key: cord-0314042-3u2pdiqd
authors: Arcucci, Valeria; Ishaq, Musarat; Roufail, Sally; Dredge, B. Kate; Bert, Andrew G.; Hackett-Jones, Emily; Liu, Ruofei; Pillman, Katherine A.; Fox, Stephen B.; Stacker, Steven A.; Goodall, Gregory J.; Achen, Marc G.
title: The microRNA miR-132 is a key regulator of lymphatic vascular remodelling
date: 2021-12-23
journal: bioRxiv
DOI: 10.1101/2021.12.22.473780
sha: 96705c98cd393627753e7b3412669bac250aa86c
doc_id: 314042
cord_uid: 3u2pdiqd

Lymphangiogenesis (growth of new lymphatic vessels), and lymphatic remodelling more broadly, are important for disease progression in cancer, lymphedema and the pulmonary disease lymphangioleiomyomatosis. Multiple molecular pathways which signal for aspects of lymphangiogenesis are known but little is understood about their co-ordinate regulation in lymphatic endothelial cells (LECs). Small RNA molecules co-ordinately regulate complex biological processes, but knowledge about their involvement in lymphangiogenesis is limited. Here we used high-throughput small RNA sequencing of LECs to identify microRNAs (miRs) regulating lymphatic remodelling driven by the lymphangiogenic growth factors VEGF-C and VEGF-D. We identified miR-132 as up-regulated by both growth factors, and demonstrated that inhibiting miR-132 in LECs in vitro blocked cell proliferation and tube formation, key steps in lymphangiogenesis. We showed that miR-132 is expressed in human LECs in vivo in the lymphatics of human breast tumours expressing VEGF-D. Importantly, we demonstrated that inhibiting miR-132 in vivo blocked many aspects of lymphangiogenesis in mice. Finally, we identified mRNAs regulated by miR-132 in LECs, by sequencing after RNA-protein cross-linking and Argonaute immunoprecipitation, which demonstrated how miR-132 co-ordinately regulates signalling pathways in lymphangiogenesis. This study shows miR-132 is a critical regulator of lymphangiogenesis and a potential target for therapeutically manipulating lymphatic remodelling in disease.

The lymphatic vasculature is a critical feature of many tissues and organs which is required for 74 maintenance of tissue fluid homeostasis, immune function and absorption of fats and fat-75 soluble vitamins (1). It consists of multiple types of lymphatic vessels: initial lymphatics, which 76 take up tissue fluid; lymphatic pre-collectors and collectors which drain lymph fluid from 77 tissues; lymphatic ducts which return lymph to the venous circulation (2). Lymphatic vessels 78

can undergo remodelling involving modifications in their shape, size and molecular features. 79

A prominent form of lymphatic remodelling is lymphangiogenesis which is the formation of 80 new lymphatic vessels from pre-existing lymphatics. Lymphangiogenesis involves 81 proliferation, sprouting, migration and tube formation by lymphatic endothelial cells (LECs) 82

(3). Remodelling of the lymphatic vasculature occurs during embryonic development and in 83 several pathological conditions such as cancer, inflammation, infection, lymphedema, 84 lymphangiectasia and lymphangioleiomyomatosis (LAM) (4-7). Lymphatic remodelling can 85 be important for disease progression, e.g. lymphangiogenesis in a primary tumour can facilitate 86 the metastatic spread of cancer (3, 8, 9) . 87 88 From a molecular perspective, the remodelling of lymphatic vessels can be driven by soluble 89 growth factors which bind cognate receptors on LECs and activate molecular pathways that 90 drive cell proliferation, migration and tube formation. VEGF-C and VEGF-D are two such 91 growth factors which induce lymphangiogenesis by activating VEGFR-3, a receptor tyrosine 92 kinase localized on the surface of LECs (10-13). VEGF-C and VEGF-D are the most 93 extensively studied lymphangiogenic factors that have been shown to induce lymphatic 94 remodelling in embryonic development, cancer and other disease settings (6, 14) . The activation 95 of VEGFR-3 on LECs by VEGF-C or VEGF-D leads to activation of a variety of downstream 96 signalling pathways including the PI3K-AKT pathway (15), important for cell proliferation, 97 and pathways which play a role in tube formation by LECs (16) . Furthermore, other signalling 98 pathways important in lymphangiogenesis have recently been identified by a genome-wide 99 functional screen based on primary human LECs (17). However, the means by which such 100 signalling pathways are co-ordinately regulated in LECs to drive growth and remodelling of 101 lymphatics is not well understood. 102 treatment (6.4-fold increase) ( Figure 1D ). Similarly, levels of miR-132 were increased in 166 response to VEGF-D at every time-point, but with a maximum increase at 9 h of treatment fold increase compared with the negative control). These results confirm the findings from 168 small RNA-Seq that both VEGF-C and VEGF-D induce increased levels of miR-132 in LECs, 169 and demonstrate that the kinetics of these responses are different with peak response being 170 reached faster for VEGF-C than for VEGF-D. The previous analyses showed that miR-132 is up-regulated in LECs in vitro in response to the 185 lymphangiogenic growth factors VEGF-C and VEGF-D. Lymphangiogenesis is a complex, 186 multi-step process that involves proliferation, migration and tube formation by LECs to form 187 new lymphatic vessels (30) (31) (32) . These distinct steps of lymphangiogenesis can be modelled 188 with in vitro assays involving LECs (33) . To test the functional role of miR-132 using such 189 assays we employed an antagomiR of miR-132 (designated miR-132 antagomiR #1), delivered 190

to LECs by liposome transfection, to inhibit miR-132. This antagomiR was functionally 191 validated by transfection into LECs followed by monitoring of the mRNA for Retinoblastoma VEGF-D whereas a scrambled negative control antagomiR did not ( Figure 2A&B ). These 203 results indicate that miR-132 is required for LECs to proliferate in response to either or 206

The effect of inhibiting miR-132 in LECs on VEGF-C-and VEGF-D-driven formation of 207 vessel-like structures in vitro was also monitored. LECs were transfected with miR-132 208 antagomiR #1, or the scrambled negative control antagomiR, and used in a collagen overlay 209 tube formation assay (16). VEGF-C and VEGF-D both induced tube formation of LECs as 210 assessed by an increase in the number of branch-points of the tubes ( Figure 2C&D ). 211

Importantly, the inhibition of miR-132 with miR-132 antagomiR #1 significantly restricted 212 branching of the tubes induced by VEGF-C ( Figure 2C ,E&F) or VEGF-D ( Figure 2D ,E&F), 213

whereas the scrambled antagomiR control did not. 

