key: cord-0737804-m0y24zu1
authors: Mothes, R.; Pascual Reguant, A.; Koehler, R.; Liebeskind, J.; Liebheit, A.; Bauherr, S.; Dittmayer, C.; Laue, M.; von Manitius, R.; Elezkurtaj, S.; Durek, P.; Heinrich, F.; Heinz, G. A.; Guerra, G. M.; Obermayer, B.; Meinhardt, J.; Ihlow, J.; Radke, J.; Heppner, F. L.; Enghard, P.; Stockmann, H.; Aschman, T.; Schneider, J.; Corman, V.; Sander, L. E.; Mashreghi, M.-F.; Conrad, T.; Hocke, A.; Niesner, R. A.; Radbruch, H.; Hauser, A. E.
title: Local CCL18 and CCL21 expand lung fibrovascular niches and recruit lymphocytes, leading to tertiary lymphoid structure formation in prolonged COVID-19
date: 2022-03-27
journal: nan
DOI: 10.1101/2022.03.24.22272768
sha: e270b3e31662e460545056f25e319a0b2a23fc93
doc_id: 737804
cord_uid: m0y24zu1

Post-acute lung sequelae of COVID-19 are challenging many survivors across the world, yet the mechanisms behind are poorly understood. Our results delineate an inflammatory cascade of events occurring along disease progression within fibrovascular niches. It is initiated by endothelial dysfunction, followed by heme scavenging of CD163+ macrophages and production of CCL18. This chemokine synergizes with local CCL21 upregulation to influence the stromal composition favoring endothelial to mesenchymal transition. The local immune response is further modulated via recruitment of CCR7+ T cells into the expanding fibrovascular niche and imprinting an exhausted, T follicular helper like phenotype in these cells. Eventually, this culminates in the formation of tertiary lymphoid structures, further perpetuating chronic inflammation. Thus, our work presents misdirected immune-stromal interaction mechanisms promoting a self-sustained and non-resolving local immune response that extends beyond active viral infection and leads to profound tissue repurposing and chronic inflammation.

In order to study the pathophysiological changes that occur in the lung associated to SARS-CoV-2 infection in our cohort, we combined several microscopy techniques and spatial transcriptomics (ST) in consecutive slides of post-mortem lung tissue samples obtained from COVID-19 donors. Lung samples from non-COVID-19-related pneumonia were included as controls. All relevant patient information is shown in Table 1 . Hematoxylin/eosin (HE) -staining showed prominent alterations in the tissue composition and microanatomical structure of COVID-19 lungs, characterized by the accumulation of immune cell infiltrates and fibrotic areas that increased with disease duration (Fig. 1A, first row) . These phenomena were particularly prominent in late disease stages, culminating in a lack of alveolar spaces. In line with that, the immunofluorescence staining (IF) of adjacent tissue sections acquired by confocal microscopy showed an accumulation of collagen deposits with increasing disease duration, accompanied by an increase of CD45 + immune cells (Fig. 1A , second row). While immune cells were scattered throughout the tissue in earlier disease phases, they formed large aggregates at later disease stages. ST analysis of consecutive tissue sections validated the exacerbated collagen deposition at the transcriptional level in samples from later time points (Fig. 1A , third row). We analyzed the same lungs by multi-epitope ligand cartography (MELC) 16, 17 , a multiplex microscopy technique that enabled us to apply a 44-parameter panel for immunofluorescence histology in smaller fields of view (FOV) (Fig. S1A, B ). In accordance with the confocal data, we observed an accumulation of collagen deposits and immune cell infiltrates along with disease duration (Fig. 1A , fourth row). Additionally, multiplex microscopy allowed us to allocate the large immune cell aggregates to distinct fibrovascular areas accumulating around smooth muscle actin (αSMA) + structures (Fig. S1C ).

Based on the reported differences correlating with disease duration, we stratified the COVID-19 cases into acute (1 to 15 days of disease duration), chronic (more than 15 days) and prolonged (7 -15 weeks) (Fig. 1B) . Importantly, individuals at the latter phase had resolved infection. Accordingly, lung samples were tested negative or had only very low RNA load by qPCR, targeting the SARS-CoV-2 E gene (Table 1) . For active viral replication, subgenomic RNA (sgRNA) was used as a surrogate. We obtained a positive result in all acute cases, but in none of the chronic and prolonged lung tissues analyzed in this study (Table 1) .

However, both chronic and prolonged cases showed aggravated lung damage. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted March 27, 2022. ;  https://doi.org/10.1101/2022.03. 24.22272768 doi: medRxiv preprint Pathohistological scoring based on HE/Elastica van Gieson (EvG) staining confirmed a high correlation between stratified samples based on disease duration and lung fibrosis (Fig. 1C) .

We then pre-processed and segmented the multiplex microscopy immunofluorescence images (see methods) to obtain information of the tissue composition at the single-cell level and be able to quantify cells. While total cell numbers and CD45 + cell counts did not differ significantly between controls, acute and chronic COVID-19 groups, there was a significant increase both in cellularity and in absolute numbers of immune cells in the prolonged group ( Fig 1D) .

Thus, we concluded that severe COVID-19 disease progression in our cohort is characterized by lung fibrosis and that a prominent accumulation of immune cells occurs in prolonged cases, particularly in fibrovascular areas.

