Lymph node homeostasis and adaptation to immune challenge resolved by fibroblast network mechanics

Emergent physical properties of tissues are not readily understood by reductionist studies of their constituent cells. Here, we show molecular signals controlling cellular, physical, and structural properties and collectively determine tissue mechanics of lymph nodes, an immunologically relevant adult tissue. Lymph nodes paradoxically maintain robust tissue architecture in homeostasis yet are continually poised for extensive expansion upon immune challenge. We find that in murine models of immune challenge, cytoskeletal mechanics of a cellular meshwork of fibroblasts determine tissue tension independently of extracellular matrix scaffolds. We determine that C-type lectin-like receptor 2 (CLEC-2)–podoplanin signaling regulates the cell surface mechanics of fibroblasts, providing a mechanically sensitive pathway to regulate lymph node remodeling. Perturbation of fibroblast mechanics through genetic deletion of podoplanin attenuates T cell activation. We find that increased tissue tension through the fibroblastic stromal meshwork is required to trigger the initiation of fibroblast proliferation and restore homeostatic cellular ratios and tissue structure through lymph node expansion.

U nlike developmental systems of progressive tissue growth and maturation 1 , the homeostatic state of adult tissues is robust, maintaining form and function 2 . Secondary lymphoid organs are uniquely able to dramatically change size in response to immune challenge, adapting to the increased space requirements of infiltrating and proliferating lymphocytes while remaining structurally and functionally intact [3][4][5] . As a mechanical system, lymph nodes (LNs) continually resist and buffer forces exerted by lymphocytes entering and leaving the tissue 6 , and managing diurnal fluctuations in cell trafficking 7 . Tissue size is determined by lymphocyte numbers, highlighted by the small organ size in genetic models blocking lymphocyte development (Rag1 knockout (KO)) 8 , and the ability of the tissue to expand two-to tenfold in size to accommodate lymphocyte proliferation through adaptive immunity [3][4][5]9 . Here, we ask whether mechanical forces determine the kinetics of LN tissue remodeling through immune challenge.
LNs function at the interface of immunity and fluid homeostasis, constructing a physical three-dimensional cellular meshwork linking fluid flow, immune surveillance, and adaptive immunity 10 . The most populous stromal cell component are FRCs, which span the whole tissue generating an interconnected cellular network with small-world properties 11 , forming robust clustered nodes with short path lengths and surrounding bundles of extracellular matrix fibers 12 . It is widely assumed that extracellular matrix scaffolds are the predominant force-bearing structures determining tissue mechanics 1,2 . However, the relative contributions of the cellular structures versus the underlying extracellular network to tissue mechanics have not been addressed in a highly cellularized system undergoing such extensive expansion 2 . As LNs expand in response to immune challenge, the tissue becomes more deformable 3 . During LN expansion, the FRC network maintains connectivity through the elongation and increased spacing between FRCs, increasing mesh size of the network 3 . It is also known that CLEC-2 + dendritic cells are required to prime the stromal architecture for tissue expansion and affects LN deformability 3,4 , but the downstream impacts on the mechanical properties of the cellular network and extracellular matrix scaffolds driving the adaptation in tissue mechanics are unknown. During tissue expansion, FRCs reduce their adhesion to the underlying extracellular matrix bundles, and these matrix scaffolds become fragmented 9 . This makes LNs an ideal model system to address the relative mechanical contributions of cellular and material structures to emergent tissue mechanics.
Using molecular cell biology approaches in combination with quantitative biophysics, we asked how LN tissue mechanics is controlled and what impact tissue mechanics has on adaptive immune responses. We provide evidence that tissue tension is offset and sensed by the fibroblastic reticular network through actomyosin structures. FRCs require PDPN expression to regulate network tension through acute LN expansion, and signaling downstream of PDPN determines FRC membrane tension and regulates their response to external forces. We find that tension through the FRC network gates fibroblast proliferation and that the failure of FRCs to sense increasing forces inhibits T cell activation and proliferation. These results show that adaptive immune responses are regulated through mechanical cues sensed by the stromal architecture and identify PDPN as a key mechanical sensor in fibroblasts.

Results
The FRC network is under tension in steady-state LNs. Because fibroblasts are contractile, force-generating cells 13 , we hypothesized that the FRC network determines LN tissue mechanics. The FRC network spans the whole LN and specifically supports CD3 + T cell function in the paracortex, providing trafficking routes from high endothelial venules to B cell follicles [14][15][16][17] (Fig. 1a,b). We asked how the FRC network and associated extracellular matrix participate in tissue mechanics in the immunological steady state. We used a fibroblast-specific membrane-EGFP mouse model (Pdgfra-mGFP-CreERT2) ( Fig. 1c and Extended Data Fig. 1) to visualize the fine cellular connections through the FRC network (Fig. 1c). Following laser ablation (Fig. 1c,d, Extended Data Fig. 1e, and Supplementary Video 1), we measured a mean recoil of 0.42 μm s −1 in the FRC network of the paracortex, formally demonstrating that the reticular network is under mechanical tension (Fig. 1c,d). The severed network recoiled in all directions adjacent to the ablation site, suggesting that mechanical forces are buffered throughout the network. To investigate how cellular-scale stromal components contributed to tissue mechanics, we examined the cytoskeletal and extracellular matrix structures in the reticular network. In steady state, FRCs adhere to and enwrap the bundled collagen of the underlying conduit 18 . Therefore, recoil following laser ablation is a combined mechanical measurement of the cell and the extracellular matrix structures (Fig. 1c-f). Maximum z-projections and orthogonal views of FRCs show F-actin cables aligned proximal to the conduit and beneath the T cell-facing FRC plasma membrane ( Fig. 1f and Supplementary Video 2). These F-actin structures colocalized with phosphorylated myosin regulatory light chain (pMLC2) 19 (Fig. 1f), indicating that FRCs generate contractility and strain in steady state 20 .
