trangelo, J.-F. Eleouet, R. Le Goffic, M. Galloux, and A.-L. Spetz, “Single-stranded oligonucleotide-mediated inhibition of respiratory syncytial virus infection,”
Fron-tiers in Immunology, vol. 11, no. 3202, 2020, issn: 1664-3224. doi: 10.3389/fimmu.
2020 . 580547. [Online]. Available: https : / / www. frontiersin . org / article / 10 . 3389 / fimmu.2020.580547.
[6] E. A. van Erp, A. J. Lakerveld, H. L. Mulder, W. Luytjes, G. Ferwerda, and P. B. van Kasteren, “Pathogenesis of respiratory syncytial virus infection in balb/c mice differs between intratracheal and intranasal inoculation,” Viruses, vol. 11, no. 6, p. 508, 2019, issn: 1999-4915. doi: 10.3390/v11060508. [Online]. Available: https: / / pubmed . ncbi . nlm . nih . gov / 31163619 % 20https : / / www. ncbi . nlm . nih . gov / pmc / articles/PMC6631102/.
[7] A. L. Ji, A. J. Rubin, K. Thrane, S. Jiang, D. L. Reynolds, R. M. Meyers, M. G. Guo, B. M. George, A. Mollbrink, J. Bergenstråhle, L. Larsson, Y. Bai, B. Zhu, A. Bhaduri, J. M. Meyers, X. Rovira-Clavé, S. T. Hollmig, S. Z. Aasi, G. P. Nolan, J. Lundeberg, and P. A. Khavari, “Multimodal analysis of composition and spatial architecture in human squamous cell carcinoma,” Cell, vol. 182, no. 2, 497–514.e22, 2020, issn: 0092-8674. doi: https://doi.org/10.1016/j.cell.2020.05.039. [Online]. Available: http://www.sciencedirect.com/science/article/pii/S0092867420306723.
[8] S. Maniatis, T. Äijö, S. Vickovic, C. Braine, K. Kang, A. Mollbrink, D. Fagegaltier, Ž. Andrusivová, S. Saarenpää, G. Saiz-Castro, M. Cuevas, A. Watters, J. Lundeberg, R. Bonneau, and H. Phatnani, “Spatiotemporal dynamics of molecular pathology in amyotrophic lateral sclerosis,” Science, vol. 364, no. 6435, pp. 89–93, 2019. doi: 10 . 1126 / science . aav9776. [Online]. Available: https : / / science . sciencemag . org / content/sci/364/6435/89.full.pdf.
[9] M. Asp, S. Giacomello, L. Larsson, C. Wu, D. Fürth, X. Qian, E. Wärdell, J. Custodio, J. Reimegård, F. Salmén, C. Österholm, P. L. Ståhl, E. Sundström, E. Åkesson, O. Bergmann, M. Bienko, A. Månsson-Broberg, M. Nilsson, C. Sylvén, and J. Lunde-berg, “A spatiotemporal organ-wide gene expression and cell atlas of the developing human heart,” Cell, vol. 179, no. 7, 1647–1660.e19, 2019, issn: 0092-8674. doi: 10.1016/j.cell.2019.11.025.
[10] N. Schaum, J. Karkanias, N. F. Neff, A. P. May, S. R. Quake, et al., “Single-cell tran-scriptomics of 20 mouse organs creates a tabula muris,” Nature, vol. 562, no. 7727, pp. 367–372, 2018, issn: 1476-4687. doi: 10.1038/s41586- 018- 0590- 4. [Online]. Available: https://doi.org/10.1038/s41586-018-0590-4.
[11] J. Bergenstråhle, L. Larsson, and J. Lundeberg, “Seamless integration of image and molecular analysis for spatial transcriptomics workflows,” BMC Genomics, vol. 21, no. 1, p. 482, 2020, issn: 1471-2164. doi: 10.1186/s12864-020-06832-3. [Online]. Available: https://doi.org/10.1186/s12864-020-06832-3.
[12] A. Lannergård, A. Larsson, P. Kragsbjerg, and G. Friman, “Correlations between serum amyloid a protein and c-reactive protein in infectious diseases,” Scandinavian
Journal of Clinical and Laboratory Investigation, vol. 63, no. 4, pp. 267–272, 2003,
issn: 0036-5513. doi: 10 . 1080 / 00365510310001636. [Online]. Available: https : //doi.org/10.1080/00365510310001636.