Finally, the effect of inhibiting miR-132 on LEC migration was monitored in an in vitro 231 "wound closure" assay in which a confluent layer of LECs is "scratched" leaving a region 232 devoid of cells into which the LECs can then migrate (35). VEGF-C and VEGF-D both 233 promoted migration of LECs in this assay system ( Figure 2G&H ). To test the miR-132 234 inhibitor, LEC monolayers were transfected with miR-132 antagomiR #1, or a scrambled 235 negative control antagomiR, scratched and then incubated with Quantification of migrated LECs demonstrated that the miR-132 antagomiR did not 237 significantly alter migration of LECs promoted by VEGF-C or VEGF-D ( Figure 2G -I). 238

Overall, these data demonstrate that miR-132 is required for LEC proliferation and tube 240 formation in vitro in response to VEGF-C or VEGF-D but is not required for the migration of 241

LECs induced by either of these growth factors. Therefore, this miRNA is crucial for some, 242

but not all, cell biological aspects of lymphangiogenesis. 243 244 miR-132 is expressed in lymphatics in vivo 245 246 Based on the findings described above, we hypothesised that miR-132 levels would increase 247 in the LECs of lymphatic vessels in a lymphangiogenic environment in vivo, particularly when 248 the lymphatic remodelling is driven by VEGF-C or VEGF-D. We therefore looked for 249 expression of this miRNA in two different pathological models of lymphangiogenesis in vivo: 250 (i) Tumour xenografts in mice based on the 293EBNA-1 human embryonic kidney cell line 251 which had been genetically engineered to express full-length human VEGF-D (designated 252 "VEGF-D-293 tumours") (36); (ii) Human tumours expressing high levels of VEGF-C and/or 253

The expression of miR-132 in three VEGF-D-293 tumours was monitored by in situ 256 hybridization (ISH), and lymphatics were identified on serial sections by immunostaining for 257 podoplanin, an extensively utilized lymphatic marker broadly expressed on small and large 258 lymphatics (37) ( Figure 3A ). This revealed that approximately 70% of podoplanin-positive 259 in liver hepatocellular carcinoma, lung squamous cell carcinoma, melanoma and breast 276 invasive carcinoma. Breast invasive carcinoma was chosen for immunohistochemical analysis 277 because both VEGF-C and VEGF-D mRNAs are commonly detected at relatively high levels 278 in these tumours. Subsequent to bioinformatic analysis, immunohistochemical analysis of 279 breast cancer tissue microarrays (TMAs) containing human breast invasive carcinomas was 280 performed to identify individual tumours expressing VEGF-D; immunohistochemistry for 281 VEGF-C was not performed due to the lack of available antibodies which, in our hands, could 282 reliably detect this protein in human tissues. The breast cancer TMAs contained 120 different 283 tumour samples which included different subtypes of invasive breast cancer, specifically 284 luminal A and B breast cancer, basal-like breast cancer and human epidermal growth factor 285 receptor 2 (HER2)-amplified breast cancer. The analysis of VEGF-D by 286 immunohistochemistry (for example see Figure 3C ) demonstrated that approximately 80% of 287 HER2-enriched breast cancers showed high intensity of VEGF-D staining which was by far 288 the highest percentage for any of the tumour types analysed. These data suggested that VEGF-289 D-positive HER2-enriched breast cancers would be an appropriate choice for analysing miR-290 132 in tumour lymphatics. Therefore, two HER-2-enriched breast cancers expressing high 291 levels of VEGF-D were chosen for analysis of miR-132. Tissue sections of these tumours were 292 stained by immunohistochemistry for podoplanin to highlight lymphatic vessels, and serial 293 sections were subjected to ISH for miR-132 ( Figure 3D ). Approximately 19% (4 out of 21) of 294 podoplanin-positive lymphatic vessels detected in one of these HER2-enriched tumours were 295 also positive for miR-132, whereas none of 20 podoplanin-positive lymphatics detected in the 296 second tumour were positive for miR-132. 297 298 These data demonstrate that solid tumours in mice and humans, expressing VEGF-D, can 299 contain lymphatic vessels which are positive for miR-132. The heterogeneity of miR-132 levels 300

in tumour lymphatics which we observed could be due to variation in VEGF-D concentrations 301 in different parts of a tumour. Further, VEGF-D requires proteolytic activation to bind VEGFR-302 3 on LECs (38) so its capacity to drive expression of miR-132 in tumour lymphatics is 303 dependent on the proteases which activate it (39,40)the levels of these proteases may be 304 highly variable within and between tumours. Figure 4F ). VEGF-C also increased lymphatic vessel density ( Figure 4G ) and width 352 ( Figure 4H ), and reduced the distance between lymphatic vessels ( Figure 4I ). In contrast, when 353 VEGF-C was co-injected with miR-132 antagomiR #2, there was a statistically significant 354 reduction in all of these parameters compared with VEGF-C treatment alone or co-treatment 355

with VEGF-C and the negative control antagomiR. In summary, targeting of miR-132 inhibited 356 many aspects of lymphatic remodelling induced by VEGF-C in the mouse ear 357 lymphangiogenesis assay. 14 359 360 

To understand how miR-132 mechanistically controls the molecular pathways involved in 378 lymphangiogenesis, we identified the interactome of this miRNA in LECs using Argonaute 379