FRCs adapt to external forces throughout immune challenge. We asked how the FRC network reacts to forces exerted by lymphocytes recruited to and proliferating in the tissue following immune challenge. Immunization with incomplete Freund's adjuvant (IFA) and ovalbumin (OVA) causes LNs to expand two-to threefold over the first 5 days (Fig. 2a and Extended Data Fig. 1f). As a proxy measurement of external forces exerted, we quantified T cell packing and the spacing of network fibers through acute LN expansion. We found that after immunization, the density of CD3 + cells within tissue increased just 3 days after immunization, and we quantified 10-20% more T cells per 100 μm 2 (Fig. 2b,c). This increase in T cell packing density was maintained at day 5 (Fig. 2b,c), suggesting that the FRC network would experience higher external forces following  Fig. 2 | FRCs adapt to changing external forces throughout immune challenge. a, LN mass change after IFA/OVA immunization. One-way analysis of variance (ANOVA) with Tukey's multiple comparisons, ****P = 1.00 × 10 −6 . Each point represents one LN at day 0 (n = 10), day 3 (n = 11), and day 5 (n = 12). b, T cell packing in the LN during inflammation for PDPN (FRCs, yellow), CD3 (T cells, blue), and DAPI (nuclei, gray). Scale bars, 25 µm. c, Quantification of CD3 + T cell nuclei per 100 µm 2 . Box plot indicates median, interquartile range, and minimum/maximum. Dotted line represents median of day 0. One-way ANOVA with Tukey's multiple comparisons, *P = 0.0362, ****P = 0.0003. n indicates individual image ROI on day 0 (n = 17), day 3 (n = 15), and day 5 (n = 20) from three independent LNs. d, FRC gap analysis during inflammation. PDPN (FRCs), binary, and circle overlay. Scale bars, 50 µm. e, Quantification of circle radius. Box plot indicates median, interquartile range, and minimum/maximum. Dotted line represents median of day 0. One-way ANOVA with Tukey's multiple comparisons, ****P = 2.00 × 10 −6 . Each point represents a circle radius. n indicates individual image ROI on day 0 (n = 12), day 3 (n = 15), and day 5 (n = 15) from three independent LNs. f, Laser ablation of the FRC network throughout inflammation. PDGFRα + mGFP + (FRCs) ablation ROI (white box) and cut site (red dotted line). Scale bars, 50 µm. Recoil displacement (arrowheads) with pre-(green) and postcut (magenta) overlay. Scale bars, 10 µm. g, Recoil curves of network displacement (μm) (mean ± standard error of the mean (s.e.m.)). h, Initial recoil velocity (μm s −1 ) after IFA/OVA immunization. Box plot indicates median, interquartile range, and minimum/maximum. Kruskal-Wallis test with Dunnett's test, *P < 0.05. g,h, n indicates an ablation on day 0 (n = 30), day 3 (n = 37), and day 5 (n = 32) for three independent experiments.
immune challenge. We next quantified the spacing of the FRC network fibers using our gap analysis algorithm 3 . Consistent with previous observations, we find that the spaces between network fibers were increased as the FRC network stretches (Fig. 2d,e) and that FRC network integrity and connectivity remains intact. If the FRC network balances mechanical forces in LNs, then we would expect tension in the FRC network to change throughout LN expansion. Surprisingly, at day 3 after immunization, initial recoil velocity decreased by 29% to 0.30 μm s −1 (Fig. 2f-h and Supplementary Videos 3 and 4) despite a 1.5-fold increase in tissue mass (Fig. 2a) and increased packing of T cells (Fig. 2b,c). In contrast, at day 5 after immunization, mean initial recoil velocity was 60% higher than in the steady state at 0.71 μm s −1 (Fig. 2f-h and Supplementary Videos 3 and 5). Therefore, FRC network tension did not correlate with either T cell packing density or network stretching, indicating that tissue tension is not solely determined by external forces and that cell-intrinsic factors must also impact FRC network mechanics.
Actomyosin contractility sets FRC network tension. We next sought to determine the cellular determinants of FRC network tension. Laser ablation methods cannot discriminate between cellular and extracellular matrix components 21,22 , as the FRC network is tightly associated with the underlying conduit 18,23,24 . We therefore examined the cytoskeletal structures in FRCs within the tissue context and their relationship to the underlying matrix through acute immune challenge. We compared cytoskeletal and matrix structures of FRCs at day 3 (Fig. 3a, lower tension) versus day 5 (Fig. 3b, higher tension). The majority (75-85%) of F-actin structures were located proximal to the basement membrane (asterisks, Fig. 3a We have previously reported that during acute LN expansion, the conduit network becomes fragmented 9 , whereas the cellular network remains connected 3,4,11 . We reported that polarized microtubule networks via LL5β control FRC-matrix deposition during early LN expansion 9 . Laminins, collagens, and basement membrane components become fragmented as LNs expand causing disruption of conduit flow 9 . We compared the conduit matrix using second harmonic signals at day 3 and day 5 after immune challenge (Extended Data Fig. 2) and found that disruption in the conduit occurred at day 5, correlating with increased tension. We measured increased fibril thickness in areas of disruption at day 5 (Extended Data Fig. 2) and that fibers appeared straighter in draining LNs, consistent with fibers being pulled taught before breaking. To determine whether FRCs formed connections in the absence of underlying matrix, we genetically labeled individual FRCs via sporadic Cre-recombinase activation in Pdgfra-mGFP-CreERT2 mice to express the R26R-Confetti 25 conditional allele (Pdgfra iR26R-Confetti ). We observed individual FRCs spanning sections of disrupted conduit (Fig. 3d), suggesting that the matrix structure may not contribute to the increased network tension at day 5 ( Fig. 2f-h) and that the intact cellular network balances tissue tension. Contractility and tension through actin filaments is regulated through interactions with myosin 26,27 . We quantified myosin light chain 2 phosphorylation (T18, S19), within PDPN + structures, to determine contractile cables in FRCs stained positive for pMLC (Fig. 3e). However, at day 3, pMLC + F-actin cables were reduced and in some areas absent (Fig. 3e). The pattern of reduced actomyosin contractility in FRCs at day 3 mirrored the reduced tissue tension measured at the same time point (Fig. 2f-h). We therefore hypothesized that the actomyosin cytoskeleton is a major contributor to FRC network tension. We tested this pharmacologically using the ROCK inhibitor Y27632 to inhibit phosphorylation of myosin light chain. Pre-treatment of LNs with ROCK inhibitor reduced tissue tension to basal levels in steady state and following immunization (Fig. 3f,g and Extended Data Fig. 3a-d). Following ROCK inhibitor treatment (Fig. 3h), FRCs remained connected and F-actin cables were visible (arrowheads, Fig. 3h,i), but the proportion of contractile pMLC + F-actin cables in both steady-state and immunized LNs was reduced (Fig. 3j), confirming the correlation between actomyosin contractility and FRC network tension.