[13] J. L. Ather, O. Dienz, J. E. Boyson, V. Anathy, E. Amiel, and M. E. Poynter, “Serum amyloid a3 is required for normal lung development and survival following influenza infection,” Scientific Reports, vol. 8, no. 1, p. 16 571, 2018, issn: 2045-2322. doi: 10.1038/s41598-018-34901-x. [Online]. Available: https://doi.org/10.1038/s41598-018-34901-x.
[14] N. Borregaard, O. E. Sørensen, and K. Theilgaard-Mönch, “Neutrophil granules: A library of innate immunity proteins,” Trends in Immunology, vol. 28, no. 8, pp. 340– 345, 2007, issn: 1471-4906. doi: https://doi.org/10.1016/j.it.2007.06.002. [Online]. Available: http://www.sciencedirect.com/science/article/pii/S1471490607001561.
[15] J. Mellors, T. Tipton, S. Longet, and M. Carroll, “Viral evasion of the complement system and its importance for vaccines and therapeutics,” Frontiers in Immunology, vol. 11, no. 1450, 2020, issn: 1664-3224. doi: 10.3389/fimmu.2020.01450. [Online]. Available: https://www.frontiersin.org/article/10.3389/fimmu.2020.01450.
[16] A. M. Risitano, D. C. Mastellos, M. Huber-Lang, D. Yancopoulou, C. Garlanda, F. Ciceri, and J. D. Lambris, “Complement as a target in covid-19?” Nature Reviews
Immunology, vol. 20, no. 6, pp. 343–344, 2020, issn: 1474-1741. doi: 10 . 1038 /
s41577 020 0320 7. [Online]. Available: https : / / doi . org / 10 . 1038 / s41577 020 -0320-7.
[17] A. Bukreyev, L. Yang, and P. L. Collins, “The secreted g protein of human respiratory syncytial virus antagonizes antibody-mediated restriction of replication involving macrophages and complement,” Journal of Virology, vol. 86, no. 19, pp. 10 880– 10 884, 2012. doi: 10.1128/jvi.01162-12. [Online]. Available: https://jvi.asm.org/ content/jvi/86/19/10880.full.pdf.
[18] J. C. Tam, S. R. Bidgood, W. A. McEwan, and L. C. James, “Intracellular sens-ing of complement c3 activates cell autonomous immunity,” Science, vol. 345, no. 6201, p. 1 256 070, 2014, issn: 0036-8075 (Print) 0036-8075. doi: 10 . 1126 / science.1256070.
[19] M. J. Kim, D. H. Shim, H. R. Cha, K. Y. Moon, C. M. Yang, S. J. Hwang, K. W. Kim, J. H. Park, C. G. Lee, J. A. Elias, M. H. Sohn, and J. M. Lee, “Chitinase 3-like 1 protein plays a critical role in respiratory syncytial virus-induced airway inflammation,” Allergy, vol. 74, no. 4, pp. 685–697, 2019, issn: 0105-4538 (Print) 0105-4538. doi: 10.1111/all.13661.
[20] K. A. Shirey, L. M. Pletneva, A. C. Puche, A. D. Keegan, G. A. Prince, J. C. G. Blanco, and S. N. Vogel, “Control of rsv-induced lung injury by alternatively activated macrophages is il-4rα-, tlr4-, and ifn-β-dependent,” Mucosal Immunology, vol. 3, no. 3, pp. 291–300, 2010, issn: 1935-3456. doi: 10 . 1038 / mi . 2010 . 6. [Online]. Available: https://doi.org/10.1038/mi.2010.6.
[21] M. G. Nair, I. J. Gallagher, M. D. Taylor, P. Loke, P. S. Coulson, R. A. Wilson, R. M. Maizels, and J. E. Allen, “Chitinase and fizz family members are a generalized feature of nematode infection with selective upregulation of ym1 and fizz1 by antigen-presenting cells,” Infection and immunity, vol. 73, no. 1, pp. 385–394, 2005, issn: 0019-9567 1098-5522. doi: 10.1128/IAI.73.1.385- 394.2005. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/15618176%20https://www.ncbi.nlm.nih.gov/pmc/ articles/PMC538942/.