High Throughput Sequencing after Cross-Linked Immunoprecipitation (Ago HITS-CLIP) 380

technology. This approach allows the global identification of mRNAs regulated by a miRNA 381 in a cell. After transfection with a molecular mimic of the miRNA, RNA-protein complexes in 382 the cell are cross-linked, Ago (the protein on which miRNAs pair with mRNA targets) is 383 immunoprecipitated along with mRNAs bound to it, and these mRNAs are isolated for 384 sequencing to identify mRNAs which are enriched due to the presence of the mimic ( Figure  385 5A) (43). We transfected human primary LECs with a miR-132 mimic (a chemically modified 386 double-stranded RNA molecule that mimics endogenous miR-132), or a scrambled mimic 387 negative control, and then performed Ago HITS-CLIP. From this experiment, we identified 388 approximately 500 mRNAs putatively regulated by miR-132 as CLIP peaks enriched in miR-389

Datafile 1 and Supplementary Figure 1 ). Some mRNA targets identified by the Ago HITS-391 CLIP were previously validated targets of miR-132, for instance mRNAs for the phosphatase 392 PTEN (44) and the cyclin-dependent kinase inhibitor p21 (45). These targets validated the 393 experiment and provided some immediate insight into mRNAs regulated by miR-132 in LECs.

In order to analyse the results from the Ago HITS-CLIP, we performed pathway and gene 396 identifying pathways, from those highlighted by pathway analysis, which were most likely to 404 be involved in lymphangiogenesis and lymphatic remodellingthese pathways are listed in 405 Figure 5B . To evaluate the reliability of our HITS-CLIP results, we transfected LECs with the 406 miR-132 mimic and checked by RT-qPCR whether eight of the mRNAs, randomly chosen 407 from mRNAs identified by Ago HITS-CLIP and predicted by our bioinformatic analyses to be 408 involved in lymphagiogenesis, were down-regulated. These mRNAs were for Claudin-11, isoform of the catalytic subunit of protein phosphatase 2 (PP2A)), RHOB, CTGF (connective 411 tissue growth factor) and Nidogen-1. Figure 5C shows that the levels of all eight mRNA targets 412 were down-regulated by the miR-132 mimic, as expected, with the reduction being statistically 413 significant for seven, namely the mRNAs for Claudin-11, Nectin-2, Thrombospondin-1, 414 GSK3B, PPP2CB, RHOB and CTGF. These results indicate a high degree of reliability for the 415 Ago HITS-CLIP experiment. 416 417

The analyses described above identified multiple pathways via which miR-132 co-ordinately 418

regulates key steps in lymphangiogenesis and lymphatic remodelling, including LEC 419 proliferation and tube formation ( Figure 5D ). The down-regulation by miR-132 of mRNA for 420

Claudin-11 in LECs ( Figure 5C leading to down-regulated expression of these proteins, would co-ordinately enhance activity 439 of this key lymphangiogenic signalling pathway via four distinct control points ( Figure 5D ). 440

Finally, RHOB is a negative regulator of lymphangiogenesis which acts by restricting LEC 441 sprout formation (56) ( Figure 5D ). The mRNA for RHOB is down-regulated in LECs by miR-442 445 446 properties. This distinct signalling was previously observed by Karnezis and co-workers who 478

found that VEGF-C and VEGF-D were able to stimulate different signalling pathways in 479 lymphatic vessels (57,58). Both VEGF-C and VEGF-D can bind VEGFR-2, as well as 480 VEGFR-3, on LECs, and it has been reported that they can promote formation of VEGFR-2 481 homodimers, VEGFR-3 homodimers and VEGFR-2/VEGFR-3 heterodimers (59-63). Each of 482 these receptor complexes could potentially induce distinct patterns of signalling (61) We showed that miR-132 was the most up-regulated miR in LECs in response to three hours 495 of exposure to VEGF-C, and was significantly up-regulated by VEGF-D over the same time 496

period. However, the kinetics of the responses to these proteins differed in that the peak in 497 miR-132 level in response to VEGF-C occurred much sooner than for VEGF-D. The reason 498

for the different kinetics of these responses is unknown but could relate to different affinities 499

of VEGF-C and VEGF-D for VEGF receptor homodimers versus heterodimers, and different 500 affinities for cell surface non-VEGFR molecules such as neuropilins, integrins and heparan and tube formation by LECs in vitro, but not for migration of these cells. This suggests that the 503 VEGF-C-and VEGF-D-driven pathways signalling for LEC migration are distinct from those 504 controlling proliferation and tube formation, and that there is a lack of miR-132 targets in the 505 migratory pathways. Targeting miR-132 in vivo in mice blocked many morphological features 506

of VEGF-C-driven lymphatic remodelling, including vessel branching, and increases in vessel 507 width and density, likely reflecting the involvement of LEC proliferation and tube formation 508 in multiple aspects of lymphatic vessel remodelling. miR-132 has also been found to stimulate 509 remodelling of blood vessels in vivo in a mouse model of breast carcinoma (28). Taken 510 together, these results indicate that miR-132 is a "master regulator" of remodelling across the 511 blood and lymphatic vasculatures, likely exploiting similar molecular mechanisms in both 512

In addition to unravelling the crucial function of miR-132 in lymphatic remodelling, our study 515 identified mechanisms by which this miRNA regulates this complex process. The miR-132-516 mRNA interactome was identified in LECs by Ago HITS-CLIP indicating that miR-132 517 controls multiple molecular pathways critical for VEGF-C-or VEGF-D-driven lymphatic 518 remodelling, including pathways important for the biological processes of LEC proliferation 519 and tube formation, as well as lymphatic branching which is dependent on sprout formation; 520 these processes were shown to be dependent on miR-132 in our in vitro and in vivo studies 521 (Figures 2A-D and 4D ). More specifically, a miR-132 mimic down-regulated the mRNAs for 522