Next, we further investigated the endothelial compartment in our cohort. In line with previous studies, we observed drastic lung vasculopathy in severe COVID-19 disease 5, 14 , which presented with a breakdown of the endothelial barrier. CD31 staining was absent in 5 out of 8

FOVs analyzed in the acute lungs but could be detected, although with an aberrant pattern, in later disease phases ( Fig. 2A ). In line with the previous reports pointing to a failed remodeling of the alveolar epithelium after SARS-CoV-2 infection 3,5 , we observed a lack in the prototypical monolayer structure in the pancytokeratin (PCK) staining ( Fig. 2A) . Indicative of progressive lung fibrosis, we also detected an increase in αSMA signal and the reticular fibroblast marker ER-TR7 with disease duration ( Fig. 2A ). In addition, computational analysis of the multiplexed microscopy image data 17 , based on dimensionality reduction and unsupervised clustering ( Fig. S3A -C), revealed that the absolute numbers of endothelia, epithelia and fibroblasts per FOV were increased, particularly in the prolonged samples (Fig. S3D) .

Consistent with the loss of CD31 in acutely infected lungs, we detected a downregulation of the endothelial-related transcripts PECAM1, CLDN5 and VWF by ST (Fig. 2B ) and several vessels with detachment of endothelial cells in the same disease group by electron microscopy (Fig. 2C) . The vasculopathy was associated with a prominent coagulopathy, with thrombotic vessels (Fig. 2D , video S1) and coagulation pathways highly abundant in acute lungs (Fig.   2E ), as evidenced by gene set enrichment analysis (GSEA) of the ST data. We additionally observed transcriptional pathways of endothelial activation induced upon COVID-19 in the lungs. Heme scavenging and vessel remodeling processes were upregulated in later disease phases (Fig. 2E ) in the perivascular areas (Fig. 2F ). Further supporting the leakage of plasma from disrupted vessels into the tissue, we detected an increase in Fe 3+ positive cells (using Prussian blue staining) within COVID-19 lungs along with disease duration (Table 1) . Those cells were not restricted to the alveolar space, but also present in perivascular areas. nor immunohistochemically using antibodies against SARS-CoV-2 nucleocapsid (Fig. S2B ).

Hence, we conclude that indirect endothelial dysfunction and leakage in acute COVID-19 lungs lead to heme scavenging along with vessel remodeling within the activated fibrovascular areas, which function as seed-points of fibrosis.

Based on the upregulation of the heme scavenging pathway observed around activated fibrovascular areas, together with the recently demonstrated accumulation of pro-fibrotic macrophages expressing the heme scavenger receptor CD163 in the lungs of severe COVID-19 10 , we next interrogated our ST data for the presence of macrophage-derived, pro-fibrotic candidate genes enriched in those activated fibrovascular areas.

One of the clusters that appeared in the ST analysis, C3, showed a macrophage is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint (Fig. S4A-C) . Interestingly, the highest DEG within C3 was CCL18 ( Fig.   S4A -C), a chemokine known to be produced by alternatively activated macrophages, inducing fibrosis via crosstalk with fibroblasts in idiopathic pulmonary fibrosis (IPF) 18 . Single nuclei RNA sequencing (snRNAseq) data obtained from some of the same samples and donors (Table 1) 19 demonstrated that CD163 + macrophages in COVID-19 are the cellular source of CCL18 (Fig. S5A) . The macrophage population upregulated CD163 in COVID-19 lungs compared to controls, and showed an enrichment with CCL18 expression in later disease phases, compared to the acute lungs.

Consistent with the fact that CD163 and CCL18 belong to the same ST cluster (C3), the expression of both transcripts co-localized in distinct lung microenvironments that neighbored those areas assigned to C5, C6 and C9 throughout the tissue sections ( Fig. S4A-C) . These stroma-enriched, vascular-associated clusters were allocated to large vessel landmarks and the fibrotic areas around them (Fig. 3A ). While these clusters were transcriptionally distinct from each other, they partly shared a fibrotic signature (Fig. S4C) , and could reflect different maturation states of the ongoing fibrosis with a transition from microenvironments characterized by a more endothelial-dominated towards a mesenchymal signature, a process known as endothelial to mesenchymal transition (EndMT) 20 . Accordingly, C5 showed features of constitutive connective tissue associated with intermediate-to-large vessels 1 , while C9 also included endothelial-related genes, suggestive of recently generated, capillary-rich fibrous tissue 21 , and showed a strong expression of CCL21 ( Fig. S4A-C and Fig. 3A) . CCL21 was hardly detectable in acute COVID-19 samples, but highly upregulated in later disease phases, together with COL1A1, COL3A1 and COL4A1, the fibroblast-related transcripts ACTA2, PDGFRA and CTHRC1 and the pro-fibrotic mediator TGF-β (Fig. S5B ). GSEA revealed that the spatial distribution pattern of pathways related to tissue fibrosis mirrored that of CD163, CCL18 and CCL21 and overlapped with the tissue areas represented by C5, C6 and C9 ( Fig.   3 A, B), in contrast to the epithelia-associated clusters C1 and C7 and to C2, which showed an endothelial signature, but lacked contractility-related genes or stromal hallmarks ( Fig. 3A and Fig. S4A-C) .