[22] D. C. Look, M. J. Walter, M. R. Williamson, L. Pang, Y. You, J. N. Sreshta, J. E. Johnson, D. S. Zander, and S. L. Brody, “Effects of paramyxoviral infection on airway epithelial cell foxj1 expression, ciliogenesis, and mucociliary function,” The
American journal of pathology, vol. 159, no. 6, pp. 2055–2069, 2001, issn:
0002-9440 1525-2191. doi: 10 . 1016 / S0002 - 0002-9440(10 ) 63057 - X. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/11733356%20https://www.ncbi.nlm.nih.gov/pmc/ articles/PMC1850590/.
[23] S. A. Khader, J. Rangel-Moreno, J. J. Fountain, C. A. Martino, W. W. Reiley, J. E. Pearl, G. M. Winslow, D. L. Woodland, T. D. Randall, and A. M. Cooper, “In a murine tuberculosis model, the absence of homeostatic chemokines delays granuloma formation and protective immunity,” The Journal of Immunology, vol. 183, no. 12, pp. 8004–8014, 2009. doi: 10.4049/jimmunol.0901937. [Online]. Available: https: //www.jimmunol.org/content/jimmunol/183/12/8004.full.pdf.
[24] J. C. Lo, R. K. Chin, Y. Lee, H. S. Kang, Y. Wang, J. V. Weinstock, T. Banks, C. F. Ware, G. Franzoso, and Y. X. Fu, “Differential regulation of ccl21 in lym-phoid/nonlymphoid tissues for effectively attracting t cells to peripheral tissues,” J
Clin Invest, vol. 112, no. 10, pp. 1495–505, 2003, issn: 0021-9738 (Print) 0021-9738.
doi: 10.1172/jci19188.
[25] J. Rangel-Moreno, J. E. Moyron-Quiroz, L. Hartson, K. Kusser, and T. D. Randall, “Pulmonary expression of cxc chemokine ligand 13, cc chemokine ligand 19, and cc chemokine ligand 21 is essential for local immunity to influenza,” Proceedings of
the National Academy of Sciences, vol. 104, no. 25, pp. 10 577–10 582, 2007. doi:
10.1073/pnas.0700591104. [Online]. Available: https://www.pnas.org/content/pnas/ 104/25/10577.full.pdf.
[26] H. Fleige, S. Ravens, G. L. Moschovakis, J. Bölter, S. Willenzon, G. Sutter, S. Häussler, U. Kalinke, I. Prinz, and R. Förster, “Il-17–induced cxcl12 recruits b cells and induces follicle formation in balt in the absence of differentiated fdcs,” Journal
of Experimental Medicine, vol. 211, no. 4, pp. 643–651, 2014, issn: 0022-1007. doi:
10.1084/jem.20131737. [Online]. Available: https://doi.org/10.1084/jem.20131737. [27] J. E. Moyron-Quiroz, J. Rangel-Moreno, K. Kusser, L. Hartson, F. Sprague, S. Goodrich, D. L. Woodland, F. E. Lund, and T. D. Randall, “Role of inducible bronchus associated lymphoid tissue (ibalt) in respiratory immunity,” Nature Medicine, vol. 10, no. 9, pp. 927–934, 2004, issn: 1546-170X. doi: 10.1038/nm1091. [Online]. Avail-able: https://doi.org/10.1038/nm1091.
[28] H. Fleige, B. Bosnjak, M. Permanyer, J. Ristenpart, A. Bubke, S. Willenzon, G. Sutter, S. A. Luther, and R. Förster, “Manifold roles of ccr7 and its ligands in the induction and maintenance of bronchus-associated lymphoid tissue,” Cell Reports, vol. 23, no. 3, pp. 783–795, 2018, issn: 2211-1247. doi: https://doi.org/10.1016/ j.celrep.2018.03.072. [Online]. Available: http://www.sciencedirect.com/science/ article/pii/S2211124718304339.
[29] B. Xu, N. Wagner, L. N. Pham, V. Magno, Z. Shan, E. C. Butcher, and S. A. Michie, “Lymphocyte homing to bronchus-associated lymphoid tissue (balt) is mediated by l-selectin/pnad, alpha4beta1 integrin/vcam-1, and lfa-1 adhesion pathways,” J Exp
Med, vol. 197, no. 10, pp. 1255–67, 2003, issn: 0022-1007 (Print) 0022-1007. doi:
10.1084/jem.20010685.