Claudin-11 and Nectin-2 in LECs which would co-ordinately modulate two distinct signalling 523 pathways to favour tube formation (51,52) ( Figure 5D ). Likewise, the miR-132 mimic down-524 regulated, in LECs, the mRNAs for Thrombospondin-1 and proteins that negatively regulate 525 the PI3K-AKT pathway, which would co-ordinately modulate two signalling pathways to 526 favour LEC proliferation (15,53) ( Figure 5D ). The down-regulation we observed of the mRNA 527 for RHOB in LECs, in response to the miR-132 mimic, would favour formation of lymphatic 528 sprouts (56), another key step in lymphatic remodelling. These findings indicate that miR-132 529 can (i) co-ordinately modulate an important signalling pathway for lymphatic remodelling at 530 multiple control points (e.g. the PI3K-AKT pathway for LEC proliferation); (ii) co-ordinately 531 modulate multiple signalling pathways to promote a biological process in lymphatic 532 remodelling (e.g. the PI3K-AKT and Thrombospondin-1 pathways involved in LEC 533 proliferation); (iii) co-ordinately modulate multiple processes in lymphatic remodelling (e.g.

Our finding that miR-132 is a key regulator of lymphangiogenesis could have clinical 537 implications as it may be a useful therapeutic target for diseases in which lymphangiogenesis 538 and lymphatic remodelling play a role in disease progression, such as cancer, lymphoedema 539 and lymphangioleiomyomatosis. miR-132 could potentially be targeted systemically in vivo by 540 nucleic acid-based therapeutics. Such therapeutics have faced several hurdles, especially 541 related to the challenges in achieving efficient delivery to organs other than the liver, and 542 overcoming off-target effects and chemistry-dependent toxicity. Nevertheless, up to 2021, 11 543 oligonucleotide-based therapies had been approved for clinical use (68), and a lipid-544 nanoparticle formulated, nucleoside-modified synthetic RNA for immunisation against SARS-545

CoV-2 (the BNT162b2 mRNA Covid-19 vaccine) (69) recently achieved regulatory approval. 546

Also, many more therapeutics based on nucleic acids are being assessed in clinical trials (68). 547

These developments are encouraging for the feasibility of targeting miR-132 in the clinic. As 548 miR-132 has the ability to regulate entire networks of pathways by controlling the expression 549 of many mRNA targets in multiple pathways, it could be an appropriate target for 550 comprehensively inhibiting lymphangiogenesis and lymphatic remodelling. The targeting of a 551 specific growth factor, such as VEGF-C or VEGF-D, or a specific growth factor receptor, such 552 as VEGFR-3, might encounter inherent or acquired drug resistance because alternative 553 lymphangiogenic growth factors or receptors are expressed in a disease prior to treatment, or 554 become up-regulated in response to therapy. In contrast, targeting miR-132 may be less likely 555 to encounter resistance given this miRNA targets multiple lymphangiogenic signalling normalised to mRNA for human β-actin, and miRNA levels were normalised to levels of hsa-639 miR-423-5p, hsa-miR-345-5p and hsa-let7g-5p. Data were analysed using the comparative CT 640 method as described (71). 641 642 in cell monolayers. Experiments were carried out on duplicate plates, one whose end-point was 666

immediately after the scratch (the t0 plate) and one whose end-point was after 24 h of treatment 667 with growth factors as described below (the t24 plate). Following the scratch, cells in the t24 668 plate were washed once with PBS and then treated with growth factors for 24 h. At the assay 669 end-points, both plates were washed once with PBS and fixed with 1% PFA for 2 h at room 670 temperature. Cells were then washed in PBS, permeabilised and stained using a PBS solution 671 containing 2% BSA, 0.2% Triton X-100 (cat #T9284, Sigma-Aldrich) and phalloidin-Alexa 672 488 (1:100) (A12379, Thermo Fischer Scientific). Images of scratches at t0 and t24 were 673 acquired using a high throughput screening microscope (Arrayscan XTI, Thermo Fisher 674 Scientific). For each well, 9 adjacent fields of view were captured and then stitched together as 675 a montage image using a 4x or 20x objective. Images of scratches were analysed using image 676 analysis software (MetaMorph, Molecular Devices, Sa Jose, CA). The analysis protocol 677 consisted of several processing steps -images were smoothened and a filter was applied to 678 create a binary image that separates cells from background, and scratch areas were then 679 measured. Resulting data were used to calculate migration area of cells according to the 680 following formula: Migration area = scratch area in t0 plate -scratch area in t24 plate.

LECs were plated in black-walled 96-well imaging plates (Costar ®, Corning) at 1 x 10 4 cells 685 per well, and transfected with miR mimics or antagomiRs. After 24 h, cells were serum starved, 686 and the rest of the assay was performed 48 h post-transfection. To prepare a 10 ml overlay 687 collagen mix (1 mg/ml collagen), 50 μl of 1 N sodium hydroxide (Sigma-Aldrich), 3.33 ml of 688 10 x PBS, 950 ml Milli-Q water, 2 ml of bovine collagen I (5 mg/ml) (Gibco), and 6.67 ml 689 serum starvation medium were mixed on ice. The collagen mix (100 μl Formalin-fixed paraffin-embedded tissues were used for in situ hybridization (ISH) employing 720 locked nucleic acid (LNA) probes labelled with digoxigenin (DIG) at both the 5′-and 3'-ends 721 (Qiagen). ISH was performed with probes specific for miR-132-3p (10 nM) or U6 snRNA (2 722 nM), or with a "scrambled" negative control probe (10 nM), according to manufacturer's 723

instructions. The scrambled control probe consisted of a random sequence that was not 724 complementary to any known miRNA. Tissue sections were deparaffinized in histolene, 725 rehydrated and treated with proteinase K (15 μg/ml) at 37°C for 10 min. DIG-labelled probes 726

were denatured at 94°C for 5 min. Tissue sections were then coated with probes diluted in 727 hybridization buffer, coverslips applied to avoid evaporation and slides incubated at 56°C for 728 1 h in a hybridization oven (HybEZ, Oven, Advanced Cell Diagnostics, Newark, CA). 729 Stringent washes were then performed at 56°C: once with 5 x SSC, twice with 1 x SSC, and 730 three times with 0.2 x SSC buffers. Each washing step was carried out for 5 min. Slides were 731 then washed in PBS for 5 min at room temperature. Tissue sections were blocked by incubation 732

with Antibody blocking solution (see Table 2 ) in a humidified chamber for 15 min. Anti-733