These data show that mechanical forces are generated by increased packing of lymphocytes in the FRC meshwork and resisted by actomyosin through the FRC network (Fig. 2). However, there is not a simple linear relationship between the increases in external forces caused by T cell packing and tissue tension measured through the FRC network. Tissue tension varies throughout responses to immunogenic challenge, and we find that increased tissue tension at day 5 occurs independently of extracellular matrix integrity. We have previously reported that 5 days after immunization, the whole, intact LN is more deformable 3 , yet we now measure increased tension through the fibroblastic reticular network increases ( Fig. 2f-h). Because the elasticity of gels is known to scale as the inverse of mesh size 28 , the combination of the fragmentation of the matrix (Fig. 3c,d) and increasing mesh size (Fig. 2d,e) of the FRC network may explain increased LN deformability through acute expansion.
FRCs must elongate contact with their network neighbors as they spread and expand the stromal architecture 11,34 . FRC elongation could be achieved by increasing the surface area to cell volume ratio through exocytotic pathways or unfolding membrane reservoirs 35,36 . We found that FRCs contain active EHD2 + caveolae structures 37 that could contribute to tension-sensitive plasma membrane reservoirs (Fig. 4d). We tested, using osmotic shock (Extended data Fig. 5a,b), whether regulation of membrane tension through the CLEC-2/PDPN signaling axis impacted how rapidly FRC uses existing membrane reservoirs (Fig. 4e,f). We found that CLEC-2 engagement (Fig. 4e, Extended Data Fig. 5c, and Supplementary Videos 9 and 11) or knockdown of PDPN (PDPN KD) ( Fig. 4f and Supplementary Video 10) permitted more rapid cell expansion in hypotonic conditions but did not alter total membrane availability (Extended Data Fig. 5d and Supplementary Video 12). In vivo, we observed differentially labeled FRCs in Pdgfra iR26R-Confetti mice extending cell-cell contacts at day 3 ( Fig. 4g), suggesting morphological adaptations for cell extensions. We therefore conclude that CLEC-2/PDPN binding plays a dual role in FRC network remodeling, changing cell surface mechanics of FRCs and reducing actomyosin contractility 3,4 ( Fig. 3f), which together permit LN expansion while maintaining FRC network integrity. FRCs are mechanically sensitive in vitro. Existing studies report that FRC proliferation lags behind lymphocytes 3-5 , but it is not known how FRC proliferation is triggered or spatially regulated 5 . Five days after IFA/OVA immunization, the number of proliferating FRCs (Ki67 + ) doubled compared to steady state (Fig. 5a,b). Ki67 + FRCs were sporadically located throughout the paracortex, and we observed no specific proliferative niche surrounding blood vessels or beneath the capsule (Fig. 5a). This finding suggested that entry into the cell cycle is not spatially regulated or limited to an FRC subpopulation. Five days after immunization, pressure from T cells (Fig. 2a-c) increased FRC network tension ( Fig. 2f-h), which is resisted by actomyosin contractility (Fig. 3f). We hypothesized that increased mechanical tension may gate FRC proliferation. Because cell responses in tissues are driven by a complex combination of chemical and mechanical cues, we measured FRC proliferation in response to mechanical stimuli alone in vitro using a reductionist system of polyacrylamide gels ranging in stiffness 38 (Fig. 5c). Using automated tracking of FRC nuclei, we found that FRCs proliferated more rapidly on stiffer substrates (Fig. 5c,d), indicating that FRC proliferation is mechanically sensitive. Inhibition of ROCK using Y27632 (Fig. 5e), which blocked increased tissue tension in response to immune challenge in vivo, also blocked the mechanically determined changes to FRC proliferation in vitro (Fig. 5e). Knockout of PDPN expression also attenuated the impact of increased stiffness on FRC proliferation (Fig. 5f). These data show that even in the absence of cell-cell communication with immune cells and soluble chemical cues in the tissue, FRC proliferation is mechanically sensitive and stimulated by stiffer external environments (Fig. 5c,d), sensed by PDPN signaling and actomyosin contractility (Fig. 5e,f).
We have previously reported that PDPN is upregulated in FRCs following immune challenge 32 by CLEC-2 binding 32 . Additionally, PDPN expression in immortalized FRC cell lines, cultured on rigid plastic, is higher than in primary cells 3 , but the regulation of PDPN expression remains poorly understood. Because we have measured increased tissue tension during acute LN expansion, which correlates with increasing PDPN expression, we investigated whether PDPN expression might be mechanically regulated. We measured cell surface expression of PDPN on primary FRCs (CD140α + CD31 − ) using flow cytometry after cell culture on rigid plastic and found that inhibiting actomyosin using Y27632 reduced PDPN expression by approximately 30% after 3 days of culture and more (75%) after 6 days (Fig. 5g,h). These data indicate not only that PDPN is a mechanical sensor in FRCs in vitro but also that PDPN expression is impacted by mechanical forces.

PDPN deletion blocks FRC mechanical adaptation in vivo.
We next sought to test the role of PDPN in LN mechanics in vivo. Mice with full knockout of PDPN fail to develop LNs and die shortly after birth due to circulatory defects, so PDPN function has not been tested specifically in FRCs 39 . We generated Pdgfra-mGFP-CreERT2 × Pdpn fl/fl40 mice (Pdgfra mGFPΔPDPN ) to conditionally delete PDPN in FRCs (Fig. 6a). FRCs are constitutively labeled with membrane-targeted GFP (Extended Data Fig. 1), which in control mice (Pdgfra-mGFP-CreERT2) also express PDPN (Fig. 6b). Reduced PDPN expression was quantified by flow cytometry in FRCs in steady-state LNs of Pdgfra mGFPΔPDPN mice (Fig. 6b,c). We saw no gross alterations to FRC network structure 7 days following tamoxifen treatment (Fig. 6b).
We next immunized control and Pdgfra mGFPΔPDPN mice (Extended Data Fig. 6) and found that mice lacking PDPN + FRCs exhibited attenuated acute LN expansion (Fig. 6d), phenocopying CD11c ΔCLEC2 mice 3 . We examined the FRC network structure in Pdgfra mGFPΔPDPN reactive LNs. The constitutive expression of membrane-targeted EGFP enabled us to identify FRCs independently of PDPN expression to directly compare PDPN + and PDPN − FRCs side by side in situ (Fig. 6e). We observed F-actin cables along the underlying conduits, adjacent to the basement membrane (perlecan), in both PDPN + (asterisk) and PDPN − (arrowhead) FRCs (Fig. 6f), and we observed no clear difference in pMLC staining of F-actin cables in Pdgfra mGFPΔPDPN LNs in steady state (Fig. 6f). We next tested the role of PDPN in LN mechanics through immune challenge using laser ablation. In contrast to our experiments using ROCK inhibitor (Fig. 2), deletion of PDPN did not alter FRC network tension in the steady state. However, Pdgfra mGFPΔPDPN LNs did not exhibit lower network tension at day 3 or increased tension at day 5 ( Fig. 6g-i and Supplementary Videos 13 and 14), indicating that PDPN is required for the FRC network to adapt to mechanical forces through immune challenge. We conclude from these data that PDPN functions as a mechanical sensor for remodeling of the FRC network in vivo.

