Chikungunya virus infection disrupts lymph node lymphatic endothelial cell composition and function via MARCO

Infection with chikungunya virus (CHIKV) causes disruption of draining lymph node (dLN) organization, including paracortical relocalization of B cells, loss of the B cell-T cell border, and lymphocyte depletion that is associated with infiltration of the LN with inflammatory myeloid cells. Here, we find that during the first 24 h of infection, CHIKV RNA accumulates in MARCO-expressing lymphatic endothelial cells (LECs) in both the floor and medullary LN sinuses. The accumulation of viral RNA in the LN was associated with a switch to an antiviral and inflammatory gene expression program across LN stromal cells, and this inflammatory response, including recruitment of myeloid cells to the LN, was accelerated by CHIKV-MARCO interactions. As CHIKV infection progressed, both floor and medullary LECs diminished in number, suggesting further functional impairment of the LN by infection. Consistent with this idea, we find that antigen acquisition by LECs, a key function of LN LECs during infection and immunization, was reduced during pathogenic CHIKV infection.


INTRODUCTION
Chikungunya virus (CHIKV), a mosquito-transmitted arthritogenic alphavirus, remains a persistent threat to global health 10 years after spreading to the Americas and 20 years since epidemic-level outbreaks occurred in the Indian Ocean region (1,2).CHIKV disease typically presents with acute fever, rash, and severe arthralgia, and up to 60% of patients remain chronically afflicted with arthritis and arthralgia for months to years post infection (2,3), constituting a high socioeconomic burden (3).Studies in both animal models (4)(5)(6)(7)(8) and human patients (9,10) suggest that chronic CHIKV disease is associated with persistence of viral RNA and antigen in cells within joint-associated tissue.
In prior studies using an immunocompetent mouse model of CHIKV infection, we found that CHIKV evades the B cell response to establish viral persistence in joint-associated tissues.
These studies revealed that peripheral lymph nodes are important for the generation of the CHIKV-specific B cell response and control of CHIKV infection (6,7,11,12).Closer examination of the lymph node response during CHIKV infection revealed that wild-type (WT) CHIKV infection disrupts the structure and function of the draining lymph node (dLN), the first secondary lymphoid organ to encounter virus following infection (13), in contrast to infection with the attenuated CHIKV 181/25 strain, which does not establish persistent infection in mice (11).This disruption of dLN organization is mediated by an early influx of inflammatory myeloid cells that contribute to diminished lymphocyte recruitment and retention, disruption of the B-T cell border, relocalization of B cells, and poor germinal center formation (12,13).However, the specific cell types that interact with CHIKV and promote inflammation in the dLN remain to be elucidated.
The architecture and cellular organization of LNs are essential for the development of effective immune responses against viral antigens (14).The organization of LNs is coordinated by lymph node stromal cells (LNSCs), including fibroblastic reticular cells (FRCs), lymphatic endothelial cells (LECs), and blood endothelial cells (BECs).These cell populations provide a physical scaffold for immune cell migration and produce signals to regulate migration, adhesion, localization, function, and survival of hematopoietic cells.FRCs and LECs can be grouped into distinct subsets based on transcriptional signatures and regional localization within the LN (15)(16)(17)(18)(19)(20).LECs are among the first cells in the LN to encounter viruses, cells, and antigens draining into the LN via the afferent lymphatics (21)(22)(23).In addition, LN LECs play an active role in enhancing immune responses through internalization and retention of antigen during vaccination and infection (24)(25)(26).
Recently, we found that CHIKV dissemination within an infected host is restricted by LNs and the scavenger receptor MARCO (macrophage receptor with collagenous structure), and that viral particles co-localize with MARCO + LECs in the dLN (27).Single cell mRNA sequencing (scRNA-seq) of LNSCs confirmed that viral RNA in the dLN accumulated largely in a subset of LECs that express MARCO termed MARCO + LECs (27).As LECs are important regulators of LN tissue organization and function (28)(29)(30)(31), we hypothesized that the interaction between CHIKV and MARCO + LECs promotes LN inflammation previously shown to impair B cell responses during CHIKV infection.
In this study, using scRNA-seq, we identified MARCO-expressing floor LECs that line the subcapsular sinus (SCS) as a site of early viral RNA accumulation.These findings, together with our prior observations that CHIKV RNA subsequently accumulates in MARCO + LECs that line the medullary sinuses, suggest that CHIKV targets multiple subsets of MARCO-expressing LECs in the LN.CHIKV infection caused dramatic alterations to the gene expression program of LNSCs, including LECs and other LNSC subtypes, that was characterized by an inflammatory gene transcriptional response, and this early inflammatory response was accelerated by CHIKV-MARCO interactions.Quantification of LN LEC subsets throughout the acute phase of CHIKV infection revealed reduced numbers of both floor and medullary LECs at later times post-infection, which was MARCO-dependent.Evaluation of LN LEC function during CHIKV infection revealed that WT, but not attenuated, CHIKV infection impairs antigen acquisition by LN LECs in a MARCO-dependent manner.Collectively, these findings identify a role for the scavenger receptor MARCO in regulation of LN inflammation and support a model by which CHIKV targeting of MARCO-expressing LECs initiates an inflammatory response that rewires the transcriptional program of LNSCs, alters the composition of specialized LEC subtypes, and impairs known LEC functions.