Digoxigenin-AP Fab fragments (Cat # 11093274910, Roche, Basel, Switzerland) were diluted 734 1:800 in Antibody Dilutant Solution (Table 2) , applied to tissue sections and incubated for 1 h 735 at room temperature in a humidified chamber. Slides were washed three times with PBST for 736 3 min and tissue sections incubated with AP substrate (Table 2) for 2 h at 30°C. The reaction 737 was stopped by washing slides with KTBT buffer (Table 2 ) twice for 5 min. Slides were then 738 washed twice in Milli-Q water for 1 min, and tissue sections counterstained with Nuclear Fast 739

Red (N8002, Sigma-Aldrich) for 10 min before carefully rinsing in running tap water for 10 740 min. Tissue sections were then dehydrated and mounted with Eukitt mounting medium (cat # 741 03989, Sigma-Aldrich) and coverslips. Images were acquired using an Olympus BX61 742 microscope and a SPOT Model 25.4 2 Mp Slider digital camera (Diagnostic Instrument Inc. 743

Sterling Heights, MI) using SPOT Software Version 5.0. 744 Immunohistochemical staining was performed for human or mouse podoplanin, human VEGF-749 D and mouse LYVE1 (see Table 3 for primary antibodies). Formalin-fixed, paraffin-embedded 750 tissue sections were deparaffinized in histolene and rehydrated. For all antigens, slides were 751 incubated in 0.1 M citrate buffer pH 6.0 Target Retrieval Solution (cat # S1699, Dako, 752

Productionsvej, Denmark), and antigen retrieval was performed by heating in a pressure cooker 753 (Dako) at 125°C for 3 min. Slides were left to cool at room temperature for 30 min and washed 754 with PBS three times for 5 min. After antigen retrieval, blocking of endogenous peroxidase 755 activity was performed by incubating tissue sections in 3% H2O2 in methanol at room 756 temperature for 20 min, followed by rinsing with PBS three times for 5 min. Sections were 757 blocked by incubation with Tris-NaCl blocking buffer (TNB) (0.1 M Tris-HCl, pH 7.5, 150 758 mM NaCl, and 0.5% w/v blocking reagent, PerkinElmer) for analysis of Podoplanin, or with 759 serum-free protein block (cat # X0909, Dako) for analysis of VEGF-D or LYVE-1 in a 760 humidifying chamber for 1 h. Tissue sections were incubated with the appropriate primary 761 antibody diluted in TNB (for podoplanin) or Antibody Diluent (cat # S0809, Dako) (for VEGF-762 D or LYVE 1) in a humidified chamber at 4°C overnight. The following day, slides were 763 washed with TNT buffer three times for 5 min and then incubated at room temperature for 1 h 764 with secondary antibodies as follows: (i) For mouse podoplanin, a biotinylated anti-hamster 765 secondary antibody raised in goat (Vector Laboratories, Burlingame, CA) diluted 1:300 in TNB 766 was used; (ii) For human podoplanin and human VEGF-D, a biotinylated anti-mouse secondary used; (iii) For mouse LYVE1, a biotinylated anti-rabbit secondary antibody raised in goat 769 (Vector Laboratories) diluted 1:300 in Antibody Diluent (Dako) was used. Slides were then 770 washed with TNT buffer three times for 5 min, before peroxidase staining was conducted for 771 30 min with a peroxidase staining kit (VECTASTAIN® ABC Reagent, cat # PK-4000, Vector 772 Laboratories) as described by the manufacturer. Slides were then incubated with 3,3'-773 diaminobenzidine working solution (SK-4100, Vector Laboratories) for 5 to 10 min, as 774 described by the manufacturer, to visualise colour change. The colour reaction was stopped by 775 immersing slides in Milli-Q water for 5 min. Finally, tissue sections were cover slipped with 776 AquaMount (Merck) and left to dry. Images were acquired using an Olympus BX61 SCID/NOD/gamma (NSG) mice, 6-12 weeks old, were obtained from the Animal Care and 2 h to allow mice to recover. At the end of experiments (7 days after injection) mice were killed 807 by asphyxiation with CO2, and ears were harvested for immunohistochemical analysis. 808 809 Ears were fixed in 4% PFA in PBS overnight at 4°C on an orbital shaker. The following day, 810 ears were washed twice in PBS and kept in a solution of 0.3% Triton X-100, 0.05% sodium 811 azide in PBS until dissection. Prior to dissections, ears were carefully shaved with a multi-812 blade razor, and the frontal side of the ear was then separated from the dorsal side using forceps. 813

Ears were then washed extensively by vortexing in wash solution (0.3% Triton X-100 in PBS). 814

The cartilage was then peeled from the internal part of the frontal side of the ear and another 815 round of washing was performed (by vortexing in wash solution). The fat was then scraped 816 from the internal part of the frontal side of the ear using curved forceps. Ears were then washed 817 for the last time (by vortexing in wash solution) and pinned on a silicone-based matrix (cat # 818 761028, Sylgard, Sigma-Aldrich) in a 6-well plate. Wash solution was poured into the wells 819 and ears were kept in this solution until whole-mount immunostaining. 820 821 Sample sizes for this assay were based on previous in-house experience showing that two 822 independent experiments, each with study group sizes in the range of 8-10, were appropriate 823 from the perspectives of feasibility and statistical comparison. Quantitative analyses for this 824 assay were conducted in a blinded fashion so the investigator was unaware of which sample 825 was from which study group. 826