We next questioned if the activated vasculature was actually able to respond to the CCL18 signal via the cognate receptor CCR8 22 , as well as to CCL21 via CCR7. We indeed detected strong CCR8 and CCR7 expression in SMA + and ER-TR7 + fibroblasts lining large vessels with disrupted endothelia (Fig. 3C ). In addition, several cells of the vascular wall were positive for all three markers, CD31 and SMA and ERTR7, another sign of EndMT [23] [24] [25] . However, while these stromal cells were the main subset expressing the CCR8 receptor within the COVID-19

. CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted March 27, 2022. ;  https://doi.org/10.1101/2022.03.24.22272768 doi: medRxiv preprint lung samples analyzed, the CCR7 signal was not restricted to the stromal compartment.

Instead, it was also observed in cells with a hematopoietic morphology (Fig. 3C) . Therefore, our results show that activated fibrovascular niches in prolonged COVID-19 lungs host heme scavenging CD163 + macrophages, favoring a specialized chemokine milieu including CCL18 and CCL21, which can modulate the local stromal composition and, possibly, the local hematopoietic pool.

We next aimed to further characterize the hematopoietic cells expressing CCR7 within COVID-19 lungs. We detected broad CCR7 expression in large T cells clusters occurring within distinct fibrovascular niches in lung samples from the prolonged group (Fig. 4A, Fig. S6A ). Such T cell aggregates were found in 7 out of 12 FOVs of the prolonged lung samples analyzed by multiplex microscopy, while T cells were scattered without forming clusters in the other disease groups. Both CD4 + and CD8 + T cell counts were significantly enriched in the prolonged lungs Since CCL21 and its receptor CCR7 support T cell migration and homing to lymphoid tissues, but also the recruitment of activated T cells to the lung 26 , we aimed to compare the local T cell response in the effector site to that of the paired draining lymph nodes (dLNs) (see Table   1 for paired organs analyzed; Fig is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint Finally, and based on the occurrence of T cell aggregates within fibrovascular niches in the prolonged COVID-19 lungs, we compared the expression profile of T cells in distinct lung niches within those samples (Fig. 4G) . We observed the highest expression of ICOS, indicative of a T follicular helper phenotype with increased capability to interact with B cells, and the proliferation marker Ki67, in the CD4 + T cell population located within the fibrovascular niche, in comparison to their epithelial niche counterparts. Notably, also CD16 and PD1 expression were upregulated in CD4 + T cells within the fibrovascular niche. Co-expression of all these markers was confirmed at the single cell level within those areas (Fig. S6A ), highlighting the role of this microenvironment for imprinting an unconventional T cell phenotype.

Altogether, we conclude that while T cell activation persists in the draining lymph nodes in prolonged COVID-19 disease, lung fibrovascular niches, which expand and upregulate CCL21, provide a space where CCR7 + T cells accumulate and acquire an unconventional, exhausted T follicular helper-like phenotype with proliferative capacities.

Some of the large immune cell aggregates observed in the prolonged lung samples appeared particularly densely packed, expressed high levels of CD45, were associated to collagen + structures in the fibrovascular space and sometimes, directly attached to bronchi / bronchioli, as shown in (Fig. 5A ). Immune cells contained in this airway were also expressing CD45, albeit at lower levels than the cells in the fibrovascular space. ST analysis revealed a unique transcriptional fingerprint within the area where the CD45 hi cell cluster was located (Fig. 5A,   B ). In line with our multiplex microscopy results, T cell-specific transcripts such as CD3D and TCF7 accumulated within that area, where also MKI67 was transcribed and where CCL21 was also enriched (Fig. 3A , lower row). In addition, CCL19, LTB and CXCL13 (Fig. 5B ) were particularly elevated in and restricted to that area. These cytokines play a crucial role in the organization of lymphoid structures 27 . Along that line, the B cell master regulator PAX5 and CCR6, which has been related to B cell activation and memory formation 28, 29 , were exclusively . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted March 27, 2022. ; https://doi.org/10.1101/2022.03.24.22272768 doi: medRxiv preprint expressed there (Fig. 5B) . IL7R, HLA-DRA transcripts were broader expressed across the tissue section, but they were also found enriched within that region (Fig. 5B) , indicating the presence of survival cues and occurrence of antigen presentation. Of note, macrophagerelated transcripts, such as MARCO and IL1B were not expressed within the lymphocyteenriched area, but rather within the infiltrated airway structures (Fig.S7 ). On the other hand, the master regulator of plasma cell identity PRDM1, together with the immunoglobulin transcripts IGHG1, IGHD, IGHA1 and IGHM were very prominently expressed in airwaysurrounding regions and lower levels of IGHD, IGHA1 and IGHM were found within the lymphoid structure (Fig. 5B ). While IGHA1 was robustly detected within all lung samples analyzed, IGHA2 transcripts were mainly absent, in accordance with published data from blood 30 and lung 31 . Similar large, dense and structured immune cell aggregates, consisting of both T and B cells, could be observed in 2 out of 3 prolonged lungs analyzed by ST.

Consistently, we also observed a significant increase in the absolute numbers of B cells / plasma cells in the prolonged COVID-19 group compared to control tissues and earlier disease stages by multiplex microscopy (Fig. 5C ).

Altogether, our data demonstrate that CCL21 production in fibrovascular niches eventually contributes to the formation of tertiary lymphoid structures in prolonged COVID-19 lungs.