CHIKV RNA accumulates in MARCO-expressing LN LECs
Previously, using confocal microscopy and scRNA-seq analysis of dLN cells, we discovered that MARCO + LECs lining the medullary sinuses internalize virus particles and harbor CHIKV RNA at 24 h post-infection (27).Notably, analysis of LECs captured from mock-and CHIKV-infected LNs indicated that the number of floor LECs was reduced in CHIKV-infected compared with uninfected control LNs (27).Floor LECs line the inner layer of the SCS and have intimate contacts with antigen-detecting SCS macrophages (15,29), which makes them the first LNSCs to encounter cells and foreign antigens entering the LN and suggests these cells could interact with CHIKV prior to MARCO + LECs in the downstream medullary sinus.CHIKV replication peaks between 24-72 h following infection of WT C57BL/6 mice (6,7,11,27,32), and can be directly cytopathic (33).Based on these observations, and previous reports that some floor LECs express MARCO (15,16,20), we hypothesized that floor LECs interact with CHIKV at earlier times post-infection leading to their reduction by 24 h post-infection.To test this idea, LNSCs from the dLN of mock-and CHIKV-infected mice were profiled by scRNA-seq at 8 h post-infection.Similar to our previous analysis, single cell suspensions of LNs were enriched for CD45 -stromal cells within the LN by negative selection before processing for scRNA-seq (Supplemental Figure 1) (27).In comparison with our prior analysis of LNSCs at 24 h (27), fewer viral RNA reads were detected at 8 h.Thus, we enriched the cDNA libraries for CHIKV RNA using our previously described resampling and resequencing method (34).After enrichment, we clustered cells and identified LN cell types using our established methods (27) (Supplemental Figure 2).Next, we assessed the percentage of reads aligning to the CHIKV genome among total CHIKV-and mouse-specific reads for each individual cell (CHIKV score, Figure 1A).Analysis of cells harboring viral RNA revealed several CHIKV + cell types consisting of different endothelial cell and fibroblast populations (Figure 1B), a finding that is consistent with the known primary tropism of CHIKV for non-hematopoietic cells (8,35).Among the CHIKV + cell types identified, floor LECs and MARCO + LECs had the highest CHIKV scores and the greatest proportion of CHIKV + cells (Figure 1C-1D), with the floor LEC population containing the highest proportion of CHIKV + cells.
Since CHIKV RNA was predominantly detected in MARCO-expressing LECs at 24 h postinfection (27), we next analyzed Marco expression in the CHIKV RNA + and RNA -floor LECs at 8 h post-infection.CHIKV RNA + floor LECs (CHIKV + ) exhibited greater Marco expression than CHIKV RNA -floor LECs (CHIKV -) (Figure 1E).Floor and MARCO + LECs are transcriptionally similar but can be distinguished by Madcam1 expression in floor LECs, which is absent in MARCO + LECs (15).Both CHIKV RNA + and CHIKV RNA -floor LECs exhibited similar Madcam1 expression, a marker unique to floor LECs (36)(37)(38), suggesting that these cells are indeed floor LECs and not mis-annotated MARCO + LECs (Figure 1F).Collectively, these studies reveal that CHIKV RNA accumulates in two subsets of LECs during the first 24 h of infection and suggest that CHIKV interactions with MARCO are important for viral capture and internalization by endothelial cells in the LN.

CHIKV RNA + LECs show signs of active CHIKV RNA replication
At 24 h, CHIKV RNA-high cells identified by scRNA-seq had attributes consistent with decreased cell viability and with virus-mediated transcriptional shutoff, including expression of fewer host genes and an increased fraction of reads aligning to mitochondrial genes (27), suggesting that LECs support active CHIKV RNA replication.One marker of active CHIKV RNA replication is the production of a positive-sense subgenomic mRNA that encodes the viral structural polyprotein (39).To provide further evidence of viral RNA replication in LECs, we calculated the ratio (sgRNA ratio) of reads aligning to the viral sgRNA to reads aligning to the fulllength viral genome.Consistent with the localization of CHIKV RNA-high cells in our previous study (27), at 24 h cells with the highest sgRNA ratio were found predominantly within the MARCO + LEC subset and a cluster of endothelial cells that we were unable to further annotate due to the low number of expressed host genes (unassigned-LEC) (Supplemental Figure 3A-D).When we further characterized cell types with the highest CHIKV sgRNA ratio at 24 h, we observed a negative correlation between the sgRNA ratio and the number of mouse genes expressed by MARCO + LECs and unassigned-LECs (Supplemental Figure 3E).In addition, we also identified a positive correlation between the sgRNA ratio and the percentage of mitochondrial reads per cell (Supplemental Figure 3E).Notably, the unassigned-LECs have a higher CHIKV sgRNA ratio, a higher percentage of mitochondrial reads, and a lower number of expressed mouse genes compared with MARCO + LECs, suggesting these cells could be severely injured MARCO + LECs (Supplemental Figure 3E).To further evaluate LEC viability during CHIKV infection, we used flow cytometry to assess the viability of LECs in the dLN of mock-and CHIKVinfected mice at 1 d post-infection, which revealed diminished LEC viability in the dLN of CHIKVinfected mice (Supplemental Figure 3F), consistent with our scRNA-seq data.Overall, these results show that cells with high levels of viral sgRNA have indications of reduced viability, suggesting that CHIKV RNA replication occurs in LECs that capture and internalize CHIKV particles and leads to cell injury or death.

Lymph node sinus alteration during pathogenic CHIKV infection
Our analyses indicated that CHIKV RNA accumulates in multiple subsets of MARCOexpressing LN LECs during infection.To further investigate the fate of these cells, we evaluated Lyve1 and MARCO expression in the dLN during infection with the attenuated CHIKV 181/25 strain, which does not disrupt dLN cellular organization (13), and its parental strain, the pathogenic WT CHIKV AF15561, at 8, 24, and 48 h post-infection using immunofluorescent confocal microscopy.At 8 and 24 h post-infection, Lyve1 and MARCO expression were similar across dLNs from mock-, CHIKV 181/25-, and WT CHIKV-infected mice (Supplemental Figure 4A-4B).Lyve1 signal was observed in both subcapsular and medullary sinus regions, supporting the annotation of both floor and MARCO + LEC subsets in the 8 h post-infection scRNA-seq dataset and consistent with previous reports supporting the specificity of Lyve1 expression for floor (low level) and MARCO + (high level) LECs (15,16,40).MARCO signal was localized predominantly to the LN medullary sinuses, consistent with reports indicating MARCO expression on LN LECs is regionally distinct (15,16).However, by 48 h post-infection, Lyve1 signal was greatly reduced and MARCO signal was undetectable in dLNs from WT CHIKV-infected mice (Figure 2A-B and Supplemental Figure 4C).Higher magnification imaging of the subcapsular and medullary sinuses of LNs from mock-and WT CHIKV-infected mice using Lyve1 (floor and MARCO + LECs) and CD36 (ceiling LECs) revealed marked expansion of both sinuses in the dLN of mice infected with WT CHIKV (Figure 2C).To investigate the cellular composition of the expanded sinuses, LN sections from mock-and WT CHIKV-infected mice were stained for CD11b, as prior studies identified localization of inflammatory monocytes to the medullary region of the dLN at 24 h post-CHIKV infection (12), the fibroblast marker ERTR-7 to visualize the LN capsule, and DAPI to identify cell nuclei.Indeed, the expanded LN sinuses observed in the dLN of WT CHIKV-infected mice contained numerous CD11b + cells (Figure 2D), supporting the association of inflammatory cellular infiltrates with alteration of resident cells in the LN sinus.