Mouse ears were blocked overnight at 4°C on an orbital shaker in blocking solution (5% goat 829 serum, 0.2% BSA, 0.05% sodium azide and 0.3% Triton X-100 in PBS) and primary antibody 830 treatment was then performed. Ears were stained with LYVE1 antibody (rabbit anti-mouse 831 LYVE1 antibody, cat # 70R-LR003, Fitzgerald, North Acton, MA), diluted 1:1,000 in PBS, to 832 identify lymphatic vessels. This primary antibody was incubated with ears overnight at 4°C on 833 an orbital shaker. Ears were then washed three times (each time for 1 h at room temperature) 834 in washing solution (0.3% Triton X-100 in PBS), and were stored overnight at room 835 temperature in 0.3% Triton X-100, 0.05% sodium azide in PBS. Next day, ears were transferred 836 into a solution containing secondary goat anti-rabbit 488 antibody (cat # A11034, Invitrogen) 837 diluted 1:250 in 0.3% Triton in PBS, and incubated overnight at 4°C on an orbital shaker, 838 protected from light. The next day, ears were washed three times (each time for 1 h at room 839 temperature) in washing solution (0.3% Triton X-100 in PBS) and then stored overnight at 840 room temperature in 0.3% Triton X-100, 0.05% sodium azide in PBS. On the following day, 841 ears were vigorously vortexed for 1 min in a washing solution (0.3% Triton X-100 in PBS) and 842 mounted using aqueous anti-fade mounting medium (Vectashield, cat # H1000, Vector 843

Laboratories) onto slides on which frames (In situ frames, cat # 0030127-536, Eppendorf, 844

Hamburg, Germany) had been applied. Coverslips were then applied and ears imaged during 845 the following week using a confocal microscope (FV3000 Olympus, Tokyo, Japan). For each 846 ear, four fields surrounding the Matrigel plug were imaged and analysed using MetaMorph and 847

AngioTool software (73). In broad terms, the Ago-HITS-CLIP method was performed as described previously (74) with 852 the following major modifications: after cross-linking, cells were lysed on the plate, and both 853 3' and 5' linkers were ligated to the Ago-bound RNAs prior to elution from the IP beads. More 854 specifically, LECs were transfected in 10 cm cell culture dishes with 20 nM miRVana mimic 855 (miR-132 or negative control) using RNAiMAX (Invitrogen); two technical replicates each of 856 two biological replicates i.e. four samples per condition. After 24 h, cells were rinsed with ice-857 cold PBS and UV irradiated with 800 mJ/cm 2 , 254 nm, in ice-cold PBS using a UV Stratalinker 858 (Stratagene). Cells were lysed with 1 X PXL (1 X PBS, 0.1% SDS, 0.5% deoxycholate, 0.5% Igepal) containing EDTA-free Complete protease inhibitor cocktail (PIC; Roche). DNA was 860 digested with 5 μl Turbo DNAse (Ambion AM2238) at 37°C for 10 min. RNA was partially 861 digested with RNase 1 (Ambion AM2295) by adding 10 μl of 1:75 diluted RNase 1 per 1 ml 862 of lysate at 37°C for 5 min. Lysates were centrifuged at 55,000 g for 22 min at 4°C and 863 supernatant transferred to a fresh tube. AGO-RNA complexes were immunoprecipitated using 864 mouse IgA2 monoclonal anti-Ago2 antibody 4F9 (75) (hybridoma sourced from University of 865 Florida ICBR). Antibodies (15 μg) were conjugated to 20 μl protein L Dynabeads 866 (ThermoFisher, 88849) before resuspending the washed beads with 1 ml of prepared lysate at 867 ~800 g/ml and rotating for 2 h at 4°C. Bound AGO-RNA complexes were washed twice each 868 consecutively with ice cold 1 X PXL, 5 X PXL (5 X PBS, 0.1% SDS, 0.5% sodium 869 deoxycholate, 0.5% Igepal), and 1 X PNK (50 mM Tris-Cl pH 7.5, 10 mM MgCl2, and 0.5% 870 Igepal). Beads were first treated with T4 PNK (NEB, M0201L; 20 U in 80 μl reaction volume) 871

in the absence of ATP (37°C, 850 rpm for 20 min) to dephosphorylate 3' RNA ends followed 872 by washes with 1 X PNK, 5 X PXL, and two washes with 1 X PNK at 4°C. The 3' 873 preadenylated linker (NEBNext 3'SR adaptor for Illumina; /5rApp/AGA TCG GAA GAG 874 CAC ACG TCT /3AmMO/) was ligated to the RNA fragments on bead using T4 RNA ligase 875 2 truncated KQ (NEB M0373) at 16°C, overnight with shaking. Beads were washed 876 consecutively with ice cold 1 X PXL, 5 X PXL, and twice with 1 X PNK. Bound RNAs were 877 then labelled with P32 γ-ATP using T4 PNK, 25 min at 37°C, followed by addition of 2.5 µM 878 ATP, 5 min at 37˚C. Beads were washed twice each with ice-cold 1 X PNK+EGTA (50 mM 879

Tris-HCL pH7.5, 20 mM EGTA, 0.5% Igepal) and 1 X PNK. The 5' RNA linker (5'-blocked 880 and containing a 10 nt UMI 881 (/5AmMC6/GUUCAGAGUUCUACAGUCCGACGAUCNNNNNNNNNN3') was ligated to 882 the RNA fragments on bead using T4 RNA ligase (NEB M0437) for 2 h at 24˚C with shaking. 883

Beads were washed 3 times with ice-cold 1 X PNK+EGTA. 884 885 AGO-RNA complexes were eluted with 40 μl 1 X Bolt LDS sample buffer (ThermoFisher), 886 1% β-mercaptoethanol at 70°C for 10 min. Samples were separated through Bolt 8% Bis-tris 887

Plus gels (ThermoFisher) using BOLT MOPS SDS running buffer at 165 V for 75 min. 888