FRC mechanics impacts immune outcomes in vivo.
Because LN expansion is an integral part of adaptive immunity 10,24,41 , we asked how LN tissue mechanics impacted immune outcomes. LN expansion was significantly attenuated in Pdgfra mGFPΔPDPN following immunization with CFA/OVA measured by both tissue mass (Fig. 7a) and cellularity (Fig. 7b). PDPN deletion in FRCs resulted in both fewer stromal and immune cell populations 5 days after immunization (Fig. 7c). Both B cell and T cell populations were reduced (Fig. 7d,e), but the ratio of T cells to B cells was not affected (Fig. 7f). Because FRCs primarily support T cell populations in the paracortex 42 , we quantified the T cell subsets ( Fig. 7g and Extended Data Fig. 7). We found that naive CD4 + and naive CD8 + cells were similarly increased in LNs 5 days after immunization in both control and Pdgfra mGFPΔPDPN LNs, suggesting that recruitment and trapping of naive lymphocytes (CD62L + , CD44 − ) from the circulation was unaffected by PDPN deletion in FRCs (Fig. 7h-j). Upregulation of CD25 was not significantly affected (Extended data Fig. 8a), suggesting that antigen presentation was not inhibited. However, approximately 50% fewer CD4 + and CD8 + cells expressed CD44 following immunization in Pdgfra mGFPΔPDPN LNs, and effector memory T cells (CD62L − , CD44 + ) were specifically constrained (Fig. 7h-j). These data suggest that FRCs in Pdgfra mGFPΔPDPN LNs lack the ability to adapt and stretch to accommodate increasing T cell numbers and as a result constrain effector T cell populations.
Because deletion of PDPN in FRCs attenuated LN expansion, we also examined the impact of PDPN-dependent mechanical perturbation on stromal populations. We found that blood endothelial cells proliferated similarly in control and Pdgfra mGFPΔPDPN LNs, suggesting that angiogenesis in the LN is not dependent on mechanical signals in FRCs (Fig. 7k). However, the number of lymphatic endothelial cells was reduced (Fig. 7l). When we examined FRCs, gating on either GFP + or alternatively CD140α + CD31 − stroma, we measured approximately 50% fewer FRCs 5 days after immunization in Pdgfra mGFPΔPDPN LNs (Fig. 7m). The reduction in FRCs could be a cause or a consequence of the attenuated LN expansion (Fig. 7a,b). However, as FRC proliferation is mechanically sensitive in vitro (Fig. 5), we hypothesize that reduced FRC network tension in Pdgfra mGFPΔPDPN mice (Fig. 6g,h) may inhibit FRC proliferation in vivo.

Mechanical tension gates FRC proliferation in vivo.
We have shown that PDPN is a mechanical sensor in FRCs impacting cell-intrinsic mechanics and morphological adaptation in response to immune challenge (Figs. 4 and 5). However, the CLEC-2/PDPN signaling axis regulates an array of immunologically relevant pathways in FRCs in addition to cell mechanics, dendritic cell migration 43 , ECM deposition 9 , and high endothelial venule function 16 , and we cannot definitively conclude that the impact on immune outcome we measure in Pdgfra mGFPΔPDPN LNs results solely from mechanical perturbation. Therefore, to examine the specific impact of FRC network mechanics on adaptive immune function, we manipulated tissue tension through pharmacological inhibition of actomyosin contractility. We reduced tissue tension in vivo through pharmacological inhibition of ROCK 19 in wild-type (C57BL/6 J) mice (Fig. 8a). We found that in contrast to Pdgfra mGFPΔPDPN mice, increases in tissue mass and cellularity were unaffected by ROCK inhibition (Fig. 8b). Total immune cell numbers (CD45 + ) were unaffected, but there was a trend toward a reduction in stromal populations at day 5 after immunization (Fig. 8c). B cell and T cell numbers were similar between PBS-treated controls and Y27632-treated samples, and T cell activation was also unaffected (Fig. 8d,e and Extended Data Fig. 9a,b). We examined the expression of Ki67 in T cells within the tissue context and confirmed that the initiation of T cell proliferation was not affected by actomyosin inhibition (Extended Data Fig. 8b,c). However, we found that inhibition of actomyosin selectively constrained T cell proliferation, changing the T cell to B cell ratio in reactive LNs (Fig. 8f). Proliferation of endothelial cells was unaffected by actomyosin inhibition, whereas the FRCs were significantly reduced 5 days after immunization ( Fig. 8g and Extended Data Fig. 9c,d). This leads us to conclude that the fibroblastic stromal architecture is reactive to the physical space requirements of lymphocytes. Indeed, previous studies identified a robust ratio between fibroblastic stroma and T cells, maintained as the tissue expands 5 . A mechanical cue for stromal cell proliferation would maintain the steady-state ratio of fibroblastic stroma to lymphocytes independently of the kinetics or scale of the immune reaction, ensuring a supportive immune microenvironment for lymphocyte populations 42 . We examined the relationship between FRC network tension and FRC proliferation in steady-state and reactive LNs and how these were altered by mechanical pathways (Fig. 8h). We found that higher network tension coincides with higher FRC number and that this association can be disrupted by either inhibition of actomyosin contractility (Fig. 3) or deletion of PDPN from FRCs (Fig. 6). Together, our data demonstrate that the fibroblastic structure of the LN is the active mechanical component during tissue expansion. Using the dynamic cellular network rather than the more rigid ECM to respond to changing lymphocytes numbers in the tissue is an elegant mechanical system that can proportionately respond to lymphocyte requirements.

Discussion
Other studies have shown that tissue scale properties emerge from cellular scale mechanics in the transition from developing to adult tissues 44 . We have directly addressed this concept in immunologically relevant adult mammalian tissue during homeostasis and immune challenge. We show that the fibroblastic reticular cellular network deploys molecular signals controlling cellular mechanical properties to collectively determine tissue scale mechanics of LNs.