Pathogenic CHIKV infection alters LN LEC composition
To evaluate whether loss of Lyve1 and MARCO signal at 48 h post-infection corresponded to the loss of specific LECs or an altered composition of LEC subsets, dLN stromal cells were evaluated at 1, 2, and 5 d post-infection by flow cytometry.After excluding non-viable cells, LNSCs were segregated within all CD45 -cells based on expression of CD31/PECAM-1 and podoplanin (PDPN) (15,36,37): CD31 + PDPN -blood endothelial cells (BECs), CD31 -PDPN + fibroblastic reticular cells (FRCs), and CD31 + PDPN + LECs.LECs subsets were defined using mannose receptor C-type 1 (MRC1), intercellular adhesion molecule 1 (ICAM1), integrin subunit alpha 2B (ITGA2B), and the scavenger receptor CD36, which have been used successfully in other studies to discriminate medullary (MRC1 + ICAM1 + ), floor (MRC1 -ICAM1 + ITGA2B + ), and ceiling (MRC1 - ICAM1 -CD36 + ) LEC subsets (36,41) (Figure 3A).The total number of BECs, FRCs, and LECs was similar between mock-, CHIKV 181/25-, and WT CHIKV-infected LNs at 1 and 2 d postinfection (Figure 3B).At 5 d post-infection, while the number of BECs increased to a similar extent following infection with either CHIKV 181/25 or WT CHIKV (Figure 3B), FRC and LEC numbers increased solely in CHIKV 181/25-infected LNs (Figure 3B), suggesting that WT CHIKV infection alters the proliferation or survival of these LNSC subtypes.Further segregation of LECs into medullary, floor, and ceiling subtypes revealed that both the percentage and number of medullary LECs was reduced during WT CHIKV infection compared with CHIKV 181/25 infection, whereas the number of ceiling LECs was unchanged at 5 d post-infection (Figure 3C).The percentage and number of floor LECs was also reduced during WT CHIKV infection compared with CHIKV 181/25 infection (Figure 3C).Notably, there was a small but significant reduction in the percentage and number of dLN floor LECs between mock and WT CHIKV-infected mice at 1 d post-infection, suggesting that these cells, which interact with the virus early after infection, could be damaged as a result.These data suggest that WT CHIKV infection impairs the expansion and/or maintenance of specific regional LEC subsets.Notably, medullary and floor LECs include all the MARCO-expressing LECs in the LN, which are LN cell types predominantly targeted by CHIKV early after infection (Figure 1B-D) (27), suggesting that these changes could be due to CHIKV-MARCO interactions.

Alteration of LN LEC composition is dependent on CHIKV-MARCO interactions
Our prior studies identified a role for the scavenger receptor MARCO in early viral accumulation in the dLN and restricting early viral dissemination to distal tissues (27,42).Furthermore, our data indicates that CHIKV RNA accumulates predominantly in subsets of LECs that express MARCO.Thus, we hypothesized that CHIKV-MARCO interactions promote LN sinus alteration and altered LEC subset composition in the dLN of WT CHIKV-infected mice.To assess this, dLNs were evaluated at 48 h post-infection by confocal microscopy following CHIKV infection of WT and MARCO -/-mice.In contrast to the sparse Lyve1 signal in the dLN of WT mice, the dLN of CHIKV-infected MARCO -/-mice exhibited more robust Lyve1 expression (Figure 4A).Higher magnification imaging of the medullary sinus highlights the difference in Lyve1 expression in the dLN, where Lyve1 signal is substantially diminished in WT mice while MARCO -/-mice maintain robust Lyve1 expression (Figure 4B).These data suggest that MARCO promotes the loss of Lyve1 expression or LECs in the dLN during CHIKV infection.Furthermore, analysis of LN LEC subsets at 5 d post-infection revealed that in WT CHIKV-infected MARCO -/-mice, medullary and floor LEC populations were increased compared with WT CHIKV-infected WT mice (Figure 4C-D), indicating that the altered composition of LECs during CHIKV infection is MARCO-dependent.
In addition, the percentage of ceiling LECs was similar in WT CHIKV-infected MARCO -/-mice to CHIKV 181/25-infected WT mice in contrast to the higher percentage of ceiling LECs in WT CHIKV-infected WT mice (Figure 4C), suggesting that the change in the proportion of ceiling LECs may be a component of the LNSC response to CHIKV infection.Importantly, the composition of LN LEC subsets is similar in naïve WT and MARCO -/-mice (Figure 4E-F), supporting the conclusion that disruption of LN LEC composition during CHIKV infection is MARCO-dependent.Overall, these data demonstrate a role for MARCO in the alteration of specific LN LEC populations during pathogenic WT CHIKV infection.