Complexes were then transferred to nitrocellulose (Schleicher&Schuell, BA-85) by wet 889 transfer using 1 X BOLT transfer buffer with 10% methanol. Filters were placed on a phosphor 890 screen and exposed using a Typhoon imager (GE). 130-180 kDa regions (corresponding to digestion (2 mg/mL proteinase K, 100 mM Tris-HCl pH 7.5, 50 mM NaCl, 10 mM EDTA, 893 0.2% SDS) at 50°C for 60 min on a Thermomixer (1,200 rpm) followed by extraction with acid 894 phenol (ThermoFisher, AM9712) and precipitation with 1:1 isopropanol:ethanol. RNA was 895 pelleted by centrifugation then separated on a 15% denaturing polyacrylamide gel (1:19 896 acrylamide, 1 X TBE, 7 M urea). The wet gel was wrapped in plastic wrap and exposed to a 897 phosphor screen and imaged using a Typhoon. Gel slices were cut corresponding to the 898 expected size of the cross-linked RNA (70 -280 nt) and eluted by the "crush and soak" method 899 as previously described (76). Reverse transcription was performed using SR-RT primer (IDT, 900 AGACGTGTGCTCTTCCGATCT) with SuperScript IV. Products were amplified for 12 901 cycles using a common forward primer (NEBNext SR primer for Illumina) and barcoded 902 reverse primers for each sample (NEBNext Index primers for Illumina). PCR products were 903 purified using 1.8 volumes of Axygen AxyPrep magnetic beads (MAG-PCR-CL), separated 904 on an 8% acrylamide (19:1), 7 M urea, TBE semi-denaturing gel, stained with SYBR Gold 905 nucleic acid gel stain (ThermoFisher) and imaged on a ChemiDoc (BioRad). Products 906 corresponding to an insert size of ~25 -70 nt were excised from the gel and extracted by the 907 "crush and soak" method as above. Library quality and quantity was assessed by Bioanalyzer 908 (Agilent) and qPCR, pooled and sequenced on an Illumina NextSeq 500 (1 x 85bp Annotation obtained from miRBase (www.mirbase.org; (80)) was used to obtain miR-level 930 read counts using htseq-count (81) and read counts were normalised to reads per million (RPM) 931 using the total number of reads mapping to microRNAs. 932 933 HITS-CLIP. HITS-CLIP libraries were prepared for LEC cells transfected with miR-132 934 mimic or a negative control mimic, consisting of two biological replicates which were each 935 separated into two technical replicates and sequenced to produce 85 bp single-ended reads with 936 average sequencing depth of ~22 million reads. FASTQ files were analysed at various stages 937 for quality and content with FastQC v0.11.5 (77) and raw reads were adapter trimmed and 938 filtered using cutadapt v1.6 (78) with an adapter sequence of 939 AGATCGGAAGAGCACACGTCTGAACTCCAGTCA, error rate of 0.2, overlap of 5 and 940 minimum length of 18. Reads derived from PCR duplication were collapsed using Unique 941 Molecular Identifiers (UMIs) with UMI-tools (v0.5.3; (82)) by first using the 'extract' method 942 with default parameters to cut the 10 bp UMIs from the 3' end of the reads allowing an edit 943 distance of 1. To address cases where 5' adapters had concatemerized during preparation of the 944 libraries, a second round of adapter trimming was performed to remove the 5' adapter using 945 cutadapt with the same parameters as above but an adapter sequence of 946 GTTCAGAGTTCTACAGTCCGA. Filtered reads were mapped against the human reference 947 genome (hg19) using the Tophat2 alignment algorithm (version 2.1.1 with default parameters) 948 (83), returning an average alignment rate of ~45%. Subsequently, UMIs were used to collapse 949 PCR duplicate reads using the UMI-tools 'dedup' method with default parameters. To identify 950 enriched regions of the genome, technical and biological replicates were pooled using the 951 Picard Tools function MergeSamFiles (84) and quality filtered using samtools (-q 5) (85). Peak 952 calling was then performed separately for each strand using MACS2 peak caller (version 2.1.1) 953 (86) with the negative control mimic sample as control (settings: -f BAM -g hs --keep-dup all 954 --nomodel --shift -15 --extsize 50 -B --call-summits --slocal 0 --llocal 0 --fe-cutoff 2 -q 0.05) 955 and the resulting peak files from each strand were merged. HITS-CLIP peaks and alignments 956 were visualized and interrogated using the Integrative Genomics Viewer v2.8.0 (87) and positions of DNA sequences matching a 6 nt portion of the miR-132 seed region 960 (AACAGT). Predicted mRNA targets of miR-132 from Ago HITS-CLIP were subjected to 961 gene ontology analysis with QuickGo (https://www.ebi.ac.uk/QuickGO/) and pathway analysis 962 with The Database for Annotation, Visualization and Integrated Discovery (DAVID) using 963 default parameters. Sequencing read and peak files for the HITS-CLIP dataset are available 964 from the NCBI Gene Expression Omnibus under the accession GSE190833 (for Reviewer 965 access go to https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE190833 and enter 966 token kfwtowawhjizhub into the box). 