Post-acute COVID-19 sequelae include various pulmonary and extra-pulmonary symptoms 32 .They are not restricted to severe cases and represent a considerable burden on individuals post-infection, as well as on health care systems. Long-COVID syndrome is usually diagnosed if symptoms persist >2 months 33 , but the mechanisms promoting and sustaining lung damage remain elusive. Recent reports highlighted the existing correlation between enriched cytotoxic lymphocytes and B cells within the airways of post-COVID-19 patients and impaired lung functionality 11, 12 . In this study, we stratify the samples based on disease duration and include a prolonged group (7 -15 weeks of disease duration post-infection), in which active viral infection could neither be reported in the lungs nor in the LNs. Yet, those individuals showed persistent lung damage along with the increase in lymphocyte infiltration, as reported for Long-COVID previously 11, 12 . Thus, the prolonged group in this study temporally and symptomatically meets the Long-COVID criteria currently in use. Despite the non-negligible limitations of post-mortem studies, we think the mechanisms described here might also help to shed light into the pathophysiology of the Long-COVID syndrome.

By combining advanced microscopy approaches coupled to single-cell computational analysis with ST techniques in autopsy tissues, we delineated the fibrovascular niche, as a site where . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted March 27, 2022. a dysregulated immune response leads to tissue repurposing in prolonged COVID-19 disease.

In line with the idea that endothelial dysfunction is a key pathogenic mechanism in COVID-19 13, 14 and that endothelial cells are major participants in and regulators of inflammatory reactions 34 , we identify EndMT as a mechanism driving fibrosis. When activated endothelial cells undergo EndMT, they are transcriptionally reprogrammed, their tight cellular junctions are disrupted and they turn into fibroblast-like cells. Consequently, they lose the expression of cell adhesion proteins, such as CD31, while mesenchyme-specific factors, including α-SMA, are upregulated 35 . All these features are evident in our samples ( Fig. 2A-C, Fig. 3A -C, Fig.   S4C ). In addition, we find that fibroblasts in COVID-19 lung fibrovascular niches express the chemokine receptor CCR8 (Fig. 3C ). Its ligand, CCL18, has previously been linked to pulmonary inflammation and fibrosis 36 and it is the biomarker most consistently associated with negative outcomes in IPF 37 . It was previously reported to be enriched in SARS-CoV-2 RNA + myeloid cells in the lungs of COVID-19-deceased individuals 3 . We have recently shown that alternatively activated, pro-fibrotic macrophages accumulate in fibrotic COVID-19 lungs 10 . Here, we demonstrate that CD163 + macrophages accumulating in COVID-19 lungs produce CCL18, and spatially allocate this chemokine to the fibrotic patches (Fig. S5A, Fig.   3A ). It remains to be evaluated whether the reported systemic increase in CCL18 during COVID-19 38 is associated with fibrosis in other tissues beyond the lung, and whether increased CCL18 serum levels are associated to Long-COVID.

Our results also demonstrate an enrichment in CCL21 in prolonged COVID-19 ( CCR7, the receptor for CCL21, is expressed by IPF fibroblasts in contrast to normal lung fibroblasts, and CCR7 + IPF-derived fibroblasts respond to CCL21 with activation, migration, survival, and proliferation 39, 40 . We show here that myofibroblasts in COVID-19 lungs also express CCR7 and, thus, can respond to the elevated CCL21 signals within the fibrovascular niches, particularly in the prolonged phases. On the other hand, although CCL21 is well known for its role recruiting CCR7 + T cells to secondary lymphoid organs to organize the adaptive immune response 41, 42 , CCL21 expression is also essential for the recruitment of effector T cells into the lungs 26 . In line with that, our results show broad expression of CCR7 within the T cell population in the lungs of prolonged COVID-19, time-point in which absolute numbers of both CD4 + and CD8 + T cells are found significantly increased compared to all other disease groups analyzed (Fig. 4A, B ). In addition, the marked upregulation of CD16 in both lung CD4 + and CD8 + T cell populations, together with Eomes and Granzyme B in CD4 + T cells is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint When comparing the T cell phenotype of the lung to that of the paired dLNs, our data point to a shift towards exhaustion within the inflammatory site, as opposed to a more conventional activated status in the secondary lymphoid organs ( Although the function of the exhausted population in the fibrovascular niches is unclear, we noticed co-expression of PD1 and ICOS, a phenotype linked to T follicular helper cells.

Furthermore, the lymphoid accumulations show a strong upregulation of the CCR7 ligands CCL19 and CCL21 and LTB (Fig. 5B ). These chemokines are essential for the formation of inducible bronchus-associated lymphoid tissue (iBALT) upon lung viral infection and inflammation [47] [48] [49] . Additionally, lymphotoxin beta (LT-β) has been shown to further enhance CCL21 expression upon airway challenge, functioning as a positive feedback loop to attract additional CCR7 + effector T cells 26 . The expression of the B cell attracting chemokine CXCL13 50 , another important factor in iBALT organization 51 , is restrictively enhanced in the same lung area (Fig. 5B ) and co-localizes with PAX5, along with evidence for a mucosal B cell profile, local B cell activation and antigen presentation. Thus, the data presented here strongly suggest the formation of iBALT in prolonged COVID-19 lungs, which contributes to the expansion of adaptive immune responses at sites of inflammation, thereby linking these tertiary lymphoid structures to lung pathology and fibrosis. Our data bring the activation of fibrovascular niches together with the generation of ectopic lymphoid follicles, consistent with their repurposing intoan inflammatory neo-tissue 34 . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint 