[30] M. E. Schmidt and S. M. Varga, “The cd8 t cell response to respiratory virus infec-tions,” Frontiers in immunology, vol. 9, pp. 678–678, 2018, issn: 1664-3224. doi: 10.3389/fimmu.2018.00678. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/ 29686673%20https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5900024/.
[31] A. Sountoulidis, A. Liontos, H. P. Nguyen, A. B. Firsova, A. Fysikopoulos, X. Qian, W. Seeger, E. Sundström, M. Nilsson, and C. Samakovlis, “Scrinshot enables spatial mapping of cell states in tissue sections with single-cell resolution,” PLOS Biology, vol. 18, no. 11, e3000675, 2020. doi: 10 . 1371 / journal . pbio . 3000675. [Online]. Available: https://doi.org/10.1371/journal.pbio.3000675.
[32] M.-A. Rameix-Welti, R. Le Goffic, P.-L. Hervé, J. Sourimant, A. Rémot, S. Riffault, Q. Yu, M. Galloux, E. Gault, and J.-F. Eléouët, “Visualizing the replication of respiratory syncytial virus in cells and in living mice,” Nature Communications, vol. 5, no. 1, p. 5104, 2014, issn: 2041-1723. doi: 10.1038/ncomms6104. [Online]. Available: https://doi.org/10.1038/ncomms6104.
[33] G. Taylor, “Animal models of respiratory syncytial virus infection,” Vaccine, vol. 35, no. 3, pp. 469–480, 2017, issn: 1873-2518 0264-410X. doi: 10.1016/j.vaccine.2016. 11.054. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/27908639%20https: //www.ncbi.nlm.nih.gov/pmc/articles/PMC5244256/.
[34] K. L. Stokes, M. H. Chi, K. Sakamoto, D. C. Newcomb, M. G. Currier, M. M. Huckabee, S. Lee, K. Goleniewska, C. Pretto, J. V. Williams, A. Hotard, T. P. Sherrill, J. Peebles R. Stokes, and M. L. Moore, “Differential pathogenesis of respiratory syncytial virus clinical isolates in balb/c mice,” Journal of virology, vol. 85, no. 12, pp. 5782–5793, 2011, issn: 1098-5514 0022-538X. doi: 10 . 1128 / JVI . 01693 - 10. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/21471228%20https://www. ncbi.nlm.nih.gov/pmc/articles/PMC3126300/.
[35] G. Taylor, E. J. Stott, M. Hughes, and A. P. Collins, “Respiratory syncytial virus in-fection in mice,” Inin-fection and Immunity, vol. 43, no. 2, pp. 649–655, 1984. [Online]. Available: https://iai.asm.org/content/iai/43/2/649.full.pdf.
[36] P. M. Cook, R. P. Eglin, and A. J. Easton, “Pathogenesis of pneumovirus infections in mice: Detection of pneumonia virus of mice and human respiratory syncytial virus mrna in lungs of infected mice by in situ hybridization,” J Gen Virol, vol. 79 ( Pt 10), pp. 2411–7, 1998, issn: 0022-1317 (Print) 0022-1317. doi: 10.1099/0022-1317-79-10-2411.
[37] T. D. Randall, “Chapter 7 - bronchus-associated lymphoid tissue (balt): Structure and function,” in Advances in Immunology, S. Fagarasan and A. Cerutti, Eds. Academic Press, 2010, vol. 107, pp. 187–241, isbn: 0065-2776. doi: https://doi.org/10.1016/ B978-0-12-381300-8.00007-1. [Online]. Available: http://www.sciencedirect.com/ science/article/pii/B9780123813008000071.
[38] S. Y. Foo and S. Phipps, “Regulation of inducible balt formation and contribution to immunity and pathology,” Mucosal Immunology, vol. 3, no. 6, pp. 537–544, 2010, issn: 1935-3456. doi: 10.1038/mi.2010.52. [Online]. Available: https://doi.org/10. 1038/mi.2010.52.