Organ-specific lymphatic vasculature: From 1012 development to pathophysiology

Lymphatic Vessel Network Structure and Physiology

1017 (2014) Lymphangiogenesis and lymphatic vessel remodelling in cancer

The lymphatic system in health and disease

Lymphangioleiomyomatosis: current understanding and potential 1022 treatments

Emerging Roles for VEGF-D in Human 1024

Vascular endothelial growth 1026 factor signaling in development and disease

Multiple roles of lymphatic vessels in 1028 tumor progression

Proteolytic processing regulates receptor specificity and activity 1033 of VEGF-C

Vascular endothelial growth factor D (VEGF-D) is a ligand 1036 for the tyrosine kinases VEGF receptor 2 (Flk1) and VEGF receptor 3 (Flt4)

Growth Factor C in the Morphogenesis of Lymphatic Vessels

Vasculature in the 21(st) Century: Novel Functional Roles in Homeostasis and Disease

Isolated lymphatic endothelial cells transduce growth, survival and migratory signals 1049 via the VEGF-C/D receptor VEGFR-3

Lymphatic Endothelial Cell Tube Formation Assay to Identify Novel Kinases Involved 1053 in Lymphatic Vessel Remodeling

Genome-wide functional analysis reveals central signaling regulators of 1059 lymphatic endothelial cell migration and remodeling

About miRNAs, miRNA seeds, target genes and target pathways

MicroRNAs: target recognition and regulatory functions

Genome-wide identification of miR-200 targets reveals a 1069 regulatory network controlling cell invasion

The role of MicroRNAs in human cancer

1077 and Boshoff, C. (2010) miR-132 regulates antiviral innate immunity through 1078 suppression of the p300 transcriptional co-activator

MiR-19a enhances cell proliferation, 1083 migration and invasion through improving lymphangiogenesis via targeting 1084 thrombospondin-1 in colorectal cancer

1086 (2015) miR-128 downregulation promotes growth and metastasis of bladder cancer 1087 cells and involves VEGF-C upregulation

The Roles of

Non-Coding RNAs in Tumor-Associated Lymphangiogenesis

MicroRNA-132-mediated loss of p120RasGAP activates the 1093 endothelium to facilitate pathological angiogenesis

The lymphatic vasculature in disease

Vascular 1101 remodeling in cancer

Lymphangiogenesis: in vitro and in vivo models

2011) miR-132 and miR-212 are increased in 1106 pancreatic cancer and target the retinoblastoma tumor suppressor

VEGF-D promotes 1112 the metastatic spread of tumor cells via the lymphatics

Isolation and characterization of 1115 dermal lymphatic and blood endothelial cells reveal stable and functionally specialized 1116 cell lineages

Biosynthesis of Vascular Endothelial Growth Factor-D involves proteolytic processing 1120 which generates non-covalent homodimers

Plasmin activates the 1123 lymphangiogenic growth factors VEGF-C and VEGF-D

Proprotein convertases promote processing of VEGF

D, a critical step for binding the angiogenic receptor VEGFR-2

Vascular endothelial growth factor-d modulates caliber and function of initial 1132 lymphatics in the dermis

A system for quantifying the 1135 patterning of the lymphatic vasculature

Argonaute HITS-CLIP 1137 decodes microRNA-mRNA interaction maps

De-repression of FOXO3a death axis 1140 by microRNA-132 and -212 causes neuronal apoptosis in Alzheimer's disease. Hum 45

Multiple microRNAs modulate p21Cip1/Waf1 expression by directly targeting its 3' 1144 untranslated region

Systematic and integrative 1146 analysis of large gene lists using DAVID bioinformatics resources

The Gene Ontology resource: enriching a GOld mine

Structure and function of claudins

Endothelial cell-to-cell junctions-molecular 1157 organization and role in vascular homeostasis

Nectin-2 (CD112) Is Expressed on Outgrowth Endothelial Cells and 1162

Regulates Cell Proliferation and Angiogenic Function

1167 (1999) Inhibition of angiogenesis by thrombospondin-1 is mediated by 2 independent 1168 regions within the type 1 repeats

AKT hyper-phosphorylation associated with PI3K mutations in 1171 lymphatic endothelial cells from a patient with lymphatic malformation

RhoB controls coordination of adult angiogenesis and lymphangiogenesis following 1177 injury by regulating VEZF1-mediated transcription

VEGF-D promotes tumor metastasis by regulating prostaglandins produced by the 1185 collecting lymphatic endothelium

A novel vascular endothelial growth factor

VEGFR-3) and KDR (VEGFR-2) receptor tyrosine 1189 kinases

Vascular endothelial growth factor D (VEGF-D) is a ligand 1192 for the tyrosine kinases VEGF receptor 2 (Flk1) and VEGF receptor 3 (Flt4)

Ligand-induced vascular endothelial growth factor 1196 receptor-3 (VEGFR-3) heterodimerization with VEGFR-2 in primary lymphatic 1197 endothelial cells regulates tyrosine phosphorylation sites

VEGF receptor 2/-3 heterodimers detected 1202 in situ by proximity ligation on angiogenic sprouts

1205 (2013) The propeptides of VEGF-D determine heparin binding, receptor 1206 heterodimerization, and effects on tumor biology

Neuropilin-2 1212 interacts with VEGFR-2 and VEGFR-3 and promotes human endothelial cell survival 1213 and migration

Structural basis for VEGF-C binding to neuropilin-2 and 1216 sequestration by a soluble splice form

Advances in oligonucleotide 1221 drug delivery

Safety and efficacy of the BNT162b2 mRNA Covid-19 vaccine

Three-1231 dimensional CRISPR screening reveals epigenetic interaction with anti-angiogenic 1232 therapy

Monoclonal antibodies to vascular endothelial growth factor-D block its interactions 1238 with both VEGF receptor-2 and VEGF receptor-3

Post-transcriptional Gene Regulation by MicroRNA-194 Promotes Neuroendocrine 1246 Transdifferentiation in Prostate Cancer

1248 (2006) Detection of the argonaute protein Ago2 and microRNAs in the RNA induced 1249 silencing complex (RISC) using a monoclonal antibody

CLIP: crosslinking and immunoprecipitation 1252 of in vivo RNA targets of RNA-binding proteins

FastQC: A Quality Control Tool for High Throughput Sequence

Cutadapt removes adapter sequences from high-throughput 1257 sequencing reads

Fast and accurate long-read alignment with Burrows-1259

The microRNA Registry

UMI-tools: Modelling sequencing errors 1264 in Unique Molecular Identifiers to improve quantification accuracy

TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions 1268 and gene fusions

Project Data Processing

Integrative genomics viewer

Simple combinations of lineage-1281 determining transcription factors prime cis-regulatory elements required for 1282 macrophage and B cell identities