The authors declare no competing interests. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint Figure S4 . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted March 27, 2022. ; https://doi.org/10.1101/2022.03.24.22272768 doi: medRxiv preprint Figure S7 . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted March 27, 2022. ; https://doi.org/10.1101/2022.03.24.22272768 doi: medRxiv preprint

The lung and lung draining lymph node (dLN) samples included in this study have been Table 1 . Quantitative real-time PCR for SARS-CoV-2 was performed on RNA extracts with RT-qPCR targeting the SARS-CoV-2 E gene. Quantification of viral RNA was performed using photometrically quantified in vitro RNA transcripts as described previously 54 . Total DNA was measured in all extracts by using the Qubit dsDNA HS Assay kit (Thermo Fisher Scientific). The RT-qPCR analysis was replicated at least once for each sample.

As a correlate of active virus replication in the tested tissues, subgenomic RNA (sgRNA) was assessed by using oligonucleotides targeting the leader transcriptional regulatory sequence and a region within the sgRNA encoding the SARS-CoV-2 E gene, as described previously 55 . All information is presented in Table 1 . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted March 27, 2022. ; https://doi.org/10.1101/2022.03.24.22272768 doi: medRxiv preprint

Lung FFPE tissue blocks were taken at the day of autopsy and fixed for 24 h in 4% paraformaldehyde at room temperature. Routine histological staining (HE, EvG and Prussian blue) was performed according to standard procedures.

Immunohistochemical for nucleocapsid staining was performed 1:4000 with #HS-452 011 (Mouse;

Synaptic systems) on a Benchmark XT autostainer (Ventana Medical Systems) with standard antigen retrieval methods (CC1 buffer, pH 8.0, Ventana Medical Systems) using 4-μm-thick FFPE tissue sections. Immunohistochemistry sections were evaluated by at least two board-certified (neuro-)pathologists with concurrence. To biologically validate the immunohistological stainings, control tissues harboring or lacking the expected antigen were used. Staining patterns were compared to expected results as specified in the Fig. S2B , a semi-quantitative scoring was applied (0=no positive cell, +=single positive cells, ++=some positive cells, +++abundant positive cells and debris) (see Table 1 and Fig. S2B ).

Fe3 + positive cells were scored using Prussian blue staining by two independent pathologists semiquantitatively (0=no positive cell; +=single cells; ++=some cells and iron depositions; +++=abundant positive cells and iron depositions).

Images were analyzed using a Zeiss Axiolab 5 microscope.

Frozen autopsy lung samples were thawn, fixed with 2.5% glutaraldehyde and 2% formaldehyde in 0.1 M sodium cacodylate buffer and embedded in Epon according to a routine protocol including en bloc staining with uranyl acetate and tannic acid 52 . Large-scale digitization of ultrathin (70 nm) sections was performed using a Zeiss Gemini 300 field-emission scanning electron microscope in conjunction with a STEM detector via Atlas 5 software at a pixel size of 3-4 nm as previously described 56 . Regions of interest from the large-scale datasets were saved by annotation ('mapped') and then recorded at very high resolution using a pixel size of 0.5-1 nm. Other images of virus particles in autopsy lung tissue of the same patient were previously published [57] [58] [59] . We tried to optimize the sample handling for best preservation of the tissue, however effects mediated by autolysis and suboptimal fixation conditions cannot be excluded. We also analyzed two large-scale datasets of autopsy lung that we published previously 52 (see also www.nanotomy.org for open access of these two datasets).

Cryosections of lung tissue were cut at 7 mm on a MH560 microtome, transferred onto is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted March 27, 2022. ;  https://doi.org/10.1101/2022.03.24.22272768 doi: medRxiv preprint (PBS, 5% BSA and 0.01%Triton) and incubated at room temperature for 60 min. The following antibodies were used for staining: anti-Collagen I-PE, anti-Collagen IV-FITC, anti-CD45-Alexa

Fluor 647 and DAPI. After staining, samples were washed three times with PBS for 5 min and mounted with ProLong gold mounting medium (Life Technologies, Waltham, MA). Images were acquired by confocal microscopy using a Zeiss LSM 880 microscope with × 20/0.5 NA (air) objective at room temperature.

Fresh frozen lung and dLN tissue was cut in 5 µm sections with a MH560 cryotome (ThermoFisher, Waltham, Massachusetts, USA) on cover slides (24 x 60 mm; Menzel-Gläser, Braunschweig, Germany) that had been coated with 3-aminopropyltriethoxysilane (APES).

Samples were fixed for 10 minutes at room temperature (RT) with freshly opened, electron microscopy grade 2% paraformaldehyde (methanol-and RNAse-free; Electron Microscopy Sciences, Hatfield, Philadelphia, USA). After washing, samples were permeabilized with 0.2% Triton X-100 in PBS for 10 min at room temperature and unspecific binding was blocked with 10% goat serum and 1% BSA in PBS for at least 20 minutes. Afterwards, a fluid chamber holding 100 μl of PBS was created using "press-to-seal" silicone sheets (Life technologies, Carlsbad, California, USA; 1.0 mm thickness) with a circular cut-out (10 mm diameter), which was attached to the cover slip surrounding the sample. Prior to each MELC experiment fresh washing solution consistent of PBS with 5% MACS BSA (Miltenyi Biotec) and 0.02% Triton X-100 was prepared. The sample was placed on the sample holder and fixed with adhesive tape followed by accurate positioning of the binning lens, the light path, as well as Köhler illumination of the microscope.