Inflammatory gene expression in LNSCs during CHIKV infection
We next evaluated differential gene expression in LNSCs from mock-and CHIKV-infected mice at 8 and 24 h post-infection using our scRNA-seq datasets.UMAP projection of all LNSCs from mock-and CHIKV-infected mice at both 8 and 24 h post-infection revealed that LNSCs from CHIKV-infected mice at 8 h clustered strongly with LNSCs from mock-infected mice whereas LNSCs from CHIKV-infected mice at 24 h were strongly segregated from the other populations (Figure 5A-5B) and this was consistent when coloring the UMAP projection by cell type (Figure 5B).We next identified gene ontology terms (biological process) for genes upregulated in each cell type.To identify the predominant gene expression programs upregulated during the first 24 h of CHIKV infection, we clustered gene ontology (GO) terms for each timepoint into distinct modules based on similarity.From this analysis the primary gene expression module upregulated at 8 h post infection consists of factors associated with the innate immune response (Figure 5C), including Bst2 which is broadly upregulated in most cell types at 8 h and can promote retention of CHIKV particles at the host cell membrane to prevent virus release (43,44).We also detected upregulation of Zbp1, a key factor in sensing cytosolic DNA during virus-induced cell damage, and Irf7, another key factor in the induction of anti-viral cytokines such as interferon-β (45,46) (Figure 5D).When we compare changes in gene expression between the 24 h and 8 h timepoints we identify a similar innate immune response module and observe further upregulation of Bst2, Zbp1, and Irf7 (Figure 5D-E).By 24 h post-infection, we also detected increased expression of genes involved in a broader inflammatory response, including Ccl2, Cxcl9, Cxcl10, and Ccl7, which were upregulated across the major LNSC subsets (Figure 5E-F).However, minor differences in the degree of expression of specific genes were noted, with Ccl7 upregulated to a lower degree in MARCO + LECs and perivascular cells (PvCs) than FRCs, which is consistent with fibroblasts being a primary CCL7-producing stromal cell type (Figure 5F) (47,48).In addition, CCL2 is a potent chemoattractant for monocytes, which previous studies demonstrated are detrimental to LN structure and function (12,49); the high expression of Ccl2 detected in LECs suggests these cells may contribute to early recruitment of inflammatory monocytes.We also evaluated expression of important LNSC homeostatic chemokines including Ccl21a, Il7, Cxcl13, and Ccl19 (Figure 5G).The expression level of these genes was largely unchanged or diminished in LNSCs of CHIKV-infected mice when compared to mock-infected mice (Figure 5G), indicating that the primary effect of CHIKV infection on the transcriptome of LNSCs is activation of antiviral and inflammatory gene expression programs.

MARCO expression triggers a rapid LN inflammatory response
The presence of inflammatory CD11b + cells in the expanded sinuses of CHIKV-infected LNs (Figure 2B-C), retention of Lyve1 signal in MARCO -/-mice (Figure 4A), and high expression of monocyte chemoattractant Ccl2 in LNSCs at 24 h post-infection (Figure 5F) suggest that CHIKV-MARCO interactions promote LN inflammation via regulation of inflammatory chemokine expression and recruitment of inflammatory monocytes.To investigate this hypothesis, inflammatory chemokine mRNA expression (Ccl2, Cxcl1, Cxcl9, and Cxcl10) was assessed in whole LNs from WT and MARCO -/-mice at time points (8,12, and 16 h post-infection) prior to the dominant type I IFN response observed at 24 h (Figure 5) (12).Consistent with our scRNA-seq analysis, little to no upregulation of these chemokines was observed in CHIKV-infected WT or MARCO -/-mice at 8 h post-infection.However, by 12 h post-infection, the expression of Ccl2, Cxcl1, Cxcl9, and Cxcl10 in the dLN of CHIKV-infected WT mice, but not MARCO -/-mice, was significantly increased in comparison with mock-infected mice (Figure 6A).By 16 h post-infection, chemokine expression was increased in both CHIKV-infected WT and MARCO -/-mice in comparison with mock-infected mice, however, Ccl2, Cxcl1, and Cxcl9 expression remained significantly higher in the dLN of WT mice (Figure 6A).To further address whether direct CHIKV-MARCO interactions and subsequent viral internalization promote early inflammatory chemokine expression in the dLN, chemokine expression in the dLN was assessed at 12 h post-infection in WT mice infected with WT CHIKV or CHIKV E2 K200R , which lacks interaction with MARCO (27,42,50).The expression of Ccl2 and Cxcl1, was significantly increased in the dLN of WT CHIKV-infected mice compared with CHIKV E2 K200R -infected mice (Figure 6B), whereas differences in Cxcl9 and Cxcl10 expression were not statistically significant.Concurrent with the lower inflammatory chemokine expression in MARCO -/-mice, we detected less CHIKV RNA in the dLN of MARCO -/-mice at 8 and 12 h post-infection (Figure 6C).These data suggest that CHIKV-MARCO interactions promote inflammatory gene expression in the dLN.
To determine if MARCO also promotes infiltration of the LN with CD11b + cells during CHIKV infection, we evaluated accumulation of inflammatory monocytes in the dLN of WT and MARCO -/-mice using flow cytometry (Figure 6D).Consistent with higher Ccl2 expression in WT mice at 12 and 16 h post-infection, the percentage (Figure 6E) and total number of (Figure 6F) inflammatory Ly6C hi monocytes (CD45 + CD11c neg CD11b + Ly6C hi Ly6G neg ) in the dLN was significantly greater in WT mice at 12 and 16 h post-infection compared with MARCO -/-mice.By 24 h post-infection, the percentage of inflammatory monocytes remained significantly higher in WT mice than in MARCO -/-mice (Figure 6E) although the difference in monocyte numbers was not statistically significant (Figure 6F).These data suggest CHIKV-MARCO interaction induces a rapid early pro-inflammatory response, recruiting pathogenic monocytes which cause disruption of dLN cellular organization and impair B cell responses (12).