We have quantified tissue tension through the FRC network. Because by definition forces must be balanced in steady state 2 , these data gave us an indirect measurement of the forces exerted by lymphocytes. Upon immune challenge, LNs expand to accommodate increasing numbers of lymphocytes, first trapped from the circulation and then proliferating in response to antigen-specific activation. We quantified increased T cell packing and FRC network spacing during LN expansion, but unexpectedly, these increases did not correlate with increased tension through the FRC network. The uncoupling of tissue size with tissue tension at day 3 after immunization provided evidence that FRC network tension is not solely determined by the external forces. Rather, FRCs can actively and intrinsically adapt their cellular-scale mechanics, and we show that the LN becomes mechanically permissive to expansion through PDPN signaling. Box plots indicate median, interquartile range, and minimum/maximum. One-way ANOVA with Tukey's multiple comparisons, ****P = 1.00 × 10 −6 , ***P = 0.000567, **P = 0.002087. n indicates LNs at day 0 (n = 12), day 0 ΔPDPN (n = 12), day 5 (n ≥ 6), and day 5 ΔPDPN (n = 8) over two independent experiments. c-e, CD45 + , CD45 − , CD19 + and CD3 + cell numbers. Box plots indicate median, interquartile range, and minimum/maximum. One-way ANOVA with Tukey's multiple comparisons, ****P = 1.00 × 10 −6 , ***P < 0.001, **P = 0.00989, *P = 0.027535. f, Ratio of CD19 + and CD3 + cell numbers. Box plots indicate median, interquartile range, and minimum/maximum. One-way ANOVA with Tukey's multiple comparisons, ****P = 1.00 × 10 −6 . Each point represents one LN. g, Flow cytometric gating. Representative dot plots and percentages of CD45 + , CD45 − , CD19 + , CD3 + and CD3 + CD4 + and CD3 + CD8 + subpopulations. h, Representative flow cytometric gating comparing control and Pdgfra mGFPΔPDPN CD3 + CD4 + , CD3 + CD8 + subpopulations. i,j, CD3 + CD4 + and CD3 + CD8 + subpopulation cell numbers. Box plots indicate median, interquartile range, and minimum/maximum. One-way ANOVA with Tukey's multiple comparisons, ****P = 1.00 × 10 −6 , **P = 0.0058, *P < 0.05. Each point represents one LN. k-m, Stromal cell numbers for blood endothelial cells (BECs) (k), lymphatic endothelial cells (LECs) (l), and GFP + and CD140α + FRCs (m). Box plots indicate median, interquartile range, and minimum/maximum. One-way ANOVA with Tukey's multiple comparisons, ****P = 1.00 × 10 −6 , **P = 0.0043, *P < 0.05. Fig. 6 | Deletion of PDPN in vivo attenuates mechanical adaptation of the FRC network in response to immune challenge. a, Tamoxifen and immunization strategy for Pdgfra mGFPΔPDPN mice. Draining LNs (Dr) and non-draining LNs (nDr). b, LN paracortex maximum z-projection. GFP (Pdgfra mGFP , green), PDPN (FRCs, magenta). Scale bars, 50 µm (zoom, 30 µm). c, Representative histograms of surface protein expression for PDPN in control and Pdgfra mGFPΔPDPN steady-state LNs. d, LN mass after CFA/OVA immunization. Box plot indicates median, interquartile range, and minimum/maximum. Two-way ANOVA with Dunnett's test, *P = 0.030. n indicates LNs on day 0 (n = 10), day 0 ΔPDPN (n = 9), day 3 (n = 6), day 3 ΔPDPN (n = 5), day 5 (n = 4), and day 5 ΔPDPN (n = 4). e, Representative image of GFP + (green), PDPN − (magenta) FRC cell body in day 5 after immunization in Pdgfra mGFPΔPDPN LNs from two independent experiments. Scale bar, 25 µm. f, Representative images of actomyosin and ECM structures within control and Pdgfra mGFPΔPDPN FRCs 5 days after immunization. Arrowheads and asterisks mark PDPN − and PDPN + FRCs respectively; perlecan (magenta), PDPN (yellow), F-actin (cyan), and pMLC (red). Scale bars, 10 µm. g, Recoil curves of network displacement (µm) (mean ± SEM) for control and Pdgfra mGFPΔPDPN mice. h, Initial recoil velocity (µm s −1 ) after CFA/OVA immunization in control and Pdgfra mGFPΔPDPN mice. Box plot indicates median, interquartile range, and minimum/maximum. Two-way ANOVA with Sidak's multiple comparisons, *P < 0.05, ****P = 1.00E^− 6 . Each point represents an ablation. i, Laser ablation of the FRC network throughout inflammation in control and Pdgfra mGFPΔPDPN mice. PDGFRα + mGFP + (FRCs) ablation ROI (white box) and cut site (red dotted line). Scale bars, 50 µm. Recoil displacement (arrowheads) with pre-(green) and postcut (magenta) overlay. Scale bars, 10 µm. g-i, n indicates ablation at day 0 (n = 48), day 0 ΔPDPN (n = 44), day 3 (n = 18), day 3 ΔPDPN (n = 18), day 5 (n = 28), and day 5 ΔPDPN (n = 21) over three independent experiments. Through this study, we addressed the relative contributions of inert extracellular matrix versus cell structures to tissue mechanics. Because the extracellular matrix of the LN conduits become disrupted through acute LN expansion 9 , leaving only the cellular network intact, we are able to conclude that the cellular structures are sufficient to resist and remodel in response to the forces of LN expansion. It is a common assumption that the matrix structures of tissues provide physical guidance for cellular organization 2 . Here, we question that assumption and show that the cytoskeleton and cellcell connections are sufficient to maintain and remodel LNs during tissue expansion. Cell-matrix adhesion in other tissue contexts is reinforced via forces through the cytoskeleton 45 . We conclude that reduced FRC-matrix adhesion would occur during the acute phase of LN expansion as a consequence of reduced FRC contractility. Indeed, we have previously reported that inhibition of actomyosin contractility in FRCs reduces focal complexes and impacts the tethering of microtubules to sites of cell-matrix adhesion 9 . In future studies, we should now consider the nature of the cell-cell junctions in the FRC network, as these may also play a key role in maintaining FRC network integrity and contribute to tissue mechanics.