[39] J. C. Hogg, F. Chu, S. Utokaparch, R. Woods, W. M. Elliott, L. Buzatu, R. M. Cherni-ack, R. M. Rogers, F. C. Sciurba, H. O. Coxson, and P. D. Paré, “The nature of small-airway obstruction in chronic obstructive pulmonary disease,” New England Journal
of Medicine, vol. 350, no. 26, pp. 2645–2653, 2004. doi: 10.1056/NEJMoa032158.
[Online]. Available: https://www.nejm.org/doi/full/10.1056/NEJMoa032158. [40] J. E. Johnson, R. A. Gonzales, S. J. Olson, P. F. Wright, and B. S. Graham, “The
histopathology of fatal untreated human respiratory syncytial virus infection,”
Mod-ern Pathology, vol. 20, no. 1, pp. 108–119, 2007, issn: 1530-0285. doi: 10.1038/
modpathol . 3800725. [Online]. Available: https : / / doi . org / 10 . 1038 / modpathol . 3800725.
[41] A. Auais, B. Adkins, G. Napchan, and G. Piedimonte, “Immunomodulatory effects of sensory nerves during respiratory syncytial virus infection in rats,” Am J Physiol
Lung Cell Mol Physiol, vol. 285, no. 1, pp. L105–13, 2003, issn: 1040-0605 (Print)
1040-0605. doi: 10.1152/ajplung.00004.2003.
[42] C. Johansson, “Respiratory syncytial virus infection: An innate perspective,” F1000Research, vol. 5, pp. 2898–2898, 2016, issn: 2046-1402. doi: 10.12688/f1000research.9637.1.
[Online]. Available: https://pubmed.ncbi.nlm.nih.gov/28105323%20https://www. ncbi.nlm.nih.gov/pmc/articles/PMC5224685/.
[43] P. J. Openshaw and C. Chiu, “Protective and dysregulated t cell immunity in rsv infection,” Curr Opin Virol, vol. 3, no. 4, pp. 468–74, 2013, issn: 1879-6257 (Print) 1879-6257. doi: 10.1016/j.coviro.2013.05.005.
[44] T. C. Mangodt, M. A. Van Herck, S. Nullens, J. Ramet, J. J. De Dooy, P. G. Jorens, and B. Y. De Winter, “The role of th17 and treg responses in the pathogenesis of rsv infection,” Pediatr Res, vol. 78, no. 5, pp. 483–91, 2015, issn: 0031-3998. doi: 10.1038/pr.2015.143.
[45] T. Stuart, A. Butler, P. Hoffman, C. Hafemeister, E. Papalexi, W. M. Mauck, Y. Hao, M. Stoeckius, P. Smibert, and R. Satija, “Comprehensive integration of single-cell data,” Cell, vol. 177, no. 7, 1888–1902.e21, 2019, issn: 0092-8674. doi: https: //doi.org/10.1016/j.cell.2019.05.031. [Online]. Available: http://www.sciencedirect. com/science/article/pii/S0092867419305598.
[46] I. Korsunsky, N. Millard, J. Fan, K. Slowikowski, F. Zhang, K. Wei, Y. Baglaenko, M. Brenner, P.-r. Loh, and S. Raychaudhuri, “Fast, sensitive and accurate integration of single-cell data with harmony,” Nature Methods, vol. 16, no. 12, pp. 1289–1296, 2019, issn: 1548-7105. doi: 10 . 1038 / s41592 - 019 - 0619 - 0. [Online]. Available: https://doi.org/10.1038/s41592-019-0619-0.
[47] S. Aibar, C. B. González-Blas, T. Moerman, V. A. Huynh-Thu, H. Imrichova, G. Hulselmans, F. Rambow, J.-C. Marine, P. Geurts, J. Aerts, J. van den Oord, Z. K. Atak, J. Wouters, and S. Aerts, “Scenic: Single-cell regulatory network inference and clustering,” Nature Methods, vol. 14, no. 11, pp. 1083–1086, 2017, issn: 1548-7105. doi: 10.1038/nmeth.4463. [Online]. Available: https://doi.org/10.1038/nmeth.4463. [48] O. Franzén, L.-M. Gan, and J. L. M. Björkegren, “Panglaodb: A web server for
ex-ploration of mouse and human single-cell rna sequencing data,” Database, vol. 2019, 2019, issn: 1758-0463. doi: 10.1093/database/baz046. [Online]. Available: https: //doi.org/10.1093/database/baz046.