Pathogenic CHIKV infection impairs foreign antigen acquisition by LECs
A key function of LN LECs is the ability to acquire and retain foreign antigen to promote long-lived adaptive immunity following both viral infection and vaccination (21,(24)(25)(26).Prior studies showed that fluorescently-labeled ovalbumin (ova) or influenza nucleoprotein (NP) is specifically acquired by LN LECs upon subcutaneous injection of mice experiencing an active viral infection or when delivered with an adjuvant, such as polyI:C (24-26).To evaluate the functional capacity of LN LECs to acquire antigen during CHIKV infection, we immunized CHIKV 181/25-or WT CHIKV-infected mice with 10 µg ova-488 in both calf muscles (20 µg/mouse total) at 3 d post-infection, at which point the dLN of WT CHIKV-infected mice is substantially disorganized (13), and assessed the proportion of ova + LNSCs 2 and 7 days later (Figure 7A).
As a positive control, we immunized naïve mice in both calf muscles with 10 µg ova-488 and 5 µg polyI:C per calf (24,25).Consistent with prior studies (24,25), ova acquisition was highly specific to LN LECs, as little to no ova was detected in BECs or FRCs (Supplemental Figure 5A-B), confirming that we could measure LEC antigen acquisition by this method (25).We evaluated ova + LNSCs in both the popliteal (first footpad draining LN) and iliac (next draining LN in sequence) LNs (51) to determine if any differences in ova acquisition were associated with impaired lymphatic drainage.Comparison of LNSC populations in the popliteal LN of ovaimmunized/CHIKV-infected mice revealed that the number of LECs in WT CHIKV-infected mice was diminished at both 2-and 7-days post ova injection in the popliteal LN compared with CHIKV 181/25-infected mice (Figure 7B).FRC and BEC numbers were not significantly different between WT CHIKV-infected mice and CHIKV 181/25-infected mice at both 2-and 7-days post ova immunization in the popliteal LN (Figure 7B), suggesting the adverse effects of WT CHIKV are specific to LECs.By gating on ova + LECs in naïve mice as a negative control, we found that both the percentage and number of ova + LECs was reduced during WT CHIKV infection compared with CHIKV 181/25 infection in the popliteal LN (Figure 7C-D).The difference in ova + LEC number was greater at both 2-and 7-days post-ova in the popliteal LN between CHIKV 181/25 and WT CHIKV infection than total LEC number alone, suggesting differences in ova acquisition within the popliteal LN cannot be attributed solely to differences in total LEC numbers.Notably, similar findings were observed in the iliac LN (Supplemental Figure 5C-E