MELC image acquisition was performed as previously shown 16, 17 . In short, we generated the is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted March 27, 2022. ;  https://doi.org/10.1101/2022.03.24.22272768 doi: medRxiv preprint followed by acquisition of the fluorescence images 3-D stack (+/-5 z-steps) ; (iii) photobleaching of the fluorophore; and (iv) a second auto-focusing step followed by acquisition of a post-bleaching fluorescence image 3D stack (+/-5 z-steps). In each four-step cycle up to four fluorescence-labeled antibodies were used, combining PE, FITC, APC and DAPI and images were acquired for two fields of view (FOV) in order to maximize area and cell numbers analyzed for each experiment and sample.

The antibodies used are listed in the supplementary information table. The antibodies were stained in the indicated order.

Image pre-processing was conducted as previously described 17 . In short, the reference phase-contrast image taken at the beginning of the measurement was used to align all images by cross-correlation. Afterwards, the signal of the bleaching image in each cycle and focal plane was used to subtract the background and correct the illumination of the fluorescence image obtained in the same cycle and focal plane 60 .Thereby, tissue autofluorescence and potential residual signal from the previous cycle were removed. In case of uncoupled antibodies and to account for unspecific signal of the secondary antibody used, we subtracted the fluorescence image of the secondary antibody stained and acquired before the corresponding primary antibody, instead of the bleaching image. Then, an "Extended Depth of Field" algorithm was applied on the 3D fluorescence stack in each cycle 61 . Images were then normalized in Fiji, where a rolling ball algorithm was used for background estimation, edges were removed (accounting for the maximum allowed shift during the autofocus procedure) and fluorescence intensities were scaled to the full intensity range (16 bit => 2 16 ). The 2-D fluorescence images generated were subsequently segmented and analyzed.

Segmentation was performed in a two-step process as previously described 17 , a signalclassification step using Ilastik 1.3.2 62 and an object-recognition step using CellProfiler 3.1.8 63 . Ilastik was used to classify pixels into three classes: nuclei, membrane, and extracellular matrix (ECM). A probability map for each class was generated. Classification of images regarding membranes and ECM was performed by summing up a combination of images, using markers expressed in the respective compartments, while only the DAPI signal was used to classify nuclei. The random forest algorithm (machine-learning, Ilastik) was trained by manual pixel-classification in a small region of one data-set (approx. 6% of the image) and applied to the rest without re-training. CellProfiler was subsequently used to segment the . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted March 27, 2022. ;  https://doi.org/10.1101/2022.03.24.22272768 doi: medRxiv preprint nuclei and membrane probability maps and to generate nuclei and cellular binary masks, respectively. These masks were superimposed on the individual fluorescence images acquired for each marker, in order to extract single-cell information (mean fluorescence intensity, MFI) of each marker per segmented cell and their spatial coordinates.

Data was transformed using the hyperbolic arcsine function with a scale argument of 0.2. 

We compared the phenotype of the CD4 + and CD8 + T cell clusters located in the fibrovascular niche to the ones located in the epithelial niche. We defined the niche as the area of a 32 pixel-radius (equivalent to 10 µm; the average size of a hematopoietic cell) around each endothelial or epithelial cell, similar to previous analysis 17 . We extracted the MFI of each T cell-related marker for the sum of all fibrovascular niches and epithelial niches and displayed the expression profile as heat map showing the z-score values.

Visualisation of gene expression in lung tissue was performed using 10x Visium spatial gene expression kit (10x Genomics) following manufacturers protocol. The four capture areas is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint 68 , where the normalized enrichment scores can be visualized on the spatial feature plot in Seurat. NES range was clipped to the first and last five percentile to enhance visibility.

Each lung tissue sample was minced and placed in digestion medium (500 U/ml Collagenase, I.5 U/ml Dispase and 1 U/ml DNAse) for 1 h at 37°C. Cells were filtered through a 70 μm strainer and enzymatic reaction was stopped by cold RPMI with 10 % fetal bovine serum and 1 % L-Glutamine. Cells were washed with 50 ml cold RPMI with 10 % fetal bovine serum and 1 % L-Glutamine and red blood cells were lysed using Red blood cell lysis solution (MiltenyiBiotec). Finally, cells were filtered using a 40 μm FlowmiR Cell Strainer (Millipore) and re-suspended in PBS supplemented with 2 % fetal bovine serum at the concentration of 10.000 cells/μl for scRNA-Seq. The single-cell capturing and downstream library constructions were performed using the Chromium Single Cell 3ʹ V3.1 library preparation kit according to the manufacturer's protocol (10x Genomics). Full-length cDNA along with cell-barcode identifiers were PCR-amplified and sequencing libraries were prepared. The constructed libraries were either sequenced on the Nextseq 500 using 28 cycles for read 1.55 cycles for read 2, and 8 index cycles, or on the Novaseq 6000 S1 using 28 cycles for read 1, 64 cycles for read 2, and 8 index cycles, to a median depth of 36000 reads per cell.