Our study provides evidence that FRC proliferation in vivo is mechanically sensitive. A mechanical cue resolves the conflicting reported kinetics 3-5 of FRC proliferation observed with different adjuvants, as an immune challenge causing rapid increases in lymphocyte numbers would increase FRC network tension sooner and induce FRC proliferation earlier. A mechanical trigger for proliferation is also consistent with our observations that there is no specific niche of proliferative FRCs around blood vessels or under the capsule; instead, proliferative FRCs were seen sporadically throughout the FRC network. A mechanical mechanism is an ideal measurement system to ensure that the ratio of FRCs to lymphocytes is maintained independently of LN size. What remains unresolved is how neighboring FRCs maintain connections and how forces might be transmitted through these unstudied cell junctions.
proliferation. Deletion of PDPN on the other hand did not alter steady-state tissue tension. Instead, we find that PDPN expression by FRCs is required to adapt FRC cell-intrinsic mechanics, leading us to conclude that PDPN is a key mechanical sensor. Targeting FRC network mechanics through either actomyosin contractility or PDPN expression attenuated FRC proliferation in vivo. However, when we compared the impact of these mechanical perturbations on T cell activation, we found that only PDPN deletion impacted LN expansion and T cell activation and constrained T cell proliferation. We suggest that by inhibiting actomyosin contractility directly, we are able to permit sufficient stretching of the FRC network to allow space for lymphocyte proliferation. In contrast, PDPN is required for FRCs to adapt and stretch, and failure to do so constrains LN expansion. Because it is known that PDPN has additional functions in FRCs (as a ligand for dendritic cell trafficking 43 , roles in maintaining high endothelial venule integrity through crosstalk with platelets 16 , and as a binding partner for the key chemokine CCL21 46 ), our in vivo studies here are intentionally short-term to specifically test the function of PDPN in tissue mechanics. In future studies, it will be interesting to test the role of PDPN in longer-term assays and ask what other tissue functions require this key signaling molecule. It will also be important to examine the role of PDPN in lymphoid tissue function in disease.
There are several missense mutations reported in PDPN in patients with diffuse large B cell lymphoma (https://www.cbioportal.org), and it is possible that PDPN expression levels and function are relevant to other pathological states.
In summary, this study further highlights the essential role for fibroblastic stroma in lymphoid tissue homeostasis 41 . Beyond their known roles in lymphocyte trafficking and production of growth and survival factors for lymphocyte activation, we now show that FRCs are also key cellular mechanical sensors. How external mechanical forces cooperate with cell intrinsic physical properties to control cell and tissue function in vivo is a relatively new research field 47 . We address this paradigm to show that FRCs can change their cell-intrinsic mechanical properties in response to changing external forces, allowing the tissue to maintain structure and function through rapid expansion in response to immune challenge. This brings together molecular and cell biology with biophysics approaches to provide mechanistic insights into lymphoid tissue remodeling and integrates tissue mechanics into our understanding of immune function.

online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41590-022-01272-5.

Methods
Mice. Experiments were performed in accordance with national and institutional guidelines for animal care and approved for the Laboratory for Molecular and Cell Biology by the institutional animal ethics committee review board, European research council, and the UK Home office. Breeding of animal lines were maintained off site at Charles River Laboratory. Wild-type C57BL/6 J mice were purchased from Charles River Laboratories. Generation of novel mouse model of Pdgfra-mGFP-CreERT2 was designed as follows: EGFP with a membrane tag (N-terminal 0-20 amino acids of neuromodulin GAP-43) was inserted into the Pdgfra (CD140α) gene locus in combination with a CreERT2 cassette (linked to mGFP with a P2A self-cleavage peptide 48 ) using Cyagen CRISPR-Cas9 technology. Pdgfra-mGFP-CreERT2 mice were crossed with R26R-Confetti 25 or Pdpn fl/fl animals 40 to generate Pdgfra-mGFP-CreERT2 x R26R-Confetti (Pdgfra iR26R-Confetti ) and Pdgfra-mGFP-CreERT2 x Pdpn fl/fl (Pdgfra mGFPΔPDPN ) mouse models, respectively. Activation of the Cre recombinase was achieved through the administration of tamoxifen (22 mg ml −1 ) resuspended in corn oil. Tamoxifen was dosed (82 µg g −1 ) intraperitoneally on 3 consecutive days a week before immunization. Females and males aged between 8 and 15 weeks were used for experiments, unless stated otherwise. Animals were assigned experimental groups at random. Immunization model. Mice were immunized via subcutaneous injection into the right flank, proximal to the hip, with 100 μl of an emulsion of OVA with IFA or complete Freunds adjuvant (CFA) where stated (100 µg OVA per mouse) (Hooke Laboratories). Where stated, mice were treated with Y27632 (Tocris, 1254). Next, 10 µl 1 mg ml −1 Y27632 dissolved in PBS was injected subcutaneously into the right flank, proximal to the hip; 10 µl sterile PBS was used as injection control. Y27632/ PBS injections were given on 3 consecutive days 24 h after IFA/OVA injection. After 5 or 3 days, mice were culled, and inguinal LNs from both flanks (naive and inflamed) were extracted for paired histological studies, flow cytometry analysis, or ex vivo laser ablation.

Ex vivo cultures.
Preparation for ex vivo live LN laser ablation was optimized following previously established methods 14,[49][50][51] . UltraPure low-melting-point agarose (Thermo Fisher Scientific) was prepared in PBS at 3% w/v and maintained at 37 °C. LNs were embedded into agarose and left for 5 min to set on ice. LN blocks were secured by superglue to cutting stage and placed into ice-cold PBS cutting chamber. Leica Vibratome (VT1200S) 200-μm-thick sections were cut at a rate of 0.3 mm s −1 and 1.5 mm amplitude until the LN was completely sectioned. Collected sections were placed into RPMI 1640 containing 10% fetal bovine serum, 1% penicillin and streptomycin (P/S) (Thermo Fisher Scientific, 15140122), and 1% insulin-transferrin-selenium (ITS) (Thermo Fisher Scientific, 41400045) at 37 °C, 10% CO 2 and imaged within 24 h. Live sections recovered for 1 h before being used for live imaging.