DISCUSSION
Previous work demonstrated that LN cellular organization and immune responses are disrupted during CHIKV infection (12,13), however, the specific virus-host interactions that promote these aberrant LN responses remained unknown.Recently, we found that CHIKV particles are internalized by MARCO + medullary LECs in the LN (27).Building on these findings, here we show that CHIKV RNA accumulates within floor and medullary LN LECs, that this accumulation is associated with expression of the scavenger receptor MARCO by these cells, and that these cells may support active viral RNA replication.Moreover, CHIKV infection was associated with the rapid induction of an antiviral and inflammatory gene expression program across LNSCs that was accelerated by CHIKV-MARCO interactions.In addition, we found that as CHIKV infection progressed, both floor and medullary LECs diminished in number in a MARCO-dependent manner.This unique viral targeting and/or capture by LN LECs raised questions about how virus interactions with LNSCs influence the function of these cells in immunity.Indeed, we found that acquisition of vaccine antigen by LECs was reduced following pathogenic CHIKV infection in a MARCO-dependent manner, suggesting that virus-LNSC interactions can influence subsequent secondary responses.
Both previous work and our data here indicate that CHIKV RNA accumulates within specific subsets of LECs in the dLN (27).That is, CHIKV RNA preferentially accumulated in LECs expressing higher levels of the scavenger receptor MARCO (27), emphasizing the critical role of MARCO in facilitating CHIKV targeting of LECs, consistent with our recent report that MARCO can facilitate viral internalization in vitro (50).The accumulation of CHIKV RNA in MARCO + floor LECs and medullary MARCO + LECs also reflects the transit over time of viral particles through the LN sinuses, and prompted the question of whether these cells are permissive to viral replication.In fact, we identified an increased ratio of CHIKV sgRNA within MARCO + LECs at 24 h post-infection and both a lower number of expressed host genes and a higher percentage of mitochondrial reads in those cells, suggesting active viral RNA replication.These data contrast with a prior report indicating LNSCs are not targets of CHIKV infection (13).One explanation could be that there is some viral RNA replication within LECs but the infection is ultimately abortive due to little to no virion production and release.Indeed, expression of the antiviral factor Bst2/tetherin was increased at both 8 and 24 h post-infection in LECs, and we observed robust expression of type I IFN-stimulated genes at 24 h post-infection, which together could limit virus replication.
Alternatively, due to the small population of LECs that interact with CHIKV, CHIKV + cells may be difficult to detect by methods less sensitive than RNA sequencing.Although CHIKV can interact with a multitude of cell surface proteins for attachment on many cell types, including endothelial cells, other studies have predominantly identified fibroblasts, skeletal muscle cells, and macrophages as targets for active CHIKV replication (8,35,52,53).Indeed, signs of active viral RNA replication in LNSCs are unique and have rarely been identified.Instead, much research has focused on LN resident macrophages, which interact with and can be targeted by viruses such as Zika virus, vaccinia virus, and vesicular stomatitis virus (54)(55)(56).However, lymphocytic choriomeningitis virus (LCMV) targets murine FRCs within lymphoid organs, and importantly, LCMV infection of FRCs was higher for the clone 13 strain, which establishes a persistent infection, than the Armstrong strain, which is cleared efficiently by CD8 + T cells (57).In contrast, while multiple studies have shown that FDCs can trap HIV-1 and maintain infectious virions, representing a potential viral reservoir during chronic infection, FDCs are not permissive to HIV-rather than the result of viral targeting of FDCs (58,59).While more investigation is needed, viral targeting of LNSCs by CHIKV and LCMV, two viruses associated with chronic disease, suggest that viral targeting of LNSCs may have a critical role in the establishment of chronic viral infection.
A key marker of the floor and medullary LEC populations that harbor CHIKV is Lyve1, which distinguishes them from other LEC subsets (15,16,20,26,40).LNs from WT CHIKV but not attenuated CHIKV 181/25-infected mice displayed a reduced and spatially altered expression of Lyve1 by 48 h post-infection as well as a loss of MARCO expression, suggesting that WT CHIKV infection disrupts Lyve1 + LECs.In lymphatic vessels, Lyve1 expression is negatively regulated by inflammatory cytokines (60).Notably, WT CHIKV infection of MARCO -/-mice did not disrupt Lyve1 expression, suggesting that the perturbations of Lyve1 expression observed in WT mice could be driven by MARCO-promoted inflammatory responses.Consistent with this idea, we found that the presence of MARCO accelerated and promoted inflammatory gene expression  (12,13).These changes were specific to floor and medullary LECs, the LEC subsets that harbor CHIKV RNA based on scRNA-seq.Notably, both floor and medullary LEC numbers were restored in LNs from WT CHIKV-infected MARCO -/-mice, suggesting that these changes are a consequence of CHIKV targeting of MARCO-expressing LECs.Unexpectedly, we observed a reduced proportion of ceiling LECs among total LECs during CHIKV 181/25 infection compared with WT CHIKV at 5 d post-infection, yet total ceiling LEC numbers were similar.It is possible that both Cd36 and Icam1 expression are altered in ceiling LECs by cytokines and chemokines expressed in the dLN in the first 5 d post-infection, biasing our gating analysis.Furthermore, we also observed a reduced proportion of ceiling LECs in WT CHIKV-infected MARCO -/-mice similar to that seen in CHIKV 181/25-infected WT mice, supporting the idea that this change is part of a functional LN immune response.One caveat to comparison of our flow cytometry and confocal microscopy data is that the major LEC subsets were identified using different surface proteins.
Thus, there could be microenvironment-dependent changes in expression of key markers of LEC subsets that complicate interpretation of the cell populations identified by flow cytometry, such as upregulation of ITGA2B on LECs in response to inflammation (37,61).Notably, comparison of LEC subsets during homeostatic and response-to-stimuli conditions using scRNA-seq suggested that floor LECs undergo the greatest transcriptional alteration following stimulus with the TLR7 agonist imiquimod, with moderate alteration observed in medullary LECs, and the least changes observed in ceiling LECs (40).However, we note that total LEC numbers in the dLN, as determined by CD31 and PDPN, were reduced during WT CHIKV infection compared with CHIKV 181/25 infection, suggesting that the major changes in LEC subsets are accurate.A previous report investigating the mechanisms dictating expansion and contraction of LN LECs during an immune response demonstrated that type I IFN and PD-L1 both inhibit early LEC division and that a decrease in PD-L1 was sufficient to increase LEC proliferation (62).Given the early inflammatory response in the dLN that occurs during WT CHIKV infection, signaling by type I IFNstimulated genes, such as PD-L1, may prevent expansion of LN floor LECs.
LEC functions include maintenance of LN homeostatic chemokine gradients, acquisition and storage of antigen to promote memory CD8 + T cell responses through antigen exchange with migratory DCs, and maintenance of peripheral self-tolerance (24-26, 28, 63, 64).We found that the capacity of LN LECs to acquire antigen after a secondary immunization was specifically impaired during WT CHIKV infection.In contrast, LEC antigen acquisition in mice infected with CHIKV 181/25 was similar to that observed in naïve mice stimulated with polyI:C, consistent with data showing that viral infection can also induce LEC antigen acquisition (24).Antigen acquisition was markedly reduced 2 days after immunization in WT CHIKV infected mice in a MARCOdependent manner, and retention of that antigen also may be impaired as the number of ova + LECs remained constant between 2-and 7-days post-immunization in CHIKV 181/25-infected mice but decreased in WT CHIKV-infected mice.Overall, these data provide evidence that LNSCtargeting viruses that disrupt the function of the cells represent a challenge for vaccination campaigns as patients recently infected with such a virus may need to delay immunization to generate stronger, more protective vaccine-specific responses.Importantly, while CHIKV induces both LEC alteration and dysfunction in a MARCO-dependent manner, MARCO itself is not required for optimal LEC function as LECs from uninfected WT and MARCO -/-mice stimulated with polyI:C acquired antigen equally well, and the composition of LN LEC subsets was also similar in naïve WT and MARCO -/-mice.
In summary, our work demonstrates that CHIKV interactions with specific subsets of LECs expressing the scavenger receptor MARCO is associated with remodeling of the LNSC transcriptome, extensive LN inflammation, and dysfunction of LN LECs.These findings suggest CHIKV-LEC interactions contribute to impaired downstream LN function and impaired adaptive immunity during CHIKV infection.
Viruses.CHIKV AF15561, AF15561 E2 K200R and 181/25 were generated as previously described (65).Briefly, plasmids were linearized by NotI digestion and used as a template for in vitro transcription with SP6 DNA-dependent RNA polymerase (Ambion).RNA transcripts were electroporated into BHK-21 cells and 24 h later, cell culture supernatant was collected and clarified by centrifugation (1,721 x g), aliquoted, and stored at -80°C.Viral titers were determined by plaque assay or by quantification of RNase-resistant viral genomes by RT-qPCR as previously described (7,13).
Mouse Experiments.Mice were bred in specific-pathogen-free facilities at the University of Colorado Anschutz Medical Campus.All mouse studies were performed in an animal biosafety level 3 laboratory.WT C57BL/6J (Jax# 000664) mice were acquired from Jackson Laboratories, and congenic MARCO -/-mice were provided by Dawn Bowdish (McMaster University) (66).Mice were anesthetized with isoflurane vapors and inoculated with the indicated dose of virus in a 10 μl volume via subcutaneous (s.c.) injection into the rear footpad.For scRNA-seq and flow cytometry experiments, mice were inoculated with an equal dose of virus in both rear footpads; for microscopy, mice were inoculated in a single footpad.As sex-based difference have not been observed in the CHIKV infection model, WT male mice were purchased commercially and were age matched and distributed randomly across groups.Based on prior studies of LN inflammation during CHIKV infection, mice 4 weeks of age were used in all experiments (12,13).Experimental animals were humanely euthanized at defined endpoints by exposure to isoflurane vapors followed by bilateral thoracotomy.Single-cell library preparation.Cells were subjected to single-cell droplet-encapsulation using the Next GEM Chip G Kit (1000127) and a 10× Genomics chromium controller housed in a BSL3 laboratory.We targeted recovery of 10,000 cells per replicate.Single-cell gene expression libraries were generated using the Next GEM single-cell 30 GEM library and gel bead kit v3.1 (1000128) and single index kit T set A (1000213).Sequences were generated with an Illumina NovaSEQ 6000 instrument using S4 flow cells and 300 cycle SBS reagents.We targeted 50,000 reads per cell, with sequencing parameters of Read 1:151 cycles; i7 index: 10 cycles; i5 index: 0 cycles; Read 2: 151 cycles.