The Cell Ranger Software Suite (Version 3.1.0) was used to process raw sequencing data with the GRCh38 reference. Single-cell RNA sequencing data analysis was performed in R (version 3.6.1) with Seurat (version 3.1.1). Cells with at least 500 and less than 5000 detected genes and less than 10% mitochondrial content were combined from each library. Library depth (total number of UMIs) was regressed out when scaling data and libraries from different donors were integrated using IntegrateData. Cluster annotation was performed using the Human Lung Cell Atlas reference dataset 69 and Seurat's TransferData workflow as published elsewhere 19 .

Whole-mount staining and optical clearing of human lung tissue About 3 mm PFA-fixed human lung cubic samples were processed for tissue permeabilization and decolorization based on a recently published approach 70 . All steps were performed permanently shaken on a tube rotator (MACSmix TM , Miltenyi Biotec). We incubated the . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted March 27, 2022. ; https://doi.org/10.1101/2022.03.24.22272768 doi: medRxiv preprint samples first in 5mL 25% w/v Quadrol solution (N,N,N′,N′-Tetrakis-(2-hydroxypropyl)ethylendiamin, Sigma-Aldrich) for 2 days at 37°C, followed by 5mL 25% w/v Quadrol and 10% w/v CHAPS solution ((3-[3-cholamidopropyl)dimethylammonio]-1-1propanesulfonate, Sigma-Aldrich) for 5 days at 37°C 70 . After initial decolorization and permeabilization, we followed the SHIELD protocol for PFA-fixed human samples 71 . Therefore, samples were incubated in SHIELD-OFF solution at 4 °C for 3 days and then placed in SHIELD-ON Buffer and SHIELD-Epoxy Solution in a ratio of 1:1 at RT for 24h. SHIELD preservation is then completed and we washed the samples for 1 day in PBS with three times refresh at 4°C. Next, we actively cleared the samples for 3 days using stochastic electrotransport (SmartClear Pro, LifeCanvas Technologies) and performed the same washing step as before. 

We slightly glued the human lung samples on a plastic plate and immersed the sample in 100% Easy Index solution in a 5x5x45mm quartz cuvette (msscientific Chromatographie-Handel GmbH). For image acquisition, we immersed the smaller cuvette containing the lung sample in the bigger quartz cuvette (LaVision Biotec), filled with fused silica matching liquid with a RI of 1,459 (Cargille Laboratories).To acquire light-sheet images we used an Ultramicroscope II (LaVision BioTec) coupled to an Olympus MVX10 zoom body providing a zoom ratio from 0.63x -6.3x and a 2x dipping objective (Olympus MVPLAPO2XC/0.5 NA [WD= 5.6mm, corrected]) resulting in a total magnification ranging from 1.36x -13.56x. The Ultramicroscope II featuring a lateral resolution of ~4 µm was equipped with an Andor Neo sCMOS Camera with a pixel size of 6.5x 6.5 µm² and a LaVision BioTec Laser Module featuring the following filter sets: ex 488 nm, em 520/50 nm; ex 561 nm, em 620/60 nm; ex 785 nm, em 835/70 nm. We used the 488 nm ex for generation of autofluorescent signals and detection of nuclear Sytox-AF488 signal, the 561 nm ex for detection of ER-TR7-AF546 signal and the NIR 785 nm ex for detection of CD3-iF790. We imaged the longest wavelength first to . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted March 27, 2022. ; https://doi.org/10.1101/2022.03.24.22272768 doi: medRxiv preprint avoid photobleaching during imaging and performed a linear z-adaption of the ex. 785 nm laser. We used a total optical zoom of 8,61x and a z-step of 5 µm for all acquisitions. Bigger tile scans were acquired using 10% overlap along longitudinal x-axis and y-axis of the human lung sample and the laser power was adjusted depending on the intensity of the signal (in order to not reach the saturation.

We separately acquired 16-bit grayscale TIFF images for each channel by light sheet microscopy with the ImSpector software (LaVision Biotec). Tiff stacks were converted (ImarisConverterx64) into Imaris files (.ims) and stitched by Imaris Stitcher. We used Imaris 

De-identified human/patient spatial transcriptomics data have been deposited at Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/), under record GSE190732, and are available for reviewers. Microscopy data, as well as any additional information required to reanalyze the data reported in this paper will be shared by the lead contact upon request. This study did not use any original code.

Perivascular stromal cells: Directors of tissue immune niches

Causes of death and comorbidities in hospitalized patients with COVID-19

COVID-19 tissue atlases reveal SARS-CoV-2 pathology and cellular targets

Imbalanced Host Response to SARS-CoV-2 Drives Development of COVID-19

Lung epithelial and endothelial damage, loss of tissue repair, inhibition of fibrinolysis, and cellular senescence in fatal COVID-19

Healing after COVID-19: Are survivors at risk for pulmonary fibrosis?