Laser ablation. LN sections were transferred to glass-bottom 35-mm MatTek dishes (P35G-1.5-20-C) with a small volume of RPMI media containing 1% P/S and 1% ITS. A glass coverslip was placed on top to secure the section. Sections imaged on a Zeiss LSM 880 inverted multiphoton microscope with the imaging chamber maintained at 37 °C, 10% CO 2 . Where stated, sections were treated with 100 µM Y27632 (Tocris, 1254) diluted in ex vivo culture media for at least 1 h before imaging and ablation. Sections were imaged with a Plan-Apochromat ×40 oil objective (NA 1.3), 1,024 × 1,024 resolution, and ×4 digital zoom for ablation ROIs. Laser ablation of the FRC network was achieved by using a pulsed Chameleon Vision II TiSa with laser power 75-80% (coherent) tuned to 760 nm. Ablation was performed on small, manually defined linear ROIs between FRC connections away from cell bodies. Ablation was performed in a single z-plane at the center of the FRC connection in a vibratome slice. Time-lapse videos were recorded over 25 s on a single channel photomultiplier tube (PMT) detector 512 × 512 pixels, with 521-ms scan speed per frame to capture recoil of the network. Recoil of the network was calculated by manually measuring displacement of the FRC network between two points located away from the ablation site, with initial recoil velocities calculated from the displacement one frame after the cut, as in other studies 22,52 .
Immunostaining of tissue sections. LNs that were used for sectioning and immunofluorescence were fixed in Antigen fix (DiaPath) for 2 h at 4 °C with gentle agitation. LNs were washed for 30 min in PBS before being applied to 30% w/v sucrose 0.05% sodium azide solution at 4 °C overnight. LNs were dipped into Tissue-Tek optimum cutting temperature compound before being embedded into molds containing optimum cutting temperature compound. A maximum of six LNs were embedded into a single block for comparative analysis. LNs were sectioned on the Leica cryostat at a thickness of 20 µm.
For immunofluorescent staining, sections were permeabilized and blocked using 10% normal goat serum (NGS), 0.3% Triton X-100 in PBS for 2 h at 20-25 °C. Primary antibodies were diluted according to Supplementary Table 1 in 10% NGS, 0.01% Triton X-100 in PBS, and the mix was centrifuged at 15,000 g for 5 min at 4 °C. Sections were incubated with primary antibodies overnight at 4 °C. Sections were then brought to 20-25 °C and washed three times for 15 min each in 0.05% PBS-Tween 20. Sections were then blocked using 10% NGS, 0.3% Triton X-100 in PBS for 2 h at 20-25 °C. Secondary antibodies were then prepared as primary antibodies with dilutions given in Supplementary Table 1 and were applied to the sections for 2 h at 20-25°C. This was followed by two 15-min washes of 0.05% PBS-Tween 20 and a final wash of 15 min in PBS. Sections were then mounted using mowiol mounting media.
For staining of 200-µm-thick vibratome sections, slices were first fixed in Antigen fix (DiaPath) for 2 h at 4 °C with gentle agitation. Sections were placed into 0.1 M Tris-HCl, pH 7.4, on ice for 30 min. Sections were permeabilized using IHC buffer containing 0.5% BSA, 2% Triton X-100 (Sigma-Aldrich) in 0.1 M Tris-HCl, pH 7.4, for 20 min at 4 °C with gentle agitation. Primary antibodies were diluted into IHC buffer, according to Supplementary Table 1 Unless otherwise stated, confocal images were acquired using Leica TCS SP8 STED 3X or the Leica TCS SP5 using HCX Plan-Apochromat ×40 (NA 1.25) and ×63 (NA 1.4) oil lenses. Images were captured at 1,024 × 1,024 pixels, three-line average onto hybrid pixel (HyD) or photomultiplier tube (PMT) detectors. Fluorophore excitation and acquisition was performed in a sequential and bidirectional manner. Imaging regions were manually defined, and z-stacks (15-40 µm) with regular z-intervals ranging from 0.5 µm to 1 µm (depending on the sample) were acquired using a motorized stage. Tile scans were automatically stitched (numerical) using Leica imaging software. To quantify T cell packing, ECM fiber thickness and pMLC-positive and perlecan positive F-actin fibers, Fiji (ImageJ) was used on acquired z-stacks and maximum projections. For T cell packing (×40 lens), an in-house Fiji (ImageJ) macrocleared nuclei inside PDPN + stain. Then, a single z-plane of DAPI (nuclei) had despeckle and gaussian blur (sigma = 2) applied. Nuclei were then detected and counted using thresholding and watershed segmentation. For F-actin fiber analysis (×63 lens), hand-drawn ROIs of F-actin fibers were applied to PDPN maximum projections to check and count the number of F-actin fibers that were within the PDPN + FRC network. Then, F-actin fiber ROIs were applied to pMLC and perlecan channels to count the number and percentage of pMLC + fibers or perlecan-aligned fibers. The FRC gap analysis used an in-house MATLAB script 3 . Briefly, PDPN fluorescence maximum projections (×40 lens) were converted into a binary mask before a circle-fitting algorithm consecutively fit the largest circle possible within the gaps in the network that did not overlap with other fitted circles. Each circle was given a radius, and the distribution of circles with radius >12 µm were plotted. The MATLAB script is available at https://doi.org/10.5522/04/8798597.v1.
Flow cytometry of LNs. LNs were carefully dissected and weighed and placed into RPMI 1640 media on ice. LNs were then processed as previously described 3,53 . Briefly, LNs were placed into a digestion buffer containing collagenase D (250 µg ml −1 ) (Millipore Sigma), dispase II (800 µg ml −1 ) (Thermo Fisher Scientific), and DNase I (100 µg ml −1 ) (Sigma-Aldrich). LNs were gently digested in a water bath at 37 °C, removing and replacing the cell suspension every 10 min until completely digested. Cell suspensions were then centrifuged at 350 g for 5 min. The cells were resuspended in PBS, consisting of 1% BSA (Sigma-Aldrich) and 5 mM EDTA (Sigma-Aldrich), and were then filtered, counted, and resuspended at 10 × 10 6 cells ml −1 . Then, 2.5 × 10 6 cells were seeded and stained for surface and intracellular markers for a stromal cell or T cell panel (Supplementary Table 1). Cells were blocked with CD16/CD32 Mouse Fc block (BD Biosciences) and then stained with primary antibodies for 20 min at 4 °C. For intracellular staining of Ki67 cells, cells were fixed and permeabilized using FOXP3 fix/perm buffer as specified by the manufacturer (BioLegend). Samples were run on the Fortessa X20 flow cytometer (BD Biosciences) at the University College London Cancer Institute. Data were analyzed using FlowJo software.