Preparation of single-cell suspensions
CHIKV-specific library enrichment.The scRNA-seq libraries for the 8 h timepoint were enriched for molecules aligning to the CHIKV genome according to a previously published method (34,67).Specifically, the CHIKV genome was PCR amplified in 3 fragments (primer sequences: CHIKV-R3 5ʹ-AAAAACAAAATAACATCTCCTACGTC-3ʹ) and labeled with biotin-dUTP using the same primers before sonicating to generate ~150 bp fragments for hybridization.Denatured and diluted biotin-dUTP-labeled CHIKV genome fragments were hybridized to the concentrated scRNA-seq libraries separately.Streptavidin capture beads were mixed with the hybridized libraries and washed to remove unbound DNA.Libraries were amplified directly from the cleanedup beads and sequenced.FASTQ files for each replicate were processed using the cellranger count pipeline (v5.0.1).Reads were aligned to the mm10 and CHIKV AF15561 (EF452493.1)reference genomes.
To quantify CHIKV RNA levels for each cell identified in the enriched library, we calculated a CHIKV score (Figure 1A and C), which is the number of CHIKV reads aligning to the CHIKV genome divided by the total mouse reads and CHIKV reads for each cell.To visualize this metric on UMAP projections (Figure 1A), a pseudo count (smallest non-zero value / 2) was added to each value plotted.
Single-cell RNA-seq gene expression analysis.FASTQ files for each replicate were processed using the cellranger count pipeline (v5.0.1).Reads were aligned to the mm10 and CHIKV AF15561 (EF452493.1)reference genomes.Initial filtering of gene expression data was performed separately for the 8 h and 24 h timepoints using the Seurat R package (v4.2.0).Gene expression data for each biological replicate were combined into a single Seurat object.CHIKV reads were excluded from the gene expression matrices so they would not influence downstream processing (dimensionality reduction, clustering) of the mouse expression data.
Previously published scRNA-seq data for the 24 h timepoint was processed as previously described (27).CHIKV-low and -high cells were identified by filtering cells to only include those with >5 CHIKV reads.K-means clustering was then used to independently group each biological replicate into CHIKV-low and -high populations.Cells with 5 CHIKV reads or less were included in the CHIKV-low population.Cells were filtered based on the number of detected mouse genes (>250 and <6000) and the percent mitochondrial reads (<20%).Genes were filtered to only include those detected in >5 cells.Potential cell doublets were removed using the DoubletFinder (v2.0.3)R package using an estimated doublet rate of 10%.Due to the ability of CHIKV to inhibit host transcription, CHIKV-high cells with a low number of detected mouse genes (<250) or a high fraction of mitochondrial reads (>20%) were not filtered and remained in the dataset for downstream analysis.The sgRNA ratio (Figure S2B, D, E) was calculated by dividing the number of sgRNA (position 7567-12036) reads by the number of 5ʹ (position 1-7566) reads.To visualize this metric on UMAP projections (Figure S2B), before calculating the sgRNA ratio, a pseudo count of 1 was added to the sgRNA and 5' counts for each cell plotted (to eliminate division by 0).
Cells from the 8 h timepoint samples were filtered based on the number of detected mouse genes (>250 and <8000) and percent mitochondrial reads (<20%).Genes were filtered to only include those detected in >5 cells.The cell calls made by the cellranger pipeline (10X Genomics) for the 8 h CHIKV replicate 2 sample were not accurate based on analysis of UMI counts and likely included a substantial number of empty droplets.To account for this, a cutoff of 800 UMI counts was used to remove potential empty droplets.Counts from the CHIKV-capture libraries were then added to the object for all cells passing our filtering cutoffs.Due to the very few cells with detectable CHIKV RNA at 8 h, CHIKV + cells were classified as any cell with at least 1 CHIKVcapture read aligning to the CHIKV genome.
For both the 8 h and 24 h timepoints, mouse gene expression reads were normalized by the total mouse reads for the cell, multiplied by a scale factor (10,000), and log-transformed (NormalizeData).Normalized mouse counts were scaled and centered (ScaleData) using the top 2,000 variable features (FindVariableFeatures).The scaled data were used for PCA (RunPCA) and the first 40 principal components were used to identify clusters (FindNeighbors, FindClusters) and calculate uniform manifold approximation and projection (UMAP) (RunUMAP).
To ensure accurate and consistent cell type annotations, we integrated the 8 h and 24 h datasets based on timepoint (8 h and 24 h) and sample (mock-and CHIKV-infected) using the R package, Harmony (v0.1.1)(68).We then re-clustered the cells using the integrated data and generated an initial set of broad cell type annotations using the R package, clustifyr (v1.8.0) (69) and reference data from Immgen (70).These annotations were checked for accuracy and further refined using known cell type markers including, Cd19 (B cells), Cd3e (T cells), Hbaa1 (erythrocytes), Pdpn, and Pecam1.To identify perivascular cells (PvCs), fibroblasts were reclustered, integrated, and clusters were annotated using published reference data (20).To identify endothelial cell subsets, endothelial cells were re-clustered, integrated, and clusters were annotated using published reference data (15).LEC and BEC annotations were further refined using known marker genes including Marco, Pdpn, and Pecam1.Visualization of integrated UMAP projections suggest that broad LN cell type and endothelial cell type annotations were consistent across conditions (Supplemental Figure 2A-B).In addition, strong correlation with the published reference data (Supplemental Figure 2C) and expression of key endothelial cell marker genes (Supplemental Figure 2D) further support the accuracy of our cell type annotations.