Post-COVID-19 pulmonary fibrosis: Novel sequelae of the current pandemic

Severe COVID-19 Is Marked by a Dysregulated Myeloid Cell Compartment

The spatial landscape of lung pathology during COVID-19 progression

SARS-CoV-2 infection triggers profibrotic macrophage responses and lung fibrosis

Immuno-proteomic profiling reveals aberrant immune cell regulation in the airways of individuals with ongoing post-COVD-19 respiratory disease

Immune signatures underlying post-acute COVID-19 lung sequelae

Endothelial dysfunction and immunothrombosis as key pathogenic mechanisms in COVID-19

Pulmonary Vascular Endothelialitis, Thrombosis, and Angiogenesis in Covid-19

Long-term cardiovascular outcomes of COVID-19

Multiplexed fluorescence microscopy reveals heterogeneity among stromal cells in mouse bone marrow sections

Multiplexed histology analyses for the phenotypic and spatial characterization of human innate lymphoid cells

A vicious circle of alveolar macrophages and fibroblasts perpetuates pulmonary fibrosis via CCL18

Human Lungs Show Limited Permissiveness for SARS

Due to Scarce ACE2 Levels But Strong Virus-Induced Immune Activation in Alveolar Macrophages

Endothelial to mesenchymal transition represents a key link in the interaction between inflammation and endothelial dysfunction

Fibroblasts as confederates of the immune system

Identification of human CCR8 as a CCL18 receptor

Epithelial-tomesenchymal and endothelial-to-mesenchymal transition from cardiovascular development to disease

Histone deacetylase 9 promotes endothelial-mesenchymal transition and an unfavorable atherosclerotic plaque phenotype

Endothelial to mesenchymal transition contributes to arsenic-trioxideinduced cardiac fibrosis

Differential regulation of CCL21 in lymphoid/nonlymphoid tissues for effectively attracting T cells to peripheral tissues

Lymphoid neogenesis in chronic inflammatory diseases

Early CCR6 expression on B cells modulates germinal centre kinetics and efficient antibody responses

CCR6-Dependent Positioning of Memory B Cells Is Essential for Their Ability To Mount a Recall Response to Antigen

Single-cell multi-omics analysis of the immune response in COVID-19

SARS-CoV-2 in severe COVID-19 induces a TGF-βdominated chronic immune response that does not target itself

High-dimensional characterization of post-acute sequelae of COVID-19

Long COVID: current definition

Evolving functions of endothelial cells in inflammation

TGF-β-Induced Endothelial to Mesenchymal Transition in Disease and Tissue Engineering

Complex regulation of pulmonary inflammation and fibrosis by CCL18

Genetic variation in CCL18 gene influences CCL18 expression and correlates with survival in idiopathic pulmonary fibrosis: Part A

TGF-β and CCL18 as Indicators for Predicting and Monitoring the Development of Pulmonary Fibrosis in Patients with COVID-19

Idiopathic pulmonary fibrosis fibroblasts migrate and proliferate to CC chemokine ligand 21

Therapeutic targeting of CC ligand 21 or CC chemokine receptor 7 abrogates pulmonary fibrosis induced by the adoptive transfer of human pulmonary fibroblasts to immunodeficient mice

CCR7 coordinates the primary immune response by establishing functional microenvironments in secondary lymphoid organs

The chemokine SLC is expressed in T cell areas of lymph nodes and mucosal lymphoid tissues and attracts activated T cells via CCR7

BDCA-3)+ dendritic cells (DCs) represent a unique myeloid DC subset that cross-presents necrotic cell antigens

Transforming growth factor-β-regulated mTOR activity preserves cellular metabolism to maintain long-term T cell responses in chronic infection

Untimely TGFβ responses in COVID-19 limit antiviral functions of NK cells

Stromal cell contributions to the homeostasis and functionality of the immune system

Inducible bronchus-associated lymphoid tissue (iBALT) in patients with pulmonary complications of rheumatoid arthritis

Pulmonary expression of CXC chemokine ligand 13, CC chemokine ligand 19, and CC chemokine ligand 21 is essential for local immunity to influenza

B cell-attracting chemokine 1, a human CXC chemokine expressed in lymphoid tissues, selectively attracts B lymphocytes via BLR1/CXCR5

Regulation of inducible BALT formation and contribution to immunity and pathology

Olfactory transmucosal SARS-CoV-2 invasion as a port of central nervous system entry in individuals with COVID-19

B cell depletion and signs of sepsis-acquired immunodeficiency in bone marrow and spleen of COVID-19 deceased

Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR

Virological assessment of hospitalized patients with COVID-2019

Preparation of Samples for Large-Scale Automated Electron Microscopy of Tissue and Cell Ultrastructure

Why misinterpretation of electron micrographs in SARS-CoV-2-infected tissue goes viral

Using EM data to understand COVID-19 pathophysiology -Authors' reply

COVID-19 and the central and peripheral nervous system

Analyzing proteome topology and function by automated multidimensional fluorescence microscopy

Generation of all-in-focus images by noise-robust selective fusion of limited depth-of-field images

Ilastik: Interactive Machine Learning for (Bio)Image Analysis

CellProfiler: Image analysis software for identifying and quantifying cell phenotypes

Integrated analysis of multimodal single-cell data

R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing

Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression

Fast gene set enrichment analysis

Easy single cell analysis platform for enrichment

A molecular cell atlas of the human lung from single-cell RNA sequencing

Cellular and Molecular Probing of Intact Human Organs

Protection of tissue physicochemical properties using polyfunctional

The authors are most grateful to the patients and their relatives for consenting to autopsy and subsequent research. We thank Lars Philipsen for his assistance with multiplex microscopy and Gudrun Holland for her support in acquisition of the electron microscopy data. The authors