Cell lines and primary cell culture. Immortalized FRCs (control FRC) were generated as described by Acton et al. 3,32 . PDPN was stably knocked down (PDPN KD FRC) in the parental cell line by transfection of a PDPN shRNA lentivirus. PDPN was completely depleted from the parental cell line (PDPN KO FRC) using CRISPR-Cas9 genetic deletion. In all experiment where exogenous PDPN mutant cell lines (PDPN WT, PDPN T34A, and PDPN S167A-S171A) were used, protein production was induced by the addition of 1 µg ml −1 doxycycline for 48 h. Cell lines were maintained at 10% CO 2 , 37 °C in Dulbecco's modified Eagle medium (Thermo Fisher Scientific) supplemented with 10% fetal bovine serum, 1% P/S, and 1% ITS unless otherwise stated. In vitro experimental groups were defined by the genotype of the cell lines. Cells were treated with recombinant CLEC-2-Fc or Control-Fc supernatant 3,32 for approximately 2 h, where indicated, to assay the effect of CLEC-2 signaling through PDPN. Primary cell isolation from murine LNs was performed as previously described 53 and cultured as with cell lines. Primary cells were then treated with 20 µM Y27632 (Tocris, 1254) for indicated timepoints.
Culture media was changed every 3 days, upon which fresh ROCK inhibitor was applied. Cells were collected and analyzed by flow cytometry.
Tether pulling and trap force measurements. Trap force measurements were performed using a home-built optical tweezer and a 4-W 1,064-nm laser quantum Ventus within a ×100 oil immersion objective (Nikon, NA 1.30, CFI Plan Fluor DLL) on an inverted microscope (Nikon, Eclipse TE2000-U) equipped with a motorized stage (PRIOR Proscan). The optical tweezer was calibrated following previous studies 30,31 . The trap force calibration was performed in every experiment with typical calibration trap stiffness of k∼0.114 pN nm −1 . Measurements were performed using concanavalin A-coated (50 μg ml −1 ) carboxyl latex beads, (1.9 μm diameter, Thermo Fisher Scientific). Beads were incubated on a shaker with concanavalin A for 2 h before the experiment. Beads were applied to the culture media, manipulated by the optical trap, and brought into contact with the cellular membrane and typically held for 2-5 s to allow binding to membrane. Bead position was recorded every 90 ms in brightfield before and during tether formation. Cells were maintained in the trap at 37 °C and had CO 2 flowing into the chamber. CLEC-2 or CTRL supernatant was added 2-4 h before measurement of trap force. Trap force (F t − pN/µm) was then calculated based on the trap calibration (k), bead position (Δx), using a homemade Fiji macro 31 and the equation F T = kΔx.
Osmotic swelling assay. Osmotic shock was performed in accordance with previously reported protocols 54 . By altering the osmolarity of a solution, cells swell or shrink. Osmolarity was estimated using osmolarity calculations. Isotonic solution was prepared with 137 mM NaCl, 5.4 mM KCl, 1.8 mM CaCl 2 , 0.8 mM MgCl 2 , 20 mM HEPES, 20 mM D-glucose and pH to 7.4 with NaOH. Hypotonic solutions were prepared by diluting the isotonic preparation in milliq water that is 50 mOsm is a 1/6 dilution of isotonic solution. To control for the dilution of ions in solution and account for the effect this may have on swelling, the 50 mOsm solution was restored to 330 mOsm using D-mannitol at 280 mM, acting as the true isotonic control (ISO). Cells were dissociated from cell culture with sterile Dulbecco's phosphate-buffered saline (Thermo Fisher Scientific) + 2 mM EDTA and placed onto individual 35-mm MatTek dishes and allowed to settle for 30 min. After 30 min, the cells remain rounded on the coverslip. Cells were then treated with either ISO, HYPO 50 mOsm, or Extreme HYPO 0 mOsm for 1 h. Phase contrast images of cell swelling were captured every 30 s using a ×20 air objective on a Nikon Ti inverted microscope with a motorized stage controlled by NIS-elements software. Diameter of individual swelling cells were calculated using manual circular ROIs. The ratio of swelling was then calculated by dividing all diameters (d) by the initial diameter (d0). Area under the curves were calculated for the first 20 min of the swelling response.
Immunoblot. Equal cell numbers were grown to confluency, and protein was isolated using 300 μl 4× Laemmli lysis buffer (Bio-Rad) and cell lifters (Thermo Fisher Scientific). Lysates were then sonicated for 20 s followed by 10 min at 95 °C. Then, 1% β-mercaptoethanol (143 mM stock, Sigma-Aldrich) was added to samples to reduce oligomerized protein structures. Electrophoresis gels (10%) were loaded with the same quantity of lysates and run for 60 min at 110 V. Transfer to polyvinylidene fluoride (PDVF) membranes was carried out at 65 V for 2 h at 4 °C. Membranes were blocked for 2 h at 20-25°C with 5% skim milk powder (Sigma-Aldrich), 1% BSA in TBS and stained with primary antibodies (Supplementary Table 1) overnight at 4 °C in 1:5 diluted blocking buffer. Membranes were then washed thoroughly with TBS 0.05% Tween 20 and incubated with horseradish peroxidase-conjugated secondary antibodies (Supplementary Table 1) for 1 h at 20-25 °C in 1:5 blocking buffer. After washing with TBS 0.05% Tween 20, membranes were visualized using ECL-horseradish peroxidase reaction and imaged using Image Quant 5000 (GE Life Sciences).
Linear unmixing and Imaris rendering. Imaging of Pdgfra iR26R-Confetti LN sections was carried out using lambda mode and chameleon laser at 900 nm to acquire multi-channel images. Widefield images and z-stack intervals of 0.5 µm were obtained, for an approximate thickness of 30-60 µM. The emission wavelengths for each fluorophore were set on Zeiss Zen black software spectral unmixing function by selecting labeled cells within the confetti imaging. The second harmonic of the two-photon laser detected the ECM conduit structure in the LN. Rendering of PDPN, CFP, and YFP in Figs. 1, 3, and 4 and supplementary movies was achieved using Imaris surface tools.
Quantification and statistical analysis. Prism7 Software (GraphPad) was used to perform multiple statistical analyses, including appropriate tests that were performed as indicated in figure legends. Data collection and analysis were not performed blind to the conditions of the experiments. Data distribution was assumed to be normal, but this was not formally tested. In general, comparison of multiple groups was performed using one-or two-way ANOVA with Tukey's multiple comparisons depending on the dataset. Comparisons of two data sets were mostly performed using two-tailed Mann-Whitney tests.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
There are no restrictions on data availability. Data, code, or reagents are available upon request. Numerical source data files for all figures are provided in Excel supplementary data files and listed in the inventory. Image source data files for all figures are supplied in TIFF format in the supplementary data files and listed in the inventory. Source data are provided with this paper.