Differentially expressed genes were identified for each cell type for mock vs 8 h and 8 h vs 24 h using the Seurat package.To allow for equal comparison with the 8 h timepoint, the top two replicates (based on cell number) from the 24 h timepoint were used for identifying differentially expressed genes.Genes were considered upregulated if the average log2 fold change was >0.15 for 8 h and >0.25 for 24 h for all replicates and the largest p-value for all replicates was <0.05.Gene ontology terms (Biological Process) were identified for the top 200 upregulated genes (sorted by maximum p-value for replicates) for each cell type using the R package, clusterProfiler (v4.4.4) (71).Terms were filtered to only include those with an adjusted p-value <0.05 and at least 3 or 10 upregulated genes overlapping the term for the 8 h and 24 h timepoints, respectively.Terms with <10 or >750 genes were excluded from the analysis.Terms identified for each cell type were combined and clustered into 5 modules based on the pairwise overlap between terms using the clusterProfiler package.Enrichment scores were calculated by dividing the fraction of upregulated genes overlapping the term by the fraction of background genes overlapping the term.For Figure 5C  Antigen acquisition by LNSCs.Antigen acquisition was evaluated using fluorescently labeled ovalbumin (ova) as previously described (24).Ovalbumin (A5503, Sigma-Aldrich) was decontaminated of lipopolysaccharide using a Triton X-114 detoxification method and tested with Pierce LAL chromogenic endotoxin quantitation kit (88282, Thermo Fisher Scientific).Ovalbumin was labeled using an Alexafluor 488 succimidyl ester labeling system (A20100, Thermo Fisher Scientific).Mice were inoculated with 20 µg AF488-labeled ovalbumin via intramuscular injection into both calf muscle (10 µg per calf), and popliteal and iliac LNs were collected for analysis of ova + LNSCs by flow cytometry.
Statistical Analysis.Non-sequencing data were analyzed using GraphPad Prism version 10.1.1.software.Data were evaluated for statistically significant differences using a two-tailed, unpaired t test, and either a one-way or two-way analysis of variance (ANOVA) test followed by Tukey's multiple comparison test.A P-value < 0.05 was considered statistically significant.
) although the magnitude of the effects was reduced compared to the popliteal LN, suggesting that the detrimental effects of WT CHIKV infection on LN structure and function are reduced as cells and antigen move farther downstream in the lymphatic drainage network.Consistent with the role of MARCO in alteration of LN LEC composition and enhanced inflammation during WT CHIKV infection, we observed robust antigen acquisition by LECs in WT CHIKV-infected MARCO -/-mice, similar to that observed in CHIKV 181/25-infected WT mice (Figure 7E-F), indicating that MARCO promotes the disruption of both LEC composition and function during CHIKV infection.Importantly, antigen acquisition by LECs in uninfected mice was MARCO-independent (Figure 7G-H), further indicating that CHIKV infection impairs LEC function via MARCO.Overall, these data suggest that pathogenic CHIKV infection impairs the ability of LECs to acquire antigen upon a secondary foreign challenge, which could have implications for the strength and success of downstream adaptive immunity to respond to that challenge.
in the LN.Diminished inflammatory gene expression in MARCO -/-mice was associated with reduced viral RNA accumulation in the dLN, suggesting MARCO may accelerate inflammation by facilitating viral RNA accumulation in LN LECs.Furthermore, we observed numerous CD11b + cells within the disrupted LN sinuses at 48 h post-infection, and the presence of MARCO contributed to the accumulation of inflammatory monocytes in the dLN.Despite the altered expression and spatial distribution of Lyve1 expression observed by confocal microscopy, total LEC, medullary, and floor LEC numbers remained similar in LNs from mock-, CHIKV 181/25-, and WT CHIKV-infected mice at 48 h post-infection when evaluated by flow cytometry.However, by 5 d post-infection, we detected decreased numbers of medullary and floor LECs in LNs from WT CHIKV-infected mice compared to those from mice infected with CHIKV 181/25, which induces less LN inflammation and does not disrupt LN cellular organization for single-cell mRNA sequencing.LNs were pooled into individual replicates (3 replicates; LNs from 5 mice pooled per replicate) and mechanically homogenized using a 22G needle in Click's medium (Sigma-Aldrich) supplemented with 5 mg/mL liberase DL (Roche, 05401160001) and 2.5 mg/mL DNase (Roche 10104159001) and incubated for 1 h at 37°C.Cell suspensions were enriched for CD45 -cells by labeling cells with PE-conjugated anti-mouse CD45 (30-F11), CD140A (APA5), and Ter119 (Ter119) monoclonal antibodies and subsequent depletion of PE-labeled cells using Miltenyi anti-PE microbeads (130-048-801) and MACS LS (130-042-401) columns according to the manufacturer's instructions with the following modifications: (i) We used 25% of the recommended volume of anti-PE microbeads, and (ii) we subjected the CD45 -enriched cell fraction to a second MACS LS column.Cells were enumerated using a hemacytometer.Cell fractions throughout the procedure were analyzed for cell depletion and enrichment of CD45 -cells by flow cytometry by staining with fixable LIVE/DEAD dye (Invitrogen, L34955) and antibodies against the following cell surface antigens: CD45 (30-F11), CD31 (390), PDPN (8.1.1),B220 (RA3-6B2), TCRβ (H57-597), CD11b (M1/70), and Ly6C (HK1.4) obtained from BioLegend, BD Bioscience, or eBioscience.Following staining, cells were washed and fixed in 1% paraformaldehyde (PFA)/1% FBS, and data were acquired on a BD LSR Fortessa X-20 flow cytometer.Data analysis was performed using FlowJo analysis software (Tree Star).

Study approval.
Animal experiments were performed with the approval of the Institutional Animal Care and Use Committee of the University of Colorado School of Medicine (Assurance Number: A3269-01) under protocol 00026.

Figure 2 .
Figure 2. WT CHIKV infection disrupts LEC marker expression and elicits infiltration of LN sinuses.(A-D) WT mice were mock-inoculated (n = 3) or inoculated in the footpad with 10 3 PFU

Figure 5 .
Figure 5. LNSCs exhibit a dominant pro-inflammatory response after 8 h post CHIKV

Isolation of cells from LNs and flow cytometry. LNs
and 5E, cell types are only shown if they have at least one upregulated gene overlapping any of the terms plotted.
Gene expression analysis by RT-qPCR.LNs were homogenized in TRIzol reagent (Life Technologies).RNA was isolated using the PureLink RNA Mini kit (Thermo Fisher) with oncolumn DNase treatment.Gene expression was quantified by RT-qPCR using Taqman gene expression assays (Thermo Fisher).Expression of each gene was normalized to 18S and analyzed as fold change over mock samples.